Social Exclusion: Short and Long Term Causes and Consequences

AIEL Series in Labour Economics For further volumes: http://www.springer.com/series/7370 Giuliana Parodi Dario Sciul...

Author: Giuliana Parodi | Dario Sciulli

42 downloads 1902 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

AIEL Series in Labour Economics

For further volumes: http://www.springer.com/series/7370

Giuliana Parodi Dario Sciulli •

Editors

Social Exclusion Short and Long Term Causes and Consequences

Prof. Giuliana Parodi Dipartimento di Metodi Quantitativi e Teoria Economica Università ‘‘G. d’Annunzio’’ di Chieti-Pescara Viale Pindaro 42 65127 Pescara Italy [email protected]

ISSN 1863-916X ISBN 978-3-7908-2771-2 DOI 10.1007/978-3-7908-2772-9

Asst. Prof. Dario Sciulli Dipartimento di Metodi Quantitativi e Teoria Economica Università ‘‘G. d’Annunzio’’ di Chieti-Pescara Viale Pindaro 42 65127 Pescara Italy [email protected]

e-ISBN 978-3-7908-2772-9

Springer Heidelberg Dordrecht London New York Library of Congress Control Number: 2011936144 Ó Springer-Verlag Berlin Heidelberg 2012 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Cover design: eStudio Calamar, Berlin/Figueres Printed on acid-free paper Physica-Verlag is a brand of Springer Springer is part of Springer Science+Business Media (www.springer.com)

Acknowledgments

This book contains a collection of papers, most of which have been presented at the thematic session on ‘‘Social Exclusion’’ of the XXV Annual Conference of the Italian Association of Labour Economists (AIEL), held at the University ‘‘G. d’Annunzio’’ of Chieti-Pescara in September 2010. All the papers, before being published, have been submitted to a double blind peer-review process. This has been possible thanks to the contributions of the referees: all the papers have benefited from their valuable suggestions. We are also indebted with various institutions that have provided us the financial support that has made possible this publication. We would like to thank: the Italian Association of Labour Economists, the Camera di Commercio, Industria, Artigianato e Agricoltura di Chieti, the Camera di Commercio, Industria, Artigianato e Agricoltura di Pescara, the Dipartimento di Metodi Quantitativi e Teoria Economica of the Università ‘‘G. d’Annunzio’’ di Chieti-Pescara, the Facoltà di Scienze Manageriali of the Università ‘‘G. d’Annunzio’’ di ChietiPescara and the Università ‘‘G. d’Annunzio’’ di Chieti-Pescara. Referees Luca Beltrametti, Università di Genova, Italy Roger A. Bowles, University of York, UK Floro Ernesto Caroleo, Università Parthenope di Napoli, Italy Sergio Destefanis, Università di Salerno, Italy Donata Favaro, Università di Padova, Italy Renata Livraghi, Università di Parma, Italy Emanuele Millemaci, Università di Messina, Italy Catia Nicodemo, Universitat Autonoma de Barcelona, Spain Carmen Pagliari, Università G. d’Annunzio di Chieti-Pescara, Italy Giuliana Parodi, Università G. d’Annunzio di Chieti-Pescara, Italy Matteo Picchio, Universiteit van Tilburg, The Netherlands Cristina Salvioni, Università G. d’Annunzio di Chieti-Pescara, Italy Dario Sciulli, Università G. d’Annunzio di Chieti-Pescara, Italy v

vi

Acknowledgments

Antonello Scorcu, Università di Bologna, Italy Marcello Signorelli, Università di Perugia, Italy Peter J. Sloane, University of Swansea and IZA Bonn, UK Massimo Tamperi, Università Politecnica delle Marche, Italy Francesco Timpano, Università Cattolica del Sacro Cuore - Sede di Piacenza -, Italy Ainura Uzagalieva, Universidade dos Açores, Portugal Robert J. Waldmann, Università di Roma Tor Vergata, Italy

Contents

1

Introduction and Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Giuliana Parodi and Dario Sciulli

Part I

The Structural and Long Term Causes of Social Exclusion

2

Health, Lifestyle and Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . Gianluigi Coppola

3

A Comparative Analysis of Literacy Rate in Contributing to Social Exclusion Insights. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Edgardo Bucciarelli, Carmen Pagliari, Fabrizio Muratore and Iacopo Odoardi

4

5

1

17

35

Education and Socioeconomic Mobility in Post-Communist Countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alina Veraschagina

67

Analysing Industrial Accidents in European Countries Using Data Envelopment Analysis . . . . . . . . . . . . . . . . . . . . . . . . Eugenia Nissi and Agnese Rapposelli

93

Part II

Groups Structurally Vulnerable to Social Exclusion

6

Social Exclusion and Offending . . . . . . . . . . . . . . . . . . . . . . . . . . Roger Bowles

105

7

Disability and Social Exclusion . . . . . . . . . . . . . . . . . . . . . . . . . . Peter J. Sloane and Melanie K. Jones

127

vii

viii

8

Contents

Civilian Disability Pensions as an Antipoverty Policy Instrument? A Spatial Analysis of Italian Provinces, 2003–2005 . . . . . . . . . . . Massimiliano Agovino and Giuliana Parodi

Part III 9

149

Economic Crisis, Labour Markets and Social Exclusion

Measuring the Long Wave. Unemployment, Discouragement and Semi-Employment in Italy, During and After the Crisis . . . . Leonello Tronti and Riccardo Gatto

171

10

Poverty and Unemployment: The Cases of Italy and Spain . . . . . Tindara Addabbo, Rosa García-Fernández, Carmen Llorca-Rodríguez and Anna Maccagnan

11

Labour Market Transitions During the Financial Crisis in Italy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marco Lilla and Stefano Staffolani

221

Labour-Market Dynamics and Unemployment: New Scenarios in the Italian Economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Carlo Lucarelli and Chiara Mussida

239

12

199

Contributors

Tindara Addabbo Dipartimento di Economia, Università di Modena e ReggioEmilia, Modena, Italy, e-mail: [email protected] Massimiliano Agovino Dipartimento di Metodi Quantitativi e Teoria Economica, Università ‘‘G. d’Annunzio’’ di Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy, e-mail: [email protected] Roger Bowles York Criminal Justice Economics York, UK, e-mail: roger. [email protected] Edgardo Bucciarelli Dipartimento di Metodi Quantitativi e Teoria Economica, Università ‘‘G. d’Annunzio’’ di Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy, e-mail: [email protected] Gianluigi Coppola Dipartimento di Scienze Economiche e Statistiche, Università di Salerno Salerno, Italy, e-mail: [email protected] Rosa García-Fernández Métodos Cuantitativos para la Economía y Empresa, Universidad de Granada Granada, Spain, e-mail: [email protected] Riccardo Gatto ISTAT Rome, Italy, e-mail: [email protected] Melanie K. Jones School of Business and Economics, Swansea University Swansea, UK, e-mail: [email protected] Marco Lilla Dipartimento di Economia, Università Politecnica delle Marche, Ancona, Italy, e-mail: [email protected] Carmen Llorca-Rodríguez Departamento de Economía Internacional y de España, Universidad de Granada, Granada, Spain, e-mail: [email protected] Carlo Lucarelli Ufficio Formazione e Lavoro, ISTAT, Rome, Italy, e-mail: [email protected] Anna Maccagnan Dipartimento di Economia, Università di Modena e ReggioEmilia, Modena, Italy, e-mail: [email protected]

ix

x

Contributors

Fabrizio Muratore Scuola Superiore, Università ‘‘G. D’Annunzio’’ di ChietiPescara, Pescara, Italy, e-mail: [email protected] Chiara Mussida Dipartimento di Scienze Economiche e Sociali, Università Cattolica del Sacro Cuore, Sede di Piacenza, Italy, e-mail: chiara.mussida@ unicatt.it Eugenia Nissi Dipartimento di Metodi Quantitativi e Teoria Economica, Università ‘‘G. d’Annunzio’’ di Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy, e-mail: [email protected] Iacopo Odoardi Scuola Superiore, Università ‘‘G. D’Annunzio’’ di ChietiPescara, Pescara, Italy, e-mail: [email protected] Carmen Pagliari Dipartimento di Metodi Quantitativi e Teoria Economica, Università ‘‘G. d’Annunzio’’ di Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy, e-mail: [email protected] Giuliana Parodi Dipartimento di Metodi Quantitativi e Teoria Economica, Università ‘‘G. d’Annunzio’’ di Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy, e-mail: [email protected] Agnese Rapposelli Dipartimento di Metodi Quantitativi e Teoria Economica, Università ‘‘G. d’Annunzio’’ di Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy, e-mail: [email protected] Dario Sciulli Dipartimento di Metodi Quantitativi e Teoria Economica, Università ‘‘G. d’Annunzio’’ di Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy, e-mail: [email protected] Peter J. Sloane School of Business and Economics, Swansea University, Swansea, UK, e-mail: [email protected] Stefano Staffolani Dipartimento di Economia, Università Politecnica delle Marche, Ancona, Italy, e-mail: [email protected] Leonello Tronti Università di Roma Tre, Rome, Italy, e-mail: [email protected] Alina Veraschagina Dipartimento di Economia e Diritto, Università di Roma La Sapienza, Rome, Italy, e-mail: [email protected]

Chapter 1

Introduction and Overview Giuliana Parodi and Dario Sciulli

1.1 The XXV AIEL Conference The XXV Conference of AIEL (Associazione Italiana degli Economisti del Lavoro) was held in the Università d’Annunzio in Chieti-Pescara in September 2010. The theme chosen for the thematic session was ‘‘Social Exclusion’’, in line with the choice made by the European Union to make 2010 the ‘‘European year to fight poverty and social exclusion’’. The European Commission defines social exclusion for societies and for individuals: ‘‘for societies: disintegration and fragmentation of social relations and hence of social cohesion… for individuals… a progressive process of marginalization leading to economic deprivation and various forms of social and cultural disadvantage’’. From another perspective, the socially excluded can include individuals who are unable to participate in the basic economic and social activities of the society in which they live. In the Conference, the themes under investigation have ranged from causes to consequences of social exclusion, focusing on disability, gender gaps, youth, income distribution, discrimination, accidents at work, and literacy rates (LR). Selected works that were presented at the Conference have been invited to be part of this book, to which distinguished authors, not present at the Conference, have also been invited to contribute. The book is organised in three parts: the first and the second parts deal with structural, long-term causes of social exclusion (i.e., health, education, disability,

G. Parodi (&) D. Sciulli Dipartimento di Metodi Quantitativi e Teoria Economica, Università d’Annunzio di Chieti-Pescara, Pescara, Italy e-mail: [email protected] D. Sciulli e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_1, Springer-Verlag Berlin Heidelberg 2012

1

2

G. Parodi and D. Sciulli

crime); the third part deals with short-term causes of social exclusion and, in particular, the effects of the most recent economic crisis. In most of the papers, social exclusion is assessed in terms of unemployment, income, and education. The first part deals with the long-term, structural causes of social exclusion and, in particular, the problems associated with its possible effects on health (Coppola), education (Bucciarelli, Pagliari, Muratore, Odoardi, and also Verashchagina), and work conditions (Nissi and Rapposelli). The second part deals with vulnerable groups who are particularly at risk of social exclusion [i.e., former convicted persons (Bowles) and disabled persons (Sloane and Jones, Agovino and Parodi)]. Finally, the third part discusses problems connected with the mechanisms of the labour market, one of the main direct or indirect causes of social exclusion, with special emphasis on the effects of the most recent economic crisis on social exclusion, including unemployment (Tronti and Gatto), poverty (Addabbo, GarcíaFernández, Llorca-Rodríguez and Maccagnan), and transitions within the labour market (Lilla and Staffolani and also Lucarelli and Mussida). In what follows, we briefly discuss the approaches used to investigate social exclusion; we also revise the main effects of the most recent economic crisis on the labour market and policy measures put (or not put) into effect to overcome its effects. We then present a brief summary of the contributions that appear in the three parts of the book. Finally, we draw some conclusions.

1.2 Main Approaches to the Discussion of Social Exclusion Social exclusion is a term that has jumped to the forefront of the literature since it was first used in France in the 1970s. Since then, it has been widely adopted as a topic of economic policy, both by governments and by researchers; for example, in 1997, the British government set up a Social Exclusion Unit (SEU), and the London School of Economics set up the Research Centre for the analysis on Social Exclusion (CASE); in 2007, the Social Inclusion Unit was set up in Australia; within the European Union, the definition of social exclusion has been sharpened over time (Nolan and Whelan 2010). Social exclusion is interpreted both as a static situation of deprivation and as a process leading to it. Multidimensional poverty and social exclusion are closely related concepts, but they do not overlap. Both of these concepts refer to Sen’s intuition of capability failures, but in the multidimensional approach to poverty, failure is viewed in terms of shortfalls from a threshold for each function, and in the social exclusion approach, failure is interpreted as an inability to participate and, therefore, identifies functioning failures. These two different approaches refer to quantitative and qualitative aspects, respectively, of social exclusion (Chakravarty and Zoli 2009). Initially, the threshold was defined in terms of income. Social exclusion came very close to the definition of poverty; subsequently, it was defined in terms of observed and unobserved heterogeneity. At the individual level, shortfalls in the

1 Introduction and Overview

3

exclusion or deprivation score of a person depend on the number of characteristics from which that individual is excluded. With respect to heterogeneity, the literature differs both with respect to the variables considered relevant hindrances to take part in societal life, and with respect to the weights to be attributed to them. There is a consensus on some of the variables to be included in the definition, such as housing, health and education. Other authors include financial difficulties, basic necessities, durables, social contacts, and dissatisfaction. Authors’ choices vary in terms of whether the effect among these variables is multiplicative or additive. If multiplicative, being excluded in just one dimension is sufficient to create social exclusion. Moreover, some dimensions may be considered more relevant than others, so the same or different weights can be attributed to shortcomings in each characteristic. Gender effects of social exclusion have been investigated, even though only few significant differences have been found between genders. Furthermore, few significant differences have been found between cohorts with respect to social exclusion; therefore, the findings about the factors affecting social exclusion appear to be robust. With respect to time, social exclusion shows a dynamic or longitudinal dimension: duration of the time spent in deprivation is considered important in defining social exclusion, even though the literature differs in deciding how long permanence in deprivation must be in order to allow for a definition of social exclusion, or to lead to social exclusion. True state dependence, and therefore persistence, has been investigated with respect to social exclusion. In fact, stochastically socially excluded people appear to experience a shorter duration in social exclusion compared with the structurally socially excluded (Adato et al. 2004). The inter-generational legacy of social exclusion has also been investigated, overall and with respect to specific variables. Research has found that traditional elements of transmission of disadvantage, such as class, play a minor role when compared with other more specific variables. Moreover, there are indications that some legacies of disadvantage perpetuate themselves across generations in a segmented way, i.e., bad housing perpetuates itself, as do low educational achievements, demographics and family behaviours, personality behaviours and mental health (Nolan and Whelan 2010). Society’s social exclusion can be interpreted as the aggregate of individuals’ social exclusion. Some axiomatic measures of social exclusion have been proposed (Chakravarty and Zoli 2009), where societies’ overall social exclusion has been derived from the individuals’ social exclusion. A country’s social exclusion is determined by the weighted social exclusion of the various groups of excluded people within it. In policy terms, the elimination of exclusion in one group would reduce social exclusion in the country for Europe, see for instance Chackravarty and D’Ambrosio (2003). The policy implications of these findings do not always follow a single trend: short-term intervention(s) are needed if persistence appears as the main factor explaining social exclusion, while long-term intervention(s) are needed to deal with structural factors that affect social exclusion. In the European Commission Report on Social Exclusion (2010), an account is given of EU structural

4

G. Parodi and D. Sciulli

interventions and the financial intervention of individual member states. The European Union has tackled the problem of social exclusion with various initiatives. In 2006, the Social Policy Agenda, 2006–2010, was established and renewed by the Social Agenda presented in 2008. The structural European Funds provide direct financial intervention, and in particular, the European Social Fund for the 2007–2013 period allocates € 756 billion to the Member States. Furthermore, the EU launched specific programs for categories of people particularly at risk of social exclusion, such as the 2007–2013 PROGRESS (Programme for Employment and Social Solidarity), which is financed with € 743 million; programmes for youths; the HELIOS programme for disabled people; programmes for the elderly; and programmes for gender equality. In addition, the European Union takes an active role in raising awareness towards the problems of social exclusion and poverty with diversified initiatives: it ran an investigation on people’s attitudes towards social exclusion (e.g., 2007, Eurobarometer) and nominated 2010 as the year to fight poverty and social exclusion, which entailed financing several important initiatives in the Members States. At the level of individual countries, the 2010 European Commission Report provides information on various policies that address poverty and social exclusion, in connection with the labour market, education, health, housing, and social networks. The labour market is the only area where Member States have agreed on a common benchmark for 2010: to encourage at least 25% of the long-term unemployed into training, re-training, work practice, employment or another employability measure, combined, where appropriate, with on-going job search assistance. In all other areas, Member States pursue their own independent policies. In 2006, the EU27 on average devoted 27% of their respective GNP to social protection; these transfers cover both prevention policies (such as pensions and health) and alleviation policies (such as unemployment benefits and housing). As an indicator of the structural effect of these policies, we can look at the reduction in the at risk of poverty rate after social transfers (excluding pensions), which is 27% for the entire EU 27, with a particular strong effect on children under 18, for whom the reduction in the risk of poverty rate rises to 39%. This aggregate figure conceals great heterogeneity among countries; for instance, the risk reduction after social transfer is particularly low in Mediterranean countries, such as Italy, Greece, and Spain.

1.3 Economic Crisis, Labour Markets and Social Exclusion Economic conditions are an important factor in determining poverty and social exclusion. During economic downturns, people are exposed to higher unemployment and, possibly, to fiscal austerity applied by governments in order to control deficit and public debt levels. Moreover, without adequate policies, economic downturns are likely to produce unbalanced outcomes that affect disadvantaged people and that lead to higher inequality, poverty and social exclusion.

1 Introduction and Overview

5

Leschke and Watt (2010), focusing on the role of the 2008–2009 economic crisis on incomes, public budget and labour markets, have highlighted the implications of economic crisis regarding social effects. According to the European Commission statistics (2010), more than 80 million Europeans, including 20 million children, live at risk of poverty and social exclusion. The economic crisis has made things worse by putting an estimated 5 million more people out of work, potentially exposing many more families to poverty and social exclusion. European labour markets were strongly affected by the 2008–2009 economic crisis; it caused a severe output decline that was quickly transmitted to the labour market, producing a reduction in employment and working hours and, in many countries, a strong increase of unemployment rates and underemployment. These effects represent a challenge for labour markets and for the social policies of European governments to mitigate the welfare loss of their citizens. Many European countries entered the recession with low unemployment rates, partly because of the effects of structural labour market reforms implemented since the 1980s. In any case, unemployment has increased strongly in the European Union during the 2008–2009 economic downturn, reaching 9.6% in March 2010 (EUROSTAT 2010). However, the rise in unemployment has been quite heterogeneous across European countries. According to the OECD (2010) analysis, job losses were particularly high in countries (e.g., Spain, Ireland, USA) where the housing market contributed the most to the economic downturn. On the contrary, unemployment has only slightly increased in countries where the economic crisis was mainly explained by a decline in exports (e.g., Germany, the Netherlands, Slovak Republic, and Japan). Moreover, job losses have been distributed heterogeneously among worker groups and industries. The economic crisis has more strongly affected construction, mining and manufacturing sectors; young, low-skilled and temporary workers have been the most affected at the individual level by the economic downturn. Statistics indicate that youth unemployment is two to three times greater than the average unemployment rate, and this effect is especially prevalent among low-qualified workers, for which exit from the labour market and/or the loss of earnings should be carefully avoided. Similar considerations apply to immigrants. Heterogeneous effects at worker level also reflect the structure of the economy and labour market institutions. For example, Bentolila et al. (2010) have found that Spain would have avoided 40% of new unemployment if it had adopted French employment protection legislation. It follows that, while flexibility policies have strongly contributed to increase employment in Europe until the economic recession, flexible labour markets during economic crises are associated with higher unemployment rates, especially for specific worker groups (e.g., youth unemployment). The 2008–2009 recessions has also affected the unemployment dynamics of many European countries; it has caused an increase in unemployment inflows and a reduction in unemployment outflows, contributing to an increase in the unemployment pool and in the average duration of unemployment. In this case as well,

6

G. Parodi and D. Sciulli

unemployment dynamics have affected various countries and various worker groups differently. European governments have adopted various measures with the aim of reducing the social impact of the economic downturn. In this sense, even though the economic recovery had begun in 2009, many European countries have continued to expand or hold constant resources to finance unemployment benefits and re-employment assistance. Unemployment benefit systems should be designed in such a way that the job search effort is not negatively affected. Income support policies adopted at European level have also included minimum-income benefits as well as income support for low income workers. Active labour market program (ALMPs) spending has scaled up when compared with passive spending in the last decade, but it seems to be acyclical (OECD 2009) in absolute terms, with a small number of exceptions. Some countries have also reinforced expenditures for public employment services. Moreover, some specific measures, such as employment subsidies [including shortterm work (STW) schemes] and direct public job creation, have been expanded in various European countries. For example, participation in STW schemes has increased in many European countries, playing an important role in preserving (permanent) jobs, through the encouragement of work sharing to reduce layoffs. Furthermore, STW has reduced the loss of permanent employment by over 200,000 in Germany and by about 120,000 in Italy (OECD 2010). However, as STW just preserves regular employment, it contributes to increased labour market segmentation against temporary and part-time workers. In this context, Italy has been extensively affected by the 2008–2009 economic downturn, even though its situation appears to be better than many other European countries. During the recession, the Italian GDP has fallen by 6%, but the unemployment rate increased to just 8.6% in 2010 (ISTAT 2011), which is below the EU average rate. This result has been achieved through the extensive use of STW schemes, which have helped to mitigate the negative effect of the economic crisis on the permanently employed. However, the economic recovery does not seem particularly brilliant, and the effect of the crisis is likely to persist in the next months. In Italy, the manufacturing sector has suffered most, mainly because of a reduction in exports. Moreover, the rise in unemployment has not been homogenous across workers groups. The male unemployment rate has risen more than the female unemployment rate, primarily because of the reduction in labour market participation of women. The youth unemployment rate has risen to 28.9%, opening questions about the effectiveness of the structure of labour market policies in Italy.

1.4 Brief Summary of the Papers in the Book We now briefly summarise the content of the book, presenting the summary of the contributions according to the order in which they appear: (Sect. 1.4.1) deals with general causes of social exclusion (i.e., health, education, and work conditions); (Sect. 1.4.2) deals with some vulnerable groups who are most likely to be

1 Introduction and Overview

7

associated with social exclusion (i.e., former criminals and disabled people); and (Sect. 1.4.3) deals directly with the impact that the most recent crisis has had on social exclusion through the labour market.

1.4.1 The Structural and Long-Term Causes of Social Exclusion Coppola’s model predicts that the effect of income growth on health can be ambiguous, providing an explanation to the observed ‘‘health poverty trap’’. This model provides micro foundations to Solow’s model of economic growth by linking growth to health via considering the effect of lifestyle on growth. This model centres on the concept of the lifestyle returns to scale that expresses the elasticity of health with respect to income: a bad lifestyle implies that an increase in income increases the consumption of commodities that are pernicious to health. At the aggregate level, the Growth Model presented uses the Solow Growth Model with a constant saving rate, diminishing returns of capital and labour, labour augmenting technology, and constant returns to scale. The model develops a health multiplier, on the assumption that health is labour augmenting; therefore, individual choices about health affect society as a whole via externalities. However, lifestyle can also be an important channel to transmit economic growth to health, depending on technological progress. Unlike economic growth, the rate of health growth may be positive, null or negative, depending on the sign of the Lifestyle Returns to scale: if it is negative, economic growth negatively affects health. Bucciarelli, Pagliari, Muratore, and Odoardi address the important issue of investigating the connection between social exclusion and LR by analysing data across 30 very diversified countries, some of which are OECD members, collected for the period between 2007 and 2009. The topic under investigation is important, as the LR and variables related to social exclusion can be seen as proxies of human and social capital, which are essential for long-term development. Knowledge of the variables that affect education is essential for policymaking purposes. The paper has a comprehensive description of the statistical methodology that is used in the investigation (i.e., multivariate regression analysis, maximum likelihood with VARIMAX rotation, factor transformation matrix, factor analysis and multivariate regression models with maximum likelihood components, and cluster analysis). The paper is divided into several sections. Initially, it analyses LR as a function of three groups of variables: variables strictly related to the educational situation, i.e., school enrolment at three different levels and children out of school; public intervention in education, expressed both in terms of general public expenditures and in terms of GDP; and macro variables, such as real growth rate and long-term unemployment rate. Because of the heterogeneity of the data, links

8

G. Parodi and D. Sciulli

among the variables are not very evident. But, after applying the various methods of analysis, it is found that LR can be affected by three components with the nature of social exclusion variables, i.e., a synthetic indicator of social exclusion, an indicator regarding educational level, and the GDP real growth rate. Most of the countries analysed, including almost all of the OECD ones, show similar levels of macro variables related to education, and instead differ on social exclusion. This result means that all countries with medium or high levels of income per capita show good levels of enrolment and attendance at school but that not all of them can deal effectively with social problems. Finally, a detailed discussion of policy towards literacy in the USA and Italy is discussed. Veraschagina addresses the question of the role of education on social mobility in 12 post-communist countries of Central Europe and the former Soviet Union. Three hypotheses are tested: during the socialist era, the relation between the education level of parents and their children weakened; the current levels of educational mobility in post-communist countries are higher than in their Western counterparts; and the transition to market economy caused an increase in educational persistence. Despite heterogeneity among countries, educational persistence appears to decline sharply until the 1950s; however, evidence from later years does not confirm this trend. To test the effect of this transition, identification of the break was necessary. The CUSUM and the Chow test were used for this purpose, and they showed that possibly not a single break occurred throughout the countries but that additional breaks in the data referring to educational persistence took place, on average about 10 years before the transition itself occurred. Moreover, lower educational persistence in post-communist versus mature countries is confirmed by the evidence only for the generation born in the 1950–1960s. The question of the sustainability of educational mobility was also investigated: the increase of private returns to education (PRE), which the data show after the transition, is expected to be associated with higher educational mobility. This result is not confirmed by the evidence. A possible explanation is the shift in the remuneration system, heavily relying on in kind transfers in the pre-transition period, towards a completely monetary system of remuneration after the transition. Taking this aspect into account, it is possible that the PRE has actually not been altered. Unfortunately, the data do not allow an assessment of in kind transfers, so the explanation provided here is a convincing speculation. Nissi and Rapposelli apply the nonparametric technique Data Envelopment Analysis to investigate comparative efficiency in terms of low numbers of accidents at work to EUROSTAT data collected for 15 European countries in 2005, in the sectors of manufacturing, construction and distribution trades. These sectors were chosen because Europe registers the greatest number of accidents. Under the traditional hypothesis of variable returns to scale, and under the specific assumption of an input oriented approach, the paper develops the idea of undesirable outputs, which only recently has been introduced in the literature and which has mainly been applied to environmental studies. The results of this analysis show that countries differ from each other with respect to efficiency in a different way, according to the sector considered. In the manufacturing sector, four countries are

1 Introduction and Overview

9

efficient (i.e., Germany, Ireland, Luxemburg, and the United Kingdom). They make a total of four efficient countries in the construction sector as well; in the distribution trades sector, Belgium and France join the other countries on the efficient frontier, and Germany returns to being efficient, making a total of seven efficient countries. The overall conclusion suggests that in the construction and manufacturing sectors, the efficiency of reducing accidents at work can improve considerably, while in the distribution trade sector, nearly 50% of the countries considered are already efficient. Ireland, Luxemburg, and the United Kingdom are efficient in all sectors considered, and Ireland and the United Kingdom are frequently cited as exemplary efficient Decision Making Units, as they have a low number of accidents at work.

1.4.2 Groups Structurally Vulnerable to Social Exclusion The fragile groups considered in this part of the book include former convicted and disabled people. Bowles uses various cross-sectional and longitudinal British data sets to explore the relationship between crime and social exclusion, concentrating the investigation on the human capital approach model of offending. According to this model, crime is the result of a very high discount rate, leading to underinvestment in oneself. High discount rates can be explained in terms of a lack of planning skills and an incapacity to anticipate the future, with special reference to an inability to assess the dynamic consequences of offending. The formation of this kind of individual preferences may be explained by variables that are intertwined with the concept of social exclusion. This model is well suited to explain the offending behaviour of young people, who actually represent a large majority of offenders. The variables that the data show to be correlated with offending are experience of disadvantage as children, poor attendance at and exclusion from school, exclusion from mainstream education, leaving school by the minimum school leaving age, pre-custody drug use, and other crime victimisation rates. Furthermore, reconviction rates appear to be related to prospects of unemployment after release, and to problems of accommodation; the probability of reconviction falls with age. There is also evidence of a comparatively high incidence of mental health issues among prisoners. Of course, with all these factors, one cannot say whether they are contributory factors to crime, or consequences of it, or whether they both have a common cause. The policy implications of this model go well beyond those of the punitive model of crime, and recommend both ‘prevention’ policies aimed at raising aspirations and encouraging investment in human capital, and ‘resettlement’ policies. A problem connected with this model is that the predicted benefits of intervention spill over a wide range of potential savings in costs, which affect various government departments, and it may be difficult to take all of them into account if, for the purpose of financing, the benefits have to be assessed in terms of crime prevention only.

10

G. Parodi and D. Sciulli

Sloane and Jones present a very thorough review of the literature on disability that is likely to become a standard tool of reference in the field. They pay particular attention to disabled people and the labour market, commenting on a variety of datasets, ranging from the European ECHP and the Australian HILDA to the British BHPS. The review touches on most aspects related to disability: variability of the percentage of people who declare themselves disabled, the probability of employment and seriousness of disability, evidence on disability and low income from work, both with cross-sectional and with longitudinal data, and a discussion of additional expenditure requirements for disabled people. The definitional aspect of disability is thoroughly investigated, with special emphasis on work limited disability, non-work-limited disability, and the non-disabled. Other issues also thoroughly investigated are problems associated with the age of onset of disability, with self-reporting disability, with the simultaneous relations between health and labour supply and vice versa, and with state dependence. A thorough review of the literature using a UK background is also provided. The problem of possible discrimination against disabled persons is investigated, with special attention on disability and productivity, the importance of education and disability on employment and earnings, work limited and non-work-limited disability, part-time employment and self-employment, gender, possible job mismatches, and job satisfaction. A thorough review of policies related to disability, including income support, employment quota systems, sheltered employment, anti-discrimination legislation, and employment policies, is also provided. Individually targeted employment policies are presented as the most promising policies to help disabled individuals to retain and gain employment for which they are most suited. Agovino and Parodi investigate whether Italian civilian invalidity pensions are used as a form of income support, analysing data at the level of individual provinces. Their analysis proceeds in two steps: a panel analysis and a Generalized Method of Moments where the spatial variable is introduced. The first step of the analysis aims at identifying possible differences in disabling health conditions in various Italian provinces. For this purpose, information about the discharges from hospitals of people suffering from possibly invalidating illnesses is analysed. The results of this analysis show that there is no significant difference in this variable across provinces. Subsequently, the analysis is developed on the link between civilian invalidity pensions and indicators of social malaise such as poverty. The poverty rate appears always significant as a regressor of civilian invalidity pensions, even when the spatial lag is introduced in order to take into account similarities between provinces which may go well beyond administrative boundaries, because of historical and socio-economic factors. These findings reveal a degree of flexibility in the allocation of civilian invalidity pensions, which appears to be inconsistent with the strict rules defined by the law, and suggest discretionary interpretation according to the locally prevailing socio-economic conditions. The unsuitability of civilian invalidity pensions as a form of income support is stressed.

1 Introduction and Overview

11

1.4.3 Effects of the Recent Crisis on the Labour Market Papers dealing with the effects of the 2008–2009 economic downturn on labour markets and its socio-economic consequences include some case-studies from Tronti and Gatto, Addabbo, Garcia-Fernandez, Llorca-Rodriguez and Maccagnan (Addabbo et al. hereafter), Lilla and Staffolani and, finally, Lucarelli and Mussida. Specifically, Tronti and Gatto investigate the 2008–2009 recession in the context of the Italian labour market, focusing especially on the problems that concern the unemployment measurement and its consequences in understanding the phenomena. Addabbo et al. study the impact of unemployment on the probability of being income poor and on the difficulties of accessing medical and dental visits treatments in Italy and Spain to draw lessons that can be extended to the economic crisis. Finally, both Lilla and Staffolani and Lucarelli and Mussida, focus on unemployment dynamics in Italy clarifying the role of the 2008–2009 recession on labour market transitions. Tronti and Gatto stress the relevance of considering various unemployment measures to avoid shortcomings in evaluating the socio-economic impact of the 2008–2009 recession. According to official statistics, according to the standards defined by the International Labour Office, the unemployment rate has risen to 8.6% in 2010, as a consequence of the strong economic downturn. Nevertheless, the actual impact of the recession may be greater if the meaning of unemployment is extended to include the concept of the labour underutilisation rate. Specifically, Tronti and Gatto show that, considering relevant phenomena such as the discouragement effect, with specific STW schemes providing for working zero hours (Cassa Integrazione Guadagni), labour hoarding and semi-employment (atypical workers), the resulting labour underutilisation rate is almost double the standard unemployment rate. This finding has relevant implications when we consider the impact of the crisis on the socio-economic conditions of individuals and households. Moreover, it changes the debate about the adequacy of official definitions of unemployment in providing a reliable measure of the phenomenon and its implications (see also Brandolini et al. 2006). Addabbo et al. analyse the effect of being unemployed on the probability of being income poor and on the difficulties of accessing medical and dental visits treatments in Italy and Spain using 2007 IT-SILC/ES-SILC data. The authors also describe the different unemployment insurance systems and the effect of the economic crisis on both labour markets. Overall, Addabbo et al. find that being unemployed increases the probability of being income poor and, partially, the difficulties of accessing to medical and dental treatments in both countries. In Italy, the effect is stronger for the previously self-employed now unemployed while in Spain it is stronger for the unemployed who were never employed before. These findings are relevant in analysing the possible effects of the 2008–2009 economic downturn on households’ well being, considering the structure of welfare systems with specific attention to the unemployment insurance systems. Unemployment dynamics in Italy during the last few years have been studied in two papers. The evidence from both of the papers is quite consistent, even though

12

G. Parodi and D. Sciulli

some differences emerge, probably because of some differences in model specifications and because of the data sets used. The results seem to reveal a strong duality in the Italian economy, as the recession has had a strong impact on disadvantaged categories. Overall, young workers and immigrants have suffered more than other groups during the jobs crisis because recent flexibility policies have banished them from temporary employment. This finding raises questions about the role of labour market policies and supports in Italy. Specifically, Lilla and Staffolani use the ISTAT Labour Force Survey micro-data to evaluate the consequences of the 2008–2009 economic crisis on the Italian labour market by analysing the determinants of individuals’ transitions between occupational states. They use two different strategies to identify the determinants of transitions: first, they consider the ‘‘perceived’’ employment condition, available at the individual level for the previous year; and, second, they use the ‘‘ecological inference’’ approach, i.e., they get quarterly transition probabilities for the ‘‘official’’ employment condition. The estimation results confirm that during the crisis young workers, men, immigrants, and individuals in the north of Italy faced the major risks of becoming unemployed, while women, middle-aged and northern individuals had major troubles in finding a job. With respect to the family types that were analysed in the paper, the most affected individuals belong to ‘‘couples without children’’ and to ‘‘single parents without children’’. They experienced severe reductions in employment stability along with an increase in the transitions to unemployment. Finally, Lucarelli and Mussida provide a picture of the labour market transitions in Italy for the 2004–2009 period, with specific attention on the effects of the 2008–2009 economic downturns. The authors find that particular characteristics reduce the negative impact of the economic downturn. The individual characteristics that increase the chances of remaining in employment emerge as being male, no longer belonging to the ‘‘young’’ age group, and holding a high educational qualification. Moreover, they find that full-time workers have a lower probability of exiting employment. The same effect emerges for occupations requiring high qualifications. With regard to outflows from inactivity, women appear less likely to exit the nonlabour force. This finding indicates a strong discouragement effect for women, especially at the beginning of their working career. Holding low qualifications and unfavourable labour market conditions inevitably exacerbate the situation. All of these findings suggest a greater negative impact for disadvantaged labour market groups, which leads to a widening of labour market duality. The authors discuss changes in government labour policy promoted to fight the negative effects of the economic crisis on labour market outcomes.

1.5 Conclusions The contributions to the book have emphasised the circular aspects of social exclusion. Factors that generate social exclusion are themselves an effect of social exclusion. Groups that are traditionally considered fragile, such as disabled people

1 Introduction and Overview

13

and formerly convicted people, are particularly vulnerable to social exclusion. Articles on the 2008–2009 economic downturn make clear that regarding social exclusion, the crisis has affected a much wider range of people, including youths, low-skilled workers and immigrants, in terms of income, job loss, job finding, even participation. The result is a widening of socio-economic duality. The relevance of definitions of socio-economic outcomes for measurement and evaluation of the crisis have also been stressed, and their consequences for interventions of policymakers in fighting socio-economic effects of the economic downturn. It has been said above that social exclusion is both an indicator of discomfort for the people who suffer it and a threat to social cohesion, so policy interventions to address social exclusion are required, on both grounds. However, both the situation of specific disadvantaged groups and the asymmetric effect of the crisis open questions about the effectiveness of the welfare system in addressing and mitigating social exclusion. Given these premises, a revision of the welfare structure should be considered, taking into account, for example, the changes in labour market institutions introduced in the last few decades. Finally, it seems desirable that the European Union should impose to member countries parameters to be respected, and not only targets and guidelines, in terms of social indicators, so that countries too remote, for instance, from the Lisbon (original and revised) targets may find themselves under threat of sanctions.

References Adato M, Carter MR, May J (2004) Sense in sociability? social exclusion and persistent poverty in South Africa, University of Wisconsin-Madison, Department of Economics and Agricultural Economics, Staff paper, No. 477 Bentolila S, Cahuc P, Dolado JJ, Le Barbanchon T (2010) Two tier labour markets in the great recession: France vs. Spain. Paper presented at the XXV simposio de la asociaciòn espanola de economia, Madrid, December 2010 Brandolini A, Cipollone P, Viviano E (2006) Does the ILO definition capture all unemployment? J Eur Econ Assoc 4(1):153–179 Chakravarty SR, D’Ambrosio C (2003) The measurement of social exclusion, DIW, German Institute for economic research, Berlin, Discussion papers 364 Chakravarty SR, Zoli C (2009) Social exclusion orderings, Working paper series, University of Verona European Commission (2010) Combating poverty and social exclusion: a statistical portrait of the European union 2010. Eurostat statistical books, http://ec.europa.eu/social EUROSTAT (2010) Official statistics. http://ec.europa.eu/eurostat ISTAT (2011) www.istat.it Leschke J, Watt A (2010) How do institutions affect the labour market adjustment to the economic crisis in different EU countries, ET UI working paper no. 2010-04, European trade union institute Nolan B, Whelan CT (2010) Using non monetary deprivation indicators to analyze poverty and social exclusion: lessons from Europe? J Policy Anal Manage 29(2):305–325 OECD (2009) Employment outlook: tackling the jobs crisis. Paris OECD (2010) Employment outlook: moving beyond the jobs crisis, Paris

Part I

The Structural and Long Term Causes of Social Exclusion

Chapter 2

Health, Lifestyle and Growth Gianluigi Coppola

Abstract In this article, I attempt to explain why lifestyle may have a positive impact on economic growth. First, I consider the ways in which health affects a consumer’s utility, and I then define a Health Production Function for which health is the output and consumer good is the input. In this approach, the Lifestyle Return to Scale (LRS) parameter is defined. The first result is that an increase in a consumer’s personal income may have a positive or a negative effect on health. That is, health may be a normal or an inferior good, depending on the Lifestyle Return to Scale value. According to this result, I compute a health multiplier and then modify the Solow Growth Model in which health is labour-augmenting. The result is a model in which the Lifestyle Return to Scale positively affects per capita income and per capita income growth.

Lifestyles Growth JEL Classification I10 O40 Keywords Health

2.1 Introduction At the macro level, stylized facts indicate substantial differences in per capita income and health status among countries and regions. This may imply that low per capita incomes negatively affect health and vice versa.

G. Coppola (&) Dipartimento di Scienze Economiche e Statistiche, Universita’ Degli Studi Di Salerno and C.E.L.P.E, Salerno, Italy e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_2, Springer-Verlag Berlin Heidelberg 2012

17

18

G. Coppola

In the last 20 years, the literature on economic growth has focused primarily on the role of human capital accumulation, while health has occupied only a marginal role in economic analyses. Second, if the income and health differences among countries are significant, those among regions are even stronger and are essential to economic growth. The literature on health and health economic growth has not considered the impacts that lifestyle has on Economic Growth. The aim of this theoretical paper is to develop the relationship between Health and Growth by accounting for consumer lifestyles. Contoyannis and Jones’s hypothesis (2004) introduced a micro-model of consumer choice to define lifestyles more accurately and to explain the effects of consumers’ choices on Health. At the macro level, Weil (2005) asked if the forces driving differences among regions were primarily derived from health or income. In this context, I provide an answer by computing a health multiplier (Sect. 2.2) and developing a simple modified Solow growth model in which health is labouraugmenting (Sect. 2.3). Thus, this model includes the relationship among income, lifestyle and health status first obtained at the micro level. The first important result is that an increase in a consumer’s personal income may have a positive or a negative effect on health if the consumer has a ‘‘good’’ or a ‘‘bad’’ lifestyle, respectively. At the macro level another result of the model is that lifestyle may be crucial for growth: a ‘‘good’’ lifestyle can generate a positive impact on economic growth, while a ‘‘bad’’ lifestyle may also negatively affect growth. The model also explains why health improvement positively affects income, while increasing income may have a lower effect on health (Weil 2005).

2.2 Some Empirical Evidence The empirical evidence can be divided into two categories: Long-Run ‘‘stylized facts’’ and facts related to government’s measures against the unhealthy habits of the citizens. The long-run ‘‘stylized facts’’ show increases in per capita income and life expectancy. From 1820 to 2001, the per capita World GDP grew from $667 to $5.709 (Maddison 2003). In the same period, world life expectancy at birth increased from 28.5 years in 1820 to 65.2 years in 1990. The differences among regions, in terms of both per capita GDP and life expectancy at birth, have also increased (Riley 2005). In 1998, the per capita GDP of the United States was 20 times that of Africa, while in 1820, it was only 3 times larger. In the period from 1800 to 1820, the differences among regions, in terms of life expectancy at birth, were relatively low compared to the period from 1990 to 2001. In the first period, the difference between the highest and the lowest regional life expectancies was only 9.2 years. This gap widened to 26.3 years in the period from 1990 to 2001 (Riley 2005).

2 Health, Lifestyle and Growth

19

These disparities may be explained by differences in Public Health, Medical diagnostics and care, diet and, more generally, lifestyle. As regards the second aspect, the government’s measures, it exists a lot of evidence of the importance that people’s lifestyle assumes. For example, in England, many local governments offer incentives to encourage consumers to lead healthier lifestyles. In Dundee, smokers are offered £12.50 a week by the NHS if carbon monoxide testing shows that they have quit. In Essex, pregnant women can claim a £20 food voucher from the NHS when they stop smoking for one week, £40 after four weeks and another £40 after one year. Brighton offers children £15 to stop smoking for 28 days. Overweight patients in Kent are offered incentives for losing weight. In the US and other countries, incentives are offered for weight loss, compliance with diabetes treatments, or regularly testing negative for sexually transmitted diseases.1 Moreover, in Japan, a national law against obesity came into effect in 2008. Under this law companies and local governments must measure the waistlines of Japanese people between the ages of 40 and 74 during annual checkups. In Italy, phrases warning consumers of the damage caused by smoking are printed on cigarette packs, which are sold by a state monopoly. On some packages, the following sentences appear: ‘‘Smoking while pregnant harms your baby; smoking kills’’. These examples demonstrate government concern with lifestyles and the implementation of policies to change consumption habits. In so doing, many governments heavily influence individual’s choices. What determines this intrusive government interest? By reasoning backward, government interest produces a conflict between individual and social choices that originates in two factors: the rising cost of health care in many Western nations and the assumption that bad lifestyles may negatively affect labour productivity. Zargosky (2005) demonstrated a large negative association between Body Mass Index (BMI) and White females’ net worth, a smaller negative association for Black women and White males and no relationship for Black males in the US. He also found that individuals who lose small amounts of weight experience small changes in their net worth, but those who lose large amounts of weight have improved financial positions. For households in Sierra Leone, Strauss (1986) found a highly significant relationship between caloric intake and labour productivity, providing solid support for the nutrition-productivity hypothesis. The marginal effect on productivity decreases drastically as calorie consumption increases but remains positive at moderately high levels of intake. One result from this situation is a decrease in the effective price of food that is more significant for households that consume fewer calories. In general, it can be argued that health has a positive effect on the labour productivity of individuals. Thus, lifestyle choices, such as smoking and drinking,

1

Financial Times Cash incentives seen as helping nation’s health, 11 April 2009.

20

G. Coppola

of individuals are of interest to society and, also to firms because they affect labour productivity. Ultimately, lifestyle choices generate externalities, a term that indicates possible conflicts that have not been resolved by the market. Externalities affect labour productivity and healthcare costs above and beyond the level for which companies are responsible. This relatively new concept of externalities is explained by Sassi and Hurst (2008): ………Lifestyle choices, as many other forms of consumption, may produce external effects. There are immediate externalities that derive directly from acts of lifestyle consumption, such as passive smoking, violent and disorderly behaviour associated with alcohol abuse, or traffic accidents resulting from reckless driving. There are also deferred externalities, which are generated through the link between lifestyle choices and chronic diseases. Once chronic diseases emerge, and in some cases even before they emerge (e.g., when important risk factors such as hypertension or obesity begin to manifest themselves), the individuals affected will become less productive, possibly entirely unproductive, they will make a more intensive use of medical and social services, which may be publicly funded, they may require care by members of the family and friends. Conversely, a reduced life expectancy may mean a less prolonged use of publicly funded medical and social services at the end of life, as well as reduced pension payments, which are not themselves externalities, but would translate into a less onerous fiscal burden and therefore less distortion in the way the economy works. All of these phenomena involve externalities (negative the former, positive the latter) on society at large, family and friends, which can be attributed at least to some extent to the lifestyle choices originally made by the individual. The extent to which externalities can be associated with lifestyle choices depends, of course, on the strength of the link between lifestyles and disease, i.e., by the increase in the risk of developing a chronic disease associated with adopting a particular lifestyle.

These arguments are not new. John Stuart Mill (1859) wrote the book On Liberty to fight against laws that would limit individual freedom. In the nineteenth century in Great Britain, social degradation phenomena such as alcoholism were prevalent. Several social movements asked the government to implement prohibitionist measures to halt these phenomena. Mill contrasted these ideas, arguing that what happens inside a person’s body or mind is that person’s private business, not the business of society and certainly not the business of the government ………over himself, over his own body and mind, the individual is sovereign.

2.3 A Micro-Model In the following section, I define lifestyle and develop a micro-funded model that explains the relationship between health and income and the effects of income on health.

2 Health, Lifestyle and Growth

21

First, let us suppose that an economy produces 3 goods: 2 commodities for consumption, x and z, and Capital, K. The saving rate, s, is exogenous and constant. According to the Grossman model (1972), health capital and the demand for health have been widely modelled in the economic literature. Contoyannis and Jones (2004) developed a static model of lifestyle and health production. In this model, (1) income is assumed to be endogenous, but there is no direct influence of lifestyle or health on wages; (2) health affects consumer utility, unlike Grossman’s dynamic model (1972) in which health is considered to be a stock that produces flows of pecuniary and non-pecuniary benefits as a result of investments; (3) health is a result of a production function in which the inputs are a vector of goods, a vector of exogenous influences on health and a vector of unobservable influences on health; and (4) the money budget constraint and the time constraint close the model. As a result of these conditions, the following are maximised: consumer utility, using a Lagrangian function, the Marshallian demand for goods and the level of consumer Health. Contoyannis and Jones (2004) set the Health Production Function (HPF) equal to H ¼ hðC; XU ; UH Þ

ð2:1Þ

where H is a measure of individual health; C is a vector of M goods; XU is a vector of exogenous variables that influence health; and UH is a vector of unobservable influence on health. In this article, the model of Contoyannis and Jones (2004) is augmented to produce a model with 2 equations: (1) the consumer utility function and (2) the health production function.

2.3.1 The Consumer Utility Function The consumer utility function is assumed to be a Cobb Douglas function in which health, h, is an input. For this reason, health affects the consumer utility function (among others, Kip Viscussi and Evans 1990). The other 2 inputs are the commodities x and z. The utility function can be written as U ðh; x; zÞ ¼ ha xb zd

ð2:2Þ

Where a; b and dare the elasticities of h, x and z, respectively. Equation a 0 may be considered as the self-assigned weight of a consumer to his/her own health. If a ¼ 0; health is not important to the consumer. On the contrary, if a [ 0; then health is important. b; d \ [ 0: If b\0 or d\0; x or z is not a good but a ‘‘bad’’ for the consumer (Varian, 1992).

22

G. Coppola

This is a static model. There is no rational addiction, but positive elasticity values indicate that a consumer knows the commodity’s ophelimity.

2.3.2 The Health Production Function Contoyannis and Jones (2004) assume that the utility of a consumer depends on a set of goods C, Health H, XU ; a vector of observable exogenous influences on U and lU , a vector of unobservable influences or U, defined by U ¼ uðC; H; XU ; lU Þ

ð2:3Þ

Furthermore, HPF depends on a set of goods C, a vector of observable exogenous influences on H, XH ; and a vector of observable exogenous influences on H, lH : The vector H is defined by H ¼ hðC; XH ; lH Þ

ð2:4Þ

It is assumed that the consumption of a commodity can improve, worsen or have no effect on the health of a consumer. Consumption can worsen health in the case of smoking, alcohol and drugs. For simplicity, let us assume that every commodity can only better or worsen the health of a consumer. Further, no commodities can positively impact health in small quantities and negatively impact health in stronger doses.2 It is also assumed that x improves health, while z worsens health. The commodity x may be defined as the virtuous, or sustainable, good, and z may be defined as the harmful good. Health also depends on the initial level of health ðh0 Þ; public health ðWÞ; time t and a stochastic component e: The Health Production Function is hðx; z; h0 ; w; t; eÞ ¼ xq zc h0 we/t ee

ð2:5Þ

The function can be split into two parts: xq zc can be interpreted as a consumer’s activity, and h0 we/t ee can be attributed to other factors. To simplify the model, we use the relation X ¼ h0 we/t ee : The HPF is then hðx; z; h0 ; w; t; eÞ ¼ Xxq zc

ð2:6Þ

In this health production function, there is one input, x, with a positive marginal productivity. This assumption is compatible with the neoclassic production function theory (Gravelle and Rees 1992). The term ðq cÞ is equal to the elasticity of scale and can be positive, negative or null. Let h ¼ q c: Each input is

2

The ancient Romans said ‘‘In Medio stat Virtus’’. In the model that hypothesis doesn’t matter for each single good.

2 Health, Lifestyle and Growth

23

assumed to exhibit a decreasing return; thus, 0\q\1; 0\c\1; and, therefore, 1\h\1: Sassi and Hurst (2008) related individual lifestyles to individual behaviours that affect health. Contoyannis and Jones (2004) also defined lifestyle ‘‘as a set of behaviours which are considered to influence health and are generally considered to involve a considerable amount of free choice’’. If h [ 0; consumption increases positively affect health, while for h\0; consumption increases negatively affect health. With h ¼ 0; consumer behaviour has no effect on health. For this reason, the parameter h may be defined as the Lifestyle Return to Scale (hereafter LRS). Before maximising the consumer utility function, hðx; z; h0 ; w; t; eÞ ¼ Xxq zc is substituted into U ðh; x; zÞ ¼ ha xb zd to yield U ðh; x; zÞ ¼ Xxaq zac xb zd

ð2:7Þ

U ðh; x; zÞ ¼ Xxaqþb zdac

ð2:8Þ

or

The elasticity with respect to x becomes aq þ b; and the elasticity with respect to z becomes d ac: The commodity x (or z) will be consumed only if aq þ b [ 0 (or d ac [ 0;) that is, if the relative elasticity is positive. The other properties of the utility function are the same as before. Hence, the choice to consume x (or z) depends on 3 parameters: (1) the elasticity of utility with respect to x (or z) b (or d), that is, the weight conferred by the consumer to the commodity x (or z); (2) a, the importance of health to the consumer; and (3) the benefit (or damage) of x (or z) on health b (or c ). For example, an individual will consume a medicinal (x) only if the positive health impact ðaqÞ is given a value greater than the elasticity with respect to x (b). On the contrary, a consumer will not drink alcohol if alcohol is not preferred ðd\0Þ or if alcohol is enjoyed ðd [ 0Þbut he assumes the negative impacts on health ðacÞ to be greater than the elasticity of alcohol’s utility ðd ac [ 0Þ: A consumer may decide to drink alcohol even if the dangerous health effects are known. Further, consumers who are aware of the damage of smoking may continue to smoke.3 Following this approach, alcohol consumption depends also on factors other than the level of a consumer’s education. Generally, by including health in the consumer utility function, the consumption of commodities that benefit health increases, while the consumption of goods that cause damage decreases because a [ 0.

3 See for example Berger and Leigh (1989) and Kenkel (1991) for the relationship between schooling and health. See also Avitabile (2009) for the relationship between health and information.

24

G. Coppola

2.3.3 The Utility Maximization Problem: The Optimal Choice of x, z and h Let X ¼ 1: The consumer’s budget constraint is px x þ pz z ¼ cy; where px ; pz are the prices of the goods; y is the per capita income used for consumption, y ¼ YL ; c is the average propensity to consume ð0\c\1Þ; and L is the population. The consumer maximizes utility when maxx;z xaqþb zdac such that px x þ pz z ¼ y.4 Recall that aq þ b [ 0 and d ac [ 0: Optimally solving the Lagrangian maxx;z L ¼ Uðx; zÞ kðpx x þ pz z yÞ; where k is the Langrage Multiplier, the quantities of commodities consumed are (Mas-Colell et al. 1995) x¼

aq þ b cy b þ d þ að q cÞ p x

ð2:9Þ

z¼

d ac cy b þ d þ aðq cÞ pz

ð2:10Þ

The weight of health, a; increases the consumption of ‘‘virtuous’’ goods and reduces the consumption of harmful goods. Optimally, the health level is aq þ b cy q d ac cy c ð2:11Þ h¼ b þ d þ að q cÞ p x b þ d þ aðq cÞ pz or

q

c

ð pz Þ c ð2:12Þ ðcyÞðqcÞ ðpx Þq q aqþb Equation 2.11 is the health demand function, where bþdþa ðqcÞ and c dac are the shares of good x and good z, respectively, weighted for bþdþaðqcÞ h¼

aq þ b b þ d þ að q cÞ

d ac b þ d þ að q cÞ

their relative health elasticities. The level of health and the price of virtuous good are negatively correlated. If the price of good x increases (or decreases), it worsens (or improves) the level of health. Conversely, health improves (or worsens) if the price of z increases (or decreases). The health elasticity with respect to income is q c ¼ h; the parameter LRS. Unlike the parameters that can have only one sign, this parameter may be positive or negative. If q c ¼ 0; income growth does not affect the level of health. If q c\0; income negatively affects health. If q c [ 0; income positively affects health. 4

This approach may be considered as a generalization of Wagstaff’s model (1986). See Appendix for details.

2 Health, Lifestyle and Growth

25

Thus, income growth does not always positively affect health. The sign and the degree to which income affects health depend on the parameter h: A proxy or Index of a consumer’s Lifestyle (LI) may be given by the weighted average of the quantity of commodities consumed for the consumer’s health elasticity. This variable follows the relationship 1\LI\1 and is given by aq þ b d ac LI ¼ q c ð2:13Þ b þ d þ að q cÞ b þ d þ að q c Þ Therefore, LRS h is a crucial variable in the model because it indicates the attitude of a consumer, based on preferences and opportunities, toward leading a particular lifestyle. Thus, h ¼ q c partially and indirectly reflects consumer preferences because the health production function contains only those commodities that consumers prefer or can purchase.5

2.4 Comparative Static: The Health Multiplier In the previous section, the effects of income on health were described. Assuming the existence of a representative agent, Eq. 2.11 can be rewritten as h Y ð2:14Þ h¼t L q c c ðpZ Þ aqþb dac where m ¼ bþdþa c ðqcÞ bþdþaðqcÞ ðpx Þq A production function with a constant return to scale and for which both technology and health are labour augmenting is assumed. This may be a Cobb Douglas Production Function (i.e. Weil 2005; Sala-i-Martin 2005). Y ¼ K a ðAhLÞ1a

ð2:15Þ

or

5

Three issues should be highlighted here. First, in this simple consumer model, choices are made between two commodities. In reality, a commodity may be not consumed for three reasons, the first two of which were outlined previously: (1) the consumer does not like a commodity; (2) even if a commodity is liked, the health damage caused by the commodity may be greater than the commodity’s utility, preventing consumption of the commodity; and (3) the relative price of a commodity may be greater than income, preventing consumption of the commodity. In the first two cases, the commodity is not consumed as a result of free choice. In the second case, this choice may be difficult. In the third case, price and income limits restrict access to the commodity. In this article, we consider only the case in which individuals consume both commodities.

26

G. Coppola

Y ¼ L

a K ðAhÞ1a L

ð2:16Þ

From the system given by Eqs. 2.14 and 2.15, the impacts of a ‘‘health shock’’ ðDmÞ and an ‘‘income shock’’ ðDAÞ on health and income can be quantified. Solving this system yields the effects in terms of elasticity. The results are reported in Scheme 1 for 0\h\1: Scheme 1 On health

On income 1\h\1

d log h 1 ¼ d log m 1 hð1 aÞ d log h hð1 aÞ ¼ d log A 1 hð1 aÞ

Health shock Income shock

In terms of elasticity, the health multiplier is equal to

d log Y ð 1 aÞ ¼ d log m 1 hð1 aÞ d log Y ð 1 aÞ ¼ d log A 1 hð1 aÞ

1 1hð1aÞ

for health and

ð1aÞ 1hð1aÞ

for income. For 0\h\1; both multipliers are positive and greater than one. If a lifestyle is positive, a health shock will more strongly affect health and the growth of labour productivity. The effect of a health shock on income depends positively on both LRS and income elasticity, with respect to labour or labour income share. hð1aÞ The effect of an income shock on income is equal to 1h ð1aÞ ; while the effect

ð1aÞ on health is equal to 1h ð1aÞ : In this case, for 0\h\1; both effects are greater than zero. The question of whether a health shock more strongly affects income or an income shock more strongly affects health (Weil 2004) can now be answered: both technological and health shocks produce the same effect on income. This effect is ð1aÞ 1 equal to 1h ð1aÞ ; which is greater than ð1 aÞ because 1hð1aÞ [ 1:

A health shock has an impact on health equal to 1hð11aÞ ; which is greater than ð1aÞ 1hð1aÞ because 0\h\1: ð1aÞ 1hð1aÞ ; which is greater than

an income shock

A health shock has an impact on

hð1aÞ income equal to an income shock on health 1h ð1aÞ : The effect that a reduction of the price of commodity x has on health can be quantified. The multipliers for income and health are

d log h d log h d log t 1 q ¼ ¼ d log px d log t d log px 1 hð 1 a Þ

ð2:17Þ

2 Health, Lifestyle and Growth

27

d log Y d log Y d log t ð1 aÞ ¼ ¼ q d log px d log t d log px 1 hð1 aÞ

ð2:18Þ

The effect of a price shock for x on health is equal to 1hð11aÞ q: It depends both on the multiplier

1 1hð1aÞ

and negatively on the parameter q; which is the

ð1aÞ health elasticity of x. The impact on income is 1h ð1aÞ q; which is the product of

the impact of the price of x on health, 1hð11aÞ q; and the impact of health on income, ð1 aÞ. This impact has a negative sign because a reduction in the price of x positively affects health and income. Scheme 2 lists these health multipliers. Scheme 2 Health shock Income shock

On health dh 1 ¼ Yh dm 1 hð1 aÞ dh hð1 aÞ h ¼ dA 1 hð1 aÞ A

On income dY ð1 aÞ Y ¼ dm 1 hð1 aÞ t ð1 aÞ Y dY ¼ dA 1 hð1 aÞ A

2.5 A Growth Model with Health In the comparative static framework, the level of population is constant. In this section, I present a Growth Model that includes health as an input factor and also considers lifestyle. Interactions between health and economic growth are complex (Morand 2005). Listing all of these interactions is beyond the scope of this paper. One of the positive effects of economic growth on health is the possibility of consuming higher quantities of better quality goods. Another consequence of economic growth is that technological progress positively impacts medical technology and care. Therefore, health may have a positive impact on Growth through several channels. First, health may positively affect labour productivity (Marshall 2006). In Chapter V, titled ‘‘The Health and Strength of the Population’’, in Book 4 of the Principles of Economy, Marshall wrote that strength, physical, mental and moral, are the basis of industrial efficiency, on which the production of material wealth depends; while conversely the chief importance of material wealth lies in the fact that, when wisely used, it increases the health and strength, physical, mental and moral, of the human race.

For Streeten (1994), one of the reasons for promoting human development is that a well-nourished, healthy, educated, skilled and alert labour force is the most important productive asset.

28

G. Coppola

In the Neoclassic Economic Growth literature, many models consider health as a factor of growth. Lòpez-Casasnovas et al. (2005) and Rivera and Currais (1999a) used a conditional convergence regression in which the growth of per capita income is a function of steady-state determinants. Assuming that health is an important determinant of an enhanced labour force, they showed that health affects income growth both positively and significantly. In another article (Rivera and Currais 1999b), investment in health significantly explained variations in the output of human capital, even in countries with presumably high levels of health. Heshmati (2001) extended the model of Mankiw et al. (1992) by incorporating health. The results show that Health Care Expenditures positively affect economic growth and the speed of convergence. For Morand (2005), increasing longevity may incentivise agents to increase investments in capital and human capital and thereby reinforce economic growth. One of the main goals of this article is to consider the effects of individual lifestyles on economic growth. Let us now consider a Solow Growth Model (Solow 1956) with a constant saving rate (s), diminishing returns of capital ð0\a\1Þ and labour, Labour-augmenting technology and constant returns to scale. We assume a Cobb Douglas production function: Y ðtÞ ¼ K ðtÞa ðAðtÞLðtÞÞ1a

ð2:19Þ

where K(t), A(t), L(t) are capital, level of technology, and labour, respectively. Let us assume all the hypotheses of the Solow’s Growth Model. Technological progress and the population growth rate are exogenous and d ln AðtÞ d ln LðtÞ constant: ¼ g; ¼ n: dt dt Assuming that health is a labour-augmenting factor (Weil 2005; Sala-i-Martin 2005), the production function becomes Y ðtÞ ¼ K ðtÞa ðAðtÞhðtÞLðtÞÞ1a

ð2:20Þ

h Y h ¼ ty ¼ t h ¼ tY h Lh ; L

ð2:21Þ

With h

this becomes 1a Y ðtÞ ¼ K ðtÞa AðtÞtY ðtÞh LðtÞð1hÞ

ð2:22Þ

1h1a a ð1aÞ Y ðtÞ ¼ K ðtÞ1hð1aÞ AðtÞtLðtÞð1hÞ

ð2:23Þ

or

Equation 2.23 can be rewritten as

2 Health, Lifestyle and Growth

29

ð1hÞð1aÞ a 1 1 1hð1aÞ Y ðtÞ ¼ K ðtÞ1hð1aÞ AðtÞ1h t1h LðtÞ Equation 2.24 shows a constant return to scale because Substituting a1 ¼

a 1hð1aÞ

and a2 ¼

ð1hÞð1aÞ 1hð1aÞ

This

is a

Solow’s

Model,

and

aþð1hÞð1aÞ 1hð1aÞ

¼ 1.

¼ 1 a1 yields

1a1 1 1 Y ðtÞ ¼ K ðtÞa1 AðtÞ1h t1h LðtÞ

d ln A2 ðtÞ dt

ð2:24Þ

ð2:25Þ

the ‘‘new’’ technological rate

is

1

d ln AðtÞ1h dt

1 1h g

¼ ¼ At equilibrium, the Income growth rate and per capita income growth rate are d ln YðtÞ 1 gþn ð2:26Þ ¼ dt 1h d ln YðtÞ LðtÞ dt

¼

1 g 1h

ð2:27Þ

The income level is 1 Y ðt Þ 1 ¼ ðAð0ÞvÞ1h e1hgt LðtÞ

s g 1h þ n

a ð1hÞð1aÞ

ð2:28Þ

The Health growth and Health level are d ln YðtÞ d ln hðtÞ 1 LðtÞ ¼h g ¼h dt dt 1h 1 1h

h h 1h 1hgt

hð t Þ ¼ t ð A ð 0Þ Þ e

s g 1h þ n

ð2:29Þ

ð1hahÞð1aÞ ð2:30Þ

The first result is that even if the Solow model with health remains an exogenous growth model, the parameter LRS positively affects per capita income growth and the level of income per capita at the steady state. For example, if h ¼ 0:5; the income growth rate is equal to 2g; while a negative LRS ðh\0Þ results in a per capita income growth rate that is less than that of technological progress. Thus, a ‘‘good’’ lifestyle can improve economic growth, while a ‘‘bad’’ lifestyle can slow growth. The second result is that health increases more slowly than per capita income. Unlike economic growth, the rate of health growth may be positive, null or negative, depending on the sign of the LRS parameters. For h\0; economic growth negatively affects health, which worsens.

30

G. Coppola

Fig. 2.1 The effect of LRS h on the growth rate of per capita income with a constant technological growth rate (g = 2%)

The second scenario may be the case of a ‘‘health-poverty trap’’. Such a trap was noted in Russia; male life expectancy, which can be considered a health indicator, plummeted by 7 years from 1989 to 1994 because of high levels of alcohol consumption (UNDP 2010). The results of the model can be explained differently. Per capita income growth depends on the product of the technical progress parameter and LRS, both of which are exogenous. Technical progress is considered to be ‘‘manna from heaven’’. In fact, the aim of the endogenous growth theory is to identify those factors and mechanisms that could be controlled by the government to ensure higher and more durable economic growth. In the model presented in this article, another exogenous parameter, the Lifestyle Return to Scale, h, impacts economic growth. In the introduction, we presented several cases in which governments have attempted to control lifestyles. This type of governmental behaviour has several possible explanations. Equation 2.27 can be re-written as g ð2:31Þ h ¼ 1 YðtÞ d ln LðtÞ dt

Suppose that the government has established a target for its economic growth rate, denoted by ^y ; that can be controlled by technological progress g. If ^y [ g; the economic growth rate fixed by the government is greater than technical progress, and the government can strive for higher growth by controlling or trying to modify lifestyles, which are one of the channels that transmit the effects of economic growth to health. Conversely, if technological progress is high, lifestyles may not be important because technological progress can ensure a high level of economic growth. Further, health growth is equal to economic growth multiplied the parameter h. As shown in Fig. 2.1, with a fixed a technological rate g, the effects of economic growth on health improvement depend on the value of h. Lower values of h, result in weaker links between economic production and health improvements. This relationship could become negative for h \ 0. Hence, economic growth is a

2 Health, Lifestyle and Growth

31

necessary but not sufficient condition for improving health. If a government’s priority is to improve health rather than economic growth, then LFS must be positive. In this scenario, the conflict between public and private interests can become stronger.

2.6 Conclusions In this article, I attempted to formalize what Jean Anthelme Brillat-Savarin, the author of Philosologie of taste Brillat-Savarin 2004, wrote two hundred years ago: (1) animals feed themselves; men eat, but only wise men know the art of eating, and (2) the destiny of nations depends on the manner in which they are fed. The crucial hypotheses of the model are that (1) individuals can rationally choose to consume goods that negatively affect health, (2) individuals are co-producers of their health and (3) health positively affects labour productivity. First, I developed a consumer micro-model with health and two goods, both of which are positively correlated to the Consumer’s Utility. Health is the output of a consumer’s production function with the two commodities as inputs. The first commodity has a positive impact on health, while the second one has a negative impact. The result is that the elasticity of consumer health with respect to income, referred to as Lifestyle Return to Scale and denoted by the parameter h is equal to the algebraic sum of the health elasticity with respect to commodities. It may be positive, negative or neutral. In opposition to health’s role as a normal good, as reported by Wagstaff (1986), health can also be an inferior good in this model. Second, I computed health multipliers. The impacts of a health shock on health and income depend on labour share and are higher if the Lifestyle Return to Scale is positive. Third, the micro-behaviour function was introduced in the Solow growth model in which the return to scale is constant. The most important results are that (1) Lifestyle Return to Scale affects economic growth (the growth of income per capita is higher than the technical progress if LRS is positive), and (2) health improvement depends on the parameter LRS, h, it is lower than economic growth, and it may be negative, even if economic growth is positive. In fact, the existence of a health poverty trap in which economic growth diminishes health can be demonstrated. In conclusion, lifestyle is another aspect of society that governments can attempt to control or regulate.For this reason, the Aristotelian concept of intermediates is useful in the definition of virtue. In Nicomachean Ethics (Aristotle 2009), Aristotle explains that drink or food which is above or below a certain amount destroys the health, while that which is proportionate both produces and increases and preserves it (Nicomachean Ethics, Book 2, Chap. 3).

32

G. Coppola

The absence of this notion of intermediate in individual behaviours negatively affects society, and possible responses may be social exclusion and government interventions in the private sphere, as condemned by Mill (1859), the intensities of which may vary until an ethical Hegelian state is established. Acknowledgements I would like to thank seminar participants at the University of Salerno and the University of Pescara. I am grateful for many helpful comments, received during the course of those presentations. Thanks also to Adalgiso Amendola, Alberto Bennardo, Dimitrios Christelis, Floro Ernesto Caroleo, Marcello D’Amato, Sergio Destefanis, Fernanda Mazzotta, Niall O’Higghins, Carmen Pagliari and Giuliana Parodi for their very useful suggestions. The usual disclaimer applies.

A.1 Appendix A.1.1 A Generalization of Wagstaff’s Model Starting with Michael Grossman’s Model (1972) and Wagstaff (1986) developed a one-period model of demand for health. The four hypotheses of the model include the following: (1) an individual’s health is determined by the consumption of health inputs hð xÞ ¼ xq ; (2) preferences are non-lexicographic: individuals desire health but not above everything else; (3) individuals also consume other comdU modities that have a positive cost for consumers, so U ¼ uðh; zÞ with dU dh ; dz [ 0 2

2

and d Udhðh;zÞ ; d Udzðh;zÞ \0; and (4) consumers have limited economic resources or budget constraints: px x þ pz z ¼ Y; where px and pz are the prices of commodities x and z, respectively, and Y is the income. Assuming a Cobb Douglas Utility function and a Health production function hð xÞ ¼ xq ; the Wagstaff Model can be formulated with the following formulas: U ðh; zÞ ¼ ha zd

ð2:32Þ

hð xÞ ¼ xq

ð2:33Þ

p x x þ pz z ¼ Y

ð2:34Þ

where 0\a\1 and 0\d\1 are the utility elasticities with respect to x and z, respectively, and 0\q\1 is the elasticity of h with respect to x. This is a special case of the Consumer’s model (Sect. 2) with b ¼ 0: The commodity x is not in the Consumer’s utility function with ðc ¼ 0Þ; thus, z does not affect health. The solutions can be obtained from two different methods. The first was proposed by Wagstaff: max U ðh; zÞ ¼ ha zd s:t: px hq þ pz z ¼ Y h;z

ð2:35Þ

2 Health, Lifestyle and Growth

33

In this case the Budget Constraint is not linear. The consumer chooses between health and z. The second possible solution is max U ðx; zÞ ¼ xqa zd s:t: px x þ pz z ¼ Y x;z

ð2:36Þ

The consumer chooses the quantities of x and z that maximize utility. Both methods yield the same solutions: x¼

aq Y d þ aq px

d Y d þ aq pz q aq Y h¼ d þ aq px z¼

ð2:37Þ ð2:38Þ ð2:39Þ

The main differences include the following: (1) in the Wagstaff model, Health can only be a normal good because dh dy [ 0 (conversely, in the model proposed in this paper, Health may also be an inferior good), and (2) this result depends on the lifestyle of the consumer.

References Aristotle (2009) Nicomachean Ethics (trans: Ross WD). World Library Classical Books Avitabile C (2009) Does conditionality matter for adults’ health? Evidence from a randomized experiment. CSEF Working Paper 222 Berger MC, Leigh JP (1989) Schooling, self-selection, and health. J Human Res 24(3):433–455 (Summer, 1989) Brillat Savarin JA (2004) The phisiology of taste. Kessinger Publishing, US Contoyannis P, Jones AM (2004) Socio-economic status, health and lifestyle. J Health Econ 23:965–995 Gravelle H, Rees R (1992) Microeconomics, 2nd edn. Longman Group UK Limited, England Grossman M (1972) On the concept of health capital and the demand for health. J Political Econ 80(2):223–255 Heshmati A (2001) On the causality between GDP and health care expenditure in augmented Solow growth model. SEE/EFI Working paper series in Economics and Finance 423 Kenkel D (1991) Health behaviour, health knowledge and schooling. J Political Econ 99:287–305 Kip Viscussi W, Evans WN (1990) Utility function that depend on health status. Am Econ Rev 80(3):353–374 Lòpez-Casasnovas et al (2005) Health and economic growth: findings and policy implications In: López-Casasnovas G, Rivera B, Currais L (eds) Cambridge MIT press, Cambridge Maddison A (2003) The world economy: historical Statistics. Paris Development Centre of the Organisation for Economic Co operation and Development, Paris Mankiw NG, Romer D, Weil N (1992) Contribution to the empirics of economic growth. Quarterly Journal of Economics 107:407–437

34

G. Coppola

Marshall A (2006) Principles of Economics Cosmio Inc. original published (1890) by Prometheus Book Mas-Colell A, Whinston M, Green J (1995) Microeconomic theory. Oxford University Press, Oxford Mill JS (1859) On liberty. Republished by Agora Publication inc (2003) Morand OF (2005) Economic growth, health, and longevity in the very long term: facts and mechanisms. In: López-Casasnovas G, Rivera B, Currais L (eds) Health and economic growth: findings and policy implications. Cambridge MIT press, Cambridge Riley JC (2005) Estimates of regional and global life expectancy 1800–2001. Population Dev Rev 31(3):537–543 Rivera B, Currais L (1999a) Economic growth and health: direct impact or reverse causation? Appl Econ Lett 6(11):761–764 Rivera B, Currais L (1999b) Income variation and health expenditure: evidence for OECD countries. Rev Dev Econ 3(3):258–267 Sala-i-Martin X (2005) On the health–poverty trap. In: López-Casasnovas G, Rivera B, Currais L (eds) Health and economic growth: findings and policy implications. Cambridge MIT press, Cambridge Sassi F, Hurst J (2008) The prevention of lifestyle related chronic diseases: an economic framework. OECD Publishing 32 Solow R (1956) A contribution to the theory of economic growth. Q J Econ 70:65–94 (February) Strauss J (1986) Does better nutrition raise farm productivity? J Political Econ 94(2):297–320 Streeten P (1994) Human development: mean and ends. Am Econ Rev 84(2), In: Papers and Proceedings of the 106th annual meeting of the American Economic Association, May 1994, pp 232–237 UNDP (2010) Human development report. Second printing Palgrave Macmillian, London Varian H (1992) Microeconomic analysis, 3rd edn. Norton, New York Wagstaff A (1986) The demand for health: a simplified Grossman model. Bull Econ Res 38:1 Weil DN (2004) Economic growth. Addison-Wesley, Reading Weil DN (2005) Accounting for the effect of health on economic growth. NBER Working Paper w1145 Zargosky J (2005) Health and wealth: the late-20th century obesity epidemic in the US. Econ Human Biol 3(2):296–313

Chapter 3

A Comparative Analysis of Literacy Rate in Contributing to Social Exclusion Insights Edgardo Bucciarelli, Carmen Pagliari, Fabrizio Muratore and Iacopo Odoardi

Abstract Our contribution aims to analyze the relationship between the phenomena of social exclusion and literacy levels, and to consider the significant implication of this relationship on economic growth. The goal, which is reached by analyzing cross-country data, is firstly to describe the situation of social exclusion with the use of specific socio-economic variables, and secondly to compare the levels of education and training for each considered country. These two phenomena are mutually influenced, as a low level of literacy in affecting the employment status precludes the possibility to enter and operate freely in society, while poverty and persistent social exclusion of a person or family make difficult to address appropriate educational and training paths. Therefore, our study has rejoined two issues which influence almost all decisions adopted by policy makers, especially in the Western world. The opening issue is the level of education, which should constitute the human capital of a country, through appropriate investment, and the second one is the relational condition of social dynamics, which highlight the so-called social capital. Together these two types of intangible capitals constitute a strong support for the long-term development of a country. Our quantitative analysis is also addressed to detect differences and peculiarities among the

E. Bucciarelli (&) C. Pagliari Dipartimento di Metodi Quantitativi e Teoria Economica, Università d’Annunzio di Chieti-Pescara, Pescara, Italy e-mail: [email protected] C. Pagliari e-mail: [email protected] F. Muratore I. Odoardi Scuola Superiore, Università d’Annunzio di Chieti-Pescara, Pescara, Italy e-mail: [email protected] I. Odoardi e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_3, Springer-Verlag Berlin Heidelberg 2012

35

36

E. Bucciarelli et al.

different national realities, with the ultimate purpose to recognize which socioeconomic variables affect more directly the processes of education. Keywords Social exclusion Literacy rate Maximum likelihood Hierarchical cluster JEL Codes C82

Education

Human capital

I21 O50 O57 Y10

3.1 Introduction and Related Literature The level of education for an individual represents, especially in modern Western societies, a potential effective indicator of the level of working capacity, the socalled productivity, but also implies other less observable, but equally significant skills. These are the essential knowledge for living in society as dynamic players, without incurring the risk of being excluded from the traditional and contemporary activities of society. But today there are many limitations to the normal training of an individual, and in some cases, to the groups of disadvantaged people, for example, the problem of poverty which affects the most vulnerable groups such as the children, the elderly, and the ethnic minorities. Children living in families at social risk or below the poverty line see foreclosed the best channels of education and training, and find obstacles in the first years of their training school. This situation could affect children’s subsequent paths of training and prevent them from achieving adequate levels of education. For adequate levels of education we mean the level of training which is considered optimal from the perspective of each culture and society in order to access to stable employment as well as to participate in society. However, if households living below the poverty line are the most at risk, we must remember those with only one parent, those who live in neighbourhoods with little social organization, and even those belonging to ethnic or religious minority. Moreover, the strong positive correlation between illiteracy and social exclusion is underlined by the wide set of definitions of literacy which have been given over time. An example is provided by the General Conference of UNESCO 1978 which provides several definitions of literacy. The first one states that ‘‘[…] a person is literate who can with understanding both read and write a short simple statement on his everyday life’’, while the second one is more complex: ‘‘[…] a person is functionally literate who can engage in all those activities in which literacy is required for effective functioning of his group and community and also for enabling him to continue to use reading, writing and calculation for his own and the community’s development’’. The two definitions range from the simple ability of writing and reading to the reasoning skills of adequately using the knowledge acquired. We are not referring merely to the increasing of more cognitive abilities which allow to reach higher levels of understanding, but also to the manner in which those who have these skills can

3 A Comparative Analysis of Literacy

37

exploit them to coexist in the stable and balanced society they belong to. This particularly includes the definition offered by the International Adult Literacy Survey (see Statistics Canada and OECD 2000): ‘‘The ability to understand and employ printed information in daily activities, at home, at work and in the community—to achieve one’s goals, and to develop one’s knowledge and potential’’. The definition used by OECD (2010) adds that ‘‘[…] differences in levels of literacy matter both economically and socially: literacy affects, inter alia, labour flexibility and quality, employment, training opportunities, income from work and wider participation in civic society’’. Furthermore, the IALS Final Report (Statistics Canada and OECD 2000) shows a number of relationships about literacy: countries with higher literacy scores have higher labour force participation and shorter work hours whereas countries with a high proportion of adults with low prose skills have lower GDP per capita. There is a correlation between a low GDP per capita of a country and the rising proportion of adults with low prose skills. Public intervention may partially cope and withstand the mentioned problems related to the lack of literacy, by trying to obtain a plentiful and robust level of national human capital in the long run. This is now a fundamental part of many economic studies concerning growth and development. But the complex phenomenon of educating and training individuals cannot be regarded only as a general educational teaching base of general educational foundations. This means not only prepared trainers, and adequate public and private investment, but also a rational project in order to develop in all individuals a sense of common rules through education (see, for example, Grossman and Kim 1997). These are the foundations for building a solid social capital which is itself a determinant of socio-economic development processes (see, among others, Gradstein and Justman 2002, who have also included the education processes in the relationship between social capital and economic growth). People who cannot reach satisfactory levels of literacy either by choice or compulsion are the weakest in society and go toward the phenomenon of social exclusion. The problems caused by poor literacy can be observed through two joint events. The first one concerns the limited prospects of finding safeguarded employment opportunities, the second one considers the defective contribution which people offer to the socio-economic system through their labour productivity. The difficulty in finding a job involves a certain separation from society. This separation deprives individuals of their ability to fully exploit the possibilities offered them by contemporary world. The individuals, though not in poverty, will not have the opportunity to be engaged in specified conducts which are characteristics of each society, as constitutional elements of general needs. Therefore, this path is degenerative, and leads to conditions which are close to relative poverty which tends to become a phenomenon that is passed between generations. In fact, poor parents cannot guarantee the optimal education for their children (see, among others, Grossmann 2008; Galor and Moav 2004), and socially excluded parents represent a source of social exclusion for their children. The education level of a person can be found either through the years of school attended, or through the qualifications acquired. In particular, in this work, we observe the

38

E. Bucciarelli et al.

influence which some typical variables, related to the processes of social exclusion, can have on the average level of education of a country. It is known that the economic conditions may affect the ability to achieve high average levels of education, but also the strength of education and social capital constitute a lever of development (see, among others, Temple and Johnson 1998). Gradstein and Justman (2002) analyze the importance of a broad common cultural basis to start those basic functions of effective interaction between individuals. The spread of education, and, in general, of literacy may be a fundamental policy goal in trying to establish virtuous processes from an economic standpoint. Essentially, the main aim of our empirical study is to demonstrate how certain variables,1 which characterize social exclusion, have obvious influences on the literacy rate. In order to pursue this aim, we gather data concerning 30 countries and then we briefly compare the effect of the above mentioned variables on the level of economic development of the countries considered. In our particular case study, the literacy rate represents the percentage of people with the ability to read and write without specifying the level of education. The framework of the paper is organized as follows. In Sect. 3.2, we start by showing the methodology used. Then we proceed in Sect. 3.3 with a preliminary empirical investigation on literacy rate, and on the expectancy variables. In Sect. 3.4, the results of the multivariate regression model are presented. The findings of the factor analysis with maximum likelihood method and VARIMAX rotation are shown in Sect. 3.5. Furthermore, Sect. 3.6 focuses on multivariate regression model for literacy rate with maximum likelihood components. In Sect. 3.7, we represent and analyze the countries in a hierarchical cluster with variables related to literacy. In Sect. 3.8 we focus our attention on specific American literacy policies which are particularly significant for our study. We present our conclusions in Sect. 3.9.

3.2 Methodology Applied for Empirical Analysis on Literacy Rate for 30 World Countries In this section we analyze literacy rate by using the least-squares method, the factor analysis with maximum likelihood method and the hierarchical cluster. The method of least squares is a standard approach to the approximate solution of over determined systems (Moser 1996; Freund 2003), i.e. sets of equations in which there are more equations than unknowns. Least squares means that the overall solution minimizes the sum of the squares of the errors made in solving every single equation. The most important application is in data fitting: the best fit in the

1 We have collected the cross-country dataset from UNESCO and World Bank. For the crosscountry analysis we provide, see also Levine and Renelt (1991, 1992); de Gregorio and Lee (2002); Hoover and Perez (2004).

3 A Comparative Analysis of Literacy

39

least-squares sense minimizes the sum of squared residuals, a residual being the difference between an observed value and the value provided by a model (Björck 1996). We consider a linear regression model: hence, the model comprises a linear combination of the parameters: f ðxi ; bÞ ¼

m X

bj /j ðxi Þ

ð3:1Þ

j¼1

where the coefficient /j is functions of xi. Letting: Xij ¼

of ðxi ; bÞ ¼ /j ðxi Þ obj

ð3:2Þ

Then we can see in case (3.2) the least square estimate (or estimator, in the context of a random sample), b is given by: b ¼ ðX T XÞ1 X T y

ð3:3Þ

Consequently, we consider the response variable as a linear function of the regressors: yt ¼ b0 þ b1 x1 þ þ bk xk þ et

ð3:4Þ

In our paper, we analyze literacy rate (LR) as a function of school enrolment consisting in pre-primary (PRE), primary (PRI), secondary (SEC) and tertiary (TER); GDP real growth rate (GDP); long term unemployment rate (LUR); public spending on education as percentage of GDP (PSE) and public spending on education as percentage of government expenditure (PGE); children out of school (COS). In our analysis, we consider different levels of education: pre-primary, primary, secondary, and tertiary education. These levels are generally found in similar ways in many countries, and generally go from kindergarten to the so called higher education. The analysis regard 30 countries2 from different continents, and we consider mostly countries of Organizations for Economic Cooperation and Development (OECD). The model we provide is the following: LR ¼ b0 þ b1 PRE þ b2 PRI þ b3 SEC þ b4 TER þ b5 GDP þ b6 LUR þ b7 PSE ð3:5Þ þ b8 PGE þ b9 COS Therefore, in the next study we apply factor analysis in order to reduce the high number of variables of model (3.5). Factor analysis is a statistical method used to

2 The countries considered by this research are: Australia, Austria, Belgium, Bulgaria, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Republic of Korea, Latvia, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, Turkey, United Kingdom, United States.

40

E. Bucciarelli et al.

describe variability among observed variables in terms of a potentially lower number of unobserved variables called factors (Mardia and Kent 1980). In other words, it is possible, for example, that variations in three or four observed variables mainly reflect the variations in a single unobserved variable, or in a reduced number of unobserved variables. Factor analysis searches for such joint variations in response to unobserved latent variables. The observed variables are modeled as linear combinations of the potential factors, plus an error term. The information gained about the interdependencies between observed variables can be used later to reduce the set of variables in a dataset. Factor analysis is related to principal component analysis (PCA) but it is not identical, because PCA performs a variancemaximizing rotation of the variable space and it takes into account all variability in the variables. In contrast, factor analysis estimates how much of the variability is due to common factors (communality). The two methods become essentially equivalent when the error terms in the factor analysis model (the variability not explained by common factors, see below) can assume the constant variance (Shore and Sheppard 1996). In fact, PCA involves a mathematical procedure that transforms a number of possibly correlated variables into a smaller number of uncorrelated variables called principal components. The first principal component accounts for as much as possible of the variability in the data, and each succeeding component accounts for as much as possible of the remaining variability. PCA is mathematically defined as an orthogonal linear transformation that transforms the data to a new coordinate system such that the greatest variance by any projection of the data comes to lie on the first coordinate (called the first principal component), the second greatest variance on the second coordinate, and so on. PCA is theoretically the optimum transform for given data in least square terms (Boersma and Weenink 2001). For a data matrix, XT, with zero empirical mean (the empirical mean of the distribution has been subtracted from the data set), where each row represents a different repetition of the experiment, and each column gives the results from a particular problem, the PCA transformation is given by: ð3:6Þ Y T ¼ X T W ¼ VRT where the matrix R is an m-by-n diagonal matrix with non-negative real numbers on the diagonal and WRT V is the singular value decomposition (svd) of X. In our analysis we tried to apply principal component analysis but we do not find good results. Therefore we use maximum likelihood method with VARIMAX rotation considering an extraction based on eigenvalue with eigenvalue greater than 0.5. Indeed, in this kind of choice we find a good response of components. Before trying this method, we analyze literacy rate also considering an extraction base on eigenvalue greater than one but we discover a low result. In fact maximum likelihood estimation (MLE) is a popular statistical method used for fitting a statistical model to data, and providing estimates for the model’s parameters (Besset 2001). The method of maximum likelihood corresponds to many well-known estimation methods in statistics. The sample mean is then the maximum likelihood estimator of the population mean, and the sample variance is a close approximation to the maximum likelihood estimator of the population variance. For a fixed set of data

3 A Comparative Analysis of Literacy

41

and underlying probability model, the method of maximum likelihood selects values of the model parameters which maximize the likelihood function. Maximum likelihood estimation gives a unified approach to estimation, which is welldefined in the case of the normal distribution and many other problems. In applying MLE we suppose that there is a sample x1, x2, …, xn of n independent and identically distributed (i.i.d.) observations, coming from an unknown distribution f0(). It is however known that the function f0 belongs to a certain family of distributions ff ðjhÞ; h 2 Hg; called the parametric model, so that f0 ¼ f ðjh0 Þ: The value h0 is unknown and is referred to as the ‘‘true value’’ of the parameter. It is desirable to find some ^ h (the estimator) which would be as close to the true value h0 as possible. Both the observed variables xi and the parameter h can be vectors. The variables xi may be non-i.i.d., in which case the formula below for joint density will not separate into individual terms; however the general principles would still apply. To use the method of maximum likelihood, one first specifies the joint density function for all observations. For i.i.d. sample this joint density function will be f ðx1 ; x2 ; . . .; xn jhÞ ¼ f ðx1 jhÞ f ðx2 jhÞ f ðxn jhÞ

ð3:7Þ

We may extend the domain of the density function so that the density is also a function of the parameter h. Then, for a given sample of data with observed values x1, x2, … xn, the extended density can be considered a function of the parameter h. This extended density is the likelihood function of the parameter: Lðhjx1 ; . . .; xn Þ ¼ f ðx1 ; x2 ; . . .; xn jhÞ ¼

n Y

f ðxi jhÞ

ð3:8Þ

i¼1

However, in general, the likelihood function is not a probability density. In fact, it does not need to be an additive function, thus it is not a probability measure. In practice it is often more convenient to work with the logarithm of the likelihood function, ln L; called the log-likelihood, or its scaled version, called the average log-likelihood: n X 1 ln f ðxi jhÞ ‘^ ¼ ln L ð3:9Þ ln Lðhjx1 ; . . .; xn Þ ¼ n i¼1 Indeed, ‘^ estimates the expected log-likelihood of a single observation in the model. The method of maximum likelihood estimates h0 by finding a value of h ^ that maximizes ‘ðhjxÞ: This method of estimation is a maximum likelihood estimator (MLE) of h0: ^ ^ 1 ; . . .xn Þ ð3:10Þ hmle ¼ arg max ‘ðhjx h2H

By applying maximum likelihood estimation, we can identify a point estimate referred to each country considered in the analysis. Furthermore, for the application of maximum likelihood we use VARIMAX rotation (Jakeman 2008). We know that in quantitative methods a VARIMAX rotation is a change of coordinates

42

E. Bucciarelli et al.

used in principal component analysis and factor analysis which maximizes the sum of the variances of the squared loadings. That is, it seeks a basis that most economically represents each individual, so that each country can be well described by a linear combination of only a few basic functions: 0 !2 1 p k X P k X X X c 4 2 RVARIMAX ¼ arg max@ ð3:11Þ ðKRÞij ðKRÞij A R p j¼1 i¼1 j¼1 i¼1 where c ¼ 1 for VARIMAX. Variance maximizing rotation is often used in surveys to see how groups of countries measure the same phenomenon. In our case how the 30 countries considered distance themselves. Ultimately, we use hierarchical clustering, which is a method of cluster analysis which seeks to build a hierarchy of clusters. Strategies for hierarchical clustering generally fall into two types, but we use only agglomerative method that is a bottom up approach, in fact each observation starts in its own cluster, and pairs of clusters are merged as one moves up the hierarchy. In general, the merges and splits are determined in a greedy manner. The results of hierarchical clustering are usually presented in a dendrogram. A measure of dissimilarity between sets of observations is required in order to decide which clusters should be combined (for agglomerative) (Székely and Rizzo 2005). In most methods of hierarchical clustering, this is achieved by use of an appropriate metric (a measure of distance between pairs of observations), and a linkage criteria which specifies the dissimilarity of sets as a function of the pairwise distances of observations in the sets. We chose Euclidean distance: rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ X ð3:12Þ ðai bi Þ2 ka bk2 ¼ i

We use the linkage criteria maximum or complete linkage clustering that determines the distance between sets of observations as a function of the pairwise distances between observations: max fdða; bÞ : a 2 A; b 2 Bg

ð3:13Þ

We apply the method of hierarchical cluster in order to identify any consistency between the countries considered, and part of the same cluster as well as the heterogeneity between clusters.

3.3 Preliminary Empirical Investigation on the Literacy Rate and Expectancy Variables The following analysis refers to the 30 countries some of which are OECD members, over the period of time 2007–2009. The countries we chose to include in our cross-country analysis are those that represent manifold and diversified

3 A Comparative Analysis of Literacy

43

Table 3.1 Average, lowest and highest values for dependent and expectancy variables Indicator Mean Lowest Highest Literacy rate GDP r.g.r. Pre-primary Primary Secondary Tertiary l.t.u.r. p.e.p. GDP p.e.p. GOV c.o.s.

98.25 0.003 0.891 1.058 0.785 0.413 1.760 0.051 0.132 0.030

88.70 (TR) -0.041 (LV) 0.159 (TR) 0.549 (PT) 0.183 (NL) 0.01 (NL) 0.045 (MX) 0.019 (AT) 0.034 (PT) 0.0002 (LU)

99.80 0.045 1.233 1.577 1.479 0,968 5.988 0.086 0.281 0.416

(LV) (PL) (ES) (ES) (AU) (U.S.) (SK) (JP) (LV) (U.S.)

Mean, maximum and minimum values for dependent and expectancy variables. In parentheses is given the international symbol of the various countries

socio-economic contexts. However, the principal context studied is that of the most advanced economies, which can be divided in numerous subgroups of remarkable interest. Indeed, in the European case it is possible to observe the substantial differences between the Scandinavian and the Mediterranean countries, or the dissimilarities in the American States with Anglo-Saxon or Latin origin, in theme of economy, education and society. In further investigations, we do not exclude to consider and compare socio-economic contexts characterized by a very dissimilar situation, like rich and poor countries, but it is essential to consider the relevant problem of quality and updating of so numerous variables, as those used in this paper. The first analysis performs the calculation of average values, lowest and highest values reported for each considered variable. Variables that we analyze are: literacy rate,3 GDP real growth rate (GDP r.g.r.), pre-primary, primary, secondary and tertiary schooling as percentage of total population by every education levels of every country,4 long term unemployment rate (l.t.u.r.), public spending as percentage of GDP (p.e.p. GDP) and public spending as percentage of government expenditure (p.e.p. GOV), children out of school (c.o.s.) as percentage of total children out of school. In the next page we show the results obtained from descriptive analysis (Table 3.1). Literacy rate has a mean value equal to 98.254%, this average is referred to the 30 countries considered in our analysis. The minimum is 88.7% and is represented by Turkey (TR), the maximum value is represented by Latvia (LV) and is almost 100%. GDP real growth rate has its minimum value in Latvia, the same regards other countries such as Hungary (HU) and Ireland (IE) which show negative

3 The literacy rate represents the percentage of people with the ability to read and write, and in this case without specifying the level of education. 4 In our analysis we take into consideration different levels of education, such as pre-primary, primary, secondary and tertiary education: these levels are generally found in similar ways in many countries, and represent usually from pre-primary until the so called higher education (tertiary).

44

E. Bucciarelli et al.

values, while the highest refers to Poland (PL). The average GDP real growth rate is near to zero, because in recent years there has been stagnation and recession of the economy. The several education levels get the minimum value by Turkey, Portugal (PT) and Netherlands (NL) in the four categories considered. Slovakia (SK) shows the highest value reported for long unemployment rate. Furthermore, Austria (AT) and Portugal are the countries with lowest usage rates of public spending in education. Finally the variable related to children out of school shows its maximum value in United States (US) and this percentage is always conditioned by a high population abundance, and we note simultaneously a high degree of school drop-outs. In the next section we begin the analysis of the literacy rate applying multivariate regression method, maximum likelihood method and clustering analysis to highlight the differences in education levels in recent years in the countries considered.

3.4 Results from the Multivariate Regression Model A first analysis of the variables related to education highlights a high variability in the data due to different density of population of countries considered. For example, United States has only in the pre-primary school more than 12 million of people, whereas the U.S. population exceeds 300 million inhabitants in 2009 (U.S. Census Bureau data). This indicates that data are based also on country’s population size. The same is true for Mexico, Japan and France, which show a high abundance of individuals enrolled in different educational levels. In contrast, countries which show less members are represented by Iceland, Ireland, Latvia, Luxembourg and Slovenia. This variability is also confirmed by symmetry and kurtosis indices showing high amounts; this indicates an asymmetry of these distributions compared to the normal one. We consider the parameter as normal when most of data are distributed in average value, however in our results values are distributed more even in areas of tail. Hence, literacy rate and children out of school reveal an asymmetry of information, but it is only apparent as there are slight differences in their percentages. Some countries have 99% of schooling, but many others are below 90%. While long term unemployment rate, public spending and GDP real growth rate report almost a normal distribution with low variability. By examining interdependence between variables we notice strong groupings of countries in relation to education levels and economic growth. Considering different levels of education relate to each other, we note high direct relations. However, we note that in most countries the literacy rate has percentages between 90 and 99% while the long run unemployment rate shows a clear grouping of countries in the lowest percentages of literacy rate. This indicates that with an increasing of the degree of education levels the long-run unemployment rate decreases, as with a higher level of education attained, individuals find more job opportunities. Our aim here is to analyze literacy rate relationship with other variables presented in Sect. 3.3. For this purpose we use a multivariate regression analysis and the

3 A Comparative Analysis of Literacy

45

mathematical model we propose is the (3.4), and in terms of variables is the (3.5). Before applying the (3.5) we perform other types of investigations in order to understand the influence of literacy rate compared to single macro groups5 under analysis: we have first investigated literacy rate as a function only of education variables, then literacy rate only with variables related to social exclusion. Observing our results of statistical tests concerning the significance of the parameters and considering the best fitting to data, we choose to consider model as a whole. Indeed, we use also education variables as percentages of total education in different countries in order to identify the optimal model applied; resulting data are not satisfactory for the significance of the parameters in any kind of linear combination we calculate. Thus, the combination of economic growth, education levels and social exclusion variables leads to a better result than analytical models considered singly. We analyze the presence of multicollinearity using the tolerance analysis that is equal to:tol ¼ k R2 . When the values of this analysis are close to zero there is a combination of other endogenous variables covered by our study. Moreover, we use the variance inflation factor (VIF) that is the reciprocal of tolerance, in order to achieve further confirmation of the result found with the tolerance analysis. The VIF has a threshold value equal to 0.10 and we can check if there is collinearity among independent variables considered. Tests for investigating endogeneity have not been reported in this study since they have provided results in the standard and not particularly interesting for our goals. In addition, an important test we apply to detect the presence of autocorrelation between independent variables is the Durbin-Watson statistics which is used in the erratic component (et) of a regression analysis. The related formula is equal to: PT ðet et1 Þ2 ð3:14Þ d ¼ t¼2PT 2 t¼1 et We are also interested in applying the heterogeneity test which concerns the system of variables we have selected in order to describe the phenomenon of social exclusion in terms of illiteracy as our data set is quite multifaceted (many variables, different from each other and variability between countries). For identifying any heterogeneity we use the ANOVA test (analysis of variance), which locates the dispersion of observed values compared with an expected value, so that we can detect the possible homogeneity between the variables; the resulting variances should be quite similar to each other otherwise it will confirm the hypothesis of heterogeneity of the system. Finally, all the explanatory variables should be uncorrelated with the residual terms (uncorrelated residual values), that is the residue of an observation must be independent from the residual value resulted from another observation. The results are shown in Tables 3.2 and 3.3.

5

For macro groups we mean groupings of the endogenous variables based on their similarities.

0.138 0.064 0.023 1 0.124 -0.020 0.114 0.043 0.163

0.119 -0.016 1 0.023 -0.316 -0.152 -0.003 0.295 0.020

Secondary

-0.094 1 -0.016 0.064 -0.160 -0.008 -0.125 0.068 -0.353

GDP r.g.r. Pre-primary Primary Secondary Tertiary l.t.u.r. p.e.p. GDP p.e.p. GOV c.o.s.

1 -0.094 0.119 0.138 0.004 0.132 -0.233 -0.057 -0.118

Table 3.2 Correlation index among explanatory variables GDP r.g.r. Pre-primary Primary 0.004 -0.160 -0.316 0.124 1 0.277 0.299 -0.310 0.491

Tertiary 0.132 -0.008 -0.152 -0.020 0.277 1 -0.010 -0.294 -0.136

l.t.u.r. -0.233 -0.125 -0.003 0.114 0.299 -0.010 1 -0.190 0.348

p.e.p. GDP

-0.057 0.068 0.295 0.043 -0.310 -0.294 -0.190 1 0.052

p.e.p. GOV

-0.118 -0.353 0.020 0.163 0.491 -0.136 0.348 0.052 1

c.o.s.

46 E. Bucciarelli et al.

3 A Comparative Analysis of Literacy Table 3.3 Univariate ANOVA Sum of squares GDP r.g.r.

Pre-primary

Primary

Secondary

Tertiary

l.t.u.r.

p.e.p. GDP

p.e.p. GOV

c.o.s.

b. g. w. g. Total b. g. w. g. Total b. g. w. g. Total b. g. w. g. Total b. g. w. g. Total b. g. w. g. Total b. g. w. g. Total b. g. w. g. Total b. g. w. g. Total

0.006 0.004 0.009 0.960 0.490 1.450 0.731 0.176 0.906 0.962 2.009 2.971 1.002 2.521 3.523 18.158 42.584 60.743 0.003 0.006 0.009 0.061 0.044 0.105 0.017 0.197 0.214

47

Df

Mean square

F

Sig.

13 19 32 13 19 32 13 19 32 13 19 32 13 19 32 13 19 32 13 19 32 13 19 32 13 19 32

0.000 0.000

2.331

0.046

0.074 0.026

2.859

0.019

0.056 0.009

6.079

0.000

0.074 0.106

0.700

0.742

0.077 0.133

0.581

0.840

1.397 2.241

0.623

0.807

0.000 0.000

0.848

0.612

0.005 0.002

2.021

0.079

0.001 0.010

0.126

1.000

The acronym b.g. stands for between groups, while w.g. identifies within groups

First to apply the model, in Table 3.2 we report as preliminary analysis the correlation index among the explanatory variables. This analysis helps us to find the possible connections between variables and in almost all cases we find a weak correlation between variables. In fact, the results are negative and positive but in any way close to zero, and these values do not confirm concordances or discordances for each variables, but a substantial indifference. We confirm in this first analysis an independence between variables, since we note no specific connections between them. We observe only a positive correlation relating to tertiary level of education and children out of school (0.491), although this discordance is not so strong to verify a relationship between them. Moreover, in Table 3.3, in order to stress the possible heterogeneity between variables we calculate the analysis of variance (ANOVA test) for all independent variables to get the significant values comparing with a F (Fisher) test. In this case we note a significant result between groups for GDP real growth rate, pre-primary and primary levels of schooling, and public spending as percentage of GDP.

48 Table 3.4 Determination index and Durbin-Watson test

E. Bucciarelli et al. R

R-square

Adjusted R square

Std. error of the estimate

Durbin-Watson test

0.544

0.295

0.020

0.022

2.094

Furthermore, in Table 3.4 we note a low determination index which stands at 0.295 and this value does not give a normal adaptation for observed data and in addition we report the value of Durbin-Watson test that is equal to 2.094. In this case the test confirms the presence of no autocorrelation in the residuals. In Table 3.5 we present the set of parameters resulting for each variable we analyze and we show the standard error for each parameter. The results reveal that most of coefficients are positive and so there is a direct relationship between the type of endogenous variables and the literacy rate. This means that with the increasing of students enrolled in pre-primary and primary levels of education the level of literacy rate in each country grows more than exponentially. Parametric values obtained are small for those coefficients related to variables of social exclusion. Finally, the significance of coefficients refers to pre-primary education with a margin of error of 0.001. It can be seen as a significant and high parameter obtained by the constant. And, further, in Table 3.5 we have the results of collinearity diagnostics and every tolerance values are higher than 0.5 for each independent variable. These results confirm the presence of no multicollinearity in our observed values. Therefore, this confirms that independent variables we consider do not experience any kind of linear combination and do not show any kind of influence between them (Table 3.6). We observe that parameter estimates in secondary and tertiary education levels tend to be equal to zero. Effectively in secondary level of school the value is -0.005 and in tertiary level is -0.0003. Hence, in these two cases there are not increases in the literacy rate, and so the dependent variable is constant during the time for the latter two cases. The two parameter estimates do not confirm an inverse relation because they are closer to zero, so it means that low percentage from secondary and tertiary education levels have a constant effect on the literacy rate. The lowest levels (pre-primary and primary) obtain a substantial growth in the literacy rate. Consequently the literacy rate is most influenced by these previous categories (pre-primary and primary levels of schooling) because they represent the most widespread activity of education. In fact, we refer to OECD countries wherever these levels are reached by almost the entire population as compulsory. In a dichotomic context, in which one is able or not to read and write, to be literate or illiterate, the lowest levels of education have a greater influence on the amount of literacy rate and therefore they are more representative of our analysis. For example, imagine such a hypothetical backward economy, in which there is only a low stock of physical capital, coeteris paribus: in this state of the world it will be sufficient even the simplest investments to increase the stock of capital (quantity rather than quality) in order to enable the first processes of economic growth. Then the economy can move to investments with higher quality in capital (e.g. ICTs) and/or in the labour component (e.g. human capital,

3 A Comparative Analysis of Literacy

49

Table 3.5 Multivariate analysis of literacy rate and collinearity statistics Model Coefficients Collinearity statistics Indicators

Constant GDP r.g.r. Pre-primary Primary Secondary Tertiary l.t.u.r. p.e.p. GDP p.e.p. GOV c.o.s.

b

Std. error

t

Tolerance

VIF

95.560 20.932 4.860 2.508 -0.005 -0.0003 0.072 15.315 8.806 4.577

3.653 24.391 1.977 2.547 0.070 0.001 0.311 25.631 7.698 6.474

26.160 0.858 2.458 0.985 -0.071 -0.03 0.231 0.598 1.144 0.707

0.836 0.825 0.795 0.910 0.514 0.794 0.762 0.754 0.522

1.197 1.213 1.257 1.099 1.946 1.259 1.312 1.327 1.915

Mean square

F

Sig.

5.012 4.676

1.072

0.419

Table 3.6 Multivariate ANOVA of literacy rate Model Sum of squares Df Regression Residual Total

45.110 107.552 152.662

9 23 32

knowledge-based economy, etc.) as well as in our study we find that the so-called higher education affects other variables and has not a strong effects on the literacy rate, such as productivity at work and the ability to produce innovations and, overall, to promote a general economic development. In general, the results concerning parameter estimates are not satisfying because we do not find any significance with variables related to education or those linked with social exclusion (except for the pre-primary level of schooling) and this requires to gather only the most interesting information and sufficient to describe the subject under analysis. So, these results lead us to apply other analysis, and in particular we choose the factor analysis. Indeed, we note there is a redundancy of information due to the presence of variables that are quite similar to each other: e.g. in the case of education we have a number of variables available which explain it and thus with factor analysis we try to represent the education with a synthetic indicator. Consequently, as a first step we use the maximum likelihood method. In the next section we report the findings.

3.5 Factor Analysis with Maximum Likelihood Method and VARIMAX Rotation In the following contents the factor analysis is carried out, and we apply the maximum likelihood method in order to identify new components of the model [see (3.10)]. Therefore, we use this method with VARIMAX rotation [see (3.11)]

50 Table 3.7 Initial and extraction values with maximum likelihood method

E. Bucciarelli et al. Indicator

Initial

Extraction

GDP r.g.r. Pre-primary Primary Secondary Tertiary l.t.u.r. p.e.p. GDP p.e.p. GOV c.o.s.

0.164 0.175 0.205 0.090 0.486 0.206 0.238 0.246 0.478

0.999 0.379 0.238 0.139 0.745 0.269 0.206 0.338 0.924

considering an extraction based on eigenvalue, with the latter greater than one. The results are shown in Table 3.7. In Table 3.7 we describe the initial and extraction of each variable according to the maximum likelihood method. We note that initial value of each variable increases with extraction. In particular, there is a good extraction in variable referred to GDP real growth rate and children out of school. Once we calculate the extraction we try to reduce variability of the phenomenon and eliminate an unavailing information in the distribution with maximum likelihood method to try to separate the meaning of the literacy rate from potential redundant relationships. The set of findings is shown in Table 3.8. In Table 3.8 we report initial eigenvalues, the extraction as sum of squares and the rotation method we use: this method shows a normal dispersion of the variability of the phenomenon under study. We find four components which allow us to explain almost 67% of the variability of literacy. Once we calculate eigenvalues we report the results of the factor matrix. In factor matrix (see Table 3.9) we show the importance of each component on every endogenous variable analyzed. The first component explains in particular the dynamics of GDP (GDP r.g.r.), while the second one relates to public spending of GDP (p.e.p. GDP). The third is related to tertiary education level. The fourth is related with pre-primary and secondary education levels. Then we apply VARIMAX rotation and we obtain the results as shown in Table 3.10. VARIMAX rotation (see Table 3.10) is similar to the results obtained from factor matrix, but results of variables are different: in fact, the first component is related to the long term unemployment rate (l.t.u.r) and the public expenditure of GDP (p.e.p. GDP). While the second one concerns the GDP real growth rate (GDP r.g.r.). The third is related to tertiary education level and the fourth refers to the importance of education in pre-primary, secondary and tertiary schooling. Finally, we calculate the factor transformation matrix. The results are shown in Table 3.11. In the factor transformation matrix (see Table 3.11) we report results of factor matrix with VARIMAX rotation considering every principal component; thus, we note a little change in the distribution of the phenomenon. Once we calculate the

3 A Comparative Analysis of Literacy

51

Table 3.8 Initial eigenvalues, extraction sums of squared and rotation sums of squared Factor Initial eigenvalues

1 2 3 4 5 6 7 8 9

Extraction sums of squared loadings

Rotation sums of squared loadings

Total % of variance

Cumulative %

Total % of variance

Cumulative %

Total % of variance

Cumulative %

2.12 1.59 1.23 1.07 0.86 0.74 0.61 0.48 0.29

23.58 41.29 55.01 66.91 76.47 84.65 91.40 96.78 100.00

1.13 1.56 1.18 0.36

12.56 29.95 43.07 47.09

1.20 1.18 1.15 0.71

13.31 26.39 39.17 47.09

23.58 17.71 13.72 11.90 9.56 8.18 6.75 5.38 3.22

Table 3.9 Factor matrix

12.56 17.39 13.12 4.02

13.31 13.08 12.78 7.92

Factor Indicator

1

2

3

4

GDP r.g.r. Pre-primary Primary Secondary Tertiary l.t.u.r. p.e.p. GDP p.e.p. GOV c.o.s.

0.999 -0.093 0.119 0.138 0.003 0.132 -0.234 -0.057 -0.120

0.002 -0.381 -0.022 0.189 0.596 -0.063 0.359 -0.020 0.947

0.000 0.048 -0.473 -0.006 0.619 0.492 0.147 -0.543 -0.118

-0.004 0.472 0.016 0.291 0.083 -0.075 -0.033 0.200 0.004

Table 3.10 VARIMAX rotation

Factor Indicator

1

2

3

4

GDP r.g.r. Pre-primary Primary Secondary Tertiary l.t.u.r. p.e.p. GDP p.e.p. GOV c.o.s.

0.064 -0.029 -0.462 -0.035 0.608 0.504 0.140 -0.568 -0.112

0.976 -0.093 0.155 0.070 -0.149 0.109 -0.292 -0.035 -0.256

0.047 -0.593 0.024 -0.016 0.366 -0.047 0.280 -0.085 0.757

0.200 0.133 -0.018 0.364 0.468 -0.026 0.150 0.085 0.522

components with maximum likelihood method we apply multivariate regression analysis using components resulting from factor analysis. In the next section we report results of this new analysis.

52

E. Bucciarelli et al.

Table 3.11 Factor trasformation matrix

Factor

1

2

3

4

1 2 3 4

0.064 0.013 0.988 -0.137

0.977 -0.157 -0.078 -0.122

0.046 0.796 -0.096 -0.595

0.199 0.584 0.088 0.782

Table 3.12 Determination index and Durbin-Watson test

R

R square

Adjusted R square

Std. error of the estimate

Durbin-Watson test

0.404

0.263

0.044

2.1356769

2.005

3.6 Multivariate Regression Model for Literacy Rate with Maximum Likelihood Components Factor analysis is helpful in order to use resulting components for regression multivariate analysis of literacy rate. The first analysis is conducted considering four components. The results are shown in Table 3.12. In Table 3.12, we note a normal value for determination index, which is equal to 0.263. We consider also that in this model we include all components extracted by maximum likelihood method. We remark that Table 3.13 shows a significant parameter only for components 1, 3 and 4 (the constant is also significant). The value of Durbin-Watson test it is equal to 2.005. In this case the test confirms the presence of no autocorrelation in residuals values. Also in Table 3.13 we have the results of collinearity diagnostics and the tolerance values are in all cases independent and higher than 0.5. This confirms the presence of no multicollinearity in our values. The three components above mentioned have a significance for a margin of error of less than 0.05. This result is satisfactory for our analysis, because we are able to explain the dependence of literacy with only three variables. Consequently, we repeat multivariate regression analysis considering only the above mentioned three components. In Table 3.14, we show the value of determination index of the model. The result is similar to that reported in the previous analysis, standing at 0.263 but it is related to the use of only three components. In Table 3.15 parameters are significant for all three components calculated with tolerance of less than 0.05. Overall, we report a normal result of analysis, while these three components are sufficient to explain the phenomenon of literacy rate. As a result we consider that component 1 is a synthetic indicator of social exclusion of the countries considered, component 3 is an indicator regarding education level, while component 4 is the economic factor represented by GDP real growth rate. The chart in Fig. 3.1 reassumes the values of literacy rate, expected and residual values concerning the multivariate regression model.

3 A Comparative Analysis of Literacy

53

Table 3.13 Multivariate regression model for literacy rate with maximum likelihood components and collinearity statistics Model Coefficients Collinearity statistics Constant Regr. factor Regr. factor Regr. factor Regr. factor

1 2 3 4

b

Std. error

t

Sig.

Tolerance

VIF

98.255 -0.560 -0.023 -0.967 0.778

0.372 0.444 0.384 0.491 0.551

264.286 -1.261 -0.060 -1.970 1.412

0.000 0.218 0.953 0.059 0.169

0.990 0.983 0.815 0.808

1.011 1.018 1.226 1.238

Table 3.14 Determination index and Durbin-Watson test R R-square Adjusted R-square Std. error of the estimate

Durbin-Watson test

0.404

2.005

0.263

0.077

2.0986646

Table 3.15 Multivariate regression model for literacy rate with maximum likelihood principal components and collinearity statistics Model Coefficients Collinearity statistics Constant Regr. factor 1 Regr. factor 3 Regr. factor 4

B

Std. error

t

Sig.

Tolerance

VIF

98.255 -0.560 -0.964 0.774

0.365 0.437 0.480 0.537

268.947 -1.283 -2.008 1.441

0.000 0.210 0.054 0.160

0.990 0.823 0.820

1.011 1.215 1.219

Fig. 3.1 Literacy rate model with observed, predicted and residual values

54

E. Bucciarelli et al.

In Fig. 3.1 expected values of literacy rate are very similar to observed data. Determination index confirms a normal approximation of theoretical to real data. Moreover, the success of the model is confirmed by residual values which tend to zero and have very small variations, furthermore we note that residuals values are also distributed randomly: this indicate a good result of analysis. Finally, on horizontal axis we report international acronym of the countries: each value belongs to literacy rate values. In the figure we note three countries having a literacy rate lower than other countries considered, and we can see three light spikes downwards due to smaller values. The three countries are: Mexico, Portugal and Turkey. Once we identify the presence of a lower literacy rate for these countries we perform a clustering to show differences between countries.

3.7 Hierarchical Cluster with Maximum Likelihood Components for Literacy Rate In this section we analyze literacy rate with regard to component 1 and 3 extracted within the factor analysis by using maximum likelihood method. We consider a hierarchical cluster applying an Euclidean distance and a complete linkage clustering which are presented in methodology section as (3.12) and (3.13). By means of the dendogram (not reported here) we find 3 important clusters. The results are graphically described in Fig. 3.2. Figure 3.2 shows that the cluster analysis is applied to components 1 and 3 which have been extracted by using maximum likelihood method. The first component regards social exclusion variables such as long term unemployment rate and children out of school, while the third component represents education levels which consists in pre-primary, primary, secondary and tertiary schooling. It is possible immediately to notice that the U.S. position is characterized also thanks to their historic effort towards innovation and investments in human capital, and at the same time they can keep good positions in the field of social exclusion. Most of the countries concerned, almost all OECD, show similar levels of macro variable related to education, and instead differ for social exclusion. This means that since all countries with medium or high levels of income per capita always are able to have good levels of enrollment and attendance at school, but not everyone can deal effectively with social problems. In general, even if the theme of social exclusion is extremely complex, and it is possible to observe it even with the use of other variables, the countries on the left side of the figure reveal better results in this case because they are affected by the recent past in which the macroeconomic results have influenced some variables related, for example, to employment levels, which are certainly among the main causes of exclusion. An additional cause for exclusion taken by us into account is the rate of school non-attendance, it is certainly, as fully demonstrated by the specific literature, a strong limit for integration, and often precludes the possibility, during adulthood, to reach adequate

3 A Comparative Analysis of Literacy

55

Fig. 3.2 Hierarchical cluster with maximum likelihood components for literacy rate

levels of human capital: these are indispensable to achieve job and economic security, especially in advanced and competitive socio-economic systems such as those in analysis. Figure 3.2 also shows a cluster consisting of a single country that is Slovakia. This country presents a low literacy rate compared to other countries, and a low level of education. A second cluster is composed by Mexico, Portugal, Turkey, Greece, Belgium, Bulgary, Germany and Italy: these countries are located in two different areas, because they are closer to other countries considered, but in the same cluster because they have lower education levels than other countries, and different kinds of social exclusion component. The third cluster is represented by the greater part of countries, which report normal levels of education and social exclusion. In particular, we note United States in the third cluster and placed on the left-top of Fig. 3.2. It confirms a high level of consideration about the phenomenon of literacy rate with respect to the other countries: we have the highest level of education than other countries with a low degree of social exclusion. Hence, policy implications should take account of these complex relationships, and not only promote general literacy, but also support highest and specialized levels of education and training, especially encouraging synergies between education, social capital, and labor market. Indeed, in order to endorse the long-run economic growth and development we cannot consider only the mentioned fundamental aspects of literacy which an advanced society must have. Even those countries that today have low levels of social exclusion, as it has been calculated by us, if they do not make continuous and future investments in human capital, public and private, they will not enjoy advantages or benefits on globalized market quotas or niches which can provide a strong support to the national economy. On the other hand, contexts in which widespread education is more safe, because of established public and private institutions, cannot ignore social issues and they have to favor the so-called ‘‘social capital’’ of the entire population, as well as the acquisition of civil behavioral rules in addition to education and training, and

56

E. Bucciarelli et al.

especially to integrate the welfare of society in the necessary measures to take off economic processes. In conclusion, we can assert that literacy rate, considered as a factor of social exclusion, can be certainly affected by the social exclusion variables here considered. This conditioning cannot occur early in the analysis, because of the heterogeneity of data has led us to different results. As a consequence, overall both the results of multivariate regression model, and those related to maximum likelihood method and cluster analysis confirm an exclusion of Turkey, Mexico and Portugal due to their relatively lower education than other countries involved in the study. The condition of relative low education levels in to the last three countries leads them on a multidimensional process of progressive social disruption. Ultimately, among the possible cases we explain in the next section the case of the United States and main critical issues concerning Italy.

3.8 Policy Insights From Social Exclusion, Literacy and Skillfulness: The Italian Criticalities and the Case of the United States After having described in the previous sections the relationship between the phenomena of social exclusion and literacy levels, and having discussed both the situation of social exclusion with the use of specific socio-economic variables, and the levels of education and training with other variables for each considered country, it is now important to focus our attention on the combination of literacy policy and social exclusion. Recently there has been a general trend in economic research to reckon that any socio-economic problem can be blamed on poor literacy rates. If literacy rates could be increased, then every country would be a more enlightened, productive country. The theory is that if people are more skilled in reading, writing, and other so called basic skills then that will allow workers to be more productive. So that over the last decade, recommendations to increase literacy and human capital have become more frequent, and now they are a constant of the economic policy debate and beyond. In fact, our empirical evidences, as well as those of numerous other studies leave no doubt. For example, in terms of human capital stock, if we consider the rate of people holding a college degree, Italy looks like a backward country and continues to move back towards the more advanced countries. Recent studies conducted Statistics Canada and OECD (2003, 2005) show that the proportion of graduates in Italy compared to the population, even if increased from 9 to 14% between 1998 and 2008, has grown less than the other countries considered. In 2008, the gap between Italy and other countries has increased from 12 to 14% points in relation to the OECD average, and from 10 to 11 in relation to the EU average. In policy terms, this means that the discrepancy for the young Italian population is nowadays higher than for the

3 A Comparative Analysis of Literacy Table 3.16 Average literacy and numeracy scores of 16–65 year old, by country (SC and OECD 2003)

57

Literacy

Numeracy

Country

Score

Country

Score

Norway Canada Switzerland United States Italy

293 281 274 269 228

Norway Canada Switzerland United States Italy

290 272 270 261 233

people aged 55–64 years. However, the average years of education are still rather a vague measure of the total endowment of human capital. Even in light of the foregoing, a most appropriate measure is based on the collection of knowledge and skills of the population. Well, the proficiency of the Italian population over the age of compulsory education is lower than the proficiency of all countries collected in the latest survey edited by Statistic Canada and the OECD in 2003 and entitled ‘‘International adult literacy and life skills survey’’. This survey is an international comparative study conducted to provide participating countries with information about the skills of their adult populations. The survey measures the literacy and numeracy skills of a representative sample of 16–65 year old from five participating countries (Canada, Italy, Norway, Switzerland, and the United States). In this case, literacy is defined as the skills which are necessary in order to understand and use information coming from a text and other written formats. Numeracy refers to the skills required to manage mathematical demands of diverse situations. In the OECD’s issue, prose literacy and document literacy scores are combined into a single literacy score measured on a scale from 0 to 500 points. Numeracy scores also range from 0 to 500; U.S. adults have an average literacy score of 269 and a score of 261 in numeracy (see Table 3.16). The United States outperform Italy in literacy and numeracy, but is outperformed by Canada, Norway, and Switzerland in both skill areas. According to this survey, the Italian adult population, taken as a whole, does not have an appropriate functional literacy for the needs of an advanced country. For example, approximately 80% of Italians aged between 16 and 65 have a level of command on the mother tongue assessed substantially inadequate. In comparison, in countries with higher skills (e.g. Norway), this quantity does not exceed 30%, while it does not go beyond 50% in countries at intermediate positions (such as Canada, United States, and Switzerland). If these data reflect the relatively recent development of mass schooling in Italy, it seems that this is not the only cause of the inappropriate functional literacy of Italian people. In fact, the same OECD survey shows how in the age group between 16 and 25, Italy reveals a gap with other countries which is only marginally lower than that characterizing the group aged between 46 and 65. For younger generation, in the four PISA surveys conducted by the OECD between 2000 and 2009 (OECD 2001, 2006, 2009), Italy has always shown a significant delay in the skills of the 15 year old students in each of the areas investigated.

58

E. Bucciarelli et al.

If we compare PISA and other international surveys related to periods before the school situation, it appears evident that this delay comes from a gradual deterioration in the quality of the various school levels. The evidence is congruent with the common perception of a good quality of the Italian primary schools, and with a disappointing quality of the three years post elementary course. In effect, the elementary school in Italy appears to be more effective in facilitating the learning of reading and science, but not math, independently of the set of countries taken as a reference. In the Italian secondary school, the results in science are worse than the international average, and the outcomes in mathematics are much worse. Therefore, investment in knowledge is needed, especially in Italy. But it should also investigate into what and how to increase the skills which our society will ever be in need of. In this regard, especially in Italy, the debate has been ongoing for several years, and it involves the world of educational research, but also that of socio-economic studies. For example, one can mention the project ‘‘Partnership for twentyfirst Century Skills’’, a joint initiative of the Federal Government and some major U.S. companies, which is designed to act as a catalyst to help plug the gap between what individuals learn at various levels of education (education system) and what will be immediately useful to them in their workplace. It will be fundamental to integrate the mastery of concepts related to the traditional knowledge (languages, mathematics, science, economics, history, geography, law, art) with the twenty-first century skills which contribute to social inclusion: the practice of right thinking, the ability to solve problems, creativity, the positive attitude towards innovations, the capacity to communicate effectively, the openness to collaborate and work in teams. Moreover, it is desirable to rethink the traditional relationship between knowledge and know-how. For example, the overcoming of teaching methods based on a substantial separation between the learning of knowledge and their subsequent use in working life may require the introduction of significant changes in school organization as well as in methods of teaching. However, the initial situation of the Italian education system does not facilitate this change, because of the dynamics of schooling attainments and employment contracts which make it difficult to recruit new teachers trained in developing the twenty-first century skills, and because of the scarcity of funds earmarked for school policies. Nevertheless in Italy there is a need of training as well as of an effective system of incentives and valid selection criteria which can differentiate and provide appropriate recognition of merit and preparation. Ultimately, the revitalization of the Italian school system is a central theme of economic policy, and not just a matter of public finance and/or a subject of sectoral issue because investment in literacy and human capital goes beyond what is done in school-based educational processes. In Italy, indeed, the formative and educational aspects of apprenticeship are characterized by the gap between training activities arranged by the regions and business reality. In fact, there is also no practice of certification of skills acquired in the workplace. Therefore the Italian apprenticeship often becomes a mere bargaining tool which reduces labour costs and makes the relationship between employer and employee temporary. The difficulty to develop a real and effective

3 A Comparative Analysis of Literacy

59

training programs in the system of firms also depends on the particular nature of the Italian production system focused on small and micro enterprises. We must understand that more and more investments in literacy and human capital are a crucial asset to be cultivated throughout lifelong learning. This aim can be achieved, especially in countries such as Italy, particularly through radical changes in the educational provision. The literature on poverty and social exclusion has paid much attention to problems of local communities and this is highly compatible with the focus on area in some of the exclusion literature centered on literacy. Social exclusion is of increasing interest because it has gained a primary role in official documents and in the political debate in Europe and, more recently, in Australia, Canada and the United States. Hence, after having provided a conceptual foundation and having given some guidance such as the application of the concepts suggested here, it is now significant to highlight the public policy concerning this issue. Certainly social exclusion has long been seen as a multi-dimensional concept, and we consider central in the analysis we carried out to observe now the literacy policy in the American case among the 30 countries here considered. We believe that analyses of the social exclusion of American people, as well as other populations in the world, already delve far beyond income into other areas such as education, health, housing and social environment, including their overlaps with income poverty. The U.S. is therefore already collecting and analyzing a great deal of data on different aspects of social exclusion in which people and above all children have the potential to be excluded and considering how these overlap with each other. In fact, as we stated before, social exclusion has long been considered as a multi-dimensional concept and we wonder whether this is good or a bad news for the use of ‘exclusion’ in the U.S. socio-economic system, and we think that on the one hand it is bad because one cannot look at social exclusion as something that will drive completely new collection and analysis of data on various areas of people’s lives, as it has arguably done in some European countries. On the other hand it is good, as the data are there and there is much analysis on which to build. Those signing-up to the ‘intellectual’ motivation for the use of ‘exclusion’ would argue that the existing analyses and policies are no substitute for what could yet be attained. Those subscribing only to the ‘political’ reason might argue that the banning of exclusion would allow the existing work to penetrate yet further into the policy world. If social exclusion is to gain ground as a concept in the U.S. then those who seek to push it will have to think hard about the geographical definition of the society from which people can be excluded, and how this relates to the level at which anti-exclusion policy operates. In particular, we wonder if it is better to persist with a national characterization of social exclusion and, for the insights of our analysis, of school drop-outs. For example, as in some European countries, large differences in state-level incidence of cash poverty among American people result from switching from a national to a state-level poverty line, when defined in conventional ‘European’ terms (Micklewright 2002). Scholars such as Rainwater et al. (2001) show the effect of moving from a line of half the national median income to one of half the state median. The average absolute difference in child

60

E. Bucciarelli et al.

poverty rates is 4.1% points and the correlation between the two rates is 0.53. New Jersey and Arkansas, the richest and poorest states respectively with median incomes 25% above and 25% below the national figure, see their child poverty rates rise from 14 to 22% (New Jersey) and fall from 26 to 14% (Arkansas). Furthermore, there was a more than fourfold increase in permanent literacy and education exclusions between 1990 and 1997. Primary school exclusions have been rising faster from a low level, and sixty per cent of the excluded come from unemployed homes. Those in care are more likely to be excluded as are those with special needs literacy and education. In particular, in order to evaluate U.S. literacy policy, and then to suggest a set of guidelines to improve it, a reflection on how the meaning of literacy has changed from the 1930s to nowadays is fundamental.6 In fact, the development of literacy policy is affected by the way policy makers, employers, unions, and government define literacy. It is important to note with Fingeret (1992) that definitions of literacy share an emphasis on reading, writing, and, usually, computation skills but differ in their descriptions of the extent of skills and the criteria for application. For example, in one setting persons who can sign their name may be considered literate, whereas in an another setting literacy is viewed as the ability to read a complex manual and apply the information to a task (Fingeret 1992, p. 3).

The absence of a univocal definition of literacy as well as the fact that literacy is a social construct is underlined both by studies aimed to identify parameters to measure literacy performance and by researches focused on pedagogical issues. From a pedagogical point of view, Gillespie Marilyn in Many literacies: modules for training adult beginning readers and tutors (1990) gives evidence of the fact that the programmes to fight illiteracy have to consider the fact that learners do not share the same idea of literacy. From a statistical point of view, Bormouth in Reading literacy: its definition and assessement (1974) underlines how the first step to develop models for identifying performance criteria concerning literacy is to analyze the concept of literacy for the purpose of identifying the parameters that must be specified in literacy definitions. Following the study developed by Fingeret (1992), it is possible to identify four main views of literacy: (1) literacy as skills which considers literacy as ‘‘a set of discrete skills that exist regardless of context. This conception of literacy leads to focus on simple encoding and decoding skills: sounding out words and studying lists of letter sounds, syllable sounds, and words in isolation’’ (Fingeret 1992, p. 5); (2) literacy as tasks. This notion described by Northcutt et al. (1975) underlines the ability to use skills coming from literacy in order to cope with different tasks such as filling an application form without taking into account the setting where the literacy task is accomplished; (3) Literacy as social and cultural practices underlines the influence of the setting on succeeding in the literacy task; for example, filling an application form for a checking account is a different task from sitting in a literacy 6

For a wider historical and social description of literacy in U.S. it is useful to consult Carl F Kaestle et al. (1993).

3 A Comparative Analysis of Literacy

61

class so that ‘‘the ability to do the task is considered stable across situations and does not require only individual skill achievement’’ (Fingeret 1992, p. 6); and (4) Literacy as critical reflection and action highlights the importance to take into account both the ability to interpret the meaning of the written text as well as that to elaborate critically the information coming from the text. This approach can be represented by the works of Schwabacher (1989), Breen and Sobel (1991), and Mc-Clellan-Cason et al. (1992). The literacy policies have been influenced by these four different ways to define literacy so that governments developed their programs to fight illiteracy by starting from different conceptions of literacy. The historical survey A political review of international literacy meetings in industrialized countries presented by Jean-Paul Hautecoeur (1997) on the basis of the reports of 20 seminars and international conferences linked to UNESCO gives evidence of how literacy policies changed both in western Europe and north America from 1981 to 1994 without achieving a general consensus. Concerning the aim of our paper, it is relevant to focus our attention on the different literacy policies adopted by the USA government. On a general level, it is possible to note that the concept of literacy and the following literacy policy established by U.S. governments from 1930s to 1980s changed after 1991 when the National Literacy Act was signed by President George Bush. The main characteristics of the period from 1930s to 1980s were represented by the idea of literacy as skills and as tasks, by the close link between literacy goals and economic achievement, and the lack of a strong central governmental coordination of the literacy policies. Cook (1977) underlines that literacy is defined as the simple ability to read and write a text by the U.S. Census Bereau in the 1930s and 1940s. Moreover, from the outset in the 1960s, the enabling legislation for the federal literacy program contained language supporting literacy development as a means toward other ends. Conceived of as part of the larger war on poverty, literacy education was supposed to bring about personal economic self-sufficiency as well as contribute to the eradication of poverty in the United States (Fingeret 1992, p. 15).

The period from 1970s to 1980s puts the basis to the new tendency promoted by the National Literacy Act. In fact, in the 1970s the Right to Read was strongly asserted, and during 1980 s ‘‘several new organizations emerged to fulfill various functions related to information dissemination, staff development, technical assistance, research, and advocacy’’ (Fingeret 1992, p. 17) concerning literacy and, at the same time, the Presidential Literacy Initiative was introduced. These social and political events represent the background to the National Literacy Act which was one of decisive steps toward implementing AMERICA 2000 strategy. The National Literacy Act was significant because it ‘‘has the potential to change the focus from a crisis orientation to a long-term commitment. The act establishes the National Institute for Literacy, with relationship to the Departments of Education, Labor, and Health and Human Services. The act also provides for creating state or regional resource canters under the governor’s office rather than in the state department of education. The National Literacy Act, provides the potential

62

E. Bucciarelli et al.

leadership for a new infrastructure that addresses literacy education as an ongoing responsibility of a coordinated set of institutions’’ (Fingeret 1992, pp. 16–17). The National Literacy Act was a first decisive answer to the growing crisis of illiteracy in America described by Jonathan Kozol in his work ‘‘Illiterate America’’ (1986) because it faced the problem by promoting both a long-term strategy and a central coordination of the different regional and national efforts against illiteracy. This policy was confirmed in 1998 by the Adult Education and Family Literacy Act (AEFLA; Title II of the Workforce Investment Act) with provides States with funding for a variety of services to help adults develop basic skills such as reading, writing, math, English language competency, and problem solving. After having shown the positive potential of the literacy policy introduced since 1990s, it is important to focus our attention on a consideration which can minimize one of its possible negative potentials. Beder (1991) explains that ‘‘adult literacy education is a tapestry of diversity’’ (p. 152). He makes reference to the fact that learners come from different cultures and communities so that the central effort of coordinating the different literacy programmes have to try to preserve this diversity and not to promote cultural standardization.7 Hence it is important to involve the community where the learners live in the literacy education. For example, the National Illiteracy Action Project 2007–2011 promoted by the non-profit Talking Page Literacy Organization aims to create Community Literacy Collaborations. In fact, illiteracy is not only a school educational problem, but also a community problem. From this point of view, it is important to develop a participatory curriculum development process which takes into account and respects the cultures of the learners (Auerbach 1989, 1990) as well as to promote scientific researches which questions the interactions about teachers and learners as well as the nature of literacy in families, communities, and workplaces.

3.9 Concluding Remarks The aim of our study has been to examine how certain variables which characterize the various processes of social exclusion have an impact on literacy rates in terms of education in a large number of countries. According to social exclusion theory, risks for illiteracy are positively associated with involuntary social, economic, and cultural exclusion from society. In this paper, for analyzing the complex phenomenon of social exclusion in terms of literacy a social exclusion framework has been used in terms of children out of school, levels of education, and exclusion from the labour market. The available literature on literacy related to social exclusion has been reviewed to explore the available evidence on

7 This consideration is especially true for educational programmes defining literacy as critical reflection and action because they can risk to develop a way of thinking based on the cultural background of the teachers and not on the cultural affiliation of the learners.

3 A Comparative Analysis of Literacy

63

associations among literacy, exclusion, and socio-economic outcomes. Starting from the assumptions of many studies and researches, the paper addresses the question of whether participation in education and labour programs reduces literacy inequities through promoting social inclusion. Equality of opportunity, including educational and labour market exclusion is only partially open to national policy interventions. It is also important how national education systems offer opportunities of starting and continuing education. A more sophisticated comparative research into this area is needed. Nevertheless it is known that the complex phenomenon of social exclusion has a great influence on the economic performance, and specifically on the socio-economic development of a country through a variety of channels. We focus on the average education level of each country considered. The education level and also the social capital are influenced by several variables which can be traced by social exclusion itself. Certainly the concepts of poverty and social exclusion are deeply interconnected. Therefore when individuals or larger shares of population suffer because of socio-economic conditions, such as widespread deprivation and unemployment, it is more difficult to undertake and maintain optimal paths of education and training, and this could foreclose the goal of creating an adequate level of human capital. This occurs more in Western countries, where competition is high, and where human capital has a crucial role in the dynamics of growth and development. On this regard, literacy plays a crucial role in the construction of individuals’ ability to understand and use reading, writing and numeracy skills, which provide the foundation for citizenship, lifelong learning, work, and freedom in contemporary society. Literacy demand is growing in number and complexity, and will continue to be a foundation for citizenship, lifelong learning, work, and freedom in the future. Moreover, the lack of an adequate literacy, related to the average level of each social community, increases the risks of acquiring a protected working system, but also the need to learn how to live civilly with others. In particular, literacy and education are a foundation of a civil society which wants to grow permanently and in the long run in terms of social, cultural, and economic prosperity. So that a national policy should try to promote social inclusion through direct incentives for the weaker groups of society. The variables which we consider for our crosscountry analysis cover the economic aspects of the school enrolment rates at various levels, but also the public investments dedicated to education. We know that conditions which hinder educational processes may result from economic households difficulties, which in turn are often influenced by general economic conditions, especially for people already at risk. It is also interesting to take into account the number of students enrolled at various educational levels and the economic outlook of households as indicators of widespread change in terms of behaviour toward education. According to the findings of an initial analysis, we do not verify a strong relationship between literacy and social exclusion: the basic problem is due to the heterogeneity of the data related to the countries considered. Factor analysis with maximum likelihood method has been useful to indentify the main components in the estimate of literacy rate. It has been necessary to add a variance maximum rotation to transform parts of the model. Finally, we identify

64

E. Bucciarelli et al.

three components which can explain the overall dispersion of the phenomenon of literacy rate: the first component (component No. 1 in the analysis) is a synthetic indicator of social exclusion, the second one (component No. 3) is an indicator regarding the education level, while the third component (component No. 4) represents the economic factor identified by the GDP real growth rate. Then, we have obtained a phenomenon composed of nine initial variables which have become three unique variables. Finally, we group the 30 countries considered in three key clusters which have naturally excluded those countries with education and social differences from the other ones. Countries which mainly suffer social exclusion conditions are Mexico, Turkey and Portugal due to their literacy rate which is lower than that of the other countries. Ultimately, our analysis has served to recognize the differences between countries in terms of literacy at the social dimension. Indeed, we believe that social exclusion is one of the various causes of literacy rate of a country. The empirical evidence has highlighted the heterogeneity of different countries, but we have also identified the presence of different continental reality as the U.S. and some European countries which have strong similarities regarding literacy aspects. The future studies we intend to develop in this research field will consist on the investigation of the relationship between literacy skills and the socio-economic characteristics of individuals. For example, we intend to analyze the relationship between parents’ education and skill of youth, how transmitted deprivation conditions affect social exclusion dynamics, and if different age or linguistic groups manifest different skill levels, and specifically if males and females perform differently. Furthermore, it is our intention to study at what kinds of jobs people at various literacy levels work and what wages they earn. Finally, we want to observe what is the relationship between skills and the socio-economic characteristics of nations, for example how new technologies affect skillfulness and general health. In fact, it is possible to analyze the skills of the adult labour force which characterize the growing areas of an economic system. Acknowledgements The responsibility for the contents of this article lies entirely with the authors and it should not be attributed to their affiliated institutions. The authors wish to thank Giuliana Parodi, Dario Sciulli, Ainura Uzagalieva, two anonymous referees and the participants at the XXV National Conference of Labour Economics, AIEL, Pescara (Italy), September 9–10, 2010 for suggestions, useful comments and the support received.

References Auerbach ER (1989) Toward a social-contextual approach to family literacy. Harward Educational Rev 59(2):165–181 Auerbach ER (1990) Making maening, making change. A guide to partecipatory curriculum development for adult ESL and family literacy. English Family Literacy Project, University of Massachsetts, Boston Beder H (1991) Adult literacy: issues for policy and practice. Krieger Publishing Co., Malabar

3 A Comparative Analysis of Literacy

65

Besset D (2001) Object-oriented implementation of numerical methods. An introduction with java and smalltalk. Morgan Kaufmann, San Fransisco, USA Björck Å (1996) Numerical methods for least squares problems. SIAM Boersma P, Weenik D (2001) Praat, a system for doing phonetics by computer. Tech. Rep. 132, Institute of Phonetic Sciences of the University of Amsterdam Bormouth HD (1974) Reading literacy: its definition and assessment. Reading Research Q 9:7–66 Breen M, Sobel D (1991) Popular oral history and literacy: a handbook. Storylinks: Learning through Dialogue, Toronto, Ontario Cook WD (1977) Adult literacy education in the united states. International Reading Association, Newark, DE de Gregorio J, Lee J-W (2002) Education and income inequality: new evidence from crosscountry data. Rev Income Wealth 48(3):395–416 Fingeret H (1992) Adult literacy education: current and future directions, an update. ERIC clearinghouse on adult, career, and vocational education, Information series, N. 355 Freund R (2003) Statistical methods. Academic Press, London Galor O, Moav O (2004) From physical to human capital accumulation: inequality and the process of development. Rev Econ Stud 71:1001–1026 Gillespie M (1990) Many literacies: modules for training adult beginning readers and tutors. Amsherst Center for International Education, University of Massachusetts at Boston, Boston Gradstein M, Justman M (2002) Education, social cohesion, and economic growth. Am Econ Rev 92(4):1192–1204 Grossman HI, Kim M (1997) Human capital and predation: a positive theory of educational policy. Working papers 97–30, Department of Economics, Brown University Grossmann V (2008) Risky human capital investment, income distribution, and macroeconomics dynamics. J Macroeconomics 30:19–42 Hautecoeur JP (1997) A political review of international literacy meetings in industrialized countries 1981–1994. Int Rev Educ 43(2–3):135–157 Hoover KD, Perez SJ (2004) Truth and robustness in cross-country growth regressions. Oxford Bull Econ Stat 66(5):765 Jakeman A (2008) Environmental modelling, software and decision support. Elsevier, Amsterdam Kaestle CF, Damon-Moore H, Tinsley K (1993) Literacy in the United States: readers and reading since 1880. Yale University Press, London Levine R, Renelt D (1991) Cross-country studies of growth and policy. Methodological, conceptual, and statistical problems. Working paper 608. Macroeconomic Adjustment and growth division, Country Economics Department, World Bank, Washington, DC Levine R, Renelt D (1992) A sensitivity analysis of cross-country growth regressions. Am Econ Rev 82(4):942–963 Mardia K, Kent J (1980) Multivariate analysis. Academic, London Mc-Clellan-Cason J, Gaines J, Hinton S, Munsey J, Sligh A, Thompson P, White C (1992) Remembering lonsdale, our community, our home: a project of the lonsdale improvement organization. Lonsdale Improvement Organization, Knoxville, TN Micklewright J (2002) Social exclusion and children: a European view for a U.S debate. LSE STICERD research paper N. CASE 051 Moser W (1996) Linear models. A mean model approach. Academic, London Northcutt N, Selz N, Shelton E, Nyer L, Hickok D, Humble M (1975) Adult functional competency: a summary. Division of Extension University of Texas at Austin, Austin OECD (2001) Knowledge and skills for life: first results from PISA 2000. Paris OECD (2003) PISA 2003 assessment framework—mathematics, reading, science and problem solving, knowledge and skills. Paris OECD (2006) PISA 2006—science competencies for tomorrow’s world. Paris OECD (2009) PISA 2009 assessment framework—Key competencies in reading, mathematics and science. Paris OECD (2010) Adult literacy. Centre for effective learning environments. OECD, Paris

66

E. Bucciarelli et al.

Rainwater L, Smeeding TM, Coder J (2001) Poverty across states, nations and continents. In: Vleminckx K, Smeeding TM (eds) Child well-being, child poverty and child policy in modern nations: what do we know? The Policy Press, Bristol Schwabacher S (1989) Transforming communities through creating a community literature. Presented to the California Association for Bilingual Education, Anaheim, California, February Shore L, Sheppard (1996) The precision of instrumental variables estimates with grouped data. Industrial relations section, Department of Economics, Princeton University Statistics Canada (SC) and organization for economic cooperation and development (OECD) (2000) Literacy in the information age: final report of the international adult literacy survey. Ottawa and Paris Statistics Canada (SC) and organization for economic cooperation and development (OECD) (2003) Adult literacy and lifeskills survey, Ottawa and Paris Statistics Canada (SC) and organization for economic cooperation and development (OECD) (2005) Learning a living: first results of the adult literacy and life skills survey, Ottawa and Paris Székely GJ, Rizzo ML (2005) Hierarchical clustering via joint between-within distances: extending ward’s minimum variance method. J Classification 22:151–183 Temple J, Johnson PA (1998) Social capability and economic growth. Q J Econ 113(3):965–990 UNESCO (1978) Estimates and projections of illiteracy, current study and research in statistics. UNESCO, Paris, p 154

Chapter 4

Education and Socioeconomic Mobility in Post-Communist Countries Alina Veraschagina

Abstract Patterns of intergenerational educational mobility are studied in twelve post-communist countries of Central Europe and the former Soviet Union (FSU). No clear trend in educational inheritance emerges over the recent 50 years, covering both the period of socialism and transition to a market economy. If any, we find the decrease in intergenerational persistence up until the generation of the 1950s. In subsequent years no further decline is observed. On the contrary in a number of states the correlation between parents’ and children’s schooling got stronger, further increasing over the period of transition. Keywords Education Intergenerational mobility Eastern Europe Transition to a market economy JEL Codes I24

Post-communist countries

D31 J62 P27

4.1 Introduction Education is considered a strategic resource in a modern knowledge-based economy, as well as the main prerequisite for socioeconomic mobility. Given so much emphasis on increasing educational attainments, one would expect educational mobility1 1

Educational mobility is understood as a change in educational status across generations. It is considered to be high if the highest level of education achieved by a child does not closely relate to that of his/her parents. Educational persistence, by contrast, characterizes a situation where the educational choices of children are affected by those of their parents. A. Veraschagina (&) Dipartimento di Economia e Diritto, Università di Roma La Sapienza, Rome, Italy e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_4, Ó Springer-Verlag Berlin Heidelberg 2012

67

68

A. Veraschagina

to be also on the rise. The available literature suggests this has not always been the case (Hertz et al. 2007). Assuming random distribution of talents between children, the development and further use of these talents would require equal access to educational resources, independently of the family background. High educational persistence can thus be seen as an impediment to growth, whereby part of the available human capital is not exploited fully. This chapter investigates the trends in intergenerational educational mobility in twelve post-communist economies2 in Central and Eastern Europe (CEE) and the former Soviet Union (FSU) for which little empirical evidence has been made available thus far. These countries witnessed a spectacular increase in educational attainments during the socialist era fuelled by education offered free of charge. The specific question that we ask is whether it had been accompanied by an increase in educational mobility. Some of the most recent studies, including Hertz et al. (2007) and Pfeffer (2008), do not find considerable change in intergenerational educational mobility over the twentieth century in a wide range of both developed and developing countries, as if there are intrinsic forces keeping it relatively stable. Pfeffer (ibid.) goes so far as to call educational mobility patterns and rates pervasive characteristics of nations. It would thus be important to understand whether communist regimes managed to overcome this limit. One of the factors that might hinder the increase of intergenerational education mobility alongside the general rise of educational attainments is the tendency of parents to provide their children with at least the same level of education as their own.3 This holds true in any setting, be it capitalist or socialist economy. Parents’ involvement in the education careers of children takes different forms depending on how the education system operates. In contexts where education is free, the number of positions at higher levels is usually rationed. Thus higher-status parents in socialist countries might have strived to facilitate entry into higher or higher quality education for their offspring. Moreover, as long as the average educational attainments increase, the share of parents who would favour educational persistence would also increase. The transition to market economy can be seen as another countervailing factor. With the launch of market-oriented reforms and the concurrent near abolition of free education, the role of the family background would gain importance and educational mobility would decline as a result. Testing these hypotheses is made difficult by the fact that investigation of the communist period affords a long time-perspective but is limited by lack of data. Conversely data are now available for the transition and post-reform periods, 2

These comprise Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, Slovenia, Romania, Russia and Ukraine. 3 Breen and Goldthorpe (1997) modeled this behavior within the framework of formal rational action theory, despite the fact that the parents’ decision is often taken irrespective of the abilities that the offspring shows.

4 Education and Socioeconomic Mobility in Post-Communist Countries

69

for which, however, the time span may still be too short to reconstruct a clear trend. Different research methods have been proposed to overcome data limitations, at least in part, and some empirical evidence on the intergenerational educational mobility in post-communist economies is already available. In one of the first studies to appear, Ganzeboom and Nieuwbeerta (1999) considered six Eastern European countries, including Bulgaria, Czech Republic, Hungary, Poland, Slovakia and Russia. The authors concluded that the effect of parents’ education decreased by about half from 1940 to 1985, but remained at the high end of the spectrum in international comparisons.4 This would imply that the socialist systems were unable fully to exploit the potential of education in order to facilitate intergenerational mobility. Since then, a number of country-specific studies have appeared (Beblo and Lauer 2004, for Poland; Gang 1996, and Varga 2006, for Hungary; Saar 2010, for Estonia; Hazans et al. 2008, for the Baltic States), and practically all of them claim that parent’s education continues to have a strong positive effect on children’s educational attainments in post-communist economies. Most of the studies nevertheless confirm that some increase in educational mobility took place, especially in the post-World War II period. However, there is no consensus as to its order of magnitude or how long the temporary improvement lasted. There is even less consensus on the direction of the current trends or on current levels of educational mobility, since both depend on how country-specific institutions developed during the reforms. Several recent studies claim that intergenerational educational mobility may have declined over transition, this being the case for Bulgaria (Hertz et al. 2009) and Russia (Gerber and Hout 2004). In what follows we shall re-consider these issues for a much larger number of Eastern European countries than examined by any of the existing studies, and we shall use the same methodology across the countries. Our specific focus is on three hypotheses, respectively that (1) the relationship between education of parents and their children weakened during the socialist era; (2) current levels of education mobility in post-communist countries are higher than in their Western counterparts, and (3) transition to a market economy has caused an increase in educational persistence. We shall look at both pre- and post-reform periods. The divide between the two is more marked for the FSU countries, where the launch of market-oriented reforms coincided with the break-up of the Soviet Union. For other countries in Eastern Europe, the timing of reform and the starting conditions varied. Some of these countries had preserved elements of a market culture, and thus were better prepared for the great transformation, while for others the transition came as a shock after almost half a century of domination of socialist system. As a result, the patterns of intergenerational educational mobility are also likely to vary among

4 The coefficients obtained by regressing the education of children against that of their parents were at the level of 0.4–0.6 for Bulgaria, the Czech Republic, Hungary, Poland, and Slovakia. Only for Russia were they found to be lower, 0.28 for men and 0.33 for women.

70

A. Veraschagina

transition countries, despite the fact that earlier they all adhered to socialist ideals and currently all accept the market economy. The chapter is structured as follows. In the next section we describe the data and methodology used to approach the issue. We then present the results and discuss the main findings. The final section makes some concluding remarks.

4.2 Data and Measures Used for the Analysis We use education as a proxy for status in a society, and in order to trace its transmission across generations, we look at how the education of parents and children relate to each other. We rely on two basic measures for this purpose: the coefficient obtained by regressing the years of education of an individual against that of his (her) parents, and the correlation between the two levels of education.5 As emphasized in Hertz et al. (2007) these two measures yield different pieces of information. The regression coefficient shows the change in the expected level of education of children in response to a one-year change in the education of their parents. The correlation coefficient measures the correspondence between one standard deviation change in parents’ education and one standard deviation difference in the schooling of their children.6 Hertz et al. (2007) document a substantial decline in the regression coefficients for a set of the 42 countries considered, which indicates a weakening of the statistical association between the education of children and that of their parents, not a weaker causal relation. However, the explanatory power of parents’ education for the education of the next generation (R2 in the bivariate regression) remains fairly stable, which is also reflected in the fact that the correlation coefficient between the education of children and that of their parents is stable around a value of 0.4. The study thus concludes that parental schooling by itself now explains as much of the variance of children’s schooling as ever. This holds true for some post-communist countries which had been investigated among others. Given the distinctiveness of these countries with regard to education, it is worth focusing on a larger number of them and asking more specific questions. This is what we do in this chapter. The 12 ex-socialist countries that we consider are listed in Table 4.1, which also reports the data source and the size of the sample used for each country. The sources of data include the European Social Survey (ESS) and the EU Statistics on Income and Living Conditions (EU-SILC). For the countries covered 5 For this study we only use the information about the education of parents and their children, no other regressors enter the specification. Note that our main focus is not on the causal relation, but rather on association between the two main entries. 6 The relation between the two measures is as follows: rsc ¼ bcs ðrc0 =rc1 Þ; where the indexes c and s stand for cohort and schooling, rc0 and rc1 are standard deviations of schooling in two successive generation.

4 Education and Socioeconomic Mobility in Post-Communist Countries

71

Table 4.1 Data used for the analysis Country Year Dataset Number of observations usedc (out of total) Men/Women Czech Republic Estonia Hungary Latvia Lithuania Poland Slovakia Slovenia Bulgaria Romania Russia Ukraine

2005 2005 2005 2005 2005 2005 2005 2005 2006 2006 2006 2006

EU-SILCa EU-SILC EU-SILC EU-SILC EU-SILC EU-SILC EU-SILC EU-SILC ESSb ESS ESS ESS

5.751(8.628) 5.570 (9.643) 9.611(14.791) 4.770 (7.913) 6.251(9.929) 23.699 (37.671) 8.394 (12.879) 5.356 (23.862) 965 (1.400) 1.389 (2.139) 1.539 (2.437) 1.352 (2.002)

2.768/2.983 2.593/2.977 4.570/5.041 2.119/2.161 2.825/3.426 11.223/12.476 3.966/4.428 2.612/2.744 356/609 656/733 641/898 532/826

Imposing age limits for the youngest and the oldest cohorts and dropping observations with missing education implied a reduction in the size of the sample as shown in brackets (column 4) a The EU-SILC b ESS c The size of the samples is restricted by the number of observations with complete information on own and parental education

by the EU-SILC, we use the years of education reported in the main survey, while drawing information on the education of parents from the special 2005 module on the intergenerational transmission of poverty. The module includes a question about the highest level of education attained by the father and the mother. The ESS provides similar type of information as the EU-SILC. The education of parents in both of the surveys was reported in the form of the highest level achieved. Levels of education were thus converted into years by exploiting observations for which both years and level were reported individually (see Tables 4.7–4.8 of the Appendix). We first carried out the estimation on a year-by-year basis for all the countries. Because the noisy picture obtained was not easy to interpret, and because the results may have been influenced by unequal sizes of age-groups, we decided to rely mainly on by-cohort estimates. The data were divided into nine 5 year birth year cohorts as described in Table 4.2. Because the youngest and the oldest respondents were excluded, the age interval of our observations spanned the ages from 26 to 66 in EU-SILC, and 25 to 69 in ESS. Table 4.3 reports the range in years of education for both the parents and their children. Note that for most of the countries zero values are not allowed. This depends on the classification used in the questionnaire, but should not be a problem in a context where everybody is expected to obtain at least basic level of education.7 Parent education is measured by the average value for the mother and 7

This was probably the motivation behind the coding for education, and it may have given rise to an upward bias for older generations.

1935–1939 1940–1944 1945–1949 1950–1954 1955–1959 1960–1964 1965–1969 1970–1974 1975–1979 Total

– 777 792 803 664 606 628 715 766 5751

– 696 642 741 820 822 708 608 533 5570

– 1284 1087 1342 1295 1019 1115 1189 1280 9611

– 664 490 627 681 658 626 524 500 4770

Latvia – 802 718 843 1008 993 749 588 550 6251

Lithuania – 2078 2752 3605 3698 3092 2692 2771 3011 23699

Poland

Source: Own calculations using EU-SILC 2005 and ESS 2006 as specified in Table 4.1

1 2 3 4 5 6 7 8 9

Table 4.2 Observations by countries and age cohorts No Cohort Czech Republic Estonia Hungary – 826 932 1238 1212 1190 904 935 1157 8394

Slovakia – 598 575 756 741 690 683 640 673 5356

Slovenia 108 132 128 123 125 91 99 84 75 965

Bulgaria

98 177 166 188 130 131 191 182 126 1389

Romania

141 116 177 202 193 175 189 183 163 1539

Russia

180 126 156 191 150 129 144 134 142 1352

Ukraine

72 A. Veraschagina

4 Education and Socioeconomic Mobility in Post-Communist Countries Table 4.3 Years of education Country Years of education

Czech Republic Estonia Hungary Latvia Lithuania Poland Slovakia Slovenia Bulgaria Romania Russia Ukraine

73

Average years of education

Parents (min/max)

Children (min/max)

Parents

Children

Cohort 1

Cohort 9

Cohort 1

Cohort 9

4/15 2/15 2/15 2/15 2/15 2/15 4/15 2/15 3/16 2/18 4/18 3/17

4/15 2/15 2/15 2/15 4/15 2/15 4/15 2/15 1/22 0/25 3/22 0/25

8.9 6.0 6.0 6.1 4.0 4.1 7.1 5.1 6.8 4.7 6.2 5.7

10.2 11.5 10.0 10.4 10.9 8.7 10.3 8.4 10.4 10.7 12.4 12.1

10.1 10.5 9.2 9.1 10.1 7.9 10.3 8.1 10.6 7.9 10.9 10.3

10.8 10.9 10.7 10.0 11.5 11.0 11.1 10.6 11.9 12.7 13.7 12.5

Source: Own calculations using data as specified in Table 4.1

the father. When the information is missing for either parent, the remaining value is treated as the average value for the couple in order to maximize the number of observations. All of the datasets involved in this study lack information on children who live outside the household. At the same time, parent education is reported by children independently of whether they lived in or out of the household, or whether they were still alive or not. Whilst this ensures wider coverage, recollection by children may be problematic. Table 4.3 also reports average years of education for the first and the ninth cohorts, with separate records for parents and children. The figures for children are often twice as high as those for parents, providing evidence of a considerable increase in educational attainments in the countries under consideration in the second half of the twentieth century.

4.3 Empirical Findings We first estimated the two basic measures of educational persistence as described in the previous section. For the six countries in common with the study of Ganzeboom and Nieuwbeerta (1999), the values obtained for the correlation and regression coefficients were broadly comparable, with the sole exception of Bulgaria. For this country we found much higher educational persistence, as can be seen from Table 4.4. Our findings are probably driven by the sharp decline in

74

A. Veraschagina

Table 4.4 The relation between parents’ and children’s education Country All Men Czech Republic Estonia Hungary Latvia Lithuania Poland Slovakia Slovenia Bulgaria Romania Russia Ukraine

Women

Correl.

Coeff.

Correl.

Coeff.

Correl.

Coeff.

0.380 0.331 0.461 0.389 0.358 0.391 0.329 0.402 0.626 0.508 0.402 0.312

0.519 0.308 0.419 0.448 0.292 0.409 0.335 0.458 0.665 0.557 0.336 0.309

0.386 0.339 0.434 0.395 0.337 0.385 0.304 0.335 0.614 0.466 0.368 0.326

0.532 0.299 0.375 0.448 0.272 0.389 0.303 0.355 0.618 0.538 0.313 0.356

0.383 0.335 0.487 0.390 0.387 0.398 0.353 0.463 0.629 0.562 0.438 0.291

0.523 0.316 0.460 0.444 0.314 0.426 0.366 0.554 0.689 0.605 0.369 0.295

Source: Own calculations using the data as specified in Table 4.1

intergenerational mobility in post-socialist Bulgaria also documented in Hertz et al. (2009).8 Overall, no clear pattern emerges for the trend in educational inheritance over the past 50 years (see Fig. 4.1), which is contrary to the expectations raised by Ganzeboom and Nieuwbeerta (1999). If anything, we find a decrease of intergenerational persistence until the generation of the 1950s. In later years there appears to be no further decline, on the contrary, in a number of states the effect of family background grows stronger. In all likelihood, the earlier decline is the outcome of the policy of massively expanding education implemented by practically all governments of the Eastern Bloc in the first half of the twentieth century. The graphs in Fig. 4.1 also highlight the between-countries differences in the levels of educational mobility. The Central European countries oscillate around values of 0.3–0.5 for correlation and regression coefficients, while former members of the Soviet Union tend to record lower values, around 0.2–0.49: this is specifically the case of Lithuania, Russia and Ukraine (see Table 4.4 for average values of correlation and regression coefficients across cohorts). The recent trends are of particular interest for testing the effect of transition on intergenerational mobility. This is a feasible task with the data at hand, since both EU-SILC 2005 and ESS 2006 include people born in the 1970s, who are expected to have completed their education career in the mid 2000s, be it at secondary or 8 The correlation between the education of parents and that of their children in Bulgaria almost doubled from 1995 to 2000. Moreover, educational attainments declined in absolute terms for children from families with lower levels of parents’ education. Hertz et al. (2009) claim that this was an economically-driven structural change caused by the contraction of public spending on education and the decline in its quality, the increase in out-of-pocket costs, the fall in the number of schools, and the rise in unemployment among those with secondary education. 9 Some overlapping between the two groups is possible within the limits of the confidence intervals, although it is not consistent throughout the years considered.

4 Education and Socioeconomic Mobility in Post-Communist Countries Czech Republic

75

Estonia

1

1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2 coefficient

coefficient

correlation

0

correlation

0 1942 1947

1952 1957

1962 1967

1972 1977

1942 1947 1952 1957 1962 1967 1972 1977

Hungary

Lithuania

1

1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2 coefficient

correlation

0

coefficient

correlation

0 1942 1947 1952 1957

1962 1967 1972 1977

1942 1947 1952 1957 1962 1967 1972 1977

Latvia

Poland

1

1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2 coefficient

correlation

0

coefficient

correlation

0 1942 1947 1952 1957 1962 1967 1972 1977

1942 1947 1952 1957 1962 1967 1972 1977

Fig. 4.1 Intergenerational educational regression coefficients and correlations. Source: Own calculations using the data as specified in Table 4.1

higher levels. Most importantly, the sample includes also people who obtained their degrees during transition and whose educational choices may have been affected by the ongoing changes. In order to test the effect of the transition we had to identify the timing of the possible structural break in the data. This task was not straightforward, since the reform was launched at different points in time and at different paces throughout Eastern Europe. Graphical evidence about the time pattern of the estimated coefficients (Fig. 4.1) is not particularly suggestive, since no isolated and major break emerges clearly over the period considered. We therefore decided to test for all possible breaks in the data by using first the CUSUM test and then the Chow test. The CUSUM test is based on the cumulated sum of residuals (Brown et al. 1975) and verifies the occurrence of structural breaks without prior knowledge of the precise timing of the breaks. The test is performed by plotting the cumulated sum of residuals against time as in Fig. 4.4 of the Appendix. The structural break occurs if the CUSUM plot falls outside the predefined boundaries.

76

1

A. Veraschagina Slovenia

1

0.8

0.8

0.6

0.6

0.4

0.4 0.2

0.2

coefficient

correlation

0

Bulgaria

1942 1947 1952 1957 1962 1967 1972 1977 1

Romania

0.8

0.8 0.6

0.6

0.4

0.4 0.2

0.2

coefficient

correlation

0

coefficient

correlation

0 1937 1942 1947 1952 1957 1962 1967 1972 1977

1937 1942 1947 1952 1957 1962 1967 1972 1977 1

coefficient correlation

0 1942 1947 1952 1957 1962 1967 1972 1977

1

Slovakia

Russia

1

0.8

0.8

0.6

0.6

0.4

0.4

Ukraine

0.2

0.2

coefficient

correlation

correlation

coefficient

0

0 1937 1942 1947 1952 1957 1962 1967 1972 1977

1937 1942 1947 1952 1957 1962 1967 1972 1977

Fig. 4.1 (continued)

When the plot gets close to the boundaries, it is generally interpreted as a sign of possible model instability in that period. In our case it refers to the instability of the regression coefficient bt, i.e. the coefficient of parents’ education, our only regressor for the years of education of children. CUSUM squared is a complementary version of the CUSUM recursive residuals. It is based on the cumulated sum of squared residuals. The two tests are complementary in the sense that CUSUM squared is more appropriate to test for haphazard rather than systematic changes in regression coefficients. We performed the CUSUM test on the whole time series of matched (average years of) education of children and their parents, starting from the generation of the late 1930s—early 1940s and ending with the generation of the late 1970s. This procedure thus enabled us to identify all the possible breaks in the data over the 40 year period considered and not only those caused by the transition.

4 Education and Socioeconomic Mobility in Post-Communist Countries Table 4.5 The incidence of structural breaks Country Structural breaks in the data detected by Czech Republic Estonia Hungary Latvia Lithuania Poland Slovakia Slovenia Bulgaria Romania Russia Ukraine

77

Confirmed by the Chow test

CUSUM

CUSUM squared

No break

Instability 1962–1970

1967 (1%)

1969–1970 1967 1973 1965 1967–68 No break Instability 1962–1966 Instability 1977–1982 1980 Instability 1954–1957 No break

1960–1962, 1968–1970 No break 1958–1959, 1972–1973 1954–1955, 1962–1963 1956–1957, 1963 Instability 1958–1963 No break

1960(1%), 1968 (5%) No break 1958 (1%), 1972 (1%) 1954 (1%), 1963 (1%) 1956 (1%), 1963 (1%) 1958 (5%) 1962(1%)

1961, 1977, instability 1960–1980 1960, instability 1960–1968 Instability 1966–1970

1961 (10%), 1977 (5%) 1960 (1%) 1954 (1%)

Instability around 1945

1945 (1%)

Note: Here we refer to the year of birth of a child Source: Own calculations

As exemplified in Fig. 4.4, the main change in fact occurred around the beginning of the 1960s (note that birth years are displayed by the figure). Table 4.5 in turn lists the detected years of instability or of structural breaks. Once break years had been identified through the CUSUM tests, the statistical significance of the related findings could be ascertained using the Chow test (see column 4 in Table 4.5). In the majority of cases the Chow test confirmed the occurrence of breaks as detected by the CUSUM squared. Note that, sometimes, significance (at the conventional 5% level) was achieved by the CUSUM squared plot, but not by the CUSUM of the recursive residuals. According to Brown et al. (1975), in such cases instability is due to change in the residual variance rather than to shifts in the values of the regression coefficients. Moreover, the timing of detected breaks as in Table 4.5 suggests that, although transition generated some instability in the patterns of intergenerational transmission of education, there had probably been prior events that brought about major changes.10 The search for possible explanations for the trends identified in the data cannot ignore the historical context. We have already found that educational persistence started to increase after a short period of post-World War II improvement. In that period, the Cold War escalated and was epitomized by the construction of the Berlin Wall. Fear and uncertainly about the future were characteristic features of that time, leaving imprints on everyday life. In such a climate, people might have become more self-oriented and less prone to share equally resources, at least with regard to education. 10

Recall that the most important transformations caused by the economic transition in Eastern Europe took place in the late 1980s to early 1990s (Roland 2000).

78

A. Veraschagina

There may be other reasons for more self-oriented behavior. One of them is self-protection, as in the case of the young males in the Soviet Union who entered higher education also to avoid army service at the time of the war in Afghanistan. The war began in 1979 and continued for a long nine-year period during which it never stopped being a threat for youths who feared being sent to the war zone. The generation of the early 1960s that was about to enter university at the turn of the 1970s was affected the most. For many parents, higher education became a way to shelter their sons, since entering college allowed postponement of military service.11 Clearly, the sons of influential parents could resort to this option more easily, which tended to reinforce educational persistence. This specific factor may help explain why educational mobility started to decline in the FSU, and why sons attracted more investment in education than daughters. But it does not rule out the more general explanations that will be explored in the next section. To summarize our findings thus far, of the three hypotheses that we tested, the first two were partially rejected, while the third was weakly confirmed. Specifically, our expectations of a weakening in educational persistence during the communist era did not receive full support from the data. The initial decline during the post-World War II period was followed by a setback some 10 years before transition started.12 We also expected to find higher mobility, i.e. lower regression and correlation coefficients, in post-communist versus mature market economies, but this too was only partly supported by the data. Among the countries considered, only Lithuania, Russia and Ukraine came close to the 0.2 mark at some point (Fig. 4.1). The remaining countries consistently displayed middle range values. This has been also found to hold for market economies by other studies employing the same methodology (Hertz et al. 2007). Our final expectation concerned the effect of transition on educational mobility, and the data suggest that this further declined during the economic reforms. The next section offers some evidence in support of these findings.

4.4 Discussion In the light of the fact that so much effort was devoted by communist countries to promoting socioeconomic mobility, and that education was the main road to mobility, the overall outcome was modest: only the generations between the late

11

After World War II and until 1984, which covers the first five years of the war in Afghanistan, college students were exempted from military service. Because of the lack of soldiers, this privilege was abolished in 1984 for all the universities but the very best [http://www.allpravo.ru/ library/doc6934p0/instrum6935/print6943.html]. The status quo was restored in 1989, the year in which the war in Afghanistan ended. 12 Note, here we refer to the year when a person acquired the highest level of education, while previous tables and graphs referred to the year of birth. We assume that the distance between these two moments is of the range of 17–20 years.

4 Education and Socioeconomic Mobility in Post-Communist Countries

79

Fig. 4.2 Low (left panel) versus high (right panel) mobility cases

1930s and the 1950s were apparently able to take advantage of temporary increases in educational mobility. Overall, the case of post-communist economies raises the question of whether high educational mobility is sustainable over time. If we abstract from the economic and political context shared by the countries of Eastern Europe at that time, and which may in itself counter social mobility, what other factors may have worked against high educational mobility? The issue of sustainability is explored in Fig. 4.2 by means of a scattergram of the paired values of parent and child education in two hypothetical cases of mobility, respectively low (left panel) and high (right panel). High mobility is characterized by a greater dispersion of values, with at least two implications. First, the fact that average educational attainment increases may lead to a decrease in educational mobility because of lower dispersion. Put simply, when everybody is highly educated, there is little room for improvement.13 Second, high mobility also includes cases of downward mobility whereby highly educated parents fail to ensure transmission of high educational levels to their children. The latter circumstance is not easy to accept for parents, all the more so when the number of children per family decreases. As argued by Hirsh back in 1976, education is a partly positional good in the sense that its value depends on both absolute and relative values consumed. In his view, the relative quantity and quality of education matters for access to the limited number of high status positions to which individuals may aspire. When education is made available to everybody for free, it loses the meaning of the main feature of socioeconomic status. By contrast, the privatization of education sphere is expected to strengthen competition for a positional good (Adnett and Davies 2002).

13

Note that in the extreme case where everybody holds the highest degree, mobility would go down to zero. While this is not realistic, it makes the point about the effect of an overall increase in education levels and the kind of contraction in the gap between the education of children and their parents evidenced in Table 4.3.

.6

RO

.5

HU

PLLT CZ

.4

Educational persistence

BG

EE

RU

LV SK

.3

Fig. 4.3 Returns to education and educational persistence. Source: Own calculations

A. Veraschagina

.7

80

UA SI

5

6

7

8

9

10

Returns to education

In order for investment in education not to be discouraged, it is important that the educated labour force should be able to find jobs where the skills previously acquired are adequately exploited and remunerated.14 Providing everybody with equal opportunities to get education and/or increasing the average educational level is therefore a necessary but not sufficient condition to sustain educational mobility. On the labour supply side, moreover, the more education is freely available, the longer the investment period may become before conspicuous returns are reaped, which implies longer postponement of working life. This amounts to an increase in the opportunity cost of education, especially for children from disadvantaged families, and it may therefore contribute to slowing down educational mobility. A decrease in educational mobility that originated on the supply side may have been reinforced by the reduction in PRE on the demand side of the labour market. The latter was probably accentuated by the deterioration of the economic situation which prompted the reforms. Direct evidence on pre-transition PRE is scanty. One exception is the study on Romania by Andren et al. (2004), which provides an estimate of returns to education during the 50 year period from 1950 to 2000. According to this study, an initial increase in PRE (see Table 4.3, ibidem), lasted until the 1960s and was followed by a decline until the early 1990s. This is consistent with the pattern of educational mobility that characterizes Romania, where educational persistence rose from the late 1950s onwards (see Fig. 4.1). Note that (PRE) are expected to have different effects on intergenerational income mobility (IIM) and educational mobility (IEM): 14

An educated labour force is a luxury that not every country can afford in large numbers. In an open economy, people can always migrate in search of higher returns to their education. Unless a country pursues well-designed education and migration policies, increasing human capital may turn into a loss. In former planned economies migration was limited, as well known, and the full effect of higher education was felt within the country.

4 Education and Socioeconomic Mobility in Post-Communist Countries

81

PRE # ¼ [ IIM "

ð4:1Þ

PRE # ¼ [ IEM #

ð4:2Þ

The first expectation is widely supported by estimated earnings functions,15 while Fig. 4.3,16,17 provides evidence that the second expectation may also hold for transition countries. The figure illustrates how increasing PRE are pushing mobility up, and vice versa.18 The latter case is often referred to as an incentive trap, the logic behind it being that low returns to education create little incentive for children with poor backgrounds to spend their effort on schooling. Because of this incentive trap, there may have been a substitution effect following the decrease in PRE (that we have hypothesized to occur before the transition) whereby people started to attribute more importance to income than to education as a way to obtain higher social status.19 An indirect evidence on the same effect can be drawn from a study by Titma et al. (2003), where fathers’ education was found to play modest role for achieving higher societal status shortly before the collapse of the USSR. A very important implication is that education in post-communist countries is increasingly becoming an investment good while in the past it was partly treated as consumption good.20 Change in this respect especially affects the new generations, those exposed to a market culture for which schooling is increasingly driven by economic calculus.21 This may further reinforce the relation between PRE and 15

IIM will be higher in a given generation if there are lower returns to human capital for children or if children’s human capital is less sensitive to parental earnings (see e.g. Solon 2004; Blanden et al. 2005). To show that, assume that earnings are a function of human capital in a given generataion (Wt = /tHt+ut). It is also plausible to assume that children’s human capital accumulation relates to parental income (Ht = utWt-1+vt). Thus we can obtain the following intergenerational mobility function: Wt = /t utWt-1+xt, where xt = /t vt +ut.Then the intergenerational income mobility expressed by the join parameter /t ut will be higher if there are lower returns to human capital for children (captured by /t), or if children’s human capital is less sensitive to parental earnings (lower ut). 16 Note: We use returns to education as reported in Table 4.6. They were deliberately selected for the mid 1990s and are expected to affect the education choices of the youngest cohorts. Mobility in turn is measured as an average of the correlation coefficients between the education of children born in the 1970s and the education of their parents. 17 For a set of countries considered, returns to education explain about one third of the variation in educational persistence. 18 Because of the few countries considered, the revealed positive relationship between PRE and IEM may not be particularly indicative. This effect was actually found to be weak in a recent study by Chevalier et al. (2009). 19 Education is one of the main determinants of earnings, yet it explains only about one third of the variation in earnings (Bowles et al. 2001). 20 Studying, studying and studying, Lenin’s famous slogan was a form of life guide instilled in the minds of people from early childhood. Education was perceived primarily as a means to achieve a comprehensively developed personality (Pastuovic 1993). 21 This explains a shift in interest to specialties not particularly in demand under central planning, with finance, economics and law leading the list.

82

A. Veraschagina

Table 4.6 Selected findings on returns to education in transition countries Country Author Data source Year

PRE

Czech Republic Estonia Hungary Latvia Lithuania Poland Slovakia Slovenia Bulgaria Romania Russia Ukraine

7 7.9 8.2 6.7 7.1 7 6.6 10 5.3 6.7 7 9.7

Flabbi et al. (2008) Hazans (2003) Flabbi et al. (2008) Hazans (2003) Hazans (2003) Flabbi et al. (2008) Flabbi et al. (2008) Flabbi et al. (2008) Flabbi et al. (2008) Andren et al. (2004) Flabbi et al. (2008) Brainerd (2000)

ISSP LFS ISSP LFS LFS ISSP ISSP ISSP ISSP IHS ISSP HS

1996 2000 1996 2000 2000 1996 1998 1997 1997 1996 1997 1994

PRE–private returns to education. BHSIE–Belarusian Household Survey of Income and Expenditure, ISSP–International Social Survey Programme, LFS—Labour Force Survey, IHS— Integrated household Survey, HS—Household Survey

educational mobility. If the state is unable to sustain PRE at reasonably high levels, the potential of education as a channel of socioeconomic mobility will be weakened. Table 4.6 reports the PRE values used to construct Fig. 4.3. The values have been drawn from different studies and might not be fully comparable, but their levels are apparently as high as PRE in other European countries (for a survey see Psacharopoulos and Patrinos 2004). It is interesting to compare the levels of PRE with the measures of intergenerational educational mobility reported in Sect. 4.4. Across countries, higher PRE associate with higher educational mobility, as also shown in Fig. 4.3. For each country, however, high returns to education in later stages of the transition often go together with middle-to-high and yet decreasing educational mobility (see Fig. 4.1). One explanation for these apparently contradictory findings is that transition reversed the relationship between PRE and IEM. An alternative explanation, which we favour, is that estimated PRE do not fully capture actual returns before and during the transition. Estimates of returns to education available in the literature on former planned economies have been derived by applying the standard approach developed for market economies. How appropriate is this approach to estimating PRE in the pre-reform period? The main problem, as we shall see, is that in socialist countries the overall ‘pay’ packet included important nonmonetary components. The latter are disregarded by the standard estimation approach, thus causing underestimation. For example, the elites often enjoyed free access to health services or housing, the latter alone could reduce household expenditures by about one third. The demise of planned economies ended this redistribution in-kind in favour of the elites. With the liberalization of wage setting, monetary returns to education are likely to increase because

4 Education and Socioeconomic Mobility in Post-Communist Countries

83

the pressure towards equality of (monetary) income ceases and the best educated receive comparatively higher wages. However, if this is accompanied by the loss of in-kind benefits, then overall returns to education may not increase. Thus, estimates of PRE before or during the transition period must be treated with caution, especially for countries where in-kind benefits were large. An additional shortcoming of many existing estimates of PRE for countries in transition is that they fail to account for the fact that the private cost of education generally rises with the progressive expansion of the market. If this were instead taken into account, the likelihood of decreasing rather than increasing returns during transition would definitely be higher. There is still no consensus in the literature on this matter (for a recent survey see Flabbi et al. 2008; Brunello et al. 2010). But the indirect evidence, including the decreasing intergenerational educational mobility found by this study, reinforces the idea that PRE may have not increased during transition.

4.5 Conclusion The expectations that educational persistence might have decreased substantially during the socialist era are not entirely supported by this study. We find that an increase in educational mobility until the generation of the 1950–1960s was followed by a decline. We also find that economic transition cannot be held responsible for the mobility decline, since much of it generally took place some 10 years before the reform was launched. More than one reason can be cited as to why this had happened. The first reason, which we may call ‘intrinsic’, is that high levels of intergenerational educational mobility are inherently difficult to sustain. High mobility necessarily entails some downward mobility, whereas parents generally tend to oppose the latter, because they do not accept the prospect of having children with lower education than their own. Other reasons are specific to socialist countries. Since the latter were initially able substantially to increase educational levels for all at low cost, education lost the meaning of the main attribute of socioeconomic status in the central years of socialism, which in turn discouraged investment in education. This was particularly the case of children from disadvantaged families, for whom the opportunity cost of working at earlier age was a crucial factor. The higher the average educational level, the longer it is necessary to invest in schooling in order to gain a comparative advantage. Thus further improvement increasingly became the privilege of children from well-off families. In addition, the Soviet system pursued egalitarian policies featuring compressed earnings scales, which lowered the returns to education. As a result, the idealistic pursuit of education as a consumption good lost its attraction and people began to

84

A. Veraschagina

view it as an investment good. If education does not pay enough, why invest in it? Widespread disincentives started to roll educational mobility back. The deterioration in the economic situation that preceded the transition to a market economy was the result of persistent inefficiencies inherent to a centralplanned economic system (Roland 2000). This reinforced the downward pressure on wages exerted by the egalitarian ideology, and it is likely to have further reduced PRE. In the pre-reform period, in fact, further lowering of PRE and the contraction of mobility rates may have sustained each other. The challenge is to explain what happened after the demise of socialism during transition to a market economy. According to some of the first studies PRE increased during the early transition period (see e.g. Newell and Reilly 1999; Münich et al. 2005), while further developments are not so clear. The most recent studies suggest that the evidence of a rising trend in PRE during late transition is rather weak (Flabbi et al. 2008; Brunello et al. 2010). Our own findings are that intergenerational mobility declined, and this evidence may be taken to imply that the relationship between private returns and educational mobility reversed during transition. However, there are reasons to doubt that such a reversal took place. This is because the estimated increase in PRE in transition countries may have been exaggerated on two counts: because many of the benefits were in-kind under socialism and are not included in available estimates, and because the latter also ignore the rise in the costs of schooling after transition. Further research is thus needed to account fully for changes in returns to education and educational mobility during this period. Whatever answer is given to the last question, the future prospects of intergenerational educational mobility will hinge critically on each country’s ability to ensure adequate levels of actual returns to education via the labour market. Acknowledgements I am grateful to Sam Bowles and Tom Hertz firstly for getting me involved into this field of research and then providing guidance. Special thanks go to Francesca Bettio for careful supervision of this work which made part of the Doctoral thesis. I also wish to thank the participants of the IX GDN conference, the XXV AIEL conference and the BEROC seminar for useful comments, in particular, Tom Coupe, Randall Filer, Francesco Pastore and Maksim Yemelyanau, and the two anonymous referees for suggestions that allowed improving the chapter. All possible imperfections remain my own responsibility.

A.1 Appendix See Tables 4.7–4.8. See Fig. 4.4.

8 10 12 15

579

4317 90 749

– 5751

– 4

Years

– 16

Nobs

Hungary

Latvia

Lithuania

Poland

8

2 4

– 5570

2745 10 579 12 1500 15

661

18 67

2 4

4 9611

5462 10 34 12 1436 15

2429 8

61 185 8

2 4

10 4770

2426 10 565 12 929 15

176

35 629 8

– 4

20 6251

2117 10 1847 12 1480 15

645

– 142

35 23699

Slovakia

Slovenia

8

2 4

8

– 4

12 8394

6244 10 – – 1471 15

658

– 9

8

34 5356

3061 10 348 12 652 15

163

21 2 1077 4

Years Nobs Years Nobs Years

15175 10 875 12 3520 15

1

139 3954

Nobs Years Nobs Years Nobs Years Nobs Years Nobs

Estonia

*Corresponds to ISCED values 5 (not leading directly to an advanced research qualification) and 6 (leading to an advanced research qualification) Source: Own elaboration on the basis of EU-SILC 2005

Pre-primary education Primary education or first stage of basic education Lower secondary or second stage of basic education (Upper) secondary education Post-secondary non-tertiary education First and second stage of tertiary education* Missing Total

Table 4.7 EU-SILC 2005, education coding Highest education level achieved Czech Republic

4 Education and Socioeconomic Mobility in Post-Communist Countries 85

86 Table 4.8 ESS 2006, education coding Highest education level achieved Bulgaria Not completed primary education Primary education Lower secondary education Upper secondary Post secondary, non-tertiary education Tertiary education Total Romania No school Primary school General school, lower secondary Vocational and apprenticeship High school (upper secondary) Post-high school University degree Post-graduate degree Total Russia Primary of first stage of basic education Lower secondary, second stage of basic Upper secondary Post secondary, non-tertiary First stage of tertiary Second stage of tertiary Total Ukraine Not completed primary education Primary education Not completed secondary education Completed secondary education Secondary technical education First stage of high education Completed high education (specialist, master, post-graduate, scientific degree) Total Source: Own elaboration on the basis of ESS 2006

A. Veraschagina

Nobs

Years

10 29 199 478 66 182 965

3 4 8 11 14 16 –

21 106 221 341 372 131 161 8 1389

0 4 8 11 12 15 17 18 –

40 132 401 515 427 24 1539

4 8 11 13 15 18 –

5 51 70 370 515 41 306 1358

3 7 8 10 13 15 17 –

4 Education and Socioeconomic Mobility in Post-Communist Countries

CUSUM

0

0

CUSUM squared

CUSUM squared

Czech Republic

CUSUM

87

1

0

year

Estonia

CUSUM

CUSUM

1942

1979

0

0

year

1979

CUSUM squared

CUSUM squared

1942

1

0

year

Hungary

CUSUM

CUSUM

1942

1979

0

0

year

1979

CUSUM squared

CUSUM squared

1942

1

0

year

latvia

CUSUM

CUSUM

0

0

1942

1942

1979

year

CUSUM squared

1942

1979

Fig. 4.4 Graphical illustration of the CUSUM test

year

1979

CUSUM squared 1

0

1942

year

1979

88

A. Veraschagina Lithuania

0

0

CUSUM squared

CUSUM squared

CUSUM

CUSUM

1

0

year

1979

Poland

CUSUM

CUSUM

1942

0

0

year

1979

CUSUM squared

CUSUM squared

1942

1

0

year

1979

0

0

1942

year

CUSUM

1942

Fig. 4.4 (continued)

0

1942

Slovenia

0

0

year

1979

1979

1

1979

CUSUM

year

CUSUM squared

Slovakia

CUSUM squared

CUSUM

CUSUM

1942

CUSUM squared

1942

year

1979

CUSUM squared

1

0

1942

year

1979

4 Education and Socioeconomic Mobility in Post-Communist Countries Bulgania

0

0

CUSUM squared

CUSUM squared

CUSUM

CUSUM

89

1

0

year

1987

Romania

CUSUM

CUSUM

1940

0

0

year

1987

CUSUM squared

CUSUM squared

1940

1

0

year

1987

1940

Russia

CUSUM

CUSUM

0

0

year

1987

CUSUM squared

CUSUM squared

1940

1

0

year

1981

Ukraine

CUSUM

CUSUM

1940

0

0

year

1981

CUSUM squared

CUSUM squared

1940

1

0

1940

Fig. 4.4 (continued)

year

1981

1940

year

1981

90

A. Veraschagina

References Adnett N, Davies P (2002) Education as a positional good: implications for market-based reforms of state schooling. Br J Educ Stud 50(2):189–205 Andren D, Earle J, Sapatoru D (2004) The wage effects of schooling under socialism and in transition: evidence from Romania, 1950–2000. W.E. Upjohn Institute Staff Working Paper No. 04–108 Beblo M, Lauer C (2004) Do family resources matter? Educational attainment during transition in Poland. Econ Transition 12(3):537–558 Blanden J, Gregg P, Machin S (2005) Educational inequality and intergenerational mobility. In: Machin S, Vignoles A (eds) What is the good of education? The economics of education in the UK. Princeton University Press, Princeton and Oxford Bowles S, Gintis H, Osborne M (2001) The determinants of earnings: a behavioral approach. J Econ Literature XXXIX:1137–1176 Brainerd E (2000) Women in transition: change in gender wage differentials in eastern Europe and FSU. Ind Labor Relat Rev 54(1):139–162 Breen R, Goldthorpe J (1997) Explaining educational differentials. Towards a formal rational action theory. Rationality Soc 9:275–305 Brown B, Durbin J, Evans J (1975) Techniques for testing the constancy of regression relationship over time. J R Stat Soc Ser B 37:149–172 Brunello G, Crivellaro E, Rocco L (2010) Lost in transition? The returns to education acquired under communism 15 years after the fall of the Berlin wall. IZA Discussion Paper No. 5409 Chevalier A, Denny K, McMahon D (2009) A multi-country study of inter-generational educational mobility. In: Dolton P, Asplund R, Barth E (eds) Education and inequality across Europe. Edward Elgar, Cheltenham/Northampton Flabbi L, Paternostro S, Tiongson E (2008) Returns to education in the economic transition: a systematic assessment using comparable data. Econ Educ Rev 27(6):724–740 Gang I (1996) Who matters most? The effect of parent’s schooling on children’s schooling. Rutgers University, Department of Economics, Working Paper no. 199613 Ganzeboom H, Nieuwbeerta P (1999) Access to education in six eastern European countries between 1940 and 1985. Results of a cross-national survey. Communist Post-Communist Stud 32:339–357 Gerber T, Hout M (2004) Tightening up: declining class mobility during Russia’s market transition. Am Sociological Rev 69(5):677–703 Hazans M (2003) Returns to education in the Baltic countries. GDNet Knowledge Base Working Paper No. DOC16801 Hazans M, Trapeznikova I, Rastrigina O (2008) Ethnic and parental effects on schooling outcomes before and during the transition: evidence from the Baltic countries. J Popul Econ 21(3):719–749 Hertz T, Jayasundera T, Piraino P, Selcuk S, Smith N, Verashchagina A (2007) The inheritance of educational inequality in developing countries: international comparisons and fifty-year trends. B E J Econ Anal Policy 7(2). Article 10 Hertz T, Meurs M, Selcuk S (2009) The decline in intergenerational mobility in post-socialism: evidence from the Bulgarian case. World Dev 37(3):739–752 Münich D, Svejnar J, Terrell K (2005) Returns to human capital under the communist wage grid and during the transition to a market economy. Rev Econ Stat 87(1):100–123 Newell A, Reilly B (1999) Rates of return to education in the transitional economies. Educ Econ 7(1):67–84 Pastuovic N (1993) Problems of reforming educational systems in post-communist countries. Int Rev Educ 39(5):405–418 Pfeffer F (2008) Persistent inequality in educational attainment and its institutional context. Eur Sociological Rev 24(5):543–565

4 Education and Socioeconomic Mobility in Post-Communist Countries

91

Psacharopoulos G, Patrinos HA (2004) Returns to investment in education: a further update. Educ Econ 12(2):111–134 Roland G (2000) Transition and economics: politics, markets and firms. The MIT Press, Cambridge Saar E (2010) Changes in intergenerational mobility and educational inequality in Estonia: comparative analysis of cohorts born between 1930 and 1974. Eur Sociological Rev 26(3):367–383 Solon G (2004) A model of intergenerational moblity variation over time and place. In: Corak M (ed) Generational income mobility in North America and Europe. Cambridge University Press, Cambridge, UK Titma M, Tuma NB, Roosma K (2003) Education as a factor in intergenerational mobility in soviet society. Eur Sociological Rev 19(3):281–297 Varga J (2006) The role of labour market expectations and admission probabilities in students’ application decisions on higher education: the case of Hungary. Educ Econ 14(3):309–327

Chapter 5

Analysing Industrial Accidents in European Countries Using Data Envelopment Analysis Eugenia Nissi and Agnese Rapposelli

Abstract There has been increasing interest in improving working conditions and in reducing occupational accidents and diseases in the European Union. This paper examines the performance of fifteen European countries, in respect of this issue, in three economic sectors—manufacturing, construction and distribution trades—by means of the non-parametric approach to efficiency measurement, represented by Data Envelopment Analysis (DEA). A linear programming framework is therefore used to construct a production frontier which allows measurement of relative efficiency among national institutions in the sample considered. Keywords Occupational safety and health Accidents at work efficiency Data envelopment analysis Undesirable outputs JEL Codes J28

Technical

J81 L60 L74 L81 C61

5.1 Introduction Safety and health at work is now one of the most important and most highly developed aspects of the European Union’s policy on employment and social affairs. Nowadays, the development and implementation of holistic approaches

E. Nissi (&) A. Rapposelli Dipartimento di Metodi Quantitativi e Teoria Economica, Università d’Annunzio di Chieti-Pescara, Chieti, Pescara, Italy e-mail: [email protected] A. Rapposelli e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_5, Ó Springer-Verlag Berlin Heidelberg 2012

93

94

E. Nissi and A. Rapposelli

and strategies towards occupational safety and health (OSH) becomes more and more important to further improve the working conditions in the EU Member States. For the European Union, in fact, the improvement of working conditions and the prevention of workplace accidents are amongst the primary objectives to pursue, as stipulated in the Treaty of Rome (Article 136) and confirmed by the Framework Directive 89/391. Again at the Lisbon European Council in March 2000, the objective that the European Union set itself was ‘‘creating more and better jobs’’ (Commission of the European Communities 2002). However, according to Eurostat data, 5,720 people die each year due to injuries at work and 3.2% of employees suffered at least one injury. An accident at work is defined as an external, sudden, unexpected, unintended and violent event during the execution of work or arising out of it, which causes damage to the health or loss of life of the employee. For qualification as an accident at work to apply, there must be a causal relationship (direct or indirect relationship of cause and effect) between the violent event and the work. There are many methods of preventing or reducing industrial accidents, including anticipation of problems by risk assessment, safety training, control banding, personal protective equipment, respiratory equipment, safety guards, mechanisms on machinery, safety barriers, etcetera. The current Community strategy aims to achieve a 25% cut in accidents at work across the EU by 2012. A key concept and fundamental pillar for reaching the objectives of this Community strategy is the development and implementation of coherent national strategies in the EU member states. Occupational safety and health policy is not only a matter of laws and regulations. In order to achieve measurable improvements of working conditions and a reduction of occupational accidents and diseases it is necessary to combine them with a variety of other instruments, such as social dialogue, good practices, awareness raising, corporate social responsibility, economic incentives and mainstreaming. Hence, the performance of national institutions needs to be evaluated in terms of their ability to maximise macroeconomic objectives while minimising accidents at work. So there is an increasing need for tools that allow proper measurement of the performance of organizations in respect to this issue. In this context, the aim of the present paper is to measure the technical efficiency of fifteen European countries for the year 2005 in respect to the number of workplace accident in three economy sectors—manufacturing, construction and distribution trades. To this purpose, the efficiency analysis is conducted by using the non-parametric approach to efficiency measurement, represented by Data Envelopment Analysis (DEA). Except for a few studies (Beriha et al. 2010; El-Mashaleh et al. 2009; Qu et al. 2010), efficiency analysis using DEA method have not been conducted frequently in this field. However, while in traditional DEA models we have two categories of factors (inputs and outputs), now we consider a third kind of factor, an undesirable output, represented by the number of accidents at work, that could be generated from the production process.

5 Analysing Industrial Accidents in European Countries

95

Ordinary efficiency measures are not suitable in contexts where at least one of the variables that have to be radially contracted or expanded is not a ‘‘good’’. In the standard DEA models, decreases in outputs are not allowed and only inputs are allowed to decrease (similarly, increases in inputs are not allowed and only outputs are allowed to increase). Hence, our objective is to adapt the DEA technique to the problem at hand, where outputs do not refer only to ‘‘goods’’, but we have also ‘‘undesirable’’ outputs. To effect the rankings, we therefore implement a new model of DEA, where undesirable and desirable outputs will be treated differently. The paper is organized as follows. Section 5.2 reviews the theoretical background about technical efficiency and DEA method, Sect. 5.3 presents the data used and lists the results obtained and Sect. 5.4 gives the conclusions.

5.2 Data Envelopment Analysis Methodology The term efficiency is widely used in economics and refers to the best use of resources in production. Modern efficiency measurement began with Farrell (1957), who drew upon the work of Debreu (1951) and Koopmans (1951) and introduced a measure for technical efficiency. According to them, technical efficiency is producing the maximum amount of output from a given amount of input, or alternatively producing a given output with minimum quantities of inputs. Farrell suggested measuring the efficiency of a firm in terms of distance to the best unit on the production frontier, that is represented by the production function of the efficient units. The efficiency frontier is unknown, and it must be estimated from sample data. Drawing inspiration from his argument, two approaches were developed for estimating the efficiency of organisational units: a non-parametric one represented by Data Envelopment Analysis (DEA) and a parametric one represented by Stochastic Frontier Analysis (SFA). The units under evaluation are also called Decision Making Units (DMUs). DMUs are homogeneous organisational units: they perform the same function, by using the same types of resources to produce the same kinds of goods or services. Each DMUs represents an observed correspondence of multiple input–output levels. Data Envelopment Analysis is a non-parametric method for assessing the relative efficiency of Decision Making Units. Rather than explicitly stating the functional form of the best practice frontier, DEA measures efficiency relative to a deterministic frontier using linear programming techniques to envelop observed input/output vectors as tightly as possible. The basic DEA models measure the technical efficiency of a DMU in terms of the maximal radial contraction to its input levels (input orientation) or expansion to its output levels feasible under efficient operation (output orientation). The first DEA model, proposed by Charnes et al. (1978) and known as CCR, assumes the DMUs to be assessed to operate within a technology where efficient production is characterised by constant returns to scale (CRS). Under input

96

E. Nissi and A. Rapposelli

orientation the relative efficiency of a DMU j0 is obtained from the following linear model: e0 ¼ min h0 subject to h0 xij0

n X

kj xij 0;

i ¼ 1; . . .; m

ð5:1Þ

j¼1 n X

kj yrj yrj0 ;

r ¼ 1; . . .; s

ð5:2Þ

8j

ð5:3Þ

j¼1

kj 0;

where yrj is the amount of the rth output to DMU j, xij is the amount of the ith input to DMU j, kj are the weights of DMU j and h0 is the shrinkage factor for DMU j0. The linear programming problem must be solved n times, once for each unit in the sample, for obtaining a value of h for each DMU. The value of h0 obtained is termed the technical input efficiency of DMU j0 and it is bounded between 0 and 1: a technical efficient unit, according to Farrell (1957) definition, will have a score of unity, while inefficient ones will have a score less than unity. The technical efficiency of DMU j0 can be also determined under output expansion orientation, whose objective is to maximise outputs while using no more than the observed amount of any input. Due to the CRS assumption, the relative efficiency score h0 of the output-orientated model relates to that of the input-orientated model via e0 ¼ 1=h0 . Subsequent papers have considered alternative sets of assumptions, such as Banker et al. (1984), who modified the basic CCR model to permit the assessment of the productive efficiency of DMUs where efficient production is characterised by variable returns to scale (VRS). The VRS model, known as BCC, differs from the basic CCR model only in that it includes in the previous formulation the convexity constraint: n X

kj ¼ 1

ð5:4Þ

i¼1

The presence of the convexity constraint in the BCC model reduces the feasible region for DMUs, which, in general, results in an increase of efficient units; otherwise CRS and VRS models work in the same way. In general, under the VRS assumption the model orientation (input or output) affects the projection point on the frontier and the resulting efficiencies may not be the same. Thus, for inefficient DMUs we may have e0 6¼ 1=h0 , although the subset of efficient DMUs is the same irrespective of model orientation. However, although this method has been extensively applied to many areas of economics as an instrument of efficiency measurement, few authors have used it to

5 Analysing Industrial Accidents in European Countries

97

take into account undesirable outputs. It was mentioned already in the seminal work of Koopmans (1951) that the production process may also generate undesirable outputs. Undesirable outputs are prominent in the ecological context (‘‘environmental harmful effects’’ or ‘‘harms’’, Thore and Freire 2002), such as pollution emissions generated in air or in water, waste, poisonous metals dumped into the soil, but they may as well appear in non-ecological applications (Smith 1990) including health care (complications of medical operations) and business (tax payments). As mentioned earlier, classical DEA models measure the relative efficiency of a DMU described by its input and output quantities in terms of maximal radial contraction to its input levels or expansion to its output levels feasible under efficient operation, but this is not valid any longer in contexts where also ‘‘bads’’ have to be considered (Chung et al. 1997; Dyckhoff and Allen 2001; Seiford and Zhu 2005). In the literature several approaches for incorporating undesirable outputs in DEA models have been proposed, but a general protocol is not clear (Scheel 2001). We must underline that efficiency scores, and rankings, may change for different approaches (Dyson 2001). We propose a modified DEA model that incorporates undesirable outputs as inputs, while also seeking to minimise them (Coli et al. 2008, 2011). These factors will be included directly into the linear programming problem, just like inputs that have to be radially reduced. Hence, in this application the general DEA formulation includes the following constraint: h0 htj0

n X

kj htj 0;

t ¼ 1; . . .; z

ð5:5Þ

j¼1

where hij is the amount of the tth input to DMU j and the multiplier h shrinks both inputs and environmental variables in an equi-proportional manner.

5.3 Data and Results We applied the efficiency concept seen before to 15 European countries for the year 2005. For our analysis, we considered three non-financial business economic sectors, according to NACE (Nomenclature statistique des activités économiques dans la Communauté européenne) definition: manufacturing, construction and distribution trades. We decided to analyse only these three sectors, because they register the greatest number of accidents in Europe. In an assessment of comparative performance the first and the most important stage is the identification of the input and output variables. In order to model relative efficiency of a set of DMUs it is necessary to define a production function which captures the key points of the production process. In our specific case, we defined a model characterised by a single input, the number of persons employed, and a single desirable output, the value added (in Euros) for each sector. Moreover,

98 Table 5.1 DEA efficiency scores by European countries for the year 2005

Table 5.2 Summary statistics for DEA efficiency scores

Table 5.3 Spearman rank correlation coefficients

E. Nissi and A. Rapposelli DMU

Manufacturing

Construction

Distribution trades

Belgium Denmark Germany Greece Spain France Ireland Italy Luxembourg Netherlands Austria Portugal Finland Sweden United Kingdom

0.7436 0.4120 1 0.2652 0.6836 0.8844 1 0.6974 1 0.8077 0.5569 0.1469 0.4763 0.8186 1

0.6155 0.7322 0.5461 0.4764 0.4755 0.6137 1 0.4781 1 0.7501 0.5972 0.1887 0.6080 1 1

1 0.9899 1 0.4826 0.6737 1 1 0.7586 1 0.9480 0.8651 0.3657 0.9690 1 1

Manufacturing Construction Distribution trades Mean Minimum Maximum Standard deviation

0.6995 0.1469 1 0.2671

0.6721 0.1887 1 0.2338

0.8702 0.3657 1 0.2007

Manufacturing Construction Distribution trades Manufacturing 1 Construction 0.666 Distribution trades 0.828

1 0.783

1

as mentioned in the introduction, the application of efficiency techniques to this context has motivated the inclusion of a special kind of output, an undesirable output, represented by the number of industrial accidents resulting in three days or more off work. All the data has been obtained from Eurostat. The non-parametric efficiency measures are computed by using the modified input-oriented DEA model under a variable returns to scale assumption, because of the large variation in size of the units (VRS assumes that changing inputs will not result in a proportional change in outputs). The linear program associated with the model is solved using DEA-Solver, a software developed by Kaoru Tone (Cooper et al. 2000). DEA technique provides very detailed information about the analysed DMUs, providing individual efficiency scores for each of them, peer groups and production and consumption objectives for those that are inefficient. Table 5.1 shows, for

5 Analysing Industrial Accidents in European Countries

99

Table 5.4 Reference sets Peer set— Frequency to manufacturing other DMUs

Peer set— construction

Frequency to other DMUs

Peer set— Frequency to Distrib. trades other DMUs

Germany Ireland Luxembourg United Kingdom

Ireland Luxembourg Sweden United Kingdom

9 1 6 8

Belgium Germany France Ireland

5 1 0 4

Luxembourg Sweden United Kingdom

0 1 1

0 9 3 6

each sector evaluated, the efficiency ratings obtained from the input-orientated BCC model. In particular, evaluation of DMUs by means of our model shows that in the manufacturing sector four DMUs are BCC-efficient: Germany, Ireland, Luxembourg and United Kingdom. Besides, many countries do not have very high ratings. In the construction sector, we register the entrance between the top performers of one more country, Sweden (in the manufacturing sector it was quite close to the frontier having an efficiency score equal to 0.8186). On the other hand, Germany returns to be inefficient and several of the others countries receive very low ratings. Finally, we examined the distribution trades sector. Seven of the units form the efficient frontier: we can observe that two more countries (Belgium and France) move on the efficient frontier and Germany returns to be efficient. Besides, one country—Denmark—is very close to the frontier having the efficiency rating of 0.9899. The remaining countries are sub-efficient but they do not show very low ratings. Table 5.2 presents a summary of the efficiency ratings for all sectors analysed. We can see that the distribution trades sector shows a higher average efficiency score and displays less variability than other sectors. On the basis of these results we proceeded to a correlation analysis among the efficiency measures obtained for the three different sectors. We observed quite high Spearman rank correlation coefficients between the technical efficiency rankings (Table 5.3). An important additional aspect of the discriminating power of DEA is its ability to identify a meaningful group of peer units which have similar characteristics to those of an inefficient DMU. Table 5.4 displays the frequency with which efficient countries appear in the peer group of the inefficient ones. In the manufacturing sector, Ireland and the United Kingdom appear quite frequently in the reference sets (9 and 6 times, respectively). In other sets, the most frequent units are Ireland (9 times), United Kingdom (8) and Sweden (6) in the construction sector and Belgium (5) and Ireland (4) in the distribution trades sector. On the other hand, Luxembourg and Germany are not likely to be better role models for less efficient units to emulate.

100

E. Nissi and A. Rapposelli

5.4 Conclusion In this work we evaluated the performance of 15 European countries for the year 2005 in three economy sectors—manufacturing, construction and distribution trades—in respect to the number of workplace accidents. To this purpose, we have obtained measures of technical efficiency by means of the non-parametric approach to efficiency measurement, represented by Data Envelopment Analysis. The results provided by the BCC DEA model shows that there is not a substantial difference between manufacturing and construction sectors: DMUs are not operating at a very high level of efficiency and there is room for improvement in several countries. The mean efficiency score is 0.6995 and 0.6721, respectively. In the distribution trades sector, instead, seven countries are efficient and the mean efficiency score is 0.8702. In addition, we can observe that Ireland, Luxembourg and United Kingdom are always located on the efficiency frontier, for each sector evaluated. Furthermore, Ireland and United Kingdom appear very frequently in the reference sets of efficient units. However, we must point out that the efficiency degree obtained by each unit is relevant only in the context analysed, so, only relative to the chosen model and to the sample examined: if we include a new DMU in the sample or if we assume different model specifications, we could obtain different efficient units or different efficiency degrees. Finally, it must be remembered that the efficiency analysis applied in this work can be improved. First of all, we could include additional key variables or we could apply the model proposed to further application studies, for comparing the performance in other territorial systems, such as Italian regions, European countries, etc., and also in other sectors, such as the transportation and logistic ones. Moreover, we could carry out a performance analysis over time (Sengupta 2000), when data become available.

References Banker RD, Charnes A, Cooper WW (1984) Some models for estimating technical and scale inefficiencies in Data Envelopment Analysis. Manage Sci 30:1078–1092 Beriha GS, Patnaik B, Mahapatra SS (2010) Benchmarking on safety performance using Data Envelopment Analysis. AIMS International Conference on Value-based Management Charnes A, Cooper WW, Rhodes E (1978) Measuring the efficiency of decision making units. Eur J Oper Res 2:429–444 Chung YH, Fare R, Grosskopf S (1997) Productivity and undesirable outputs: a directional distance function approach. J Environ Manage 51:229–240 Coli M, Nissi E, Rapposelli A (2008) Performance measurement by means of Data Envelopment Analysis: a new frontier for undesirable outputs. In: Mantri JK (ed) Research methodology on Data Envelopment Analysis. Universal-Publishers, Boca Raton Coli M, Nissi E, Rapposelli A (2011) Monitoring environmental efficiency: an application to Italian provinces. Environ Model Softw 26:38–43

5 Analysing Industrial Accidents in European Countries

101

Commission of the European Communities (2002) Adapting to change in work and society: a new Community strategy on health and safety at work 2002–2006, Communication from the Commission, Brussels Cooper WW, Seiford LM, Tone K (2000) Data Envelopment Analysis: a comprehensive text with models, applications, references and DEA-Solver software. Kluwer Academic Publishers, Boston Debreu G (1951) The coefficient of resource utilization. Econometrica 19:273–292 Dyckhoff H, Allen K (2001) Measuring ecological efficiency with Data Envelopment Analysis (DEA). Eur J Oper Res 132:312–325 Dyson RG, Allen R, Camanho AS, Podinovski VV, Sarrico CS, Shale EA (2001) Pitfalls and protocols in DEA. Eur J Oper Res 132:245–259 El-Mashaleh MS, Rababeh SM, Hyari KH (2009) Utilizing Data Envelopment Analysis to benchmark safety performance of construction contractors. Int J Project Manage 28:61–67 Farrell MJ (1957) The measurement of productive efficiency. J R Stat S Ser A 120:253–290 Koopmans TC (1951) An analysis of production as an efficient combination of activities. In: Koopmans TC (ed) Activity analysis of production and allocation, Cowles Commission for Research in Economics, Monograph n.13, Wiley, New York Qu J, Wu L, Liao X, Xie Y, Song B, Yan G (2010) DEA with weight constraints for evaluation of work safety supervision. The Ninth Internatioanl Symposium on Operations Research and its applications (ISORA 10), China Scheel H (2001) Undesirable outputs in efficiency valuations. Eur J Oper Res 132:400–410 Seiford LM, Zhu J (2005) A response to comments on modeling undesirable factors in efficiency evaluation. Eur J Oper Res 161:579–581 Sengupta JK (2000) Dinamic and stochastic efficiency analysis: economics of Data Envelopment Analysis. World Scientific Publishers, London Smith P (1990) Data envelopment analysis applied to financial statements. Omega 18:131–138 Thore SA, Freire F (2002) Ranking the performance of producers in the presence of environmental goals. In: Thore SA (ed) Technology commercialization: DEA and related analytical methods for evaluating the use and implementation of technical innovation. Kluwer Academic Publishers, Boston

Part II

Groups Structurally Vulnerable to Social Exclusion

Chapter 6

Social Exclusion and Offending Roger Bowles

Abstract Social exclusion is costly for both its victims and for mainstream society which may experience negative externalities from excluded or marginal groups and contribute resources to support them. Offending behaviour (the principal area of interest in this paper) is costly to its victims and to society and is closely intertwined with exclusion. Offending is often found in association with other indicators of exclusion such as unemployment, problematic substance misuse and mental and physical health issues. The paper uses a generic model of offending, based on the notion of human capital development, viewed from the wider context of social exclusion. It seeks to avoid two criticisms of standard economic models of crime, namely that they tend to be static rather than dynamic and also that they omit any treatment of adolescence during which preferences and planning skills are still in their formative phase. The paper uses data from a number of contemporary English surveys to test the principal conjectures on which a more general, economics-based approach to offending might be developed. The paper concludes that an approach that considers the role of social exclusion provides a better understanding of offending than one that treats crime purely as an enforcement issue. Keywords Offending Social exclusion Unemployment Reconvictions JEL Codes I24

Poverty Adolescent development

I32 I38 J71 K42

R. Bowles (&) York Criminal Justice Economics, York, UK e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_6, Ó Springer-Verlag Berlin Heidelberg 2012

105

106

R. Bowles

6.1 Introduction Social exclusion is costly for both its victims, who typically enjoy a lower quality of life than others, and for mainstream society which may experience negative externalities from excluded or marginal groups and contribute resources to support them through provision of public services and social security payments. Offending behaviour (the principal area of interest in this paper) is costly to its victims and to society at large as citizens and governments seek to prevent crime and to enforce criminal law. Social exclusion and offending are closely intertwined, offending being either or both a symptom and a product of exclusion. Although offending is frequently associated with social exclusion, of course, the two things are not synonymous: some offences are committed by individuals who are not excluded and many who are excluded do not offend. Offending, nevertheless, is often found in association with other indicators of exclusion such as unemployment, problematic substance misuse and mental and physical health issues: Cunliffe and Shepherd (2007), Ministry of Justice (2010a) and Vennard (2007). The question then becomes whether there is any kind of causal link between social exclusion and offending and, if so, whether there is a particular direction of causation and whether the two phenomena are mutually reinforcing. Crime reduction is one of the key domestic policy goals of governments in many countries, with the result that developing a better understanding of the relationship between social exclusion and offending is an important research objective. This paper sets out to use models from economics as a means of illuminating behaviour that is more often approached through other disciplines such as criminology, sociology and psychology. Much of the economic literature to date concerned with crime uses an approach that seems rather sterile when viewed from a social exclusion perspective. The purpose of this paper is to set out a generic model, based on the notion of human capital development, which is able to treat offending in the wider context of social exclusion.

6.1.1 Economic Analysis of Crime Some economists looking at crime have treated both poverty and offending as products of common underlying causes such as a lack of effort (Ehrlich 1973), and this argument could probably be extended to encompass the broader notion of social exclusion. The argument, in effect, is that social exclusion is endogenous and reflects choices that individuals have made in the light of their endowments and proclivities. The unwillingness of economists to contemplate ‘paternalistic’ intervention means that decisions about lifestyle and behaviour are treated as purely private matters. The object of criminal law in these circumstances is to provide a set of enforcement probabilities and punishments that minimises the aggregate social costs of crime and its control. In the standard Becker model of

6 Social Exclusion and Offending

107

crime, offending is a static decision individuals make in the light of information about the size of the potential gain relative to the risk of getting caught and punished: Becker (1968). The policy maker chooses an enforcement budget (including spending on police and public order) in order to influence the production of negative externalities and thereby to minimise the aggregate social cost of crime plus crime control: Bowles et al. (2008). But there are (at least) two key elements in this characterisation that are open to challenge. First, ‘endowments’ (for these purposes) have to include the characteristics of the parents and carers of children and young people. A lot of offending is done by adolescents, whose cognitive development and life experience is still at a comparatively early stage. The influence of home background on the attitudes of these young people as they develop is bound to be substantial. Although judged mature enough to be of a criminally responsible age many of these young people will not have developed the capacity to anticipate the full consequences of their actions for themselves and others. They may still lack the planning and anger management skills to behave responsibly towards other citizens and to understand the negative externalities they may be imposing. For the high offending risk age range of roughly 13–18 there is a tension between treating individuals as being fully responsible from a criminal perspective1 and as being still in need of guidance from parents, educators and others as to appropriate bounds of behaviour and the impact of their actions on others. Making parents or carers vicariously responsible for the crimes of adolescents may not be a very successful policy strategy, since many may themselves be poor role models. Secondly, the deterrence model is a static one that neglects dynamic effects that may modify optimal enforcement policy. This is particularly significant in discussion of the relationship between social exclusion and offending. Offences can be regarded as one-off events whereas social exclusion is a state of being that continues into the future. A conviction may influence the degree of exclusion an individual faces in the short term. But the record of this conviction will remain and may have untold consequences for the exclusion the individual feels into the future. The ‘knock-on’ effects of a criminal conviction for the ability to find a job, accommodation, access to credit and other markets tend to be cumulative and mutually sustaining. A conviction, for example, may mean that an applicant finds it very difficult to get through any initial screening of job applications. This heightens the risk of long-term unemployment which in turn means that the individual has less to lose from further offending.

1 The age of criminal responsibility is currently 10 in England and Wales. There is, however, wide variation even across the EU where in some countries it is lower, and some where it is substantially higher, such as the Scandinavian countries where it is 15.

108

R. Bowles

6.1.2 Implications The implications that flow from the proposition that social exclusion and offending are both products of inadequate effort on the part of individuals who have not bothered to invest sufficiently in human capital are unpalatable to many social researchers. It suggests a punitive, enforcement-orientated criminal law enforcement approach rather than a co-operative approach based on the argument that measures to re-include offending citizens may be welfare-increasing. Social exclusion can be seen almost as a policy tool since it enables a degree of ‘punishment’ to be imposed without incurring the resource costs associated with locking offenders in prison or providing community-based supervision. Although this kind of approach has the advantage that it respects individual tastes, preferences and choices it is clear that as an enforcement device social exclusion imposes high costs of its own. Once individuals come to feel excluded their incentives to seek reintegration may be weakened and over the longer term this may prove very costly in terms of lost output. The literature on ‘million dollar blocks’ (small areas of housing with a high concentration of offenders that absorb large amounts of public resources and contribute little by way of productive activity) is indicative of the concern that the punitive approach may give rise to concentrations of individuals for whom there is little prospect of breaking a repetitive cycle of offending, imprisonment and release followed by further offending: Gladwell (2006) and Martinez and Burt (2006). The cost of imprisoning offenders is just a part of the larger cost of managing a group that may become increasingly dysfunctional from a social perspective. These wider costs might include the costs of dealing with neglected or abused children, run-down social housing, high rates of unemployment, substance misuse and domestic violence and so on: Elliott-Marshal et al. (2005) and Harper et al. (2005). There are other features of offending careers that cause many non-economists working in the criminal justice field to be sceptical about deterrence-based ‘solutions’. For one thing, the peak rates for offending come during adolescence and the offending rate for males at this stage of development is much higher than for females. This implies that there may be a biological basis for a lot of offending. A corollary is that a lot of offending occurs at a stage in the life course when cognitive skills are still developing and when young people have not developed their capacity to anticipate, and take responsibility for, the consequences of their actions. On this view the best response may be to intervene with youngsters who offend by providing them with behavioural treatment and by supporting parents who lack the capacity or inclination to help their children to develop these skills. Looked at from a social exclusion perspective this is to argue that children growing up in a deprived, poor area with a high crime rate and with parents who are themselves excluded may lack the cognitive skills and associated resilience to avoid getting drawn into a feckless, offending lifestyle. Their counterparts in affluent areas may,

6 Social Exclusion and Offending

109

on average, expect better living conditions, more supportive parents and teachers and thus have a wider choice of career pathways ahead. The development of the relationship between these various elements as children and young people grow into adulthood is a particular focus here. We start from the stylised facts that offending rates are at their highest during adolescence and that reconviction rates amongst adult offenders tend to be highest for those whose first offending was recorded at a particularly early age. Some individuals who offend while young desist from offending as they mature into adults.2 But others do not make such a successful transition and remain locked in worlds where many elements of social exclusion are present. Social exclusion, rather like a black hole in space, can be easy to fall into as you get close to it and then very difficult to escape once you are in it. The paper uses empirical data from a number of different sources to explore the relationship between offending and social exclusion. That the two phenomena are closely intertwined is not in dispute. High crime rates are a characteristic of deprived neighbourhoods: at least in the UK the crime rate represents one of the very criteria applied in defining and assessing the degree of deprivation in an area.3 But, in many respects more importantly, offending may not just be a symptom of exclusion so much as a result of it. A report on re-offending by the Social Exclusion Unit (2002), for example, concluded that: Before they ever come into contact with the prison system, most prisoners have a history of social exclusion, including high levels of family, educational and health disadvantage, and poor prospects in the labour market.

This suggests that what is of interest here is not so much the relationship between offending and social exclusion per se, but the characteristics which might dispose a young person to grow up to find themselves socially excluded and also to become involved in offending. The ideas and empirical regularities on which we rely are not new, particularly the link between poverty and crime which has been widely remarked. The novelty lies more in trying to articulate the role that economic forces might be playing in driving both exclusion and offending.

6.2 Model The model underpinning the arguments in this paper is set out in Bowles and Florackis (2011). A high subjective discount rate is likely to be associated with both under-investment in human capital and a disregard for the downside reputational risks associated with offending. It may also be associated with an increased 2

For a glimpse of the statistical analysis of offending careers see Soothill et al. (2009). Crime is one of the seven domains on which the English Index of Multiple Deprivation is constructed. For further details of the construction of the IMD see: http://www.communities. gov.uk/publications/communities/indicesdeprivation07 3

110

R. Bowles

likelihood of becoming involved in problematic substance abuse. By potentially providing a common cause for both offending and social exclusion it is a useful means of exploring further the relationship between the two phenomena. In essence the argument is that certain kinds of household and neighbourhood influences may result in a lack of engagement with education that leaves the young person at a real disadvantage in labour markets, and by extension in other markets such as credit and housing markets. Mutual reinforcement between the excluding influences may find expression in a high probability of being unemployed, of not having adequate accommodation, of involvement in substance misuse and, critically, of involvement in offending and anti-social behaviour. Poor physical and mental health may accompany, or be exacerbated by, the other symptoms. We do not seek to provide a new theory of offending but to use the model as a basis for utilising some of the many datasets in the public domain to explore the links and relationships between offending and its covariates implicit in the human capital approach to offending. In particular we try to construct the elements of social exclusion that are common to the offending behaviour of both children and young people and of adults. The dynamics of the growing up process are not discussed in depth, but it seems clear that a model needs to be able to say something about the evolution of offending behaviour over the life course. Studies of those convicted of crime provide plenty of evidence of the co-occurrence of these various indicators of exclusion: Social Exclusion Unit (2002) and Harper et al. (2005). Offenders commonly exhibit poor employment histories, inappropriate accommodation, high prevalence of substance misuse and comparatively high incidence of physical and mental health issues. Not only does this mean that offenders are prone to being socially excluded. It also makes it difficult for them to become re-included. Many of the components of exclusion are mutually re-enforcing, so that well-intended interventions that address some but not all of an offender’s issues may fail. To make support to offenders allembracing may sometimes be effective but it is very costly and may, in any event, be unwelcome to the offender society is trying to resettle. The model of social exclusion on which we rely is depicted in Fig. 6.1. Its cornerstone is the low investment in human capital lying at its heart. This is seen as a product in part of personal psychology and preferences but a product also of non-supportive households and of negative attitudes that may characterise a deprived neighbourhood. The non-supportive household may comprise parents or carers who are frequently absent or disinterested in children. As well as lack of care this may manifest itself in an incapacity or unwillingness to provide children with life skills and support for the development of cognitive skills, such as planning skills and the management of anger and impulsivity. If financial resources are tight then the household may well be unable to supply financial support for investment in any further education (for 16-18 year olds) or higher education (beyond the age of 18). In many cases, however, the damage will have been done before this stage. For example failing to ensure that children attend school regularly may mean that the children never acquire the basic qualifications that are needed as a pre-requisite for further study.

6 Social Exclusion and Offending Fig. 6.1 Model of human capital formation and offending

111 Nonsupportive household

Low effort: peer pressure, truancy, exclusion

Poor employment prospects

Low accumulation of human capital

Cognitive skills under-developed

Increased probability of substance misuse and/or poor physical & mental health

Increased probability of offending

6.3 Social Exclusion and Youth Offending From a life course perspective social exclusion may be a temporary or a permanent feature. For some children and young people brought up in conditions of poverty there is the possibility of escaping social exclusion later by means of educational achievement and/or success in the labour market. For others the prospects may remain poor. But a more affluent background is not a guarantee of insurance against social exclusion. A combination of substance misuse, unemployment and offending (for example) can result in exclusion for those with the wealthiest of starts. Nevertheless it remains the case that social exclusion is much more commonly experienced in adulthood by those brought up in a deprived area. Evidence in support of this contention can be derived in various ways. In England and Wales increasingly detailed empirical studies of the characteristics of those who have been imprisoned point consistently to the over-representation of certain characteristics of childhood among prisoners.

6.3.1 Surveying Prisoner Crime Reduction A recent report concerned primarily with reconviction risks in England and Wales (Ministry of Justice 2010b, following an interim report, Stewart 2008) incorporates findings from the Surveying Prisoner Crime Reduction (SPCR) source. This is a ‘large, general purpose longitudinal cohort survey of 3,849 prisoners (..) sentenced to between one month and four years’. Sample 1 (the part of interest here) comprised 1,435 prisoners, based primarily on those sentenced in 2005–2006. These prisoners were asked about background factors such as childhood experiences and data were collected on their convictions’ history from the Police National Computer, the most authoritative source of secondary data on offending. From the data in Table 6.1 it can be seen that prisoners are much more likely than a random sample of the population at large to have experienced various kinds of disadvantage as children. A high proportion experienced abuse and a very high proportion witnessed violence in the home. Nearly 40% had been brought up in a

112

R. Bowles

Table 6.1 Childhood experience of prisoners Yes No Experienced abuse as a child Observed violence in the home as a child Respondent taken into care* up to the age of 17 ‘Has anyone in your family ever been found guilty of a nonmotoring criminal offence other than yourself?’ ‘Did you regularly play truant or skip school without permission?’ ‘Were you ever expelled or permanently excluded from school?’

Total responses

Yes (%)

412 1,013 1,425 579 847 1,426 347 1,079 1,426 519 881 1,400

28.9 40.6 24.3 37.1

838

587 1,425

58.8

596

830 1,426

41.8

Source: Adapted from Ministry of Justice (2010b): Tables 2.2, 2.4, 2.6, 2.9, 2.22 and 2.24 can be inspected at: http://www.justice.gov.uk/publications/docs/spcr-full-tables-paper-5-2-prisonersbackgrounds-reconviction-a.xls * This includes living with foster parents or in an institution

household where another family member had been convicted of a criminal offence. Spending formative years in such environments appears to leave many children susceptible to becoming offenders themselves in later life. An immediately striking feature is the high proportion (24.3%) who had been taken into care as children. It is not easy to make an exact comparison with the population at large but, as an indicator, in 2001 there were 58,900 children and young people in care4 out of a total population of 11.28 million in the relevant age range (0–18). Those in care thus represented just 0.52% of the population. Reasons for being in care vary widely as does the kind of care provided, some living with foster parents others in residential accommodation funded by the government. But, as a group, these young people represent a high risk of offending in the future and of continued social exclusion. A further finding from Table 6.1 is that a high proportion (nearly 60%) had played truant from school and a high proportion (over 40%) had been excluded from school. A finding from Table 2.23 in Ministry of Justice (2010b), not included in Table 6.1 here, is that the reconviction rate for those prisoners who had regularly played truant was 61.1% compared with 38.8% for those who had not. The link between offending, truancy and exclusion from school was already wellestablished before the 2010 report. Bowles et al. (2005), for example, used truancy rates and exclusion from school rates to assess the impact of a crime reduction programme implemented in schools because of the link known to exist between offending and absence from school. The rates of truancy and exclusion exhibited by prisoners serve to reinforce the observation of the link with offending behaviour. As an indicator of how high these proportions are, compare them with the rates reported in Table 6.2, which is based

4

The data were taken from the following site: http://www.education.gov.uk/rsgateway/DB/ VOL/v000509/contents.shtml

6 Social Exclusion and Offending

113

Table 6.2 Suspension from school and the offending propensity Have you ever been suspended from a school? Have you been arrested during Yes the past 12 months? No Total Arrested (%)

Yes

No

Total

Suspended (%)

46 299 345 13.3

100 3,597 3,697 2.7

146 3,896 4,042 3.6

31.5 7.7 8.5 –

Source: Author’s analysis based on data from OCJS: (see disclaimer at end)

on findings from a quite different, non-prisoner survey. The Offending, Crime and Justice Survey (OCJS) was the first national longitudinal, self-report offending survey for England and Wales. This survey5 asked a sample of children and young people (aged between 10 and 25, drawn from the general public) a series of questions about their involvement in offending and drug use and about the environment in which they live. The OCJS was run annually over a four year period (2003–2006). Table 6.2 reports a cross-tabulation of those saying they had been arrested during the previous 12 months and those who said they had ever been suspended from school. The proportion of those arrested who had been suspended at some stage (31.5%) was much higher (more than four times as high) than for those who had not been suspended (7.7%). Likewise, of those who had ever been suspended the proportion arrested was much higher (13.3%) than it was for those who had not been suspended (2.7%). Evidence from the latest report of the Youth Survey (conducted annually for the Youth Justice Board) is also consistent with the notion that truancy is correlated with the offending propensity. Of those young people in mainstream education who had never played truant (just over two-thirds of the valid sample), 12% had committed an offence. This rose to 36% for those who had played truant on no more than two occasions, and 49% for those who had played truant more than twice.6 The inescapable conclusion is that poor attendance at school and exclusion from school are significant risk factors for offending. For purposes of testing our model, however, the conjecture is that it is the effect on the formation of human capital that is mediating at least some of this influence: a correlation between exclusion and offending could potentially be explained in other ways. Some indirect evidence on the relationship between school behaviour and social exclusion is provided by data from the 2009 Youth Survey: Youth Justice Board (2010). Table 6.3, based on these data, shows that pupils who had been excluded from mainstream schools and moved to Pupil Referral Units (PRUs) (typically on the basis of behaviour that was disruptive to other pupils in mainstream schools)

5

The datasets are held at the Economic and Social Data Service at the University of Essex, available via:http://www.esds.ac.uk/findingData/ocjsTitles.asp 6 Youth Justice Board (2010) at p. 139.

114

R. Bowles

Table 6.3 Exclusion from mainstream education Living in single parent household Father in full-time employment Mother or step/foster mother in full-time employment

Mainstream education (%)

Pupil Referral Unit (%)

23

45

68

42

41

30

Source: Youth Survey 2009: Youth Justice Board (2010)

tended to have a different home environment from pupils in mainstream schools. The proportion of PRU pupils living in a single parent household was higher and the proportion with a working parent was lower, suggesting they suffered greater social exclusion than those in mainstream education. It is helpful at this stage to return to the SPCR data. Table 6.4 reports the ages at which prisoners had been first arrested and at which they had left school. The great majority of prisoners had left school by the minimum school leaving age, currently 16 in England and Wales (having been 15 at the time when some of the older prisoners in the sample would have left school). Only 14.6% of the prisoners had stayed on at school beyond the age of 16. Leaving at or before the statutory age substantially constrains the scope for investing in human capital. Some may have gone on to vocational training schemes on leaving school but many will not have. Empirical data on earnings show that lifetime earning streams are significantly higher for those getting a minimum set of examination passes at age 16: Bowles et al. (2005). Unfortunately the SPCR survey did not ask about the quality of educational attainment by prisoners: it asked just a generic question about whether they had ‘any qualifications’. Table 6.4 indicates also that 55% of prisoners in the SPCR sample had been arrested by the age of 16. As with much other evidence in the field this cannot be used to establish anything reliable about the sequence of events or the direction of causality: for example it does not help establish whether truancy tends to pre-date offending or vice versa. Offending often goes unobserved and, in any event, the conjecture is that both offending and weak attendance are symptoms with a common cause.

6.4 Reconviction and Social Exclusion A quite different way of looking at offending in relation to the life course and its relationship with social exclusion is to use data from adult offending careers to look back from an adult’s offending career towards possible precursors of their offending when young. In terms of Fig. 6.1, this has a focus on the right hand part

6 Social Exclusion and Offending

115

Table 6.4 Age at first arrest and at leaving school Up to 13 14 Age at first arrest Age at leaving school

Number % Number %

231 17.6 0 0

490 37.3 258 18.6

15

16

419 30.1

594 45.2 510 36.7

17

83 6.0

18 & over

Total

120 8.6

1,315 100 1,390 100

Source: adapted from Tables 2.7 and 2.26 of Ministry of Justice (2010b) Table 6.5 Ministry of Justice findings on precustody drug use and reconviction rates

Problem staying off drugs before custody?

One year reconviction rate (%)

N

Proportion of the sample (%)

No Yes All

46 75 58

2,825 2,073 4,898

42.3 57.7 100

Source: Adapted from Table 4 of May et al. (2008)

of the diagram: it is concerned with offending behaviour given a low level of investment in human capital. Until recently offending data and data on other aspects of an offender’s situation were kept largely separate, so that offending careers tended to be viewed, in isolation, as a sequence of offences comprising a convictions’ record. Pressure to find ways of reducing the proportion of offenders being reconvicted, along with improved databases on offenders, has resulted in increasing interest in exploring offending careers in relation to a variety of social and other variables: Ministry of Justice (2010a, b) and National Audit Office (2010). From the published findings from this work it is possible to apply a degree of hindsight to offending careers. Three resettlement surveys of prisoners were conducted in 2001, 2003 and 2004: May et al. (2008). A representative sample of 4,898 prisoners in the last three or four weeks of custody was asked about their circumstances before and during prison. These data were later matched with reconviction data from the Police National Computer to generate a dataset with which reconviction rates could be modelled. Table 6.5 indicates that 57.7% of the offenders had had problems staying off drugs before custody. This group had a 12 month reconviction rate of 75% compared with 46% for those who had not had such problems. Of more direct interest from a human capital perspective is a second table taken from May et al. (2008) identifying reconviction rates by reference to expected employment after release. As with the previous table, the interest lies more in the proportions of prisoners in each category than in differences in the reconviction rates, revealing though these are. Table 6.6 indicates that a little under a quarter of the sample had a paid job to go to while over half expected to be looking for a job or course, not an enviable task on release from prison. The comparatively high reconviction rate for all groups in the sample gives an indication of the difficulty in general of persuading individuals to give up offending once they have started.

116

R. Bowles

Table 6.6 Ministry of Justice findings on employment status and reconviction rates Expected employment or other status One year reconviction N Proportion of the after release rate (%) sample (%) Paid job to go to Training or education to go to Looking after home or family Long-term sick or disabled Looking for job or course Do not want to work or train Other including retired Total

45 56 53 58 62 75 64 58

1,105 282 53 413 2,683 103 259 4,898

22.6 5.8 1.1 8.4 54.8 2.1 5.3 100

Source: Adapted from Table 8 of May et al. (2008)

Table 6.7 Logistic regression analysis of 12-month reconviction outcomes Variable Odds ratio Aged 18–20 Aged 21–24 Aged 25–29 Aged 30–39 Aged 40 ? (reference category) Drug problem before custody Both accommodation and employment problems Victim awareness course Other variables not reported here include number of previous convictions and offence type

7.35 3.43 2.01 1.51 1.00 1.87 1.43 0.70 N.a.

Significant at 5% level Yes Yes Yes Yes N.a. Yes Yes Yes N.a.

Source: adapted from Table 2 of May et al. (2008)

A third table (Table 6.7) adapted from the same paper summarises findings from a multiple logistic regression analysis based on a binary reconviction dependent variable. It provides estimates of the relative strength of influences on the likelihood of an individual being reconvicted within 12 months of release from prison. The pattern of the coefficients on the age group dummies shows that the reconviction risk falls sharply as the individual progresses through young adulthood. The findings provide further evidence that those with a drug problem before custody are substantially more likely to be reconvicted. The conjectured link between offending and social exclusion7 finds further support, the combination of accommodation and employment problems being associated with a significantly heightened reconviction risk. Again however, it is important to be wary of

7

Earlier evidence of such a link was documented by Niven and Olagundoye (2002) and Niven and Stewart (2005).

6 Social Exclusion and Offending Table 6.8 Prisoner mental health

117 Yes No

Total Yes (%)

Mental health illness or depression 177 1,254 1,431 12.4 as a longstanding illness? 241 1,192 1,433 16.8 Treated/counselled for a mental health or emotional problem in the year before custody? Source: Ministry of Justice (2010b) Tables 4.1 and 4.6

implying causation here. The employment (or accommodation) outcomes may share with the reconviction risk a common cause. It is instructive to consider which of the list of variables originally included in the model8 was excluded in this final specification, since many involved exclusion characteristics. For example no education measures (such as whether the offender had any qualifications) survived. However, the number of convictions (by deciles) was included and it is quite likely that there would, for example, have been a nonnegligible correlation between educational status and ranking in the convictions’ distribution. So it might not be safe to conclude from the findings that educational status plays no role in reconvictions. The conclusion, rather, is that education plays no further significant part once allowance has been made for the number of convictions. Models of this kind are useful for assessing reconviction risk but are less useful for identifying the underlying influences on offending decisions. It has been long established that offenders are more likely than average to suffer from various forms of mental ill health. The standard reference, based on a psychiatric morbidity survey in England and Wales conducted in 19979 (Singleton et al. 1998), found that 7% of male sentenced prisoners, 10% of men on remand and 14% of female prisoners had a functional psychosis (such as schizophrenia or manic depression) in the year prior to interview. In addition, the proportion of respondents with significant neurotic symptoms, such as anxiety, depression and phobias, in the week before interview were far higher than found in a similar household survey and ranged from 39% of male sentenced to 75% of female remand prisoners. Further evidence of there being a comparatively high incidence of mental health issues amongst prisoners, as summarised in Table 6.8, comes from the SPCR survey.10 Accommodation is another indicator of the degree to which individuals are socially excluded. As with other elements of exclusion poor accommodation is likely to be associated with offending whether as a contributory factor or as a consequence. Imprisonment is likely to interrupt tenancy agreements and the consensus is that offenders leaving prison with no accommodation are likely to return to past friends and associates with the risk that they will return to substance 8

A full listing of the variables included can be found in Table 1 of May et al. (2008). The findings are summarised at: http://www.statistics.gov.uk/ssd/surveys/survey_of_psychiatric _morbidity_among_prisoners.asp 10 Op. cit. 9

118

R. Bowles

Table 6.9 Accommodation status prior to imprisonment House or flat owned by prisoner or partner House or flat rented by prisoner or partner House or flat part-owned by prisoner or partner Paying board in a house or flat owned by someone else Paying board in a house or flat rented by someone else Living rent-free in a house or flat owned by someone else Homeless/Temporary accommodation Homeless/Sleeping rough Living with family member Other Total

Number

percent

163 491 34 162 117 176 96 125 49 21 1,434

11.4 34.2 2.4 11.3 8.2 12.3 6.7 8.7 3.4 1.5 100

Source: Ministry of Justice (2010b) Table 2.29

misuse and offending. Arranging alternative forms of housing, supplied by the voluntary or statutory sectors, requires knowledge and cognitive skills that offenders may lack. Managers responsible for offenders post-release therefore often find it worthwhile to provide accommodation support. But as well as being a result of imprisonment, poor accommodation may increase the likelihood of offending. Peer pressure to be involved in substance misuse and/or offending is likely to be greater where an individual is sharing accommodation with others involved in these activities. In more extreme cases, offenders who are homeless may feel they have little to lose from offending and imprisonment. Prison provides a daily routine plus regular meals and clothing and thus may provide refuge of a sort from a harsh street life. Evidence from the SPCR, summarised in Table 6.9, is that fewer than half of prisoners were living in a regular house or flat that was owned or rented by themselves or a partner. This gives a sense of the transient, chaotic lives many were leading before being imprisoned. As was pointed out above, imprisonment is likely to threaten such stable tenancies as individuals have established. One of the ‘hidden costs’ of imprisonment results from the likelihood either that other members of a household have to re-locate or that accommodation goes unused. A neighbourhood’s crime rate is one of the components of its Deprivation Index score,11 and it is well-established (if in some ways slightly circular) to say that crime rates are higher in more deprived areas. A simple way of demonstrating the relationship between offending and the local environment is to compare crime victimisation rates for those who themselves offend with those who have not

11

Other components (in addition to the Crime domain) of the Index of Multiple Deprivation in England and Wales are income, employment, health and disability, education/skills/training, barriers to housing and services and living environment.

6 Social Exclusion and Offending

119

Table 6.10 Offending and victimisation rates

Were you a victim of crime in the last year? Did you commit an offence during the last year?

No Yes Total

No

Yes

Total

Yes (%)

1,925 424 2,349

1,574 917 2,491

3,499 1,341 4,840

45.0 68.4

Source: Offending, Crime and Justice Survey, 2005, 3rd edition, UK Data Archive

offended. From Table 6.10, based on data from the OCJS of 2005,12 it can be seen that 68.4% of those who had offended had themselves been a victim of crime compared with 45.0% of those who had not.

6.4.1 Anomalies The empirical regularities and correlations demonstrated seem all to be broadly consistent with the conjecture that a high propensity to offend is closely associated with an array of other indicators of social exclusion and, in particular, with less than average accumulation of human capital. However, there are some aspects of offending careers which remain unexplained using such an approach. Two of the more obvious ones are (a) that males are convicted substantially more frequently than females13 and (b) that the likelihood of conviction (and reconviction) declines very appreciably and varies predictably (although non-linearly) with age14: Ministry of Justice (2010b) and Bowles and Florackis (2007). Figure 6.2 illustrates the grounds for supposing that there is a biological basis of some sort for the offending propensity. Offending by girls reaches a peak at 15 and declines thereafter. For boys the propensity is still rising at age 17, although the rate of increase with age is slower from 16 to 17 than it is for 15 to 16. In the criminology literature these tendencies are ascribed to differences in gender development and to the idea that as individuals age they become more likely to settle down with family responsibilities which reduce their propensity to offend. This is puzzling to the extent that a career path based on low investment in human capital during adolescence along with involvement in offending is not a promising foundation on which to build a successful labour market career in later

12

Home Office. Research, Development and Statistics Directorate. Offending Surveys and Research, National Centre for Social Research and BMRB. Social Research, Offending, Crime and Justice Survey, 2005 (computer file). 3rd Edition. Colchester, Essex: UK Data Archive [distributor], December 2008. SN: 5601. 13 During the second quarter of 2010, for example, males represented 85.2% of those sentenced in all courts in England and Wales. 14 Detailed sentencing statistics for England and Wales are available from: http:// www.justice.gov.uk/sentencingquarterly.htm

120

R. Bowles 45,000 Male Female

40,000 35,000 30,000 25,000 20,000 15,000 10,000 5,000 -

10

11

12

13

14

15

16

17

18

Fig. 6.2 Youth crime disposals by age and gender, 2008/2009. Source: Graph based on youth justice statistics from the Ministry of Justice website: http://www.justice.gov.uk/publications/ youth-justice-statistics.htm

life. Social exclusion does not disappear with age but it seems that offending, and possibly substance misuse, does at least decline with age. Bowles and Florackis (2011) develop a model in which a decline with age in the propensity to offend is attributed to a reduction in the subjective discount rate. As individuals develop psychologically through adolescence and into early adulthood their planning skills and capacity to anticipate the future improve. The downside risks of offending may then feature more strongly, and may be compounded as judges increase the punishment tariff with repeated offending. This conjecture is argued to be consistent with the age-profiles of offending in Fig. 6.2 on the grounds that females mature at an earlier age than males.

6.5 Policy Options Encouraging young people to attend school and to pass examinations does not sound very path-breaking but seems to be the most direct means available of opening up labour market opportunities and thereby promoting resilience against social exclusion. Many criminal justice professionals are convinced that the best form of investment in crime reduction is early intervention in the form of (crime) prevention work with young people: Ashford (2007). In England and Wales youth crime prevention work is aimed both at younger children (in the 8–12 age range) and at adolescents in the 13–17 age range. The prevention work recognises the importance of school in providing a sound education and qualifications that can provide a foundation for acquiring job market skills. Evidence from adult offenders shows a heavy over-representation amongst those who achieved few qualifications and those who attended poorly or were excluded from school on the basis of bad behaviour.

6 Social Exclusion and Offending

121

The problem, of course, is that the policy instruments available may not be very effective. Persuading a young person who receives little support from their family (who may themselves be socially excluded because of substance use, unemployment, ill health or whatever), who has rebellious friends and little interest in schoolwork that it is worthwhile studying hard is likely to be a hard sell. For some in this position the wretchedness may itself act as a positive trigger and fire a determination to escape to a better life. But for many others the response is a negative one likely to condemn them to a continuing life in which they are excluded from mainstream society, Economists examining youth crime prevention programmes have found that the potential for very high returns in some individual cases may be sufficient to justify investment even where the proportion of successes may be small and uncertain: Cohen (1998). The stakes are high in individual cases because different choices at critical points in the life course may have effects that literally last a lifetime. It is difficult to construct convincing empirical analyses of these propositions because of the lengthy real-time requirements of cohort studies that cover childhood as well as adult offending careers and because of the difficulties in establishing credible counterfactuals. It is easier, but less compelling, to find anecdotes and case studies that illustrate positive outcomes from interventions in individual instances. A different way of looking at the policy issue is to consider instead how the wealth of information about the influences on offending might be used to help inform an economic analysis of resettlement policy. Using Fig. 6.1 as the focus for this discussion would suggest a basic split between ‘prevention’ policy aimed at raising aspirations and encouraging investment in human capital on the one hand and ‘resettlement’ policy on the other. Prevention policy can be further split down into three strands: operating on individual preferences and cognition; operating on household support and operating on attendance at, and behaviour in, school. Elements of all three are to be found in contemporary English crime and youth crime prevention programmes. Cognitive behaviour treatment programmes15 have been used with individual young people at risk of offending and Restorative Justice programmes have been used in some cases in an effort to get young offenders to consider the consequences of their actions for others. Efforts to help parents and carers improve the degree of support they provide to children and young people have taken the form of both voluntary and mandatory ‘parenting programmes’. Schools have been encouraged to take seriously the control of truancy and in many cases have begun working in partnership with the police to facilitate the identification of children believed to be at risk of offending. Encouraging young people to invest seriously in their own futures has always been one of the objectives of educators and the implication of the findings marshalled for this paper is that the highest priority should be afforded to this goal in the future.

15

For a review of the rationale and content of such programmes see Hollin and Palmer (2006).

122

R. Bowles

Resettlement programmes, likewise, have traditionally respected many of the findings here. ‘Holistic’ approaches recognising the wide range of issues faced by offenders, particularly prisoners at the point of release, have been based on the idea that an offender’s issues cannot be tackled piecemeal. Programmes based on concerted action in relation to a range of issues including substance misuse, mental health, accommodation, cognitive and training issues have been deployed in efforts to reduce reconviction rates: Dawson and Cuppleditch (2007) and Vennard (2007). One barrier to this development has been the potentially wide range across which may be spread the improved outcomes resulting if and when these programmes are successful. The traditional metric has been the impact on reconviction rates. Analysts are becoming aware that success can bring many collateral benefits: Metropolitan Police (2010). For example, although it has been common for analysts of drug treatment programmes to take account of the crime reduction impact of their work they are increasingly coming to take account of other benefits such as improved quality of life for clients, reduced pressure on health resources, reduced pressure on facilities for looking after neglected children and improved employment outcomes: Marsden et al. (2008). Likewise there remains scope for analysts of resettlement programmes to include a wider range of outcome measures in their measurement of benefits, since offenders are typically making heavy use of public services outside the criminal justice system; Davies et al. (2009). The heavy focus on reconviction rates at one or two years, which has dominated the English debate on prisoner resettlement and other projected interventions, is too narrowly based to be an effective measure of the financial returns on projects. Recent work by economists on drug treatment programmes (such as Davies 2009) has incorporated health status and health service utilisation measures along with the cost of reconvictions. There is scope to extend this work into other outcome measures including employment rates and to apply the same range of outcome measures in studies of offender programmes as well as health and drug programmes. An implication of this argument that the consequences of convictions run more widely is that convictions themselves may be regarded as destroying human capital insofar as they reduce the likelihood of a legitimate sector employment career. There are echoes of this kind of research agenda also in areas such as local government where many spending programmes (such as social care, housing, education and police) have an impact on various groups in the population including offenders. Efforts to identify the costs of providing support to socially excluded groups such as the elderly or offenders are potentially worth consulting for the information they may hold on the wider costs of offending. It is worthwhile noting at this point that the standard analysis of optimal enforcement policy against offenders omits the wider costs associated with offending. For example the standard Home Office approach to estimating the economic and social costs of crime (Brand and Price 2000; Dubourg et al. 2005)16 does not make allowance for

16

See also Cohen (2005)

6 Social Exclusion and Offending

123

the possibility of convictions destroying human capital via labour market discrimination against those with criminal records. It therefore runs the danger of under-estimating the scale of benefits delivered by successful crime reduction programmes, since standard economic evaluation methodology relies on ‘cost of crime’ estimates to value programme benefits: Dhiri and Brand (1999) and Bowles (2010). With substantial pressure on public budgets throughout Europe and beyond it is important that the costs and benefits as well as the cash flow implications of criminal justice policy decisions are viewed realistically and thoroughly.

6.6 Concluding Remarks This paper has explored the relationship between offending and social exclusion from an economic perspective. It argues that making use of the notion of social exclusion provides a richer model (than the standard economics of enforcement model) for analysing the context in which offending takes place. By giving the formation of human capital a pivotal role in the development process it is able both to incorporate home and social influences on the propensity to under-invest in human capital and also to provide a basis for analysing decisions (later in life) by convicted offenders to re-offend. Using recently published data from a variety of English surveys it suggests that the contemporary policy debate about criminal justice in Europe, particularly Britain, is more readily explicable using such an approach than it is when viewed from the perspective of standard, static economic models. Social exclusion is viewed as a co-product, with offending, of unwillingness on the part of young people to invest in their own human capital. This choice may itself be encouraged by an environment of deprivation and by inappropriate support of children and young people. There is no simple way of inducing greater levels of investment at this stage but there are many possible channels through which encouragement can potentially be supplied. Recent literature on labour market outcomes has begun to look more closely at the various influences on the development of cognitive and non-cognitive skills during childhood (Carneiro et al. 2007; Cunha et al. 2006; Cunha and Heckman 2008; ter Weel 2008) and there would seem to be good grounds for applying this approach to the analysis of offending careers. For adult offenders with a record of convictions effective crime reduction may require a range of measures including drug treatment programmes, job-related training, provision of accommodation with positive role models, support for employers taking on ex-offenders and the choice of appropriate incentive structures within the social security system. Intervening at this later stage may enable better targeting of resources but it is likely to be very costly and does not seem to be very effective. This puts a premium on producing a better understanding of the early development of offending careers and of the social exclusion with which they are often embedded.

124

R. Bowles

These kinds of conclusions do, however, have some potentially unpleasant implications for the perspective of the standard laissez faire model of economics where individual preferences are sovereign. Particularly where children and young people are involved there are real difficulties in arguing that society can intervene legitimately in family life to influence adolescent development and there by produce socially preferred outcomes. Acknowledgements The original data creators, depositors or copyright holders, the funders of the Data Collection and the UK Data Archive bear no responsibility for the analysis or interpretation of the OCJS data referred to in this paper. The author would like to thank two anonymous referees for very helpful comments on an earlier version of the paper.

References Ashford R (2007) Towards a youth crime prevention strategy. Youth Justice Board, London Becker GS (1968) Crime and punishment: an economic approach. J Political Econ 76(2):169–217 Bowles R (2010) Valuing the benefits from criminal justice interventions. In: Roman J, Dunworth T, Marsh K (eds) Cost benefit analysis and crime control. Urban Institute, Washington DC Bowles R, Florackis C (2007) Duration of the time to reconviction: evidence from UK prisoner discharge data. J Crim Justice 35:365–378 Bowles RA, Florackis C (2011) Impatience reputation and offending. Appl Econ 40 Bowles R, Pradiptyo R, Garcia Reyes M (2005) Safer school partnerships. Youth Justice Board, London Bowles R, Faure M, Garoupa N (2008) The scope of criminal law and criminal sanctions: an economic view and policy implications. J Law Soc 35(3):389–416 Brand S, Price R (2000) The economic and social costs of crime. Home Office, London Carneiro P, Crawford C, Goodman A (2007) The impact of early cognitive and non-cognitive skills on later outcomes. Centre for the economics of education, London school of economics, Discussion Paper no. 92 Cohen M (1998) The monetary value of saving a high risk youth. J Quant Criminol 14:5–33 Cohen M (2005) The costs of crime and justice. Routledge, Oxford Cunha F, Heckman JJ (2008) Formulating, identifying and estimating the technology of cognitive and non-cognitive skills. J Human Res 43(4):738–782 Cunha F, Heckman JJ, Lochner L, Masterov D (2006) Interpreting the evidence on life cycle skill formation. In: Handbook of the economics of education, vol 1. Elsevier, Amsterdam, pp 697–812 Cunliffe J, Shepherd A (2007) Re-offending of adults: results from the 2004 cohort. Home Office, London Home Office Statistical Bulletin 06/07 Davies L, Jones A, Vamvakas G, Dubourg R, Donmall M (2009) The drug treatment outcomes research study (DTORS): cost-effectiveness analysis, Research report 25, Key implications. Home Office, London. http://www.homeoffice.gov.uk/rds/pdfs07/horr03c.pdf. Dec 2009 Dawson P, Cuppleditch L (2007) An impact assessment of the prolific and other priority offender programme. Home Office, London Dhiri S, Brand S (1999) Analysis of costs and benefits: guidance for evaluators, crime reduction programme–guidance note 1. Home Office, London Dubourg R, Hamed J, Thorns J (2005) The economic and social costs of crime against individuals and households 2003/04. Home Office On-Line Report 30/05, London Ehrlich I (1973) Participation in illegitimate activities: a theoretical and empirical investigation. J Political Econ 81(3):521–565

6 Social Exclusion and Offending

125

Elliott-Marshal R, Ramsay M, Stewart D (2005) Alternative approaches to integrating offenders into the community. In: Harper G, Chitty C (eds) The impact of corrections on re-offending: a review of ‘what works’. Home Office, London Home Office Research Study 291 Gladwell M (2006) Million-dollar murray: why problems like homelessness may be easier to solve than to manage. New Yorker, Issue of 2006-02-13 and 20 Harper G, Mann L-H, Taylor S, Niven S (2005) Factors associated with offending. In: Harper G, Chitty C (eds) The impact of corrections on re-offending: a review of ‘what works’, 3rd edn. Home Office, London Home Office Research Study 291 Hollin CR, Palmer EJ (eds) (2006) Offending behaviour programmes: development, application and controversies. Wiley, Chichester Home Office (2010) Drug-misusing offenders: results from the 2009 cohort for England and Wales. Home Office, London. http://rds.homeoffice.gov.uk/rds/pdfs10/misc0410.pdf Marsden J, Farrell M, Bradbury C, Dale-Perera A, Eastwood B, Roxburgh M, Taylor S (2008) Development of the treatment outcomes profile. Addiction 103:1450–1460 Martinez TE, Burt MR (2006) Impact of permanent supportive housing on the use of acute care health services by homeless adults. Psychiatric Serv 57(7):992 ps.psychiatryonline.org May C, Sharma N, Stewart D (2008) Factors linked to reoffending: a one-year follow-up of prisoners who took part in the resettlement surveys 2001, 2003 and 2004. Ministry of Justice, London Research Summary 5 Metropolitan Police (2010) An evaluation of the diamond initiative: year one findings. Strategy, Research and Analysis Unit (SRAU) of the Metropolitan Police for London Criminal Justice Partnership. http://lcjb.cjsonline.gov.uk/area23/library/Diamond%20Initiative/Diamond%20 Initiative.pdf Ministry of Justice (2010a) Reoffending of adults: results from the 2008 cohort. Ministry of Justice, London. http://www.justice.gov.uk/compendium-of-reoffending-statistics-and-analysis.pdf. Mar 2010 Ministry of Justice (2010b) Compendium of reoffending statistics and analysis, Ministry of Justice Statistics bulletin, Nov 2010 National Audit Office (2010) Managing offenders on short custodial sentences. Report by the Comptroller and Auditor General. HC 431, Session 2009–2010 Niven S, Olagundoye J (2002) Jobs and homes: a survey of prisoners nearing release, Home Office Findings 173. http://www.homeoffice.gov.uk/rds/pdfs2/r173.pdf Niven S, Stewart D (2005) Resettlement outcomes on release from prison in 2003, Home Office Findings 248. http://www.homeoffice.gov.uk/rds/pdfs05/r248.pdf Singleton N, Meltzer H, Gatward R, Coid J, Deasy D (1998) Psychiatric morbidity among prisoners: summary report. Social Survey Division of Office of National Statistics for Department of Health Social Exclusion Unit (2002) Reducing re-offending by ex-prisoners. Social Exclusion Unit, Office of the Deputy Prime Minister, London Soothill K, Francis B, Fitzpatrick C (2009) Understanding criminal careers. Willan Publishing, Cullompton Stewart (2008) The problems and needs of newly-sentenced prisoners: results from a national survey. Ministry of Justice Research series 16/08. www.justice.gov.uk/publications/problemsneeds-prisoners.htm ter Weel B (2008) The non-cognitive determinants of labor market and behavioural outcomes: introduction to the symposium. J Human Res 43(4):729–737 Vennard J (2007) Chapter 11: The resettlement of prolific offenders: policy and practice. In: Hucklesby A, Hadley-Dickinson L (eds) Prisoner resettlement: policy and practice. Willan Publishing, Cullompton Youth Justice Board (2010) Youth Survey 2009. Youth Justice Board, London

Chapter 7

Disability and Social Exclusion Peter J. Sloane and Melanie K. Jones

Abstract This paper examines the recorded incidence of disability across European countries and draws attention to the considerable measurement problems involved in the economic analysis of the phenomenon. However, the distinction between work-limited and non-work-limited disability turns out to be particularly helpful in understanding labour market outcomes. Finally, policy alternatives for increasing the degree of social inclusion of the disabled population are evaluated. Keywords Disability JEL Codes I1

Measurement Work-limitations Government policy

J2 J7

7.1 Introduction The objective of improving the relative position of disadvantaged groups has become a major concern of the European Union, as well as elsewhere. Thus, since 1999 in particular there has been concerted action to ban discrimination on the P. J. Sloane (&) M. K. Jones School of Business and Economics, Swansea University, Swansea, UK e-mail: [email protected] M. K. Jones e-mail: [email protected] P. J. Sloane National Institute of Labour Studies, Flinders University, Adelaide, Australia P. J. Sloane IZA, Bonn, Germany

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_7, Springer-Verlag Berlin Heidelberg 2012

127

128

P. J. Sloane and M. K. Jones

grounds of gender, racial or ethnic origin or belief, age or sexual orientation, as well as disability. Yet, there is reason to believe that the degree of disadvantage is more severe for a substantial number of disabled individuals compared to any of the other groups including, in particular, on the basis of non-participation in the labour market and levels of family income (see Eurostat 2001). Disability poses a number of problems for the economist in terms of analysis. It takes several forms and varies considerably in terms of its severity. Unlike the case of most other groups (age also being an exception) we cannot assume that there are no negative effects on productivity. Certain individuals may be born with a disability, while others may suffer from its onset at a later age and the causes may include the effects of the working environment. There are particular problems of definition and measurement, since the probability of someone reporting disability may depend on the nature of the available support mechanisms and the generosity of transfer payments. As the foreword to an OECD Report (2003) points out, disability policy faces ‘‘two twin but potentially contradictory goals’’ (page 3). One goal is to minimise the possibility that the disabled are excluded by ensuring that as many of the disabled as possible are able to participate in the labour market. The second goal is to ensure that the disabled have income protection, so that their standard of living is not reduced too dramatically if they are unable to work. The OECD Report suggests that the problem of attempting to reconcile these two goals has yet to be satisfactorily resolved. In this paper we first present some facts on the incidence of disability in various European countries. We then examine a number of measurement problems in analysing disability. Next, using the UK example we show how the distinction between work-limited and non-work-limited disability can be used in empirical work to cast light on a number of issues. Finally, we consider alternative policies for achieving greater social inclusion for disabled workers.

7.2 The Facts One might expect that the incidence of disability would be similar in countries with similar health status among their inhabitants. Yet this does not appear to be the case as illustrated in the figures in Table 7.1. These figures are derived from an ad hoc module of the Labour Force Survey conducted in 2002 (see Dupré and Karjalainen 2003). Respondents aged 16–64 were asked an identical question, namely whether they had a longstanding health problem or disability lasting for six months or more or expected to last six months or more. One in six individuals across the European Union as a whole (15.7%) stated that they did have a long term health problem or disability defined in this way and 78% of those who reported severe disability were outside the labour force compared to only 27% of those without any long term health problem or disability. Dupré and Karjalainen (2003) document several commonalities across countries. The incidence of disability is much higher for those with low education and rises with age in all

7 Disability and Social Exclusion Table 7.1 Percentage incidence of disability among the adult population (16–64) in 2002—25 countries

129

High Finland UK Netherlands France Estonia Czech Republic Portugal Sweden Denmark Slovenia Belgium Norway

Low 32.2 27.2 25.4 24.6 23.7 20.2 20.1 19.9 19.9 19.5 18.4 16.4

Austria Cyprus Luxembourg Hungary Germany Ireland Greece Spain Malta Lithuania Slovakia Italy Romania

12.8 12.2 11.7 11.3 11.2 11.0 10.3 8.7 8.5 8.4 8.2 6.6 5.8

Source: EU Labour Force Survey Ad Hoc Module, 2002, adapted from Dupré and Karjalainen (2003). High (low) refers to countries above (below) the unweighted EU average

countries. For example, in the EU 15 countries 19.0% of those educated to ISCED 0–2 are disabled, compared to 15.6% with ISCED 3 and 12.1% of those with ISCED 4–6.1 Using these data they are also able to identify the cause and type of the disability. Overall, only 16.9% of the sample had such a condition since birth, and for only 18.4% was the cause work-related. The most common forms of disability were musculoskeletal problems, especially with the back or neck legs or feet, followed by heart, blood pressure and circulation problems. However, in a number of countries mental illness has become an increasingly important type of disability, especially when considering disability benefit claimants (see OECD 2006, 2007, 2008). In Finland, Denmark and the Netherlands it is now more important than muscular-skeletal problems in terms of inflow on to disability benefits among young and prime age individuals, and overall it accounts for over 40% of the total inflow (OECD 2006). In the UK, too, the prevalence of mental illness has increased sharply, reaching 40% of all disability benefit recipients by 2006 (OECD 2007). Consistent with this, 22.5% of those with mental health problems were in employment in the UK in 2002, or less than half the rate for those with muscular-skeletal conditions. The incidence of mental health problems has also increased in Australia and the Netherlands and to a lesser extent in Norway and Poland (OECD 2006, 2007, 2008). Despite these commonalities incidence varies dramatically across countries and appears generally to be higher in Northern Europe than in Southern Europe. This may reflect the relative generosity of welfare payments in the former relative to the latter countries. In countries such as Finland, Luxembourg and Spain disability

1

Further, previous analysis, has shown that education is a particularly important determinant of labour force participation among the disabled. See later discussion for more details.

130

P. J. Sloane and M. K. Jones

Table 7.2 Employment rate by the severity of disability (where available), percentage of 20–64 population, late 1990s Disabled Non-disabled 5 4 Ratio of 1–4 1 All 2 Severe 3 Moderate Australia Austria Belgium Canada Denmark France Germany Italy Korea Mexico Netherlands Norway Poland Portugal Spain Sweden Switzerland UK US OECD (19)

41.9 43.4 33.5 56.3 48.2 47.9 46.1 32.1 45.9 47.2 39.9 61.7 20.8 43.9 22.1 52.6 62.2 38.9 48.6 43.9

31.4 23.9 21.1 N.A. 23.3 36.4 27.0 19.4 13.4 N.A. 26.5 N.A. N.A. 27.6 15.1 33.8 N.A. 19.3 26.4 N.A.

46.9 50.2 40.0 N.A. 55.1 55.5 52.9 37.9 51.5 N.A. 46.4 N.A. N.A. 55.3 26.5 69.0 N.A. 46.8 58.8 N.A.

76.6 71.8 61.7 78.4 79.4 66.6 69.0 53.8 61.7 61.1 67.0 85.8 71.2 74.0 54.2 75.8 79.1 73.9 83.9 70.8

54.7 60.4 54.3 71.8 60.7 71.9 66.8 60.0 74.4 77.3 60.0 71.9 29.2 59.3 40.8 69.4 78.6 52.6 57.9 62.0

Source: Adapted from Table 3.3, Transforming Disability into Ability: Policies to Promote Work and Income Security for Disabled People, OECD 2003

benefits tend to be used as a form of early retirement, so a high proportion of the disabled are over 55 (OECD 2007, 2008). Norway has one of the highest proportions of workers on disability benefits in the OECD and most policies are focused on benefits rather than work inclusion strategies (OECD 2006). There are, however, some anomalies in the case of neighbouring countries. Thus, while disability incidence in the UK is 27.2% in Ireland it is only 11.0%, and in Portugal incidence is 20.1% compared to only 8.7% in Spain. This raises the question of the relationship between disability incidence and labour force participation. If low incidence reflects setting a high standard for qualifying as disabled, this should impact negatively on labour force participation among those recorded as disabled, as there will be fewer individuals with moderate disabilities who find it easy to participate. The reverse may be true for countries with a low threshold for disability. However, if the threshold for self-reporting disability is related to the standards set to qualify for out of work disability benefits then a low threshold for disability may be associated with fewer incentives to participate. Hence in this case the outcome is indeterminate. In Table 7.2 employment rates are provided by degree of severity of disability relative to the non-disabled for 19 OECD countries. In all countries the probability

7 Disability and Social Exclusion

131

of employment is much higher for the moderately disabled relatively to the severely disabled, though we do not know the extent to which the scales of these correspond across countries. We must also allow for the fact that employment rates for the non-disabled vary substantially across countries. Ignoring Poland, which is an outlier, the ratio of employment rates between disabled and non-disabled groups (presented in column 5) vary from 40.8% in Spain to 78.6% in Switzerland. Thus, the Spanish example is consistent with a high threshold for defining disability being associated with a low employment rate. We can in general conclude that in all countries the employment rate for the disabled is substantially lower than that of the non-disabled. The above table does not control for differences in hours of work between the disabled employees and non-disabled employees. According to Eurostat (2001) in EU14 19.4% of the severely disabled work 30 h or less compared to 14.9% of the moderately disabled and 11.1% of the non-disabled. In part this will be influenced by the fact that for some of the disabled their disability affects the amount of work they can do. There is abundant evidence that the disabled suffer from low incomes relative to the non-disabled. First, there is a body of evidence which shows that, all else constant, the disabled earn less when in work compared to the non-disabled (DeLeire 2001; Jones et al. 2006). However, as noted above, the disabled also have a lower propensity to work and a tendency to work fewer hours. Further OECD (2003) reports that the total personal income of unemployed disabled persons is on average only half of that of disabled people in employment. Table 7.3 compares the average income of working age people with a disability with that of the income of the working age population or of the non-disabled, according to the availability of the data, for a number of countries, where income data are available. While some of the definitions vary between countries making cross country comparisons problematic, in every case the incomes of disabled individuals are consistently lower than those without a disability. Further insights can be obtained by analysing longitudinal data. Thus, Bardesi, Jenkins and Rigg (2000) show that most of the disadvantage suffered by the disabled can be accounted for by the low economic status of those who become disabled. Using the BHPS 1991–1998 they show that the income of working age men before the onset of a disability is just 10% higher than working age men who already have a disability. Though the post disability gross income of disabled working age men is 20% lower than their non-disabled equivalents their actual reduction in net income is only 8%, cushioned as it is by pension benefits and tax reductions. In the case of Germany, Lechner and Vasquez-Alvarez (2009) utilise the German Socio-Economic Panel 1984–2002 to examine the effects of becoming disabled on labour market outcomes. They find no significant evidence of reductions in income or increases in unemployment due to disability onset which they attribute to the effectiveness of the German social security system. It is suggested, however, that disability onset is associated with additional costs, with the implication that even with the same level of income as the non-disabled they would have a lower standard of living. Thus, one study found that without

132

P. J. Sloane and M. K. Jones

Table 7.3 Relative income of working age population with a disability, circa 2005 Country Source Definition Date Percentage Norway Poland Switzerland Finland Denmark Ireland Netherlands

OECD Disposable income per equivalent person relative to 2004 91 (2006) the non-disabled 2004 83 2002 90 OECD Disposable income per equivalent person relative to 2005 91 (2008) the entire working age population 88 71 OECD Disposable household income relative to the entire 2004 87 (2008) working age population Luxembourg OECD Disposable household income per equivalent 2004 89 (2007) person relative to the non-disabled Spain 86 UK 73 Australia Gross household income per equivalent person 2003 65 relative to the non-disabled Source: OECD 2006, 2007 and 2008 Notes: All figures refer to individuals of working age

using an equivalence scale 23% of households in the UK with disabled household members had less than 60% of the median income, but when adjustments were made for additional expenditure requirements associated with disability, such as transportation and medical services, this figure rose to over 47% (see Braithwaite and Mont 2008).

7.3 The Measurement Problem The above suggests that there may be differences in the extent to which individuals with a given degree of health problems are likely to report themselves as disabled in different countries. The first question is to establish what is meant by the term disability. There is no single definition. As Thornton and Lunt (1997) point out definitions vary from one country to another and may be formulated for specific purposes such as • to determine who will benefit from national policies towards the disabled • to outline the characteristics of a group with rights under the law (e.g. the right not to be discriminated against) • to define the target group for particular policy initiatives (such as who will count towards meeting a quota, where this applies) • to determine individual eligibility for specific services or measures (such as qualification for sheltered employment). Whilst a common definition could apply to all of these this is not necessarily always the case and this in itself may influence answers to common survey questions. OECD (2003) reports that many of those who report themselves as disabled do not receive any disability benefits, while many of those on disability

7 Disability and Social Exclusion

133

benefits do not report themselves as disabled. Further, many of those classifying themselves as severely disabled and who do not work receive no benefits.2 This could occur through lack of fulfilment of insurance requirements or failing a household means test. The two requirements for a satisfactory definition of disability are that the condition should not be short-term and it should be more than a minor ailment. Thus, in the case of the UK, respondents to the Labour Force Survey are asked first; ‘Do you have any health problems or disabilities that you expect will last more than one year?’; second, ‘Does this health problem affect the kind of paid work you might do?’ and third, ‘Does this health problem affect the amount of paid work you might do?’ According to the UK Labour Force Survey 2009, of those who had a disability just over half said it affected the kind of work they could do and 43% said that their disability affected the amount of work they could do, and this was slightly higher for women than for men. If positive answers are given to the first and either the second or third question one can classify an individual as work-limited disabled. The large majority of this group are limited in both the amount and kind of work they can do. This enables one to distinguish between the work-limited disabled, the non-work-limited disabled and the non-disabled and this three-way split turns out to be particularly informative as is shown below. However, there are some caveats. The precise nature of and the order in which questions are asked influence the responses. For example, after such changes in the 1997 UK LFS, 24% fewer respondents reported a long-term disability which affected the kind of work they might do, and of those responding positively a greater proportion were economically inactive (Cousins et al. 1998). The Household Income and Labour Dynamics in Australia (HILDA) Survey is particularly rich in its coverage of disability.3 Further, the panel nature of these data has the advantage of facilitating analysis of transitions in self-reported disability status. Using the first five waves of HILDA Oguzoglu (2007) finds that 71% of men and 73% of women always report themselves as non-disabled, whereas 8% of men and 5.5% of women always report themselves as disabled. Roughly 18% of both men and women exhibit an irregular pattern of limitation. Thus, for many, disability is a temporary phenomenon. A further consideration on which data are often lacking in Europe is age of onset of disability. Mavromaras et al. (2007), using the HILDA data, suggest that

2

According to Eurostat (2001) in the mid 1990s sickness and disability benefits covered 48% of those reporting a severe disability. In Belgium, Spain and Denmark such benefits covered more than half this group but the proportion in Greece, Italy, Luxembourg, Austria and Portugal were in the range 27–36%, while the UK had the maximum coverage of 65%. 3 The HILDA survey asks ‘Do you have a long-term health condition, impairment or disability that restricts you in your everyday activities, and has lasted or is likely to last, for 6 months or more?’. Respondents are then asked to list which of 17 types of disability affects them. They are also asked whether these conditions limit the type or amount of work they can do and to pick a number between 0 and 10 to indicate how much their conditions limit the amount of work they can do.

134

P. J. Sloane and M. K. Jones

mature age onset is associated with the worst labour market outcomes. In particular more than half of males and more than 60% of females who develop disability in mature age do not participate further in the labour market. This may be because individuals adapt to their disabilities better when they are younger. Using the same data, Oguzoglu (2009) finds that recovery from a severe limitations shock (in terms of returning to the labour market) takes an average of 5 years for a woman and 3 years for a man. There is, however, a wider measurement error problem as individuals may use different response scales to self-report disabilities. As we have seen the fraction of workers reporting disability is very different across countries with similar economic (and health) development. Kapteyn, Smith and van Soest (2007) use ‘‘vignette’’ questions on internet surveys in the US and Netherlands, and ask respondents to evaluate on the same scale as they normally use the severity of work disability problems in hypothetical scenarios. The answers imply that work disability is almost 20% points higher among those aged 45–54 in the Netherlands than the US, the two countries compared in the study, despite evidence from elsewhere that the Dutch are healthier than Americans. In fact this study suggests that about half the difference in reported rates of disability can be attributed to response scale differences. It seems likely that different response scales are an important cause of the differences in reported incidence of disability across European countries, though Datta Gupta, Kristensen and Pozzoli (2009) have questioned the assumption of response consistency which forms the basis of this approach. Much of the previous literature has been concerned with the problem of justification bias or the possibility that some individuals may claim to be disabled in order to justify their inferior labour market status or non-participation in the labour market. The empirical evidence on the extent to which this is common is mixed and may be partly overcome by using specific health measures, which, while also self reported may be less prone to such biases. There is a further complicating factor when considering work-limiting disability in that those who are out of the labour market may consider a wider range of occupations when making this judgement than those who say they are work-limited, but who are in the labour market. Another form of endogeneity arises from the fact that as well as poor health having an effect on the propensity to work, some work situations may themselves cause poor health. This suggests that a simultaneous equations model (or IV approach) may be required in order to obtain efficient estimates of the impact of health on labour supply. There is again disagreement in the literature about the importance of this form of endogeneity and as noted above less than 20% of disability appears to be work related. However, using data from HILDA, Cai (2010) shows that while good health has a positive and significant effect on labour force participation both in the case of men and women, employment has a general negative effect on men’s health and a positive effect on women’s health. This implies the justification hypothesis can be rejected for men but not for women.

7 Disability and Social Exclusion

135

Notwithstanding the difficulties involved when using international data, several studies have used the European Community Household Panel Survey (ECHP) 1994–2001, which contains two binary measures of health/disability. The first asks respondents; ‘Do you have any chronic physical or mental health problems, illness or disability?’ A second question asks; ‘Are you hampered in your daily activities by this physical or mental health problem, illness or disability?’ They are then asked if this is severely, to some extent, or not at all. However, we do not know whether or not the problem is long term or whether it has an impact on work activities. Disappointingly there are no direct questions on the timing of disability onset but individuals who become disabled during the panel can be identified and this is important for identifying the causal impact of disability on labour market outcomes. Despite these limitations Hernandez-Quevedo, Jones and Rice (2007) use ECHP data to test for individual heterogeneity and state dependence in disability incidence. Heterogeneity appears to account for half of the unexplained variation in the presence of any health limitations and the estimate of state dependence (or the effect of previous disability on current status) is large and significant in all countries. A second paper to use the same data-set is Hagan, Jones and Rice (2009). They utilise a discrete time proportional hazards model to examine the relationship between health and retirement and argue that since they use self reported measures of health recorded prior to the exit decision this should reduce the problem of justification bias. They find marked cross country heterogeneity in the magnitude of responses to health shocks. Ireland, Portugal and Spain are the countries where health shocks have the greatest impact. Ireland has a generous early retirement benefit scheme and Portugal the highest ratio of disability pension related expenditure to old age pension expenditure. However, the magnitude of the effect in Spain is surprising given the low incidence of disability there, but is possibly explained by the use of disability benefits as a form of early retirement as discussed earlier.

7.4 Disability and Labour Market Outcomes: A UK Example The above discussion suggests that comparative analyses across countries are fraught with difficulties because measures of incidence may be non-comparable. For this reason much of the literature focuses on a single country and our discussion is restricted to the example of the UK. Differences between the disabled and non-disabled, as we have seen, are substantial in relation to employment rates, but smaller in terms of wage rates when in employment. Both gaps may be due to both observable (factors such as education) and unobservable (factors such as work commitment) characteristics in addition to the presence or otherwise of disability. In examining wage differences it also needs to be recognised that there is a potential for selection bias, as those in the labour market may not be a random subset of the total population.

136

P. J. Sloane and M. K. Jones

First, in considering whether there is discrimination against the disabled allowance has to be made for the possibility that the disabled are less productive than the non-disabled, though this may vary according to the severity and type of disability and will likely also vary with job demands. Following DeLeire (2001) and Jones et al. (2006) it is possible to make use of the distinction between worklimited and non-work-limited disability to cast light on this, using a conventional decomposition analysis. If we assume that the disabled who report no work limitations do not have lower productivity as a consequence of their impairment relative to the non-disabled, we can interpret the unexplained residual in a Oaxacatype decomposition as an estimate of discrimination. Further, if we then assume that the degree of discrimination is identical for the two groups of disabled workers, the unexplained residual for the work-limited disabled can then be regarded as an estimate of the lower productivity of the work-limited disabled relative to the non-disabled, which is not captured by the measured personal characteristics included in the empirical model. The fact that outcomes for the non-work-limited disabled approximate those of the non-disabled rather than the work-limited disabled suggests that the productivity effect is important. The unexplained component between the non-work-limited and non-disabled, that is, the measure of discrimination is found to be very small. These findings are in line with those of DeLeire for the US. However, a potential problem here is that the characteristics of the two groups may differ, such that one group is clearly identifiable from the other and possibly more subject to discrimination. Indeed, there are significant and important differences in the distribution of types of disabilities between work-limiting and non-work-limiting disability (Jones et al. 2006).4 One of the other interesting features to emerge from the analysis of Jones et al. (2006) is the importance of education. In Table 7.4 we present the results of highest educational qualification on employment and earnings. Possessing higher level educational qualifications has a highly significant positive impact on the likelihood of being in employment for all groups, though less obviously so for the more highly educated males within the non-work-limited group. Similarly it has a highly significant positive effect on earnings, with the exception of the less welleducated men and women within the non-work-limited group. More revealing are the marginal effects in the employment probits in Table 7.5. Education has the strongest effect of any of the explanatory variables on the likelihood of being in employment and the effect is much stronger for the work-limited disabled. Thus, a 4

When specifications are estimated with controls for the type of health for both disabled groups all variables turn out to be positively significant compared to those with mental health problemsthe omitted category. Mental health is, therefore, more problematical, both for entering the labour market and for obtaining earnings comparable to those of other workers. This distinction has recently been emphasised by Longhi, Nicoletti and Platt (2010) who extend the work of DeLeire (2001) and Jones et al. (2006) by redefining productivity differences to incorporate co-disabilities and sickness absence and splitting the disabled into those with physical and mental disabilities. Using data on men in the UK LFS over the period 1997-2008 their quantile regression results show a residual wage gap, or discrimination, only for those with mental health problems and this only at the top of the wage distribution.

7 Disability and Social Exclusion

137

Table 7.4 The influence of education on employment and earnings by disability status, UK LFS 2003 Work-limited disabled Non-work-limited disabled Non-disabled Males Degree Other higher education A level O level Other Lambda Observations Females Degree Other higher education A level O level Other Lambda Observations

Earnings

Employ

Earnings

Employ

Earnings

Employ

0.343*** (6.61) 0.200***

0.946*** (14.09) 0.860***

0.374*** (12.11) 0.166***

0.170* (1.89) 0.013

0.399*** (27.63) 0.203***

0.433*** (11.68) 0.452***

(3.89) 0.138*** (3.57) 0.120*** (2.88) 0.103*** (2.70) 0.010 (0.18) 2,409

(11.49) 0.587*** (12.32) 0.620*** (10.96) 0.421*** (7.77)

(0.13) 0.323*** (4.22) 0.218** (2.49) 0.310*** (3.39)

4,834

(13.44) 0.135*** (10.77) 0.066*** (5.10) 0.069*** (5.11) 0.029 (1.16) 21,389

(9.58) 0.437*** (13.19) 0.481*** (13.66) 0.434*** (11.27)

7,780

(5.20) 0.094*** (3.68) 0.039 (1.42) 0.020 (0.70) 0.041 (0.74) 3,899

0.483*** (9.25) 0.310***

1.030*** (15.27) 0.960***

0.338*** (10.83) 0.210***

0.722*** (9.19) 0.739***

0.327*** (24.06) 0.184***

0.720*** (23.41) 0.842***

(6.44) 0.207*** (4.99) 0.171*** (4.62) 0.137*** (3.78) 0.156*** (3.00) 2,478

(15.33) 0.763*** (14.10) 0.664*** (14.02) 0.538*** (10.38)

(7.20) 0.047* (1.83) 0.042* (1.77) 0.017 (0.73) -0.022 (0.57) 3,764

(9.12) 0.638*** (9.17) 0.640*** (10.25) 0.422*** (6.24)

(13.74) 0.094*** (8.23) 0.034*** (3.20) 0.030*** (2.65) -0.058*** (3.83) 22,627

(23.91) 0.544*** (19.24) 0.561*** (21.72) 0.402*** (13.69)

7,938

5,309

27,302

33,023

Source: Adaptation of Tables 2, 3, 4 and 5 from Jones et al. (2006). Wage equations also include controls for region, occupation, industry, absence, ethnicity, marriage, experience, experience squared, small firms, part-time employment, tenure and tenure squared, sector, overtime and housing tenure. Employment probit model also includes controls for region, ethnicity, marriage, housing tenure, age and age squared, presence of dependent children in household and of another earner. *** ** , and * denote significance of the coefficients at the 1, 5 and 10% level respectively

work-limited disabled man with a degree is 35% points more likely to be in employment than his equivalent without any qualifications, while this effect is only 9% points for a non-disabled male. The usefulness of the distinction between work-limited and non-work-limited disability is also shown in studies of part time and self employment. Thus, Jones (2007) notes that disabled workers in the UK are much more likely to work

138

P. J. Sloane and M. K. Jones

Table 7.5 Marginal effects associated with the influence of education on employment by disability status, UK LFS 2003 Work-limited disabled Non-work-limited disabled Non-disabled Males Degree Other higher education A level O level Other Females Degree Other higher education A level

O level Other

0.353*** (13.80) 0.322*** (11.08) 0.203*** (11.86) 0.223*** (10.34) 0.147*** (7.37)

0.034** (2.02) 0.003 (0.13) 0.064*** (4.49) 0.042*** (2.73) 0.057*** (3.91)

0.090*** (13.47) 0.087*** (12.13) 0.094*** (14.51) 0.098*** (16.12) 0.087*** (13.67)

0.388*** (15.48) 0.361*** (15.24) 0.281*** (13.54) 0.238*** (13.52) 0.193*** (9.87)

0.189*** (11.96) 0.189*** (12.28) 0.176*** (11.08) 0.185*** (11.59) 0.123*** (7.05)

0.205*** (29.18) 0.221*** (33.81) 0.164*** (22.09) 0.175*** (23.89) 0.124*** (15.35)

Marginal effects (evaluated at the sample means of the data) accompany employment probit models presented in Jones et al. (2006) Tables 2 and 3 respectively. Z statistics reported in parenthesis. *** ** , and * denote significance of the marginal effects at the 1, 5 and 10% level

part-time than the non-disabled, but more particularly the unexplained disability gap in part-time employment for work-limited disabled workers is not due to employer discrimination, but rather reflects differences in the preferences for parttime employment among this group. Part-time employment affected 11.3% of work-limited disabled men, compared to 5.0% of the non-disabled and 5.25% of the non-work-limited disabled. For women the figures were 49.4, 39.2 and 39.5% respectively. Similarly, Jones and Latreille (forthcoming) find for the UK that those whose impairment is not work-limiting have self employment rates which are very similar to those of the non-disabled, while self employment rates among the work-limited disabled are significantly higher. They report for men that 21.3% of work-limited disabled, compared to 16.6% of the non-work-limited disabled and 17.4% of the non-disabled are self employed. For women the figures are 9.3, 6.3 and 7.3% respectively. Similar results are reported by Pagan (2009) for various European countries, using the European Community Household Panel (ECHP) 1995–2001. For almost all European countries analysed, self-employment (as a percentage of total employment) is higher for disabled individuals compared with

7 Disability and Social Exclusion

139

non-disabled individuals. The greatest differentials are found in Southern European countries such as Greece and Portugal. A further consideration for those disabled workers is whether they are able to obtain employment which utilises their skills and experience, that is, a good ‘match’. Mismatch may result not simply from discrimination, but from reduced productivity amongst the disabled narrowing job choices. In addition, there may be spatial constraints on job search as a result of transport limitations or hours or accommodation constraints. Regardless of the source of the constraint on employment opportunities this increases the risk that a disabled individual will accept mismatched employment. However, despite this, the earlier literature on skill mismatch has largely ignored the issue of disability. Jones and Sloane (2009) attempt to remedy this omission by making use of the UK Workplace Employment Relations Survey 2004, which contains questions both on disability, whether worklimited or not, and on self-assessed skill mismatch measured on a five point scale. Descriptive data show that the disabled and particularly the work-limited disabled are significantly more likely to say they are over-skilled than the non-disabled. This is confirmed in a multi-nomial logit model of the determinants of skill mismatch in which both the non-work-limited and work-limited disabled are found to be more likely to be mismatched (with respect to both under-skilling and overskilling), with marginal effects stronger for the work-limited disabled. In Table 7.6 (which is adapted from Jones and Sloane 2009) we consider the implications of skill mismatch for hourly earnings for each of the three disability groups. Both specifications show that in terms of hourly earnings the non-worklimited disabled are not significantly different from those of the non-disabled. However, the work-limited disabled earn nearly 6% less than the non-disabled. In addition there is a penalty to over-skilling of over 3%, but no penalty to being under-skilled. When separate equations are run for the three groups the pay penalty to being over-skilled ranges from just over 2% for the non-disabled, to 6% for the non-work-limited disabled and over 10% for the work-limited disabled. Jones and Sloane (2009) also consider the effects of skill mismatch on the job satisfaction of each group in terms of seven facets of satisfaction and a constructed index of satisfaction. In each of these cases the job satisfaction of the work-limited disabled is significantly lower than that of the non-disabled in terms of the descriptive statistics, but in only three cases is this so for the non-work-limited disabled. (One factor which we do not consider here is the possibility that increased workplace disability diversity is associated with lower employee wellbeing among those with no reported disability, see Haile 2009). Table 7.7 presents some key results from the ordered probit and random effects ordered probit models for the determinants of job satisfaction estimated by Jones and Sloane (2009). In both specifications the disabled have significantly lower job satisfaction than the non-disabled, as do those who are either under-or over-skilled. Finally noting that the disabled appear to have less control over their work than the non-disabled Jones and Sloane (2009) construct an index of worker control which combines the ability to influence tasks, the pace of work, how the individual does the work, the order of work and the time of arrival and finish of work. This index was found to

140

P. J. Sloane and M. K. Jones

Table 7.6 Determinants of hourly earnings in Britain, WERS 2004 OLS Random effects GLS

Non-work-limited disabled Work-limited disabled Underskill Overskill Observations

All

All

Non-disabled

Non-work-limited disabled

Work-limited disabled

-0.000 (0.01) -0.057*** (4.75) 0.006 (0.45) -0.033*** (6.32) 18,156

0.001 (0.16) -0.047*** (4.18) -0.004 (0.30) -0.026*** (5.25) 18,156

0.000 (0.03) -0.022*** (4.19) 15,988

-0.038 (0.80) -0.060*** (2.96) 1,343

-0.002 (0.04) -0.104*** (4.01) 825

Source: Adapted from Table 5 of Jones and Sloane (2009). Specifications with random effects take into account that there are individuals employed within the same workplace. Wage equation also includes controls for gender, tenure, qualifications, part-time employment, union membership, marital status, ethnicity, workplace size, sector, an index of worker control, age, temporary contract, presence of children and a full set of regional, occupational and industry dummies. *** ** , and * denote significance at the 1, 5 and 10% level

have a negative impact on the probability of being mismatched, and a positive effect on both wages and job satisfaction. Therefore, they argue that giving workers greater control over their own work can have considerable benefits.

7.5 Policy Issues 7.5.1 Income Support The number of individuals on disability benefits has been increasing in a number of countries. Thus, in the European Union there were approximately 26 million claimants in 2005, equivalent to 5% of the employed labour force, which amounted to an increase of 5 million over the start of the decade (Benitez-Silva et al. 2010). In the US, Autor and Duggan (2003) note that the share of non-elderly adults receiving Social Security Disability Insurance rose by 60% between 1984 and 2001. They attribute this to reduced screening stringency, declining demand for less skilled workers and an increase in the earnings replacement ratio as a consequence of increased earnings dispersion. In turn this led to a one half percent reduction in the unemployment rates of non-elderly adults. Gruber (2000) utilises the fact that in Canada there are separate arrangements for disability insurance in Quebec and the rest of Canada. The latter programme increased the replacement rate of such benefits by 36% in 1987, whilst benefits remained constant in Quebec. He uses this policy difference to estimate the elasticity of labour force non-participation with respect to disability insurance benefits (using a difference-in-

7 Disability and Social Exclusion

141

Table 7.7 Determinants of job satisfaction in Britain, WERS 2004 Ordered Random effects ordered probit model probit All All Non-disabled Non-work-limiting Work-limiting disabled disabled Non-work-limited disabled Work-limited disabled Underskilled Overskilled Control index Observations

-0.087*** (2.92) -0.216*** (5.77) (1.73) -0.473*** (12.20) -0.300*** (18.46) 0.134*** (56.41) 18,028

-0.076** (2.47) -0.214*** (5.58) (1.22) -0.462*** (11.63) -0.300*** (17.98) 0.140*** (56.54) 18,028

(0.76) -0.442*** (10.30) -0.309*** (17.37) 0.139*** (52.22) 15,894

(1.78) -0.538*** (3.68) -0.284*** (4.50) 0.146*** (15.37) 1,318

(0.24) -0.717*** (4.17) -0.230*** (2.75) 0.144*** (11.67) 816

Source: Adapted from Table 6 of Jones and Sloane (2009). Specifications with random effects take into account that there are individuals employed within the same workplace. Equation also includes controls for gender, tenure, qualifications, part-time employment, union membership, marital status, ethnicity, workplace size, sector, age, temporary contract and a full set of regional, occupational and industry dummies. *** ** , and * denote significance at the 1, 5 and 10% level

difference approach) and finds this to be in the range 0.28–0.36. In Europe similar trends have been observed. Thus, in the UK the number of working age individuals claiming disability benefits rose from under 1 million in the mid 1970s to 2.7 million by 2003 (Beatty and Fothergill 2004). Within this figure the proportion of females raised from one-fifth to two-fifths. McVicar (2008) suggests that the explanations are consistent with the North American literature. Three recent papers have emphasised that disability insurance and unemployment insurance are substitutes. Bratsberg, Fevang and Roed (2010) observe that countries with low unemployment often have high rates of disability and this is particularly the case in Northern Europe. Rates of disability in Denmark, Finland, Norway, Sweden and the Netherlands are within the top six out of 28 OECD countries and each has low rates of unemployment relative to the OECD average. They note that in Norway the fraction of the working age population which receives health related social security benefits rose from 15.2 to 20.4% between 1993 and 2006, and the ratio of permanently disabled to the number of unemployed almost quadrupled at a time when the proportion of the adult population reporting themselves as in good health was rising. Koning and van Vuuren (2010) estimate that the degree of substitution between disability insurance and unemployment in the Netherlands over the period 1993–2002 is such that about one quarter of disability insurance enrolments constitute hidden unemployment. Finally, Benitez-Silva, Disney and JimenezMartin (2010) report that disability insurance programmes have been used to cushion recessions across most OECD countries. Variations in regional unemployment rates explain about 12% of the variation in self reported health problems

142

P. J. Sloane and M. K. Jones

and the elasticity of disability benefit claims to the local unemployment rate is 0.16–0.17. That is, a doubling of the unemployment rate is associated with 16–17% more claimants to disability benefits. Thus, the measurement of disability is not independent of policies impinging on those who are, or who are thought to be, disabled.5 In the light of these findings it seems that ensuring as many of the disabled as possible are in employment is a more appropriate policy than attempting to increase income support which may be a much more expensive strategy and one which is less well aligned in terms of social inclusion. As Burkhauser and Daly (2002) note, disability income transfer programmes have to overcome potential problems of moral hazard and work incentives. This is exacerbated by the fact that disability eligibility is not well defined.

7.5.2 Quotas Historically mandatory employment quotas have been widely used to assist the disabled to obtain gainful employment. Wuellrich (2010) suggests that a third of all OECD countries use them, with quotas ranging from 2% of employees in Spain to 7% in Italy. In cases of non-compliance employers may be required to pay a tax on any unfilled places, ranging from 0.25 to 4% of the monthly payroll. However the compliance rate is only 50% on average with a range of 25% in Spain to 67% in France, where a quota system was promoted under a 1957 Act. Certain employees may count more than once in reaching the target depending on age, severity or type of disability. In Austria those employing 25 or more are required to hire one registered disabled person for every 25 non-disabled employees. Employers are required to contribute to a levy fund which is used to reward employers who exceed the quota. In Germany those employing 16 or more are obliged to ensure that at least 6% of the workforce consists of severely disabled individuals, with very severe cases counting three times the norm. In Greece those employing 50 or more must have at least 7% of their workforce disabled, but effectiveness is limited by the fact that over 50% of the workforce is self employed and over 80% of enterprises have less than ten employees. Until comparatively recently there were no penalties for non-compliance. Italy has had a quota system since 1947 with different quotas for separate categories where more than 35 are employed, but with exemptions such as those for people over 55. Ireland has a public sector quota of 3%, while the Netherlands attempted to achieve a quota of between 3 and 5% by voluntary means (for a discussion of all of the above and eighteen countries in total see Thornton and Lunt 1997). In general it seems quotas have been set too low to bite and in the main the law seems to be disregarded, but a 5

Much less attention has been paid to the possibility that individuals who are in fact disabled may in fact not describe themselves as such for fear of being stigmatised. One effect of legislation and other policy measures to assist the disabled may be to remove or reduce the stigma, so that the recording of disability goes up.

7 Disability and Social Exclusion

143

further problem is how to define whether someone meets the quota, given the definitional problems outlined above. One response to this would be to increase the cost of non-compliance or increase the size of wage subsidies offered to employers who take on disabled workers.6 Thus, Wuellrich (2010) examined a non-compliance tax increase of 30% in Austria on firms not complying with the quota and found this led to a 2.5% increase in the average number of disabled workers in employment. Yet, quotas, even where they are implemented stringently appear to be a blunt instrument for improving the employment prospects of disabled workers since they may be set at too high or low a level. In the former case, some industries may be less able to hire disabled workers than others, while in the latter case rising proportions of disabled workers may render the quota redundant.

7.5.3 Sheltered Employment Employment in sheltered work environments is used in most OECD countries to provide work opportunities for disabled people. Visier (1998) provides information on its incidence in Europe in the early 1990s, though the extent to which incidence may have changed since then is not known. In this evidence the Netherlands stands out as having the highest share of the economically active population in sheltered employment, but this ratio is generally small. OECD (2006) notes that often such work is designed to serve a therapeutic function rather than achieve work task and responsibility functions. In France the quota employment obligation on employers may be met through contracts with sheltered workshops. In Italy there are over 300 sheltered workshops, but some of these involve no payments to the disabled. Visier (1998) suggests that sheltered employment would be more successful in integrating participants into normal jobs if working conditions were more similar to those in the regular labour market. There are also questions about the level of disability required to qualify for sheltered employment and whether this should be permanent or temporary as in Norway. Subsidies paid to employers of sheltered employment may require the implementation of reintegration services. However, sheltered employment is unlikely to achieve the same degree of social inclusion as normal employment.

7.5.4 Anti-Discrimination Legislation Disability anti-discrimination legislation is a relatively recent development. It was only in 2000 that the European Union issued a directive protecting the rights of disabled people in the fields of employment and training, but this is leaves member states with great discretion over how the mandates are implemented. 6

Wage subsidies may be offered to both employers and employees as in Austria, but employer subsidies are more common.

144

P. J. Sloane and M. K. Jones

Comprehensive legislation was passed in the US in 1990, Australia in 1992, the UK in 1995 and Ireland in 2005. Such legislation generally requires employers to make reasonable accommodations to remove any workplace disadvantages such as modifying hours of work and physical features of the workplace. In principle this attempts to more directly address the work-limiting nature of disability and the heterogeneous nature of requirements highlighted above. However, as noted by Acemoglu and Angrist (2001) these adjustments may have potential negative effects on employers through the imposition of additional costs. It also raises questions about who is covered and who is not. The more specific differences in provisions between countries are outlined in Pillinger (2003). In Germany and Spain for example, the anti-discrimination clause only applies to the severely disabled, defined in percentage terms. In some countries the reasonable accommodation duty has not been included in national legislation (e.g. Estonia, Italy and Poland). In other cases it is not clear whether failure to provide reasonable accommodation is to be treated as a form of unlawful discrimination (e.g. France, Hungary and Latvia), while in other cases it is not considered as such (e.g. Finland, Germany, Luxembourg and Portugal). Further, the amount of compensation offered to employees who have been found to be discriminated against varies across countries. Studies of the effects of the legislation on the employment of the disabled have been carried out in the US and the UK. In the US the results are conflicting with several studies finding negative effects (such as Acemoglu and Angrist 2001) and others insignificant effects (such as Hotchkiss 2004). In the UK, Bell and Heitmueller (2009) are unable to find any positive effects from the introduction of the Disability Discrimination Act (DDA). The extent to which this legislation can lead to a substantial improvement in the relative position of the disabled remains to be determined.

7.5.5 Employment Policy More and better targeted employment services seem essential if one is to maximise the employment potential of the disabled. A major reason for poor labour market outcomes for the disabled is their low level of academic qualifications. Further, OECD (2007) notes that in many countries the education gap between the disabled and non-disabled is widening rather than narrowing. Raising the education qualifications of the disabled should therefore, be a priority. In addition individual vocational rehabilitation, training and employment support are required. The OECD concludes that what is important is close contact between the job seeker with a disability and the employment advisor and a proactive approach in relation to the employer after a job has been found. The UK is cited as an example of good practice in its Pathways to Work Programme. Here all new disability benefit claimants are obliged to attend an initial work focused interview with a Personal Advisor at their local Jobcentre Plus. There are normally five further interviews with the same advisor at monthly intervals designed to overcome barriers to

7 Disability and Social Exclusion

145

employment and develop an action plan specifying what activities are to be undertaken. For an evaluation see Dorsett (2008). He concludes that the policy was successful in increasing employment in a cost effective manner and may have reduced the extent to which individuals’ health conditions limited their ability to engage in day-to-day activities. However, he counsels that greater priority needs to be given to those suffering from mental illness. In addition there is a need to encourage employees to hire, utilise effectively and retain disabled employees. Hiring and retention of workers with disabilities can be assisted by work subsidies, accommodation schemes and personalised support from public employment services. As for labour utilisation, as noted above, Jones and Sloane (2009) found the disabled (particularly those who were work-limited) were significantly more likely to be skill mismatched than the nondisabled. This suggests that employers should be encouraged both to increase the discretion that disabled employees are given in the way they perform their job and more formally to assess the capabilities of such workers.

7.6 Conclusion In this paper we have reviewed, from a European perspective, the facts relating to the incidence of disability, the considerable measurement problems in analysing the phenomenon, and the various policies adopted in Europe to assist the disabled. The extent to which individuals self-report disability varies considerably between EU countries. In part, this may reflect differences in health and functional limitations amongst the population but it is also a result of cultural norms and the incentives generated by institutional and policy frameworks. Within the UK, a country with a high prevalence of self-reported disability, we summarise evidence of substantial labour market disadvantage in terms of participation, relative wages, hours of work and skill mismatch for those whose disability is classified as worklimiting. If the aim of policymakers is to enhance social inclusion through labour market participation, then the important question is: what factors cause a disability to be work-limiting? This, however, is a difficult question since it will reflect an individual’s own experience, aspirations and perceptions, the social and policy environment, as well as the fundamental interaction between functional limitations and job demands. We have reviewed five different types of policy which have been implemented to assist the disabled. In the context of social inclusion, income support should not be offered to those who are able to participate in the labour market, though where to draw the line is likely to be problematical given the measurement issues and the changing nature of the disability itself. Quotas seem to be potentially inefficient and sit uneasily with equal opportunities policies. Sheltered employment is provided on a small scale in most countries and is unlikely to provide the same degree of social inclusion as normal employment. Further, there is little evidence that more recently introduced anti-discrimination legislation has provided major gains

146

P. J. Sloane and M. K. Jones

for the disabled. This leaves us with individually targeted employment policies as the most promising policies to help disabled individual retain and gain employment for which they are most suited. A critical starting point, therefore, is identifying non-disabled individuals from disabled individuals who are capable of work and those who are incapable of work since the nature of support needs to differ considerably. However, fundamental in ensuring the effectiveness of such policies is an awareness that disability reporting and the reporting of a disability as work-limiting will, in part, depend on the incentives generated by such a policy and its relationship to other related support. In trying to resolve such a difficult issue, a number of considerations are important. First, whatever the nature of the policy, it should be made clear who is eligible and stringent assessments maintained to ensure the policy covers its target audience. For the case of identifying work-ability this will require a comprehensive individual assessment of capabilities and of potential job demands. Secondly, policies need to be carefully structured and targeted to minimise the influence they have on reporting the existence or extent of disability by providing support that is particularly beneficial for the target group. While few would argue against out of work financial support for those with the most limiting of conditions, appropriate incentives for employment should be generated for those who are able to work. Key in this respect is the recognition that disabled individuals whose disability limits their work may need support in work in order to maintain the incentive to work for those who are able to do so. Recent changes to disability benefits in the UK have moved in this direction. For example, Incapacity Benefit, the main form of income support for those out of work was replaced in 2008, for new entrants, by the Employment and Support Allowance. A key element in this change is the focus on the ability to work. This includes a medical ‘work capability’ assessment which evaluates the extent to which the disability limits work activity. For those who are deemed as capable of work, work-focused interviews and support are provided to help individuals regain employment. As part of the assessment process considerable attention needs to be focused on raising the educational qualifications of the disabled as this has a critical role in improving the employability of this group. Recent work has shown that the incidence of mental health problems has become a larger proportion of the total incidence of disability and that this group has greater disadvantage in accessing employment and benefitting from it. Individual employment support may be able to reach these groups that require special consideration.

References Acemoglu D, Angrist JD (2001) Consequence of employment protection? the case of the Americans with disabilities act. J Political Economy 19:915–950 Autor DH, Duggan MG (2003) The rise in the disability rolls and the decline in unemployment. Quart J Econ 118(1):157–205

7 Disability and Social Exclusion

147

Bardesi E, Jenkins S, Rigg J (2000) Disability, work and income: a British perspective. ISER, University of Essex, UK October Beatty C, Fothergill S (2004) The diversion from ‘Unemployment’ to ‘Sickness’ across British regions and districts. Sheffield Hallam University, UK Bell D, Heitmueller A (2009) The disability discrimination act in the UK; helping or hindering employment among the disabled. J Health Econ 28(2):465–480 Benitez-Silva H, Disney R, Jimenez-Martin S (2010) Disability, capacity for work and the business cycle; an international perspective. Economic Policy 25(63):485–536 July Braithwaite J, Mont D (2008) Disability and poverty, Discussion Paper No. 0805, World Bank Bratsberg B, Fevang E, Røed K (2010) Disability in the welfare state: an unemployment problem in disguise? IZA Discussion Paper No. 4897, Bonn, April Burkhauser RV, Daly MC (2002) US disability policy in a changing environment. J Econ Perspect 16(1):213–224 Winter Cai L (2010) The relationship between health and labour force participation: evidence from a panel data simultaneous equation model. Labour Econ 17:77–90 Cousins C, Jenkins J, Laux R (1998) Disability data from the LFS: Comparing 1997–98 with the Past, Labour Market Trends, June pp 321–340 Datta Gupta N, Kristensen N, Pozzoli D (2009) External validation of the use of vignettes in cross-country health studies, IZA Discussion Paper No. 3989 DeLeire T (2001) Changes in wage discrimination against people with disabilities: 1948–93. J Human Resour 36(1):144–158 Dorsett R (2008) Pathways to work for new and repeat incapacity benefits claimants; evaluation synthesis report for the department for work and pensions, Report No. 525, National Institute for Economic and Social Research Dupré D, Karjalainen A (2003) Employment of disabled people in Europe in 2002, Statistics in Focus, Population and Social Conditions, Theme 3, Eurostat Eurostat (2001) Disability and social participation in Europe, Luxembourg Gruber J (2000) Disability insurance benefits and labour supply. J Political Economy 108(6):1162–1183 Hagan R, Jones AM, Rice N (2009) Health and retirement in Europe. Int J Environ Res Public Health 6(10):2676–2695. doi:0.3390yerph6102676 Haile G (2009) Workplace disability diversity and job related well-being in Britain; A WERS 2004 Based Analysis, IZA Discussion Paper 3993 Hernandez-Quevedo C, Jones AM, Rice N (2007) Persistence in health limitations; a European comparative analysis, HEDG Working Paper 07/03 University of York Hotchkiss JL (2004) A closer look at the employment impact of the Americans with disabilities acts. J Human Resour 39:887–911 Jones MK (2007) Does part-time employment provide a way of accommodating a disability? Manchester School 75(6):695–716 Jones MK, Latreille PL Disability and self employment: evidence from the UK LFS, Applied Economics (forthcoming) Jones MK, Sloane PJ (2009) Disability and skill mismatch, IZA Discussion Paper No. 4430, Bonn, September Jones MK, Latreille PL, Sloane PJ (2006) Disability, gender and the labour market. Oxford Economic Papers 80(8):823–843 Kapteyn A, Smith JP, Van Soest A (2007) Vignettes and self reports of work disability in the United States and the Netherlands. Am Econ Rev 97(1):431–473 Koning PWC, van Vuuren DJ (2010) Disability insurance and unemployment insurance as substitute pathways. Appl Econ 42(5):575–588 Lechner M, Vasquez-Alvarez R (2009) The effect of disability on labour market outcomes in Germany. Appl Econ 43(4):1466–4283 July, i First Longhi S, Nicoletti C, Platt L (2010) Interpreting wage gaps of disabled men; the roles of productivity and discrimination, ISER Working Paper No. 2010–19, June

148

P. J. Sloane and M. K. Jones

Mavromaras K, Oguzoglu U, Black D, Wilkins R(2007) Disability and employment in the australian labour market, Final Report to the Australian Government Department of Employment and Workplace Relations, June McVicar D (2008) Why have UK disability benefit rolls grown so much? J Econ Surveys 22(1):114–139 OECD (2003) Transforming disability into ability: policies to promote work and income security for disabled people, Paris OECD (2006) Sickness disability and work: breaking the barriers, vol 1. Norway, Poland and Switzerland, Paris OECD (2007) Sickness disability and work: breaking the barriers, vol 2. Australia, Luxembourg, Spain and the United Kingdom, Paris OECD (2008) Sickness disability and work: breaking the barriers, vol 3. Denmark, Finland, Ireland and the Netherlands, Paris Oguzoglu U (2007) Dynamics of work-limitations and work in Australia, IZA Discussion Paper No. 2867, June Oguzoglu U (2009) Severity of work disability and work, IZA Discussion Paper No. 4328, July Pagan R (2009) Self employment among people with disabilities; evidence from Europe. Disability Soc 24(2):217–229 Pillinger J (2003) Managing diversity in public health and social care in the interests of all citizens, report II, Disability. European Social Network, Dublin Thornton P, Lunt N (1997) Employment policies for disabled people in eighteen countries: a review, Cornell University ILR School Visier L (1998) Sheltered employment for persons with disabilities. Int Labour Rev 137(3):347– 365 Wuellrich J-P (2010) The effects of increasing financial incentives for firms to promote employment of disabled workers. Econ Letters 107:173–176

Chapter 8

Civilian Disability Pensions as an Antipoverty Policy Instrument? A Spatial Analysis of Italian Provinces, 2003–2005 Massimiliano Agovino and Giuliana Parodi

Abstract The purpose of this paper is to investigate whether civilian disability pensions have been used as an antipoverty measure in Italy. We applied a two-step analysis to Italian provincial data for the years 2003–2005. We implemented a classic panel analysis, followed by a two-step GMM (Generalised Method of Moments) analysis in which we introduced the spatial variable. The analysis shows that the number of civilian disability pensions is not related to disabling disease, but it is related to the unemployment rate in some areas and to the rate of poverty everywhere. These results robustly hold when the spatially lagged dependent variable is introduced among the regressors. The spatial variable allows us to take into account the local dimension and the possible social, historical, and cultural links among provinces that go beyond administrative boundaries. In discussing the results, we stress that the figures reflect the number of civilian disability pensions granted, not those requested. Moreover, the national legislation on the attribution of civilian disability pensions is administered locally; therefore, its application may reflect degrees of discretionary interpretation. We propose that there is room to interpret the use of civilian disability pensions as an antipoverty policy instrument in areas characterised by economic difficulties. However, we suggest that civilian disability pensions are particularly unsuited to play the role of an assistance policy instrument; once granted, they are seldom withdrawn despite possible changes in the financial situation of the recipient.

M. Agovino (&) G. Parodi Dipartimento di Metodi Quantitativi e Teoria Economica, Università d’Annunzio di Chieti-Pescara, Chieti, Pescara, Italy e-mail: [email protected] G. Parodi e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_8, Ó Springer-Verlag Berlin Heidelberg 2012

149

150

M. Agovino and G. Parodi

Keywords Disability JEL Codes J14

Pensions Models with panel data Spatial models

I38 C33 C21

8.1 Introduction The purpose of this paper is to investigate whether civilian disability pensions have been used as an antipoverty measure in Italy. This analysis is performed using provincial-level data. We introduce the concept of a local ‘‘market’’ for civilian disability pensions, show the dual structure in the distribution of civilian disability pensions between the north and the south of Italy, and we attempt to test a stylised model that may explain it. We develop our analysis stressing the concept of a local ‘‘market’’ for civilian disability pensions. In Italy, the observed number of disability pensions corresponds to the number of accepted requests for civilian disability pensions. The set of national rules that dictates the attribution of civilian disability pensions is administered locally by institutions such as provincial medical committees and doctors operating at nearby medical centres. Therefore, the criteria chosen locally in applying the national rules might take into account characteristics of local labour markets, thus generating an uneven distribution of the number of recipients of civilian disability pensions. The literature addressing the growing number of individuals of working age receiving disability pensions has not yet investigated the reasons for the uneven spatial distribution of these pensions. For example, there is a high number of disability pensioners in northern England and the southern United States (McVicar 2006). In a pioneering work, Beltrametti (1996) describes the differential growth and distribution of invalidity benefits over time and across Italian regions. He also discusses their role as an inefficient instrument of income redistribution. Several factors explain the spatial concentration of disability pensions, such as the concentration of dangerous economic activities that may create accidents leading to disability benefits; poverty (Smith 1998; Currie and Madrian 1999) and unemployment (Disney and Webb 1991), as they can affect health and therefore disability; the concentration of the ageing population (McVicar 2006); doctors’ discretionary behaviour, as such behaviour may be influenced by local labour market conditions in certifying disability (Ritchie et al. 1993); and generally poor conditions in the labour markets themselves, which lead to an increase in the demand for disability pensions. Regarding Italy, Baldacci and De Santis (2003) compare ISTAT and Bank of Italy data and report that the number of disability pensions paid out to individuals of working age is about twice the number of disabled people of working age. Furthermore, Baldacci and De Santis demonstrate a relationship between unemployment rate and disability pensions, and they comment on the use of disability pensions as an improper instrument of income support. This relationship mirrors

8 Civilian Disability Pensions as an Antipoverty Policy Instrument?

151

earlier results from Castellino (1976), Beltrametti (1996) and Baldacci and Milan (1998). Silva et al. (2010) test a similar hypothesis using European and American microdata. Interestingly, they distinguish between ‘‘work disability’’ and ‘‘health disability’’ and explore possible interactions between the two. Work disability is affected by socio-economic factors (e.g., poverty rates and unemployment rates), while health disability is associated with health problems of the recipient of invalidity benefits. Several factors affect work disability at the geographical level, such as the unemployment rate, the poverty rate, and the distribution of wages in labour markets. The last factor is important because disability pensions act as the reservation wage. In Italy, despite the centralised system of wage determination, there are large interregional wage differentials because some workers receive wages above the contractual minimum. In addition, there is flexibility in describing workers’ occupations and, therefore, in determining their contractual minimum wage. In this work, we concentrate on poverty rates because data regarding salaries and occupations are not available at the provincial level. Silva et al. (2010) investigate possible interactions between health and work disability such as the effects of unemployment-related stress on health and attempts to fraudulently receive invalidity benefits when unemployed. In our work, we follow the methodology developed by Silva et al. (2010) except that we use poverty rates rather than unemployment rates as indicators of work disability.1 No significant relationship emerges between the two variables (health disability and poverty rates), even though indicators of health disability are controlled for in the test of the relationship between the number of recipients of disability pensions and the poverty rate. Regression results often indicate a negative relationship between indicators of health disability and the poverty rate, except for specific types of medical conditions. A further test was conducted for these specific medical conditions, and the results allow us to conclude that health disability is not a primary factor to explain the uneven distribution of disability pensions in Italian provinces. These results continue to hold when the spatially lagged variable is introduced, which accounts for similarities among regions that go beyond administrative boundaries. It is suggested that the results can be explained by the different explicit or implicit rules that govern the attribution of invalidity benefits in the various geographical areas, which is a theme that we develop in this paper. The paper is organised as follows: Sect. 8.2 presents the characteristics of the disability pension system in Italy, the data used in the analysis, and some preliminary results. Section 8.3 develops a spatial analysis to test for spillovers in neighbouring provinces. Section 8.4 presents results of the econometric tests of the hypothesis that disability pensions are used as an antipoverty policy instrument. Section 8.5 presents some further results that check for spatial significance and impact of the poverty rate. Section 8.6 presents concluding remarks.

1 In our case, we use poverty rates and rates of population over 55 as control variables (McVicar and Anyadike-Danes 2007; Disney and Webb 1991; Autor and Duggan 2003; McVicar 2006; Beltrametti 1996).

152

M. Agovino and G. Parodi

8.2 Characteristics of the Disability Benefit System in Italy, Data Used, and Some Preliminary Results In Italy, there are three ways to financially support people with disability that are generally referred to as disability pensions. Invalidity pensions are paid to people with a reduced ability to work, provided that they have reached a certain age and that they have made a certain number of national insurance contributions over the years. Indemnity pensions are paid for accidents at work or for professionally related diseases. Civilian disability pensions are not connected with national insurance contributions; they are paid to disabled people on the basis of their physical characteristics (e.g., people affected by blindness, deafness, or other types of impairments). These pensions are also paid to people with no income or insufficient income after the age of 65 (Ministero del Lavoro e delle Politiche Sociali 2006, 2008). In a more general version of this paper investigating all three types of disability pensions, we find that civilian disability pensions are particularly relevant to the topic under investigation; therefore, we focus on civilian disability pensions in the following analysis. We use data on the 103 Italian provinces, corresponding to the European level NUTS-3 over the period from 2003 to 2005, and we concentrate on individuals of working age (aged 15–64).2 ISTAT provides provincial data on unemployment rates for people aged 15–64, poverty rates, percentage of population over 55, and rates of various illnesses. INPS provides data on the number of people receiving disability benefits Provincial data are used as we explicitly introduce the spatial element in our analysis, as it is well known that administrative data aggregate individuals on the basis of arbitrary geographical boundaries reflecting political and historical situations. The choice of the spatial aggregation unit is therefore essential as different choices may lead to different results in the estimates (Arbia 1988). Regional data cannot be considered ‘‘independently generated’’ (Anselin 1988; Anselin and Bera 1998) because of spatial similarities of neighbouring regions; thus, standard estimation procedures3 can provide biased estimators of the parameters. Aggregating data at the provincial level will allow spatial effects, such as spatial spillovers, to be properly modelled (Arbia et al. 2002; Arbia 2005). In this section, we provide a preliminary investigation of variables that may explain the prevalence4 of civilian disability pensions over the various provinces, by observing the plots of the various variables considered.

2

This choice excludes from our data income support transfers to the poor aged over 65 with insufficient incomes, which are included under the category of civilian disability pensions. 3 That is, estimates that do not take into account spatial dependence. 4 Prevalence measures the proportion of ‘‘events’’ in a given population at a given time. In other words, prevalence measures the proportion of individuals in a given population who suffer from a specific illness. In our case, the concept of disability replaces that of illness.

8 Civilian Disability Pensions as an Antipoverty Policy Instrument?

153

Fig. 8.1 Standard deviation map: civilian disability pensions, and poverty rates (2005)

In Fig. 8.1,5 we plot prevalence rates for civilian disability pensions and poverty rates as a Standard Deviation Map. From visual inspection of the maps, two clusters emerge for the civilian disability pensions and poverty rates; both variables show values above the average in the south and below the average in the north. This first impression confirms the hypothesis that poverty affects both the demand for and the supply of civilian disability pensions; where economic conditions are difficult, individuals seem to apply for more civilian disability pensions, and these pensions also seem to be granted more easily. The maps of Fig. 8.1 also show that high poverty rates are largely concentrated in the south. 5 We only show maps for 2005 as those of previous years do not add any relevant information due to spatial persistence. The number in parenthesis next to each range indicates the number of provinces for which the prevalence rate falls within that range.

154

M. Agovino and G. Parodi

The last variable that we consider in this preliminary investigation is the distribution of the health of the population at the provincial level, which we interpret as an indicator of health disability. If we observe a concentration of health disability in specific provinces, this could be an explanation for a possible concentration of civilian disability pensions. For lack of more precise variables, we consider the rates of discharge6 from hospitals of individuals affected by a variety of illnesses7 as indicators of the health of the population at the provincial level. The scatterplots of the various illnesses, not shown here for brevity, suggest that the distribution of illnesses does not follow a north–south geographic pattern; therefore, the plots do not allow us to conclude that the higher number of civilian disability pensions in the south can be explained in terms of higher rates of health disability in the south. Given the results of this preliminary investigation, we proceeded to apply econometric techniques to identify factors that can explain the concentration of civilian disability pensions in the south.

8.3 Spatial Correlation: Further Analysis Figure 8.1 in Sect. 8.2 has shown two rather marked clusters of the prevalence rate of civilian disability pensions. In this section, we investigate spatial correlation as a stepping stone towards further analysis. In particular, we want to test an hypothesis of ‘‘spatial persistence’’; we examine whether neighbouring provinces tend to display similar rates of civilian disability pensions (Cracolici et al. 2007).8

6

We define as discharge rate the percentage of individuals discharged from a hospital after a period of hospitalisation. The discharge rate refers to the last contact with the institution where the individual has been hospitalised. 7 Previous empirical works on aggregate data have never considered variables related to the state of health as precise as ours. For instance, Nolan and Fitzroy (2003) use the number of visits to the hospital and the mortality rate as health indicators; Autor and Duggan (2003) also use the mortality rate, as do many other authors; and Stapleton et al. (1998) use a measurement of the incidence of AIDS. Note that the mortality rate is criticised in the literature when used as a regressor to explain the incidence of disability (McVicar and Anyadike-Danes 2007). 8 On the contrary, we talk of a uniform spatial structure.

8 Civilian Disability Pensions as an Antipoverty Policy Instrument?

155

Table 8.1 Moran I index calculated over prevalence rates of civilian disability pensions Year Variable Moran I index Standard deviation p-value 2003 2004 2005

Civilian disability pensions Civilian disability pensions Civilian disability pensions

0.8380 0.8227 0.8148

0.0703 0.0674 0.0701

0.0010* 0.0010* 0.0010*

*Significant at 0, 01

For this purpose, we used the Moran I index,9 the most traditional indicators of spatial correlation. Table 8.1 shows the values of the Moran I Index for the prevalence rates for civilian disability pensions over the years from 2003 to 2005. These values show positive spatial autocorrelation, which will have to be accounted for during the econometric estimation in Sect. 8.5. We then applied another instrument of spatial statistics, the Local Indications of Spatial Autocorrelation (LISA), which allowed us to test for local spatial autocorrelation and to identify local clusters where adjacent areas show similar values.10 The LISA cluster map allows us to identify four types of spatial autocorrelation in two categories: spatial clusters (negative values next to negative values, and positive values next to positive values) and spatial outliers (negative values next to positive values, and positive values next to negative values). In other words, this analysis shows the areas with a local Moran statistic significantly different from zero classified by the kind of spatial correlation. The four types of spatial correlation, High–High, Low–Low, Low–High, and High–Low, correspond to the four quadrants of the Moran I Index. The areas characterised by spatial correlation High–High and Low–Low suggest spatial clusters (given the similarity of values) while the areas classified as High–Low and Low–High signal spatial outliers (given the dissimilarity of values). We only show results for the year 2005; previous years add nothing to the discussion because of the spatial persistence of the data. 9

The Moran I Index is similar to the correlation coefficient: it varies between zero and one, -1 and +1. When I equals zero, there is no spatial autocorrelation; when I is close to -1 or to +1 there is high negative or positive spatial autocorrelation, respectively. The I index has the following main characteristics when compared with the coefficient of correlation:

1. It takes one, and not two, variable into account. 2. It incorporates the weights (wij) that index the relative areas. In our case, these are organised in a matrix of inverse distances expressed in km. This choice of weights allows us to formalise the hypothesis that the relationships among individual areas tend to decrease in strength as distances between these areas increase. 3. It is appropriate to think of it as ‘‘the correlation between neighbouring values on a variable’’ (O’Sullivan and Unwin 2003). 10 The statistics to measure the degree of spatial autocorrelation at local level (LISA) allow us to identify the contribution of each province to the overall autocorrelation (Moran I Index); therefore, we can investigate the variation in spatial autocorrelation across the whole area. By focusing on each individual province, these techniques can be used to identify spatial clusters.

156

M. Agovino and G. Parodi

Fig. 8.2 LISA cluster map (2005)

In Fig. 8.2, two clusters emerge for civilian disability pensions. Dark areas (High–High areas) mark provinces with the highest concentration of recipients of civilian disability pensions; they identify provinces in the south of Italy. Grey areas mark provinces with the lowest concentration of recipients of civilian disability pensions; they identify provinces in the north of Italy. This result seems to suggest some autonomy for local agents in applying the national rule when granting civilian disability pensions.

8.4 Specification of the Model Following the methodology in Silva et al. (2010), we want to investigate whether work disability, indicated by the relationship between civilian disability pensions and poverty rates, is contaminated by health disability.11 For this purpose, we developed our analysis in steps: in a first step, we regressed discharge rates of specific illnesses on dummies for provinces (fixed effects) and the provincial poverty rate, where discharge rate of specific illnesses is interpreted as a proxy for health disability. If we find a positive relationship, it is possible that the relationship between work disability and poverty is partially explained by a relationship between health disability and the poverty rate. This implicit hypothesis has to be tested to show that civilian disability pensions have been used as an antipoverty policy instrument both across time and space and that civilian disability pensions are not explained by variations in health disability. In a second 11

If work disability is linked to socio-economic conditions (e.g., poverty rate) and if health disability is only partially linked to work disability, then health disability is, as a consequence, related to socio-economic conditions (the poverty rate) with a circular link.

8 Civilian Disability Pensions as an Antipoverty Policy Instrument?

157

step, we regressed the prevalence rates of civilian disability pensions on the poverty rate controlling for the unemployment rate and the rate of ageing population. In the third step, we tested the robustness of the relationship between civilian disability pensions and the poverty rate by repeating the regression mentioned in the second step, but adding the discharge rates of the various illnesses as regressors. Only if the coefficient on the poverty rate is still significant and positive can we draw the conclusion that civilian disability pensions are used as an antipoverty policy instrument. As we use panel data, we had to decide whether to use fixed or casual effects in the first step; the estimates could be very different in our case, as the time considered is short and the number of considered units is large. It is advisable to choose the fixed effect model when the sampled individuals (in our case Italian provinces) cannot be assumed to be casually extracted from the population. The Hausman test allowed us to choose between the two models. If the test rejects the null hypothesis, the fixed effect model is preferred (Baltagi 2005; Verbeek 2009). From Table 8.2, the results from the Hausman test show that it is correct to use the fixed effect method. They also show that the relationship between discharge rates for various illnesses and the poverty rate is in most cases significantly negative, or not significant; therefore, there is no evidence that health disability is positively influenced by the poverty rate. However, a positive and significant relationship between health disability and poverty is found for three groups of medical conditions: glands, circulatory system, and congenital malformation. Table 8.3 shows the results of the test associated with the second step of our investigation: the estimation with fixed effects. Table 8.3 shows that the coefficient on the poverty rate is significantly positive at the national level in columns I, II, III and IV. At the territorial level of disaggregation, the results show that the coefficient of the rate of poverty is positive and significant only for the south of Italy whether or not we include rates of discharge for the various medical conditions. These results clearly indicate that civilian disability pensions are not homogeneously distributed across the country; moreover, they indicate that socio-economic variables, such as the poverty rate, are significantly correlated with the attribution of civilian disability pensions in the south. This result suggests that the national rule regarding the attribution of civilian disability pensions has a margin for interpretation by local institutions. These findings allow us to conclude that from 2003 to 2005 in the south of Italy, civilian disability pensions were used as an instrument of income support for the poor.

8.5 Geographical Distribution of Civilian Disability Pensions: Further Analysis We now move to the third step of our investigation. According to the theory, units spatially close to each other show a higher probability of similar behaviour when compared with units that are spatially distant (Huckfeldt 1986; Vasquez 1995;

158

M. Agovino and G. Parodi

Table 8.2 Estimates with fixed effects (FE) and random effects (RE): relationship between the discharge rates of various illnesses and the poverty rate Dependent variable: % of discharges for the Constant Coefficient on the Hausman various illnesses poverty rate test Liver Diabetes Genital urinary apparatus Breathing apparatus Glands Blood diseases Muscular system Skin diseases Circulatory system Morbid conditions Tumours Congenital malformations Nervous system

19.64611* (68.25) 22.40132* (58.96) 153.5469* (143.96) 137.0665* (150.79) 63.59959* (112.14) 24.63499* (75.21) 130.6292* (146.61) 34.02201* (105.84) 292.8374* (59.85) 83.19982 (93.00) 195.4509* (61.34) 20.4439* (104.59) 172.8303* (70.84)

-1.71601 (-0.72) 1.908563 (0.61) 0.7475317 (0.08) -20.50656** (-2.72) 17.6263* (3.75) 3.394031 (1.25) 12.32411 (1.67) -6.698444** (-2.51) 43.47461* (3.27) -1.001942 (-0.13) 10.42277 (1.12) 5.123178* (3.16) 13.30279 (0.66)

42.73* 25.47* 31.97* 30.40* 32.78* 17.33* 13.87* 59.60* 1.30 59.56* 0.20 21.90* 7.93*

*, **, and ***: 1, 5, and 10%; (): t-statistics

Cardoso and Faletto 1979). The Moran I Index (Table 8.1 above) has indicated a rather evident positive spatial correlation for civilian disability pensions. So far our estimates have considered individual economic units to be entirely self-contained; this fundamental hypothesis is rather doubtful, as provinces are units defined by administrative rules, which ignore historical and economic similarities. Therefore, it is likely that spillovers go beyond administrative boundaries, with decreasing effects as distances increase. We introduced spatial effects into the relationship between the prevalence rate of civilian disability pensions and the poverty rate in order to take into account spatial effects of different provinces.12 These effects are represented by a spatially 12

By estimating a model ignoring spatial clustering, an inefficient estimate of the parameters is obtained. The standard errors are underestimated, and a type I error is made. Luckily, local or global measures of spatial autocorrelation are estimated to decide whether the data show spatial dependence, and here, the Moran I index indicates a clear process of positive spatial autocorrelation to be considered in further estimates.

Genital urinary apparatus

Diabetes

Liver

Age [55

Poverty rate Centre

Poverty rate North

Poverty rate South

Poverty rate

Unemployment rate Centre

Unemployment rate North

Unemployment rate South

Unemployment rate

Constant

0.2620* (6.08)

0.5130* (4.32)

-3.806* (-4.75) -0.0118* (-3.13)

Table 8.3 Fixed effects estimates (FE) Dependent variable: FE (I) prevalence rate of civilian invalidity pensions

0.1913* (4.59) -0.0031 (-0.92) -0.0095* (-2.91) -0.00026 (-0.24)

0.2611** (2.26)

-2.962* (-3.89) -0.0108* (-3.14)

FE (II)

0.2600* (6.30)

-0.0176* (-4.60) 0.0348* (2.96) 0.0136 (0.86) 0.4772* (4.19)

-3.852* (-5.01)

FE (III)

0.1980* (4.74) -0.0034 (-1.01) -0.0079** (-2.39) -0.00069 (-0.62)

-0.0138* (-3.77) 0.0145 (1.31) 0.0146 (0.99) 0.3355* (2.97)

-3.168* (-4.15)

FE (IV)

0.5035* (4.00) 0.5155 (1.10) 0.7363 (1.24) 0.2601* (5.97)

-3.773* (-4.66) -0.0118* (-3.10)

FE (V)

0.2968** (2.45) -0.1182 (-0.28) 0.0579 (0.11) 0.1951* (4.65) -0.0030 (-0.89) -0.0099* (-3.00) -0.00006 (-0.06)

-3.044* (-3.96) -0.0105* (-3.03)

FE (VI)

0.4950* (4.14) -.1046 (-0.22) 0.9572 (1.65) 0.2577* (6.22)

-0.0176* (-4.61) 0.0396* (3.21) 0.0164 (1.01)

-3.816* (-4.95)

FE (VII)

(continued)

0.3037** (2.58) -0.4491 (-1.04) 0.3170 (0.61) 0.1913* (4.68) -0.0020 (-0.60) -0.0085** (-2.63) -0.0001 (-0.13)

-0.0150* (-4.15) 0.0218*** (1.94) 0.0130 (0.89)

-3.070* (-4.10)

FE (VIII)

8 Civilian Disability Pensions as an Antipoverty Policy Instrument? 159

0.3779 21.20*

FE (I)

*, **, and ***: 1, 5, 10%; (): t-statistics

R2 overall Hausman Test

Nervous system

Congenital malformations

Tumours

Morbid conditions

Circulation system

Skin diseases

Muscular system

Blood diseases

Glands

Table 8.3 (continued) Dependent variable: prevalence rate of civilian invalidity pensions Breathing apparatus FE (III)

0.4475 155.47*

FE (II)

-0.0017 (-1.62) 0.0055** (2.33) 0.00041 (0.15) 0.0046* (4.34) -0.0096* (-3.01) -0.00099 (-1.46) -0.0007 (-0.66) 0.0039* (3.66) 0.0081*** (1.78) -0.0011* (-2.86) 0.4857 148.08*

-0.00166 (-1.49) 0.0047*** (1.99) 0.0008 (0.30) 0.0042* (3.85) -0.0082** (-2.77) -0.0010 (0.114) -0.0012 (-1.14) 0.0031* (3.08) 0.0077 (1.68) -0.0009** (-2.33) 0.5020 120.69*

FE (IV)

0.3790 145.55*

FE (V)

-0.0017 (-1.58) 0.0057** (2.40) 0.0003 (0.11) 0.0048* (4.45) -0.0098* (-3.05) -0.0010 (-1.57) -0.0007 (-0.62) 0.0037* (3.52) 0.0077 (1.69) -0.0011** (-2.75) 0.4758 152.03*

FE (VI)

0.3986 153.98*

FE (VII)

-0.0012 (-1.18) 0.0050** (2.16) 0.0014 (0.51) 0.0041* (3.87) -0.0106* (-3.38) -0.0011 (-1.66) -0.0008 (-0.74) 0.0041* (3.90) 0.0070 (1.57) -0.0009* (-2.23) 0.4931 158.83*

FE (VIII)

160 M. Agovino and G. Parodi

8 Civilian Disability Pensions as an Antipoverty Policy Instrument?

161

lagged vector of the prevalence rate of civilian disability pensions in the provinces. The purpose is to test whether the prevalence rate of civilian disability pensions in a given province depends on the weighted average of the prevalence rate of civilian disability pensions in other provinces. Including the spatially lagged variable among the regressors makes estimators based on OLS biased and inconsistent, regardless of the properties of the error term (Anselin 1988). Hence, we employed a ‘‘Two-Step feasible GMM Estimation’’ (Baum et al. 2002). This method provided a suitable set of tests to determine the validity of the variables used to instrument for the spatially lagged dependent variable.13 The results of this test are shown in Table 8.4. In Table 8.4, the estimates with GMM disaggregated to the macro-areas show that the poverty rate is significant only for the south of Italy (columns V and VII); however, the coefficient on the poverty rate loses significance when the regressors for health disability are introduced in the estimates (columns VI and VIII). This result seems to suggest that civilian disability pensions have not been used as an antipoverty policy instrument. However, column IX refines the analysis by excluding the discharge rates for gland and circulatory system problems and congenital malformation, which are the three variables that appeared to be significant in Table 8.2. We estimated the fixed effect model of the relationship between poverty and health disability without these three variables. When these three regressors are excluded, the poverty rate becomes significant once again. This result suggests that we cannot discount the hypothesis that civilian disability pensions are used as a poverty relief policy instrument. In fact, these results suggest that only for the three medical conditions excluded from column IX, civilian disability pensions take health disability into account, and not the socioeconomic conditions of the area. Furthermore, for all the other types of medical conditions considered, civilian disability pensions do not deal with health disability; they are instead used as antipoverty policy instruments. This conclusion is confirmed by the fact that the various pathologies except those excluded from column IX are either scarcely significant or not significant in explaining the prevalence rate of civilian disability pensions. In addition, the significance and importance of the spatially lagged variable shows spatial dependence among local markets at provincial level, indicating that an increase in the prevalence rate of civilian disability pensions in a given province is associated with an increase in the same variable in neighbouring provinces; this was also shown by the Moran I Index above. Additionally, the LISA cluster map above shows that the behaviour of economic units is not spatially stable, but

13

The literature suggests using spatial lags of the regressors as instruments (Anselin 1988).

Blood diseases

Glands

Genital urinary apparatus Breathing apparatus

Diabetes

Liver

Age [55

Poverty rate Centre

Poverty rate North

Poverty rate South

0.103* (4.55)

0.202* (2.18)

-0.009* (-4.21)

Unemployment rate

Unemployment rate South Unemployment rate North Unemployment rate Centre Poverty rate

I

Dependent variable: prevalence rate of civilian invalidity pensions

Table 8.4 Two-step feasible GMM estimation

0.073* (3.03) 0.0023 (0.90) -0.0001 (-0.06) -0.0013** (-2.11) -1.70e-06 (-0.00) -0.0003 (-0.22) 0.0003 (0.21)

0.149 (1.34)

-0.008* (-3.78)

II

0.1075* (4.59)

-0.0075* (-3.09) -0.0177** (-2.05) -0.0033 (-0.53) 0.1893** (2.02)

III

0.0719* (2.97) 0.0027 (1.09) -.0007 (-0.34) -0.0014** (-2.31) -0.0003 (-0.56) 0.0001 (0.08) -0.0001 (-0.07)

-0.0076* (-3.47) -0.0107 (-1.18) -0.0011 (-0.21) 0.1308 (1.16)

IV

0.224** (2.22) -0.232 (-1.14) 0.219 (0.71) 0.111* (4.83)

-0.008* (-3.71)

V

0.170 (1.37) -0.249 (-1.09) -0.013 (-0.05) 0.075* (3.21) 0.0029 (1.18) -0.0010 (-0.52) -0.001*** (-1.82) -0.0003 (-0.50) 0.0003 (0.20) -0.0003 (-0.22)

-0.0072* (-3.42)

VI

0.273* (2.39) -0.228 (-1.20) 0.181 (0.55) 0.1006* (4.25)

-0.0071* (-2.92) -0.012 (-1.39) -0.005 (-0.77)

VII

0.1706 (1.37) -0.266 (-1.27) 0.0160 (0.05) 0.075* (3.22) 0.0029 (1.17) -0.0009 (-0.45) -0.0012*** (-1.82) -0.0002 (-0.40) 0.0002 (0.13) -0.0001 (-0.10)

-0.0075* (-3.46) -0.008 (-0.91) -0.0019 (-0.30)

VIII

(continued)

-0.00002 (-0.02)

0.2057** (1.91) -0.2889 (-1.37) 0.0716 (0.24) 0.0750* (3.32) 0.0032 (1.23) -0.00085 (-0.56) -0.0011*** (-1.80) -0.00024 (-0.38)

-0.0074* (-3.27) -0.0078 (-0.89) -0.0021 (-0.34)

IX

162 M. Agovino and G. Parodi

Anderson-Rubin wald (Chi-sq)

Anderson-Rubin wald (F)

Kleibergen-Paap rk wald statistic

Uncentered R2 Shea’s partial R2 test of instrument strength of spatial lag F-test on the joint nullity of the instruments Kleibergen-Paap rk LM statistic

Spatial lag

Nervous system

Congenital malformations

Tumours

Morbid conditions

Circulatory system

Skin diseases

Dependent variable: prevalence rate of civilian invalidity pensions Muscular system

Table 8.4 (continued)

0.878* (19.17) 0.8085 0.9887

-0.0009 (-1.36) -0.005** (-2.21) 0.0001 (0.40) 0.0011 (1.21) 0.0013** (2.27) 0.0029 (1.07) 0.0005** (2.40) 0.944* (16.64) 0.8333 0.9934 5079.97 [0.0000] 82.75 [0.0000] 22504.82 [0.0000] 61.29 [0.0000] 271.53 [0.0000]

0.858* (22.06) 0.8072 0.9963

4484.52 [0.0000] 101.40 [0.0000] 47865.90 [0.0000] 49.37 [0.0000] 526.95 [0.0000]

1825.38 [0.0000] 84.59 [0.0000] 17626.30 [0.0000] 41.59 [0.0000] 401.57 [0.0000]

III

II

I

2029.24 [0.0000] 76.53 [0.0000] 11421.41 [0.0000] 41.16 [0.0000] 231.69 [0.0000]

-0.0007 (-1.08) -0.0057** (-2.28) 0.0002 (0.46) 0.0009 (1.01) 0.0015** (2.48) 0.0026 (0.98) 0.0005* (2.56) 0.926* (15.53) 0.8338 0.9822

IV

2185.33 [0.0000] 97.35 [0.0000] 21102.05 [0.0000] 54.81 [0.0000] 529.30 [0.0000]

0.862* (22.65) 0.8105 0.9904

V

2298.72 [0.0000] 82.34 [0.0000] 12938.18 [0.0000] 46.90 [0.0000] 263.98 [0.0000]

-0.0005 (-0.82) -0.0058** (-2.31) 0.0001 (0.33) 0.0010 (1.12) 0.0014** (2.37) 0.0018 (0.69) 0.0006** (2.55) 0.929* (16.58) 0.8359 0.9841

VI

2876.36 [0.0000] 81.20 [0.0000] 15271.40 [0.0000] 65.00 [0.0000] 345.10 [0.0000]

0.883* (19.20) 0.8112 0.9882

VII

2062.83 [0.0000] 75.63 [0.0000] 11738.73 [0.0000] 40.86 [0.0000] 232.51 [0.0000]

-0.0005 (-0.83) -0.006** (-2.40) 0.0001 (0.26) 0.0009 (1.05) 0.0014** (2.40) 0.002 (0.77) 0.0006* (2.73) 0.926* (15.56) 0.8362 0.9825

VIII

(continued)

2016.60 [0.0000] 78.79 [0.0000] 16546.60 [0.0000] 46.71 [0.0000] 261.45 [0.0000]

0.00061** (2.63) 0.924* (16.73) 0.8356 0.9880

0010 (1.15) 0.0015* (2.69)

-0.00054 (-0.81) -0.0056** (-2.30)

IX

8 Civilian Disability Pensions as an Antipoverty Policy Instrument? 163

II

71.77 [0.0000] 5079.973 24.58 13.96 10.26 8.31 5.835 [0.1199] 2.380 [0.1229] 3.455 [0.1777]

I

82.11 [0.0000] 4484.524 38.54

20.88

14.78

11.65

13.733 [0.1321] 7.924 [0.1604] 5.808 [0.2139]

(): t-statistics; []: p-value; *, **, and ***: 1, 5, and 10%

Kleibergan-Paap rk wald F-statistic Stock-Yogo weak ID test critical values: 10% maximal IV size Stock-Yogo weak ID test critical values: 15% maximal IV size Stock-Yogo weak ID test critical values: 20% maximal IV size Stock-Yogo weak ID test critical values: 25% maximal IV size Hansen J statistic (overidentification test of all instruments) Hansen J statistic (eqn. excluding suspect orthog. conditions) C statistic (exogeneity/orthogonality of suspect instruments)

Dependent variable: prevalence rate of civilian invalidity pensions Stock-Wright LM S statistic (Chi-sq)

Table 8.4 (continued)

12.625 [0.1254] 6.538 [0.1624] 6.087 [0.1927]

11.07

14.01

19.71

71.65 [0.0000] 1825.377 36.19

III

5.441 [0.2450] 2.026 [0.1547] 3.415 [0.3319]

8.84

10.98

15.09

60.68 [0.0000] 2029.242 26.87

IV

13.130 [0.1075] 5.981 [0.2006] 7.148 [0.1282]

11.07

14.01

19.71

83.11 [0.0000] 2185.326 36.19

V

5.526 [0.2375] 1.863 [0.1722] 3.663 [0.3003]

8.84

10.98

15.09

71.25 [0.0000] 2298.725 26.87

VI

4.549 [0.3367] 1.254 [0.2627] 3.295 [0.3483]

8.84

10.98

15.09

68.17 [0.0000] 2876.361 26.87

VII

5.687 [0.2238] 2.086 [0.1486] 3.601 [0.3079]

8.84

10.98

15.09

60.14 [0.0000] 2062.825 26.87

VIII

5.726 [0.2205] 1.438 [0.2304] 4.288 [0.2320]

8.84

10.98

15.09

63.00 [0.0000] 2016.602 26.87

IX

164 M. Agovino and G. Parodi

8 Civilian Disability Pensions as an Antipoverty Policy Instrument?

165

generates specific spatial regimes: a cluster of provinces for the north and one for the south of Italy. Positive spatial autocorrelation and spatial heterogeneity support our claim that ‘‘spatial persistence’’ exists.14 Despite the high value of the R2, we can exclude multicollinearity between the poverty rate and the various discharge rates because of the significance of the regressors.15

8.6 Conclusions The main purpose of this work has been to investigate the hypothesis that civilian disability pensions are used as an antipoverty instrument in Italy, using provinciallevel data. The results confirm that this hypothesis is true in the south of Italy, as we find a positive and significant relationship between the prevalence rate of civilian disability pensions and poverty rate, even when controlling for indicators of health disability. These results robustly hold when the spatially lagged dependent variable is introduced among the regressors. The positive spatial autocorrelation and spatial heterogeneity indicate ‘‘spatial persistence’’ and spatial relevance in the diffusion process of civilian disability pensions. In addition, the spatial variable allows us to take into account the local dimension, and the possible social, historical, and cultural links among provinces that go beyond administrative boundaries. The rules governing the attribution of civilian disability pensions are clearly defined and uniform at the national level. However, the application of these rules is

14

Notes on the tests in Table 8.4: the Shea partial R2 appears to be rather high, while the F-test of joint nullity of the instruments rejects the null hypothesis; this allows us to conclude that the instruments are valid and relevant. The rank tests, Kleibergen-Paap rk LM statistic and Kleibergen-Paap rk Wald statistic, reject the null hypothesis and allow us to conclude that the model is identified. The Anderson-Rubin Wald (F), Anderson-Rubin Wald (Chi-sq) and StockWright LM S statistic (Chi-sq) tests reject the null hypothesis and reassure us of the robustness of the instruments that we used. In addition, we observe that the Kleibergan-Paap rk Wald Fstatistic is much bigger than the value of the Stock-Yogo, shown four rows later; thus, we conclude that the instruments are strong. As a general conclusion, the estimated model is identified and contains valid, relevant and strong instruments. Finally, to test whether the instruments are uncorrelated with the residuals, we calculate the Hansen (J-test) of overidentification and orthogonality, as well as the C statistic of exogeneity of the instruments. As all of the tests do not reject the null hypothesis, the instruments can be considered exogenous (Baum et al. 2002). 15 The calculation of the Variance Inflation Factor (VIF) allows us to control for multicollinearity and to exclude it, as for each variable, VIF \10; the highest VIF, found for the variable ‘‘genital urinary apparatus’’, is 5.21, which is well below the threshold. The average VIF is 2.99. The tolerance associated with each variable, 1/VIF, is [0.1, which allows us to exclude multicollinearity (a tolerance value \0.1 is comparable with a VIF equal to 10). Finally, the conditional number, which is generally used to assess the global instability of the regression coefficients, suggests stability, as it is 7.7270 where 10 is the threshold value.

166

M. Agovino and G. Parodi

deferred to local medical and administrative committees, which have discretion in interpreting these general rules. These institutions may take into account local socio-economic conditions; this fact can explain the distribution of civilian disability pensions across provinces. Finally, civilian disability pensions are not a flexible instrument; once granted, they are very rarely withdrawn, even though such withdrawal is technically possible. Therefore, it would be improper to use civilian disability pensions as an income support instrument, as they often are invariable with respect to possible changes in the economic conditions of the recipients. More suitable instruments ought to be used as income support measures.

References Anselin L (1988) Spatial econometrics: methods and models. Kluwer Academic Publishers, Boston Anselin L, Bera A (1998) Spatial dependence in linear regression models with an introduction to spatial econometrics. In: Ullah A, Giles DEA (eds) Handbook of applied economic statistics. Marcel Dekker, New York Arbia G (1988) Spatial data configuration in statistical analysis of regional economics and related problems, advanced statistical theory and applied econometrics. Kluwer Academic Publisher, Boston Arbia G (2005) Spatial econometrics, statistical foundations and applications to regional convergence. Springer Verlag, Ney York Arbia G, Basile R, Mirella S (2002) Regional convergence in Italy 1951–1999: a spatial econometric perspective. ISAE Working Papers 29, ISAE–Institute for Studies and Economic Analyses, Rome, Italy Autor DH, Duggan MG (2003) The rise in the disability rolls and the decline in unemployment. Q J Econ 118:157–206 Baldacci E, De Santis G (2003) Disability pensions in Italy: the lawand the numbers. In: Prinz C (ed) European disability pension policies. European Centre, Vienna, pp 225–252 Baldacci E, Milan G (1998) Gli effetti di redistribuzione territoriale della spesa pensionistica di invalidità. In: Rossi N (ed) Il lavoro e la sovranità sociale, 1996–1997. Il Mulino, Bologna Baltagi BH (2005) Econometric analysis of panel data. Wiley, Ney York Baum CF, Schaffer ME, Stillman S (2002) Instrumental variables and GMM: estimation and testing. Boston College Economics Working Paper 545 Beltrametti L (1996) Le pensioni di invalidità: effetti redistributivi tra le regioni italiane (1951– 1993) in Politica Economica n. 3, pp 391–403 Cardoso FH, Faletto E (1979) Dependency and development in Latin America. University of California Press, Berkeley, CA Castellino O (1976) Il labirinto delle pensioni. Il Mulino, Bologna Cracolici MF, Cuffaro M, Nijkamp P (2007) Geographical distribution of unemployment: an analysis of provincial differences in Italy. Growth Change 38(4):649–670 Gatton College of Business and Economics, University of Kentucky Currie J, Madrian BC (1999) Health, health insurance and the labor market. In: Ashenfelter RO, Card D (eds) Handbook of labor economics, vol 3C. Elsevier, Amsterdam, pp 3309–3416 Disney R, Webb S (1991) Why are there so many long-term sick in Britain? Econ J 101:252–262 Huckfeldt R (1986) Politics in context: assimilation and conflict in urban neighborhoods. Agathon Press, New York

8 Civilian Disability Pensions as an Antipoverty Policy Instrument?

167

McVicar D (2006) Why do disability benefit rolls vary between regions? A review of the evidence from the USA and the UK, regional studies. Taylor Francis J 40(5):519–533 July McVicar D, Anyadike-Danes M (2007) Panel estimates of the determinants of British regional male incapacity benefits rolls 1998–2006. Applied Economics (fort) Ministero del Lavoro e delle Politiche Sociali (2006) Rapporto di Monitoraggio sulle politiche sociali, I trasferimenti monetari per invalidità, pensioni sociali, integrazioni al minimo, pensioni di guerra e ai superstiti, Direzione generale per la gestione del Fondo Nazionale per le politiche sociali e monitoraggio della spesa sociale Ministero del Lavoro e delle Politiche Sociali (2008) Rapporto di Monitoraggio sulle politiche sociali, I trasferimenti monetari per invalidità, pensioni sociali, integrazioni al minimo, pensioni di guerra e ai superstiti, Direzione generale per la gestione del Fondo Nazionale per le politiche sociali e monitoraggio della spesa sociale Nolan M, Fitzroy F (2003) Inactivity, sickness and unemployment in Great Britain: early analysis at the level of local authorities. Mimeo, University of Hull O’Sullivan D, Unwin DJ (2003) Geographic information analysis. Wiley, Hoboken, NJ Ritchie J, Ward K, Duldig W (1993) GPs and IVB. Research Report No. 18, Department of Social Security/HMSO, London Silva HB, Disney R, Martin SJ (2010) Disability, capacity for work and the business cycle: an international perspective. Econ Policy 25:483–536 Smith J (1998) Socioeconomic status and health. Am Econ Rev 88:192–196 Stapleton D, Coleman K, Dietrich K, Livermore G (1998) Empirical analysis of ID and SSI application and award growth. In: Rupp K, Stapleton D (eds) Growth in civilian disability pensions. W.E. Upjohn Institute for Employment Research, Kalamazoo, MI, pp 31–92 Vasquez JA (1995) Why do neighbours fight? Proximity, interaction, or territoriality. J Peace Res 32(3):277–293 Verbeek M (2009) Econometria. Zanichelli, Bologna

Part III

Economic Crisis, Labour Markets and Social Exclusion

Chapter 9

Measuring the Long Wave. Unemployment, Discouragement and Semi-Employment in Italy, During and After the Crisis Leonello Tronti and Riccardo Gatto

Abstract The official data indicate a much less severe impact of the international crisis on employment and on the incomes of wage-earners and pensioners in Italy than might have been predicted. But the standard measure of unemployment agreed at international level has some shortcomings. For one thing, the ‘‘objective’’ definition of unemployment is a poor fit with the Italian labour market. And for another, the considerable increase in labour hoarding and the emergence of ‘‘semi-employment’’—the alternation of brief periods of work with periods of unemployment or economic inactivity—make the state of the market hard to assess. The paper offers a more in-depth examination of the impact of the crisis, using labour force survey data to count labour underutilization, funded and unfunded labour hoarding, the discouraged, and semi-employed workers. The picture that emerges is one of a ‘‘long wave’’ of available labour underutilization: the Italian labour market has to cope with an increase in joblessness broadly conceived of about 1,400,000—more than twice the rise in official unemployment. This wave will inevitably last for years to come, until the pools of implicit and explicit unemployment are reabsorbed and jobs are created for those now unoccupied and the new entrants. So now is the time for reform of unemployment benefits, not just to provide income support for those caught up in the long wave of joblessness, but also to permit quicker reorganization of production. The reform will have to highlight the diverse interests and roles of government and of the social partners, within the framework of a new development and industrial policy.

L. Tronti (&) Università Roma Tre, Rome, Italy e-mail: [email protected] R. Gatto ISTAT, Rome, Italy e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_9, Springer-Verlag Berlin Heidelberg 2012

171

172

L. Tronti and R. Gatto

Keywords Labour underutilization measurement Unemployment Labour hoarding Discouragement Semi-employment Unemployment benefit JEL Codes E24

J21 J64 J68

9.1 Crisis and the Labour Market Fortunately, the global crisis has had less severe effects on the public finances in Italy than in other European countries. In 2008 and 2009 the budget deficit averaged 4.0% of GDP, against 5.7% in France, 7.8% in Spain, 9.6% in the United Kingdom and 10.2% in Ireland. But the impact on economic performance has been greater. Over the two years, Italian GDP fell by 3% a year, nearly twice the 1.7% decline in the euro area. Also, prices continued to rise faster in Italy than in the rest of Europe (2.2% annually against 1.8% for the euro area). Considering the sharp fall in output, the impact of the recession on the labour market was, however, relatively modest. Total employment continued to grow through the first quarter of 2008, even though GDP had already begun to contract (Fig. 9.1). Employment growth peaked in that quarter at 1.9% over the same quarter of 2007. Job growth then slowed, giving way to a contraction in the fourth quarter. Year on year, therefore, the number of persons employed actually still grew appreciably in 2008 (by 0.8%), against a 1% fall in GDP. Then, in 2009, employment fell sharply, with 339,000 fewer persons in work; but again the relative job loss (1.6%) was much less than the huge fall in output (5.1%). Firms made ample use of income support programmes, and the number of hours of shorttime working compensation soared. Per capita labour productivity had begun to slip already in the second quarter of 2007, before the financial crisis was perceived. From then on it fell for eight consecutive quarters, with an overall loss of 4.8%. A significant portion of this deterioration was due to labour hoarding by employers, who elected to maintain staffing levels by cutting overtime, shortening hours and resorting to short-time working compensation rather than instantly adjust manpower to the demand shock. And if productivity trend had been the first harbinger of the approaching crisis, productivity also brought the first signs of recovery. Along with it, output also began to slowly resume growth. The contrast between the rate at which productivity rose and the more sluggish return of output to positive territory highlights the stubborn obstacles to employment recovery. Firms have begun a period of reorganization (which will certainly not be brief), but with demand still feeble, the resulting productivity increase had an adverse effect on an employment situation marked by ample labour hoarding. The number of people in work, in fact, has not yet responded to the growth in GDP. Indeed, in the third quarter of 2010, employment declined by a further 176,000 compared to the same quarter of 2009.

9 Measuring the Long Wave

173

104.0 103.0 102.0 Employment (LFS) 101.0 100.0 99.0 98.0

Labour productivity (GDP/FTE)

97.0 96.0 GDP 95.0 94.0 I

II

III 2006

IV

I

II

III 2007

IV

I

II

III 2008

IV

I

II

III 2009

IV

I

II III 2010

Fig. 9.1 Italy: Employment, gross domestic product and labour productivity—2006 Q1–2010 Q3 (Indices, 2006 Q1 = 100; seasonally adjusted data). Source: Istat, National Accounts and Labour Force Survey

The hardest hit sector was manufacturing, where the impact of the collapse in exports was most severe. The fall was nationwide, about half of it occurring in the North. In the service sector, by contrast, while self-employment decreased significantly, there was a marginal increase in wage employment in the Centre and North. Overall, job losses were relatively limited in the Centre and substantial in the North and especially in the South. The drop in employment was particularly steep for men, while women’s job losses were less pronounced, though still significant in relative terms. The selfemployed (craftsmen and small businessmen in services and manufacturing) and the weakest segments of the labour force (fixed-term employees and freelancers) were the hardest hit, while ‘‘core’’ staff workers were less affected. The pattern for unemployment—measured according to the standards of the International Labour Organization and Eurostat—has essentially been the mirror image of employment, though with significant differences in timing (Fig. 9.2). The number of job seekers reached a low of 1.5 million in the second quarter of 2007 and then rose steadily, as the pace of new job creation slowed. Between the second quarter of 2007 and the same quarter of 2010 the number of job seekers rose by 614,000, while between the employment peak of the first quarter of 2008 and the third quarter of 2010 the number of persons in work fell in analogous fashion, by 607,000. Overall, thus, the labour force remained relatively stable, albeit with some cyclical adjustment and a substantial shift from employment to unemployment. However, with the working-age population still growing by more than 230,000 a year, the employment slump also produced very noticeable discouragement, reflected in the fall in the participation rate from its peak of 63.5% in the second quarter of 2008 to 61.4% in the third quarter of 2010.

174

L. Tronti and R. Gatto 23,600

2,200

23,500

2,100

23,400

2,000

23,300

1,900

23,200

1,800

23,100

1,700

23,000

1,600

22,900

1,500

Unemployment (000)

Employment (000)

Employment

Unemployment 22,800

1,400

22,700

1,300 I

II III 2006

IV

I

II III 2007

IV

I

II III 2008

IV

I

II III 2009

IV

I

II III 2010

Fig. 9.2 Italy: Employment and unemployment—2006 Q1–2010 Q3. Source: Istat, Labour Force Survey

Most of the additional unemployment involved men, though it was significant among women as well. Job seekers were concentrated in the regions of the North and among job losers. There was a much less pronounced increase in unemployment in the Centre and especially in the South. The average unemployment rate rose from 6.7% in 2008 to 7.8% in 2009 and reached 8.4% in the first half of 2010. The increase was particularly severe among young people aged 15–24, and was greatest among young women in the South. The increase in joblessness was accompanied by an increase of more than 500,000 in the number of working-age economically inactive, mostly people not actively seeking work because they do not expect to find it.

9.2 Unemployment and its Measurement: ‘‘Objective’’ and ‘‘Subjective’’ Unemployment, Hidden Unemployment These figures fairly well depict the recession’s current effect on what we may call ‘‘traditional unemployment,’’ i.e. official statistics of joblessness. But it must be acknowledged that this picture is defective in two ways. One is the ‘‘objective’’ nature of the status of unemployed. The indicator that has been the international standard for two decades now entails two objective filters to a person’s inclusion among the unemployed: ‘‘active job search’’ (he must have taken at least one definite job search action)1 and a time standard (he must have taken that action no more than 4 weeks before the survey). In other words, the ‘‘objective’’ definition completely ignores the respondent’s self-perception. 1

The 13th International Conference of Labour Statisticians (1982) listed eight examples of such actions.

9 Measuring the Long Wave

175

The objective definition began to be developed at the thirteenth International Conference of Labour Statisticians, organized by the International Labour Office (now Organization) in 1982. The conference determined that in normal labour markets (in terms of job search tools, market organization and extension, level of demand and rate of self-employment), only those who had taken ‘‘specific steps in a specified recent period to seek paid employment or self-employment’’ could be considered unemployed. Subsequently, following some measurement tests using alternative standards and a highly elusive discussion in the United States,2 in 1990 the ILO specified this ‘‘recent period’’ as 4 weeks (Hussmanns, Merhan and Verma 1990).3 It is essential in this context to note that the ILO interprets this four-week standard as a reasonable compromise between the need to count only people who are actively seeking work at the time of the survey and recognition that searching for a job is not necessarily a continuous activity.4 In Italy, Istat has used the international standard since 1992, but continues to ask for the respondent’s own perception of his/her status (the ‘‘subjective criterion’’) and to record job search actions undertaken prior to the official cut-off date. Obviously, for reasons of international comparability, those not corresponding to the ‘‘objective’’ standard are not counted as unemployed in the official statistics. For years now, some Italian researchers have underscored the shortcomings of the merely objective depiction of unemployment.5 In fact, empirical analysis of the patterns of entry to employment shows that, holding constant individual characteristics (sex, age, education, etc.), there is no statistically significant distinction in job entry pattern between the ‘‘objective’’ and the ‘‘subjective’’ unemployed. Which explains a finding that is well known to those who have worked on longitudinal studies of the labour market6: namely that employment entry is quite frequent even outside the group of the ‘‘objectively’’ unemployed. However, the employment outcomes of those describing themselves as jobless but not satisfying the objective criteria or those who do not even describe themselves as unemployed are not enough to make the official measures unrealistic. This critique, in fact, 2

For a strongly critical account, see Carmignani (2009). To be sure, this indication is set out only indirectly, more as a common practice than a scientifically based recommendation: ‘‘The recent period specified for job search activities need not be the same as the basic survey reference period of one week or one day, but might be longer. The 13th ICLS did not specify the length of the job search period. It left its determination open to countries. In practice, most countries define the job search period in terms of the last month or the past 4 weeks’’ (Hussmanns 2007). 4 ‘‘The purpose of extending the job search period somewhat backwards in time is to take account of the prevailing time lags involved in the process of obtaining work after the initial step to find it was made. During these time lags persons may not take any other initiatives to find work. In particular, this may be the case of persons who can only apply for employment with one potential employer (e.g. judges) and are awaiting the reply to their application for a job’’ (Hussmanns 2007). 5 Among others, Viviano (2003); Battistin, Rettore and Trivellato (2005); Brandolini, Cipollone and Viviano (2006); Carmignani (2009). 6 For a sample of these studies, see the references in Di Laurea et al. (2006). 3

176

L. Tronti and R. Gatto

is based implicitly on the unproven claim that there is a strong, exclusive correspondence between unemployment, however measured (‘‘subjectively’’ or ‘‘objectively’’), and eventual entry into employment. Actually, though, studies of flows into employment have shown that in many cases job entry is instantaneous, with no search whatever (seasonal work, family work, etc.), just as many cases of unemployment (self-perceived or not) drag on and on and never result in employment. This discrepancy is most evident in the South, suggesting that the southern Italian labour market may still be one of those cases mentioned by the 13th ICLS, in which the objective standard is not fully appropriate, because ‘‘the conventional means of seeking work are of limited relevance, (…) the labour market is largely unorganised or of limited scope, (…) labour absorption is at the time inadequate, or (…) the labour force is largely self-employed’’. With the exception of the large scale of self-employment—a condition, in any event, that continues to be much more common in Italy than in the other euro area countries, seriously limiting the significance of many comparative studies—the other conditions are found to a varying extent in many parts of the South if not throughout it. In particular, inadequate labour absorption is especially relevant, as it sharply reduces the effectiveness, for the unemployed, of assiduous job search and thus fosters what can be called ‘‘hidden unemployment’’. This is not counted in the official statistics but nevertheless corresponds to a situation of perceived unemployment and rational job search practices reflecting the characteristics of the local labour market. It is not easy to gauge the extent to which the employment slump described above may have resulted in a ‘‘statistical submersion’’ of joblessness beyond the conventional measure. What is certain is that if in normal times the official unemployment rate represents only a part—most but not all—of those who are currently jobless but wanting to work, in times of slumping employment, when the disappearance of hiring and contract renewals undermines hopes of finding work and discourages job search, its capacity to represent the real situation is inevitably diminished. These considerations however, correct as they may be, should not be used to bolster any destructive critiques to the conventional definition of unemployment, as the benefits of accepted international standards, achieved through a long study path, are indisputably evident. But the Italian case, analysed in this paper, provides a good example of the usefulness of devising new, additional indicators. Complementary to the standard ones, these can help obtain relevant information to grasp the complexity of the phenomenon. The downturn-induced increase in the number of inactive persons (those not covered by the conventional definition of ‘‘labour force’’) can be estimated at more than half a million and most of these newly idled are people who are not actively looking for jobs because they don’t think they can find one. This gives us a first approximation for estimating hidden unemployment, i.e. the decision to take steps to seek work less frequently than needed to qualify as unemployed by the official standard, simply because those steps are less likely to lead to employment. This behaviour may not yet be an outright abandonment of the search for a job but does signal loss of confidence that it will be an effective way to find work.

9 Measuring the Long Wave

177

9.3 Labour Hoarding Another aspect to consider in assessing the impact of the crisis on unemployment beyond the standard measurement of unemployment—more frequently cited in the debate—is labour hoarding, i.e. the preservation of the employment relationship during recession by employers who reckon the cost of dismissal and subsequent rehiring is greater than that of keeping the worker on. Such a choice is all the more likely in the case of skilled workers and those in whom the employer has made a substantial training investment. Labour hoarding is certainly a concept of labour underutilization rather than unemployment, as the worker retains his job, even if working-time is reduced and can be even annulled to zero hours per week. For the worker, however, being under labour hoarding can be a status quite similar to unemployment or, at least, a clear signal of a strong increase in the probability of losing his job. This explains why some analysts would like labour hoarding to be included in the unemployment count or, at least, in some other official measure of workers’ distress. Since 1945, Italy has had a public incentive for labour hoarding in the form of a special social insurance programme, providing short-time working compensation benefits (and other European countries have since done the same).7 This programme sustains firms in economic difficulty or undergoing restructuring, maintaining employment relationships, on reduced working time or even at zero hours, and compensating workers with income support. It thus makes it more advantageous for the employer to retain staff during recession and, at the same time, gives workers more stable jobs and income, softening the social and economic repercussions of slumps and mitigating potential industrial conflict. In recent years, and all the more decisively with the onset of recession, the Italian government has extended the range of eligible beneficiaries, as well as the term of benefits (even where the beneficiaries lack the necessary national insurance contributions), through ‘‘waivers’’. This has certainly eased the economic and psychological impact of recession for beneficiaries. But at the same time it has meant lower utilization of the labour capacity of these individuals. In 2009, the unemployment funded through the programme can be estimated at the equivalent of 300,000 fulltime, year-round workers, some 210,000 more than in 2008. In 2010, the estimate, which corrects the number of hours of short-time working compensation authorized to firms by a coefficient equal to the ratio between hours authorized and hours actually paid, is about 380,000. Not all firms are eligible for short-time wage supplementation, however, even with the new ‘‘waivers’’, and not all the reduced use of available labour benefits from social shock absorbers. We can estimate the amount of labour hoarding outside the short-time working compensation programme by the Wharton

7

For a comparative analysis of programmes for labour redundancy in seven European countries from both the legal and the economic standpoint, see Carabelli and Tronti (1999).

178

L. Tronti and R. Gatto

method,8 assuming that interpolation between the cyclical peaks in productivity gives a reliable measure of the labour utilization frontier, so that underemployment can be estimated as the gap between that frontier and current productivity. By this methodology, if this non-supported labour hoarding was equivalent to some 140,000 workers in 2007, in 2008 it rose to 370,000 and in 2009, owing to steadily falling labour productivity, peaked at 960,000. As productivity recovered, in the first three quarters of 2010, it fell back rapidly, and the third quarter value was 210,000.

9.4 Unemployment and ‘‘Semi-Employment’’ The second, and even more serious reason for dissatisfaction with the limitations of standard international statistics of the labour market stems from the increasing importance of fixed-term and temporary work: fixed-term contracts, temporary employment, seasonal work, freelance collaborations, casual work, etc. Here, the traditional labour statistics risk not only overestimating unemployment but improperly representing unemployment and employment alike. As some recent studies have pointed out (Carmignani 2009; Schiattarella 2009), the temporary nature of jobs for ever-growing segments of the labour market means that, instead of the two traditional and mutually exclusive work statuses (‘‘employed’’ and ‘‘unemployed’’), there is a continuum of positions of ‘‘semi-employment’’, inevitably weakening the capacity of official statistics to accurately portray the labour market. In other words, people whose employment is intermittent simply cannot be labelled, during the time periods envisaged by the official statistics (month, quarter, year) as unequivocally and exclusively either employed, unemployed or inactive. For example, if the average duration of the employment spells of intermittent workers were around 4 months a year (as studies of semi-employment hypothesize), then at any given time a ‘‘semi-employed’’ person would have one chance in three of being officially counted as unemployed or inactive. This would do away not only with the clear conceptual distinction between employed and unemployed, but also with the ideal matching of the statistical representation of persons with the variables of the labour market. The snapshot of employment and unemployment would no longer portray given and relatively stable population groups with characteristic attitudes and economic behaviour, but merely the distribution at a given point in time of the occupational status of a shifting population that consists increasingly of persons who, depending on the time of the survey, may find themselves employed, unemployed, or inactive. Their attitudes and behaviour are accordingly quite different from those of the ‘‘permanent’’ employed and unemployed.

8

For a critical presentation, see Taylor (1970).

9 Measuring the Long Wave

179

Istat’s labour force survey finds a quarterly average of over 2.1 million fixedterm workers, which means that, counting the 400,000 freelancers as well, there are 2.5 million people (11% of all employed persons) characterized by potentially discontinuous employment. That is to say, if the ‘‘semi-employed’’ population strictly speaking, i.e. people whose fixed-term jobs do not produce steady employment, were equal to two-thirds of that quarterly potential, and if the estimate of four months’ work per year were accurate, the actual size of this labour market segment would be not 2.5 million but twice that: 5 million persons. The recognition of a semi-employed segment this large would inevitably have enormously important consequences. It would affect not only the statistical portrayal of the labour market and analysis of the behaviour of labour supply but also, for instance, projections of the exact size of the group of discontinuous workers who will claim, in the future, subsidized pensions. On the premise that a high level of semi-employment will ordinarily result in an overestimate of unemployment or inactivity (the effect, that is, will be opposite to the restrictive, objective definition of unemployment examined above), what impact may the crisis and recession have had on this labour market segment to date? Let us suggest two possible effects. One, which is certain—and reflected even in the official statistics—is the decline in discontinuous employment, which largely involves young people under 34 in manufacturing and the service sector, and especially in the education system. In 2009 alone this segment contracted considerably, by an estimated 250,000 units on average for the four quarters, counting both freelancers and fixed-term employees. If, pending more accurate measurement, we consider the proportion of discontinuous workers assumed above to be valid, this would imply a drastic reduction of about half a million people in semi-employment. The second effect—with an even greater need for empirical confirmation—is the probable shortening of average annual working time for the semi-employed. If the recession has induced firms not only not to renew a substantial portion of fixed-term contracts, but also to shorten those that they do sign, then this also implies a lower probability that a ‘semi-employed’ worker will be working at the time of the survey. Hence the shortening of the average employment duration during the year diminishes the ‘‘statistical visibility’’ of the semi-employed. This tends to decrease the detection of persons involved in at least some periods of employment during the year and, correspondingly, to increase the numbers of the officially unemployed and inactive.

9.5 Some Empirical Evidence from the Labour Force Survey’ Although the National Statistical Institutes are usually not yet equipped to provide regular data on hidden unemployment, labour hoarding and semi-employment, we can use, in the Italian case, the data from Istat’s labour

180

L. Tronti and R. Gatto

force survey to construct indicators giving a fuller and more reliable picture of an increasingly complex labour market and the employment repercussions of the recession.

9.5.1 Discouraged Workers The labour force survey gathers important information relevant to the ‘‘discouraged worker syndrome’’, in that it identifies the inactive persons who nevertheless show some interest in the labour market, insofar as they state that they do want to work and would be immediately available to do so. As a rule, these persons are not actively seeking work because they do not believe the market gives them any chance of finding a job. Quite surprisingly, in times of recession, this group generally contracts, because the perception of higher unemployment leads them to rule out all prospects of employment.9 The data on the present recession confirm this stylized fact. The number of discouraged workers decreased by 13.3%, between 2008 Q2 and 2009 Q2, from 1,254,000 to 1,465,000,10 then, in the second quarter of 2010 started again to grow (by 5.5%) (Table 9.1). Discouragement is more common among women—about two thirds of the total—although over the 3 years the gap between the sexes narrowed. The South, though accounting for the majority of discouraged workers, recorded a modest decline. Only one seventh of the discouraged reside in the Centre, but their numbers declined more sharply over the period. The smallest change was in the Northeast, where the discouragement syndrome is also least widespread.

9.5.2 Short-Time Working Compensation at Zero Hours One phenomenon on which additional information can be gathered by the labour force survey is workers receiving short-time working compensation (Cassa Integrazione Guadagni). Despite the income support benefit and the retention of their jobs, for these workers this status is a cause of malaise, and in fact some analysts feel that at least those working zero hours should be counted as unemployed. The survey does not report which type of benefit a worker is receiving (ordinary or extraordinary, for how many hours a week, how much, etc.). But we can record those ‘‘employed’’ workers who report having been absent from 9

The reflux into inactivity, however, is the dual behaviour of the attraction mechanism, stylised in the 60s by Tella (1964), for which the empirical evidence shows a strong correlation to the employment cycle. 10 All the data from the Istat labour force survey in this article refer only to the working-age population (15–64).

val. ass. var. tend. val. ass. var. tend. val. ass. var. tend. val. ass. var. tend. val. ass. var. tend. val. ass. var. tend. val. ass. var. tend.

Discouraged Total val. ass. var. tend. Males val. ass. var. tend. Females val. ass. var. tend. Northwest val. ass. var. tend.

South

Centre

Northeast

Northwest

Females

Males

Total Total

%

%

%

%

%

%

%

%

%

%

%

1.317 – 405 – 913 – 163 –

2.932 – 1.218 – 1.715 – 466 – 292 – 479 – 1.695 –

Q1

1.254 – 363 – 891 – 154 –

2.712 – 1.066 – 1.646 – 411 – 277 – 427 – 1.597 –

Q2

1.431 – 428 – 1.003 – 206 –

2.874 – 1.115 – 1.759 – 481 – 278 – 463 – 1.651 –

Q3

1.393 – 417 – 975 – 202 –

3.095 – 1.230 – 1.865 – 520 – 297 – 529 – 1.749 –

Q4

1.417 7.6 445 10.0 972 6.5 172 5.4

3.230 10.1 1.328 9.1 1.902 10.9 485 4.2 327 11.8 521 8.8 1.897 11.9

Q1

1.465 16.9 431 18.7 1.035 16.1 194 25.9

3.212 18.4 1.263 18.4 1.949 18.4 506 23.0 327 17.9 553 29.4 1.827 14.4

Q2

1.428 -0.2 455 6.5 973 -3.0 183 -11.2

3.003 4.5 1.212 8.7 1.792 1.9 471 -2.2 298 7.3 501 8.1 1.733 5.0

Q3

1.240 -11.0 403 -3.4 837 -14.2 158 -21.6

3.102 0.2 1.368 11.2 1.734 -7.0 552 6.1 305 2.7 526 -0.5 1.719 -1.7

Q4

1.343 -5.2 429 -3.7 914 -5.9 180 4.7

3.497 8.3 1.560 17.5 1.937 1.8 649 33.7 382 17.0 590 13.1 1.876 -1.1

Q1

1.270 -13.3 410 -4.8 860 -16.9 157 -18.6

3.341 4.0 1.520 20.4 1.821 -6.6 644 27.3 413 26.4 562 1.7 1.722 -5.7

Q2

1.242 -13.1 426 -6.5 816 -16.1 145 -20.5

3.263 8.7 1.529 26.2 1.734 -3.2 616 30.8 425 42.4 537 7.2 1.686 -2.7

Q3

1.213 -2.2 391 -3.1 822 -1.8 144 -9.1

3.586 15.6 1.662 21.5 1.924 10.9 724 31.2 430 41.1 638 21.2 1.794 4.4

Q4

1.357 1.1 431 0.6 926 1.3 168 -6.4

3.832 9.6 1.766 13.2 2.065 6.6 742 14.3 476 24.5 660 11.9 1.954 4.2

Q1

1.248 0.5 397 -6.8 851 4.3 162 11.7

3.291 0.8 1.510 -1.2 1.781 2.7 618 0.4 392 -7.6 569 6.0 1.712 1.5

Q3

(continued)

1.340 5.5 464 13.2 876 1.9 188 19.2

3.599 7.7 1.708 12.4 1.891 3.8 693 7.7 434 5.2 576 2.5 1.895 10.0

Q2

Table 9.1 Underutilized labour force by labour market status, sex and region—2007 Q1–2010 Q3 (thousands of persons and percentage changes on corresponding period of previous year) 2007 2008 2009 2010

9 Measuring the Long Wave 181

Q4

Q4

1.704 20.6 807

1.527 9.0 729

1.775 7.3 899

1.982 12.5 1.010

1.839 7.9 937

2.145 20.8 1.102

1.814 18.8 953

Q3

1.761 13.2 847

Q2

229 163.8 170 156.7 59 186.5 87 164.4 59 185.3 38 167.3 45 137.0

Q1

208 335.6 151 462.2 57 172.6 72 360.1 59 525.0 24 174.1 52 281.6

Q4 98 0.8 171 -6.2 800 -0.3

Q3 122 -11.3 178 -10.8 796 -12.3

Q2

2009

111 136 137 97 108 107 8.5 18.7 7.5 -18.9 -3.2 -21.7 201 213 200 183 168 187 3.4 22.0 -8.9 -14.0 -16.8 -12.3 933 922 908 802 888 818 8.8 13.8 3.5 -6.6 -4.8 -11.2 week due to short-time working compensation 52 43 48 87 172 233 -12.7 -7.5 11.5 82.2 232.7 442.9 36 25 27 66 121 173 -9.1 -9.0 -1.0 105.5 236.3 585.3 16 18 21 21 50 59 -20.1 -5.2 33.1 33.6 224.4 238.0 15 13 16 33 64 95 -24.5 -25.4 -12.5 85.0 322.4 626.7 14 7 9 21 44 66 28.1 -29.9 168.7 165.4 210.2 890.3 6 6 9 14 25 27 -47.6 11.8 50.2 102.3 303.1 326.6 16 17 14 19 38 45 -0.1 23.2 -10.9 26.7 139.2 165.9

Q1

Q3

Q1

Q2

2008

2007

val. ass. 102 115 128 119 var. tend. % – – – – Centre val. ass. 195 175 219 213 var. tend. % – – – – South val. ass. 857 810 877 858 var. tend. % – – – – Officially employed persons working zero hours during the Total val. ass. 59 46 43 48 var. tend. % – – – – Males val. ass. 40 28 27 32 var. tend. % – – – – Females val. ass. 19 19 16 15 var. tend. % – – – – Northwest val. ass. 20 18 18 18 var. tend. % – – – – Northeast val. ass. 11 9 3 8 var. tend. % – – – – Centre val. ass. 12 6 6 7 var. tend. % – – – – South val. ass. 16 14 15 15 var. tend. % – – – – Unemployed Total val. ass. 1.556 1.412 1.401 1.655 var. tend. % – – – – Males val. ass. 773 676 660 781

Northeast

Table 9.1 (continued)

2.273 14.7 1.198

201 17.2 137 12.9 64 27.4 72 11.8 56 27.6 33 30.8 40 5.0

111 3.3 186 11.3 892 0.4

Q1

2010

1.864 2.8 991

179 -13.9 123 -18.7 56 -1.2 61 -15.0 43 -26.7 32 32.8 43 -17.9

96 -21.4 175 -2.1 815 2.4

Q3

(continued)

2.093 13.8 1.121

166 -28.5 123 -28.9 44 -26.3 53 -44.1 39 -40.2 35 29.3 39 -13.3

101 -5.7 167 -11.0 885 8.1

Q2

182 L. Tronti and R. Gatto

%

%

%

%

%

%

Q3

– 740 – 257 – 147 – 238 – 759 –

Q4 – 874 – 300 – 170 – 309 – 876 –

9.5 915 16.8 298 5.6 201 12.7 314 15.1 948 15.3

Q1

Q2

– 737 – 240 – 153 – 246 – 774 –

Q1

– 783 – 282 – 179 – 273 – 822 –

2008

2007

Source: Authors’ calculations on Istat, labour force survey data

South

Centre

Northeast

Northwest

Females

var. tend. val. ass. var. tend. val. ass. var. tend. val. ass. var. tend. val. ass. var. tend. val. ass. var. tend.

Table 9.1 (continued)

Q2 19.4 897 21.7 299 24.7 184 20.3 333 35.1 888 14.8

Q3 10.5 798 7.8 272 5.8 152 3.3 292 22.8 812 7.0

Q4 15.1 877 0.2 361 20.1 187 10.5 329 6.4 898 2.5

19.3 972 6.3 405 35.7 231 14.5 397 26.5 950 0.2

Q1

2009 Q2 16.1 902 0.6 391 30.8 240 30.9 348 4.5 859 -3.3

Q3 30.6 861 8.0 398 46.3 244 61.1 334 14.6 838 3.2

Q4 22.6 1.043 19.0 494 36.9 274 46.1 428 30.0 949 5.7

18.6 1.075 10.6 501 23.9 308 33.8 441 11.0 1.023 7.7

Q1

2010 Q2 19.7 972 7.7 453 15.7 294 22.4 375 7.7 971 13.0

Q3 4.0 873 1.4 394 -0.8 253 3.8 363 8.4 854 1.9

9 Measuring the Long Wave 183

184

L. Tronti and R. Gatto

work for the entire reference week, owing to short-time working compensation— taking this as an indication of benefit recipients working zero hours.11 In the last 3 years, the aggregate grew enormously: from 59,000 individuals in the first quarter of 2007 to 179,000 in the third quarter of 2010 (Table 9.1). The biggest jump was in the second quarter of 2009, after which the aggregate continued to expand rapidly, but at a sharply decelerating pace, that turned into net decreases in the second and third quarters of 2010. Table 9.1 shows that zero-hour short-time affects both sexes, but men more extensively than women. Geographically, the sharpest increase was in the Northeast. Throughout the period, the highest number of zero-hour beneficiaries was in the Northwest. The Centre also registered a sharp rise, at rates just below the nationwide average. The South, due to its weak manufacturing structure, shows the smallest increase.

9.5.3 The Labour Underutilization Rate If we count as underutilized labour, on top of the unemployed, those out of work but receiving short-time working benefits and the discouraged, the number of persons underutilized in the third quarter of 2007 is 105.2% greater than the unemployed alone. As the recession advances, however, the number of discouraged workers declines, and with it the percentage gap between the official unemployment rate and the broader measure of underutilization. In any case, total labour underutilization rose from 2,873,000 in the first quarter of 2007 to 3,495,000 in the third quarter of 2010. This broader look at the impact of the crisis on the labour market enables us to construct an underutilization rate counting short-time working compensation plus discouraged plus unemployed as the numerator, and labour force plus short-time working compensation as the denominator. It is an indicator similar to the unemployment rate but considering the broader category of labour underutilization and consequently enlarging the concept of labour force (Fig. 9.3). This measure is obviously higher than the official unemployment rate.12 It has been above 11% since the third quarter of 2007, rising by more than 2.7% points from the second quarter of 2007 to the third quarter of 2010. It is, however, quite interesting to note that the difference between the two rates is fairly stable, around 5% points, due to the compensation between the declining share of discouraged and the hike in the number of short-time workers.

11

For the time being, as far as we know, no estimates of this phenomenon based on administrative data are available. 12 A similar rate is produced and diffused quarterly by the Bank of Italy in its ‘‘Bollettino Economico’’. For the construction details of their ‘‘rate of available unutilized labour’’, see Brandolini, Cipollone and Viviano (2006).

185

15.0

7.0 6.0

12.5 5.0 10.0

4.0 3.0

7.5

2.0 Differences ILO unemployment rate Labour underutilization rate

5.0

1.0

2.5

Differences (percentage p.ts)

Undemployment and labour underutilization rates (%)

9 Measuring the Long Wave

0.0 Q1

Q2

Q3

2007

Q4

Q1

Q2

Q3

2008

Q4

Q1

Q2

Q3

2009

Q4

Q1

Q2

Q3

2010

Fig. 9.3 Unemployment rate by ILO definition and labour underutilization rate—2007 Q1–2010 Q3 (percentage ratios on labour forces; the underutilization rate comprises, further to the ILO unemployed, the discouraged and workers in zero-hours short-time working). Source: Authors’ calculations on Istat, labour force survey data

9.6 Semi-Employment and Longitudinal Analysis 9.6.1 Fixed-Term Employees and ‘‘Semi-Employment’’ Another important area for investigation is the impact of the recession on fixedterm employment (fixed-term contracts, temporary workers, seasonal work, etc.). The time series of fixed-term employees (unlike that for short-time working) has a clear seasonal component, with a trough in the first quarter and a peak in the second. To take this into account, our analysis bears on annual averages or fourquarter changes. The total number of fixed-term employees rose, on a four-quarter basis, until the third quarter of 2008, dropped in the fourth and registered declines of more than 9% in both the second and the third quarters of 2009 (Table 9.2). By the fourth quarter, however, though still contracting, the rate of decline was much slower, and two of the three quarters of 2010 showed a positive sign. For almost this entire period fixed-term work was more common among women than men, and also the sharpest reductions came among women. Regionally, the greatest number were in the South, which is where the dip in employment was steepest. Next is the Northwest, which however recorded no fall in fixed-term employment. The easy termination of fixed-term employment relationships, however, has not induced firms to a more substantial adjustment in this segment. Comparing the number of fixed-term positions lost with the increase in zero-hour

Number % change Number % change Number % change Number % change Number % change Number % change Number % change

Q1

2.126 – 1.020 – 1.106 – 485 – 433 – 437 – 772 –

Q2

2.305 – 1.093 – 1.212 – 503 – 457 – 488 – 856 –

Q3

2.361 – 1.174 – 1.186 – 520 – 484 – 478 – 879 –

Q4 2.282 – 1.112 – 1.171 – 518 – 453 – 436 – 876 –

Q1 2.189 2.9 1.043 2.3 1.146 3.6 534 10.1 431 -0.4 462 5.8 761.0 -1.4

Source: Authors’ calculations on Istat, labour force survey data

South

Centre

Northeast

Northwest

Females

Males

Total

2.443 6.0 1.201 9.9 1.243 2.5 554 10.1 522 14.0 483 -1.1 884.4 3.3

Q2 2.406 1.9 1.233 5.0 1.173 -1.2 552 6.1 509 5.3 483 0.9 862.0 -1.9

Q3 2.255 -1.2 1.101 -0.9 1.154 -1.4 539 4.1 445 -1.8 444 1.8 827.7 -5.5

Q4 2.035 -7.0 987 -5.4 1.048 -8.5 488 -8.7 425 -1.4 404 -12.7 718.5 -5.6

Q1 2.214 -9.4 1.066 -11.2 1.148 -7.6 528 -4.7 457 -12.4 436 -9.8 793.9 -10.2

Q2 2.186 -9.1 1.091 -11.5 1.095 -6.6 500 -9.4 444 -12.7 439 -9.1 802.0 -7.0

Q3 2.174 -3.6 1.064 -3.4 1.110 -3.8 514 -4.6 428 -3.7 446 0.5 785.9 -5.1

Q4

2.047 0.6 1.018 3.1 1.029 -1.8 495 1.6 434 2.2 422 4.6 694.8 -3.3

Q1

2.200 -0.6 1.087 2.0 1.113 -3.1 499 -5.5 459 0.4 460 5.5 783.5 -1.3

Q2

2.198 0.5 1.121 2.8 1.077 -1.7 519 3.8 476 7.2 427 -2.7 774.6 -3.4

Q3

Table 9.2 Fixed-term employees by sex and region—2007 Q1–2010 Q3 (data not seasonally adjusted; thousands of persons and percentage changes on corresponding period of previous year) 2007 2008 2009 2010

186 L. Tronti and R. Gatto

9 Measuring the Long Wave

187

short-time workers suggests that, as more resources for income support were available, firms preferred a mix of solutions in which subsidized labour hoarding played a decisive role. Our analysis of semi-employment suggests that the less than predictable fall in fixed-term employment does not indicate in any way that many of these workers have a job essentially guaranteed by the possibility of contract renewal at expiration. Rather, fixed-term employment is characterized by high turnover, and the labour force survey can tell us what happens to those who lose these positions (Table 9.3). The total increase (2007 Q2–2009 Q2) in non-employment coming from fixed-term jobs was 31.9%. The increase involved men more than women, even though women outnumbered men. The sharpest rises were in the northern regions, though in absolute terms the highest numbers were in the South. The disproportion between the fall in fixed-term employment and the higher rise in the non-employment (inactivity plus unemployment) of previously employed fixed-term workers over the period was due to increased turnover, suggesting a shortening of the average duration of fixed-term contracts. The non-employed who lost a fixed-term job became mostly inactive: their number is, all over the period, about if not more than two times that of the unemployed. But the rates of growth of the unemployed coming from a fixed-term position are appreciably higher, so that their share of non-employment shows a significant increase over time (from 30.5% of 2007 Q2 to 37.4% of 20010 Q2). This evidence is consistent with a picture of accelerated turnover, in which shorter fixed-term appointments generate increasing outflows into the unemployment pool more than into inactivity.

9.6.2 Longitudinal Employment and Unemployment A different way of viewing employment status in a market in which semiemployment is increasingly important is to examine labour market participation over time, in a longitudinal framework, thus overcoming the ILO’s limited time horizon. The ILO’s aim, as we have seen, was to take a sort of snapshot of the labour market, describing it at a fixed point in time and minimizing as much as possible the job search period required to qualify as unemployed at the survey date. But in a market in which a fair proportion of people change their employment status every year, we cannot measure employment simply by answering the question ‘‘How many people are employed today?’’ We also need to answer ‘‘How many people have been employed in the course of the year?’’.13 A longitudinal reading of the survey can extend the time frame for observing employment conditions without necessarily conducting a new survey. Every

13

We have to note that the ILO itself suggested the concept of ‘‘usually active population’’, enlarging the temporal horizon of the classic definitions (see Hussmanns, Merhan and Verma 1990).

Number % change Number % change Number % change Number % change Number % change Number % change Number % change

Unemployed Total Number % change Males Number % change Females Number % change Northwest Number % change

South

Centre

Northeast

Northwest

Females

Males

Total Total

378 – 196 – 181 – 62 –

1.199 – 520 – 679 – 149 – 161 – 203 – 686 –

Q1

331 – 161 – 170 – 57 –

1.084 – 434 – 650 – 128 – 137 – 171 – 648 –

Q2

337 – 160 – 177 – 62 –

1.221 – 485 – 737 – 182 – 152 – 185 – 703 –

Q3

413 – 229 – 184 – 70 –

1.276 – 574 – 702 – 182 – 171 – 205 – 718 –

Q4

439 16.2 239 21.6 200 10.4 85 37.0

1.308 9.1 589 13.3 718 5.8 182 22.0 182 13.0 183 -9.7 760 10.9

Q1

408 23.4 208 29.0 200 18.1 69 21.4

1.199 10.6 501 15.4 698 7.4 170 32.7 159 16.2 190 11.2 680 4.9

Q2

387 15.1 184 14.7 204 15.4 67 9.0

1.303 6.7 522 7.7 781 6.0 193 5.8 162 6.9 214 16.0 734 4.5

Q3

445 8.0 239 4.6 206 12.1 91 28.5

1.333 4.5 601 4.8 731 4.1 190 4.2 192 12.6 237 15.6 714 -0.6

Q4

539 22.8 297 24.5 241 20.7 109 28.6

1.503 14.9 692 17.5 810 12.8 242 32.8 206 12.9 240 31.1 815 7.2

Q1

485 18.7 252 21.4 233 16.0 110 59.7

1.416 18.2 611 22.0 806 15.4 229 34.5 218 37.1 224 18.1 745 9.6

Q2

454 17.1 252 37.1 202 -0.9 102 50.6

1.418 8.8 613 17.3 805 3.1 238 23.4 229 41.0 206 -3.7 745 1.5

Q3

534 19.9 281 17.3 253 22.8 115 26.7

1.442 8.2 656 9.0 787 7.6 217 14.6 244 27.1 227 -4.3 754 5.6

Q4

594 10.3 339 14.1 255 5.6 114 4.8

1.578 5.0 732 5.7 847 4.5 234 -3.2 252 22.5 242 0.7 851 4.3

Q1

474 4.5 256 1.6 219 8.2 104 1.9

1.526 7.7 659 7.5 868 7.8 257 8.1 231 1.0 255 23.7 783 5.1

Q3

(continued)

534 10.1 299 18.5 235 1.0 109 -1.3

1.430 0.9 650 6.4 780 -3.2 215 -5.9 231 5.9 227 1.1 756 1.5

Q2

Table 9.3 Non-employed persons having lost a fixed-term job, by labour market status, sex and region—2007 Q1–2010 Q3 (thousands of persons and percentage changes on corresponding period of previous year) 2007 2008 2009 2010

188 L. Tronti and R. Gatto

Number % change Number % change Number % change Number % change Number % change Number % change Number % change

Number % change Number % change Number % change

821 – 323 – 498 – 87 – 111 – 137 – 485 –

753 – 273 – 480 – 71 – 105 – 114 – 463 –

32 – 57 – 185 –

Q3

885 – 325 – 560 – 120 – 115 – 135 – 514 –

37 – 49 – 189 –

Q4

863 – 345 – 518 – 111 – 121 – 128 – 502 –

49 – 77 – 216 – 869 5.8 350 8.3 518 4.1 98 11.5 125 12.3 117 -14.6 529 9.0

57 14.5 66 0.6 231 15.4

Q1

Q2

Q1

50 – 66 – 200 –

2008

2007

Source: Authors’ calculations on Istat, labour force survey data

South

Centre

Northeast

Northwest

Females

Males

Inactive Total

South

Centre

Northeast

Table 9.3 (continued)

790 5.0 293 7.4 498 3.6 101 41.8 107 1.8 111 -2.0 471 1.7

53 62.9 78 37.6 208 12.8

Q2

916 3.5 339 4.3 577 3.1 125 4.1 114 -0.6 135 0.1 541 5.2

48 30.5 79 59.6 194 2.4

Q3

887 2.8 362 5.0 525 1.3 99 -11.2 129 6.5 149 16.3 510 1.5

63 27.7 89 14.5 204 -5.6

Q4

964 11.0 395 12.7 569 9.8 133 36.3 129 3.3 136 16.7 565 6.8

76 34.2 104 56.5 250 8.1

Q1

2009

931 17.9 358 22.4 573 15.2 119 17.3 145 35.6 141 27.0 527 11.8

74 40.0 83 5.6 218 4.7

Q2

964 5.3 361 6.6 603 4.5 136 8.7 159 38.6 133 -1.6 536 -0.9

70 46.6 73 -7.3 209 8.1

Q3

908 2.4 375 3.5 533 1.6 102 3.4 152 17.4 126 -15.1 528 3.5

92 47.3 101 13.9 226 10.9

Q4

984 2.1 392 -0.7 592 4.0 120 -9.8 154 19.0 136 -0.5 574 1.6

98 28.6 106 2.4 276 10.4

Q1

2010

895 -3.9 351 -2.1 545 -5.0 107 -10.2 140 -3.5 138 -2.1 511 -3.0

92 24.6 88 6.8 245 12.3

Q2

1.052 9.1 403 11.6 649 7.6 154 12.8 160 0.8 173 29.9 565 5.5

71 1.3 82 12.6 218 4.1

Q3

9 Measuring the Long Wave 189

190

L. Tronti and R. Gatto

person in the labour force panel, in fact, is interviewed three times in a year.14 Conventionally, employment for a year is measured as the average of the four quarterly figures. That is, this traditional measure of the employment rate in a given year is the average of the number of people found to be in employment in the four quarters as a percentage of the working-age population. But taking semiemployment into account, we can calculate a different, longitudinal employment rate as the percentage ratio of those who were employed in at least one of the three interviews with the working-age population. The comparison of this indicator with the traditional one reveals the extent of semi-employment, or at least of that part of semi-employment not picked up in the traditional gauge. In a situation of stable employment someone who is employed at the first interview will have a high probability of remaining employed through the year. But in a high turnover situation, the stock of employed persons will be made up of different people each time it is surveyed, and more people will experience shorter employment spells. In this way, taking account how many brief and sporadic episodes of employment there are, the percentage of the population that is economically active is considerably higher than conventional indicators suggest. The relationship between the two indicators then should be: the nearer the standard and the longitudinal employment rates are, the more stable is the labour market, while the bigger their distance, the higher is labour market mobility, the higher the turnover and the shorter the spells of employment. Next, the longitudinal unemployment rate is the percentage ratio between the number of persons who were never employed during the year and were counted as ‘‘actively searching’’ (officially unemployed) in at least one survey and the number who were in the labour force (either employed or unemployed) in at least one survey. This indicator enables us to analyze the other portion of the labour force, people who never found even one job during the year, and so to adjust the unemployment rate for semi-employment, taking account also of those going from unemployment to inactivity and vice versa. The longitudinal unemployment rate proposed here may be higher or lower than the traditional yearly average. If changes of status are more numerous between employment and unemployment (both within the labour market) than between unemployment and inactivity (transitions into or out of the labour market), then this longitudinal rate will be lower than the conventional yearly average; conversely, if transitions between unemployment and inactivity are more numerous, the rate will be higher than the conventional one.

14

Actually, each individual who enters the survey sample is interviewed in two consecutive quarters, skipped for two quarters, and then interviewed again for two more quarters—a total of four interviews in 15 months, a sampling formula dubbed ‘‘2-2-2’’. We elected to hold the interviewing period to a year, hence to three observations, because a year is a clear reference and permits comparison with the traditional indicators expressed as yearly averages. Moreover, limiting the period to just one year enables us to include half the survey sample, while keeping the whole 15 months would cut it to a quarter, making the estimates less reliable.

9 Measuring the Long Wave

191

Table 9.4 Yearly average and longitudinal labour market indicators by sex and region— 2007–2009 (per cent) 2007 2008 2009 Total

Males

Females

Northwest

Northeast

Centre

South

Average employment rate Longitudinal employment rate Average unemployment rate Longitudinal unemployment rate Average employment rate Longitudinal employment rate Average unemployment rate Longitudinal unemployment rate Average employment rate Longitudinal employment rate Average unemployment rate Longitudinal unemployment rate Average employment rate Longitudinal employment rate Average unemployment rate Longitudinal unemployment rate Average employment rate Longitudinal employment rate Average unemployment rate Longitudinal unemployment rate Average employment rate Longitudinal employment rate Average unemployment rate Longitudinal unemployment rate Average employment rate Longitudinal employment rate Average unemployment rate Longitudinal unemployment rate

58.7 64.5 6.2 6.8 70.7 76.6 5.0 4.5 46.6 52.7 7.9 9.9 66.0 70.6 3.8 3.7 67.6 72.5 3.2 3.3 62.3 67.4 5.4 5.9 46.5 53.7 11.1 12.8

58.7 64.4 6.8 7.5 70.3 76.0 5.6 5.2 47.2 53.0 8.6 10.5 66.2 70.9 4.3 4.2 67.9 72.9 3.5 3.3 62.8 67.3 6.2 5.9 46.1 53.5 12.1 14.1

57.5 63.7 7.9 8.0 68.6 75.3 6.9 5.9 46.4 52.3 9.3 10.9 65.1 70.3 5.9 4.8 66.3 72.0 4.7 3.8 61.9 68.3 7.3 6.6 44.6 51.5 12.6 14.9

Source: Authors’ calculations on Istat, labour force survey data

The average employment rate fell by more than 1.2% points, from 58.7 to 57.5%, between 2007 and 2009 (Table 9.4). But was the share of Italians who worked really that low? Counting all those who worked at least once, the proportion of the working-age population in employment for at least part of the year—the longitudinal employment rate–rose by 5.9 points in 2007, and this differential—which measures semi-employment not picked up by the conventional definition—increased to 6.2 points in 2009. The longitudinal rate did fall during these years, but the decrease was smaller than that in the conventional employment rate. The difference indicates that there was an increase in the number of persons working only occasionally (the semi-employed), and who accordingly are not counted as employed in all the year’s surveys. Semi-employment (represented by the gap between the two indicators) diminished among women from 6.1 to 5.9 points. And in fact the average annual employment rate for women slipped from 46.6% in 2007 to 46.4% in 2009. In reality,

192

L. Tronti and R. Gatto

however, more than half of women of working age did work at least some part of the year, given that the longitudinal employment rates were respectively 52.7 and 52.3%. That is, the decline in women’s employment was disproportionately steep for those with occasional work, producing more stable conditions of non-employment. For men, by contrast, the reduction in employment resulted in more unstable and briefer employment. The average employment rate fell by a full 2.1 points, from 70.7 to 68.6%, while the longitudinal rate came down by only 1.3 points, from 76.6 to 75.3%. In the North the disparity between the average and longitudinal indicators is less pronounced than nationwide, but it nevertheless did increase under the impact of recession. The Centre recorded lower rates but wider discrepancies. This is the only Italian region where the longitudinal employment rate rose, even though, as in the rest of the country, the traditional average rate fell. The gap between the two rates accordingly widened. The region with the widest gap is the South, where it nevertheless narrowed. On a nationwide basis, the longitudinal unemployment rate is always higher than the average one. What this means is that in general Italy records more transitions between unemployment and inactivity (at the margin of the labour market) than between employment and unemployment (within the labour market). Pending better international comparison, this finding casts doubt on the widespread perception that Italian unemployment is lower than the European average. For it confirms that the Italian labour market is characterized not only by low participation but also by an extensive ‘‘grey area’’ between unemployment and inactivity. This provides additional support for the argument that the international definition of unemployment does not really fit the Italian labour market. However, a closer examination of the two indicators shows that for some market segments the longitudinal unemployment rate is lower than the standard one. In particular, this applies to men and to the Northwest in all 3 years (2007–2009), and to the Northeast and the Centre in 2008 and 2009. The exceptions with respect to the nationwide mean demonstrate the existence of labour supply segments in which the labour market attachment is however strong, even in times of recession.

9.6.3 Individual Employment Paths Following the employment history of individuals over the three interviews in the course of a year offers further interesting insights. One typical path is employment in all three surveys. It is interesting to compare the individuals in this group with those who were employed in the first of the year’s interviews. Those employed steadily through the year fell from 91.0% of those employed at the start in 2007 to 90.8% in 2008 and 90.6% in 2009, a loss of 0.4 points (Table 9.5). So the recession lowered the probability that someone employed in the first quarter of the year would also be employed in the fourth. The reduction may be due to increasingly rare job opportunities and to the shorter duration of jobs.

9 Measuring the Long Wave

193

Table 9.5 Employment status transitions—2007–2009 (per cent) 2007 2008 2009 Percentage of the persons employed in all the three quarters on those 91.0 90.8 90.6 employed in the first quarter Percentage of the persons employed in the first quarter and not employed in 5.5 5.5 5.7 the last one on those employed in the first quarter Percentage of the persons employed in a fixed-term job in the first quarter 17.2 18.8 19.4 and not employed in the last one on those employed in a fixed-term job in the first quarter Coexistence ratio: persons employed in the first quarter and not employed 104.3 112.3 124.7 in the last as percentage of those unemployed in the first quarter and employed in the last Source: Authors’ calculations on Istat, labour force survey data

The proportion of those who began the year being employed but were nonemployed in the final survey was 5.5% in 2007 and 2008, rising to 5.7% in 2009. Unsurprisingly, the situation of fixed-term employees was worse. The share of fixed-term employees observed in the first yearly survey who were not employed in the last was 17.2% in 2007, 18.8% in 2008 and 19.4% in 2009. Another interesting indicator is the ‘‘coexistence ratio’’ between those beginning the year employed and ending it non-employed (those leaving employment during the year) and those who begin jobless and end up employed. In 2007 the coexistence ratio was 104.3, fairly close to stability, but it rose to 112.3 in 2008— employment exits thus already significantly outnumbering entries—and jumped further to 124.7 in 2009, underscoring Italy’s current deteriorated job market.

9.7 The Long Wave of Unemployment and Labour Underutilization We have sought to assess the recession’s real impact on the Italian labour market, going beyond the shortcomings of the standard international definition of unemployment and analysing both unemployment that is hidden (in that the official statistics do not define it as such) and other labour underutilization phenomena such as discouragement and labour hoarding. We have also tried to take a closer look at semi-employment and at the impact of the crisis on the number and (implicitly) the duration of employment spells. Quantifying all these aspects of the jobs slump, we can derive an overview of the current situation. Compared with the pre-recession situation (ignoring time lags in the different impacts), the official fall in employment was 607,000, while the number of fixed-term workers fell by 198,000 (quarterly average) and freelance contracts by 90,000. Adopting the arguments set forth in Sect. 9.4, then, we can estimate that, actually, more than 500,000 workers left the status of semiemployment, about a third going into unemployment and the other two thirds into inactivity. So, the total fall in employment and semi-employment could have been

194

L. Tronti and R. Gatto

about 1.100,000. On the other hand, official unemployment increased by 614,000 and hidden joblessness by more than 500,000 (even though the number of discouraged workers narrowly defined fell by 217,000). Finally, the impact of labour hoarding can be estimated at about 140,000 workers on short-time working compensation at zero hours of work, net of short-time workers at more than zero working hours and of labour hoarding without resort to short-time working compensation (more than 160,000 full-time equivalent workers). All in all, then, the Italian labour market has to cope with an increase in joblessness broadly conceived, including the underutilization of labour, of about 1,400,000—more than twice the rise in official unemployment. The economy has now been showing signs of recovery for three consecutive quarters, but there is no hint of improvement in the labour market. Why is this? Ordinarily in Italy the employment cycle lags behind the economic cycle by two or three quarters. The lag depends on the mechanisms by which economic activity is translated into employment, in a series of actions ranging from adjustments in working hours and overtime to hiring and firing. In a slump as severe as the present one, however, it is simply unthinkable that this transmission can be completed relatively quickly. The recovery in employment will be unavoidably slow, because the return to economic growth requires substantial action by firms to improve productivity, which has been growing much more slowly than in Italy’s main euroarea competitors since 1995. And in turn this requires full utilization of labour, through the progressive reduction of labour hoarding, whose adverse repercussions on the job market can be avoided only if world and domestic demand for Italian produce expand rapidly, outpacing productivity. Finally, while firms are reorganizing and, hopefully, reabsorbing underutilized manpower, for years to come the younger generations coming of age will have trouble finding jobs, and youth unemployment will expand. On the hypothesis that the upturn in labour utilization begins in 2011, it is easy to calculate that excess labour hoarding and short-time working compensation will still not be superseded before 2012. And as businesses will not begin hiring again until they have made use of the hoarded labour, unemployment and joblessness will continue to increase, as those already unemployed in 2010 are joined by new labour market entrants. So in the next few years the unemployment rate is likely to rise, as it could only be brought back down to pre-recession levels by particularly rapid, sustained growth. But this tendency may be anyhow weakened by inactivity, forced by reduced employment opportunities, that could exert a downward pressure on official unemployment.

9.8 Conclusions and Policy Suggestions A first look at employment and unemployment shows that, in the Italian case, the effect of the crisis has been fairly modest, certainly less than the drop in output would suggest. But a more considered judgment requires us to take account of the

9 Measuring the Long Wave

195

shortcomings of the conventional measurement of unemployment used internationally (ILO, Eurostat) and try to determine how much joblessness may be hidden by limiting the definition of ‘‘unemployment’’ to persons who have taken concrete job search actions in the last 4 weeks. This definition is especially restrictive in a labour market like southern Italy’s, where the rareness of job opportunities, especially in recession, may induce the rational unemployed to search for work less often Equally problematic is the meaningfulness of the standard measure of unemployment given the increasingly common state of ‘‘semi-employment,’’ i.e. brief intervals of employment alternating with spells of unemployment or inactivity, and labour hoarding, i.e. firms’ deliberate underutilization of available labour. In Italy, labour hoarding is both spontaneous and encouraged by short-time working compensation incentives. Our study uses data from the labour force survey to measure the impact of the crisis on these phenomena. First, the recession increased the number of short-time working compensation beneficiaries working zero hours in the reference week by 140,000 and the number of non-employed (unemployed and inactive) by 930,000. So our broad index of involuntary labour underutilization—the conventional unemployed plus zero-hour short-time working compensation plus discouraged workers—is considerably higher than the official unemployment rate. This broader rate was 11.5% in the third quarter of 2007 and 13.5% in the third quarter of 2010. The increasing importance of all forms of fixed-term work (fixed-term contracts, temporary employment, seasonal work, freelancers, occasional work, etc.) creates several issues for the measurement of employment statuses. These problems consist not only in the risk that the conventional definition may overestimate unemployment, but also in a somewhat misleading picture of both unemployment and employment. Because more and more jobs are temporary, we now find a range of ‘‘semi-employed’’ positions between those of employed and unemployed—mutually exclusive under the conventional definition. Since the official statistics do not envisage this status, people with discontinuous employment are classified either as employed, or as unemployed or inactive. The real situation, however, belies this sharp conceptual distinction, and with it the strict match between the statistical representation of persons and the labour market aggregates. The employment and unemployment aggregates at any given point in time no longer identify distinct and relatively stable population groups (the employed and the unemployed) but only the temporary distribution of employment statuses within a population consisting increasingly of people who, depending on the precise time they are interviewed, may be employed, unemployed or inactive. The longitudinal component of the labour force survey provides a way to address these issues, because we can observe developments in the labour market by comparing standard indicators with longitudinal ones specially designed to take account of semi-employment and the possibility that periods of unsuccessful job search may also be brief. These are: the longitudinal employment rate (the percentage ratio between people employed at least once during the year and the entire working-age population); and the longitudinal unemployment rate (the ratio

196

L. Tronti and R. Gatto

between people who never worked during the year but were formally unemployed at least once and those who participated at least once in the labour force, either as employed or jobless). The principal finding to emerge from this comparison is that the proportion of the working-age population involved in the production process at some time during the year was 5.9% points higher than the standard employment rate in 2007 and the gap (a measure of the semi-employment not picked up in the official statistics) widened to 6.2 points in 2009. We also used the panel component of the survey to gauge the effects of the crisis on workers’ employment paths. The analysis focused on the worsening of employment status during the year for workers who were employed during the first quarter; the aggregates considered were total employed persons and fixed-term employees. In 2007, of those employed in the first quarter 91% were employed in the fourth quarter. This dipped to 90.8% in 2008 and 90.6% in 2009. That is, the recession made it perceptibly more likely that someone at work in the first quarter would lose his or her job in the course of the year. Our quantitative measures show that the problems of the Italian labour market, though apparently less critical than those of other European countries, will necessarily persist for a number of years to come, until the economy can reabsorb both the implicit unemployment corresponding to labour hoarding and the additional joblessness created by the 2008–2009 slump. Even if the job hemorrhage is staunched in 2011, it will still take at least two more years before existing labour hoarding is eliminated. Only later on can the economy begin to pare back unemployment, which will meanwhile have been swollen by young and immigrant new entrants, and the temporarily inactive re-entering the labour market.15 Consequently, the unemployment rate—even the official rate, which as we have seen does not fully reflect job market difficulties—is destined to stay high for many years, and possibly to rise further if employment opportunities open up again. So it is all the more urgent to reform Italy’s unemployment benefit programmes and labour policies in general. But, vital as this reform might be, before addressing it, Italy needs to move resolutely and without further delay to a drastic overhaul that can put the economy in a position to handle the challenges of new technologies and global competition. The crisis must be used to usher in a new industrial and development policy aimed at making Italian firms stronger, more flexible and more technologically advanced—a policy that can remedy the structural misalignment of prices and wages between Italy and the rest of Europe, which for so long has held back Italian growth.16 For the resumption of growth there is no other proper adjustment variable than the elimination of the rents of sheltered firms and industries. The economy must enter a new growth path in which higher wages drive domestic demand and export volumes depend on quality products at

15

For a stock-and-flow econometric analysis of the labour supply in the Italian labour market, see Di Laurea et al. (2006). 16 On this point see, among others, Tronti (2010).

9 Measuring the Long Wave

197

reasonable prices. To avoid the risk of continuing recourse to flexible labour and wages as the adjustment factors, which hark back to a no longer practicable exclusively export-led model, the right course is to build on domestic demand and wages. On the labour market, this strategy would imply moving in the ‘‘flexinsurance’’ direction, i.e. providing flexible workers with at least the same social benefit protection as workers on standard contracts, if not more substantial ones— as the European Employment Strategy originally intended. Unless it radically rethinks its development strategy along these lines, Italy will not be able to reduce unemployment or escape the trap of tight public finance and high taxes. In addition to an effective industrial and development policy, however, economic reorganization requires a coherent system of unemployment benefits, income support, worker rights and retraining for marketable skills. The two levers are complementary, as without these programmes sheltered industries can never be opened up to competition, innovation and competitiveness. If a few words on this subject can be spent as a concluding remark to a paper devoted to measuring the amplitude of the long wave of joblessness, above all, the reform of unemployment benefit programmes requires clarity on the objectives, roles and financial provisions that distinguish public action from that of social partners. Today, in fact, government action is both too little—excessively selective and arbitrary, undermining constitutionally guaranteed universality of rights—and at the same time too much—committed to finance programmes and schemes that should be mainly if not exclusively the province of private, bilateral collective agreements. The key principles governing the public interest, as far as working people are concerned, are income support for people losing their jobs, regardless of industry or firm size (as well as income for first-job seekers, especially relevant in the current situation) and programmes for enhancing workers’ human capital. For firms, the public interest is competitiveness and flexibility in production and in employment, to permit continuous adaptation to global competition. From these simple principles it follows that government must finance unemployment benefits and minimum income, as well as provide employment services and broad policies for employability addressed both to employed and to jobless workers. Social partners have different aims and objectives. Firms’ interests are retaining skilled workers through the ups and downs of the business cycle and avoiding the opposition of unions to downsizing, job mobility and innovation. Workers have different interests, namely keeping jobs and incomes during economic downturns. For this reason such instruments as financed labour hoarding, mobility programmes and outplacement should be the product of management-labour bargaining and should be funded by joint insurance schemes. Government intervention in this sphere should be quite limited and always secondary, in support of the primary role of collective agreements. It should also be limited in time, never losing sight of the general objective of the competitiveness and flexibility needed for the economy to quickly respond to the ever changing conditions of competition. The reform of Italy’s labour market policies thus requires a corresponding enhancement of training policies for employable skills, which must be run in

198

L. Tronti and R. Gatto

conjunction with income support programmes, making explicit the terms of a trade-off: government intervention in return for workers’ and businesses’ commitment to contribute, through their effort and investment, to the country’s progress. Even so, better services to boost employment and employability, along the lines of the European ‘‘flexicurity’’ model, are a necessary but not sufficient condition for reorganizing the economy, committing businesses in all industries to re-engineer the workplace to make for a more streamlined, flexible and innovative economy. A more rational and efficient system of unemployment benefits needs to be accompanied by a bold new industrial and development policy designed to foster economic growth. These are merely two sides of the same coin, and may prove successful only if launched together.

References Battistin E, Rettore E, Trivellato U (2005) Contiamo davvero tutti i disoccupati? Evidenze per l’Italia, 1984–2000, In: CONTINI and TRIVELLATO, 2005 Brandolini A, Cipollone P, Viviano E (2006) Does the ILO definition capture all unemployment? J Eur Econ Assoc 4(1):153–179 Carabelli U, Tronti L (eds) (1999) Managing labour redundancies in Europe. Labour—Special Issue, No. 1 Carmignani F (2009) Lavoro precario e statistiche del lavoro. La difficile rivincita della oggettività del soggetto, ‘‘Economia & lavoro’’, No. 3 Di Laurea D, Gatto R, Gentile M, Righi A, Spizzichino A, Tronti L (2006) La previsione della disoccupazione nelle regioni italiane attraverso il modello stock e flussi. Costruzione del database e primi risultati, ‘‘Rivista di statistica ufficiale’’, No. 1 Hussmanns R (2007) Measurement of employment, unemployment and underemployment— current international standards and issues in their application. Bulletin of Labour Statistics, No. 1 Hussmanns R, Merhan F, Verma SM (1990) Surveys of economically active population, employment, unemployment and underemployment: an ILO manual on concepts and methods, International Labour Office, Geneva Schiattarella R (2009) Lavoro discontinuo nel tempo e funzionamento del mercato del lavoro, ‘‘Economia & lavoro’’, No. 3 Taylor J (1970) Hidden unemployment, hoarded labor, and the Phillips curve. Southern Economic Journal XXXVII(1) Tella AJ (1964) The relation of labor force to employment, Industrial and labor relations review 17(3) April Tronti L (2010) The Italian productivity slowdown: the role of the bargaining model. Int J Manpow 31(7) Viviano E (2003) Un’analisi critica delle definizioni di disoccupazione e partecipazione in Italia, ‘‘Politica economica’’, No. 1. Source: Istat, National Accounts and Labour Force Survey

Chapter 10

Poverty and Unemployment: The Cases of Italy and Spain Tindara Addabbo, Rosa García-Fernández, Carmen Llorca-Rodríguez and Anna Maccagnan

Abstract This paper sets out to detect the costs of joblessness in Italy and Spain, two countries that show major differences in labour market structure and in their reaction to the crisis. We describe the different unemployment insurance systems in the two countries and how the crisis has hit the two labour markets. A multivariate analysis is then carried out to provide an initial estimate of the possible effects of the current crisis on Spanish and Italian households’ well-being by using the European Statistics on Income and Living Conditions Surveys for Italy and Spain. Our results indicate that the unemployed experience a higher degree of income poverty and costs in terms of people’s reduced likelihood of being able to access medical or dental treatment and their being able to afford a week’s holiday in both countries. These costs also change on the basis of people’s employment status prior to unemployment. Keywords Poverty JEL Codes I32

Unemployment Well-being Health Crisis

J6 J65

T. Addabbo (&) A. Maccagnan Dipartimento di Economia, Università di Modena e Reggio-Emilia, Reggio-Emilia, Italy e-mail: [email protected] A. Maccagnan e-mail: [email protected] R. García-Fernández Métodos Cuantitativos para la Economía y Empresa, Universidad de Granada, Granada, Spain e-mail: [email protected] C. Llorca-Rodríguez Departamento de Economía Internacional y de España, Universidad de Granada, Granada, Spain e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_10, Ó Springer-Verlag Berlin Heidelberg 2012

199

200

T. Addabbo et al.

10.1 Introduction The aim of this paper is to analyse the costs of unemployment on household income and wellbeing in Spain and Italy.1 Both countries are members of the Eurozone and their labour markets bear both major resemblances and differences. The two countries are characterised by a high degree of inflexibility in wage determination, rigidity in hiring and firing practices, and by very low levels of female labour force participation (World Economic Forum 2010). They both are characterised by a strong duality: fixed-term versus open-ended contracts, more accentuated in Spain. The wide use of temporary contracts in hiring young workers to avoid the much higher dismissal costs of permanent contracts and the deep recession of the Spanish economy have led its unemployment rate to be twice the European average. Furthermore, temporary employment accounted for close to 90% of all job losses in the 12 months to June 2009 (European Commission 2009, p. 16) and the youth unemployment rate in Spain was 41.6% in 2010 versus 27.8% in Italy and 20.3% in the Eurozone. To make matters worse, Spain and Italy have employment protection systems in line with the Mediterranean model and are therefore characterised by a rather low unemployment benefit coverage (Sapir 2005). However, the conditions for the receipt of benefits and the benefits duration and amounts that they offer differ widely. Hence, their capability to palliate the socioeconomic consequences of the crisis may differ too. In this regard, we will consider the effect of joblessness on household income and wellbeing and the impact of different systems of unemployment benefit on unemployment sustainability. Our focus is not only on the pecuniary dimension of well-being, but, in keeping with Sen’s multidimensional definition of wellbeing (Sen 1985), we shall also take into account indicators of deprivation in other dimensions of wellbeing. In 1997 the OECD declared that employment status is the most important factor in determining relative income and poverty and that direct taxes and public-sector income transfers substantially reduce income inequality and poverty. Moreover, there is broad agreement on how experiences such as joblessness may lead to a deterioration in levels of happiness as well as life satisfaction and health (Winkelmann and Winkelmann 1995; 1998; Goldsmith et al. 1997; Korpi 1997; Scutella and Wooden 2008; Schmitz 2010; Economou et al. 2008; Stuckler et al. 2009; Kuroki 2010). Many studies have also shown that socioeconomic status 1 Previous versions of this paper were presented at the IZA/OECD Workshop on ‘Economic Crisis, Rising Unemployment and Policy Responses: What Does It Mean for the Income Distribution?’ held in Paris in February 2010, and the AIEL National Conference held at the University G. D’Annunzio Chieti-Pescara in September 2010. This paper is based in part on the research for the International Project ‘Measuring interaction between quality of life, children wellbeing, work and public policies’ supported by the Fondazione Cassa di Risparmio di Modena (FCRMO). We would like to thank FCRMO for its support, as well as Gianna Giannelli, Daniela Mantovani, Hans-Dieter Gerner and four anonymous referees for their stimulating comments on previous versions of this paper. Usual disclaimers apply.

10

Poverty and Unemployment: The Cases of Italy and Spain

201

indicators, such as income, occupation and education, are closely correlated with morbidity and mortality (Lynch et al. 2000; Mustard et al. 1997 and Humphries and Van Doorslaer 2000). In short, our investigation tries to show how unemployment and unemployment protection systems affect household income, poverty and capacity to meet health needs in Italy and Spain. For this purpose, we will use data from the European Statistics on Income and Living Conditions Survey and analyse to what extent individuals experienced income poverty, had unmet medical or dental needs, or whose family could not afford a week’s holiday due to their social and economic status. Our results could guide decision-making underlying social policies. The remainder of the paper is organised as follows: Sect. 10.2 addresses the effect of the economic crisis on the Italian and Spanish labour markets. The third section analyses their unemployment protection systems. Our findings from the multivariate analysis on the costs of unemployment in Italy and Spain are shown in Sect. 10.4. The final section will offer conclusions and suggest policies based on our findings.

10.2 Italy and Spain: Facing the Crisis In this section we focus on the effect of the crisis on the Italian and Spanish labour markets. The comparison between Spain and Italy is of great interest, as their labour markets bear important resemblances, yet they have been hit very differently by the crisis. After having examined data taken from the European Labour force survey for 2007 and 2009, we shall try to explain the reasons for the differences in the impact of the crisis on the two economies. The data in Table 10.1 clearly show a lower level of activity in Italy than in Spain, both among men and women. The Italian male and female activity rate remains respectively 4.2 and 12.7% points below the Eurozone 2007 levels, while male activity rate is 8 points lower than in Spain, and women’s activity rate is 10.7% points lower. As a consequence of the crisis in both countries, male activity rates have slightly decreased, while female participation rates, particularly in Spain, have increased. However, it should be noted that in Italy, female activity rates have increased only in the North (from 59.7 to 60.4%) and in the Centre (from 55.8 to 57.3%). In the South, on the other hand, because of poor employment prospects, many people have stopped looking for a job, and female activity rates have decreased from 36.6 to 36.1%. In this area male rates have also decreased, from 68.4 to 66.3%. The data in Table 10.2 compare the employment rates of men and women in Italy, Spain and in the Euro area for 2007 and 2009. In 2007 these are lowest in Italy: in particular, female employment rate is 46.6% in Italy, 54.7% in Spain and 58% in the Euro area. Also male employment is lower (70.7%, versus respectively 76.2 and 73.4%). In 2007 Italy and Spain were also characterised by a much higher

202

T. Addabbo et al.

Table 10.1 Activity rates, for men and women aged 15–64 2007 2009 Italy Spain Euro area

M

F

M–F

M

F

M–F

74.4 81.4 78.6

50.7 61.4 63.4

23.7 20 15.2

73.7 81 78.5

51.1 64.8 64.6

22.6 16.2 13.9

Source: European LFS Table 10.2 Employment rates, for men and women aged 15–64 2007 2009 Italy Spain Euro area

M

F

M–F

M

F

M–F

70.7 76.2 73.4

46.6 54.7 58

24.1 21.5 15.4

68.6 66.6 71.2

46.4 52.8 58.3

22.2 13.8 12.9

Source: European LFS

Table 10.3 Unemployment rates, for men and women aged 15–64 2007 2009 Italy Spain Euro area

M

F

F–M

M

F

F–M

4.9 6.4 6.6

7.9 10.9 8.5

3 4.5 1.9

6.9 17.8 9.3

9.3 18.5 9.7

2.4 0.7 0.4

Source: European LFS

gender gap in employment than the Euro area, equal to 24.1% in Italy, 21.5% in Spain and 15.4% in the Eurozone. The fall in employment during the crisis was much higher in Spain than in Italy: male employment rates decreased in 2009 to 66.6% in Spain and to 68.6 in Italy; the female employment rate experienced a less sharp decrease in both countries, as in 2009 it was 46.4% in Italy and 52.8% in Spain. This led to a decrease in the gender gap in Spanish employment rates (Table 10.3). As regards unemployment, it should be noted that female unemployment rate in 2007 was well above the male one, both in the Spanish and in the Italian labour market. This trait is not so pronounced in the Eurozone. Unemployment rates increased sharply during the crisis, reaching in 2009 18.5 and 9.3% in Spain and Italy respectively among women, versus 17.8 and 6.9% among men. As far as age groups are concerned (Table 10.4), we may see that Italy and Spain are both characterised by very high youth unemployment rates (those aged 15–24), respectively equal to 20.3 and 18.2% in the year 2007. The Spanish unemployment rate in this age group was 37.8% in 2009, while in March 2011 it reached 44.6%. Italy shows a behaviour pattern closer to that of the Eurozone, although its unemployment rate for those aged between 15 and 24 was 25.4% in 2009

10

Poverty and Unemployment: The Cases of Italy and Spain

Table 10.4 Unemployment rates by age group 2007 Italy Spain Euro area

203

2009

15–24

25–49

50–64

15–24

25–49

50–64

20.3 18.2 15

5.8 7.4 6.7

2.5 6.1 6.1

25.4 37.8 19.7

7.4 17.1 8.9

3.7 18.3 6.8

Source: European LFS

(23.3% for men and 28.7% for women) and reached 28.6% in March 2011. It is also the age range most affected by job losses in Italy: 8.1% versus 2.7% for those between 25 and 54 years of age and versus a generation of employment in the other age group. It should be highlighted that the unemployment rate falls as the education level attained by the young people rises, yet in Italy it is higher for those with tertiary education than for those with an upper secondary and post-secondary level (29.6 versus 24.1% in 2009). Moreover, it should be noted that the percentage of the population aged 18–24 with at most lower secondary education and not in further education or training is much higher in Spain than in Italy or in the Eurozone (31.2% in 2009 versus 19.2 and 15.9% respectively), with a rate lower for women than for men in all three cases. This fact questions the effectiveness of the Spanish vocational training model to place young people in employment. Finally, it should be underlined that some regions are affected by unemployment with different intensities both in Spain and in Italy. Ceuta, Melilla, Andalucía, Extremadura and Asturias present unemployment rates much higher than the national total in Spain, and they are respectively equal to 20.3, 18.2, 12.8, 13.1 and 8.5%, compared to an average rate of 8.3% in 2007. In Italy the Southern regions suffer more intensely from unemployment, with an unemployment rate of 12.5% in 2009, versus 11.1% in 2007 (in Northern Italy, which is also the richest, unemployment rates increased from 3.5 to 5.4% over the period analysed). We have seen that, despite the fact that Italy and Spain are both members of the Eurozone and are characterised by an employment protection system in line with the Mediterranean model, the impact of the current economic crisis has been very different in the two economies. The reasons for this are diverse. Firstly, the Spanish labour market is characterised by a very high incidence of temporary employment. Since the liberalisation of temporary hiring in the labour reform of 1984, it increased significantly and in 2007 it was about twice the European and Italian levels (Table 10.5). It should be noted that the Italian labour market has a lower rate of temporary contracts compared to the Eurozone but not by a great margin. This different entity of more unstable employment should entail different costs of joblessness in the two labour markets (Table 10.6). Temporary employment is more prevalent among women than men in all cases, although this gender gap is more pronounced in Spain and Italy than in the Eurozone. It also affects younger people more, especially those aged between 15 and 24.

204

T. Addabbo et al.

Table 10.5 Temporary employment (as a percentage of total employment) by gender 2007 2009 Italy Spain Euro area

M

F

F–M

M

F

F–M

11.2 30.6 16.1

16.0 33.1 18.0

4.8 2.5 1.9

10.8 23.8 14.3

14.6 27.3 16.7

3.8 3.5 2.4

Source: European LFS

Table 10.6 Temporary employment (as a percentage of total employment) by age group 2007 2009 Italy Spain Euro area

15–24

25–49

50–64

15–24

25–49

50–64

42.3 62.8 51.3

12.2 31 14.2

6.3 15.3 6.8

44.4 55.9 49.9

11.6 25.7 12.9

5.7 12 6.3

Source: European LFS

In particular, in 2007 in Spain 62.8% of young adults were on temporary contracts, compared to 42.3% of Italians and 51.3% of the Eurozone average. In Spain, temporary employment accounted for close to 90% of all job losses in the 12 months to June 2009 (European Commission 2009, p. 16). As a consequence, the incidence of temporary employment dropped to 55.9% in 2009 among young people. Furthermore, in Spain, manufacturing activities, construction and trade sectors are those most affected by temporary employment. Moreover, unlike Italy, the Spanish labour market has been hit by one of the most extreme housing speculation bubbles in the world, leading to heavy job losses in the building sector. It should also be stressed that in 2009 in Spain, 30.5% of employees with university education had temporary jobs, which shows that investment in education does not protect the workforce against unstable work conditions in this country, with the same intensity as it does in the Eurozone or Italy (respectively 23.7 and 18.4%). Another reason why the crisis has hit Spain more hard than Italy is linked to the fact that in Italy the unemployment increase has been matched with an increase in the use of wage supplementation funds. This has limited the negative impact of the crisis on Italian poverty rates (see Addabbo et al. 2010), although the Italian unemployment insurance system is very heterogeneous and therefore likely to increase inequality amongst the unemployed. However, it should be noted that Italian labour market figures are worse than they appear from the official unemployment rates, according to the ILO definition, for they do not take into account the unemployed who were not actively seeking a job in the 4 weeks prior to the interview but who had been seeking work before (and had then become discouraged). Furthermore, neither are wage guarantee fund beneficiaries included in the official unemployment figures, as we shall discuss in the following section.

10

Poverty and Unemployment: The Cases of Italy and Spain

205

10.3 Comparison Between the Italian and Spanish Benefit Systems Spain and Italy have employment protection systems corresponding to the Mediterranean model which are characterised by rather low provision of unemployment benefits (Sapir 2005). However, the conditions for the receipt of benefits and the duration and amounts of the benefits they offer differ. Thus, their capability to palliate the socioeconomic consequences of the crisis may differ too. This section aims to show the differences in the two systems also with regard to employment status prior to the jobless spell, in order to detect the conditions that are more likely to cause severe cuts in jobless people’s income. The unemployment benefit system in Italy is characterised by inequalities derived from differences in the eligibility conditions, and in the different duration and degree of coverage (Table 10.7). The latter, given the relation of eligibility to previous employment, is on average low: in 2007, during the first 5 years of unemployment, the net replacement rate in Italy was on average 7% while the OECD average was of about 52% (OECD 2009a). Moreover, the degree of unemployment benefit coverage changes according to the type of contract, with lower coverage for collaborators and temporary workers (Bank of Italy 2009a). In Italy, the number of workers who access the wage supplementation fund has increased. According to National Social Insurance Institute data, in the second quarter of 2009 the number of hours paid by the wage supplementation fund increased by 60% on the first quarter, with the highest increase since 1985 of the number of employees drawing on this fund, yet who are not statistically computed amongst the unemployed (Bank of Italy 2009b). The number of the National Social Insurance Institute (INPS) authorised wage guarantee fund hours to be increased by 311.4% between 2008 and 2009 (INPS data). The highest increase in 2009 was in the metallurgic sector (+866%) followed by the mechanical (+449%), woodwork (+425%), trade (+410%), transport and telecommunications (+397%), mineral and non-metal minerals work (+335%), services (+335%) and extraction (+328%) (INPS data). Notwithstanding the recent extension of the wage guarantee fund system, the Bank of Italy’s simulations on EU SILC and ISTAT labour force survey data show that about 1.6 million employees or those on collaboration contracts would not have access to unemployment benefit in case of redundancy or contract interruption (Bank of Italy 2009a). The simulation by Berton, Richiardi and Sacchi (2009), based on INPS microdata, shows that between 1.5 and 2 million workers would not be covered by unemployment benefits if they lost their jobs. The inclusion of discouraged workers and wage supplementation funds beneficiaries in the pool of unemployed would have increased Italian unemployment rates in 2009 (second quarter) from 7.4 to 10.2% (Bank of Italy 2010). On the other hand, in addition to the contributory benefit, the Spanish unemployment protection system includes assistance benefits and the so-called Active

Italy

Mobility benefits

Wage suplementation funds (cigs)b

Ordinary unemployment benefita

Amount

(continued)

60% of the average gross earnings received over the last 3 months for the first 6 months, 50% for the 7th month, 40% for the following months. Maximum = €1,031.93 Reduced Requirements: At least Number of days previously worked 35% of the average daily wage for 78 days work over the previous up to a maximum of 180 days the first 120 days, 40% of the year average daily wage for the following days. Maximum = €1,031.93 80% of the average gross earnings Ordinary: Non-worked hours due to Usually 13 weeks. Maximum paid for non-worked hours. 12 months over a period of temporary reduction or suspension Maximum = unemployment 2 years of activity benefit Special: Suspension of activity due to Normally 12 to 24 months. Maximum sector or area-specific firm 36 months over 5 years restructuring Collective dismissals by firms eligible It depends on the age of recipient and Equal to CIGs for the first for benefit from the CIGs and on the location of the job. 12 months. Reduced by 20% individual dismissal of workers after 1 year. already in CIGs or under Maximum = unemployment bankruptcy proceedings benefit

Duration

Ordinary Requirements: Maximum of 8 months (12 for the Contributions for at least unemployed aged over-50) 52 weeks during the 2 year period prior unemployment

Table 10.7 The Italian and Spanish unemployment benefit systems (2008) Benefit Conditions for receipt

206 T. Addabbo et al.

Conditions for receipt Duration Amount Contribution for a minimum of It increases with contribution record. 70% of reference earnings—average 360 days in the 6 years preceding Maximum of 720 days. gross earnings over the legal status of unemployment last 180 days—for a maximum period of 180 days, then 60% of reference earnings for the remaining period. Maximum = €1356,86 Unemployed without any income of In general 6 months. Maximum of 80% of the IPREM (€413,52) any kind exceeding 75% of the 18 months (24 or 30 months for minimum inter-professional wage claimants whose contributory and in a special social situation benefit has run out and have family responsibilities) To meet special social situations like Maximum of 11 months. 80% of the IPREM (€413,52) those of disabled workers.

b

Particular conditions apply for workers in the agricultural and building sector Workers of small manufacturing firms and of most service activities are excluded c Particular conditions hold for workers in the agricultural sector d Those over 45 who have exhausted their entitlement to contributory benefit for 24 months may receive from 80 to 133% of IPREM Source: OECD, 2009b

a

Active income for job insertion

Unemployment assistanced

Table 10.7 (continued) Benefit Spain Unemployment insurancec

10 Poverty and Unemployment: The Cases of Italy and Spain 207

208

T. Addabbo et al.

Insertion Income. The coverage of contributory unemployment benefit in Spain varies depending on the contribution made to the system—work days accumulated—and the previous employment status of the unemployed, since the employee’s contribution base determines the amount of compensation. This is graduated as time out of work increases. The existing subsidies are linked to income no higher than 75% of the monthly minimum wage and provide no more that 80% of PIMEI.2 The AII, introduced in 2000 with the differentiated nature of contributory and assistance benefits, is a program to support the employability of groups with special difficulties in entering the labour market and with financial needs, such as the long-term unemployed, people with disabilities, returnees and victims of gender violence or domestic violence. In 2010 it consists of a monthly payment of €426.00, with supplements in certain cases, as well as contributions for the Social Security to healthcare benefits and family protection. According to OECD data, the net replacement rate during the first year of unemployment in 2007 was 69%, with a 5 year average of 39% in Spain compared to an OECD average of 52% and 28% (OECD 2009a, Table 1.6, p.76). The Spanish system is, therefore, more generous than the Italian one if we go by this criterion. In the first quarter of 2010 the number of beneficiaries of the assistance level of unemployment protection system in Spain increased by 82.5% over the same period in 2009. However, those covered by contributory benefits had dropped by 1.7%. The annual increase in the total beneficiaries of the protection system as a whole— contributory benefits, assistance benefits and Active Insertion Income—was 47.8% in 2009. Of these, 13.86% were foreigners benefiting mainly the contributory level, although the assistance level and the Active Insertion Income additions were at their highest since 2008. In 2009 the coverage rate of the Spanish system of unemployment protection was 75.48% (Spanish Ministry of Labour and Immigration). By sector of activity, the highest rates of yearly change in the first quarter of 2010 were recorded in Agriculture and Services on the contributory level, and in Construction and Industry on the assistance level (Spanish Ministry of Labour and Immigration). This was to affect levels of income inequality, income poverty and wellbeing in Spain since the current crisis has negatively affected these sectors of activity more intensely. It should be stressed that, according to the OECD (2009a, pp. 62–63), labour market program spending—both in Italy and in Spain—is comparable with that of the countries with a strong aggregate employment performance, but more weighted towards passive benefits. According to Eurostat 2007 data on labour market policies (Eurostat 2009, Table B.1.2 p.13), support for labour market policies in 2007 amounted to an average of 60.8% in the EU-15 countries, with an estimate of 66.8% in Spain and 63.7% in Italy.

2 Public Indicator of Multiple Effect Income replaced the minimum wage on July 1, 2004 as a benchmark for social security benefits, but the conditions of access and maintenance are still used for reference.

10

Poverty and Unemployment: The Cases of Italy and Spain

209

Having assessed the extent of unemployment in the two countries by using descriptive statistics and the differences in the unemployment benefit systems, the aim of the following section is to analyse the costs of unemployment in Italy and Spain in order to infer the costs of unemployment in the current crisis.

10.4 The Effect of Joblessness on Household Income and Wellbeing In this section we show the results of multivariate analyses carried out to estimate the effect of joblessness on household income and wellbeing and the impact of individuals’ previous employment status (and relative unemployment benefits) on their unemployment sustainability. Our focus is not only on the pecuniary dimension of wellbeing, but also on the socio-economic impact of unemployment. It is worth noting the social impact of high unemployment rates: they imply a decrease in purchasing power, a loss of human capital and the so-called discouraging effect on the long-term unemployed (Berger et al. 2009, p. 14), their social costs being broad (Sen 1997a, b). This analysis allows us to make an initial estimate of the possible effects of the current crisis on Italy and Spain. We test whether the differences between the two countries are statistically significant by applying a Hausman-White Test (White 1994) on the set of coefficient variables common to the two countries. In particular, if we use the suest STATA command proposed by Weesie (1999), which computes a Seemingly-Unrelated ClusterAdjusted Sandwich-Estimator, we find that the coefficients of the predictor variables statistically differ between the two countries. One direct cost of unemployment is loss of income. Italian unemployment benefits are very fragmented and this can produce different costs according to one’s prior employment status. The OECD (2009a) analysis on the ability of the social transfer system to alleviate poverty indicates that in Italy the alleviation of poverty focuses more on jobless householders than on working households. In Spain the impact of social transfer on the poverty rate is neutral toward these two groups. Preliminary results indicate an increase in the number of households that have experienced great difficulties in making ends meet (17% in 2008 against 15.4% in 2007); with worse figures in the South of Italy (from 22% in 2007 to 25.6% in 2008) whereas it is stable and lower in the Centre (14.3%) and North (12.6%) (ISTAT 2009c). The results of the EU SILC-European Union of Statistics on Income and Living Conditions for Spain (ES SILC 2007) point to an increase in the number of households with difficulties in making ends meet from 10.3% in 2007 to 12.2% in 2008.3 3

We do not include geographical areas disaggregation for Spain because the first level of disaggregation of Eurostat (nuts) does not reflect the heterogeneity of the Spanish labour market. To obtain significant results we should refer to the eighteen autonomous communities plus the two autonomous cities. This will be the task of a future paper.

210

T. Addabbo et al.

We extend our analysis to the increased probability of the unemployed being poor in terms of income. For this purpose we estimate a probit model using IT SILC 2007 and ES SILC 2,007 microdata (Table 10.8).4 Income poverty increases for those who are unemployed.5 In Italy, the effect of being unemployed on income poverty is higher for those unemployed who were previously self-employed, whose probability of being income poor increases by 47% if they were previously self-employed with employees and by 34% if they were self-employed without employees, followed by those never employed before (+23%) and by those who were previously employees (+14%). Taking current employment status into account, we may see that the probability of income poverty significantly increases by 13% if they are self-employed without employees, and by 6% if they are selfemployed with employees. The higher income poverty probability for the currently self-employed may be connected both to income underreporting and to the inclusion in the group of self-employed without employees of those who are in non-standard collaboration working positions facing lower levels of income. A lower but still significant effect is connected to temporary work that is found to increase the probability of poverty by 9%. In the Spanish case (see Table 10.8) the probability of being income poor significantly increases among those who have never worked before (40%). Poverty increases also among the unemployed who have previously been self-employed with employees (45%) and by 27% if they were self-employed but without employees. The same considerations as in the Italian case could be made about the income underreporting and the inclusion of non-standard collaboration working positions in the group of the self-employed. The individuals who were formerly employees present a lower probability than the former of being below the poverty threshold. Nonetheless, their probability of being below the poverty threshold increases by 21%. The probability of being income poor also increases for the inactive (15%) and the effect is significantly higher than in Italy (5%). By comparing the marginal effects of different education levels in the two countries, we may see that higher education protects more against the risk of income poverty in Italy than in Spain. Turning to the current work situation, parttime work increases the probability of poverty in Spain (+4%) whereas it decreases the probability of being poor in Italy (-8%). The latter may be connected to the greater spread of part-time work in Northern Italy, where household income is on average higher and part-time work is more often chosen by women for family reasons. In Spain, 59.1% of part-time workers accept positions with lower wage 4

For this purpose we estimate probit models, as they may be considered an appropriate response model when the dependent variable is dichotomous. Our response probability is included in the [0,1] interval, and it is defined as the standard normal cdf of a linear function of the independent variables (Greene 2008). 5 Income poverty is defined in terms of equivalised disposable income being less than the poverty threshold -60% of median equivalised disposable income. Equivalised total disposable household income is obtained by using the modified OECD equivalence scale.

10

Poverty and Unemployment: The Cases of Italy and Spain

Table 10.8 Probability of being income poor in Italy and in Spain Italy Income poor Age Age squared 0.00007 Female Married or cohabiting Sep. Divorced Widow Secondary High school Tertiary Part-time Temporary contract Self-employed with employees Self-employed without employees Unemployed previously self-employed with employees Unemployed previously self -employed without employees Unemployed previously employee Unemployed never employed before Inactive Chronic ill At least one child aged less than 5

Spain Income poor

Marginal Effects

0.009 0.044** (5.88) -0.001** 0.0001

0.025** (3.33) -

0.005

(6.73) -0.067* (2.38) -0.189** (4.54) 0.249** (4.39) 0.046 (0.52) -0.272** (6.79) -0.611** (14.74) -1.054** (16.71) -0.375** (7.32) 0.371** (8.53) 0.254** (3.90) 0.494** (12.38) 1.367**

0.466

(4.26) -0.046 (1.64) -0.087* (2.31) 0.293** (4.37) -0.466** (4.34) -0.188** (5.27) -0.428** (11.09) -0.655** (16.05) 0.197** (3.95) 0.267** (8.62) 0.860** (13.67) 0.844** (19.67) 1.378**

0.336

(2.84) 0.915**

(3.53) 1.041** (7.24) 0.507** (7.06) 0.759** (8.96) 0.242** (4.76) 0.051 (1.38) 0.279** (6.79)

Marginal Effects

211

-0.014 -0.041 0.059 0.010 -0.055 -0.122 -0.141 -0.080 0.092 0.061 0.128

0.136 0.225 0.053 0.011 0.066

(4.16) 0.767** (16.57) 1.258** (12.16) 0.658** (19.64) 0.001 (0.04) 0.103 (1.95)

0.0003**

-0.009 -0.017 0.065 -0.066 -0.034 -0.072 -0.104 0.041 0.055 0.244 0.230 0.452

0.269

0.206 0.402 0.148 0.0002 0.021 (continued)

212

T. Addabbo et al.

Table 10.8 (continued) Italy

At least one child aged from 6–14 At least one child aged 15–17 South Constant Observations

Income poor 0.259** (8.19) 0.311** (8.72) 0.771** (30.06) -1.663** (11.02) 33,423

Spain Marginal Effects 0.060 0.075

Income poor 0.205** (5.31) 0.193** (3.09)

Marginal Effects 0.043 0.041

0.185

33,423

-1.596** (11.15) 28,063 28,063

Robust z statistics in parentheses significant at 5%; ** Significant at 1% Source: IT-SILC 2007 and ES-SILC 2007 *

levels: elementary occupations (33.22%) and services/marketing sales workers (25.9%). We also find that the presence of children in the household increases the probability of being income poor in both countries: this probability being higher in Italy than in Spain. Finally, in Italy, living in the South increases the probability of poverty by 19%. The South of Italy, in fact, is characterised by deep-seated social and economic problems, such as poor economic development, lack of infrastructure, high and long-term unemployment. As a consequence of this, the probability of being income poor is much higher in this part of the country. In order to account for different dimensions of the costs of joblessness in the two countries, we analysed the probability of having unmet medical or dental needs. Previous studies have outlined the relevance of the non-pecuniary costs of joblessness (Sen 1997b; Winkelmann and Winkelmann 1998) including the costs connected to poorer mental and physical health, with an increase in costs connected to the spell of unemployment and significant differences according to gender and the previous type of job (Sen 1997b; Paul and Moser 2009). We estimated a Probit model with the probability of not having access to medical or dental visits or treatment as a dependent variable, being considered too expensive according to the EU SILC microdata for the two countries. The results of our estimate are shown in Table 10.9. The unemployed have a higher probability of not having access to medical or dental visits or treatments since they are considered too expensive in both countries, and the result differs on the basis of previous employment status. In Italy the group of unemployed who see the highest increase in this cost of unemployment is made up of those unemployed who were previously self-employed without employees (+9%), whereas in Spain the ones who bear the highest cost (considering their employment previous to current unemployment status) are those who were never employed before the

10

Poverty and Unemployment: The Cases of Italy and Spain

213

Table 10.9 Probit model on the difficulties in accessing medical and dental visits-treatments in Italy and in Spain Italy Spain

Age Age squared Female Married or cohabiting Sep. Divorced Widow Secondary High school Tertiary Part-time Temporary contract Self-employed with employees Self-employed without employees Unemployed previously self-employed with employees Unemployed previously self -employed without employees Unemployed previously employee Unemployed never employed before Inactive Chronic ill At least one child aged less than 5

Health difficulties

Marginal Effects

Health difficulties

Marginal Effects

0.019* (2.18) -0.0002 (1.48) 0.078* (2.42) -0.023 (0.48) 0.254** (3.91) 0.236* (2.54) -0.172** (3.74) -0.330** (6.90) -0.765** (11.35) -0.167** (2.98) 0.213** (4.66) -0.400** (4.81) 0.085 (1.79) 0.136

0.002

0.002

0.017

0.040* (2.38) -0.0004* (2.36) -0.027 (0.54) 0.027 (0.32) 0.439** (3.93) 0.012 (0.08) -0.041 (0.76) -0.352** (5.61) -0.552** (6.33) 0.110 (1.34) 0.215** (4.12) -0.181 (1.20) -0.056 (0.61) -0.061

-0.003

(0.31) 0.515**

0.087

(0.14) 0.436

0.032

(3.48) 0.259** (3.32) 0.028 (0.26) -0.197** (3.39) 0.457** (12.75) 0.043 (0.89)

-0.00002 0.009 -0.003 0.035 0.033 -0.019 -0.037 -0.057 -0.019 0.028 -0.035 0.010

0.036 0.003 -0.022 0.068 0.005

(1.05) 0.361** (4.67) 0.629** (4.16) 0.123** (1.99) 0.368** (6.78) -0.149 (1.45)

-0.00002 -0.001 0.001 0.031 0.001 -0.002 -0.014 -0.021 0.006 0.012 -0.007 -0.003

0.024 0.056 0.006 0.023 -0.006 (continued)

214

T. Addabbo et al.

Table 10.9 (continued) Italy

At least one child aged from 6–14 At least one child aged 15–17 South Constant Observations

Health difficulties 0.142** (3.85) 0.200** (4.67) 0.191** (6.51) -1.876** (10.82) 33,423

Spain Marginal Effects 0.017 0.026

Health difficulties -0.114 (1.53) -0.261* (2.06)

Marginal Effects -0.005 -0.010

0.023

33,423

-2.793** (8.52) 28,063

28,063

Robust z statistics in parentheses Significant at 5%; ** significant at 1% Source: IT-SILC 2007 and ES-SILC 2007 *

unemployment spell (+6%). Note that for the inactive the probability of having unmet health needs decreases in Italy, again probably showing a higher role of protection played by the family, which however, does not entail the youngest children. In fact the risk of having unmet medical and dental needs increases in Italy with the presence of children aged over six whereas the reverse holds true for Spain, and this should be borne in mind when assessing child wellbeing in terms of health status in the two countries. In order also to take other dimensions of wellbeing into account, we examined whether households can afford to pay for a week’s annual holiday away from home regardless if they want. This event may be considered a deprivation resulting in a lower development of the ability to take care of oneself and enjoy culture and the environment. The unemployed are more likely to be exposed to the risk of living in households unable to afford a week’s holiday in both countries, and in Spain this event is more frequent for the unemployed who were never employed before (29.5%) while in Italy this event is more likely for the unemployed who were previously self-employed. In both countries being employed as a temporary worker increases the likelihood of not being able to afford a week’s holiday away from home by about 11% (Table 10.10).

10.5 Conclusions As a result of the current crisis, the Italian and the Spanish labour markets have experienced an increase in unemployment rates. The financial crisis has inflicted extreme hardship on the Spanish labour market, especially in the last quarter of 2008 and the first of 2009. As a result of the heavy job losses suffered, especially

10

Poverty and Unemployment: The Cases of Italy and Spain

215

Table 10.10 Probability of not being able to afford a week’s annual holiday away from home Italy Spain

Age Age squared Female Married or cohabiting Sep. Divorced Widow Secondary High school Tertiary Part-time Temporary contract Self-employed with employees Self-employed without employees Unemployed previously self-employed with employees Unemployed previously self -employed without employees Unemployed previously employee Unemployed never employed before Inactive

No holiday

Marginal Effects

No holiday

Marginal Effects

0.026** (4.30) -0.0004** (5.06) -0.022 (1.02) -0.167** (5.36) 0.138** (2.94) 0.092 (1.21) -0.250** (7.23) -0.668** (19.06) - 1.260** (26.89) -0.171** (4.21) 0.289** (8.50) -0.447**

0.010

0.010

-0.149

0.028** (4.28) -0.0004** (5.17) -0.036 (1.54) -0.282** (8.81) 0.342** (5.78) -0.230* (2.48) -0.270** (8.67) -0.620** (18.62) - 1.055** (29.96) 0.111** (2.62) 0.312** (11.93) -0.291**

(8.31) -0.001

-0.0005

(4.43) -0.034

-0.012

(0.04) 0.913*

0.351

(0.86) 0.100

0.037

(2.35) 0.680**

0.266

(0.26) 0.510*

0.197

(3.43) 0.450**

0.175

(2.40) 0.569**

0.219

(6.94) 0.390**

0.152

(13.48) 0.759**

0.295

(4.87) -0.019

-0.007

(6.87) 0.075**

0.027

-0.0001 -0.008 -0.062 0.052 0.035 -0.091 -0.237 - 0.343 -0.063 0.111

-0.0001 -0.013 -0.102 0.130 -0.078 -0.094 -0.203 - 0.323 0.040 0.116 -0.097

(continued)

216

T. Addabbo et al.

Table 10.10 (continued) Italy

Chronic ill At least one child aged less than 5 At least one child aged from 6–14 At least one child aged 15–17 South Constant Observations

No holiday (0.45) 0.240** (8.34) 0.011 (0.33) 0.026 (0.98) 0.110** (3.60) 0.769** (36.04) -0.430** (3.53) 33,423

Spain Marginal Effects 0.091 0.004 0.010 0.041

No holiday (2.59) 0.320** (11.64) -0.004 (0.08) 0.015 (0.44) 0.010 (0.18)

Marginal Effects 0.120 -0.001 0.006 0.004

0.289

33,423

-0.377** (3.04) 28,063

28,063

Robust z statistics in parentheses significant at 5%; ** Significant at 1%

*

by men, the Spanish labour force has shrunk slightly, and the rate of youth unemployment was 37.9% in 2,009. Although the impact of the recession has been more severe in Spain, the Italian data on unemployment must be complemented with data on the beneficiaries of wage supplementation fund (who are not computed among the unemployed) to assess the effect of the crisis on the labour market more completely. Moreover, one should also take into account the high percentage of people of working age who have been discouraged from undertaking active job-hunting and are therefore not calculated in official unemployment figures. The latter should not be neglected by public policies designed to increase participation in the labour force in the two labour markets. The share of those not actively seeking work is especially high in Italy amongst women and in the South. Economic literature has widely examined the link between unemployment, income inequality, poverty and wellbeing. Interest in this subject increases in the wake of economic recessions, given the negative effects they have on labour markets. As microdata on household income for the year 2009 are not yet available, we try to infer the costs of the current crisis through an analysis of the socioeconomic costs of unemployment by using the EU SILC microdata available for the two countries. Our results indicate that the unemployed experience a higher degree of income poverty and costs in terms of the reduced likelihood of accessing medical or dental treatment or of being able to afford a week’s holiday away from home in both countries. Costs also change according to employment status prior to unemployment. When we consider the state of employment prior to the individual becoming

10

Poverty and Unemployment: The Cases of Italy and Spain

217

unemployed, we see that in Italy the higher risk of being poor is connected to being previously self-employed, while for the Spanish the probability of poverty increases more if the person has never been employed before. This difference may be connected to the higher role of protection played by the family of origin in Italy than in Spain, and this may also explain part of the higher effect of inactivity on income poverty in Spain. The unemployment insurance system in Italy is characterised by inequalities deriving from differences in the conditions of eligibility and in the different duration and degree of coverage according to the type of contract. The coverage of contributory unemployment benefits in Spain varies depending on the contributions made to the system—work days accumulated—and on the prior employment status of the unemployed, so it protects young people with less intensity, making them those most punished by unemployment. Spanish subsidies are linked to the possession of income no higher than 75% of the monthly minimum wage and provide no more that 80% of PIMEI. Finally, Spanish unemployment protection is more generous than the Italian one if we measure it as shown in OECD data. The net replacement rate during the first year of unemployment in 2007 was 69% with a 5 year average of 39% in Spain compared to 37% in Italy with a 5 year average of 7%. Together with social policies aiming to provide jobless people with better access to health services,6 we believe that the two countries’ unemployment protection systems need to be reformed. They do not produce relatively high employment rates nor do they keep the risk of poverty relatively low compared to other European systems. The extension of Wage Supplementation Fund access in Italy and the introduction of the Program for Temporary Unemployment Protection and Integration in Spain, which were measures taken as a reaction to the crisis, prove this. The choice of the exact measures to adopt will require further analysis and simulations to identify those most suited to the characteristics of the two countries, and this will be the object of future works.

References Addabbo T, García-Fernández R, Llorca-Rodríguez L, Maccagnan A (2010) Income distribution and the effect of the financial crisis on the Italian and Spanish labour markets. Materiali di discussione del Dipartimento di Economia Politica. Università di Modena e Reggio Emilia, N. 639, December 2010 Bank of Italy (2009a) Relazione annuale 2008, Roma, Banca d’Italia 6 With regard to access to health treatment, during the crisis regional governments in Italy introduced temporary exemptions of prescription charges for visits to medical specialists and tests for the unemployed or redundancy wage supplementation fund recipients and their families. This reflects the need to improve access to health services for the unemployed and is in line with our results from the multivariate analysis.

218

T. Addabbo et al.

Bank of Italy (2009b) Bollettino Economico, n. 57, July 2009 Bank of Italy (2010) Bollettino Economico, n. 59, January 2010 Berger J, Keuschnigg C, Keuschnigg M, Miesse M, Strohner L, Winter-Ebner R (2009) Modelling of Labour Markets in the European Union. Final Report. DG EMPL/D/1 ref. No. VC/2007/0344 Berton F, Richiardi M, Sacchi S (2009) Quanti sono i lavoratori senza tutele Cappariello R, Zizza R, (2009) Dropping the Books and Working Off the Books, Bank of Italy, Temi di Discussione (Working Paper), 702, January 2009 European Commission (2009) EU Employment Situation and Social Outlook. Monthly Labour Market Monitor, November Economou A, Nikolau A, Theodossiou I (2008) Are Recessions Harmful to Health After All? Evidence from the European Union, Discussion Paper 2007–18, March 2007, Centre for European Labour Market Research Eurostat (2009), Labour market policies—expenditure and participants, data 2007, Luxembourg: Publications Office of the European Union Goldsmith AH, Veum JR, Darity W Jr (1997) Unemployment, joblessness, psychological wellbeing and self-Esteem: theory and evidence. J Socio Econ 26(2):133–158 Greene WH (2002) Econometric analysis, 5th edn. Prentice-Hall, NJ Humphries KH, Van Doorslaer E (2000) Income-related health inequality in Canada. Soc Sci Med 50:663–671 Istat (2009a) Annuario Statistico Italiano 2009, Roma, Istat Istat (2009b) Rapporto annuale, La situazione del Paese nel 2008, Roma: Istat Istat (2009c) Condizioni di vita e distribuzione del reddito in Italia. Anno 2008, Famiglia e Società, Statistiche in breve, 29 dicembre 2009 Istat (2010) Occupati e disoccupati, Dicembre 2009: Stime provvisorie, Comunicato Stampa, Istat, 29/01/2010. http://www.istat.it Korpi T (1997) Is utitlity related to employment status? Employment, unemployment, labor market policies and subjective well-being among Swedich youth. Labour Econ 4:125–147 Kuroki M (2010) Suicide and unemployment in Japan: evidence from municipal level suicide rates and age-specific suicide rates. J Socio Econ 39:683–691 Lynch JW, Davey-Smith G, Kaplan GA, House JS (2000) Income inequality and mortality: importance of individual income, psychosocial environment or material pathways in health. Br Med J 320:1200–1204 Mustard CA, Derksen S, Berthelot JM, Wolfson M, Roos LL (1997) Age-specific education and income gradients in morbidity and mortality in a Canadian province. Soc Sci Med 45:383–397 OECD (1997) Income distribution and poverty in selected OECD countries in OECD Economic Outlook, 62, ABI/INFORM Global 49–59 OECD (2009a) Employment Outlook, Paris, OECD OECD (2009b) Benefits and wages: country specific files. http://www.oecd.org/els/social/ workincentives Paul KI, Moser K (2009) Unemployment impairs mental health: meta-analyses. J Vocational Behav 74(3):264–282 Sapir A (2005) Globalisation and the reform of European Social Models. Background document for the presentation at ECOFIN Informal Meeting in Manchester, Bruegel, Brussels, 9 Sept 2005 Schmitz H (2010) Why are the unemployed in worse health? The causal effect of unemployment on health. Labour Economics. doi:10.1016/j.labeco.2010.08.005 Scutella R, Wooden M (2008) The effects of household joblessness on mental health. Soc Sci Med 67:88–100 Sen A (1985) Commodities and capabilities. North Holland, Amsterdam Sen A (1997a) Inequality, unemployment and contemporary Europe. Int Labour Rev 136(2): 155–171 Sen A (1997b) The penalties of unemployment. Rome, Banca d’Italia, Roma Temi di discussione, 307

10

Poverty and Unemployment: The Cases of Italy and Spain

219

Stuckler D, Basu S, Suhrcke M, Coutts A, McKee M (2009) The public health effect of economic crises and alternative policy responses in Europe: an empirical analysis. Lancet 374: 315–323 Weesie J (1999) Seemingly unrelated estimation and the cluster-adjusted estimator. Stata Tech Bull 52:230–257 White H (1994) Estimation, Inference and Specification Analysis. Cambridge university press, NY Winkelmann L, Winkelmann R (1995) Unemployment: Where does it hurt? Discussion paper No. 1093, Centre for Economic Policy Research, London Winkelmann L, Winkelmann R (1998) Why are the unemployed so unhappy? evidence from panel data. Economica 65(257):1–15 February 1998 World Economic Forum (2010) The Global Competitiveness Report 2010–2011. Geneva

Chapter 11

Labour Market Transitions During the Financial Crisis in Italy Marco Lilla and Stefano Staffolani

Abstract The aim of this paper is to evaluate the consequences of the financial crisis on the Italian labour market by analysing the determinants of individuals’ transitions between occupational states. For this purpose, we use micro-data from the ISTAT Labour Force Survey, a repeated cross-section dataset. In order to identify the determinants of transitions, two strategies are followed: firstly, we consider the ‘‘perceived’’ employment condition, available at the individual level for the previous year. Then, following the ‘‘ecological inference’’ approach, we get quarterly transition probabilities for the ‘‘official’’ employment condition. The economic crisis has mostly penalized young, migrants, men, and people living in the North as well as couples without children and single parents with children. Keywords Crisis JEL Codes J23

Transitions Unemployment Ecological inference

J63

11.1 Introduction The financial crisis has strongly affected labour market performances. In this paper, we intend to evaluate the consequences of the crisis on the different segments of the Italian labour market, by analysing the transition probabilities M. Lilla (&) S. Staffolani Dipartimento di Economia, Università Politecnica delle Marche, Piazzale Martelli n. 8, 60121 Ancona, Italy e-mail: [email protected] S. Staffolani e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_11, Springer-Verlag Berlin Heidelberg 2012

221

222

M. Lilla and S. Staffolani

between the employment conditions of individuals, before and after the economic crisis. Our purpose is to identify the individuals whose employment perspectives have been more deteriorated by the crisis, i.e. the most vulnerable individuals in the Italian labour market. The effects of the financial crisis on the employment perspectives have been recently analysed in various economic papers for several countries. Mossfeldt and Österholm (2010) estimate the effects on the Swedish labour market, and they conclude that the NAIRU will be in equilibrium again in 2013. Boysen-Hogrefe and Groll (2010) investigate the German labour market during and after the recession. They argue that the pronounced wage moderation before the crisis and the flexibility in adjusting working time during the crisis is the reasons for the very good performance of Germany. Arpaia and Curci (2010) analyse the labour market adjustment to the 2008–2009 recession in terms of employment, unemployment, hours worked, and wages for all EU Member States. They evaluate the effects on the employment rate by considering the increase in the job separation rate and the decline in the job finding rate. Another interesting contribution is proposed by Choudhry et al. (2010). The Authors, after a review of the economic literature concerning previous crisis, conclude their econometric investigation by saying that: ‘‘Financial crisis impact on labour market indicators is significant: it negatively affects the employment rate and worsens the situation of the unemployment rate’’. The Italian Statistical Institute (ISTAT 2010b) points out that young people have suffered more from the crisis, mainly because they show the highest probability of being employed on a temporary basis (Berton et al. 2009). However, the question remains if other categories of individuals, outside young people, have seen their probabilities of finding or of loosing a job worsened because of the crisis, and if there are any categories of individuals which have not been affected at all by the crisis. Therefore, we will analyse employment state transitions at the turn of the crisis. In Italy, labour market data available in real time are unfortunately scarce. The only dataset available for our purposes comes from the ISTAT Labour Force Survey (Rilevazione Continua delle Forze di Lavoro, RCFL). This survey, which concerns about 170,000 individuals interviewed each quarter, allows researchers to use micro-data with a four months delay. In the survey, a panel dimension exists. Each household is interviewed four times, with a rotation scheme of quarter’s presences, absences and presences. Unfortunately, because of the Italian Privacy Law, ISTAT is not allowed to provide researchers with a constant identifier of the family and of the individual, so that the use of the panel structure is limited to ISTAT researchers.1

1 A longitudinal database is developed by ISTAT, but at the time we began to write this paper, the database was updated to the beginning of 2007. See Discenza and Lucarelli (2009) and ISTAT (2010a).

11

Labour Market Transitions During the Financial Crisis in Italy

223

In order to analyse transitions we therefore need some other information or to use some different methodology than the standard ones, based on transition matrix built at the individual level. We use two different approaches. In the first one, presented in Sect. 11.2, we exploit the questions of the survey referring to the backward employment condition,2 where individuals were asked about their employment condition one year prior to the survey. This information allows us to recover yearly transitions that, unfortunately, must be based on the perceived (self-declared) employment condition. Furthermore, given that the Italian labour market is characterised by a social shock absorber named ‘‘Cassa Integrazione Guadagni’’ (CIG),3 we will also present some figures on individuals employed in firms in crisis, working for a reduced number of hours or not working at all. The second approach, presented in Sect. 11.3, is based on the methodology of ‘‘ecological inference’’, that allows us to estimate transitions among occupational states at aggregate group levels, where groups are defined by individual or family characteristics (King 1997; King et al. 2004). We also investigate on household size and composition, by considering individuals belonging to different family types. The main conclusions of our work seem to suggest that the crisis worsened the labour market conditions of those individuals who had better labour market performances, i.e. males, immigrants, individuals living in the northern regions of Italy.

11.2 Labour Market Transitions 11.2.1 Perceived Versus Official Labour Market Status In the introduction we explained the reasons why the empirical analyses presented in this section are based on perceived employment condition instead of on the official definition.4 Differences between the two definitions need to be stressed. Whereas the former is based on self-declaration, the latter is based on a complex evaluation of the effective status of each individual, depending on a series of questions asked to the

2

We refer to Sect. 11.1 of the questionnaire. It is a sort of almost temporary layoff with a wage compensation mechanism, which preserves (juridically and statistically) the employment relationship, even if there is no possibility for the worker to re-entering the firm. Employees in small firms had also enjoyed CIG thank to the institute of ‘‘Cassa Integrazione in deroga’’. 4 We remind that information at time t and at time t - 1 are available for the perceived employment condition but not for the official one. 3

224

M. Lilla and S. Staffolani

Employment rate: official-perceived Unemployment rate: perceived-official

Fig. 11.1 Differences between ISTAT (official definition) and perceived (auto-declared)

respondents,5 so that perceived and official employment conditions may strongly differ for discouragement effect, underemployment, unpaid work and other reasons. Figure 11.1 highlights how these differences can be relevant at a geographical level, by showing the map of the Italian regions. In the left panel, the difference between the official employment rate and the perceived employment rate is shown. This difference is positive for all the regions, so that the perceived employment rate is lower with respect to the official one in every Italian region. Differences range from 0.5 to 2% points. However, the difference seems to be widespread over Italy. The differences between the perceived unemployment rate and the official one are in the right panel of Fig. 11.1. Note that perceived unemployment regional rates are considerably higher than the official ones: the difference ranges from 2.4 to 20% points and the highest values concern the regions of the South of Italy. Figure 11.1 therefore highlights that the official definition raises the probability of being employed and reduces the probability of being unemployed with respect to the perceived one, and that the following empirical analyses can produce different results given that they are based on different definitions.

5

Have you worked at least one hour during the last week? Have you made at least one effective action in order to find a job during the last month? The official employment condition complies with the Council Regulation (EC) No. 577/98 of 9 March 1998 on the organisation of a labour force sample survey in the European Community.

11

Labour Market Transitions During the Financial Crisis in Italy

225

Table 11.1 Transitions between employment and unemployment Unemployment to employment Employment to unemployment yearly transitions yearly transitions 1 2 3 4 5 6 2007–2008 2008–2009 Difference 2007–2008 2008–2009 Difference Males Females Immigrants Italian citizens 15–24 25–34 35–44 45–54 55–64 Primary Lower secondary Secondary Tertiary North–West North–East Center South Islands Singles Couples without children Couples with children Single parents with children Others families or NA Permanent Temporary Self-employed

27.4 24.8 44.4 24.8 29.8 28.9 25.1 21.6 12.6 15.4 24.5 30.3 32.0 41.0 47.9 34.7 17.8 18.7 26.8 28.1

22.9 21.6 29.2 21.6 23.7 24.8 23.5 17.0 11.5 14.8 20.1 24.7 32.1 30.1 31.0 29.2 18.4 16.3 18.4 24.4

-4.4 -3.2 -15.2 -3.1 -6.0 -4.2 -1.6 -4.6 -1.2 -0.6 -4.4 -5.6 0.1 -10.8 -17.0 -5.6 0.5 -2.4 -8.4 -3.7

2.5 3.1 4.7 2.5 6.7 3.7 2.5 1.9 1.5 3.9 3.3 2.3 1.8 2.1 1.6 2.7 4.1 4.6 2.6 2.1

3.7 4.0 7.2 3.5 9.7 4.8 3.8 2.8 2.2 5.7 4.8 3.3 2.1 3.3 3.0 3.4 5.1 5.8 3.5 3.5

1.2 0.9 2.5 1.0 3.0 1.1 1.3 0.9 0.6 1.7 1.5 1.0 0.3 1.2 1.4 0.6 1.0 1.2 0.9 1.3

26.6

23.3

-3.3

2.6

3.6

0.9

26.3

19.9

-6.4

3.5

5.4

1.9

20.8

20.1

-0.7

4.4

6.0

1.6

1.9 10.8 1.4

2.7 15.8 1.7

0.8 5.0 0.3

11.2.2 Transition Probabilities Table 11.1 presents the yearly transition probabilities, respectively from unemployment to employment and from employment to unemployment, for different groups of individuals belonging to different family types. Observations are divided in the pre-crisis period (2008, that means all the surveys conducted from the first to the third trimester 2008) and in the post-crisis period (all quarters for 2009, including the fourth quarter 2008) and differences between the pre and post-crisis periods transition probabilities are represented in the last column. Columns 1–3 of Table 11.1 show that the probability of finding a job at time t conditional on being unemployed at time t - 1 has decreased for each group

226

M. Lilla and S. Staffolani

we considered, except for people living in the South and for more educated individuals. Young, northern, more educated individuals, males and immigrants show higher probabilities offinding a job in both periods. In the post-crisis period, the probability of finding a job for the unemployed decreases more for males, immigrants, young, those aged between 45 and 54, individuals with secondary education, those living in the North–East. Considering the family types (Singles, Couples without children, Couples with children, Single parents with children, Other or N.A), the probability of finding a job has decreased the most for Singles and for Single parents with children. In general, it seems that the crisis has mostly reduced the probability of finding a job for those categories who used to find a job more easily before the crisis, such as males, immigrants, young and northern individuals. We now turn to the descriptive analysis of the probability of losing a job at time t conditional on having been employed at time t - 1. Columns 4–6 of Table 11.1 show that this probability increased after the crisis for all the groups. Males, immigrants, young and less educated individuals faced a major increase. The center of Italy faced the crisis with a lower waste of jobs. Temporary workers, that obviously loose their job with higher probability even in the pre-crisis period, have seen a strong deterioration of their employment perspectives: many of them are in the youngest cohorts. With respect to the family types, it seems that the crisis has increased the probability of loosing a job for whatever type of family, but less for couples with children (Table 11.2). As explained above, many situations of firms’ crisis have been faced with the CIG institute. For this reason, we present in Table 11.3 data referring to average CIG usage before and after the crisis.6 From Table 11.3 it emerges that the CIG institute concerns more men, adults and less educated individuals. The economic crisis has increased the CIG usage for all groups but those with tertiary education. There are minor differences in CIG usage between pre and post-crisis periods over the groups, even if the elderly, females and people leaving in the Center faced a minor increase.

11.2.3 Estimates of Transition Probabilities In this section we will evaluate if the differences shown above among the groups are statistically significant and, in doing so, we consider three possible employment conditions of the individuals: employment, unemployment and out of the labour force (inactivity). Table 11.4 presents the transition probabilities that we will estimate later by mean of a multinomial logit model. 6 We consider individuals working zero hours (‘‘CIG a zero ore’’), individuals working less than the contractual hours (‘‘CIG a orario ridotto’’) and we include also individuals who did not work during the previous week because of reduced firm’s activity.

11

Labour Market Transitions During the Financial Crisis in Italy

Table 11.2 Employment to unemployment yearly transitions

Males Females Immigrants Italian citizens 15–24 25–34 35–44 45–54 55–64 Primary Lower secondary Secondary Tertiary North–West North–East Center South Islands Singles Couples without children Couples with children Single parents with children Others families or NA Permanent Temporary Self-employed

227

2007–2008

2008–2009

Diff

2.5 3.1 4.7 2.5 6.7 3.7 2.5 1.9 1.5 3.9 3.3 2.3 1.8 2.1 1.6 2.7 4.1 4.6 2.6 2.1 2.6 3.5 4.4 1.9 10.8 1.4

3.7 4.0 7.2 3.5 9.7 4.8 3.8 2.8 2.2 5.7 4.8 3.3 2.1 3.3 3.0 3.4 5.1 5.8 3.5 3.5 3.6 5.4 6.0 2.7 15.8 1.7

1.2 0.9 2.5 1.0 3.0 1.1 1.3 0.9 0.6 1.7 1.5 1.0 0.3 1.2 1.4 0.6 1.0 1.2 0.9 1.3 0.9 1.9 1.6 0.8 5.0 0.3

The table clearly shows that in the post-crisis period the probability of finding a job if unemployed falls by more than 6% points, whereas the probability of loosing a job if employed raises by 1.2% points.7 Table 11.5 shows the empirical estimates of the transition probabilities. They are based on a multinomial logit model, with standard error corrected for the presence of heteroscedasticity. The initial conditions are being employed, unemployed or inactive one year before. The probability that the individual i at time t - 1 was in the state S and the same individual at time t is in the state s, defined as: pi;S;s ; for S, s = [1, 2, 3], is given by: 0

expðbS;s X i Þ

pi;S;s ¼ P3

j¼1

0

expðbS;j X i Þ

where X is a vector of variables including age, gender, citizenship, education, area of residence, family type, as well as the interaction of these variables with year 2009 (that, by construction is the ‘‘post-crisis period’’). 7

To avoid major inflows into inactivity, we excluded all the individuals aged more than 55.

228 Table 11.3 Frequency of employees in firms under stress

M. Lilla and S. Staffolani

Males Females Immigrants Italian citizens 15–24 25–34 35–44 45–54 55–64 Primary Lower secondary Secondary Tertiary North–West North–East Center South Islands Singles Couples without children Couples with children Single parents with children Others families or NA

2008

2009

Diff

2.0 1.8 2.1 1.9 1.0 1.4 2.0 2.1 2.8 2.8 2.0 1.4 2.7 2.0 1.8 1.9 1.9 2.1 2.6 2.1 2.6 3.5 4.4

3.9 2.4 3.8 3.3 2.8 3.0 3.4 3.6 3.4 4.6 4.2 2.6 2.5 3.7 3.8 2.9 3.2 2.2 3.5 3.5 3.6 5.4 6.0

1.9 0.6 1.7 1.4 1.8 1.6 1.3 1.5 0.6 1.8 2.2 1.2 -0.2 1.7 2.0 0.9 1.4 0.1 0.9 1.3 0.9 1.9 1.6

Parameters for the possible transitions are computed with the maximum likelihood method, while relative risk ratios (RRR)8 are shown in Table 11.5. The risk of being unemployed relative to being employed one year earlier or loosing their jobs (column 1) is higher for immigrants, it decreases with age and education and is lower for men and for people from the North as well as for individuals living in ‘Single parents with children’ family type. When we look at the changes due to the crisis—measured as the effect of the independent variables interacted with the dummy for the year 2009—we found that there has only been an increase in the relative risk for the North–East. With respect to the transitions from employment to inactivity (column 2), the relative risk decreases with age and education, is lower for men and higher for people living in the North–East and in the South and belonging to ‘couples’ family types. For this transition we measured an increase in the relative risk of males in 2009. With respect to the transitions from unemployment, we found (column 3) that middle and old aged faced major problems in finding new jobs as well as women, 8 RRR are equal to the ratio between two different relative risks, both calculated with respect to the base category for a marginal increase in the independent variable; as an example, the RRR for the male dummy in the first column of Table 11.5 (base category: employed, category of interest: PðY i ¼UnempljX¼MaleÞ. unemployed) should be written as: PðY i ¼EmpljX¼MaleÞ PðY i ¼UnempljX¼FemaleÞ.

PðY i ¼EmpljX¼FemaleÞ

11

Labour Market Transitions During the Financial Crisis in Italy

229

Table 11.4 Transition matrix (self-declared employment condition, percentage) Year 2008 Year 2007

Employed Unemployed Inactive Total

Employed

Unemployed

Inactive

Total

95.43 25.96 3.21 56.13

2.40 72.20 2.71 8.98

2.17 1.84 94.07 34.89

100.00 100.00 100.00 100.00

2.36 1.61 95.02 35.56

100.00 100.00 100.00 100.00

Year 2009 Year 2008

Employed Unemployed Inactive Total

93.97 19.70 2.29 53.61

3.67 78.69 2.69 10.83

Italian citizens, and individuals with the lowest educational level, coming from the Center, South and Islands areas, with no particular differences between family groups. After the crisis, the situation worsened for the middle-aged individuals, women and for individuals living in the Northern regions. There are few significant differences between groups coming from unemployment to inactivity (column 4). In particular, old aged individuals and the youngest ones (less significantly) show higher transition probabilities as well as women and individuals from couples with children. In the South and the Islands this probability is lower. The transition probabilities from inactivity to employment (column 5) are higher for the young and middle-aged individuals, men, more educated individuals, living in the North and are lower for those living in families with children. The post-crisis period shows a reduction in transitions for men and an increase for individuals in ‘Single parents with children’ family type. For the transitions from inactivity to unemployment (column 6), we found a minor relative risk ratio for old aged individuals and for those living in the North–East, while a higher one was measured for men, more educated, living in the Islands and in ‘Single parents with children’ family (even if less significant).

11.3 Transitions and ‘‘Ecological Inference’’ The ‘‘ecological inference’’ analysis (King 1997; King et al. 2004) has been used in recent years to obtain information on ‘‘group’’ behaviour when only aggregate behaviour is known.9

9

This methodology has been widely used by political science researchers, particularly in the analysis of vote flows between parties in different elections.

230

M. Lilla and S. Staffolani

Table 11.5 Multinomial logit estimates of transition probabilities, relative risk ratios (1) (2) (3) (4) (5) (6) E?U E?I U?E U?I I?E I?U Year Ref: age 15–24 Age 25/34 Age 35/44 Age 45/54 2009: Age 25/34 2009: Age 35/44 2009: Age 45/54 Men 2009: Men Migrant 2009: Migrant Ref: primary or less Junior secondary Secondary Tertiary 2009: Junior secondary 2009: Secondary 2009: Tertiary Ref: North–West North–East Center South Island 2009: North–East 2009: Center 2009: South 2009: Island Ref: Single Couples without children Couples with children Single parents with ch. Other or N.A: 2009: Couples without ch. 2009: Couples with ch. 2009: Single parents with ch, 2009: Other or N.A: Constant N r2_p Wald chi2 *

10%,

**

5%,

***

1.19

0.95

0.55**

0.56

0.53

0.84

0.59*** 0.35*** 0.24*** 0.92 1.03 1.06 0.68*** 1.13 1.51*** 1.13

0.45*** 0.27*** 0.27*** 1.17 1.18 1.19 0.29*** 1.26** 1.19 1.32

0.98 0.80*** 0.61*** 1.01 1.24** 1.21 1.53*** 0.83** 1.65*** 0.84

1.31* 1.28 1.61*** 0.74 0.90 1.03 0.31*** 0.94 0.69 0.72

2.21*** 1.79*** 0.87 0.91 0.85 1.08 1.84*** 0.77** 0.91 1.13

1.07 1.02 0.38*** 1.25* 1.02 1.35* 1.41*** 0.87 1.17 0.67**

0.64*** 0.40*** 0.36*** 1.07 1.07 0.88

0.51*** 0.43*** 0.36*** 1.24 0.86 0.84

1.29** 1.75*** 1.81*** 0.92 0.91 1.14

0.97 1.09 0.90 1.20 1.28 1.81

1.94*** 3.49*** 8.90*** 1.08 1.22 1.16

1.00 2.12*** 5.75*** 1.15 1.10 0.93

0.67*** 1.26*** 2.09*** 2.51*** 1.35** 0.95 0.96 0.89

1.38*** 0.97 1.48*** 1.27 0.91 0.92 0.80 0.78

1.44*** 0.80** 0.33*** 0.34*** 0.77* 1.21 1.56*** 1.33**

1.12 1.21 0.61*** 0.42*** 0.90 0.81 1.16 1.22

1.19** 0.80** 0.50*** 0.45*** 1.05 1.05 0.87 0.87

0.63*** 1.00 1.12 1.46*** 1.30 0.89 0.89 0.78*

0.87 0.92 1.30** 1.14 1.16 0.99 0.98 1.12 -2.14*** 93858 0.0523 2440.97

1.67** 1.97*** 1.24 1.67** 1.28 1.08 1.11 1.32 -2.69*** 93858 0.0523 2440.97

1.04 1.07 0.99 0.83 1.32 1.25 1.09 1.41* -0.81*** 17153 0.0574 1329.21

1.65 2.13** 1.25 1.54 2.45* 1.53 1.77 1.67 -2.67*** 17153 0.0574 1329.21

0.80 0.68** 0.62** 0.89 0.83 1.45 1.85** 1.58 -3.84*** 45534 0.0756 2621.18

1.16 1.10 1.54* 1.60* 1.24 1.25 1.05 1.33 -3.72*** 45534 0.0756 2621.18

1%, significant levels; E employed, U unemployed, I inactive

11

Labour Market Transitions During the Financial Crisis in Italy

231

In order to introduce the methodology, consider a traditional problem faced by ecological inference methods, represented in the following table [from King (1997), p-14]:

Black White

Democrat

Republican

No vote

? ? 130

? ? 92

? ? 483

221 484 705

Suppose that researchers are interested in evaluating the Black and White vote behaviour, but data are available only for the composition of the population (row total) and for the total vote obtained (or no vote) by the political parties. Depending on other data eventually available, scholars have proposed different methods to estimate the unknown frequencies in the table. For instance, vote data and population composition could be available at a lower level of aggregation than the one presented in the above table—say for many districts—King ( 1997) proposes a new methodology to estimate the unknown frequencies when information on vote behaviour in different ‘‘districts’’ is available, highlighting the importance of the district dimension (the grouping problem). In our case, we would like to estimate the transition probabilities among different states, defined by Z = {1; 2; …n}, by using each period marginal frequencies.10 We exemplify this methodology by assuming that each individual at time t can be employed, unemployed or out of the labour force, so that Z = {1; 2; 3}. The observed marginal probabilities, ~ pg;t ; in each group g, and the unobserved transition probabilities are described in Table 11.6. The transition probabilities lzi zj e

kzi zj

are unknown. Nevertheless, we are

dealing with a transition matrix repeated for different periods, and therefore the marginal probabilities at time t are dependent on the ones at time t - 1, according to the relationships presented in Eq. 11.1: 8 1 > p1g;t1 ð1 l12 l13 Þ þ ~ p2g;t1 k21 þ ~p3g;t1 k31 < ~pg;t ¼ ~ ~p2g;t ¼ ~ p1g;t1 l12 þ ~ p2g;t1 ð1 k21 l23 Þ þ ~p3g;t1 k32 ð11:1Þ > : ~p3 ¼ ~ 1 2 3 ~ ~ p l þ p l þ p ð1 k k Þ 31 32 13 23 g;t g;t1 g;t1 g;t1 We define sub-groups of the whole population as ‘‘district’’. These sub-groups are identified by considering individual characteristics as gender, age and so on.

10

We remind that the database ISTAT-RCFL in our hands has the structure of a repeated crosssection dataset, because the identifier for the same individual across waves is different.

232

M. Lilla and S. Staffolani

Table 11.6 Three states transition matrix Period t Period (t - 1)

1

2

3

1

1 l12 l13

l12

l13

~p1g;t1

2

k21

1 k21 l23

l23

~p2g;t1

3

k31

k32

1 k31 k32

~p3g;t1

~ p1g;t

~ p2g;t

~p3g;t

For each district the ‘‘true’’ values of lzi zj e

kzi zj

must be between 0 and 1.

The modal values of the transition probabilities for the whole population are estimated using a multivariate normal distribution, based, for each district, on Eq. 11.1. The analysis has therefore been implemented by distinguishing among the different population groups (the so called ‘‘regionalisation’’).11 With respect to the previous section, a first difference consists in the estimation of n(n - 1) transition probabilities for each period, instead of individual transition probabilities. Nevertheless, by modifying the ‘‘regionalisation’’ procedure, we can estimate the transition probabilities for whatever sub-group of the population. The reading of the results of our analysis must consider a precautionary advise: given that transitions (and permanencies) in a state depend on the number of individuals that change state inside the group (so that they depend strongly on ‘‘group’’ definition), the interest should concern the dynamics over time of the different groups more than the absolute value of transitions.12 Among the various analyses that could be done by using the ecological inference methodology, we propose: • The overall transition probabilities for five employment conditions (Z = {1; 2; 3; 4; 5}). • The subgroups transition probabilities which refers to five different type of households for three employment conditions (Z = {1; 2; 3}). With respect to the first point, individuals have been grouped in the following employment conditions13: (1) self-employed; (2) permanent employees;

11

For a two-state graphical example of the described methodology, see King (1997). Our analysis has been carried out using the plugin Ecol on the program Stata. 12 Obviously, the modal value of the change of state inside a given group is inversely related to the dimension of the groups. 13 The analysis is based both on the ISTAT official definition of employment condition and on the perceived employment condition.

11

Labour Market Transitions During the Financial Crisis in Italy

233

Fig. 11.2 Estimated probabilities of transition and permanency by employment state (percentage)

(3) temporary employees; (4) unemployed; and (5) out of the labour force individuals (inactive). The ‘‘districts’’ come from the combination of the following individual characteristics14:

14

In order to consider the sample significance and consistently with the aggregate figures from ISTAT, data were rounded to thousands.

234

M. Lilla and S. Staffolani

• Age, five groups: 15–24, 25–34, 35–44, 45–54, 55–64; • Gender. • Residence: North–West, North–East, center, south, islands. To estimate the population in each ‘‘district’’ for each Zi ; we used sample weights. Results are presented graphically in Fig. 11.2. The left panels are based on the official employment condition, the right ones on the perceived status. These two measures usually do not give very different results, with the exceptions of the permanency in the state of unemployment and the transition from unemployment toward temporary employment, respectively strongly higher and lower when measured on the perceived status with respect to the official one. The economic crisis, at least from the fourth quarter 2008 to the second quarter 2009, seems to have increased the flows between permanent employment and selfemployment, especially when status is measured by the ISTAT official definition. Transitions toward unemployment have increased at the end of 2008-beginning 2009 but not later, whereas the probabilities of permanency in unemployment and in the temporary employment have increased in the late 2009. In our second approach, in order to quantify the ‘‘vulnerability’’ of households to the financial crisis, we evaluate the individual transition probabilities considering the type of family the individual belongs to, among the following type: • • • • •

Singles. Couples with children. Couples without children. Single parent with children. Other types or not available.15

The dimensions of our ‘‘districts’’ are the same as for Fig. 11.2 (age classes, gender, regions), but the employment conditions we are interested in here are only three: employed, unemployed and out of the labour force.16 Figure 11.3 shows, in the two graphs at the top, the results in terms of flows (from employment to unemployment and vice versa) and, in the two graphs at the bottom, the permanence probabilities in employment and unemployment. All the graphs are based on the official ISTAT definition of the employment condition. Individuals belonging to ‘‘Singles’’ and ‘‘Couples without children’’ family types show a higher propensity to move from employment to unemployment and vice versa, and a lower probability to remain in the unemployment trap. These kinds of households are surely the most ‘‘dynamics’’. Conversely, individuals belonging to ‘‘Couples with children’’ are the most ‘‘statics’’, because they show a higher propensity to remain in their status: the 15

ISTAT allows the respondent to refuse the identification of the family to which he/she belongs to. Less than 7% of surveyed individuals choose this option. 16 By doing so, we preserve a sufficient reliable size for the population in each condition for each group.

11

Labour Market Transitions During the Financial Crisis in Italy

235

Fig. 11.3 Estimated probabilities of transition and permanency in employment condition by type of households (percentage)

probability of permanency is higher, both in the status of employment and in the one of unemployment, the probability of transition between states is the lowest. The economic crisis seems to have:

236

M. Lilla and S. Staffolani

• Reduced the employment opportunities for individuals in ‘‘Couples without children’’ and ‘‘Single parent with children’’ family types, because they both show a reduction in the probability of permanence in the employment condition and an increase in the probability of loosing a job. Even if an increase in the probability of leaving unemployment is displayed for them in the top-right panel of Fig. 11.3, it seems that they have been mostly damaged by the economic crisis. • Not worsened the employment opportunities for individuals in ‘‘couples with children’’ and of ‘‘singles’’. Therefore, it seems that the households composed by ‘‘Singles’’, and ‘‘Couples with children’’ suffered less the economic crisis. The former, probably thanks to their ‘‘dynamic’’ behavior in term of job instability (temporary jobs), had difficulties even before the economic crisis, so that they resulted less vulnerable to the crisis in our analyses. The latter, probably thanks to the usage of social shock absorbers as the CIG, that applies to people employed on a permanent basis, usually old aged.

11.4 Conclusions Aiming at investigating the effects of the crisis in Italy in terms of individual transitions mainly between employment and unemployment, we had to face the unavailability of up-to-date longitudinal micro data on the Italian labour market. As a consequence we tried to exploit the section of the questionnaire of the Labour Force Survey on backward employment condition. In this case, we have been forced to limit the analyses on the perceived employment condition and to look at yearly transitions. To relax this constraint, we had to employ a different strategy on the aggregate data estimated from the survey, in order to recover the flows between employment and unemployment for the official employment condition. In such a way, we have been able to estimate quarterly transitions. The period that we analysed has been characterized by a combination of increasing redundancies with decreasing job offers. Our results obviously vary as long as they differ in the definition of the object and in time reference. Notwithstanding, we confirm the findings about the bad performances for the young and we find that the different segments of the labour market seem to converge as a consequence of the crisis. Men, immigrants and North regions faced the major risks to fall into unemployment, and women, middle-aged and northern individuals had major troubles in finding a job after the crisis. With respect to the family types we considered, the most affected individuals belong to ‘‘Couples without children’’ and to ‘‘Single parent without children’’. They experienced severe reduction in the employment permanence along with an increase in the transitions to unemployment.

11

Labour Market Transitions During the Financial Crisis in Italy

237

References Arpaia A, Curci N (2010) EU labour market behaviour during the great recession, MPRA, Paper no. 22,393 Berton B, Richiardi M, Sacchi S (2009) Flex-insecurity. Perché¨ in Italia la flessibilità diventa precarietà. BOLOGNA: Il Mulino Boysen-Hogrefe J, Groll D (2010) The German labour market miracle. Nat Inst Econ Rev 214(1):38–50 19 Oct 2010 Choudhry MT, Marelli E, Signorelli E (2010) The impact of financial crises on youth unemployment rate, Quaderni del Dipartimento di Economia, Finanza e Statistica 79/2010, Università di Perugia, Dipartimento Economia, Finanza e Statistica Discenza A, Lucarelli C (2009) Dati longitudinali a 12 mesi di distanza–aspetti metodologici, ISTAT Nota Metodologica allegata alla diffusione dei files MFR, Roma, dicembre ISTAT (2010a) La mobilitÃ nel mercato del lavoro: principali risultati del periodo 2004–2008, Approfondimenti, 1 Feb 2010 ISTAT (2010b) Rapporto annuale 2009 King G (1997) A solution to the ecological inference problem: reconstructing individual behaviour from aggregate data. Princeton University Press, NJ King G, Rosen O, e Tanner MA (2004) Ecological inference, new methodological strategies. Cambridge University Press, Cambridge, MA Mossfeldt M, Österholm P (2010) The persistent labour-market effects of the financial crisis, Working paper number 117, published by the National Institute of Economic Research (NIER), March 2010

Chapter 12

Labour-Market Dynamics and Unemployment: New Scenarios in the Italian Economy Carlo Lucarelli and Chiara Mussida

Abstract The availability of new longitudinal data on the Italian labour market for the time period 2004–2009 allows us to depict new scenarios for both labour mobility and unemployment, the latter having increased in relevance since the beginning of the current economic downturn. To achieve an exhaustive picture of the Italian labour market we propose investigation in two directions: we provide both a descriptive analysis and an econometric investigation by estimating multinomial logit models. We find that as far as our analyses go, possessing some particular characteristics reduces the negative impact of the economic downturn. More precisely, individuals with work experience and high educational qualifications have a higher probability of finding a job. On the other hand, older workers with temporary contracts (and also with permanent contracts) are more likely to become unemployed. Keywords Labour mobility JEL Codes C25

Unemployment Multinomial models

C40 J60 J64

C. Lucarelli (&) Ufficio Formazione e Lavoro, ISTAT, Rome, Italy e-mail: [email protected] C. Mussida Dipartimento di Scienze Economiche e Sociali, Università Cattolica del Sacro Cuore, Milan, Italy e-mail: [email protected]

G. Parodi and D. Sciulli (eds.), Social Exclusion, AIEL Series in Labour Economics, DOI: 10.1007/978-3-7908-2772-9_12, Ó Springer-Verlag Berlin Heidelberg 2012

239

240

C. Lucarelli and C. Mussida

12.1 Introduction Analysis of the latest longitudinal data for the Italian labour market is essential for achieving a complete picture of new scenarios involving the significant economic phenomena of labour mobility and unemployment. Changes introduced in 2004 in the definitions and design of the Italian Labour Force Survey (LFS) have resulted in greater detail and precision, permitting improved in-depth investigations of the Italian labour market. Firstly, the increased frequency of the interviews, the survey now being continuous, makes it possible to obtain a more precise and accurate picture of dynamics/transitions in the Italian labour market and therefore of the overall degree of mobility. Secondly, the abandoning of the use of self-perceived criteria for defining individuals’ economic status ensures more precise and objective observation of this factor, consistent with the International Labour Organisation (ILO). This study aims to analyse labour mobility in the Italian labour market with special focus on unemployment, which assumes quite relevant proportions in our country. In particular, the data employed, which refer to the period 2004–2009, allow us to assess the impacts of the current economic downturn on the labour market. The improvements in terms of detail and precision resulting from the 2004 renewal of the National Institute of Statistics (ISTAT) LFS, together with the freshness of the data employed, guarantee that all relevant phenomena are taken into account, resulting in a thorough and up-to-date description of the Italian economy. There is much international literature on labour mobility and unemployment—see Pissarides and Wadsworth (1989); Bentivogli and Pagano (1999)—and these topics have also been extensively studied by international institutions (e.g. Comission 2009). We analyse these issues for Italy. To our knowledge, no country-specific study on labour-market dynamics and unemployment has been carried out recently for Italy. Our paper is organised as follows: Sect. 12.2 describes the characteristics and structural details of the new Italian LFS; Sect. 12.3 illustrates the descriptive analysis, with Sect. 12.3.1 providing an in-depth analysis of the characteristics of labour mobility through the period analysed (2004–2009), while Sect. 12.3.2 focuses instead on unemployment, highlighting the characteristics of the unemployed especially at the height of the economic downturn (2008–2009). Section 12.4 illustrates the econometric analysis, Sect. 12.4.1 describing the econometric approach used to estimate the determinants of labour market transitions and Sect. 12.4.2 providing a description of the variables employed in the econometric analysis, while Sect. 12.4.3 discusses the estimation results and outlines the new scenarios in the Italian labour market. Section 12.5 concludes.

12

Labour-Market Dynamics and Unemployment

241

12.2 The Continuous LFS 12.2.1 The New Survey The continuous Italian LFS conducted by ISTAT is the main source of statistical documentation on the Italian labour market. Its main purpose is to produce official estimates of the numbers of employed and unemployed people for Italy as a whole and for its most important administrative subdivisions. These estimates are employed both for cyclical and structural analysis. The survey has been conducted each year since 1959 and it has been changed many times. The most recent changes in the definitions and design of the survey occurred in 2004.1 This renewal, primarily dictated by the requirement to adapt the survey to new EU standards, was also intended to respond to the need for increased knowledge and improved survey quality. Each year, the Survey collects information on almost 280,000 households in 1,246 Italian municipalities for a total of 700,000 individuals (representing 1.2% of the overall Italian population). The reference sample of the survey consists of all household members officially resident in Italy, even if temporarily abroad. Families registered as resident in Italy who habitually live abroad and the permanent members of collective facilities (hospices, children’s homes, religious institutions, barracks, etc.) are excluded. The survey unit is that of the family, defined as a group of cohabiting people connected by marriage, kinship, affinity, adoption, guardianship or affection whose habitual domicile, but not necessarily their official residence, is in the same municipality. The restructuring of the survey has led to changes in the definitions of an employed person and a person seeking employment. While based on the principles set out by the International Labour Organisation (ILO), the changes in the classification criteria of individuals based on occupational status were dictated by the need to achieve complete harmonization with the requirements of EU regulations and full comparability with the estimates provided by other countries, such as those in the OECD area. This makes it possible, all other factors being equal, to make a comparison with the other EU countries, the U.S. and Japan with respect to the main labour market indicators such as employment and unemployment rates. The classification criteria are applied according to a hierarchy: employed people are identified first, then those classified as non-employed and the

1 The changes introduced in 2004 can be summarised as follows: increased frequency of the interviews, new criteria for classifying individual economic status consistent with the International Labour Organisation’s (ILO) definitions, wider-ranging information content, the use of a professional surveying network involving more than 300 surveyors, the use of computer-assisted instead of traditional techniques, the implementation of a sophisticated IT system supporting the carrying out of the survey. For a more detailed discussion of the characteristics of the Italian LFS (see Gazzelloni 2006; ISTAT 2009).

242

C. Lucarelli and C. Mussida

unemployed seeking employment. Lastly, among those people not classified as employed or unemployed, the inactive2 are identified. With regard to employment, the main changes concern the emphasis placed on objective information on whether a person has worked at least an hour during the reference week, and the abandoning of self-perception as the main criterion for distinguishing between the employed and the unemployed. Criteria have also been established to determine whether an individual is absent from work, the duration of absence and the wage received during absence. The main innovation in the identification of people seeking employment is the introduction of an upper age limit of 74 years. Given this limit, the unemployed aged 75 years or over are automatically classified as inactive. A three-month time limit for the possible start-up of a future economic activity has also been included.

12.2.2 Sample Structure and Longitudinal Component The sampling design of the survey is composed of two stages, with a stratification of the unit at the first stage; the first stage units are municipalities, while the second stage comprises households. Each household member is interviewed. The main difference between the two stages is that while for families a 2-2-2 rotation scheme is applied, the municipalities surveyed do not change over time. More specifically, a household was interviewed for two consecutive surveys and, after being excluded from the sample for two quarters, was interviewed for another two consecutive quarters. This is defined as a (2-2-2) rotation scheme.3 This rotation system makes it possible to maintain half the sample unchanged in two consecutive quarters and in quarters one year apart. In other words, the scheme implies a 50% overlapping of the theoretical sample to a quarter of the distance, a 25% overlapping to three quarters, a 50% to four quarters, and a 25% to five quarters. The rules for updating the sample, from which the possibility of reconstructing the longitudinal component derives, are mainly aimed at obtaining more stable estimates of the variations at 3 and 12 months’ time-distance (net flows) and to meet the accuracy constraints set by EUROSTAT, while they are not directly related to the production of transition matrices (gross flows). The latter therefore represent a ‘‘by-product’’ of the LFS. Further, the survey is not a rotating panel in the strict sense and is not referred to the overall population. It should be noted that an individual interviewed the first time in one of the sample municipalities is not re-interviewed if, in the period between the first and subsequent interview, he or she has changed residence or moved abroad.

2

For a description of the definitions ‘employed’, ‘unemployed’ and ‘inactive’, see ISTAT (2010a). 3 For in-depth details on sampling design, see Discenza and Lucarelli (2009).

12

Labour-Market Dynamics and Unemployment

243

The longitudinal component, indeed, includes only residents who did not move beyond the municipality borders in the 12 months following the first interview. Within a defined time frame, the longitudinal component does not represent the entire population but only the residents in the same municipality at the beginning and at the end of the period under consideration. This part of the population is thereafter called the ‘‘longitudinal population’’. However, the low level of mobility of the population across the country implies that only a small part of the overall population is not taken into account (only 2.2% of the individuals changed municipality of residence in a one-year time period) and it should be noted that these individuals tend to behave differently on the labour market with respect to the longitudinal population. To better understand the structure of the longitudinal data employed in this work, it is essential to stress their relationship with the cross-sectional estimates that are issued periodically by the ISTAT. Cross-sectional estimates permit the determining of possible flows between labour market states/conditions, (typically by adopting a three-state representation of the labour market, ‘employed’, ‘unemployed’ and ‘inactive’) of the overall population aged 15 and over, i.e. those individuals involved into the labour market (active population), in a given time period (quarter or year). The cross-sectional samples of the LFS only permit estimates of the distribution by labour market state (condition) of the population at the beginning and end of the period. As mentioned, part of the initial population may change residence, emigrate or die. For this part, which is not included in the longitudinal component, the condition at the beginning of the period is known. On the other hand, for the part of the population which has registered at the register office or reached 15 years of age during the observed period but is not yet contained in the longitudinal component, the condition at the end of the period is known. It thus follows that the longitudinal component based on the LFS data describes only flows of longitudinal population between the labour market states. These flows are all internal to the transition matrix. In the light of these considerations, the longitudinal data can be analysed and described by employing the following reference scheme: Figure 12.1 shows a transition matrix (identified as A) which contains the estimates by labour market condition at the beginning and at the end of the period, the flows of the longitudinal population, two vectors (C and E) on which the estimated stocks of the initial and final population by condition are placed, and two additional vectors required to adjust the longitudinal population with the total population. The first vector (B) reports the distribution of deceased people and those cancelled from register office records according to their condition at the beginning of the period. The second vector (D) displays the distribution of young people who reached 15 years of age in the period and people who registered at the register office classified by their condition at the end of the period. Within the transition matrix, the rows identify the outflows from the conditions indicated at the side toward those at the top; the columns identify the inflows into

244

C. Lucarelli and C. Mussida

Fig. 12.1 Matrix scheme: stocks and flows of the overall population

the conditions indicated at the top from the conditions indicated at the side; the main diagonal identifies the individuals that remain in the same condition at the beginning and end of the period (permanence in the initial state or condition).

12.3 Descriptive Analysis 12.3.1 Permanence and Labour Market Transitions for the 2004–2009 period The longitudinal data employed in this study refer to the first quarter of each year from 2004 (beginning of the continuous survey) to 2009. Data for the 2004–2008 period have already been released in a final version by the ISTAT (Discenza et al. 2010) whilst for 2008–2009 the estimates are provisional. In this regard, Fig. 12.2 shows the transition matrix for the last period considered. After one year, more than 90% of the employed were in the same condition, 2.4% had moved to unemployment, whilst just over 6% had moved to the stock of inactivity. Permanence in this latter stock is also significantly high, whereas the group that showed higher dynamics—especially in terms of outflows—is that of people seeking employment. For this stock, indeed, the percentages of individuals remaining in the initial state and leaving the condition are more equilibrated.

12

Labour-Market Dynamics and Unemployment

245

Fig. 12.2 Labour market transition matrix, longitudinal population qtr1 2008–qtr1 2009. Thousands (*) Longitudinal population by labour market state at the beginning of the period. (**) Longitudinal population by labour market state at the end of the period. Transition Probabilities in italics. Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

The use of data for the entire period 2004–2009 highlights the deterioration in the labour market due to the economic downturn experienced by the Italian economy. Table 12.1 displays permanence rates and transitions by labour market condition (three-state representation) through the overall 5-year period examined (percentage values). It should be noted that the last columns of the table refer to the inactive population aged 15–64. The upper limit of 64 has been selected since the quota of people aged over 65 represents in almost all cases a condition of permanence in inactivity due to retirement, or absence of interest in working or finding a job. The longitudinal picture faithfully reflects the situation of the labour market for the period analysed. In the early years, in which employment continued to grow rapidly, the percentages of permanence in employment were high and the transition towards unemployment tended to decrease; transitions to the non-labour force were mainly due to retirement. Unemployment in this initial period continued to decline and this is reflected, albeit with fluctuations, by the decrease in permanence in this state (from 32% in qtr1 2004–qtr1 2005, to 29.3% in qtr1 2006–qtr1 2007). Transitions towards employment remained at around 30% with a peak in the period qtr1 2007–qtr1 2008, while larger flows are observed from unemployment to inactivity. This is especially true for the transitions towards the so-called grey inactivity zone. The grey zone refers to the group of individuals who, despite not showing the individual characteristics required to enter the stock of unemployment, show a certain degree of attachment to the labour market. Among these, we find individuals involved in ‘inactive search’ for work (they are searching, but they have not carried out any search actions in the last 4 weeks), those seeking employment but not immediately available to start working within the next 2 weeks, and those not seeking work but who would be willing to accept a job if it were offered to them. In this period we also observe a polarization effect for the outflows from unemployment to employment, which was expanding, and towards the grey zone, especially for those willing to work but who do not have professional

2.1 1.9 1.5 1.6 2.4

Unemployed 5.9 5.5 6.1 5.8 6.2

Inactive

Author’s calculations using longitudinal ISTAT 2004/2009 LFS * Provisional data

2005 2006 2007 2008 2009*

92.0 92.6 92.4 92.6 91.4

1 1 1 1 1

Employed

2004–qtr 2005–qtr 2006–qtr 2007–qtr 2008–qtr

qtr qtr qtr qtr qtr

1 1 1 1 1

Condition at time t ? 1 30.1 32.1 30.2 33.5 28.9

Employed 32.0 33.1 29.3 31.2 32.7

Unemployed 37.9 34.7 40.5 35.3 38.4

Inactive 6.4 7.6 6.9 8.3 6.8

Employed

4.8 4.9 3.7 5.2 4.9

Unemployed

88.8 87.6 89.4 86.6 88.3

Inactive

Table 12.1 Permanence rates and transitions (12 months time-distance) by labour market condition through the period 2004–2009. Percentage values Condition at time t Employed Unemployed Inactive (15–64 age brackets)

246 C. Lucarelli and C. Mussida

12

Labour-Market Dynamics and Unemployment

247

characteristics making them attractive for the labour market; in this respect the longitudinal data is in line with the results observed for the cross-sectional data. In the last 2 years the overall situation has changed. In the first quarter of 2008, after years of continuous decline, the number of unemployed started to rise again and this translates from a longitudinal perspective into the growth of permanence in the unemployment state in the period qtr1 2007–qtr1 2008, and into a reduction of transitions towards inactivity; in contrast transitions towards employment accelerated, and the total employment stock continued to grow throughout 2008 given that there was also an increase in the inflows (to employment) from the nonlabour force. At this stage, employment was still unaffected by the economic downturn and attracted many individuals from the inactive segment. However, given that not all these inactive individuals were able to find work, there was an increase in the outflows to unemployment (5.2% in qtr1 2007–qtr1 2008 compared to 3.7% in the previous year). Although provisional, the last point of the time series provides the situation of the labour market at the trough of the economic cycle. Permanence in the state of employment underwent a slight reduction while transitions toward the other two conditions rose; the probability of the unemployed finding a job after 12 months dropped considerably, and this is also the case for the inactive population, whose permanence rates consequently increased.4

12.3.2 Focus on the Unemployed Permanence in the state of unemployment increased for both genders, but the higher growth rate is inferred for men (especially at the trough of the economic cycle). More insightful is the trend at local level. This shows that the probability of remaining unemployed in the north of the country grew significantly over the last 2 years examined; in 2009, the north-east in particular registered an increase of more than 8% points from 19.7% to over 28%, figures close to the levels for the north-west and centre of Italy. The area with the highest permanence levels is the south.

4 It should be borne in mind that this description refers to the dynamics internal to the longitudinal population which, however, is clearly affected by the latest trends observable from stock data on the labour market. Conversely, no conclusions can be drawn for individuals not belonging to the longitudinal population. As previously stated, these groups represent a small proportion of the overall population and also include a statistically insignificant quota of exits due to death. It is essential to make this distinction, as exits from the longitudinal population often conceal relocations to areas offering higher opportunities of finding a job for the unemployed, or to new jobs offering better conditions for the employed. These phenomena are extremely important but are beyond the scope of the panel-type analysis proposed in this study.

248

C. Lucarelli and C. Mussida

Fig. 12.3 Inflows into unemployment by contractual form. Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

Fig. 12.4 Outflows from unemployment by contractual form. Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

Approximately half of the workers who found themselves seeking employment after 12 months had previously held permanent employment contracts. The number of individuals working as self-employed was lower (and progressively shrinking), whereas workers under continuing temporary contracts increased, especially in the final period (Fig. 12.3). The jobless have seen their chances of finding open-end contract employment or becoming self-employed progressively reduced in recent years (Fig. 12.4). This does not mean that the economic downturn has not affected workers employed on permanent contracts, since the unemployment inflows from

12

Labour-Market Dynamics and Unemployment

249

Fig. 12.5 Unemployed: inflows from employment by age bracket. Percentage changes with respect to the previous period Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

permanent employment have increased as well, although to a lesser extent. The relevance of this impact is clarified by Fig. 12.5, which shows the inflows to unemployment from employment by age bracket. In the last period (qtr1 2008–qtr1 2009) the percentage change with respect to the previous period (12 months before) encompasses increasingly older workers (40–49 and 50 and over age brackets) who generally hold permanent employment contracts and who, as mentioned above, are also affected by the economic downturn. Nor are young workers exempt from the impacts of the downturn; indeed, young workers are strongly affected by it. A huge number of young people lose their jobs and move straight into inactivity instead of accessing/entering the state of unemployment. At the same time, opportunities for finding fixed-term employment or continuing contract work (often the same thing) have increased. Figures 12.6 and 12.7 illustrate that almost half of the new unemployment inflows coming from inactivity belong to the area furthest from the requirements of the labour market, whilst those making the transition from unemployment to inactivity enter the so-called grey zone of inactivity. In the last year, however, the number of those who have moved to the least attractive area of the labour market have increased. Given the extent of the economic downturn, those who in normal circumstances would have become inactive but would have maintained a degree of attachment to the labour market now prefer to devote themselves to other activities as they wait for the situation to improve. The downturn has resulted in more noticeable changes in the prospects of the unemployed with work experience compared to those seeking employment for the first time (Fig. 12.8). During the trough of the economic cycle (2008–2009), while first-time job seekers had the same probability of permanence and transition to other labour market states as in the previous year, those with work experience had

250

C. Lucarelli and C. Mussida

Fig. 12.6 Inflows into unemployment from grey zone and ‘‘hard’’ inactivity. Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

Fig. 12.7 Outflows from unemployment to grey zone and ‘‘hard’’ inactivity Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

much lower prospects of finding employment after 12 months in the period qtr1 2008–qtr1 2009 compared to the previous year (37.2% as opposed 31.3%). The members of the latter group face an increase in both the permanence rate in the state of unemployment (of almost 3% points) and in the rate of transition (of over 3% points) to that state of inactivity known as ‘‘hard’’, i.e. furthest from meeting the requirements of the labour market. It should be noted that although first-time job seekers were less likely to find themselves in the same condition after 12 months in the period 2008–2009, this translates into an increase in the number of the non-labour force rather than into better opportunities of finding a job. A sizeable component of the inactivity stock is composed of young people holding an educational qualification. Most likely, this group prefers to pursue

12

Labour-Market Dynamics and Unemployment

251

Fig. 12.8 Permanence rates and outflows from unemployment by work experience. Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

training and studying opportunities instead of immediately facing the increased difficulties of finding a job characterizing the Italian labour market.5 The probability of finding employment at 12 months’ time-distance is also strongly related to the duration of unemployment. Normally, those who have been seeking work for a short period (short-term unemployed) are more likely to find a job than the long-term unemployed, as the latter often do not have individual characteristics and skills that are attractive for the labour market, and therefore face higher permanence in the state of unemployment (prolonged unemployment duration). This phenomenon is observed during the period qtr1 2007–qtr1 2008 where the likelihood of finding a job for short-term job-seekers is almost double that for the long-term unemployed (42% with respect to 22%), while the permanence rate is about half (24% for the short-term unemployed compared to 40% for the long-term unemployed).6 The economic downturn disrupts this scenario (Fig. 12.9). In qtr1 2008–qtr1 2009, job opportunities for the short-term unemployed dropped sharply (-5% points) and permanence in the state of unemployment increased (+5% points). The same applies to those who declared they had been searching for employment for 7–12 months; they faced a rising probability of remaining unemployed even after 12 months (over +7% points) and a reduction in their chances of finding a job of over 11% points. These changes in unemployment permanence rates exert an impact on job search duration, short-term duration (less than 12 months) increasing by more than 5% points in the last point of the longitudinal series. However, this is not the result 5 For a detailed analysis of the attainment of educational qualifications, see ISTAT computations on data from the Ministry of Education, Universities and Research (MIUR) contained in ISTAT (2010b). 6 From the statistical point of view, the International Labour Organisation (ILO) classifies as long-term unemployed (LTU) those individuals who have been unemployed for 12 months or more.

252

C. Lucarelli and C. Mussida

Fig. 12.9 Permanence rates and transitions by unemployment duration. Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

Fig. 12.10 Unemployed by transition from short to long-term job search duration, 2007–2009. Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

of re-employment of the long-term unemployed, but rather derives from the significant increase in new inflows into the unemployment state. The spread permanence into unemployment also implies an extension of search time. Therefore, even if there is a reduced incidence of long-term unemployment duration (longterm job search), there is an increase in the transitions from short to long-term unemployment (Fig. 12.10). This is especially the case for men resident in the centre and north of Italy. This impact of the economic downturn thus results in an ‘‘unemployment trap’’, from which transition to inactivity is the most frequent exit route. Individuals in search of a job, indeed, increasingly move to the state of inactivity (from 35.3% in qtr1 2007–qtr1 2008, to 38.4% in qtr1 2008–qtr1 2009). All these elements lead to discouragement with respect to employment prospects within the labour supply: in 2009 a discouraged worker effect seems to be

12

Labour-Market Dynamics and Unemployment

253

Fig. 12.11 Permanence rates and outflows from unemployment by educational attainment level. Source: Author’s calculations using longitudinal ISTAT 2004/2009 LFS

predominant and this then leads to an increased inactivity stock. The long-term unemployed maintained low levels of transition towards employment, although there was a drop in numbers as more of the long-term unemployed moved to inactivity, both to the grey zone and to the area furthest from the labour market (permanent inactivity). In 2008–2009, therefore, there was a general shift towards a more critical situation: the unemployed who before the crisis had found few obstacles to finding a job after a short period of unemployment, experienced more difficulties, while the long-term unemployed, who were already heavily penalized in the past, turned away from the labour market in ever larger numbers to swell the various ranks of inactivity. A final important observation concerns the educational profile of the unemployed (Fig. 12.11). The worsening economic situation mostly affects the prospects of the unemployed with lower educational qualifications. Compared to the previous period, in qtr1 2008–qtr1 2009 a greater number of those who had only completed compulsory education or holding only a school diploma remained unemployed and had a lower likelihood of finding a job. Further, there is also an increase in the number of those who move directly to the most extreme condition of inactivity. Those holding a degree do not seem subject to this trend. The probability of them finding a job increased slightly in the last period despite the recession (about 2% points higher than the situation of the previous year) while a decrease of the same magnitude affected the flow into the grey zone of inactivity. Holding a degree or other higher educational qualification therefore still helps access to employment regardless of the economic downturn.7

7

For an in-depth investigation see ISTAT reports on MIUR data (ISTAT 2010b).

254

C. Lucarelli and C. Mussida

12.4 Microeconometric Analysis The purpose of this section is threefold. First, it describes the econometric model used to estimate the longitudinal data described in section for the 5-year period 2004–2009. The second part illustrates the variables employed in the analysis and explains their economic relevance as supported by literature on the labour market. Finally, emphasis is placed on the estimates obtained for the labour-market transitions, and we attempt to define where possible and with due caution an interpretative framework of labour mobility for the Italian labour market.

12.4.1 Econometric Model The literature emphasises that multi-state stochastic models provide a useful framework for the analysis of data from longitudinal studies when interest lies in the dynamic aspects of the process under investigation.8 When individuals are continuously observed over time, transitions between states are observed and parametric, nonparametric, and semiparametric methods may be used to investigate their behaviour (such as in Andersen et al. 1993). In contrast, when the subjects are seen at discrete time points—such as in panel data exact transition times are not observed and all that is known is the state occupied at each assessment of the related survey. Such data are often analysed using Markov Chains models.9 The features of the data employed in the present work, explained in depth in Sect. 12.4, allow us to use a Markov Chain approach. Estimated transition probabilities, such as those described in the previous section, are averages of heterogeneous individual transition probabilities that are likely to depend on individual characteristics as well as on the general conditions of the labour market. Let h ¼ 1. . .n be the indices for the hth individual in the sample; in this section we deal with the conditional individual transition probabilities: pij; t(h) ¼ Pr ¼ ðXt; h ¼ jjXt 1; h ¼ i; zt; hÞ;

ð12:1Þ

where Xt,h is the random variable describing the state of individual h at time t, while zt,h is a vector including individual-level covariates and economic indicators of the conditions of the labour market (these may not be the same for all individuals in the sample if, as will be emphasised below, regional-specific indicators are considered). Since we adopt a three-state representation of the labour market, it is logical to choose a Multinomial Logit model (MNL). This class of models

8

For a detailed investigation into the application of such models, see Cook et al. (2002). Aeschimann et al. (1999) explain and make use of a Markov chain approach to describe the evolution of labour market transition probabilities in the Swiss labour market. 9

12

Labour-Market Dynamics and Unemployment

255

extends ordinary logit regression from dichotomous to polychotomous dependent variables. We specify a separate model for each row of the transition matrix, i.e. we divide the sample into three sub-samples, according to their state in the labour market at the beginning of the reference period. For notational convenience we number the three states we consider from 0 to 2. The model for the transition probabilities can be written as follows: exp zth bl Pij;h ¼ P2 t l¼0 exp zh bl

ð12:2Þ

According to Theil normalisation, we set b0 = 0. Conventionally, we will assume permanence in the initial state as the baseline category. Model parameters are estimated using Maximum Likelihood. A detailed technical description of the Maximum Likelihood method in this context can be found in Gourieroux (1989) and in Cameron and Trivedi (2005). We consider only the transition from the beginning to the end of the observation period. The data structure, as explained above, does not make it possible to determine intermediate transitions.

12.4.2 Variables The analyses carried out in this paper refer to the variables described in the Appendix Table 12.6. The choice of the variables was driven both by specific econometric tests and preliminary checks, described in the following section, and by the relevance of the indicators which are widely emphasized in the literature and in the aforementioned descriptive statistics. As mentioned in the last section, we specify a separate model for each row of the transition matrix, and we estimate each model separately by assuming independence of the outflows from each of the three labour market states. The dependent variables utr, etr and ntr thus refer to the outflows from the states of unemployed, employed and inactive, respectively. Most of the variables listed in the descriptive table were estimated for all transitions, with some exceptions. The first independent variable, ‘man’, attempts to determine the impact of gender on labour market transitions. The relevance of gender is emphasized both in past literature, which analyzes aggregate data (e.g. Baussola 1985, 1988; Leoni 1984) and in more recent studies employing individual labour force data (ISTAT, for the decade 1993–2003), such as Schindler (2009); Trivellato et al. (1989). In addition, the statistics displayed in Table 12.2 stress the relevance of gender for the analysis of the current characteristics of the labour market. The individual age (which in this analysis refers to the working age with the upper limit of 74 years to reflect the new definition of ‘unemployed’ adopted in the current LFS) and age squared were both included in all estimates. The previously quoted studies, together with (among others) Bertola and Garibaldi (2002);

32.0 24.7 24.4 25.2 27.7 36.2

33.1 27.4 28.7 25.3 35.8 34.9

t1 05 t1 06

29.3 23.2 26.6 18.1 22.3 34.2

t1 06 t1 07 31.2 24.2 27.2 19.7 26.7 36.3

t1 07 t1 08 32.7 28.7 29.0 28.2 29.4 35.8

t1 08 t1 09 35.7 25.3 23.9 28.2 29.1 40.8

t1 04 t1 05

Author’s calculations using longitudinal ISTAT 2004/2009 LFS

Italy North North–West North–East Centre South

t1 04 t1 05 33.0 25.4 25.8 24.9 31.8 35.7

t1 05 t1 06 32.4 27.5 34.1 15.8 26.5 35.6

t1 06 t1 07 35.0 28.5 33.4 19.6 36.2 37.6

t1 07 t1 08

Table 12.2 Unemployed: permanence rates by geographical area and gender. Percentage values Maschi e femmine Maschi

37.6 30.5 29.5 32.5 26.8 43.2

t1 08 t1 09 28.9 24.2 24.8 23.4 26.7 31.7

t1 04 t1 05

Femmine

33.3 28.7 30.7 25.6 39.1 34.0

t1 05 t1 06

26.6 20.3 20.9 19.5 19.3 32.7

t1 06 t1 07

27.6 20.7 21.4 19.7 20.5 34.7

t1 07 t1 08

28.3 27.3 28.6 25.7 31.3 27.8

t1 08 t1 09

256 C. Lucarelli and C. Mussida

12

Labour-Market Dynamics and Unemployment

257

Picchio (2006), stress the impact of age on transitions in the labour market. Particular attention was also devoted to the issue of youth unemployment which, as previously stated, strongly characterises the Italian labour market.10 Other individual characteristics are those of citizenship and household size, the first to account for the increased weight of foreigners in the labour market, especially among the employed (ISTAT 2010b), the second so as to include the relevance of household in assessing probabilities of employment. The impact of education on labour market transitions is assessed by including specific dichotomous variables. The importance of educational qualifications is described in Sect. 12.3.2 of the present analysis and by ISTAT (2010b), both of which stress that holding a degree reduces the negative impact of the economic downturn. The structural variables, which were introduced so as to take into account the conditions in the labour market, refer to the regional unemployment rates (annual rates, ISTAT). It is important to emphasize that in addition to capturing the structural component of unemployment,11 the regional unemployment rate also summarizes the overall impact of the geographical component.12 This is another peculiarity of the Italian labour market, which is typically characterized by high geographical differentials and discrepancies, especially in terms of probability of permanence in and/or transition from the state of unemployment. This issue has also been widely discussed in the literature (among others Paggiaro 1999; Ricciardi 1991), and it is confirmed by the current analysis, as outlined in Sect. 12.3.2. The annual growth rate of regional GDP attempts to summarize the cyclical component. In relation exclusively to transitions out of the state of unemployment, information was included regarding both previous work experience—as this often increases the likelihood of finding employment (Fabrizi and Mussida 2009; ISTAT 2010b)—and duration of unemployment. The latter is a proxy of duration dependence, which describes the relationship between the probability of leaving unemployment and the duration of unemployment itself.

10

The issue of youth unemployment in Italy has been analyzed in the literature by employing data from different sources. Among these we cite Barbieri and Sestito (2008) who use the ISTAT LFS, Picchio (2008), who analyzes data from the SHIW, the Survey of Italian Households’ Income and Wealth conducted by the Bank of Italy, and Berton et al. (2008) who use data from INPS (National Institute of Social Security) administrative archives. 11 By ‘‘structural component of unemployment’’ we mean more permanent (or long-term) unemployment. For a definition of’structural component of unemployment’, see Jackman and Roper (1987). The idea of structural unemployment became popular in the 1980s (see, e.g. Sahin et al. 2010), when economists were trying to understand why unemployment was rising steadily in many European countries (including Italy). 12 Regional differentiation of the unemployment rate makes it possible to determine the impact of the geographical component on labour market transitions. The unemployment rate thus absorbs both geographical and structural components. This indicator in fact precludes the inclusion of other structural variables.

258 Table 12.3 Labour market transition matrix, outflows from unemployment, 2004–2009

C. Lucarelli and C. Mussida

Man Age Agesq Italian Famsize Experience Loweduc Compulsory Diploma Urate Dursearch Gdpgrowth Year 2005 Year 2006 Year 2007 Year 2008 Pseudo R2 Observations

UE

UN

0.293*** -0.017 0.001 0.058 0.042** 0.381*** -0.646*** -0.672*** -0.472*** -0.080*** -0.001*** 4.646* 0.102 0.006 -0.028 -0.181* 0.0468 11,352

-0.567*** -0.067*** 0.001*** 0.229** 0.016 -0.278*** 0.115 0.133 0.123 0.008 -0.001 4.503* 0.057 0.266*** 0.069 0.082 11,352

Source: Author’s calculations using longitudinal ISTAT 2004/ 2009 LFS All these variables are described in the Appendix, Table 12.6

Educational attainment title and type of work performed were included only for the outflows from employment. More precisely, a specific distinction was made between part-time and full-time work. This variable also makes it possible to assess the extent to which part-time work (which legislation over the last 15 years has attempted to encourage, especially for disadvantaged categories of the labour force, particularly women and the long-term unemployed) has reacted to the current economic downturn. The last group of variables consists of five yearly dummy variables which try to capture the behaviour of our estimates within the time period analysed, in order to indicate if there is any evidence of trend for the transition probabilities.

12.4.3 Estimates: New Scenarios in the Italian Labour Market The data described in Sect. 12.2.2 make it possible to determine the current situation in the Italian labour market and, for the last 2 years examined, i.e. 2008–2009, the impact of the economic downturn on the country. Microeconometric analysis also highlights the determinants of the dynamics shown in the transition matrix given in Fig. 12.2. Tables 12.3, 12.4, 12.5 display the estimates for the five-year period 2004–2009 from the state of unemployed, i.e.: unemployment to employment (UE) and to inactivity (UN); employed, i.e.: employment to unemployment (EU) and to

12

Labour-Market Dynamics and Unemployment

Table 12.4 Labour market transition matrix, outflows from employment, 2004–2009

Man Age Agesq Italian Famsize Loweduc Compulsory Diploma Urate Low_qual Fulltime Gdpgrowth Year 2005 Year 2006 Year 2007 Year 2008 Pseudo R2 Observations

259 EU

EN

-0.171*** -0.128*** 0.001*** -0.325 0.022 0.674*** 0.329*** 0.180*** 0.101*** 0.791*** -0.723*** 1.585 0.019 -0.221*** 0.124* 0.379*** 0.1263 138,182

-0.583*** -0.301*** 0.003*** 0.128* 0.052*** 0.849*** 0.383*** 0.224*** 0.064*** 0.521*** -0.891*** -1.983 -0.079** 0.122*** 0.117*** 0.101*** 138,182

Source: Author’s calculations using longitudinal ISTAT 2004/ 2009 LFS All these variables are described in the Appendix, Table 12.6

Table 12.5 Labour market transition matrix, outflows from inactivity, 2004–2009

Man Age Agesq Italian Famsize Loweduc Compulsory Diploma Urate Gdpgrowth Year 2005 Year 2006 Year 2007 Year 2008 Pseudo R2 Observations

NE

NU

0.905*** 0.243*** -0.003*** -0.022 -0.003 -1.226*** -1.112*** -0.760*** -0.036*** -0.925 0.204*** 0.039 0.167*** 0.013 0.1769 142,163

0.626*** 0.308*** -0.004*** -0.372*** -0.042*** -0.655*** -0.440*** -0.297*** 0.051*** 1.997 0.110** -0.211*** 0.196*** 0.043 142,163

Source: Author’s calculations using longitudinal ISTAT 2004/ 2009 LFS All these variables are described in the Appendix, Table 12.6

260

C. Lucarelli and C. Mussida

inactivity (EN); and inactive, i.e.: to employment (NE) and to unemployment (NU). The base category for results interpretation of every MNL is permanence in the state (UU, EE, and NN).13 Before the pooled estimates reported below were made, we carried out preliminary tests and analyses. Particular attention was devoted to the temporal dynamics of the transition rates (see also Tables 12.1, 12.2) and an initial estimation exercise was carried out separately for each of the 5 years examined. The temporal invariance of the impacts of selected explanatory variables for each multinomial logit model was then verified. The behaviour of estimated transitions within the time period analysed is summarized by the yearly time dummies included in each estimate. These estimates, after controlling for individual characteristics by assuming their effects as constant through time, and taking into account structural features of our labour market (unemployment rate, absorbing the structural component and GDP growth rate, representing the cyclical component), try and capture the presence of any trend in the transition probabilities. Nonetheless, in the present study the temporal trends are not referred to an excessively prolonged period. Table 12.3 shows many significant variables for the outflows from unemployment to employment, and, to a lesser extent, for those from unemployment to inactivity. The individual characteristics that facilitate successful exit from unemployment include gender, household size, previous work experience and holding a degree (or a high educational qualification). A relevant determinant for both the outflows is the unemployment rate. The latter, as stated above, captures the structural aspects of the Italian labour market and its geographical component. Unemployment in Italy, therefore, as stated in the descriptive analysis, is highly structural in nature. The cyclical component, which in the current analysis refers to the GDP growth rate, is less relevant.14 Further, employment opportunities are negatively affected by prolonged unemployment duration. This is also illustrated in Fig. 12.9. By looking at the effects of the explanatory variables on the transitions between unemployment and inactivity, we infer that the latter condition mainly characterizes women without previous work experience. Age also emerges as an important factor: it appears that the older unemployed have a lower likelihood of entering the Non-labour force pool. Transitions out of the state of unemployment do not seem to follow any trend. Men who are not young and who have high educational qualifications have a lower likelihood of leaving the state of employment (Table 12.4). Some interesting 13

A similar estimation exercise was conducted for the decade 1993–2003 (Fabrizi and Mussida 2009) on the longitudinal data of the former so-called quarterly survey adopted until 2003 by the ISTAT. 14 The cyclical and structural impact on labour market of last crisis is beyond the aim of this paper and need of a longer period after the downturn, due to the usual lagged and persistent impact of a crisis on labour market.

12

Labour-Market Dynamics and Unemployment

261

aspects emerge depending on the type of contract and educational qualification in the occupation performed. As regards the former, it appears that full-time work better withstands an economic downturn than part-time; full-time workers face a lower probability of exiting employment. The same effect is found for occupations that require high qualifications. This emphasizes the weakness of low-skilled workers employed under part-time contracts. Outflows from employment are characterized by a higher number of significant time-dummy parameters than transitions out of unemployment, especially for flows towards inactivity. However, the analysis of these coefficients, -especially those referring to flows between employment and inactivity -reveals an immediate increase from 2004 to 2005, followed by a level that remains constant over time. Among the features of interest contained in Table 12.5, it can be observed that women at the beginning of their working career (individual age effect) are less likely to exit the state of Non-labour force. This situation is aggravated for individuals holding low qualifications and by unfavourable labour market conditions summarized by the unemployment rate. A high number of relevant parameters emerge for the time trend. This suggests the existence of fluctuations in transitions in the five-year period considered. Such fluctuations are also shown in Table 12.1.

12.5 Concluding Remarks We have performed both descriptive and microeconometric analyses in order to obtain a complete picture of the new scenarios that are emerging with reference to the labour mobility in Italy, with particular emphasis on the phenomenon of unemployment. As regards the latter, the descriptive analysis suggests that conditions deteriorate for people in search of employment, especially for long-term job-seekers. Unemployment increased, affecting an increasingly large component of the male workforce, and became more widespread in northern areas of the country, especially at the height of the economic downturn (2008–2009). Having specific individual characteristics seems to reduce the impact of the crisis. In particular, previous work experience and holding a degree increase the chances of successful exit from unemployment. The microeconometric analysis confirms these findings and further stresses the structural nature of unemployment in Italy. The other econometric estimates refer to outflows from the states of employment and inactivity and make it possible to identify the major determinants, thus completing the overall picture of mobility in the Italian labour market. The individual characteristics that increase the chances of remaining in employment emerge as being male gender, no longer belonging to the ‘young’ age group and holding a high educational qualification. Interesting aspects emerge from the type of contract and educational qualification in the occupation

262

C. Lucarelli and C. Mussida

performed. Full-time workers have a lower probability of exiting employment. The same effect emerges for occupations requiring high qualifications. Finally, with regard to outflows from inactivity, women appear less likely to exit the Non-labour force and this points to a strong discouragement effect for women, especially at the beginning of their working career. Holding low qualifications, and unfavourable labour market conditions, inevitably exacerbate the situation. The findings of this work suggest that the economic downturn has had a greater effect on disadvantaged (or weak) labour market categories in terms of both employment contracts (young people employed under temporary and/or precarious contracts) and the attitude on the labour market (the unemployed and women interested in getting a job are affected by a discouragement effect). The economic downturn implies a change in government labour policy. Past legislation was intended to increase flexibility at entry level, mainly by introducing temporary and flexible contracts. During the period 1997–2008 (before the impact of the crisis was felt), according to Istat/Eurostat data overall Italian employment figures increased by 3,173 million (largely due to a total increase of 3,076 million workers under contract; of these 1,180 had temporary contracts and 1,896 million permanent contracts); in addition, the majority of existing empirical researches into the Italian context demonstrate that temporary contracts generally favour transition to a position with a permanent contract (with respect to people/young not passing for temporary contract). Recent government action has focused mainly on sustaining individual income by providing additional funds for the Cassa Integrazione Guadagni (CIG)—the use of CIG has increased dramatically over recent years and especially at the height of the economic downturn; and on passive policy such as unemployment benefits. Such intervention is useful for defending employment levels and human capital acquired with work experience (useful to firms in the recovery). Only structural intervention would possibly contribute positively to an exit from this downturn, and to recovery in employment figures. Considering other economic policies, it is necessary to reinforce support instruments at the firm level, such as facilitating bank loans (offering better conditions); indeed, the bank system has not been as efficient and supportive as has been required by the economic situation. Further, it might be useful to facilitate the return to work of young people, by offering both flexible employment contract options and specific fiscal advantages for potential employers.

Appendix See Table 12.6.

12

Labour-Market Dynamics and Unemployment

263

Table 12.6 Description of variables employed in the econometric analysis Variable Description Utr Etr Ntr Man Age Agesq Italian Famsize Experiencea Loweduc Compulsory Diploma Degree Urate Dursearcha Gdpgrowth Low_qualb Fulltimeb Time 2005–time 2008 a b

1 for UE, 2 for UN, 3 for UU (base) 1 for EU, 2 for EN, 3 for EE (base) 1 for NE, 2 for NU, 3 for NN (base) 1 if male Age of the individual, [15–74] years age bracket Individual age squared 1 if Italian, 0 if foreigner Household size 1 for unemployed with job experience 1 if illiterate or completing 5 years of schooling 1 if compulsory education completed 1 for holding diploma 1 for holding degree or PhD (omitted category) Regional unemployment rate (annual average) Search for work duration (months) Annual gdp growth rate, regional level 1 for blue collar occupation, 0 for white collar 1 for full time jobs Time dummies (yearly)

Covariates employed only for unemployment related estimates Covariates employed only for employment related estimates

References Aeschimann G et al (1999) Modelling and forecasting the social contributions to the swiss old age and survivor insurance scheme. Swiss J Econ Stat 135(3):349–368 Andersen PK et al (1993) Statistical models based on counting processes. Springer, New York Barbieri G, Sestito P (2008) Temporary workers in Italy: who are they and where they end up. Labour 22:127–166 Baussola M (1985) Flussi occupazionali e differenziali nei tassi di disoccupazione maschili e femminili. Rivista Internazionale di Scienze Sociali 3:122–145 Baussola M (1988) Labour mobility in Italy: an empirical investigation. Labour 3:113–142 Bentivogli C, Pagano P (1999) Regional disparities and labour mobility: the Euro-11 versus the USA. Labour 13(3):737–760 Bertola G, Garibaldi P (2002) The structure and history of italian unemployment. CESifo working paper series No. 907 Berton F et al (2008) Temporary jobs: port of entry, trap, or just unobserved heterogeneity? LABORatorio Revelli, working paper No. 79 Cameron AC, Trivedi PK (2005) Microeconometrics: methods and applications. Cambridge University Press, Cambridge Commission, European, Employment in Europe (2009) Directorate-general for employment, social affairs and equal opportunities, Chapter 2 Cook RJ et al (2002) A generalized mover-stayer model for panel data. Biostatistics 3(3):407–420 Discenza AR, Lucarelli C (2009) Dati longitudinali a 12 mesi di distanza -Aspetti metodologici, 2009. ISTAT, Nota metodologica allegata alla diffusione dei file MFR, Dicembre, Roma Discenza et al (2010) La mobilità nel mercato del lavoro: principali risultati del periodo 2004–2008, ISTAT, Approfondimenti, 1 Febbraio, Roma

264

C. Lucarelli and C. Mussida

Fabrizi E, Mussida C (2009) The determinants of labour market transitions. Giornale degli Economisti e Annali di Economia 68(2):233–265 Gazzelloni S (2006) La rilevazione sulle forze di lavoro: contenuti, metodologie, organizzazione. ISTAT -Metodi e Norme n. 32 -2006 -Roma Gourieroux C (1989) Econométrie des variables qualitative, 2nd edn. Economica, Paris ISTAT (2009) Forze di Lavoro -Media 2008. Annuari, Roma ISTAT (2010a) Occupati e disoccupati. Comunicato Stampa, disponibile on line. http:// www.istat.it/salastampa/comunicati/in_calendario/forzelav ISTAT (2010b) Rapporto Annuale. La situazione del Paese nel 2009 Jackman R, Roper S (1987) Structural unemployment. Oxford Bull Econ Stat 49:9–36 Leoni R (1984) Indicatori di mobilità: processi di transizione markoviani o stati di dipendenza? In: L’offerta di lavoro in Italia, Venezia, Marsilio Paggiaro A (1999) Un modello di mistura per l’analisi della disoccupazione di lunga durata. Dipartimento di Scienze Statistiche, Università di Padova, working paper No. 12 Picchio M (2006) Temporary Jobs and State Dependence in Italy. Quaderni di Ricerca, Dipartimento di Economia, Università Politecnica delle Marche, No. 272 Picchio M (2008) Temporary contracts and transitions to stable jobs in Italy. Labour 22:147–174 Pissarides A, Wadsworth J (1989) Unemployment and the inter-regional mobility of labour. Econ J 99(397):739–755 Ricciardi L (1991) La disoccupazione di lunga durata in Italia: un’analisi dell’evidenza empirica nel periodo 1977–1989. Economia Lavoro 25(2):69–94 Sahin A et al (2010) Mismatch in the labour market: UK. centre for economic policy research Schindler M (2009) The italian labor market: recent trends, institutions and reform options. IMF working paper No. 47 Trivellato U et al (1989) Analysis of labour force dynamics from rotation panel survey data. Bull Int Stat Inst 53(2):425–444