Incentives for Regional Development: Competition Among Sub-National Governments

Incentives for Regional Development Competition Among Sub-National Governments Kala Seetharam Sridhar Incentives for ...

Author: Kala Seetharam Sridhar

37 downloads 726 Views 1MB 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

Incentives for Regional Development Competition Among Sub-National Governments

Kala Seetharam Sridhar

Incentives for Regional Development

This page intentionally left blank

Incentives for Regional Development Competition Among Sub-National Governments

Kala Seetharam Sridhar

© Kala Seetharam Sridhar 2005 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London W1T 4LP. Any person who does any unauthorized act in relation to this publication may be liable to criminal prosecution and civil claims for damages. The author has asserted her right to be identified as the author of this work in accordance with the Copyright, Designs and Patents Act 1988. First published in 2005 by PALGRAVE MACMILLAN Houndmills, Basingstoke, Hampshire RG21 6XS and 175 Fifth Avenue, New York, N.Y. 10010 Companies and representatives throughout the world. PALGRAVE MACMILLAN is the global academic imprint of the Palgrave Macmillan division of St. Martin’s Press, LLC and of Palgrave Macmillan Ltd. Macmillan® is a registered trademark in the United States, United Kingdom and other countries. Palgrave is a registered trademark in the European Union and other countries. ISBN-13: 978–1–4039–4788–8 ISBN-10: 1–4039–4788–0 This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. A catalogue record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Sridhar, Kala Seetharam, 1966– Incentives for regional development : competition among sub-national governments / Kala Seetharam Sridhar. p. cm. Includes bibliographical references and index. ISBN 1–4039–4788–0 (cloth) 1. Economic development – Government policy. 2. Regional planning – Government policy. 3. Tax incentives. 4. Industrial policy. 5. Industrial promotion. I. Title. HD75.S77 2005 338.9—dc22 2005042144 10 9 8 7 6 5 4 3 2 1 14 13 12 11 10 09 08 07 06 05 Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham and Eastbourne

To my parents, husband and my little Vindhya

This page intentionally left blank

Contents List of Tables

x

List of Figures

xii

Acknowledgements

xiii

List of Abbreviations

Part I 1

xv

Introduction

Regional Development Incentives in the United States and India 3 1.1 Introduction and background 3 1.2 Motivation for the book 3 1.3 What is the black box that converts incentives to regional development? 4 1.4 Why cross-national comparison? 6 1.5 Gaps in the literature 9 1.6 Research objectives 11 1.7 Overview of the book 13

Part II

Tax Incentives: Theory and Evidence

2 Impact of Tax Incentives on Economies: Analytical Framework 2.1 Introduction and motivation 2.2 Literature review 2.3 Objectives of the model 2.4 Assumptions of the model 2.5 The model 2.6 Predictive power of the model

19 19 19 20 20 22 37

3 Competition Among American States: Evidence from Illinois Enterprise Zones 3.1 Introduction 3.2 Theoretical framework 3.3 Description of EZs in the United States 3.4 Description of enterprise zones in Illinois 3.5 Research methodology 3.6 Estimation of reservation wages 3.7 Benefit–cost analysis 3.8 Summary of the results

39 39 40 41 43 48 53 57 61

vii

viii Contents

4 Impact of Tax Incentives on the Unemployment Rate: Evidence from Ohio 4.1 The importance of the problem and the motivation for research 4.2 Overview of the chapter 4.3 A model of unemployment 4.4 Tax incentive programmes in Ohio 4.5 Implementation of the model 4.6 Data 4.7 Methodology 4.8 Results from the estimation 4.9 Policy implications 4.10 Concluding remarks 5 Benefits and Costs of Regional Development: Evidence from Ohio’s Enterprise Zone Programme 5.1 Introduction 5.2 Past literature 5.3 Ohio’s enterprise zone programme 5.4 Data 5.5 Benefit–cost methodology 5.6 Benefit–cost analysis 5.7 Policy implications

Part III

64 64 66 66 68 70 73 75 80 85 85 87 87 88 89 90 90 97 110

Infrastructure Incentives: Theory and Evidence

6 Impact of Growth Centres on Unemployment and Firm Location: Evidence from India 6.1 Introduction 6.2 The importance of and the motivation for research 6.3 Objectives 6.4 Description of the growth centres programme 6.5 Theory and model 6.6 Description of variables and data 6.7 Results from the estimation of the unemployment rate 6.8 Estimation for GCs 6.9 Concluding remarks 7 Firm Location Decisions and Their Impact on Local Economies: Evidence from India’s Growth Centres 7.1 Introduction and motivation 7.2 Objectives 7.3 Field visits to GCs 7.4 Summary of GCs visited 7.5 Impact of GC (infrastructure), and tax incentives on firm location

117 117 118 119 124 125 127 130 133 138 139 139 139 140 153 154

Contents ix

7.6 7.7 7.8 7.9 7.10 7.11

Part IV 8

The impact of GC firms on local labour markets The export orientation of firms Corporate social responsibility Firm-level implications Growth-centre-level implications Concluding remarks

157 159 161 162 164 166

Lessons Learned

Lessons Learned from the United States and India 8.1 Introduction 8.2 Lessons for the United States 8.3 Lessons for India 8.4 Is the American experience relevant for India? 8.5 Can the corporate sector help? 8.6 Concluding remarks

169 169 171 173 177 183 184

Notes

185

References

204

Index

210

List of Tables 3.1 3.2 3.3 3.4 3.5 3.6 4.1 4.2 4.3 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9

5.10

5.11

5.12

5.13

Summary of selected studies on enterprise zones Achievements of the Illinois enterprise zone programme Fiscal effects of the programme Estimation of reservation wages: results from regression Case 1: summary of net benefits from ‘new’ jobs Case 2: summary of net benefits from relocated jobs Descriptive statistics for variables Estimation of the unemployment rate Estimation of the unemployment rate with dummies for duration Switching regression model of reservation wage with sample selection Profile of Ohio’s enterprise zones, 1997 Profile of Ohio’s non-zone census block groups, 1997 Distribution of reservation wages imputed for Ohio’s enterprise zones Distribution of average net benefits and benefit–cost ratios for the (531) firms in Ohio’s enterprise zones, scenario 1 Distribution of net benefits and benefit–cost ratios in Ohio’s (143) enterprise zones, scenario 1 Distribution of net benefits and benefit–cost ratios for the (575) firms in Ohio’s enterprise zones, scenario 2 Distribution of net benefits and benefit–cost ratios for (148) Ohio’s enterprise zones, scenario 2 Distribution of net benefits and benefit–cost ratios for (198) firms in Ohio’s enterprise zones, scenario 3 (assumed elasticity 0.3) Distribution of net benefits and benefit–cost ratios for (91) firms in Ohio’s enterprise zones, scenario 3 (assumed elasticity 0.1) Distribution of net benefits and benefit–cost ratios for Ohio’s (62) enterprise zones, scenario 3 (assumed elasticity 0.3) Distribution of net benefits and benefit–cost ratios for Ohio’s (32) enterprise zones, scenario 3 (assumed elasticity 0.1) Summary of average B–C ratios in various scenarios at firm level and zone level

x

42 48 50 55 58 58 79 80 84 92 93 94 95 99 101 103 104

105

106

108

109 110

List of Tables xi

5.14 5.15 6.1 6.2 6.3 6.4 6.5 6.6 7.1 7.2 7.3 7.4 7.5 7.6 7.7

Proportion of zones with B–C ratios 1 in various scenarios Efficiency losses in various scenarios Description of data used in estimation of unemployment rate Comparison of data for areas with and without GCs Results from OLS and 2SLS estimations OLS estimation of unemployment for low- and high-unemployment areas Description of data for GCs Estimation of GC performance Registered industrial units and total employment in Karnataka’s districts, 1999–2000 Summary of manufacturing and services employment, UP districts Overview of GCs visited All firms located in GCs visited, and their location decisions Impact of GC firms on local labour market Export orientation of GC firms and suggestions for government policy Corporate social responsibility: evidence from GC firms

111 112 129 130 131 134 135 137 142 147 154 155 158 160 161

List of Figures 2.1 The distribution of wage offers 2.2 The impact of the EZ (tax incentives) with capital mobility 3.1 Distribution of unemployment rate in Illinois EZs 4.1 Determination of the unemployment rate 7.1 District map of Karnataka, India 7.2 District map of Haryana, India 7.3 District map of Uttar Pradesh, India

xii

24 36 46 70 143 144 146

Acknowledgements This book, as in all cases, is the result of cooperation from several people, without which it would not have been possible. First of all, I would like to thank the National Institute of Public Finance and Policy for facilitating the time to work on this book. I would like to thank the United States Department of Housing and Urban Development which funded the research in Chapters 2, 4, and 5, through its 1997 Doctoral Dissertation Research Grant H-21090SG. For work in these chapters, I am very thankful to Professors Chuck Adams, Don Haurin, late Professor G.S. Maddala, Anand Desai, and Mary Marvel, all at the Ohio State University, for their guidance. For work in Chapter 3 pertaining to Illinois’ enterprise zones, I am indebted to Professors Peter Fisher and Alan Peters at the University of Iowa, for their guidance and comments. I thank the Ohio Department of Development for providing data necessary for the evaluation of Ohio’s enterprise zone programme in Chapters 4 and 5. Further, I am very thankful to the various enterprise zone administrators in Ohio for patiently filling out my survey pertaining to infrastructure costs and several CRA administrators in Ohio for their discussions over the phone. Finally, I thank the erstwhile Illinois Department of Commerce and Community Affairs, and the current enterprise zone administrators in Illinois, Mr Thomas Henderson and Mr Mickey Harris, with the Illinois Department of Commerce and Economic Opportunity, for sending me current information about Illinois enterprise zones. For work in Chapters 5 and 6, I am indebted to the Indian Institute of Management in Lucknow, India for providing seed money to facilitate the research. I am very thankful to Mr Onkara Murthy at the Karnataka Industrial Areas Development Board (KIADB) for taking me around the various firms in the Hassan growth centre. Further, I am thankful to Mr K.C. Sharma with the Haryana State Industrial Development Corporation (HSIDC) for taking the time to visit several firms in the Bawal growth centre along with me. I thank Mr Alok Kumar, Joint Managing Director of the UP State Industrial Development Corporation (UPSIDC), and his colleagues, for taking the time to answer my questions relating to growth centres in UP during my visit to Kanpur. Further, I am grateful to Mr K.C. Mishra and his colleagues with the Gorakhpur Industrial Development Authority for introducing me to various firms and answering my detailed questions. I thank Mr N.P. Singh of the Satharia Industrial Development Authority for facilitating my visits to various firms in the Satharia growth centre, to enable the assessment. In all of the growth centres that I visited, I wish to thank heartily the representatives of various firms that made themselves available to answer xiii

xiv Acknowledgements

the various questions I had about their location decision and other aspects of their operation. Finally, I would like to thank Mr S. Jagadeeshan, Joint Secretary in the Department of Industrial Policy and Promotion, Ministry of Commerce and Industry, Government of India, for arranging to send me secondary data regarding growth centres in the country. I thank Dr Amaresh Bagchi at NIPFP, Mr R.K. Bajaj, Commissioner of Income Taxes, and Ms Ann Martin of Palgrave, for heartily offering to help with Palgrave’s tax form. I am very thankful to Ms Amanda Hamilton of Palgrave, the commissioning editor for this book, who patiently and promptly answered all of my questions relating to all aspects of the book. Thanks are due to Mr Keith Povey for his patience with copy-editing and proof-reading the book. I thank Ms Amita Padhwal for her assistance in formatting the book. Further, I would like to thank Sage Publications for permitting me to reprint substantial parts of my work that had been published previously in the Economic Development Quarterly, 10 (1), 69–90 in Chapter 3. I wish to thank the University of Delaware’s College of Business and Economics for providing copyright permission to reprint substantial parts of my paper published in the Review of Regional Studies, 30 (3), Winter 2000, 275–98 in Chapter 4. Further, I would like to acknowledge with thanks the copyright permission granted by the Centre for Community Economic Development at the University of Wisconsin, Madison, for allowing me to reprint substantial parts in Chapter 5, from my paper published in the Journal of Regional Analysis and Policy, 31 (2), 2001, 1–32. Thanks are due to the editorial office of the Economic and Political Weekly for permitting me to reprint parts of my paper 38 (39), 2003, 4121–30 in Chapter 7. Last but not least, I thank the incredible support and encouragement of my husband from the time of the proposal formulation to the fruition stage to writing and formatting. My little daughter, Vindhya, supported efforts of the book in her own way, without being aware of how much she was contributing. Finally, my parents and family played a big role in providing the moral support necessary for completing this work. KALA SEETHARAM SRIDHAR

List of Abbreviations 2SLS AFDC A&N Islands ARV B–C BEA CAGR CETA CPI–U CPS CRAs CRTS CSR CT DCCA DCEO DIPP EC EZs FL FY GCs GDP GIDA GIS GLS GPO HSIDC IL IN INR I–O Analysis IRBs IT KIADB LHS MAX MIN MNCs

Two-Stage Least Squares Aid to Families with Dependent Children Andaman and Nicobar Islands Annual Rental Value Benefit–Cost Bureau of Economic Analysis Compounded Annual Growth Rate Comprehensive Employment Training Act Consumer Price Index–Urban Current Population Survey Community Reinvestment Areas Constant Returns to Scale Corporate Social Responsibility Connecticut Department of Commerce and Community Affairs Department of Economic Opportunity Department of Industrial Policy and Promotion Enterprise Communities Enterprise Zones Florida Fiscal Year Growth Centres Gross Domestic Product Gorakhpur Industrial Development Authority Geographic Information Systems Generalized Least Squares Government Printing Office Haryana State Industrial Development Corporation Illinois Indiana Indian Rupees Input–Output Analysis Industrial Revenue Bonds Information Technology Karnataka Industrial Areas Development Board Left hand side Maximum Minimum Multinational Corporations xv

xvi List of Abbreviations

MO MW N/A NCR NIUA NLS OBES OH OLS PDS PSID PPPs RCs RHS SC SEZs SIC SIDA ST TIRC UA UEZ UP UPA UPSIDC US Department of HUD US VAT

Missouri Megawatts Not Applicable National Capital Region National Institute of Urban Affairs National Longitudinal Survey Ohio Bureau of Employment Services Ohio Ordinary Least Squares Public Distribution System Panel Study of Income Dynamics Public–private Partnerships Renewal Communities Right hand side Scheduled Castes Special Economic Zones Standard Industrial Classification Satharia Industrial Development Authority Scheduled Tribes Tax Incentive Review Council Urban Agglomeration Urban Enterprize Zone Uttar Pradesh United Progressive Alliance UP State Industrial Development Corporation United States Department of Housing and Urban Development United States Value Added Tax

Part I Introduction

This page intentionally left blank

1 Regional Development Incentives in the United States and India

1.1

Introduction and background

In this book, I study the effectiveness of policies to promote regional development and competition among sub-national levels of government. I study competition among sub-national governments in the United States and India to attract private investment, respectively with the use of tax and infrastructure incentives. I examine empirical evidence regarding incentives offered by state and local governments in the United States and contrast this evidence with that from incentives offered by state governments in India. Note that although the title speaks of ‘incentives for regional development’ I will only discuss financial/tax incentives and infrastructure incentives in the book. The reader should be aware that the regional development literature includes non-financial incentives also. There is a stream of literature on regional development (that is, learning regions, milieu of innovation, industrial districts),1 which is widely discussed throughout the world. Such policies have been characterized as ‘new wave’ policies (Bartik, 1991) or demand-side policies (Eisinger, 1988). They are all programmes targeted at existing firms. Those policies include services such as providing information regarding the existence of markets, export promotion assistance, assistance to small firms in the form of business incubators, hosting of trade fairs, support of high-technology, e-learning programmes, and other services which boost firm productivity and the development of the region as a whole. While the reader must be aware that there is a stream of literature in each of these areas, the focus of this book is supply-side policies that target branch plant recruitment.

1.2

Motivation for the book

The motivation for this book comes from the extensive use of financial incentives to attract firms and jobs in the United States (especially through the use of enterprise zones (EZs)) that continues to be hotly debated and 3

4

Incentives for Regional Development

aggressively pursued by states. For instance, the conventional state of Iowa in the Midwestern United States recently passed enterprise zone legislation, and as of September 2002, 332 EZs were certified in this state, with the programme’s popularity having steadily grown since 1997 (McDermott, 2002). In November 2003, I met the Mayor of the City of Chicago. He expressed the firm view that if Chicago does not offer financial incentives, it will lose out in the war for jobs. So while many sub-national governments in the United States realize the futility of such financial incentives in general, no state or local government wants to lose out in the ‘race to the bottom’. For this reason, they continue to offer such incentives. The motivation for the book also comes from the poor state of infrastructure in India’s various states, where the use of programmes such as growth centres (GCs) that provide infrastructure incentives to firms can trigger huge infrastructure reforms throughout the country. Florida (2002) showed that those US cities and regions that use tax incentives, in fact, attract manufacturing industries, and it may appear that these firms are contributing little to regional development in the present knowledgebased economy. It may be misconstrued that in the present knowledge-based economy, regional development comes from innovation, which calls for incentives other than taxes and infrastructure. Boekema et al. (2000) argue that successful firms, as well as governments, are those which have control over and access to flows of information and knowledge of technologies, markets, and organizational and managerial practices. Note, however, that infrastructure (including good roads, uninterrupted power, reliable telecom, banking and related services) is quite critical to the success of all firms – whether they are primarily manufacturing, services or knowledge-based. In the case of India, although it has emerged recently as a knowledgeintensive economy dependent on IT and IT-enabled services, it continues to be a developing economy with an underdeveloped infrastructure and is not yet a full fledged knowledge-based economy in the sense that the USA and Europe are. Because of this, what is good for the industrial economy is good for the Indian economy. Further, all the leading thinkers of the world, including economists and corporate leaders, agree that a large and sound manufacturing base is necessary to compete in the globalized market. So as Eisinger (1995) points out, there is a return to industrial recruitment among the American states. This is also true in relation to the Indian states which are pursuing entrepreneurial and knowledge-based strategies targeted at existing firms, along with financial incentives to attract new ones.

1.3 What is the black box that converts incentives to regional development? Throughout the book, regional development refers to improvement in the areas’ infrastructure, amenities, job opportunities, and, more generally, its

Regional Development Incentives 5

attractiveness as a place to live and carry out business. Given that financial and infrastructure incentives are important for regional development for one reason or another, what is the mechanism that converts tax benefits and infrastructure to regional development? Relevant aspects to consider here are: ●

●

Manufacturing industries and those that have first-mover disadvantages (for instance, information technology) are attracted by tax and infrastructure incentives, so a region can potentially enhance its economic base with the offer of such incentives. Recall that whenever the social benefits from an activity are greater than the private costs, there are grounds for subsidizing such activity. Take the case of fledgling industries such as information technology (IT) and IT-enabled services. The first movers into the industry have been at a disadvantage because of the specific kind of skills required (software programming and data processing, to give a couple of examples). Given that these skills are scarce in relation to demand, attrition in this industry is high, the first-mover firms are at a disadvantage because of the financial and other resources spent in training the manpower. So at least initially, there might be some grounds for policy to subsidize such firms, by providing tax or infrastructure incentives, since the gains from employment are substantial. The other mechanism that converts tax or infrastructure incentives to regional development is the extent to which such firms are footloose. In the context of developing countries such as India, the industries do tend to stay on once attracted to a particular region. Once the infrastructure is made available and firms do fix a location, relocations are quite rare. I have confirmed this with the Department of Company Affairs, Government of India.

In the context of the United States, footloose firms are more common, casting doubts on the ability of such incentives to translate into sustainable regional development. However, even there, many states such as Ohio have incorporated clawback of tax incentive arrangements in the event of nonperformance or relocation, should firms relocate to another region. Such arrangements, while still very incipient, do act as a check on footloose firms that could prevent the full realization of tax or infrastructure benefits by regions. So tax and infrastructure incentives do have a number of benefits that have spin-offs for regions. Infrastructure investments can increase employment through both direct and indirect effects (Alleman et al., 2002). For instance, investing in telecommunication will increase the demand for the goods and services used in their production and increase total national output. In their evaluation of the New Jersey Urban Enterprise Zone (UEZ) programme, Rubin and Armstrong (1989) take into account the direct and indirect benefits from the programme. Using the input–output model

6

Incentives for Regional Development

developed by the United States Bureau of Economic Analysis (BEA) for New Jersey in order to estimate these impacts, the study calculates the output multiplier to be 1.61 and the employment multiplier to be 2.26. This means that every $1 of direct expansion in output generated an additional $0.61 in indirect output by other companies, and every job added by the firms resulted in an additional 1.26 jobs. While this evidence is more than a decade old, Keefe (2004) finds that EZ designation raised employment growth in California’s zones by about 3 per cent each year during the first six years after designation, but did not persist in later years. Similar results are reported in Chapter 4 of this book. We have to make several observations about the relative importance of tax incentives and public (primarily infrastructure) services based on the literature. First, when the level of public (infrastructure) services is held constant, the benefits to firms of tax incentives are quite significant. Although there are several factors that are more important than taxes to business location decisions, tax incentives enable corporate planning for reinvestment. Further, whenever traditional factors (such as transport costs, raw material availability, size of, and proximity to the market) are likely to be different across states, such traditional factors are more important in influencing manufacturing firm location decisions. However, across locations (such as within a state) where traditional factors are constant, tax incentives are likely to play a more important role in influencing firm location. Further, it is not necessarily the case that incentives to industry will be harmful to the development of infrastructure. One good example is the Indian state Karnataka’s proposal to replace financial with infrastructure incentives to the auto industry. Karnataka promised training institutions, schools, colleges, office complexes, housing, a globally well-knit telecom network, roads, dedicated power and water supply necessary for the development of automobile manufacturing units, vendors and dealers in the state. While highlighting the continued importance of manufacturing in India, if other states in India and sub-national governments in other countries also follow this example, competition will enhance, rather than stifle, their infrastructure competitiveness. These examples provide some evidence that such tax and infrastructure incentives have favourable impacts on the economic base of local economies.

1.4

Why cross-national comparison?

The cross-national comparison of India and the United States is of relevance, as India has a federal structure similar to the United States, in which there are a large number of sub-national governments which compete against one another. In the United States of America, the last two decades of the twentieth century were characterized by substantial delegation of responsibilities by

Regional Development Incentives 7

national governments to their lower levels of government. This phenomenon, termed the New Federalism, began with the presidency of Ronald Reagan in the 1980s. Similarly, in India also, a reversal of concentration of central decision-making began to occur in the last two decades of the last century (Bagchi, 2003).2 In the United States, the last two decades of the twentieth century have been characterized by acceleration of competition among states and local governments with the offer of tax incentives to firms to make investment and create employment in their states and local areas. A tool of regional development that is very popular in the United States is enterprise zones (EZs). The EZ concept was born in Great Britain where the Chancellor of the Exchequer Geoffrey Howe advocated tax cuts to boost the economy. This concept was introduced in the United States by a geographer, Stuart Butler. EZs in the US are geographically targeted areas chosen for development, and firms that locate in these areas receive tax credits, exemptions and abatements. This can make considerable differences to firms because moving just a mile across state borders – or, even within a state for that matter – can save them millions of dollars. The offer of tax incentives was proliferating even among Indian states in the post-1991 period. Until recently, there were several instances of tax competition. The leaders of Karnataka and Andhra Pradesh – two leading states in South India – had been competing fiercely with each other to secure the location of Microsoft’s newest India facility. In terms of actual tax incentives also, there had been several instances of generous abatements during the 1990s in India, quite similar to ‘smoke-stack chasing’ in the United States. While states in the United States continue to offer tax incentives and many other sops, a conference of state chief ministers and the Union Finance Minister in India decided in November 1999 to stop this tax war among the Indian states.3 The decision was taken by Indian states because the offer of tax incentives, apart from affecting the general fiscal health of the states, also affects the states’ ability to provide public infrastructure services. This is especially so, given the fact that sales tax revenue accounts for nearly one-quarter of own source revenue, and roughly two-thirds of all revenue, for the majority of the Indian states. Similar concerns are also being expressed in US academic and policy circles with property tax abatements being the most important incentive at the local level. Consistent with the idea of place-oriented policies, GCs in India are like the EZs of the United States, except that they offer primarily infrastructure incentives for investors that locate there. The United States and a number of countries around the world – China, the Middle East, and countries including India have been adopting some form of place-oriented policies, be it called EZS, special economic zones or GCs. A big advantage of the ‘zone’ concept being that it can be initially tried out experimentally in a small area. If it works, well, it can be extended to other areas. If it does not, then we know

8

Incentives for Regional Development

that it doesn’t work. It thus avoids the risks and costs of trying a programme in a fully-fledged manner without knowing its effects. Further, place-oriented policies such as EZs call for the targeted development of the area rather than advocate that people should be mobile to make use of job opportunities where they exist. Place-oriented policies recognize that there are likely to be a large number of people that are immobile because of the social, psychological or economic costs of moving. These people are most likely to be those with the lowest skills. Therefore place-oriented policies such as EZs and GCs deserve an in-depth consideration for their impacts on areas that adopt them. In this book, I present evidence of place-oriented policies, EZs and GCs, from this perspective. The GCs programme was initiated by India in 1988 in order to promote the industrialization of backward areas in the country. As a result, 71 GCs were established throughout India. They were allotted to the various states on the basis of combined criteria of area, population and the extent of industrial backwardness, as long as they were located at a minimum distance from major cities in the area (the specific designation criteria are described in Chapter 6). These GCs provide basic industrial infrastructure like power, water, telecom, and banking to enable the states to attract industries. Central government funds for the programme were to be leveraged by the states for purposes of financing. The infrastructure (growth centre) approach is important to consider as it is a test of the alternative to the tax war (not place-oriented policies). Further, it is an important approach to increase rural industrialization to enable convergence where disparities exist among sub-national units (see Dreze and Sen (1996) to understand the magnitude of disparities among Indian states).4 Even in the United States, quite stark inter- and intra-state disparities exist, similar to those in India. For instance, south-eastern Ohio is relatively poor compared to northern/central Ohio. The Appalachia has traditionally been a pocket of poverty. Thus, recognizing that it is important for all sub-national governments to attract private investment and jobs, states/local governments in the United States continue to use EZs and tax incentives in order to attract private investment. Infrastructure incentives assume special importance in light of the scepticism regarding tax incentives in academic and policy circles everywhere, and the decision to stop the tax war among the Indian states. Further, note that countries everywhere are increasingly concerned about their fiscal deficits caused by financial incentives and also the extent to which such deficits stifle growth. This is especially the case if we consider (in federations) the central government’s deficit along with that of the states and local governments as a proportion of GDP. As with EZs and tax incentives in the United States, scepticism surrounds GCs in India and their effectiveness in attracting firms. Commentators’ views have varied from a perception of the GCs programme having been a

Regional Development Incentives 9

colossal failure in attracting firms to one that strongly believes in their effectiveness because of the infrastructure incentives available to firms that locate there. I present in this work, along with empirical evidence regarding the effectiveness of incentives offered in the EZs of the United States, empirical evidence of the effectiveness of GCs. Why should policy-makers in other countries care about the GCs of India? The Indian growth centre approach represents infrastructure and public service incentives, and has implications for sub-national governments that wish to attract private investment. In light of the recent decision of the Indian states to call a stop to the tax war (the so-called ‘race to the bottom’), and scepticism regarding tax incentives, there is a need to explore other, sustainable, options for sub-national governments everywhere to compete effectively for industrial investment. In summary, the similarities between India’s GCs and US EZs that make the two programmes amenable for comparison are: 1. India and US are large democracies and are federations with a decentralized, three-tier (national, state and local) government structure. 2. US EZs and India’s GCs are both geographically targeted and place-oriented, not people-oriented programmes such as those focusing on employees and on-the-job training. 3. Both the programmes offer some degree of flexibility to their respective sub-national government units to design the programmes. This gives rise to competition between the sub-national units that compete for investment. This, in fact, is the core objective of the book. For these and various other reasons explained above, place-oriented policies such as tax incentives and GCs deserve in-depth consideration for their impacts on the rest of the economy. In this book, I develop an analytical framework to analyze the impact of such place-oriented policies on the economy that adopts it. Further, I present cross-national empirical evidence regarding the effectiveness of these policies from American EZs and India’s GCs.

1.5

Gaps in the literature

There is a substantial body of literature on regional development incentives in the United States and elsewhere. Since there are several studies that provide a comprehensive literature review, I will not undertake that here, except providing a more suggestive and critical review, based on those from the United States. I review the sparse regional development literature from developing countries in somewhat greater detail in Chapter 6 when I discuss GCs. Byrnes, Marvel and Sridhar (1999) examine city-firm bargaining decisions and study the factors that explain the generosity of tax abatements, in the context of Ohio’s EZs. Fisher and Peters (1998, 2002) study, respectively, industrial

10 Incentives for Regional Development

incentives in American states and cities and state enterprise zone programmes in the United States. While Fisher and Peters (2002) study the national-level performance of EZs,5 they study in detail Ohio’s enterprise zone programme. They find that EZs as tools of growth are highly questionable. Further, a majority of the literature that studies economic development incentives in the United States does not study non-tax incentives (Steinnes, 1984; Papke, 1987, 1991; Tannenwald and Kendrick, 1995; Tannenwald, 1996, cited in Fisher and Peters, 1998). Various papers in Kenyon and Kincaid (1991) study, in the American context, whether competition among state and local governments is detrimental to public service and tax systems. They present a political science perspective of a model of federalism and the economic perspective of competition among firms in a market economy, without, however, offering detailed empirical evidence of the alternative views. Brunori (1998, 2001) deals with the problems of state taxation policy in the United States in the event of incentives, e-commerce, and globalization, and how they can be overcome. No cross-national evidence, however, is presented in the context of globalization. The past literature has looked at the effectiveness of tax incentives in attracting investment in the context of developing countries. For instance, various papers in Shah (1995) quantify the effect of investment incentives on business decisions related to production and investment in selected developing countries and then use the results to rank tax policy instruments in terms of their usefulness in promoting investment. Despite the fact that various studies in Shah (1995) examine both conceptually and empirically the impact of taxes on various business decisions, it has certain caveats. The work provides no empirical evidence of non-tax policies to promote industrial development. There is also no evidence reported from India; other developing countries’ policies are studied and evaluated. There are studies that evaluate specific fiscal, or industrial incentives in Indian states. An early example is Tulasidhar and Rao (1986), which shows employment and output loss due to tax incentives, albeit in a partial equilibrium framework. Rajaraman, Mukhopadhyay and Bhatia (1999) study fiscal incentives from the central Indian state, Madhya Pradesh, and find that fiscal incentives have a statistically insignificant impact on large and medium investment in Madhya Pradesh. None of these studies evaluate different kinds of (tax and infrastructure) incentives offered by sub-national governments in a federation, crossnationally. Such data and research are sparse in the literature. This book bridges this gap since no state-of-the-art research in the area compares the effectiveness of tax vis-à-vis infrastructure incentives in firm location decisions, and their effects on unemployment, and evaluates net benefits. It also constructs an analytical framework to understand the impacts on the rest of economy of place-oriented policies that rely on tax incentives.

Regional Development Incentives 11

1.6

Research objectives

The objectives of the research in this book are to answer several questions: 1. What are the effects of place-oriented policies such as EZs (or those that provide tax incentives) on the rest of the economy? 2. Do such incentives merely redistribute employment? 3. What are the effects of tax and infrastructure incentives on the unemployment rate of areas adopting them? 4. Are the net benefits from employment worth the costs of creating them? To answer the first of these questions, I build an analytical framework to analyze the impact of geographically targeted policies that provide tax incentives, on the economies that adopt them and on the rest of the economy, in Chapter 2. In the model I point to the cause of unemployment in the backward areas. I show the relationship between the reservation wage and unemployment rate, following Jones (1989). I then show the general equilibrium response to the tax abatement provided in EZs, in a generalized framework incorporating capital mobility, following Harberger (1962). The analytical framework developed indicates that the capital and employment impact of the tax cut on capital in the area depends on certain sets of parameters. Chapters 3, 4 and 5 respectively answer questions regarding the redistribution of employment, effect of tax incentives on unemployment rate, and comparison of the benefits and costs of EZs. Chapter 6 considers the question regarding the effect of infrastructure incentives on the rate of unemployment. Chapter 7 makes an attempt to answer these questions from a qualitative perspective, probing into aspects that could be missed in a purely quantitative or econometric approach. Regarding question two, note that the regional development literature has long argued that taxes do not affect firm location decisions. Furthermore, even if they do, they merely redistribute employment from one location to another. The thesis of the work is that poorer states/regions are justified in offering (infrastructure) incentives to attract industry and employment. If distressed areas, with the provision of incentives, were to be successful in attracting firms to invest and create employment, greater social net benefits would accrue to the area, as Bartik (1991) argues. Net benefit from a job is analogous to consumers’ surplus (in the goods market).6 Net benefit from a job is the extent to which actual wage is higher than the wage at which a person is willing to accept a job (which is referred to as a person’s reservation wage). If earnings were to be constant across areas,7 net benefits would be higher if persons are willing to accept jobs at lower wages. This is not to say that it is socially beneficial for persons to be willing to accept jobs at lower wages. But it is easy to imagine that persons in poorer, high-unemployment areas

12 Incentives for Regional Development

and poorer states, where job opportunities are difficult to come by, unemployed persons value the importance of having a job. As Bartik (1991) argues, they are likely to search more rigorously for job openings, wait longer in line for job interviews and less likely to quit a job once they obtain one. For these reasons, they would be willing to accept a job, if it becomes available, at a wage lower than the rate a person would be willing to accept in relatively richer, low-unemployment areas. Therefore, if an unemployed person in a high-unemployment area were to be offered a job, net benefits derived from this job would be higher than that from a similar job in a low unemployment area. In Chapters 3 and 5, I report evidence of this, based on data from the United States. Haurin and Sridhar (2003), using data from the United States, however, find no impact of higher local unemployment rates on individuals’ reservation wages. Nevertheless, as they suggest, it is sensible to attack clusters of high unemployment with policies that increase the demand for workers. The thesis of the book is also that infrastructure incentives, if offered by distressed areas, can become a more sustainable way to increase net benefits from employment. This is valid not only in a developing country such as India, but also in countries such as the United States, where the southern states have remained much poorer than their northern/western counterparts. Thus, we have to think about stimulating, rather than stifling competition, in the provision of infrastructure by sub-national governments. Such reforms are much needed. From this viewpoint, infrastructure incentives are crucial in influencing firm location decisions because they encourage competition among the states in the provision of infrastructure, quite critical to firms. The study thus focuses on the provision of infrastructure incentives as contrasted with the experience from tax incentives to attract private investment, as a more sustainable way of reducing regional disparities. This is especially so since the private sector is now the widely accepted driver of job growth in most countries. For instance, in India, according to data compiled from the Central Statistical Organization, aggregate employment in private sector large industries increased by an annual compound rate of 0.9 per cent during the period 1993–4 to 2001–2, whereas it declined by 0.4 per cent in the public sector during the same period. The private sector’s share in the country’s organized sector total employment increased from 25.6 to 27.6 per cent over this period, whereas the public sector’s share declined from 71 to 69 per cent in 2001–2. Furthermore, data support the continued importance of manufacturing, although the larger share of employment growth in India’s private sector was accounted for by the service industry (finance, real estate and insurance). India’s manufacturing sector witnessed increased employment growth during the period 1993–4 to 2001–2, and increased by 0.6 per cent annually compounded, raising its share in total employment from 16.9 to 17.9 per cent. In contrast, employment in public sector manufacturing fell by 3.5 per cent annually during the same period!

Regional Development Incentives 13

1.7

Overview of the book

The book is divided into four parts. Part I is introductory, consisting of Chapter 1. Part II of the book focuses on tax incentives and consists of Chapters 2, 3, 4 and 5. The focus of Part III of the book is infrastructure incentives and consists of Chapters 6 and 7. Part IV consists of the concluding chapter. The plan for the book is as follows. In Chapter 2, I outline a theoretical model that shows the effects of placeoriented policies such as EZs (that provide tax incentives) on the economy that adopts it and the rest of the economy. In the model I point to the cause of unemployment in the EZ. I show the relationship between the reservation wage and unemployment rate, following Jones (1989). I then show the general equilibrium response to the tax abatement provided in EZs, in a generalized framework incorporating capital mobility, following Harberger (1962). The analytical framework developed indicates that the capital and employment impact of the tax cut on capital in the EZ depends upon three sets of parameters: 1. Relative factor intensities of firms in the two areas; 2. The elasticity of substitution between capital and labour in firms in the areas; 3. The price elasticities of demand for goods Z and Y produced by EZ firms and non-EZ firms respectively. The framework shows that it is impossible to isolate the incidence of the tax cut given in the EZ solely to the EZ. In Chapter 3, I answer the following questions: are the local benefits of tax incentives greater than the costs? Further, are benefits greater in highunemployment areas than they are in low-unemployment areas? Next, do such policies generate net benefits for a state, even if they redistribute jobs? I present my empirical evidence of the effectiveness of the Illinois EZ programme from the Midwestern United States. I choose the Illinois programme because Illinois is one of the most important states in the Midwestern United States. Further, Illinois was one of the first states in the United States to adopt the EZ legislation once the EZ concept had been introduced in the United States from the United Kingdom in the 1980s. In this chapter, I discuss the various state EZ programmes in the USA after a summary overview of all state EZ programmes, I describe the Illinois EZ programme – its designation criteria, and tax incentives. Based on a simple OLS estimation of the reservation wages, and the imputation of benefits for Illinois, in this chapter, I find that, overall, the local benefits of creating jobs in Illinois EZs are greater than the costs of generating them. I highlight the assumptions and caveats of the empirical work. In Chapter 4, I answer the question about the effects of tax incentives on the unemployment rates of areas that adopt them. I describe Ohio’s

14 Incentives for Regional Development

enterprise zone programme. The Ohio programme is of interest as it has the dubious distinction of having nearly the largest number of EZs in the country, after Louisiana and Arkansas, and the ‘pirating’ of firms the programme has encouraged. As Fisher and Peters (2002) point out, Ohio is also one of the few states in the country that has a systematic database regarding various aspects of its EZs. I study, at Ohio’s census block group level, whether the unemployment rate is affected by the presence of a tax incentive programme. An empirical model is developed that describes the determination of unemployment in the EZ. The estimation of unemployment rate using Census data for Ohio, taking into account the treatment effects problem, shows that such programmes have a significant impact in reducing unemployment. The net impact of being an EZ for a period of seven years is approximately a three percentage point reduction in area unemployment. The results also suggest that three to five years could be the optimum period for offering incentives. In Chapter 5, I answer the question as to whether the benefits from regional development programmes such as EZs are worth their costs, using evidence from Ohio’s EZ programme. I study what happens when corrections are made to the empirical methodology used to estimate reservation wages. Then I present evidence from my analysis of Ohio’s EZ programme. Further, I present benefit–cost analyses of Ohio’s programme in various scenarios, making varying assumptions about the employment elasticity of tax incentives. I find that the net benefits of regional development can be expected to be greater than their costs. However, because of their efficiency implications, I find that it could be beneficial for high-unemployment areas to adopt tax incentive policies. Part III of the book focuses on infrastructure incentives. Chapter 6 describes India’s GC programme, symmetric to description of EZs in the US in Chapters 3 and 4. I present evidence from India’s GCs and study the effect of infrastructure incentives on the unemployment rate of districts that contain them using Census of India data (analogous to study of Ohio, in Chapter 3). I do not find GCs to have a statistically significant effect on unemployment rate. In this chapter, I also assess the performance of GCs where they exist and study whether infrastructure affects firm location decisions. I find that, among states having GCs, those having the financial resources to develop larger number of plots with infrastructure are the ones that are successful in attracting firms. Based on these findings, I conclude with implications for policy. In Chapter 7, I present qualitative evidence from my visits to GCs in India to study what could be missed based on purely quantitative data/empirical evidence. Based on the fieldwork, I find that without the infrastructure provided by the growth centre, many firms (even some representing local entrepreneurship) would not have located there. Given the scepticism about states and communities offering incentives to firms, I study the contribution

Regional Development Incentives 15

of firms to local communities where they locate. I find that these firms have favourable impacts on the local labour markets where they have located. Few of these firms export and some of them contribute socially to the communities in which they have located. I conclude with several implications for policy at the firm level and growth centre level. At the firm level, I summarize the suggestions made by various firms to improve the government’s industrial policy. At the growth centre level, I summarize implications for their designation criteria, based on the field visits. In the final chapter, I summarize the many policy implications that arise out of the work for policy-makers everywhere regarding place-oriented policies and other incentives for regional development. Finally, I highlight some areas for future research.

This page intentionally left blank

Part II Tax Incentives: Theory and Evidence

This page intentionally left blank

2 Impact of Tax Incentives on Economies: Analytical Framework

2.1

Introduction and motivation

In this chapter, I develop an analytical framework that shows the effects of place-oriented policies such as enterprise zones (EZs) on the regions that adopt them and on the rest of the economy.1 The analytical framework developed in this chapter can also be applied to GCs, wherever tax incentives are offered.2 The EZ programme has been one of the most controversial topics in the literature and in policy. Place-oriented policies such as EZs focus on the area’s development rather than arguing that people should be mobile. Such policies recognize that there are likely to be a large number of people who are immobile because of the social, psychological or economic costs of moving. The migration literature shows that immobile people are most likely to be those with the lowest levels of skills. Therefore place-oriented policies such as EZs deserve in-depth consideration for their impacts on labour and the rest of the economy. In this chapter, I develop an analytical framework to analyze the impact of the EZ on the economy that adopts it, and on labour.3

2.2

Literature review

There is a vast body of policy and empirical literature that evaluates EZs. These studies have evaluated EZs in various states of the United States – Indiana, Illinois, Ohio, Kentucky, New Jersey and California (Rubin and Armstrong, 1989; Erickson and Friedman, 1989; Seyfried, 1990; Elling and Sheldon, 1991; Redfield and McDonald, 1991; Papke, 1994; Landers, 1996; Dowall, 1996; Boarnet and Bogart, 1996). See Sridhar (1998) for details of these various studies. However, few of these studies provide an analytical framework that assess the impact of EZs on their economies. Ge (1995) and Seyfried (1990) develop analytical models which examine both the direct and indirect 19

20 Incentives for Regional Development

impacts of EZs on their regional economies. Both concentrate on production in analyzing the effect of EZs, and focus on cost minimization for firms when they locate in the EZ. They ignore the effects of the EZ on labour. It appears that none of the existing studies considers the efficiency implications of EZs or constructs an analytical framework to study their effects on labour. It is important to take into account some physical characteristics of the area, and distress characteristics pertaining to labour in the area, because most EZ programmes, at least those in the USA, specify explicit distress criteria that are used for the designation of zones. For instance, in the state of Illinois, EZs are required to document size and distress criteria for designation (these criteria are described in detail in Chapter 3). The size criteria in Illinois relate to the geographical area of the zone. A proposed EZ in Illinois must also satisfy at least one of four distress criteria relating to poverty, unemployment, low income or population loss. In this chapter I attempt to answer the following questions in the model: are EZs efficient if they are adopted by high-unemployment areas? What are the impacts of the zone on the rest of the economy? The chapter addresses these issues and develops a framework to analyze the adoption of programmes such as EZs by high-unemployment areas and examines their efficiency. It also evaluates the impact of the zone on the rest of the economy, applying some of the standard literature on analyzing the effect of tax incentives.

2.3

Objectives of the model

The objectives of the model to evaluate the EZ are as follows: 1. To describe the consequences of disequilibrium in the labour market of the high-unemployment area before it is designated as EZ, and to examine the cause of unemployment in the EZ. 2. To show that the area’s unemployment rate determines the reservation wage along with a host of other factors affecting the benefits and costs of remaining unemployed. 3. To examine whether economic rent accruing from jobs to job searchers in the EZ is higher. 4. To characterize the effect of the tax abatement on capital given in the EZ, on the EZ and the rest of the economy, taking into account full capital mobility.

2.4

Assumptions of the model

The model is based on certain assumptions: 1. There are two areas in the economy: one designated as EZ and the other being the rest of the economy, which I refer to as the non-EZ area (NEZ).

Analytical Framework 21

The NEZ may be considered as all areas that do not have an EZ programme or other programmes that abate taxes. EZ firms produce good Z, and NEZ firms produce good Y. In reality, this assumption is consistent with the fact that certain areas specialize in the production of those goods in which they have a comparative advantage. Rubin and Zorn (1985) show the comparative cost advantage different states in the United States have in different SIC category industries. In India, a typology of cities and states is emerging whereby each specializes in the output of a certain good or service (textiles, information technology and IT-enabled services and so forth). Thus, based on comparative advantage, we could easily classify areas without a loss of generality. 2. Goods Y and Z are produced by two factors of production, capital (K) and labour (L). In taking into account capital and labour, all conventional factors of production – land, labour, capital and organization – are implicitly taken into account in the model, since land can be considered to be a special kind of capital (real capital), and organizational skills a special case of labour. LZ and LY refer respectively to labour employed in the EZ and the NEZ firms. KZ and KY refer to capital employed in the EZ and NEZ firms respectively. PZ and PY are the prices of the final goods Z and Y respectively. PKZ and PLZ refer to the factor prices of capital and labour respectively in the EZ. Similarly, PKY and PLY refer to factor prices and refer to the price paid by users of capital and the price of labour (wage) in the NEZ area. Because EZ areas are, in reality, blighted, it is assumed that fLZ (the marginal product of labour in the EZ) fLY (the marginal product of labour in the NEZ). This is the result of differences in the use of capital across the areas. Specifically, KZ/LZ KY/LY, and for this reason, the marginal product of labour in the EZ is lower than in the other areas. Because fLZ fLY, the labour market is not in equilibrium when the EZ is designated. The model elaborates on this. The capital market is in equilibrium so that PKZ PKY PK. This implies that fKZ fKY and PKZ PKY.4 Consumers supply the factors in fixed amounts. 3. It is assumed that production is subject to constant returns to scale (CRTS) technology. CRTS has many interesting properties. The assumption of CRTS means that the average as well as the marginal products of the factors are dependent only on the ratio in which they are combined (Chiang, 1984; Krauss and Johnson, 1974), which, here is the capital–labour ratio. In empirical work, studies have found that a majority of the two-digit SIC industries studied were subject to CRTS. An interesting early example is Moroney (1967), in which, based upon estimates of production functions in two-digit SIC manufacturing industries in the United States, he concluded that a majority of these industries were subject to CRTS. More recent

22 Incentives for Regional Development

examples relate to plant-level data. For example, Bailey et al. (1992) used plant-level data and found that the plants they examined were characterized by constant returns to scale. Griliches and Ringstad (1971) argued that essentially constant returns were needed to rationalize the observed large dispersion of establishment sizes within a given industry. Basu and Fernald (1997) concluded that a typical two-digit industry in the United States appears to have constant or slightly decreasing returns to scale. Thus, when we look at empirical work, it appears that CRTS may not be an unrealistic assumption to make. 4. Firms in the two areas are unequally factor intensive, or KY/LY is greater than KZ/LZ (Y relatively capital intense) at any given feasible factor–price ratio. This assumption, along with CRTS, creates a production-possibility frontier that is uniformly concave to the origin. 5. The model allows for unemployment to exist.5 This is consistent with the way in which EZs are designated on the basis of certain distress criteria that includes high unemployment. LZU is the number unemployed in the EZ. The total labour force in the EZ area is LZ LZU NZ. The model also allows for unemployment to exist in the NEZ (let’s say, the natural rate of unemployment), but it is less than that in the EZ so that it does not qualify for EZ designation. In the NEZ, the total labour force NY is equal to LY LYU. LYU is the number unemployed in the NEZ. The total labour force in the economy is NZ NY N. 6. Factors are paid according to the value of their marginal products in equilibrium. Taken along with the assumption that fLZ fLY, this assumption means that wages in the EZ in the initial equilibrium are lower than in the NEZ. Capital markets, however, are in equilibrium, meaning that although different amounts of capital are used, the price of capital is the same in the two areas in the initial equilibrium. 7. The government provides a subsidy on the use of capital as well as labour in the EZ. This is equivalent to a refund of taxes paid on capital and labour to firms in the EZ. 8. The prices of goods are defined such that the value of all goods in the original equilibrium is one.

2.5 2.5.1

The model Disequilibrium in the labour market

The disequilibrium in the labour market of the EZ is the result of its high unemployment rate, which, by definition, means that a large proportion of its labour force is unemployed. Butler (1981), one of the pioneers of the EZ concept in the United States, argued that minimum wage legislation is the primary cause of unemployment in the United States and that, therefore, relaxing this minimum wage

Analytical Framework 23

constraint in the areas designated as EZs would alleviate the level of unemployment. However, this still does not explain why unemployment is concentrated in certain areas that become designated as EZs. Clearly, then, the literature on EZs does not provide a model of unemployment in the EZs, that is, does not explain the cause of unemployment in EZs. This model makes an attempt to provide that explanation. Here I elaborate on the reasons for individual unemployment in the EZ and the rate of unemployment in the area. 2.5.2

A model of unemployment

The model of unemployment that is developed here derives from the neoclassical assumption of high reservation wages. This is the explanation that individuals are unemployed if their reservation wages exceed market wages. Such an explanation appears reasonable for individuals in the EZ because of their high reservation wages relative to the market wage prevailing in the EZ. The high reservation wages in the EZ is realistic because of the benefits of remaining unemployed in the United States (Feldstein, 1978).6 The income of the unemployed is high because they are most probably recipients of unemployment compensation and probably other non-market income such as welfare payments. Given their level of skills, these individuals are unlikely to find gainful employment. This causes them to remain unemployed. The literature on job search indicates that high reservation wages relative to market wages cause unemployment, especially when unemployment benefits are generous. Feldstein (1978) shows how a combination of a high marginal tax on earnings and no tax on unemployment compensation makes the private cost of unemployment small and causes an individual to remain unemployed. To understand this effect, consider a situation in which each job searcher faces a distribution of wage offers with wr (his reservation wage) E(w) (his expected wage) ws (the maximum wage given his skills). This may be graphically shown in Figure 2.1. In Figure 2.1, the horizontal axis is the wage rate w(s) (wage given skills). The vertical axis is a frequency (refers to the number) of job offers available at the various wage rates. In the middle of the curve is shown the highest frequency and it decreases as we move away from the mean/median/mode. The curve that is obtained is the probability density function of the random variable, wage. It is normally distributed with mean at E(w). Given that the person finds a job, his/her expected wage is the weighted average of the job offers in the w(r) to w(s) range. This average wage is denoted by E(w) in Figure 2.1, which shows that the mean is also the median and the mode (which is characteristic of normally distributed random variables). We know that the individual would reject any job offer that offers w wr. Thus the area under the curve between wr and ws represents the probability

Frequency of job offers

24 Incentives for Regional Development

0

wr

E(w)

ws

Wage rate Figure 2.1 The distribution of wage offers

of the job searcher finding an acceptable job in any period (Ehrenberg and Smith, 1994). The higher this probability, the lower is the expected duration of unemployment. The benefits (such as unemployment compensation) of remaining unemployed essentially decrease the area under this curve by increasing wr (moves it further to the right) and lengthens the duration of unemployment. The unemployed in high-unemployment areas usually place a high value on the importance of having a job when compared to those in lowunemployment areas, as Bartik (1991) and Theodossiou (1992) argue, in the absence of policies that increase incomes of the unemployed (such as unemployment compensation). If such policies were to exist, they would only be willing to take up a job that offers income higher than their current income. But they are less mobile because of psychological ties to the area and the costs of relocation. These are also the ones with lower skills, less educated and likely are recipients of welfare payments.7 Therefore such individuals (who have high reservation wages relative to low market wages) tend to be concentrated in the EZ area. The market wage is low because profit-maximizing employers are willing to pay a wage only according to the skill of the workers. We know from assumption two of the model that fLZ fLY, that is, the marginal product of labour in the EZ is less than in the NEZ area. The literature on EZs points to the blight in these areas, which, according to this model, is due to the low capital–labour ratio in the EZ. The low capital–labour ratio leads to low productivity in the EZ for those employed and potential low productivity for

Analytical Framework 25

those who are unemployed. So the EZ labour force is eligible only for a lower market wage. This model offers an explanation as to why individuals with high reservation wages relative to low potential market wages tend to be concentrated in the EZ area. So prior to the designation of the area as an EZ, profitmaximizing employers do not have incentives to hire workers with low skills. It is also likely that these individuals are recipients of unemployment compensation, and are unlikely to find gainful employment given their skills, even though they do place a high value on the importance of having a job. So formally, the unemployment status (USij) of an individual i living in the jth area is determined by the difference between reservation and market wage: USij f(wijr wij)

(2.1)

where wijr is the reservation wage and wij is the market wage of the iith individual living in the jth area. The unemployment rate of the jth area may be summed over the unemployment status of i individuals living in the jth area.8 It may be written as follows:

Uj

兺 US i

ij

Population

(2.2)

Substituting for USij from (2.1):

Uj

兺 f(w i

r ij

wij)

Population

(2.3)

Thus, the aggregate unemployment rate in the jth area is dependent upon the extent to which reservation wages of individuals are higher than market wages in the aggregate. Specifically, (2.1) and (2.2) show that the unemployment rate of the EZ is high if the reservation wages of individuals are high relative to the market wage in the area. Thus, the high unemployment rate in the EZ is due to the low productivity of labour and their relatively higher reservation wages compared to the market wage. Thus, under conditions of high unemployment, EZ designation of the area acts as an important place-oriented policy to improve the blighted area in which tax abatements are provided to arrest the decline of the area. Tax abatements provide firms with incentives for investment. In section 2.5.3, I show how the tax abatement in the EZ induces in-migration of capital into the EZ due to the shifting of resources to the EZ area from the NEZ. I elaborate

26 Incentives for Regional Development

on the general equilibrium effects of tax abatement provided to firms in the EZ, first explaining the relationship between the reservation wage and the local unemployment rate and the effect of new employment on economic rents in the EZ. Next, in this section, I explain the relationship between the reservation wage and the unemployment rate. The reservation wage It is useful to recognize in a model that allows for unemployment that the standard environment in which job search is modelled involves the search for a job from a known distribution of job offers. Various models of job search ( Jones, 1989; Addison and Siebert, 1979; Ehrenberg and Smith, 1994) show that the unemployed individual’s decision to work is determined by the costs and benefits of unemployment. The model here follows Jones (1989) closely. As in Jones (1989), let be the instantaneous probability of receiving a job offer and F(w) be the distribution of job offers, b be the benefit level while remaining unemployed, and c be a measure of the costs of being unemployed. From the viewpoint of the individual job seeker, the decision to work is a function of the benefits and costs (b c) of remaining unemployed. I assume that job seekers try to maximize their utility by maximizing the income they receive from a job. Such utility-maximizing behaviour leads to the equation: rV b c

冕 max{0,W(x) V } dF(x) ⬁

(2.4)

0

In (2.4), r is the interest rate, V is the present discounted value of being unemployed, (b c) refers to the net benefits of being unemployed. The latter part of the equation represents the capital gains derived from the income from a job, which is the maximum of the net income from a job, W(x), (which is made net of the cost of being unemployed, V) if employment is found or zero. In such a situation, the optimal job search strategy displays the reservation wage property with the critical value being the reservation wage wr (Zuckerman, 1984). The reservation wage, or the lowest wage at which the unemployed are willing to supply positive labour (accept a new job), obeys: rV rW(w r ) w r

(2.5)

Equation (2.5) shows that the reservation wage wr equals imputed search income.9 Substituting for rV from (2.5) in (2.4), as in Jones (1989):

( )冕 max(x w ) dF(x)

wr b c r

⬁

wr

r

(2.6)

Analytical Framework 27

Thus the reservation wage, as in (2.6), equals the net benefits (b c) while unemployed, and a factor that depends on the expected wage in next employment adjusted for the arrival rate of job offers. Then it is possible to approximate (2.6) in some linear fashion, as in Jones (1989): wir ␣0 ␣1(bi ci) ␣2␮i ␣3i ei

(2.7)

In (2.7), bi ci is the difference between benefits and costs of remaining unemployed for the ith individual, i is the expected wage in next employment, i is the arrival rate of job offers as defined earlier, and ei is a random error term. i is a function of the various individual-specific characteristics as well as regional labour market conditions that determine the arrival rate of job offers. So i 兺j ␣3j Xij

(2.8)

The unemployment rate of the area is an important indicator of the regional labour market conditions that determines the arrival rate of job offers for individual i and so of his/her reservation wage. So, substituting for i in (2.7): wir ␣0 ␣1(bi ci) ␣2␮i ␣3Uij ei

(2.9)

Uij is the unemployment rate of the jth area ( j Z, NEZ) in which the ith individual lives. The model in (2.9) is estimable. Among plausibly many others Jones (1989) and Haurin and Sridhar (2003) estimate this relationship. Chapter 3 also estimates this relationship. Chapter 5 uses reservation wage estimates from Sridhar (1998). Moreover, a hypothesis can also be formulated with regard to the relationship between the reservation wage and the area’s unemployment rate. The actual experience of job seekers in high-unemployment areas shows that they are willing to accept lower reservation wages for some reasons: it is more likely than not that they are risk-averse. Although it is reasonable to expect that unemployed job seekers, even in high-unemployment areas, frequently begin their search with a high reservation wage, as time passes, they are likely to lower their reservation wage for reasons of family or financial hardship (Theodossiou, 1992). Thus the reservation wage of unemployed searchers can be considered to be a gradually declining function of time spent in unemployment, which is long in high-unemployment areas. This could be the case because they have family/psychological ties to the area or they are unwilling to bear the social or financial costs of relocation. The testable hypothesis that comes out of this is that unemployment rate of the area has a negative impact on the reservation wage of individuals residing in the area.

28 Incentives for Regional Development

Economic rent As in standard labour economic theory, we can define economic rent bij as the extent to which actual wages are above the reservation wage. That is: bij wij wijr, j EZ,NEZ

(2.10)

where wij and wijr are respectively the wage and reservation wage of the ith individual in the jth area,10 which could refer to the EZ area (EZ) or non-EZ area (NEZ). If the hypothesized relationship between the reservation wage and the area’s unemployment rate were true, the reservation wage (wrij) would be low in the EZ because of its high unemployment rate.11 However, wages (wZ) in the EZ are also low (compared to the NEZ) due to the low capital–labour ratio in the EZ (assumption six of the model). If a similar hypothesis were to hold in the NEZ, it would have higher reservation wage (wYr ) than the EZ. Wages in the NEZ (wY) are also high, by assumption six. So it is difficult to conclude whether the EZ or NEZ will benefit from higher economic rents. Thus, whether economic rents would be higher in high- or low-unemployment areas is an open question that can be tested in the empirical work. In the following section, I show how the tax abatement in the EZ induces in-migration of capital into the EZ from outside. I elaborate on the general equilibrium effects of tax abatement provided to firms in the EZ, and the effect of new employment on economic rents in the EZ. 2.5.3 The general equilibrium response to the tax abatement Harberger (1962) considers the effect (and the ultimate incidence) of a sector-specific corporation income tax, taking into account its general equilibrium effects on the entire economy. His analysis has become a standard framework for analyzing different kinds of taxes in the literature. I adopt his analytical framework in order to analyze the effect of the property tax abatement provided to firms in EZs that is equivalent to a tax cut on the use of capital in the EZ. In addition to a subsidy to capital, most state EZ programmes include a subsidy for labour. Reflecting this, the model includes a subsidy to labour. An equal subsidy for labour and capital is the same as a subsidy for the good produced by the sector. Therefore I decompose the subsidy to capital and labour into a subsidy for one input (that is, the tax abatement which is a subsidy to capital), and a subsidy for the output (produced by the EZ area). The subsidy for the output results in a reduction in the price of the good produced by the EZ area, which becomes (PZ CZ), where PZ is the price of the output, and CZ the subsidy on the output.

Analytical Framework 29

According to Harberger’s analysis, the tax abatement is equivalent to a tax cut which will have the immediate effect of creating a wedge between the price of capital between the EZ and the rest of the economy. Users of capital in the EZ pay a lesser price (PK AKZ) for the use of capital, where AKZ is the tax abated on capital invested in the EZ. The tax abatement has the effect of lowering the price of good Z that EZ firms produce because investors in the EZ can now produce and supply higher output at the same cost as before, because of the savings induced by the abatement. Investors in the NEZ firms still continue to pay the price on capital, PK, which includes the tax (there are no taxes abated in the NEZ). At this point, the NEZ firms will increase their supply of good Y by an amount sufficient to lower the price of Y on par with that of Z. Whether or not the NEZ firms are able to do this depends on the demand elasticity for the good (Z) produced by the EZ firms. The percentage change in the demand for Z as a result of the tax abatement, is formally derived below. The goods market: demand In a two-good economy, one would expect the demand for goods to have some price elasticity and that Z and Y would be substitutes. To see this, let us characterize the demand equation for Z as follows, following Harberger: Z f

冢P

Z

CZ PY

冣

(2.11)

Equation (2.11) shows that the demand for Z is a function of the price of Z and Y. Totally differentiating the demand for Z, dZ

⭸f

冢

PZ CZ ⭸ PY

PYdPZ PYdCZ (PZ CZ)dPY

冣

P2Y

(2.12)

In order to express the change in percentage form, we divide (2.12) by Z:

冤

冥

dZ ⭸Z 1 dPZ dCZ (PZ CZ) dPY Z PY PZ CZ Z PY P2Y ⭸ PY

(2.13)

By assumption 8 of the model, the price of goods (PZ CZ) and PY, are equal to one, and (2.13) can be written as: PZ CZ dZ Z [dPZ dCZ dPY] PY Z PZ CZ Z PY

(2.14)

30 Incentives for Regional Development

In (2.14), it may be noted that {[⭸z/⭸((PZCZ)/PY)] [((PZCY)/PY)/z]} is the demand elasticity for Z in terms of the relative prices (PZCZ/PY). So in simplified form, (2.14) may be written finally as: dZ ␧(dPZ dCZ dPY) Z

(2.15)

where ␧ {[⭸z/⭸((PZCZ)/PY)] [((PZCY)/PY)/z]} Equation (2.15) shows that the percentage change in the demand for Z depends on elasticity of demand for Z in terms of relative prices. It should be noted that the elasticity appearing in (2.15) is compensated because the marginal propensity to consume of households and the government are assumed to be equal so that the income effect in consumer demand exactly offsets that in government demand (Myles, 1995). Thus only the substitution effect is present. If | | 0, the demand for Z (produced by the EZ firms) increases more than proportionately in response to a decrease in its price. The goods market: supply For market clearance, the percentage change in demand for Z must equal the percentage change in the supply of Z. To determine the percentage change in the supply of Z, we totally differentiate the production function Z f (KZ, LZ):

dZ

⭸f ⭸f dK dL ⭸KZ Z ⭸LZ Z

(2.16)

Expressing (2.16) in percentage change form, and dividing the LHS and RHS by Z and f respectively, ⭸f ⭸f ⭸K ⭸L dZ Z Z dKZ dLZ Z Supply f f

冏

(2.17)

Multiplying and dividing through by KZ and LZ ⭸f ⭸f KZ LZ ⭸K dK ⭸L dLZ dZ Z Z Z Z Supply KZ LZ f f

冏

(2.18)

Analytical Framework 31

Equation (2.18) can be written as: dZ Z

冏

Supply

␪KZ

dKZ dLZ ␪LZ KZ LZ

(2.19)

In (2.19), it may be noted that ⍜KZ is the share of capital income in the value-added, ⍜LZ is the share of labour income, in the value added for firms in the EZ. Equation (2.19) shows that the percentage change in the supply of output by EZ firms depends upon the shares of labour and capital income in the value-added for EZ firms. The factor market: firms’ demand for factors Because of a shift in demand for good Z, increased profits accrue to EZ firms that produce Z. So the NEZ capital will have an incentive to flow into the EZ, in order to equalize returns to factors in both the areas. What happens then depends on relative factor intensities, and the firms’ demand for factors in the two areas. Following Harberger, changes in factor demands can be specified in terms of their direct elasticities of substitution. So I define elasticities of substitution, as in Tresch (1981): K 冢 L 冣 S f d log冢冣 f K d log冢冣 L S f d log冢冣 f d log

Z

Z

(2.20)

z

LZ

KZ

Y

Y

(2.21)

Y

LY

KY

SZ and SY are, respectively, the direct elasticity of substitution between capital and labour, of firms in the EZ and NEZ. With the marginal products in the EZ and NEZ firms equal to their respective price ratios, we can write:

冢f 冣 d log冢P

d log

fLZ

K

KZ

AKZ PLZ

冢f 冣 d log冢PP 冣

d log

fLY

K

KY

LY

冣

(2.22)

(2.23)

32 Incentives for Regional Development

Then, substituting for the ratio of marginal products from (2.22) in (2.20), we have:

冢KL 冣 S d log冢P Z

d log

AKZ PLZ

K

Z

Z

冣

(2.24)

Similarly,

冢KL 冣 S d log冢PP 冣 Y

d log

K

(2.25)

Y

Y

LY

Consider the LHS of equation (2.24),

冢KL 冣 K1 d冢KL 冣 K1 冤L dK L K dL 冥 Z

d log

Z

Z

Z

Z

Z

LZ

Z

Z

Z

2 Z

Z

LZ

冢dKK 冣冢dLL 冣 Z

Z

Z

(2.26)

Z

Similarly for the price ratios,

冢P

d log

K

冣

AKZ dPK dAKZ dPLZ PLZ PK AKZ PLZ

Because PK PLZ 1,

冢P

d log

K

冣

AKZ dPK dAKZ dPLZ PLZ 1 AKZ

(2.27)

Substituting (2.26) and (2.27) into (2.24) gives us:

冢dKK 冣冢dLL 冣 S 冢dP1 AdA Z

Z

K

KZ

Z

Z

Z

dPLZ

KZ

冣

(2.28)

When we do this similarly for Y, we get:

冢dKK 冣冢dLL 冣 S (dP Y

Y

Y

Y

K

dPLY)

(2.29)

Y

Equations (2.28) and (2.29) show that the percentage change in the demand for capital over labour by firms depends on the elasticity of substitution between the factors as well as change in their prices. Thus, the relative growth of demand for capital and labour will differ. While factor markets

Analytical Framework 33

continue to equate factor prices with values of marginal products, equilibrium is attained only when factor prices (wages and returns to capital) are equalized across the areas. As capital continues to flow into the EZ, demand for labour increases, as described above. This increases employment (LZ) in the EZ. However, it has to be noted that first-order effects always dominate. This means that firms that eventually locate in the EZ are more capital-intensive, although they also have increased demand for labour.12 Note that the subsidy to capital in the EZ equates the capital–labour ratios across the two areas now. With increasing employment in the EZ, there is a high probability that local residents will fill the jobs created by the new capital. This is because new jobs in the EZ are assumed to be allocated to unemployed on a random selection process. When the selection process for jobs is random, zone residents (even though they have few skills) have an equal chance of being selected from the pool of unemployed, by employers who are now subsidized. The assumption of a random selection process in the EZ when firms create jobs is supported by empirical evidence from the United States. Empirical work from the US reports that, on average, about 50 per cent of jobs created in the EZ go to zone residents. In a collection of ten case studies of statedesignated EZs, the (US Department of HUD, 1986) observed that 70 per cent of jobs created each in the Bridgeport EZ, CT and the Chicago EZ, IL, 46 per cent in the Macon EZ, MO, 19 per cent in Michigan City, IN, 30 per cent in the Tampa zone (FL) and 5 per cent of the York zone (PA), were held by zone residents. In the Louisville (Kentucky) zone, it was found that 31 per cent of the jobs created were held by persons who were either lower income or zone residents. Erickson and Friedman (1989), based on a survey of local EZ coordinators conducted by the US Department of HUD, found that the mean share of jobs held by zone residents was over 61 per cent, with a median of over 68 per cent. Data from Immergluck (1997), from the Chicago EZ, indicated that the barriers between EZ residents and jobs were dependent on some factors. He found that local employment was much higher in Latino parts of the zone and in African American neighbourhoods where there were more public sector jobs, very small firms and few manufacturers. Such data are sparse in the context of developing countries such as India. I report some evidence of this in Chapter 7 from India’s growth centres. Thus, when a substantial portion of the jobs that are created are held by zone residents the unemployment rate of the EZ falls. With an increasing use of capital and an increase in the capital–labour ratio, both labour productivity and industrial output rise. Some explanations for rising industrial output (apart from the increasing use of capital) could be due to improved managerial and organizational capabilities and enhanced labour efficiency through on-the-job training programmes in the EZ area.

34 Incentives for Regional Development

Thus, at the optimum, the unemployment rate in the EZ is lower than in the initial equilibrium, and that in the NEZ is the same as before. This is because of firms’ increasing demand for labour, made possible by the rise in the capital–labour ratio (resulting from the subsidy to capital), and the rising productivity of labour in the EZ. We have to note that, by assumption, the NEZ operates at a natural rate of unemployment. Tobin (1972) estimated that a 5 to 6 per cent natural rate of unemployment has been associated with more than 20 per cent excess capacity in the capital stock. Thus the EZ acts only as a tool to direct the excess capacity of capital in the full-employment areas to high-unemployment areas and serves to reduce their unemployment rate. The model says that capital moves across areas; this does not necessarily imply the migration of firms. But the movement of capital causes an improvement in the total economy’s overall well-being, although a change in the price of capital initiated in one area due to the existence of the EZ, is transmitted to other areas.

Market clearance For market clearance, the goods and factor markets have to remain in balance. The following has to be true for the goods market to be in equilibrium: dZ Z

冏

Demand

dZ Z

冏

(2.30) Supply

Because capital and labour are in fixed supply, the amounts of their inmigration into the EZ must be equal to the amount of their out-migration from the non-EZ region so that: dKZ dKY

(2.31)

dLZ dLY

(2.32)

The above conditions show that what capital and labour the EZ gains must be equal to the amounts of the factors lost by the non-EZ area. Comparative statics: change in capital and labour in the EZ In order to obtain the change in the capital invested (dKZ) and labour (dLZ) in the EZ at the equilibrium as in Harberger’s model, I follow the procedure described in Myles (1995) most closely. After some manipulation, we end up with a system of three simultaneous equations that can be solved simultaneously for dKZ, dLZ, and dPK.13 The simultaneous system that was used to solve for dKZ, dLZ, and dPK is shown below in matrix form.

Analytical Framework 35

冤

fKZ KZ 1 KZ 1 KY

fLZ LZ 1 LZ 1 LY

冥冤冥冤

␧(fKZ fKY) SZ (1 AKZ) SY

dKZ dLZ dPK

冥

␧dAKZfKZ SZ SZdPLZ (1 AKZ) SYdPLY

The simultaneous system, when solved, gives the solution for dKZ, dLZ, and dPK. The exact expressions obtained for dKZ and dLZ (which are of greater interest, because one of the important goals of EZs is to promote employment and investment growth in the designated areas) are quite cumbersome and so I do not report them here. They are reported in the Appendix to this chapter. It is sufficient to note that the change in capital invested and labour, dKZ and dLZ, depend on the relative magnitude of certain parameters in equilibrium. The framework provided above indicates that the effect of the tax abatement on dKZ, and dLZ is testable. More generally, the Harberger analysis, when applied to property tax abatements in EZs, indicates that the capital and employment impact of the tax cut on capital in the EZ depends on three sets of parameters: (i) Relative factor intensities of firms in the two areas; (ii) The elasticity of substitution between capital and labour in firms in the areas; (iii) The price elasticities of demand for goods Z and Y produced by EZ and NEZ firms respectively. The analysis also indicates that it is impossible to isolate the incidence of the tax cut given in the EZ and restrict it to the EZ alone. Because competitive factor markets equalize returns to capital everywhere in the economy, if investors in the EZ enjoyed an increase in the return to capital, investors everywhere will also experience the same increase. Moreover because goods and factor markets are interdependent, the changes could get transmitted to consumers in the form of changes in goods prices. Thus, Harberger’s analysis describes how the migration of capital occurs in response to changes in the rate of return to capital across areas. The general equilibrium response to the tax abatement is shown in Figure 2.2. Figure 2.2 shows that the tax abatement (tax cut) on capital leads to in-migration of capital into the EZ. Depending on the elasticity of substitution between capital and labour (SZ), the new capital leads to increase in productivity, wages and employment. This affects the unemployment rate. The unemployment rate finally determines the reservation wage. The difference between wages and reservation wages determines economic rent.

36 Incentives for Regional Development

Enterprise zone (EZ)

Rest of economy (non-EZ)

Reservation wage (w rZ)

Reservation wage (w rY)

Unemployment rate (UZ)

Unemployment rate (UY) Economic rent

Employment (LZ)

Employment (LY) Migration of capital (dKZ)

Wages (wZ)

Elasticity of substitution between capital and labour (SZ)

Price of capital less tax abatements (PK – AKZ)

Capital in-migration

Wages (wY) Capital out-migration Elasticity of substitution between capital and labour (SY)

Price of capital (no abatements) (PK)

Figure 2.2 The impact of the EZ (tax incentives) with capital mobility

More specifically, the testable hypotheses that emerge from the theoretical model are: (i) The relationship between the reservation wage and unemployment rate; (ii) The determinants of area unemployment rate being the EZ (or the existence of other incentive programmes), the duration of the programme’s existence and labour in-migration into the area. It may be noted that the context for this test is laid out in Figure 2.2. A reduction in the price of capital through tax abatements (that occurs only in EZ-designated or incentive areas) increases the capital–labour ratio in the EZ and increases labour productivity, raising wages and employment and affects the unemployment rate. Thus, zone designation, and the duration of the zone’s existence, affect the area’s unemployment rate, according to the model. Thus, one of the testable hypotheses that emerges from the model is that in an empirical examination of the unemployment rate, we include a dummy for (tax or infrastructure) incentive programmes, and the duration of the zone’s existence, in addition to other controls as exogenous variables. This is empirically tested in Chapters 4 and 6.

Analytical Framework 37

In the empirical work, one could also empirically test other hypotheses that emerge from the theoretical model. The question as to whether or not dAjz dbjz, that is, whether the cost of the abatement (EZ programme) exceeds the economic rent in the EZ, at equilibrium, forms the basis for the benefit–cost analysis in the empirical work reported in Chapters 3 and 5.

2.6

Predictive power of the model

It is a widely accepted view that a model should be judged by the accuracy with which it can predict what we observe, as well as the realism of its assumptions. The predictive power of the model that is presented here depends upon the realism of the assumptions. First, the assumption of disequilibrium in the EZ before EZ designation is consistent with what we observe. EZs are in reality abandoned areas with high unemployment rates (when we take into account the designation criteria in most of the state EZ programmes). Hence for persons in the EZ, the job search behaviour characterized by low reservation wages seems realistic due to reasons of family, other social/psychological reasons or the costs of relocation. Further, the general equilibrium response to the tax abatement is to be expected because of the inherent mobility of capital, in response to changes in its price, until it is equalized across areas. Further, the effects could also be transmitted to the goods market in both areas. Thus, the analysis indicates that it is impossible to isolate the incidence of the tax abatement to the EZ alone, which mirrors the real-world situation. A simple example of this in the model is the movement of capital. The model indicates that the migration of capital (although not necessarily of firms) serves to lower the unemployment rate in the EZs, without changing that in the other areas. Thus, when we begin with realistic assumptions regarding initial conditions in the EZ and study the implications of a tax abatement in the context of a theoretical framework, we obtain a set of hypotheses that seem both plausible and testable. Given these facts, the model that is presented here is of sufficient generality that it is applicable to most place-oriented programmes. In the forthcoming chapters, I provide empirical evidence of the effectiveness of such place-oriented programmes.

Appendix Solutions for change in employment and investment in EZ A.1 Solution to the system of equations The simultaneous equations are in matrix form, and Cramer’s rule is used to solve for dKZ and dLZ. So dKZ

兩D1兩兩D兩

38 Incentives for Regional Development

where |D| is the determinant of the matrix and |D1| is the determinant of the matrix with its first column replaced by the d vector. So

冤

冥

SYdPLY SZfLZ ␧(fKZ fKY)

LZ 1 AKZ

冤

冥

␧ S dA f SZ(fKZ fKY) (dPLZ(1 AKZ) dAKZ) LY(1 AKZ) Z KZ KZ dKZ

冤

冥

SY␧dAKZfKZ fLZSYSZ dPLZ(1 AKZ) dAKZ LY LZ 1 AKZ

冤

冥 ␧(f

SZfKZ 1 1 (1 AKZ) KYLZ LYKZ

KZ

冤K 1L

fKY)

Y Z

冥

冤

冥

fKZ fLZ 1 SZ LYKZ KZLZ

In a very similar way, dLZ

兩D2兩兩D兩

where |D| is the determinant of the matrix as before and |D2| is the determinant of the matrix with its second column now replaced by the d vector. The expression obtained for dLZ turns out to be: 1 [SZ(␧dAKZfKZ (dAKZ dPLZ(1 AKZ) ␧(fKZ fKY)))] KY(1 AKZ) SYdPLYfKZ dLZ

冤L (1S A Z

Z

冤

KZ)

冥

冥 ␧(f

SKZfKZ 1 1 (1 AKZ) KYLZ LYKZ

冤

冥

SYSZfKZ dAKZ dPLZ(1 AKZ) ␧(1 fKY) KZ LZ (1 AKZ) KZ

冤K 1L

fKY)

Y Z

冥

冤

冥

fKZ fLZ 1 SZ LYKZ KZLZ

3 Competition Among American States: Evidence from Illinois Enterprise Zones

3.1

Introduction

Until recently, supply-side regional development policy, consisting of financial and tax incentives, credits, and abatements, was believed to be ineffective. First, such policy was thought to have no effect on firms’ location decisions; secondly, it was argued that such policies were zero-sum games that redistributed jobs between places, while leaving the total number of jobs within the economy unchanged. The previous chapter developed an analytical model to study the impact of policies such as EZs not only on the economy of the region in which it is situated, but also on the rest of the economy. In this chapter, we examine empirically whether such policies produce net benefits and have favourable equity implications. Using data from the EZs of Illinois in the Midwestern United States, this chapter answers, from a local and state perspective, the challenge raised against the zero-sum nature of supply-side regional development policy. Illinois is selected because it is one of the most important states in the Midwestern United States. Furthermore, Illinois was one of the first states in the United States to adopt the enterprise zone legislation once the EZ concept had been imported from the United Kingdom in the 1980s. The previous chapter showed why reservation wages can be expected to be lower in high-unemployment areas. This chapter will focus on whether tax incentives can produce positive net benefits and have equity implications. Specifically, this chapter examines the following questions: Are benefits greater than costs when state and local governments grant tax incentives to businesses? Are the benefits from tax incentives greater in highunemployment areas, because of low reservation wages, when compared to low-unemployment areas? While the first question addresses efficiency considerations, the second question takes into account the redistributive implications of supply-side policy. 39

40 Incentives for Regional Development

The chapter is organized as follows. First, the theory behind the research is reviewed, along with the assumptions of the research; then an introductory section on EZs is followed by a description of how work in this chapter differs from earlier literature. Since EZs in Illinois have been chosen for the analysis, one section describes the Illinois programme. This background of the study area is provided in order to facilitate a broader understanding of the benefit–cost analysis of the programme presented in the chapter. The Illinois background section is followed by measures of costs and benefits chosen for the analysis, the research methodology, and the results from the regression and benefit–cost analysis. The chapter concludes by summarizing the results, drawing implications, and indicating areas for further research.

3.2

Theoretical framework

As discussed in Chapter 1, Bartik (1991) challenges the conclusion that supply-side policies were ineffective and had zero-sum effects. The bulk of the empirical evidence reviewed by Bartik for the USA supported the contention that supply-side, cost-reducing incentives had measurable effects on business location decisions, contrary to earlier evidence. Bartik’s econometric studies indicated that the long-run elasticity of business activity with respect to state and local taxes is in the range of 1.0 to 3.0 for intra-metropolitan location decisions (Bartik, 1991: 43). That is, if a small suburban jurisdiction within a metropolitan area raised its taxes by 10 per cent, it could expect in the long run a reduction in its business activity by 10 to 30 per cent. Bartik (1991) showed that a redistribution of jobs could result in net benefits to the locality adopting the policy, on grounds of equity. Bartik concluded that the benefits of redistributing jobs to high-unemployment areas were likely to be higher because of lower reservation wages, defined as the lowest price at which an individual is willing to work. The social benefits or economic rent from a job are equal to the wages paid minus the reservation wage. Since the reservation wage is inversely related to the value placed on having a job, in a high-unemployment area, many individuals who place a high value on getting a job will remain unemployed for a long time and will have low reservation wages. In such cases, the net benefit of a job – the wages paid minus the reservation wage – is high. Conversely, given the wages paid, an individual with a higher reservation wage will derive less benefit from a job. Thus, in a high-unemployment area, the benefits from a given number of jobs would be higher because of lower reservation wages and greater net benefits per job. Since this argument is plausible, this chapter provides empirical evidence of the proposition that the benefits of tax incentives are greater than their costs, which was also a testable hypothesis that emerged from the model developed in Chapter 2. In particular, this chapter provides some empirical evidence to examine whether the benefits are greater than costs in

Evidence from Illinois Enterprise Zones 41

high-unemployment areas because of low reservation wages, from the EZs of Illinois, one of the most important states in the Midwestern US. 3.2.1

Assumptions

One basic assumption underlying the research questions to be answered in this chapter is that tax incentives affect decisions about business location; a related assumption is that all location factors, with the exception of zone incentives, are similar among sites that firms are considering. In other words, it is assumed that employment generated in these zones is attributable to zone incentives. Further, it is assumed that jobs created in the zones will be filled by the local unemployed, primarily to justify local expenditure (a different version of this assumption is used in Sridhar (1998), a study which evaluates Ohio’s programme). Even though this assumption may not always be realistic, and many jobs from local growth are held by in-migrants moving between areas, as Bartik (1991) demonstrates, local job growth affects the unemployment rate and labour force participation rate in a persistent manner.1 Therefore, it can be expected that the jobs created in the zone will have some effects on the local unemployed. Given these assumptions, the research questions pertaining to supply-side policies are studied with respect to EZs in the state of Illinois. Supply-side policies, as opposed to new-wave policies, as highlighted in Chapter 1, consist of financial incentives such as industrial revenue bonds (IRBs), property tax abatements, direct state loans, and non-financial incentives (such as customized industrial training), which are primarily targeted at branch plant recruitment. Since EZs offer incentives targeted at plant recruitment, they may be considered as representing traditional – or supply-side regional – development policy. Selecting zones within a single state controls for variation in enabling legislation and the general climate surrounding the package of incentives (Reese, 1991). Furthermore, EZs in Illinois (IL) are designated on the basis of diverse distress criteria relating to income, population, and poverty, in addition to unemployment. So not all zones in Illinois are necessarily highunemployment areas. If it were, indeed, the case that all zones were highunemployment areas, the sample would be too homogeneous to test the proposition that competition encourages growth in these areas. Since zones in Illinois are either high unemployment or high-poverty, low-income or high-population loss areas, the choice of these zones would be varied enough to test the hypothesis of research in this chapter.

3.3

Description of EZs in the United States

There are two separate sets of EZ programmes that operate in the US – one at the federal level and the other at the state level. The federal programme initially

42 Incentives for Regional Development

added 11 empowerment zones and 99 communities to the list of state zones (Peters and Fisher, 2002). The United States Department of HUD re-energized the Federal Renewal Communities (RCs), Urban Empowerment Zones and Urban Enterprise Communities (ECs) Initiative in December 2001 by designating 40 urban and rural RCs and eight new urban ECs. These new designees were able to use a $22 billion tax incentive package to open new businesses, provide thousands of new jobs, and rehabilitate and build new housing, in these urban and rural areas located throughout the United States. These are federally designated areas and are administered by the US Department of HUD. Around 35 of the 50 states in the United States have state enterprise zone legislation that enables them to target their distressed areas for growth with the use of incentives. In this and forthcoming chapters dealing with empirical work relating to place-oriented programmes such as EZs and GCs, the focus is the state-administered programmes (not programmes that are centrally or federally administered) because they allow enough flexibility and variation to states in administering the programmes. Table 3.1 summarizes the studies carried out to date on state EZs. While providing a summary and critical overview, the table demonstrates that one question left unanswered by the literature thus far is whether EZs have favourable equity implications because of low reservation wages. So far, in the literature, benefits from jobs created in or redistributed to EZs have not been viewed in the light of reservation wages. Table 3.1 Summary of selected studies on enterprise zones Study

State/Zone

Measure

Technique

Finding

Keefe, 2004

California

Employment growth

Propensity score matching model

Positive

Lambert and Coomes, 2001

Louisville EZ, KY

Jobs and capital formation

Shift-share

High programme costs and low economic benefits

Boarnet and Bogart, 1996

New Jersey

Employment and municipal property values

Econometric evidence

Negative

Dowall, 1996

California

Jobs and business investment

Shift-share analysis and survey

Zone incentives have done little

Logan Jr and A. Barron, 1991

Florida

Jobs, housing

Descriptive statistics

Inconclusive

Rubin, 1991

New Jersey

Tax revenue from investment

I–O analysis

Positive

P. Grasso and S. Crosse, 1991

Maryland

Employment growth

Interrupted time series

Inconclusive

Elling and Sheldon, 1991

IL, IN, KY, OH

Ability of zones to attract firms

Multiple regression

Modest

L. Papke, 1991

Indiana

Jobs

Cost/Job

Positive

Continued

Evidence from Illinois Enterprise Zones 43 Table 3.1

Continued

Study

State/Zone a

Measure

Technique

Finding

Dabney, 1991

See note

Location

ANOVA

Negative for high-tech

Rubin and Wilder, 1989

Evansville, IN

Comparative advantage of the zone

Shift-share

Positive

US GAO, 1988

Maryland

Employment levels

Interrupted time series

Negative

CA OAG, 1988

California

Employment growth

Descriptive statistics

Positive

US HUD, 1986

Bridgeport, CT

Jobs

Local officials’ assessments

Positive

US HUD, 1986

Chicago, IL

Unemployment

Local officials’ assessments

Modest

US HUD, 1986

Dayton, OH

Unemployment rate

Local officials’ assessments

Little change

US HUD, 1986

Louisville, KY

Jobs

Local officials’ assessments

Positive for small firms

US HUD, 1986

Macon, MO

Investment, jobs

Local officials’ assessments

Positive

US HUD, 1986

St Louis, MO

Jobs to zone residents

Local officials’ assessment

Negative

US HUD, 1986

Michigan City, IN

Potential success

Local officials’ assessments

Modest

US HUD, 1986

Tampa, FL

Firm decisions to invest

Local officials’ assessments

Modest

US HUD, 1986

Thief River Falls, MN

Location decisions

Interview with firm representatives

Modest

US HUD, 1986

York, PA

Location decisions

Local officials’ assessments

Negative

Note: a. The study by Dabney examines eight EZs in different states. The locations are Baltimore, Maryland; Dayton, Ohio; Decatur, Illinois; Louisville, Kentucky; Norwalk, Connecticut; Philadelphia, Pennsylvania; St Louis, Missouri; and Topeka, Kansas.

This chapter analyzes the relationship between benefits from jobs created (or redistributed) in EZs and their reservation wages. Using a cross-section of the Panel Study of Income Dynamics (PSID) for 1987, I estimate reservation wages and net benefits from jobs created in the zones of Illinois, and compare them to the costs of the programme.

3.4

Description of enterprise zones in Illinois

At the time research for this chapter was completed, the Illinois Department of Commerce and Community Affairs (DCCA) was administering the EZ programme. Since this time, the name of the DCCA has been changed to the

44 Incentives for Regional Development

Department of Commerce and Economic Opportunity (DCEO). All Illinois municipalities and counties, individually or in combination, may apply for enterprise zone designation. EZs in Illinois can vary in size from 0.80 square kilometres (half a square mile) to 24 square kilometres (15 square miles). In 2004, the state has 95 zones (at the time the research was completed, there were 90 zones). There are some size and distress criteria an area must meet in order to be designated an enterprise zone in Illinois. 3.4.1

Size criteria for enterprise zone designation

The following are criteria for an area to be designated an EZ in this state. The criteria below are based upon amendments to the relevant Act, as found in Section 520.210, effective 14 February 2003 (obtained from the Illinois DCEO): (i)

(ii)

An enterprise zone must be a minimum of 0.80 square kilometres (onehalf square mile) and may be up to 24 square kilometres (15 square miles), excluding lakes and waterways. Where the EZ is a joint effort of three or more units of government, or two or more units of government, if located in a township divided by a municipality of 1,000,000 or more inhabitants, and where the certification has been in effect at least one year, the minimum is 0.80 square kilometres (one-half square mile) and the maximum is 21 square kilometres (13 square miles), excluding lakes and waterways. The area must be contiguous, but should not include wholly surrounded territory within its boundaries. In terms of coverage, the areas must: (a) be entirely within a municipality; or (b) be entirely within the unincorporated areas of a county, except where reasonable need is established for such zone to cover parts of more than one municipality or county; or (c) comprise all or part of a municipality and an unincorporated area of a county.

3.4.2

Distress criteria for zone designation

There are some distress and other criteria an area should also meet in order to be eligible to apply for enterprise zone designation in Illinois. A proposed EZ must satisfy at least one of four distress criteria concerning poverty, unemployment, low income, or population loss: (i)

(ii)

Unemployment: The unemployment criterion is met if the zone has an annual average unemployment rate of at least 120 per cent of the state’s average unemployment rate for the 12 months ending the previous 30 June. Anyone who is not presently employed and/or has exhausted all unemployment benefits is considered to be unemployed, for purposes of calculating the unemployment rate. Poverty: The poverty criterion is met if the poverty rate for each census tract, minor civil division or county civil division that contains any part

Evidence from Illinois Enterprise Zones 45

of the area proposed as an Enterprise Zone is at least 20 per cent, as of the 2000 Federal Census. The poverty rate is computed using the number of persons in families or who reside together as unrelated individuals who had incomes below the poverty threshold in the 2000 Federal Census, as a proportion of population. (iii) Low income: The low-income criterion is met if at least 70 per cent of the households in the zone have incomes equal to or less than 80 per cent of the median household income of the community in which the zone is located. (iv) Population loss: The population loss criterion is met if the enterprise zone suffered a population decrease of 20 per cent or more between 1980 and 2000, as determined by Federal Census data for those years.

3.4.3

Other criteria

Other criteria that were subsequently added to size and distress criteria for areas to be designated as zones in Illinois, are described below: (i) Job creation: The Illinois DCEO may designate an area as an Enterprise Zone when such designation results in the development of ‘substantial’ employment opportunities by creating or retaining a minimum of 1,000 full-time equivalent jobs due to an investment of $100 million or more, and help alleviate the effects of poverty and unemployment within or the vicinity of the zone. New units of government being added to an existing Enterprise Zone must qualify under the same qualification criteria as the existing Enterprise Zone. (ii) Closed military bases: A military base closed by the United States Department of Defense that has been properly designated as and is currently operating as a Local Redevelopment Agency. (iii) Although the DCEO does not require the applicant to use census geography boundaries as the boundaries for the EZ, census geography must be used to demonstrate how the area meets one of the eligibility criteria. The census geographies to be used have to be the smallest geographies for which data are available and which encompass the entire proposed Enterprise Zone. When an Enterprise Zone boundary splits a census tract, county civil division, or minor civil division, the rules specify that data for block groups or enumeration districts have to be entirely within the Enterprise Zone and those that include any part of the Enterprise Zone should be included in the calculation. The number of zones in Illinois increased from 90 in 1992 to 95 in 2004. This increase in the number of zones over the decade could be the result of the addition of new criteria pertaining to job creation or military bases, or more areas qualifying according to pre-existing criteria, relating to size

46 Incentives for Regional Development

or distress. It is interesting to note non-distress criteria such as jobs and closed military bases. Criteria such as ‘substantial job creation’ can be subjective and discretionary, and they suggest that potentially non-deserving areas could also secure zone designation and use tax incentives to attract firms to their area, at the cost of the exchequer. This study uses a sample of 68 zones for which data on unemployment were available at the time the research was completed. Of these 68 zones, 48 were high-unemployment areas, whose unemployment rates in 1990 were above 120 per cent of the average unemployment rate for the state during the year. The remaining zones were either average- or low-unemployment areas. Figure 3.1 shows a frequency distribution of the unemployment rates in the zones. The unemployment rate is normally distributed, with ten zones in the state having unemployment rates between 0 and 6.6 per cent (the Illinois average for 1992); ten zones with unemployment rates between 6.6 and 7.92 per cent, which is 120 per cent of the state average; and so on. The largest number of zones (36) had unemployment in the range 7.92–14.51 per cent, according to 1992 data. According to the state’s distress criteria for EZ designation, zones are considered high-unemployment areas if they have unemployment greater than 120 per cent of the state average. In terms of this criterion, 48 out of the 68 zones chosen had unemployment rates greater than 7.92 per cent as of 1992 (see Figure 3.1), and have been classified as high-unemployment areas. The variation in unemployment rates among zones was sufficient to test the hypotheses with regard to high unemployment areas. The benefit–cost analyses reported in this chapter are based on 1992 Illinois EZ programme data. Note that while current data are available for the state’s EZs, more recent data are not available for reservation wages.

Number of zones in the range

40

36

35 30 25 20 15 10

10

10

11

5

1

0 0–6.59

6.6–7.91 7.92–14.51 14.52–21.11 21.12–27.72 Range of unemployment rate (%)

Figure 3.1 Distribution of unemployment rate in Illinois EZs

0 >27.72

Evidence from Illinois Enterprise Zones 47

The reservation wage question was not asked of respondents in the Panel Study of Income Dynamics after 1987 (http://psidonline.isr.umich.edu). Note that it is the assumptions and methodology that are of interest here. The analysis can be easily replicated for recent years without loss of generality. 3.4.4

Incentives in the Illinois EZ package

The IL DCEO describes many of the tax benefits available to businesses locating in Illinois EZs. These include investment tax credit, sales tax deduction, machinery and equipment sales tax exemption, utility tax exemption (applicable to gas, electricity, and telecommunications companies), jobs tax credit on Illinois income taxes, local property tax incentives (the most important for the work in this chapter) and dividend deductions. We are interested in the property tax abatements offered to firms, as a measure of the programme costs. This is because the other incentives noted above are state-level and would be predetermined as far as local governments are concerned. Further, some state-level incentives could well be the same across several zones. Since we are interested in local net benefits, the analogous measure of costs has to be local, and the property tax abatement is the most relevant to consider as programme cost. Property tax abatement There are two types of property tax abatements in Illinois EZs: tax abatement and assessment reduction (available only in Cook County).2 Any taxing district may abate any portion of its taxes on real property located within an EZ. However, the amount cannot exceed that attributable to improvements, renovation, or rehabilitation of existing improvements. This is a local incentive, and hence the abatement rates and amounts can vary from zone to zone. Industrial property in Cook County is generally assessed at 36 per cent of its market value. Under the EZ special incentives, improvements to EZ property are assessed only at 16 per cent of market value for eight years as part of the Cook County assessment reduction incentive. As explained by the DCEO, the tax rate remains the same, but a company’s tax liability drops because the rate is multiplied by a much smaller property value. Other counties in IL assess all property at 33 per cent of its market value. In all cases, under the EZ programme, the value of abated property is subtracted from the tax base prior to calculation of the tax rate. Table 3.2 summarizes the achievements of the Illinois enterprise zone programme to date. The programme has seen a steady increase in the number of businesses investing, jobs created, and the dollar amount of actual investments made by individual firms in the zones registering the highest percentage increase over 1991–2004. Note that the actual number of ‘jobs retained’ has shown a decline over 1991–2004. ‘Jobs retained’ are jobs that would have moved out of the zones if incentives had not been provided. It is not clear what the counterfactual is – that is, what would have happened

48 Incentives for Regional Development Table 3.2 Achievements of the Illinois enterprise zone programme

Number of businesses investing Investment by companies (in $ million) Jobs created Jobs retained

Annual average, 1984–89

1991

2004

% change, 1991–2004

875

1,528

2,136

39.79

670 9,070 20,050

1,010 9,823 31,319

2,503 13,966 19,070

147.82 42.18 39.11

Source: Illinois DCEO.

if incentives had not been provided. Would jobs have moved out, or would they have remained there if incentives had not been given? Many firms use ‘moving out’ as a threat with local governments, if incentives are not given. But it is not clear if they would have actually moved even if the incentive had not been given. Because of this ambiguity, decline in the number of jobs retained is perfectly consistent with a programme doing well.

3.5

Research methodology

As noted above, research in this chapter assumes that the tax incentives available in the enterprise zone programme are responsible for creating jobs; however, this may not always be true. There are two other possibilities: 1. The programme might merely redistribute jobs from one location to another. 2. The programme may not do anything at all (neither create nor relocate jobs). Regarding the assumption of research in this chapter that incentives affect business location decisions and are responsible for creating jobs, note that Bartik’s summary of existing empirical research on the topic shows that tax incentives affect business location decisions. Based on this assumption, we are concerned with two possibilities: the programme creates jobs, or, that it merely redistributes jobs. The research methodology formulated takes these alternatives into account. Option two above (that the programme may not do anything at all) is not considered relevant. McDonald (1997) argues that there are problems with such an approach for two reasons: First, the counterfactual (the effect on employment in the absence of the programme) is not taken into account. Secondly, job creation is not a valid measure of gain from the programme. This is because job creation in IL is coincident with a firm obtaining a building permit to qualify

Evidence from Illinois Enterprise Zones 49

for the sales tax exemption on building materials (as shown in Redfield and McDonald, 1991), whereas building permits are continually issued and jobs created in the local economy. McDonald argues that it is not clear if this can be attributed to the existence of the EZ or to the tax incentives. The problems McDonald discusses relate to the counterfactual, which I do not deal with in a systematic manner here. In Chapter 4, I estimate an econometric model in which an outcome variable, unemployment rate, is dependent on a dummy for tax incentives (along with other characteristics), using Ohio’s data.3 This enables me to assess the effect of tax incentives on unemployment rate in Ohio. For the empirical work in this chapter, measures of raw costs and benefits were developed in order to assess whether the benefits from providing abatements in Illinois’ zones are greater than the costs of providing them. According to Bartik (1991, Table 7.7), the net benefits of development policies seem most sensitive to the programme cost per job created and the magnitude of real earnings effects. On this basis, measures of costs and benefits were chosen. 3.5.1

Measures of costs

Forgone property tax revenues, as a result of incentives given to businesses in the zone, were chosen as a measure of costs of the Illinois EZ programme. Property tax abatements are an important part of the state’s zone package, as reflected in the fact that, in a survey, 94 per cent of Illinois enterprise zone administrators reported that property tax abatements were available to potential investors in the zone (Elling and Sheldon, 1991: 139, Table 9.1). There are also several studies that point to the importance of property taxes in business growth and location. Because of the role that property tax abatements play in local government finance, and since there were no data on what percentage of programme costs other incentives formed in the Illinois total enterprise zone package,4 property tax abatements were chosen to reflect costs of the Illinois enterprise zone programme.5 However, it is questionable whether the forgone property tax revenues from the abatements can really be viewed as ‘costs’ of the programme. In fact, if we assume that all jobs associated with the EZs are new jobs, the relevant costs of the programme would be the fiscal effects of the new jobs, including the abatements. One such fiscal effect would be the cost of providing public services to businesses in the zone. It could be assumed that the property taxes paid would cover some or all of these costs; however, property tax revenues could be less than, equal to, or greater than the extra public-service costs from the new jobs. Consider the illustration in Table 3.3;6 here, it is assumed that with the programme, ten jobs are created, and the cost per job is $100. With this assumption, public-service costs amount to $1,000. Three different scenarios are taken into account in order to analyze the fiscal effects of the programme.

50 Incentives for Regional Development Table 3.3 Fiscal effects of the programme Without programme

Public service costs Property taxes without abatement Property tax abatement 60% of tax Taxes after abatement Cost of programmea

With programme ($) Scenario 1

Scenario 2

Scenario 3

0 0

1,000 1,000

1,000 1,500

1,000 500

0 0 0

600 400 600

900 600 400

300 200 800

Note: a. As can be seen, the cost of the programme is the difference between the public service costs and net property tax revenue (property taxes after abatement).

In scenario one, it may be noted that the property tax abatement is equal to the cost of the programme. Here, the net fiscal effect of the new jobs is zero without the abatement – that is, the extra public-service costs from new jobs just equal normal property tax revenues. With the abatement, the cost of the programme is the cost of the abatement: that is, $600. In scenario two, where property tax revenue from the new jobs (without abatement) is higher than the extra public-service costs required for their creation, the cost of the programme is, in fact, less than the abatement dollars provided, assuming the same property tax rate as in scenario one. In other words, new industry generates a fiscal surplus, which can be used, at least in part, to finance the incentives. However, where property tax revenue from new jobs is less than the extra public-service costs required, the cost of the programme will be higher than the abatement costs (scenario three). There is no empirical evidence about how costs of the enterprise zone programme in Illinois compare with those of property tax abatement costs. In general, no empirical work has compared the cost of extra public services incurred, to the tax costs (revenues) incurred for new businesses by the local government.7 In this analysis, a situation similar to scenario one is assumed to prevail; in other words, if new businesses are usually a fiscally neutral proposition for local governments, our assumption that abatement costs are equivalent to local programme costs will be accurate. It is reasonable to assume that, if businesses did not locate in the area, the cost of providing public services to them would also be non-existent, and there would be no jobs created (see US Department of HUD, 1986). On the other hand, if businesses located in the zone, we may assume, for the sake of simplicity, that the extra cost of providing public services to the new business is exactly equal to the normal property tax revenue from new jobs created by the business. In Ohio’s EZs, as we observe in Chapter 5, the firms locating there, in fact,

Evidence from Illinois Enterprise Zones 51

generated enough fiscal surplus to finance the cost of providing public services to them. 3.5.2

Measures of benefits

The number of new jobs created represents the most apparent benefits of growth; while necessarily representing a labour market perspective, they are chosen as a measure of benefit. In the actual analysis, an attempt is made to consider two extreme cases: (i) All the jobs that are ‘created’ are new (not merely relocated from somewhere); (ii) All the jobs ‘created’ are simply relocated from elsewhere.8 When jobs are new, the benefits are wages paid minus the reservation wages. In this case, the analogous measure of benefits chosen was earnings from jobs, net of estimated reservation wages. Courant (1994) argues that the mere number of jobs created should not be counted as welfare-increasing for the economy, except if they have distributional implications, which it has, in this case, because of the implications of net benefits being higher in high unemployment areas. The data on earnings by industry were obtained partly from the Census of Population and Housing (from the US Bureau of Census), and Employment and Earnings (from the US Bureau of Labor Statistics). Reservation wage, the lowest amount (wage) necessary to encourage the individual to work, was estimated as a function of the unemployment rate, using data from the Panel Study of Income Dynamics (PSID). As indicated by the model in Chapter 2, the net earnings (benefits) obtained from jobs created in the ith zone were compared to the abatement dollars spent (Ai) in the ith zone to obtain the benefit–cost ratio (B–C ratio). Mathematically, this may be represented as follows: NB(i) 兺 JijEj 兺Wir HjWjJij j

BC Ratio

(3.1)9

j

NB(i) Ai

(3.2)

In equation (3.1), NB(i) refers to net benefits from jobs in zone i, Jij refers to jobs created in the ith zone and jth sector (industry or occupation), and Ej refers to the annual average or median earnings10 in the jth sector (industry or occupation). Wir is the estimated reservation wages per hour per job in the ith zone. Hj refers to average hours worked per week in industry j, and Wj refers to average weeks worked per year in industry j.11 Reservation wages were assumed constant across all j’s (industries or occupations); they were assumed to vary across zones with varying unemployment

52 Incentives for Regional Development

rates, but constant for all jobs within a zone. It might well be the case that an unemployed person with higher education and skills may have a higher reservation wage and vice versa. Note the following, however: 1. The reservation wage estimates used to impute reservation wages for the zones also take into account education and skills (see Table 3.4). 2. Unemployment rates also differ across different jobs (industries or occupations). Since reservation wages are assumed to depend upon the unemployment rate, the assumption that reservation wages are the same for all jobs within a zone having a certain unemployment rate is reasonable. In case all the jobs are merely relocated to the zone from some other location, the benefits are not as defined in equation (3.1), but are equal to the earnings net of reservation wages in the alternative location less the earnings net of reservation wages in the zone.12 For the purposes of simplification, the alternative location was assumed to have the state’s average unemployment rate, 6.6 per cent for 1990.13 There is a net social benefit, in other words, only if the jobs provide greater benefit to those in the EZ than elsewhere, assuming that earnings are identical. Thus, assuming that the jobs relocated are identical in terms of hours and earnings, the net benefit to the state would be the reservation wage at the alternative location minus the reservation wage at the enterprise zone location. While net benefits from new jobs are represented by equations (3.1) and (3.2), net benefits from relocated jobs may be represented as follows: NB 兺Wkr HjWjJij 兺Wir HjWj Jij

(3.3)

NB 兺 HjWjJij(Wkr Wir )

(3.4)

j

j

or, j

BC Ratio

NB A

(3.5)

NB refers to net benefits to the state of Illinois as a whole; Wkr refers to the reservation wage at the kth (alternative) location. All the other terms are as defined in equations (3.1) and (3.2). It was possible to estimate net benefits from jobs, because the industry or occupation in which jobs were created in all zones was known. Because of this, it was possible to assign jobs created to an industry or occupation with a certain standard SIC (Standard Industrial Classification) code, and to use the average or median earnings for the industry in order to represent gross earnings.14 For the purpose of estimating benefits in the case, where all jobs were assumed to be new, average earnings for that industry or occupation

Evidence from Illinois Enterprise Zones 53

were used to estimate gross earnings from jobs created in the zones. For the purposes of estimating benefits in the second case (where it was assumed that all jobs were relocated from somewhere else), the reservation wage at the enterprise zone location was subtracted from that at the alternative location in order to obtain net benefits. Net benefits in this case can be positive or negative, depending upon the relative unemployment rates in the origin and area of relocation.

3.6

Estimation of reservation wages

In order to estimate reservation wages as a function of the unemployment rate for the United States, I used data from the Panel Study of Income Dynamics (PSID).15 From the overall sample of 7,000 persons, a sub-sample of 334 unemployed persons, who had valid values for their reservation wages, was chosen from the 1987 interviewing panel.16 Based on standard labour economic theory, the reservation wages is estimated by OLS with the following as exogenous determinants: PREVWG (Previous wages): This refers to the wages from previous jobs of the unemployed, and, as may be obvious, applies only to those unemployed who have been employed before; it does not apply to unemployed graduates fresh from school. It is reasonable to believe that the previous wages of an unemployed person determine his/her current reservation wage. Its expected effect on reservation wages is positive. UNRATE: This is the unemployment rate of the county of residence of the unemployed. The unemployment rate of the local area affects a person’s reservation wage through its effect on job opportunities in the local area. Its expected effect is negative, consistent with Bartik (1991). UNCOMP: This refers to the amount of unemployment compensation that affects the economic conditions of the unemployed person, and reflects the cost of unemployment. In particular, UNCOMP decreases costs of remaining unemployed for the individual, because they ensure a continuing source of income (Holzer, 1987), and increases the reservation wages. COLL: This refers to the number of years of college education. It is quite reasonable to believe that skills as reflected in college education have a positive effect on reservation wages. MARITAL (marital status): Since marital status indicates the extent of family responsibility, its effect on reservation wages is positive. The original variable for marital status in the data set contained five categories: married, never married, widowed, divorced, or separated. For the purposes of this analysis, this variable was recorded and reduced to two categories: married (one) and other (zero), to test for the effect of family responsibility on the reservation wages.

54 Incentives for Regional Development

SEX (gender): Men traditionally have lower reservation wages since they are the primary breadwinners in most, if not all, households. Hence, the expected effect on the reservation wages is negative. Sex was recorded as a zero–one variable. For purposes of this analysis, zero represents female, and one represents male. NMI (Non-market income): Non-market income consists of AFDC (Aid to Families with Dependent Children) payments, supplemental security income, social security payments, and other welfare payments. This has the same effect on reservation wages as unemployment compensation, positive, since such payments reduce the cost of staying unemployed. AGE: Age, as usual, is measured in years, and measures the effect of the individual’s age on reservation wages. Its expected effect is ambiguous because, in some cases, age may be associated with higher work experience. In such cases, age would have a positive impact on reservation wages. In other cases, older age unemployment might indicate negative effects and lead to lower reservation wages, other things remaining constant. HS (high-school degree) and COLL (college degree): The variables HS (high-school degree) and COLL (college education) enable us to test the significance of skills associated with high-school education and college education separately, on the individual’s reservation wages. These variables were recoded from the original variable in the data set for education. The original variable consisted of eight categories: zero–five grades, six–eight grades, and so forth. The variable was recoded to two dummy variables – the two dummies being high-school degree (HS) and college degree (COLL). In the recoded form, a code of zero on the variable HS meant a high-school dropout, and a code of one meant a high-school degree. Similarly, a code of zero on the variable COLL indicated no college degree, and a code of one implied a college degree. To test for the effect of college education alone, and to avoid the possibility of collinearity among certain variables, the high-school education, age and non-market income variables were removed, and a reduced model was also estimated by OLS. Both the models are presented in Table 3.4. The reduced model was used to assess the reservation wage response to the local unemployment rate. 3.6.1

Results from regression

The complete and reduced regression models obtained, and the means for the variables in the reduced model, are shown in Table 3.4. These models were estimated by OLS.

Evidence from Illinois Enterprise Zones 55 Table 3.4 Estimation of reservation wages: results from regression Dependent variable: reservation wage Variable Constant PREVWG UNRATE UNCOMP COLL MARITAL SEX NMI AGE HS F R2

Complete model 430.04 (59.19)**a 0.28 (0.03)** 10.61 (5.91)* 0.05 (0.01)** 611.01 (66.98)** 84.87 (31.54)** 21.35 (17.71) 0.01 (0.01) 0.07 (0.90) 0.29 (30.28) 51.98 0.632

Reduced model 454.73 0.28 11.94 0.05 601.27 90.39 25.78

(42.81)** (0.03)** (5.85)* (0.01)** (66.48)** (31.40)** (17.28)

Mean

$4.63/hour 6.36% $340.83 0.05 0.32 0.61b

77.86 0.626

Notes: a. Numbers in parentheses refer to standard errors; **, * respectively denote that coefficients are significantly different from zero at the 1 and 5 per cent levels, when a two-tailed test is performed of the null hypothesis that b is equal to zero. b. The means for the three dummy variables – sex, college education, and marital status – show that 61 per cent were male, 5 per cent had college education, and 32 per cent were married.

Effect of unemployment rate on reservation wages: Table 3.4 shows that the relationship between reservation wages and unemployment rate is as expected: negative, indicating that, as the local unemployment rate of the county of residence increases, the reservation wages of unemployed in those counties decrease. The magnitude of the coefficient for the unemployment rate in the complete model indicates that, for every one percentage point increase in the unemployment rate, there is a ten cents decrease in the reservation wage (with implied decimals). The magnitude of this coefficient becomes larger in the reduced model which indicates that, for every one percentage point increase in the area’s unemployment rate, there is a decrease in an individual’s (in that area) reservation wages to the extent of 12 cents. In both models, the unemployment rate is statistically significant (at the 5 per cent level), with the noted caveats (see Haurin and Sridhar (2003) for an assessment of the reservation wage response to the local unemployment rate, taking into account sample selection). Prediction of reservation wages The reduced model was used to impute reservation wages for jobs created in Illinois EZs. Since it is the reservation wage response to the unemployment rate that is relevant for purposes of this analysis, the mean values for all independent variables (shown in Table 3.4) except the unemployment

56 Incentives for Regional Development

rate were substituted in the model to obtain a simple regression model. The result was: Reservation wages 644.79 11.94 UNRATE

(3.6)

This procedure enabled the imputation of reservation wages and, subsequently, benefits for new as well as relocated jobs.17 Estimation of benefits from new jobs was done in a manner described earlier, in equation (3.1). To estimate benefits from relocated jobs (as in equation (3.3)), it was assumed, for the purpose of simplification, that the alternative location would have the state’s average unemployment rate (6.6 per cent in 1990). Assuming that earnings from jobs in a given industry or occupation are similar in intra-state locations, the reservation wage corresponding to the state’s average unemployment rate was estimated using the equation [644.79 (11.94 6.6)], which yields $5.66 per hour. This estimated reservation wage for the alternative location was converted into annual reservation wages in the manner described earlier. In order to estimate the benefits from relocated jobs, from the annual reservation wages obtained for the hypothetical alternative location (assumed to have the state’s average unemployment rate and the corresponding reservation wage), were subtracted the annual reservation wages for each zone. As a result, the resulting benefit for each zone assumed that the jobs were relocated from somewhere else and were not really new. When these net benefits from relocated jobs are summed across all zones, the result is not the local benefit to that zone, but the net benefit to the state, encompassing the zone and the alternative location. In order to illustrate the extreme case in which all jobs are assumed to be relocated, take the instance of two zones i and k. Let i be the highunemployment zone (in relation to state average) and k, the one with the state’s average (6.6 per cent) unemployment rate (the alternative location). A manufacturing firm in area k decides to relocate to zone i in order to realize tax savings, and moves its facility to zone i. Since the reservation wage in zone i (Wir the reservation wage at the enterprise zone location), because of higher unemployment, is lower than that in area k (Wkr, the reservation wage at the alternative location), the difference Wkr Wir would be positive, implying positive benefits. Thus, based on the finding that reservation wages are lower in high than in low-unemployment areas, even the redistribution of jobs from low- to high-unemployment areas would bring about net social benefits. In the case of the two extreme scenarios (when jobs are assumed to be new or relocated ), the ratio of net benefits (differently defined in each case as explained above) to abatements is taken for each zone to obtain the benefit–cost ratio. In the estimation of the benefits from relocated jobs, observe that there is no benefit to the zone, if the relocated jobs are not held by zone residents, primarily because local government expenditure is involved. But we must

Evidence from Illinois Enterprise Zones 57

distinguish between jobs to zone residents and jobs to residents with lower reservation wages.18 If the jobs created by zone firms were held by those with low reservation wages, not necessarily by zone residents, the benefits for the state as a whole derive primarily from an equity perspective, although there are no benefits to the zone. For the same reasons, it does not matter how many workers in enterprise zone firms actually live in the zone; even if they live elsewhere, their reservation wage is assumed to be that of the zone. It is assumed that workers in a high-unemployment area have lower reservation wages; the mere fact of the location of some of the employees somewhere outside the zone does not change their reservation wage. On the other hand, extraneous factors such as transport costs and house prices influence residential location decisions (to location inside a zone from outside the zone, for example). If relocation of a firm to the zone represents merely the movement of current employees from outside the zone, the relocation is not considered to constitute investment in a ‘new’ facility, and the jobs ‘created’ are not new, but only ‘retained’. This can be best illustrated by the example of a firm that threatens to relocate elsewhere if incentives are not provided. In such cases, retention of current employees and relocation to the zone from outside only represent ‘retained’ jobs (jobs which would have been lost, but for the incentives), not ‘new’ ones; there is no net benefit to the zone. However, there may be some social costs involved if firm relocations occur within the period of the contract.19 For examples of such contracts, see Byrnes, Marvel and Sridhar (1999) who use contract data from Ohio’s EZs to determine factors that explain abatement generosity. Contract agreements that are usually negotiated before the firm formally starts its operations in the zone may, in the event of firm relocation, include the clawback of some or all the incentives provided by the zone from which the firm is relocating. Further, other policy restraints imposed on firms can pertain to restrictive plant-closing legislation, or common tax-sharing arrangements (see, for example, Ledebur and Woodward, 1990; Peters, 1993; Eisinger, 1988). There are also other social costs involved in such cases, when corporatism may not be compatible with the social goals of the community (see Perrucci, 1994, for example). In this chapter, the social gain from new and relocated jobs is calculated only on the basis of wages, reservation wages in the alternative location, and that in the enterprise zone. Firms can contribute to the regional economies of where they locate, over and above the net benefits accruing to the community from jobs alone. Chapter 7 contains some evidence from India of ways in which firms contribute socially to communities where they locate.

3.7

Benefit–cost analysis

Tables 3.5 and 3.6 show the reservation wages, average net benefits, net benefits per job, and the average benefit–cost ratios, classified by zones’

58 Incentives for Regional Development Table 3.5 Case 1: summary of net benefits from ‘new’ jobs

Unemployment rate of zones, 1990 (%) Number of jobs Annual reservation wage per job ($) Gross earnings per job ($) Total net benefits ($) Average net benefits ($) Average net benefit per job ($) Abatement per job ($) Average B–C ratio Total number of zones

Lowunemployment zones

Moderateunemployment zones

Highunemployment zones

All zones

Under 8.1

8.2 to 10.9

11.0 to 20.6

0.9 to 20.6

2,841 9,336

3,396 8,552

3,481 6,896

9,718 8,261

21,763

20,657

19,372

20,597

35,965,520 2,115,620 12,169

41,543,375 2,443,730 11,939

50,763,310 2,986,100 12,301

128,272,205 2,515,140 12,136

860 14.15

879 13.58

2,231 5.51

1,358 8.94

16

16

17

49

Table 3.6 Case 2: summary of net benefits from relocated jobs Lowunemployment zones Average net benefits ($) Average net benefit per job Average B–C ratio

Moderateunemployment zones

Highunemployment zones

All zones

17,653 146 0.29

136,808 616 2.86

309,902 1,420 3.64

143,019 630 2.07

16

16

17

49

Total number of zones

unemployment category – high, medium and low – for the two cases considered for work in this chapter – new jobs and relocated jobs. Zones with a 1990 unemployment rate of less than 8.1 per cent (that is, less than 120 per cent of the state’s average unemployment rate) are classified as lowunemployment zones. Those with unemployment rate between 8.2 and 10.9 per cent are classified as having moderate unemployment, and those having greater than 11 per cent unemployment are classified as highunemployment zones. 3.7.1

Net benefits from jobs

The net benefits from jobs, using 1992 data for Illinois EZs, are reported in two scenarios representing new (scenario one) and relocated jobs (scenario two). Scenario one If all jobs were ‘new’, gross earnings are the highest on average in the low-unemployment zones and lowest in the high-unemployment zones,

Evidence from Illinois Enterprise Zones 59

consistent with assumptions two and six of the theoretical model developed in Chapter 2. When gross earnings are made net of reservation wages, however, net benefits per job are the greatest in the high-unemployment zones. This shows that the reservation wages are lower (confirmed by the sign and statistical significance of the unemployment rate in the regression shown in Table 3.4) in high-unemployment areas when jobs are created. As evidence, Table 3.5 shows that annual reservation wages per job are the lowest in highunemployment zones, and highest in the low-unemployment zones of Illinois. For the same reason, net benefits per job are high in high-unemployment zones compared to low-unemployment zones, even though earnings per job are lower (no surprise) than in the low-unemployment zones. The amount of average net benefits in the high-unemployment zones is over $2.9 million per zone (Table 3.5). The zones with high unemployment, with 3,480 jobs created, also had higher than average net benefit per job ($12,300), compared to the average for all zones. The higher overall net benefits, as well as the net benefits per job from abatements provided to the highest-unemployment zones despite lower earnings, is a result of their lower reservation wages compared to other zones. In the moderate-unemployment zones, gross earnings and net benefits per job are lower than in the low-unemployment zones. Reservation wages per job in the moderate-unemployment zones are only $780 less than in the lowunemployment areas. But gross earnings are about $1,100 less, which indicates that jobs created in moderate-unemployment areas are lower paid than those in the low-unemployment areas, which, again, should come as no surprise. A greater number of jobs are created in the moderate-unemployment than in the low-unemployment areas. On the other hand, abatement cost per job is higher in the high-unemployment areas than in the low-unemployment zones. This, while implying that it is not cost-effective to create jobs in high-unemployment areas, is due to large abatements being offered by a few zones. When all zones are taken into account, the average net benefit per job is higher than the cost per job (being $12,130), compared to the abatement cost of only $1,350. This supports the notion that if we were to assume that all jobs are new, on average, the net benefits are greater than the costs of creating them. Scenario two Table 3.6 summarizes outcomes in this scenario for zones in the three unemployment categories (high, moderate, or low) when all jobs are assumed relocated. When all jobs are assumed to relocate to low-unemployment zones from somewhere else within the state, average net benefits and benefits per job are negative, indicating a loss of benefits if jobs were redistributed from average-unemployment to low-unemployment areas. By the same token, note that the moderate- and high-unemployment areas enjoy positive net benefits even if the jobs ‘created’ in those areas were actually relocated from somewhere else (that is, from average-unemployment areas).

60 Incentives for Regional Development

Average net benefits and benefits per job are highest in high-unemployment areas (Table 3.6). Thus, there are positive net social benefits, even if all jobs are redistributed from average-unemployment to high-unemployment zones. This indicates that, if at all policy encourages relocation of jobs, from an equity perspective, they should be relocated to high-unemployment areas, where the benefits per job would be higher. 3.7.2

Benefit–cost ratios

Scenario one The benefit–cost ratio is defined as the ratio of the average net benefit to the average abatement per job. When all jobs are assumed to be new, the benefits are, on average, more than eight times the costs for all zones (Table 3.5). In the low- and moderate-unemployment areas, in fact, the benefits are more than ten times the costs. It may be seen from Table 3.5 that lowunemployment zones have the highest ratio, on average. On the other hand, high-unemployment zones have the lowest B–C ratio. Among the zones with high unemployment, the average benefit–cost ratio is lowered (in relation to the moderate-unemployment zones) primarily because of abatement costs in two zones: Monmouth, and Bloomington-Normal. Given little disparity in the net benefit per job among zones with different levels of unemployment (Table 3.5), the disparity in the benefit–cost ratios shows that the abatement cost per job is high in high-unemployment areas. This is because of zones like Bloomington-Normal and Decatur, in the high-unemployment group, which represent instances where the B–C ratio is low (respectively 0.57 and 3.10) despite high (higher than average) net benefits per job ($16,500 and $13,500, respectively). This can be attributed to the large abatements given in these two zones. I examined recent data from the Illinois DCEO on jobs created in these zones, and they continue to be a drag on jobs. The Bloomington-Normal zone created 11 jobs in 2002 and only three jobs in 2003. Monmouth zone created only eight jobs in FY 2003, with no jobs having been created in the preceding three years. Thus, given the assumptions, the analysis, although based on 1992 data, is robust even if recent data from the programme were to be used. There does exist an optimum level of incentives to be provided even in highunemployment areas, beyond which cost-effectiveness is not ensured in creating jobs. Scenario two When it is assumed, on the other hand, that all jobs are relocated to enterprise zones in Illinois from elsewhere in the state, the benefit–cost ratios are as expected. When jobs are relocated from IL’s average-unemployment areas to low-unemployment zones (those with unemployment under 8.1 per cent), the benefit–cost ratio is negative. If jobs, however, are relocated to

Evidence from Illinois Enterprise Zones 61

high-unemployment areas from elsewhere, the benefits are more than three times the costs. In moderate-unemployment areas, benefits are more than twice as much as costs. This indicates that a redistribution of jobs from an area with about the state’s average unemployment rate to areas with higher unemployment levels, in relation to the state average, can produce benefits for the state as a whole. When all 51 zones are taken into account, the benefits are twice the costs, assuming that all jobs are merely redistributed to EZs from elsewhere (that are assumed to have the state’s average unemployment rate) in Illinois. With the more favourable assumption that all jobs are new, the benefit–cost ratio is close to nine. This is to say that, even if total public costs were more than the direct abatement costs, the net benefits from the zone incentives would still be substantial.

3.8

Summary of the results

The results from this chapter show that the economic development incentives offered in the EZs of IL do not amount to a zero-sum game. On average, net benefits are expected to be about ten times the abatement costs. On average, the net benefits per job are above US$12,000 a year. And, in more than half of the zones (32 of 51), the average net benefit per job is higher than this: at about $15,910 per job. However, when all zones are taken into account, the cost per job is only about $1,350. Even assuming that jobs are only being redistributed from other areas, the benefits per job are about $1,450 on average in the high-unemployment zones of the state. With respect to the second question that was posed at the beginning – that is, whether benefits from jobs are greater than the costs of providing abatements in high-unemployment areas – the response, based on these results, is positive. First, the results show that net benefits per job are higher, despite lower earnings in high-unemployment areas, because of the low level of reservation wages. On average, the dollar amount of net benefits (about $3 million per zone) and net benefits per job (about $12,300 per job) are highest in zones with high unemployment, and the largest number of jobs are created in those areas. Only 2,840 jobs are created in the 16 low-unemployment zones, whereas nearly 3,400 jobs are created in the 16 zones with moderate unemployment, and about 3,500 jobs are created in the 17 high-unemployment areas. Even assuming that gross earnings are lower per job (as is the case in moderate- and high-unemployment zones, compared to low-unemployment zones), low-paid jobs help the unemployed to enter the labour market and can consistently improve their skills to pursue better employment (termed as hysteresis). Thus, each new job produces benefits not indicated in the dollar amount of earnings per job. Even when jobs are assumed to have merely relocated, the net benefits are three times as much as costs in

62 Incentives for Regional Development

high-unemployment areas, and twice as much as costs in moderateunemployment areas (see Table 3.6). However, from a benefit–cost viewpoint, the areas with high unemployment do not produce the highest amount of net benefits from jobs, compared to the abatements provided. This may be because of the larger abatements granted to businesses in the high unemployment zones. With the favourable assumption that jobs are ‘new’, the B–C ratio is, for example, nearly 42 in the moderate-unemployment areas, whereas in the high-unemployment areas, it is about 30. Thus, even though average net benefits and benefits per job are higher in high-unemployment areas than in low-unemployment zones (where the average B–C ratio is, for example, only 15), no clear picture emerges from a comparison of B–C ratios across the zones with differing levels of unemployment. When all of the jobs are assumed to be relocated within Illinois, taking the sum of the benefits and losses for the state as a whole, benefits are twice that of costs. Of course, the magnitude of benefits would be different for the nation as a whole, when a US perspective rather than an Illinois perspective is taken, if one is asked what the national benefits are from this Illinois programme. From a national perspective, a higher proportion of jobs would be redistributed from outside Illinois, rather than ‘new’ jobs to the country as a whole. The alternative reservation wage would be different in these two cases. In any case, these results show that reservation wages are low in highunemployment areas compared with that in low-unemployment areas. Consequently, net benefits per job are higher in the high-unemployment areas. However, the disparity between the dollar amount of net benefits and benefit–cost ratios (in zones like Bloomington-Normal and Decatur) suggests that indiscriminate provision of generous incentives, even in highunemployment zones, is likely to result in a lower cost-effectiveness of the abatement dollars provided. Rather, a strategy of targeting those labourintensive businesses, whose employment and location choices will be affected by a typical abatement package, rather than offering generous abatements, is more likely to be cost-effective. In interpreting these results, care must be exercised because compromises have been made with the data, and these may distort the results to some degree. For example, the cost estimates of the programme, as represented by the measures chosen in this chapter, are conservative, since they consider (as programme costs) only property tax abatements provided to businesses in zones. Further, the measures of benefits necessarily take a labour market perspective. Nevertheless, they have enabled a comparison between costs to create jobs and benefits that can be expected from those jobs created by the programme. The work in this chapter does not carry the analysis to the eventual conclusion of Bartik’s research: since the net benefits from a given number of jobs are higher in high-unemployment areas, does the actual competition for

Evidence from Illinois Enterprise Zones 63

jobs occur in a way that benefits such areas? This question, being whether high-unemployment areas aggressively pursue and use economic development incentives in such a way that the jobs benefit them, may lead to a more determinate relationship between the cost-effectiveness of incentives and their actual practice by regions with certain economic characteristics. We started with the proposition that the benefits of redistributing jobs to high-unemployment areas are likely to be higher because of the lower reservation wages in those areas. The results in this chapter show that, under certain conditions, high-unemployment areas do indeed have low reservation wages; for this reason, net benefits are likely to be higher in those areas. The benefits of redistribution are likely to be higher in high-unemployment areas than in low-unemployment areas. Some benefits of redistribution are illustrated in this chapter. However the intention of this chapter has not been to endorse policies which encourage redistribution itself, but rather to study, in the event of redistribution, what kinds of redistribution can result in net benefits to an area as a whole. The work in this chapter has been the first attempt to estimate reservation wages and measure net benefits from jobs; future research must make the measures of benefits and costs, hence the analysis, more perfect. If regional development policy is to be positive-sum, one possibility is to encourage indigenous entrepreneurship (an entrepreneurial climate) in EZs in order to encourage job growth, and not merely relocation (which a ‘good’ business climate may sometimes encourage). The implications of indigenous entrepreneurship in EZs are to be found only by further research. Chapter 7 provides some evidence of the extent to which place-oriented policies have promoted local entrepreneurship through the offer of infrastructure incentives in the growth centres of India. Before we study that, the next couple of chapters present evidence from Ohio’s enterprise zone programme, another state of the United States, in which the enterprise zone programme stimulated a lot of debate in academic and policy circles through the 1990s.

4 Impact of Tax Incentives on the Unemployment Rate: Evidence From Ohio

4.1 The importance of the problem and the motivation for research As discussed in earlier chapters, many regional development policies that attempt to increase employment do so by providing financial incentives to firms. These have included industrial revenue bonds (IRBs), property tax abatements, direct state loans, and customized industrial training. For various reasons, the effectiveness of such policies – specifically, tax incentives – continues to be questioned. This is because it is unclear whether taxes affect firm location decisions; and even if they do, the nature of their impact on the regional economies adopting them remains unclear. While Chapter 2 provides the theoretical framework for understanding the effects of such programmes on their regional economies, this chapter provides empirical support for the effects of tax incentives, from another Midwestern state of the USA. The purpose of this and the next chapter is to examine Ohio’s EZ programme (and other tax incentive programmes). In addition to the IL programme, Ohio’s EZ programme is chosen to provide empirical support to the various propositions in this book, for two central reasons. First, along with Illinois (which was one of the first states in the United States to adopt EZ legislation), Ohio is one of the few states in the United States to have more than 100 zones in its territory. Concern about the benefits of competition between locations has been at the forefront of policy debates in various states of the United States. In Ohio, this debate has been quite extensive because of the existence of a large number of EZs (about 330), when compared to the other states.1 Ohio’s programme contrasts sharply with New Jersey’s Urban Enterprise Zone (UEZ) programme, which allows limited designation. So far there are only 10 zones in New Jersey (Boarnet and Bogart, 1996). However, in Ohio, every area that meets at least one of 64

Tax Incentives and Unemployment in Ohio 65

six distress criteria or that meets minimum population requirements can potentially apply, and applicants cannot be denied zone designation if they satisfy either population or distress criteria. Most parts of Ohio are thus designated either as EZs or as areas providing tax incentives under other programmes. In fact, legislative discussions surrounding the EZ programme in Ohio frequently centred on the ‘pirating’ of firms in one area by neighbouring areas (Hill, 1994) that offered tax incentives. Whether these incentive programmes and the resulting tax competition is conducive for actual development to occur, remains a debate in both the literature and policy circles (Byrnes, Marvel and Sridhar, 1999). This policy debate is the policy version of the ‘zero-sum’ game argument in the literature. The advantages to the local control of these programmes both in terms of effective targeting of incentives and the importance of meeting the competition of other states have also been cited in the debates. Secondly, Ohio’s programme is one of the few EZ programmes to involve negotiation of terms between local government and individual firms in their EZs. Byrnes, Marvel and Sridhar (1999) contains a model of the bargaining between firms and local governments in Ohio’s EZs. Usually, in other state programmes, the qualified activities are determined when the programme is designed, as in the case of New Jersey’s UEZ programme, and there is thus no room for negotiation. Finally, the Ohio Department of Development has developed a systematic database regarding various aspects of Ohio’s EZ programme that enables the testing of the hypotheses that emerge from Chapter 2 of this book. The purpose of this chapter is to examine if Ohio’s EZs are effective in reducing unemployment in the areas that have adopted them. This question is important to ask because the literature and policy circles are ambiguous about the effect of these programmes. In fact, one of the explicit objectives of the EZ programme in Ohio is to reduce unemployment. In this chapter, I develop an empirical model, supporting the theoretical model outlined in Chapter 2, that shows how tax incentive programmes affect unemployment in the areas that adopt them. The model in Chapter 2 showed that tax incentive programmes can be expected to reduce the area’s unemployment rate, along with other factors, by increasing the capital–labour ratio, and labour productivity. In this chapter, I develop an empirical version of that model, testing it using data from Ohio’s incentive programmes and examining whether tax incentives are responsible for lowering unemployment in Ohio’s census block groups. Empirically, I find that tax incentive programmes have a statistically significant impact in reducing unemployment in the area that adopts them. This indicates that such programmes may not be totally detrimental to the development of areas that adopt them.

66 Incentives for Regional Development

4.2

Overview of the chapter

In the next section, I recall and explore further the analytical model that was developed in Chapter 2 for the purposes of using it for empirical work in this chapter. I then describe Ohio’s EZ programme. I then proceed to explain the data and research methodology used for the estimation. The subsequent sections focus on the results. In these sections, I report the results from the estimation of unemployment rate taking into account the treatment effects problem. The final section of the chapter evaluates the policy implications arising from the estimation.

4.3

A model of unemployment

This chapter attempts to address several gaps in the literature. It explores the cause of unemployment in the EZ, examines the effects of tax incentive programmes, and makes this model the basis for empirical work. As I highlight in Chapter 2, the model of unemployment that is developed here is based on the neoclassical assumption of high reservation wages. This means that individuals in the EZ are unemployed because they have high reservation wages relative to the market wage prevailing in the EZ. The assumption of high reservation wages is valid in the case of the United States because of the existence of various safety net programmes such as unemployment insurance and social security benefits. Generous unemployment benefits mean that an individual’s reservation wages become higher because the benefits reduce his/her costs of remaining unemployed. In fact Feldstein (1978) shows how a combination of a high marginal tax on earnings and no tax on unemployment compensation makes the private cost of unemployment small and causes an individual to remain unemployed by increasing his/her reservation wages. We know from the model outlined in Chapter 2 that the market wage in the EZ area is low because profit-maximizing employers are willing to pay a wage that only matches the skill of the workers. It is reasonable to imagine, as the model points out, that the marginal product of labour in the EZ is less than in the non-EZ area. The literature on EZs points to the blight in these areas (Ge, 1995; Erickson and Friedman, 1989; Levitan and Miller, 1992; Wilder, 1996), which, according to this model, is due to the initially low capital–labour ratio in the EZ. The low capital–labour ratio results in low productivity and wages for those employed as well as those who are unemployed. The labour in the EZ area is immobile because of its psychological ties to the area and also the costs of relocation. The migration literature has shown that those with poor skills and little education are the most likely to be immobile. The most basic data on geographical mobility that control for age group are available from the United States Current Population Reports for

Tax Incentives and Unemployment in Ohio 67

the period 1987–90. These data show that inter-state migration rates for people in the age group 30–34 were 3.4 per cent for those with nine to eleven years of education, as compared to 8.7 per cent with more than 17 years of education (Ehrenberg and Smith, 1994). One would be surprised if data from more recent years did not exhibit the same trend. The low levels of skill among the unemployed in the EZ, coupled with insufficient information regarding job opportunities outside their state (or even the local area for that matter),2 lead them to stay put. Some labour literature has modelled the unemployment status of an individual in a manner similar to that which is adopted here (Barron and Mellow, 1981; Holzer, 1986). Barron and Mellow (1981) estimate a reduced form model while attempting to answer the question as to what factors affect the subsequent labour force status of an unemployed individual. Their data are from the Current Population Survey (CPS) for May and June 1976 and from a special supplemental survey of unemployed respondents in May 1976 regarding current job-seeking activities for a sample of 1,307 individuals. They measure the reservation wages as a response to the question: ‘What is the lowest wage or salary you would accept … for this type of work?’ They estimate the probability of an unemployed individual becoming employed in a given period as a function of the relative reservation wage (the ratio of the reservation wage to the average wage in the local labour market), along with other factors. They find that a higher relative reservation wage reduced employment probability, similar to expectation in this model. In an attempt to explain the ‘shockingly high rates of unemployment which plague black youth’, Holzer (1986) estimates the wage and the duration of unemployment as a function of reservation wage and other factors. Holzer uses data from the youth cohort of the National Longitudinal Survey (NLS) in 1979 and 1980. The sample is limited to white and black non-student males, aged 16 to 21 in 1979. The reservation wage is measured in the NLS as the response to the question: ‘What would the wage or salary have to be for you to be willing to take it?’ Holzer finds that young blacks seek wages that are comparable to those of young whites in absolute terms, but the wages the young blacks seek are higher relative to what is actually received by them. He finds that the relatively higher reservation wages of young blacks contribute to the duration of their unemployment and also, to some degree, to their lower, subsequently received wages. He concludes that changes in reservation wages might help to explain the trends in the wages and employment of young blacks in recent years. Thus, the finding from the labour literature is that higher reservation wages relative to market wages lower the probability of an individual finding employment and contribute to their unemployment status, findings consistent with the model developed in this book. Formally, the unemployment status (USij) of individual i living in the jth area is determined by the

68 Incentives for Regional Development

difference between reservation and market wage, as we have learnt from the model in Chapter 2: USij f(wrij wij)

(4.1)

where wijr is the reservation wage and wij is the market wage of the ith individual living in the jth area. As the past literature and the model in Chapter 2 show, the unemployment rate in the jth area is dependent upon the extent to which reservation wages of individuals are higher than market wages in the area. The unemployment rate of the EZ is high since, for a substantial portion of the labour force, reservation wages are high relative to market wage in the area. So prior to the designation of the area as an EZ, profit-maximizing employers have little incentive to hire workers with low levels of skill. Under these circumstances, EZ designation of the area acts as an important placeoriented policy to improve the blighted area through tax abatements. Tax abatements provide firms with incentives for investment, thus increasing the capital–labour ratio, the productivity of workers, and the market wage offered to them. At the point when the market wage exceeds the reservation wage in the EZ, unemployed individuals in the area become willing to work, employment increases, and the unemployment rate falls. As the model in Chapter 2 elaborates, the tax abatement on capital also leads to in-migration of capital into the EZ, and, depending upon the elasticity of substitution between capital and labour, results in increases in the employment of labour. With increasing levels of employment, the unemployment rate in the EZ reduces to a level below its initial rate, at the optimum. At this optimum, unemployment in the non-EZ areas remains unchanged. This occurs because, as Tobin demonstrates, an economy operating at a natural rate of unemployment of 5 to 6 per cent still has (more than 20 per cent) excess capacity in its capital stock (Tobin, 1972). Thus, the EZ acts only as a tool to direct the excess capital away from the full-employment areas to high-unemployment areas.3 The model developed in Chapter 2 thus shows how tax incentives, along with wages and reservation wages, affect the unemployment rate of the area adopting it. In this chapter we empirically validate that model by estimating a reduced form model for Ohio, using census data and data on tax incentives in the state.

4.4

Tax incentive programmes in Ohio

In Ohio there are two tax incentive programmes – the EZ and the CRA. The CRA is a tax incentive programme in Ohio that provides tax incentives for property improvements that could result in job creation. The EZ is the major tax incentive programme in the state and is the focus of our study here,

Tax Incentives and Unemployment in Ohio 69

except where controls for the CRA programme are necessary and data are available. In Ohio’s EZ programme, there are two types of zones that are allowed: full authority zones and limited authority zones. Full authority zones are distress-based. They have to satisfy at least one of six distress criteria: (i) It must have had, during the most recent 12 months, 125 per cent of the state’s average unemployment rate; (ii) At least 10 per cent population loss must have occurred between 1970 and 1990; (iii) A minimum of 5 per cent of vacant, demolished commercial or industrial facilities must be prevalent; (iv) At least 51 per cent of the population must be below 80 per cent of the area’s median income; (v) There must be specific vacant industrial facilities (zone incentives apply only to those facilities); (vi) Income-weighted tax capacity of the school district must be below 70 per cent of the state average. On the other hand, it is sufficient for limited authority zones to demonstrate minimum population requirements for them to be designated as zones. These requirements are that EZs proposed within counties of population greater than 300,000 must have a minimum population of 4,000. Further, EZs proposed within counties of population less than 300,000 must have a minimum population of 1,000. As of December 1997, there were 44 distressbased zones and 278 limited authority zones in Ohio, one of the largest numbers to be found in any state in the United States. This is important because both distressed full-authority zones and nondistressed limited authority zones in Ohio can offer tax incentives to firms that locate there. This could have some implications for the targeting of areas, similar to findings based on empirical work from Illinois EZs in Chapter 3. Once a community receives EZ certification, Ohio state law permits local officials to negotiate a tax incentive agreement with a prospective firm. Once an area is certified as an EZ, firms can then enter into agreements with the local governments (having jurisdiction over the zones) and make commitments regarding investment, job creation, retention and payroll. Ohio law states that the amount and term of the tax exemption are to be negotiated between local officials and the firm. However, the law places limits on the level of incentives. It permits municipalities to exempt real4 and/or personal property5 assessed values of up to 75 per cent for up to 15 years, or an average of 60 per cent over the term of the agreement. The exemption can be provided to new investments in buildings, machinery/equipment and inventory, and improvements to existing land and buildings for a specific project. The state’s EZ law permits unincorporated areas to exempt real and/or personal

70 Incentives for Regional Development

property assessed values of up to 60 per cent for up to 15 years, or an average of 50 per cent over the term of the agreement. As in the case of municipalities, the exemption is allowed on new investments in buildings, machinery/ equipment and inventory and improvements to existing land and buildings for a specific project. All abatements and exemptions may be granted only with approval by the affected Board of Education. Note that all are incentives to capital. Since tax incentives have far-reaching consequences on the allocation of resources, as demonstrated by the model in Chapter 2, we have to study their effects.

4.5

Implementation of the model

The standard approach in the literature to assess the impact of tax incentives has been to estimate an econometric model with an outcome variable of interest. For instance, Papke (1994) estimates unemployment claims in Indiana’s EZ programme as a function of a dummy for the programme, implying the offer of incentives. I adopt a similar approach here by estimating the model shown in equation (4.1) in reduced form.6 Here the outcome variable is the unemployment rate and I estimate it as a function of other variables, including tax incentives. The estimation of unemployment rate in reduced form is consistent with the literature (Pantuosco and Parker, 1998, and other studies summarized earlier in the chapter) which shows that socio-demographic characteristics such as age, race, sex, and education, determine the unemployment rate. Here, these socio-demographic characteristics are used as exogenous variables that explain the unemployment rate, instead of wages and reservation wages, along with tax incentives, as the model in Chapter 2 shows. Figure 4.1 illustrates diagrammatically the determination of unemployment rate. It shows the reduced form of the model in which it is estimable. The basic model of unemployment in the ith

Unemployment rate of area

Aggregate wages

Tax incentive programme

Duration of programme

Aggregate reservation wages

Age, education, race, sex, occupation & industry mix Factors determining zone designation – county unemployment, vacant industrial facilities, poverty rate & population loss

Figure 4.1 Determination of the unemployment rate

Exogenous factors (age, education, race, & gender composition)

Tax Incentives and Unemployment in Ohio 71

area (census block group, as I explain later, in the section on data) may be summarized in equation (4.2): Ui Dummy for tax incentive programme (EZ or CRA7)i Duration of tax incentive programmei Manufacturing employmenti Service employmenti Employment in technical occupationsi Employment in service occupationsi Proportion African Americani Proportion malei Proportion with bachelor’s degreesi Mean agei ei

(4.2)

The unemployment rate is estimated as a function of the variables shown in equation (4.2). The hypothesis from the conceptual model presented in this book is that areas with tax incentive programmes have a lower unemployment rate than those that have no such programmes. While the theoretical effects are clear, there is an empirical complication in the context of Ohio’s EZ programme. Observe from Figure 4.1 and equation (4.2) that tax incentive programmes and the unemployment rate are determined simultaneously since high unemployment rate is a criterion for zone designation in Ohio (as full authority zone).8 Because of this problem, it is possible that causality can be reversed in a regression of unemployment as a function of tax incentive programmes. I explain the procedure to alleviate the endogeneity between zone designation and unemployment in the section on methodology. In the empirical work, if, after accounting for the treatment effects problem, it were found that the coefficient on the tax incentive variable is negative and significant, then the hypothesis that tax incentives decrease the area’s unemployment rate, would be confirmed, consistent with the model. In addition to the effect of the tax incentive programme on the rate of unemployment, I control for the time period (duration) for which the programme has been in existence in the area.9 One can imagine that the tax incentive programme reduces the unemployment rate of an area, but there could be some optimum period for which it is desirable. This effect makes it necessary to control for duration of the programme by including it as another variable. The duration of the tax incentive programme is treated as being exogenous to the model. Discussions with the Ohio Department of Development (which administers the EZ programme in the state) indicated that the timing of zone designation10 chosen by areas is a non-random process. Specifically, the duration of tax incentive programmes in Ohio depends upon factors such as pro-growth coalitions, political support, and the bureaucratic and institutional structures needed to implement the incentives in the communities. These are, however, not measurable (since there exists no quantifiable data on these characteristics at the zone level), so I am unable to explicitly control for them.

72 Incentives for Regional Development

There is another reason why duration in the sample used for empirical work in this chapter is empirically random with respect to the unemployment rate. If the duration of the tax incentive programme had systematically varied across zones and non-zones, taking zero values only for low unemployment areas, there may have been a bias in the estimates. But in the sample, this problem does not arise because there are a large number of non-zone census block groups with higher unemployment rates (than zones) for which the duration variable takes a zero value.11 The expected impact on the unemployment rate of the duration of the tax incentive programme is ambiguous. Note that the introduction of time introduces some degree of dynamism in the static model.12 One may surmise that the unemployment rate might vary in a non-linear fashion with respect to duration – at first increasing, then decreasing, or decreasing at first and then increasing. In order to track this ambiguity, the squared term of duration was included as an independent variable, along with duration. If duration of the programme had a positive impact and its squared had a negative impact on the unemployment rate, then it may be inferred that the tax incentive programme initially increases but eventually reduces the unemployment rate of the area. The impact of the tax incentive programme can be determined along with the duration variables. If the tax incentive dummy were found to be affecting the unemployment rate significantly, the duration variable would enable us to understand the time period over which the impact of tax incentives can be expected to be valid.13 In the estimation of the unemployment rate, I also control for the occupational composition of the area. Note that different occupations could have different unemployment rates as the result of different demand conditions. For instance, those employed in routine administrative and technical occupations might be less likely to lose their jobs at any given point in the business cycle than those in service occupations such as private household occupations which are much more easily substitutable. In addition, in the estimation, I include measures that reflect the area’s manufacturing base, such as the proportion of total employment in manufacturing as compared to services, in order to test for their effect on the unemployment rate. These measures reflect the importance of these sectors in the area’s economic base, and it is necessary to control for them. I expect the proportion of manufacturing employment in the area to have a positive impact, and the proportion in service employment to have a negative impact on the unemployment rate. This is because of the decreasing importance of manufacturing and the increasing significance of services in the economic base of most of the local economies. Demographic characteristics such as race, age, sex and education influence an area’s unemployment and therefore are also included in the model. I expect areas with larger proportions of African Americans to have higher levels of

Tax Incentives and Unemployment in Ohio 73

unemployment (positive effect). On the whole, African Americans are likely to have lower education, poorer skills and are more likely to be unemployed. For similar reasons, areas with a higher proportion of women are likely to have higher levels of unemployment. The expected effect is then a negative sign on the proportion male. Areas with higher average levels of education are expected to have lower levels of unemployment. This is because education implies the presence of certain skills. If job opportunities do not exist in the area for these skills, they are likely to move out of the area, reducing the labour force, in which case the unemployment rate of the area would be lower. This is consistent with the conceptual model that points to the immobility of less-educated individuals in the EZ area, and their consequent unemployment. Thus, the unemployment rate of better-educated areas is likely to be lower (due to higher mobility of the better-educated labour force). The sign on the variable representing the proportion with bachelor’s degrees is therefore expected to be negative. Finally, older persons are likely to have greater levels of work experience and will be employed. So areas with higher average age are likely to have a lower unemployment rate. The expected sign on the age variable is negative.

4.6

Data

Unemployment rates did not exist by EZ or tax incentive area in Ohio, but self-reports of personal employment at the block group level were available for 1990 from the (Ohio) Census summary tape files (STF3A), at the time this research was completed. The self-reported data on employment status in the Census are based on the responses to questions asking for detailed information regarding persons’ activity in a reference week. In the Census, all civilians 16 years and over, that are one of the following: (i) ‘at work’ as paid employees, working in their own business, profession, or farm; (ii) ‘with a job, but not at work’ during the reference week due to illnesses, bad weather, industrial dispute, vacation, or other personal reasons; are classified as employed. Those that are not in the categories ‘at work’ or ‘with a job, but not at work’ during the reference week, and (i) have looked for work during the last four weeks; (ii) are available to accept a job; (iii) those waiting to be called back to a job from which they had been laid off; are all classified as unemployed in the US Census.14

74 Incentives for Regional Development

One disadvantage of self-reported data from the Census is that it is very dependent on the reference week, which is not the same for all respondents since the enumeration is not completed in a week. Since persons can change their employment status from one week to the next, the lack of a uniform reference week may mean that the employment data will not reflect the reality of the employment situation of any given week. However, since I am not considering unemployment during any specific period in a year, the reference week is not a limitation. Nevertheless, the superiority of data on self-reported employment when compared to data on unemployment claims which Papke (1994) uses in her evaluation of Indiana’s EZ programme is clear. First, self-reported employment data cover all civilians aged 16 years and over. Although the census data are based on a sample, appropriate weighting is undertaken to ensure the representativeness of the sample.15 Data on unemployment compensation claims, however, exclude the following: (i) Persons who have exhausted their benefit rights; (ii) New workers who have not earned rights to unemployment insurance; and (iii) Persons losing jobs not covered by unemployment insurance systems in the US (including workers in agriculture, domestic services, religious organizations, self-employed and unpaid family workers). Based on self-reported data, I compute unemployment rates for Ohio’s census block groups. Census data was also the only way in which I was able to determine whether or not the census block groups had a tax incentive programme in place. I did this by overlaying a map of Ohio’s 1997 enterprise zones on that of its 1991 census block groups.16 This enabled me to determine which block groups had incentive programmes in place. I find that of the total 11,621 census block groups of the state (Ohio had a total of 11,621 census block groups in 1991), 8,300 block groups had EZs. An additional 1,000 block groups had CRAs. Thus, out of the total of 11,621 census block groups in Ohio, a total of 9,300 (or 80 per cent) block groups were tax incentive areas in 1997, leaving the remaining 2,321 block groups as non-tax incentive areas. The availability of data on other demographic characteristics in the census block group (STF3A) files made it possible to estimate these characteristics for the census block groups of the state. Relevant characteristics in addition to self-reported employment, on which data are available for persons at the census block group level, are age, whether African American, whether male or female and the level of educational attainment. This enabled estimation of the unemployment rate as a function of tax incentive programmes and other socio-demographic characteristics, using these data for Ohio’s census block groups.

Tax Incentives and Unemployment in Ohio 75

4.7

Methodology

The estimation of the unemployment rate is carried out by taking into account the treatment effects problem caused by the endogeneity of tax incentives. At the time that this research was completed, of the 322 EZs in Ohio, the majority (278 zones) were designated on the basis of population (being limited authority zones), and the rest (44) were distress-based (satisfying at least one of six criteria outlined in section 4.4). The Ohio Department of Development had no data on how many of the 44 distress-based zones were high-unemployment areas (having 125 per cent of the state’s average for the last 12 months). If a high correlation were to be assumed between the various characteristics of distress, it may be reasonably assumed that the 44 distressed zones are also high-unemployment areas. 4.7.1

Treatment effects problem

The treatment effects problem occurs because of the endogeneity of zone designation in a regression of unemployment on zone designation (or tax incentive dummy). This treatment effects (endogeneity) problem is alleviated by adopting the instrumental variable approach. Exogenous characteristics that determine zone designation (median income, prevalence of vacant commercial or industrial property, population loss, and unemployment) are used as instruments in two-stage least squares (2SLS) estimations of the unemployment rate. Note that instruments are those variables that are not influenced by zone designation, at least in the short run, but which do influence zone designation. The area’s poverty rate is used as a measure of the income criterion required for zone designation, and is used as an instrument. Observe that while the long-run poverty rate of the area might be influenced by zone designation, it does not impact zone designation directly, making it desirable as an instrument. Note that criterion (4) for designation as full-authority zone in the state refers to a certain proportion of the population being below 80 per cent of the area’s median income. Another instrument that is used is the proportion of vacant housing units in an area. This is used as a measure of the prevalence of vacant or demolished commercial or industrial facilities in the area, required for EZ designation. Data on housing units are used because only data on residential structures are reported in the census.17 The unemployment rate for the county (in which the census block group is located) at the time of zone designation (1982) is used as another instrument. The county’s unemployment rate undoubtedly influences zone designation, but is not influenced by zone designation, making it desirable as an instrument. Finally, I use two measures of net migration into the area18 – one at the census tract level and the other at the county level – as instruments that determine zone designation. These are used as measures of the population

76 Incentives for Regional Development

loss criterion for zone designation. We expect that areas that experience little in-migration are also those that experience population loss. The tract-level measure is only a dummy because the areas of census tracts changed over the period 1980–90, making direct computation of net change in population at the tract level impossible. At the county level, it is possible to compute the magnitude of the population change and so I use the net change in population over the period 1980–90 as an instrument. Since there was no data on taxable capacity at the census block group or tract level, I was unable to use measures of this variable as an instrument. However, it is easy to imagine that there is a high correlation between the poverty rate of an area and its taxable capacity. Areas that have larger proportions of their populations below the poverty level are frequently those whose school districts have their income-weighted tax capacity below the state average. Therefore the poverty rate, used as an instrument in the estimation, is a good measure of the income-weighted tax capacity criterion as well. Data on all these instruments at the block group level were obtained from the US Census, with the exception of the county-level population change for 1980–90, which was obtained from the state of Ohio census data centre. Equation (4.2) was estimated by OLS and 2SLS, by including as instruments these factors that affect zone designation. 4.7.2

Variable definitions

1. The block group’s unemployment rate (the dependent variable) is computed according to the method recommended by the Ohio Department of Development. According to this procedure, I first estimate the unemployment rate for all the counties based on the ratio of unemployed to total labour force aggregated in the 1990 census at the county level. I then estimate a similar ratio for the block groups, obtaining a raw unemployment rate for every block group. I take the ratio of block group to county unemployment rates computed in this way. This ratio multiplied by the published county unemployment rates (published by the Ohio Bureau of Employment Services (OBES)) yields the desired block group unemployment rate for use in the estimation. 2. The indicator for tax incentive programmes, is a dummy equal to one if a tax incentive programme (EZ or CRA) is in place (and zero if not). The definition of the EZ dummy is easily done because of the overlaying of map of the EZs over that of the block groups. The dummy for the CRA programme was more difficult to arrive at. I found whether or not CRA and EZ areas overlap by talking to the 282 CRAs in the state of Ohio (all certified in the pre-1994 period).19 When my discussions with the communities indicated that, in areas having both the CRA and the EZ, the CRA was non-overlapping with the EZ, I identified the census tract numbers for these cities from the census STF3A files and included them as areas having tax incentive programmes. Thus areas that had non-overlapping CRAs and EZs and those that

Tax Incentives and Unemployment in Ohio 77

did not have EZ, but used the CRA for purposes of job creation (that is, provided CRA abatements to industries) were identified as having a tax incentive programme.20 It is possible for abated residential property to exist in non-tax incentive areas, according to the way in which the variable is defined. Although abatements to residential property could result in job creation (as with the CRA programme), I do not account for these incentives. The tax abatements granted in a majority of the CRAs in Ohio are for commercial and industrial uses (based on my discussion with the CRA administrators). Not accounting for residential incentives of the CRA programme should not affect the results much. Thus, the non-tax incentive areas (with zero for the indicator dummy) in the estimation are those areas in which no tax incentives are provided to commercial/industrial property that resulted in job creation. 3. The duration for which the EZ has been in existence is measured as the time period (in years) since certification of the zone until the end of 1990.21 Although the GIS map I had of Ohio’s zones was from 1996, I defined the duration variable as of 1990 because the census block group map I had of the state was from 1990, and the unit of estimation was the census block group. Thus, zones certified in June 1990 are in place for 0.5 years, those certified in September 1990 have 0.3 years for their duration, and so on. As we expect, zones certified early on in the programme had the longest duration of existence. There are a number of zones that are certified after 1990 and the duration variable for these zones takes a zero value. For the non-EZ block groups, the duration variable takes a zero value. Since even for zone areas, zero is used to represent if it is not in place in 1990, this is a valid representation of the non-EZ areas too, since at that time they are not designated a zone. 4. Duration squared is self-explanatory, it is calculated as duration

duration for every zone. 5. The area’s manufacturing base is measured as the percentage employed in manufacturing industry jobs as a proportion of all employment in the block group.22 6. The area’s service base is the percentage employed in service industry jobs as a proportion of employment in the block group. Note that typically, the variable defined in (5) that in (6) percentage employed in other industries add up to 100 per cent.23 7. The area’s occupational composition in professional specialty, technical and administrative occupations is measured as the percentage of its employment in managerial and professional specialty occupations,24 technical, sales and administrative support occupations.25 8. The area’s base in service occupations refers to the proportion of employment in private household occupations, protective service occupations, and other service occupations. Note that, again, the variable described in (7) that in (8) percentage employed in other occupations add up to 100 per cent.26

78 Incentives for Regional Development

9. A dummy is created at the census tract level if population loss occurred in a census tract over the period 1980–90. If a census tract in 1980 is subdivided in 1990 (into a larger number of block groups, this is usually done if a tract grew over the period) or if its population increased over the period 1980–90, the population loss dummy for the census tract is determined as being zero. Otherwise it is one meaning that population loss occurred in the tract over the period. An alternative measure of the population instrument, net population change (at the county level), is computed as the change in population (in thousands) from 1980 to 1990. It is defined as the difference between the 1990 and 1980 populations for a county. If this variable is positive, the county gained population over the period 1980–90, otherwise it lost. 10. The poverty rate (multiplied by 100 for a percentage) is calculated as the ratio of the number of persons in the block group that had 1989 incomes below the poverty level to all persons in the block group. Note the difference between the way this instrument is calculated, and criterion (4) used for zone designation. Obviously they are different, and we may be reasonably sure that the instrument used here influences, but is not influenced by zone designation, at least in the short-run. 11. The percentage of vacant housing units is calculated as the ratio of vacant housing units to all housing units (occupied and vacant) in the block group (again multiplied by 100 for percentage). 12. For the county unemployment rate, the unemployment rate of the county (in which the block group is located) at the time of zone designation (1982) is used. 13. Age is measured in years. I have excluded ages below 17 in order to ensure that no children are included in the estimations, since we are concerned only with working age groups. 14. The dummy for gender is one if male, and zero if female. 15. The proportion African American is calculated as the ratio of persons with African American origin to all persons in the block group. 16. The proportion with bachelor’s degrees is computed as the ratio of those with bachelor’s degrees to all persons above 17 years in the block group. Table 4.1 shows the descriptive statistics for the variables used in the estimation. The descriptive statistics are for the 11,445 census block groups (the entire state of Ohio consists of 11,621 census block groups) used in the estimation.27 The mean for the tax incentive dummy shows that 78 per cent of areas (that is, 8,961 out of the 11,445 block groups) in the state had a tax incentive programme in place in 1990, leaving only 22 per cent (2,484 block groups) as non-tax incentive areas, a finding that reinforces the extensive nature of such programmes in the state. There is a lot of variation in the unemployment rate, with a high of 83.08 per cent (this being in Lucas County, census tract 25 and block group

Tax Incentives and Unemployment in Ohio 79 Table 4.1 Descriptive statistics for variables Variable

Block group unemployment rate Tax incentive programme Duration of programme Duration of programme squared % African American % bachelors’ degrees % male Age (years) Dummy for population loss County population change (in thousands) % manufacturing employment % service employment % professional employment % service occupation Poverty rate % vacant housing units

Mean

Standard deviation

Minimum

6.85 0.78 1.46 5.94 12.55 9.34 48.15 44.35 0.53 22.14

6.66 0.41 1.95 10.30 26.42 8.88 5.42 5.05 0.50 69.57

0 0 0 0 0 0 0 20 0 207

23.37 31.47 51.77 14.79 4.52 6.50

11.14 11.97 17.72 9.31 15.82 7.40

0 0 0 0 0 0

Maximum

83.08 1 7 49 100 100 100 73 1 92.31 100 100 100 100 100 100

number seven, which covers part of the City of Toledo EZ28). There are a few census block groups that recorded a zero per cent unemployment rate in 1990.29 On average, as of 1990, the tax incentive areas in Ohio had been in existence for only a little longer than a year, suggesting that there were a large number of new zones designated. As of 1990 some zones/tax incentive areas has been in existence for seven years, having been designated when the tax incentive programme came into existence in the state. The dummy for population loss indicates that, on average, 53 per cent of the state’s census tracts had experienced population loss over the period 1980–90. The data on net population change for the county in which the block group is located reinforce this, indicating that, on average, each county lost about 22,000 residents over the period 1980–90. On average, about 12 per cent of the population in the block groups is African American. On average, the population in the 11,445 block groups is middle-aged, with men and women about equally distributed. Finally, as one expects, the proportion with bachelor’s degrees is less than 10 per cent in all the block groups on average. Based on the average, Ohio’s economic base is largely constituted of services (which has 31 per cent of total employment), and manufacturing employment (which has 23 per cent of total employment on average). In terms of occupational composition, a majority (52 per cent) of the employment is in professional specialty, technical and administrative occupations when compared to service occupations such as household services (about 15 per cent),

80 Incentives for Regional Development

implying a higher level of occupational skills. On average, only about 4 per cent of the population in the block groups had 1989 incomes below the poverty level. Finally, the data for vacant housing units indicate that, on average, about 7 per cent of housing units were vacant at the time of the 1990 Census survey. This implies that a majority of areas would qualify for zone designation as full authority zones, since the prevalence of a minimum of 5 per cent vacant industrial/commercial facilities is required for zone designation in the case of full-authority zones. Does this mean that the large number of limited authority zones in Ohio, would in fact be justified on grounds of distress, as full-authority zones? Note from Table 4.1 that the standard deviation for the proportion of vacant housing units is greater than the mean, showing the huge variability in this variable, as a measure of distress. In the next section, I report the results from the estimation of unemployment rate taking into account the treatment effects problem.

4.8

Results from the estimation

Table 4.2 shows the results from the OLS and 2SLS estimation of the unemployment rate.30 The OLS estimation considers the tax incentive dummy to be exogenous and is a simple regression of unemployment rate as a function of all independent variables shown in equation (4.2). OLS estimates are also presented along with 2SLS estimation results, because the OLS

Table 4.2 Estimation of the unemployment rate Variable Constant Tax incentive programme Duration of programme Duration squared % manufacturing employment % service employment % professional employment % service occupation % African American Mean age Percentage male % bachelors’ degrees N R2

OLS estimates (SE)a

2SLS estimates (SE)a

12.266 (1.678)*** 0.0303 (0.1127) 0.4595 (0.1157)*** 0.0256 (0.0231) 0.0323 (0.0112)*** 0.0332 (0.0113)*** 0.0895 (0.0105)*** 0.0638 (0.0184)*** 0.0649 (0.0039)*** 0.0585 (0.0162)*** 0.0334 (0.0201)* 0.0745 (0.0135)*** 11,445 0.32

16.245 (1.984)*** 4.5263 (0.9469)*** 1.8177 (0.3031)*** 0.2173 (0.0458)*** 0.0342 (0.0113)*** 0.0260 (0.0116) 0.0901 (0.0106)*** 0.0626 (0.0186)*** 0.0616 (0.0040)*** 0.0816 (0.0177)*** 0.0296 (0.0204) 0.0904 (0.0142)*** 11,445 0.26

*** Statistically significant at the 1 per cent level. * Statistically significant at the 10 per cent level. a. SE refers to standard error of the estimate.

Tax Incentives and Unemployment in Ohio 81

estimates interestingly show the endogeneity of the tax incentive dummy. The 2SLS estimation considers tax incentives as endogenous, and uses factors that determine zone designation (or tax incentives) as instruments for the endogenous tax incentive dummy. The impact on the unemployment rate of tax incentives shown in the regressions is plausible if the assumptions of the estimation methods are taken into account. The OLS specification shows that tax incentives do not have a statistically significant impact on the unemployment rate of areas. When combined with the statistically significant duration variable and its squared, the coefficient on tax incentives in the OLS estimation may be interpreted as follows. In fact, an area with a tax incentive programme sees an increase in its unemployment rate by about 0.46 percentage points {0.0303 (coefficient on the tax incentive dummy in the OLS estimation, Table 4.2)} {0.4595 (coefficient on the duration variable in the OLS estimation) 1 year} {0.0256 (coefficient on the duration squared variable in the OLS estimation) 12} in a year. Areas that have tax incentive programmes for two years see an increase in their unemployment rate by about 0.84 points (0.0303 0.4595 2 years 0.0256 22), and so forth. Thus, with the OLS estimation, it appears as if tax incentive programmes actually increase unemployment in areas! In reality, however, the tax incentive programme is such that the causality runs in the opposite direction: areas are designated as EZs (full authority zones) or tax incentive areas because they have high unemployment rates (or have one of four other distress criteria). When the endogeneity of tax incentives is taken into account (in the 2SLS specification), the coefficient on the tax incentive dummy has a negative sign, which is statistically significant. This indicates that when their endogeneity in the zone designation process is accounted for, tax incentive programmes have some impact in reducing the unemployment rate of areas, consistent with our expectations. Further, when we consider the impact of the tax incentive along with the duration of its existence, we can say something about its net impact. The net impact of being a tax incentive area versus not being one for periods of one, two, three, four, five, six, and seven years is respectively,31 2.92 percentage points (4.5263 1.8177 1 0.2173 12), 1.76 percentage points (4.5263 1.8177

2 0.2173 4), 1.03, 0.73, 0.87, 1.44, and 2.45 percentage points respectively.32 Thus, if an EZ (or other tax incentive area) had ten unemployed persons out of a labour force of 100 persons in 1982 (when designated, with an unemployment rate of 10 per cent), the model predicts that the area’s unemployment in 1983 would be (2.92 percentage points less) 7.08 per cent. Taking into account the fact that a three percentage point reduction over a period of one year translates to about three persons obtaining gainful employment with the EZ or other tax abatement programme in place, these results are plausible.

82 Incentives for Regional Development

Given that the mean unemployment rate is 6.84 per cent (Table 4.1), should one be sceptical about the estimate, which shows that the unemployment rate moves 2.9 percentage points in the first year? Remember that the estimation of unemployment rate is carried out at the census block group level. The variation in the unemployment rate across the 11,445 census block groups is quite large, with a minimum of zero and a maximum recorded figure of 83.08 per cent (Table 4.1). Given such a large variation, it is easy to see the 2.9 percentage point reduction. Consider the example of a firm that creates 100 jobs in the first year of its existence in the zone. Assume that 97 out of these 100 new jobs are allocated to labour force that might migrate into the area with increased job growth, consistent with what the migration literature (see Bartik (1991) for a summary of this literature) has argued. The remaining three jobs are held by zone residents. Assuming that these three are among the original ten unemployed in the zone, the unemployment of the area decreases by three percentage points (from 10 to 7 per cent) in the first year. Another qualification to this result: it should be remembered that the tax incentive dummy in the regression is used as a measure of a tax abatement programme in place – either the EZ or the CRA tax incentives for job creation. Based on this, from the data, the effectiveness of EZ programme or tax incentives in general is indeed big enough to register. These results, while plausible, appear to be less in their impact than those reported by Papke (1994) in her evaluation of Indiana’s EZ programme. Papke reports a 19 per cent reduction in unemployment claims in Indiana as a result of the existence of the EZ programme. The difference in the magnitude of the impact of tax incentives between this work and Papke’s study could be due to the way in which the presence of the tax incentive programme and the unemployment rate are measured in the two studies. Papke’s variable measures the impact of the EZ considered in isolation. Here, the dummy for tax incentive programme takes into account the other tax incentive programme (CRA) in Ohio, in addition to the EZ. If the other tax incentive programme in Ohio, the CRA programme, is not accounted for, the effect of tax incentives may not have been measured adequately. Therefore, in the context of Ohio, use of the general dummy for tax incentive programmes is more appropriate than the single measure of the EZ that Papke uses. Papke also uses data on unemployment claims in Indiana whereas I use the actual unemployment rate. The superiority of self-reported data on employment when compared to that on unemployment claims is clear, as discussed earlier. From this point of view, the impact of tax incentive programmes obtained here is much more plausible than the 19 per cent reduction in unemployment claims Papke reports as being solely due to the EZ. While the reason for differences in the magnitude of coefficients across Papke’s and this study seems obvious, the results are consistent and both show

Tax Incentives and Unemployment in Ohio 83

that EZ or CRA programmes have the impact of reducing the unemployment rate of the area adopting it. In general, the other variables in the estimation (shown in Table 4.2) have the expected effects on unemployment. The area’s manufacturing base and the proportion of employment in service occupations, increase an area’s unemployment rate, consistent with expectations. The percentage of employment in technical and administrative occupations has a negative impact on unemployment, showing that the higher the percentage employed in such routine occupations, the lower the unemployment rate the area. This is what one would expect because these are services which would not decline significantly even during periods of high unemployment. The percentage of an area’s employment in the service industry has a positive impact in the OLS specification, contrary to expectation, but loses its significance in the 2SLS version. The demographic structure of an area also influences its unemployment rate. Higher proportions of African Americans increase such areas’ unemployment rate and, as expected, the effect is statistically significant in the regressions. This is because African Americans, on average, have poorer skills. It is also possible that they have family and social ties that make them less mobile than others, increasing unemployment. As expected, younger areas (that is, those where the mean age is lower) and those with a lower proportion of men have higher unemployment rates. Younger persons are likely to have less work experience and hence have a longer search for jobs, thereby increasing the area’s rate of unemployment. Men are more likely to be employed because of better matching of skills and opportunities and they, being the primary income earners in most households, are employed. The percentage with bachelor’s degrees in all the specifications has the expected sign showing that the higher the proportion of college graduates, the lower is the unemployment rate of the area to the extent of about 0.10 percentage points (in the 2SLS specification). Thus better-educated areas have lower unemployment rates, which is to be expected. The R2 indicates that the model explains 26 per cent of the unemployment rate – a fairly low figure. However, it should be remembered that the R2 is only a descriptive statistic. In cross-sectional data, a lower R2 might occur even if the model is a satisfactory one, because of the large variation across individual units of observation (Pindyck and Rubinfeld, 1991). This suggests that the R2 alone may not be a suitable measure of the explanatory power of a model. The final table (Table 4.3) replicates the specification in column three of Table 4.2 with the duration variables as dummies rather than as continuous variables, following the literature. In this instance, the same variables as in Table 4.2 are used as instruments for the endogenous tax incentive dummy.

84 Incentives for Regional Development Table 4.3 Estimation of the unemployment rate with dummies for duration Variable Constant Dummy for tax incentive programme % manufacturing employment % service employment % professional employment % service occupation % African American Mean age Percentage male % bachelors’ degrees Duration of programmes: 0–1 year Duration of programmes: 1–3 years Duration of programmes: 3–5 years Duration of programmes: 5–7 years N R2

2SLS estimates SE 18.949 (1.669)*** 6.9831 (0.8635)*** 0.0314 (0.0086)*** 0.0289 (0.0091)*** 0.0850 (0.0080)*** 0.0392 (0.0142)*** 0.0534 (0.0033)*** 0.1087 (0.0149)*** 0.0335 (0.0166)** 0.1037 (0.0117)*** 3.5911 (0.4059)*** 3.8050 (0.4024)*** 4.6721 (0.4240)*** 4.7664 (0.5926)*** 11,445 0.16

*** Statistically significant at the 1 per cent level. ** Statistically significant at the 5 per cent level.

In Table 4.3, signs on most independent variables (except that on the proportion of employment in service industry) remain as in the previous estimation, consistent with our expectations. The proportion of employment in the service industry has a positive impact on the unemployment rate, indicating that the higher the proportion of jobs in the service industry, the higher is the unemployment rate in the area. This effect is due to the fact that there are areas in the sample that have no service industry employment, but also have a low unemployment rate. This also explains why this variable has a positive impact in the OLS specification in Table 4.2. Table 4.3 shows that tax incentives continue to have the expected negative sign, which is statistically significant at the 1 per cent level. When combined with the effect of the signs on the duration dummies, the impact for areas that have been tax incentive areas for a year is a decrease in their unemployment rate to the extent of 6.9831 3.5911 1 3.39 percentage points. The impact for an area that has been providing tax incentives anywhere between one and three years is (6.9831 (3.8050 2)) 0.63 percentage points. The impacts for areas having tax incentives from three to five years and five to seven years are, respectively, increases in their unemployment rate to the extent of 12 (6.9831 (4.6721 4)) and 22 percentage (6.9831 (4.7664 6)) points respectively.33 It is a robust finding across the estimations that, initially, tax incentives have a statistically significant impact in reducing unemployment in those areas that

Tax Incentives and Unemployment in Ohio 85

adopt them, consistent with the objective for which the programme was initiated in Ohio in 1982. Beyond the initial period, however, if tax incentives are continued, the unemployment problems of areas will recur.

4.9

Policy implications

While the regressions as a whole support the hypothesis that areas with tax incentives witness a reduction in their levels of unemployment, there is less consensus regarding the duration over which the tax incentives should be offered. The regression in which durations are included as dummies, suggests that three to five years could be the optimum period for maximizing the effect of tax incentives on unemployment rate, after which it is preferable that the area ceases to offer incentives. In terms of policy action, this translates into decertifying areas that have been tax incentive (EZ or CRA) areas for longer than five years. If incentives continue for periods of longer than three to five years, it is likely that their unemployment problems will return. The reason for this is that, after a while, tax incentives could degenerate into employment redistribution games that have the effect of increasing the unemployment in those areas that lose firms. However, it should be remembered that this research does not throw any light on the zero-sum nature of tax incentive policies such as EZs. I have not undertaken any analysis, real or hypothetical (as in Chapter 3), to show the effects of redistributing employment from one area to another. Further, this chapter does not consider the question of whether the high cost of generating a few jobs with generous tax abatements is worth the modest gains in reducing unemployment, something that will be dealt with in the following chapter. It must be acknowledged that tax incentives, being place-oriented policies, are only one way of increasing employment. There might be other, more efficient ways of reducing area unemployment – for example, peopleoriented schemes such as relocation and training programmes.

4.10

Concluding remarks

In any case, increasing employment has the effect of increasing households’ labour force participation rates (measured as the ratio of employment to population). This has hysteresis effects on the labour market, as Bartik (1991) rightly argued. Hysteresis is used to explain how the electromagnetic properties of certain metals are permanently affected by the temporary application of certain magnetic forces. When applied to the labour market, this means that the increase in the labour force participation rate as the result of increased employment increases households’ employability (due to training and acquisition of skills) in the long run. From this point of view, tax incentive programmes in Ohio most probably created these hysteresis effects in

86 Incentives for Regional Development

the labour market through increasing productivity, employment, and labour force participation. It should be noted that this chapter has only considered the impact of the existence or otherwise of tax incentive programmes on the unemployment rate. It has not, however, made an examination of the counterfactual, in which identical control groups are used, and the tax incentive treatment effect is introduced in both in order to assess its effects. Given this result, the next chapter considers the benefit–cost analysis of Ohio’s EZ programme, making various assumptions about employment created in the zones.

5 Benefits and Costs of Regional Development: Evidence from Ohio’s Enterprise Zone Programme

5.1

Introduction

As highlighted in Chapter 4, the challenge that has been raised against Ohio’s enterprise zone programme is that it is a means of lobbying by regions to get zone designation. Thus, most areas in Ohio enjoy the associated benefits of being able to offer incentives even though they may not deserve to be doing so, as in the case of limited authority zones. This results in the proliferation of such programmes, ending in the ‘pirating’ of firms and jobs from one place to another, that eventually become zero-sum in their effects. This challenge, which has been the topic of many legislative discussions in Ohio and elsewhere, provides the motivation for the research in this chapter. Further, in this chapter, as in Chapter 3, I test the theoretical model we developed in Chapter 2. The question of whether or not the cost of the EZ programme exceeds the economic rent from jobs created in the EZ, at equilibrium, forms the basis for the benefit–cost analysis reported in this chapter. The objective of this chapter is to answer the questions: Are enterprise zones (EZs) efficient? And, further, are they efficient if they are adopted by high-unemployment areas? I evaluate Ohio’s enterprise zone programme and perform a benefit–cost analysis of employment created in these zones. One of the ideas underlying the evaluation of such programmes is to estimate whether the programme is successful in having some impact on the area. The standard approach to this question is to estimate an econometric model (of the outcome of interest) for small areas and include a variable that characterizes the programme under consideration. Chapter 4 takes this approach, and estimates the unemployment rate of Ohio’s census block groups as a function of the tax incentive programme and other variables that labour economics shows determine the rate of unemployment in an area. That chapter reports the finding that tax incentive programmes 87

88 Incentives for Regional Development

(the EZ and CRA programmes) have a significant impact in reducing the unemployment rate in the areas adopting them. This finding forms the basis of the benefit–cost analysis performed of Ohio’s enterprise zone programme in this chapter. That is, since Ohio’s tax incentive programmes have been effective in reducing the unemployment rates of areas that adopt them, it is also worth examining whether the dollar benefits of the EZ programme are worth their costs. In this chapter, I also test a hypothesis that Bartik (1991) developed regarding the relationship between net benefits from employment and the local unemployment rate. Net benefit, or economic rent, is the difference between the wages paid and the wage at which a person is willing to accept a job (or the reservation wage). Bartik argued that the reservation wage would be lower in high-unemployment areas because of the high value the unemployed in a high-unemployment area place on the importance of having a job. This suggests that the relationship between net benefits and unemployment rate can be expected to be positive. 5.1.1

Overview of the chapter

Sections 5.2 and 5.3 respectively review relevant literature and Ohio’s EZ programme only briefly, since Chapter 2 provided a suggestive and critical review of the literature on evaluation of EZs. Further, Ohio’s EZ programme has been described in detail in Chapter 4. Sections 5.4 and 5.5 of this chapter describe the data and methodology adopted to perform benefit–cost analyses of Ohio’s EZ programme, and how it is different from the B–C analyses of the IL programme in Chapter 3. Section 5.5 describes how I impute reservation wages for Ohio’s EZs to estimate net benefits from employment, and how I arrive at measures of programme costs. I then present the benefit–cost analysis of Ohio’s EZs making different assumptions about employment in the zones. The policy implications follow in the final section.

5.2

Past literature

There is a vast body of policy and empirical literature that evaluates enterprise zones. As noted in the earlier chapters, the evidence is mixed. Some of the studies (US Department of HUD, 1986; Erickson and Friedman, 1989; Rubin and Armstrong, 1989; Papke, 1994) are more optimistic, while others (Seyfried, 1990; Dabney, 1991; Boarnet and Bogart, 1996; Dowall, 1996) are pessimistic with regard to enterprise zone effects. Recall that Chapter 2 develops an analytical framework to understand the impact of the enterprise zone on the economy and the general equilibrium response of the tax abatement given to firms in the enterprise zone. This model provides a framework for performing benefit–cost analysis in the empirical work in this chapter. Further, Chapter 3 takes into account net

Evidence from Ohio’s Enterprise Zones 89

benefits from employment, and performs a B–C analysis of enterprise zones in Illinois. That chapter reports the net benefits from employment to be several times the costs, even if it were assumed that all employment relocated to the Illinois EZs from elsewhere, and assuming that the relocation took place from low- to high-unemployment areas. However, this result has to be viewed with caution since the reservation wage estimates (and hence the net benefit estimates) in Chapter 3 are based on a single cross-section of the 1987 panel of the Panel Study of Income Dynamics (PSID). Further, the reservation wage estimates reported in Chapter 3 do not take into account the bias arising from sample selection, since reservation wages are observed only for the unemployed, in the PSID. In this chapter, I use reservation wage estimates that are based on a pooled cross-section and time-series (1984–7) dataset and are corrected for the bias arising from sample selection. Chapter 4 describes results from the standard econometric model and reports that tax incentive programmes in Ohio did make a difference to the rate of unemployment of areas adopting them. We learnt that the unemployment rate was not the same in areas with and without the tax abatement. Further, that tax incentives caused to decrease the unemployment rate by roughly three percentage points in areas having them when compared to those without, with the appropriate caveats about the period for which such incentives can be offered. Thus, while some studies of EZs and tax incentives are more robust than others, the effectiveness of EZs appears to be an unresolved issue in the literature. In this chapter, I perform benefit–cost analysis of Ohio’s enterprise zone programme using more comprehensive measures of net benefits and costs, making different assumptions about employment at the firm level and at the zone level. I find results that are robust to several different assumptions.

5.3

Ohio’s enterprise zone programme

In the state of Ohio, in order for an area to be designated as an enterprise zone, local communities must identify the EZ’s geographical area. The defined area must meet minimum population requirements and have a single continuous boundary. In addition, the area may also fulfil certain other distress criteria that are relevant. I have elaborated on these criteria specified by state’s Department of Development, in Chapter 4 (in Section 4.4). Further, I have also described in that chapter, the tax incentives available to firms locating in EZs, and the limits to tax incentives specified by the law. Recall that Ohio law permits municipalities to exempt real and/or personal property assessed values of up to 75 per cent for up to ten years, or an average of 60 per cent over the term of the agreement. The state’s EZ law permits unincorporated areas to exempt real and/or personal property assessed values of up to 60 per cent for up to ten years or an average of 50 per cent over the term of the agreement.

90 Incentives for Regional Development

It is important for empirical work in this chapter to recall that, in Ohio’s enterprise zone programme, there are two types of zones that are allowed: full authority zones and limited authority zones. Full authority zones are distressbased, and have to satisfy at least one of the six distress criteria described in Chapter 4. As explained there, it is sufficient for the limited authority zones to demonstrate minimum population requirements to be designated as zones. Note that the only check placed on limited authority zones is that they cannot consider intra-state relocation projects unless a waiver is obtained from the Director, Ohio Department of Development. Otherwise, both full and limited authority zones are free to offer tax incentives to firms locating there. As of December 1997, there were 44 distress-based zones and 278 limited authority zones in Ohio, which was one of the few states with a large number of zones in its territory. Specifically, note the large number of limited authority zones designated on the basis of non-distress criteria. The benefit–cost analysis reported in this chapter is based on all zones (full and limited authority zones) of the state for which all data were available.

5.4

Data

In Ohio, the Department of Development administers the enterprise zone programme. The database from the Ohio Department of Development contained information regarding zone number, agreement number, agreement date, expiration date, company name, and its SIC code. The database also contained information on firms’ actual performance with respect to jobs created, retained, payroll from employment created and retained, investment, amount of investment granted exemption, terms (period) of the exemption, property (real and personal property) taxes paid and forgone, and corporation taxes paid. All these performance data were cumulative and included information on all projects in the period 1984–95. This was thus a rich database that allowed testing of the hypotheses of the theoretical model and the research questions by permitting to perform benefit–cost analysis of the programme.

5.5 5.5.1

Benefit–cost methodology Measures of benefits

Measures of programme benefits that were chosen are net benefits from employment – defined as actual wages minus the reservation wage.1 Over and above a labour market perspective attributing benefits to the employed, one might ask whether benefits to the government such as increased income tax revenue and/or decreased unemployment insurance payments are relevant to be taken into account, given that EZs are government-funded. Remember that any increases in income tax revenue or decreases in unemployment insurance are distributional in their effects, merely transferring income from

Evidence from Ohio’s Enterprise Zones 91

taxpayers to government or vice versa. They do not represent real benefits. On the other hand, net benefit from a job is similar to consumer’s surplus (in a goods market) which is the difference between the actual price and that the consumer would have been willing to pay for the good, and so is a measure of increase in welfare. As discussed in Chapter 3, McDonald (1997) argues that job creation is not a valid measure of gain. This is because job creation, at least in Illinois, is coincident with a firm obtaining a building permit to qualify for the sales tax exemption on building materials as shown in Redfield and McDonald (1991), whereas building permits are continually issued and jobs created in the local economy. McDonald argues that it is not clear if this can be attributed to the existence of the EZ or to the tax incentives. In Ohio’s programme, job creation is not tied to the issue of building permits as it is in the Illinois programme. The laws that guide the programme and eligibility in Ohio are tied to the commencement of physical construction. The law speaks of the increase in the assessed value of real property. A company locating in Ohio’s EZ discusses its proposed employment positions at the time it makes its application for the enterprise zone programme (again, an application that has nothing to do with the building permit). The company is allowed some time to create these jobs, typically three years from the date it enters into the enterprise zone agreement. Thus, note that, while the problems with job creation being coincident with building permits (as McDonald (1997) points out), are certainly applicable to Illinois’ programme, they do not apply to Ohio’s programme. In order to compute net benefits from employment (since net benefits are defined as actual wages minus reservation wage), data on reservation wages for Ohio labour force were necessary. Reservation wages estimated from the Panel Study for Income Dynamics (PSID) from Sridhar (1998) were used to impute reservation wages for Ohio’s enterprise zones.2 The PSID is a national panel data set of 6,000 American families, which has been published by the Survey Research Centre at the University of Michigan, Ann Arbor, since 1969. In the PSID, responses to the question ‘What is the lowest wage you would be willing to take home as pay?’ is measured as the reservation wage of the individual in a new job. This question was asked of respondents in the PSID consecutively for eight years from 1980 to 1987. Table 5.1, from Sridhar (1998), shows a model of the reservation wage. The model shows the elasticity of the reservation wage with respect to individual demographic characteristics, labour market characteristics (primarily the unemployment rate) and other job search characteristics (including the duration of job search) for the United States. Note some distinctive characteristics of the reservation wage estimates reported in Table 5.1, from Sridhar (1998), when compared to those reported in Chapter 3 (Table 3.4): (i) The estimates in Table 5.1 are based on data for the period 1984–7; those in Table 3.4 are based on 1987 PSID data only.

92 Incentives for Regional Development Table 5.1 Switching regression model of reservation wage with sample selection (dependent variable: log of reservation wage) Variable

Constant Age Grades completed Work experience Whether African American (1 Yes; 0 No) Male (1)/Female (0) Marital status Number of children Predicted value of duration of search Predicted value of duration of search squared Unemployment rate of county of residence Log of past wage Log of weekly minimum unemployment benefits allowable under state law Log of weekly maximum unemployment benefits Waiting (0 or 1 week) for unemployment benefits r Dependent variable R2 N

Coefficient (standard error) 13.973 (12.71) 0.0709 (0.0624) 0.1559 (0.1428) 0.1119 (0.0899) 1.4125 (1.254)

Means (standard deviation)

39.82 11.76 12.43 0.31

(11.05) (2.47) (13.82) (0.46)

0.06344 (0.6223) 0.9112 (0.9353) 0.1602 (0.1825) 0.1353 (0.0798)* 0.0012 (0.0082)

0.71 0.46 1.15 1.92 6.46

(0.45) (0.50) (1.33) (1.67) (8.32)

0.0324 (0.0543)

7.10 (2.80)

0.5502 (0.0397)*** 0.1108 (0.1189) 1.6533 (1.227) 0.7234 (0.8183) 2.7901 (2.173)

$3.17 (2.71) $23.91 (1.60) $169.13 (1.26) 0.91 (0.29) 2.14 (0.49) $3.25 (2.36)

0.40 737

Notes: * Statistically significant at the 10 per cent level. *** Statistically significant at the 1 per cent level. Source: Sridhar (1998).

(ii) The estimates in Table 5.1 take into account sample selection bias, arising because reservation wages are observed only for the unemployed, in the PSID. The OLS estimates in Chapter 3 do not correct for the sample selection bias. Note the sample selection correction factor is reported in Table 5.1, correcting for a major econometric problem. Since the PSID is a nationally representative sample (see Sridhar (1998) for a description of the weighting scheme employed in the PSID to ensure representativeness), it is assumed that the elasticities based on United States data represent the responsiveness of reservation wages in Ohio as well, to analogous socio-economic characteristics.

Evidence from Ohio’s Enterprise Zones 93

The imputed reservation wage estimates were applied to the various characteristics of persons in Ohio’s enterprise zones to predict reservation wages for them. This was possible because I overlaid a map of Ohio’s census block groups on one showing its enterprise zones.3 This enabled me to determine the census block groups that each enterprise zone in Ohio was comprised of, and to apply the block groups’ characteristics to the zones. This was especially useful because no data on socio-demographic characteristics were available by zone in Ohio. Profile of zones and non-zones Available data on socio-demographic characteristics for persons at the census block group level for Ohio from the STF3A files of the Census were whether unemployed or not, whether African American, male or female, their age, education (grades completed), marital status, and number of children. I computed the unemployment rate (based on self-reported employment status of persons), average age (of persons), mean number of children per family, mean grades completed, and mean proportion African American, proportion married, and male, for Ohio’s census block groups. I used the geographical overlaying to estimate these characteristics for Ohio’s zones and for the non-zone areas. The ultimate objective being to predict reservation wages for Ohio’s EZs, based on these characteristics, which the labour literature shows determine reservation wages, as in Table 5.1. This is much needed for estimation of net benefits from jobs. Before reporting results from the imputation of reservation wages for Ohio, I summarize profiles of the socio-economic characteristics for Ohio’s zones and non-zone areas. Table 5.2 summarizes the profile for the zones. Table 5.2 Profile of Ohio’s enterprise zones, 1997 Characteristic Unemployment rateb (%) Age (in years) Grades completed Proportion African Americanc Proportion malec Proportion marriedc Number of children per family

0th percentile a

50th percentile

100th percentile

0.73 6.57 11.27 0

6.23 44.45 12.08 0.01

16.95 49.53 13.73 0.79

0.44 0.32 0.56

0.49 0.61 0.91

0.56 0.75 1.30

Mean (standard deviation) 6.34 44.43 12.17 0.05

(2.44) (1.86) (0.47) (0.10)

0.49 (0.01) 0.59 (0.07) 0.91 (0.11)

Notes: a. The nth percentile of a distribution is the number below which n per cent of observations lie. The 0th percentile is the minimum. The 100th percentile is the maximum. b. The unemployment rate is multiplied by 100 and expressed in percentage terms. c. If these proportions are multiplied by 100, they will be expressed in percentage terms.

94 Incentives for Regional Development Table 5.3 Profile of Ohio’s non-zone census block groups, 1997 Characteristic Unemployment ratea (%) Age (in years) Grades completed Proportion African Americanb Proportion maleb Proportion marriedb Number of children per family

0th percentile

50th percentile

100th percentile

0 19.05 8.07 0

3.57 44.38 12.54 0

59.16 69.13 16.24 1.00

0.22 0 0

0.48 0.61 0.83

1.00 1.00 2.85

Mean (standard deviation) 4.66 44.63 12.72 0.08

(4.40) (5.27) (1.04) (0.20)

0.48 (0.05) 0.57 (0.15) 0.84 (0.29)

Notes: a. The unemployment rate is multiplied by 100 and expressed in percentage terms. b. If these proportions are multiplied by 100, they will be expressed in percentage terms.

Table 5.2 shows that the 280 enterprise zones in Ohio had substantial variability in their unemployment rate,4 had a middle-aged population,5 with high school (about 12 grades of school) completed on average, the distribution of women and men being equal and majority of them married. Table 5.3 shows the profile of characteristics for non-zone areas in Ohio. Table 5.3 is based on the 3,331 census block groups of the state that were not enterprise zones in 1997. It shows that some of the non-zone areas of the state had higher unemployment rates than the areas designated as zones (observe that the maximum unemployment rate for non-zone groups is 59 per cent, much higher than that for zones, 17 per cent). However, it may be noted that the average unemployment rate for the non-zone areas was lower than it was for the zone areas, indicating higher levels of distress in zones, and thereby supporting the theoretical model of Chapter 2. It may also be noted that, on average, the populations in the non-zone areas in Ohio were of the same age but slightly more educated than the population in the zone areas.6 This, if true, and with a sharper contrast than is shown by the data here, is consistent with the notion of EZs as blighted areas, as the model in Chapter 2 points out.7 From a comparison of Tables 5.2 and 5.3, note that the zones are also a little different from the non-zones in that they have slightly larger families (seen in the average number of children per family) than their non-zone counterparts. Imputation of reservation wages The imputation of the reservation wage to jobs created in Ohio’s EZs was undertaken using the estimates in Table 5.1 and the zone characteristics summarized in Table 5.2. In addition to the characteristics reported for Ohio’s enterprise zones in Table 5.2, data on the duration of unemployment,

Evidence from Ohio’s Enterprise Zones 95 Table 5.4 Distribution of reservation wages imputed for Ohio’s enterprise zones nth percentile 0th percentile (minimum) 50th percentile (median) Mean (standard deviation) 100th percentile (maximum)

Reservation wage ($ per hour) 2.60 4.09 4.88 (1.48) 16.98

past wages and work experience (see Table 5.1) were substituted from the PSID for the imputation. The minimum weekly unemployment benefit ($12.15), maximum weekly unemployment benefit ($245.44),8 and the waiting period for unemployment benefit (one week) eligibility that were applied to the estimates in Table 5.1 for the imputation, were for Ohio. The imputed reservation wage for an average Ohio zone turned out to be about $4.88 an hour in constant dollars, with 1982–4 100 (or about $10,150 annually, assuming 40 hours a week and 52 weeks a year), quite plausible. The distribution of imputed reservation wages is shown in Table 5.4, with a minimum of $2.60 and a maximum of $16.98 an hour (or about $35,000 annually, assuming 52 regular work weeks), depending on zone characteristics (all in constant dollars, with 1982–4 100). When I apply the reservation wage estimates in Table 5.1 to characteristics of the non-EZ block groups of the state, the mean reservation wage turns out to be $5.67 per hour, higher than that in the zones, as we would expect. The next step was to estimate net benefits for the employment created in Ohio’s zones. Because the reservation wages imputed from the PSID were hourly, they were converted to an annual amount (the Ohio Department of Development reports annual earnings from employment)9 assuming that persons work full-time (40 hours a week, 52 weeks a year). The earnings reported for jobs created in Ohio’s zones were made net of the imputed reservation wages to arrive at an estimate of net benefit for every job. 5.5.2

Measures of programme costs

It may be useful to note that actual property tax abatements offered to firms in the zones were used as programme costs in the benefit–cost analysis of Ohio’s programme in this chapter. Infrastructure costs are also real resource costs. Since these data were not routinely collected at the state level in Ohio in early 1998, a survey was sent to the EZ administrators of the 44 distressed Ohio EZs. The objective of the survey was to obtain some idea of the costs of providing infrastructure services to firms. It was assumed that the cost of providing public services in distressed areas define the upper limit of the cost of making infrastructure improvements for a firm that locates in any zone. One may expect the non-distressed areas (those which are designated as (limited authority) zones in Ohio) to have the basic infrastructure such as

96 Incentives for Regional Development

state highways, county roads, sewer and water lines in place. Even if they do not, it was assumed that the cost of constructing the infrastructure in these non-distressed areas would be at the most, the same, if not higher than, that in the distressed areas. Based on an analysis of 39 of the 44 distressed zones that responded to the survey, the average cost of providing basic infrastructure services to a firm that locates in a zone turned out to be $24,200. There were several zones (17 out of 39) in which the cost to the local government of providing infrastructure to firms was zero because it was either covered by federal grants or was paid for by the firm.10 It is relevant at this point to recall the hypothetical illustration in Chapter 3 (in Table 3.3) that compared property tax revenues to the cost of providing public services. Based on data from the 1,974 contracts in Ohio’s EZs formalized over the period 1984–95, I find the average tax abatement (forgone) is about $107,963. Even when the cost of providing public services (from the above-mentioned survey) is included, this abatement turns out to be $132,163. For the 1,974 contracts, the average local (personal, and real property and local corporation) tax revenue paid by a firm is roughly $3.5 million. Then the property tax revenue (even with the abatement) is, on average, several times higher than the average taxes forgone. So, according to the argument we developed in Chapter 3, it must be the case that in Ohio’s zones, the cost of the programme might well have been less than the abatement dollars provided to firms. And that the new industry generated a fiscal surplus which could be used to finance the incentives at least in part. In addition to the costs of providing infrastructure to firms, respondents were also asked about any administrative costs of negotiating the EZ agreement with the firm. Most (92 per cent of the respondents) reported that there was no cost involved in negotiating the contract with the firm. Based on those who reported administrative costs, the average cost of negotiating an agreement with a firm was around $1,000.11 The costs of infrastructure and administrative costs for providing services to EZ firms were taken into account along with property tax abatement costs and the cost of any other local incentives to estimate programme costs in the benefit–cost analysis.12 It may be noted that these measures of costs ignore commuting effects, which could be substantial, depending upon the size of the zone. The large number of zones in Ohio suggests relatively small zones. Further, it is assumed that jobs created by zone firms are held by zone residents. The empirical evidence is that, on average, about 50 per cent of the jobs created in the EZ are held by zone residents. As reported in Chapter 2, the US Department of HUD (1986) found that 70 per cent each of jobs created in the Bridgeport EZ, CT, and the Chicago EZ, IL, 46 per cent in Macon EZ, MO, 19 per cent in Michigan City, IN, and 30 per cent in the Tampa zone (FL) were held by zone residents. In the Louisville (Kentucky) zone, it was found

Evidence from Ohio’s Enterprise Zones 97

that 31 per cent of the jobs created were held by persons who were either lower income or zone residents. Based on a survey of local enterprise zone coordinators conducted by the US Department of HUD, Erickson and Friedman (1989) found that the mean share of jobs held by zone residents was over 61 per cent with a median of over 68 per cent. Immergluck (1997), based on data from the Chicago EZ, indicated that the barriers between EZ residents and jobs are dependent on some factors. He found that local employment was much higher in Latino parts of the zone and in African American neighbourhoods where there were more public sector jobs, very small firms and few manufacturers. The comparatively small size of Ohio’s zones, along with the assumption that a majority of jobs are held by zone residents, implied that substantial commuting to get to work, does not take place. In the event that commuting exists, they could reduce the benefits imputed here. The net benefits were compared to programme costs in order to estimate the benefit–cost ratio. It may be noted that the benefit–cost ratios computed in this manner refer only to local government investment, which helped to determine in which zones local government investment was worth the net benefits from employment. Given the financial autonomy of local governments in the United States, they are the appropriate unit of analysis. Moreover, this is a programme that depends to a considerable extent on the local government’s financial resources, as is clear from the description of Ohio’s EZ programme. Even if the benefits that accrue from the programme were not essentially local in the long run (for example, multiplier effects that could have spillover effects on the region’s employment), the local government incurs programme costs. Therefore it was appropriate to perform benefit– cost analysis at the level of the local government that spends its resources on the programme.

5.6

Benefit–cost analysis

Given the finding in Chapter 4 regarding the effectiveness of Ohio’s tax incentives, B–C analysis of the programme is performed and reported in three scenarios reflecting various assumptions about employment created in the zones. The assumptions which form the basis for the three scenarios are succinctly: 1. Scenario one: Net benefits estimated from all employment reported as being created and retained by the firms. 2. Scenario two: Net benefits estimated from employment reported as being created (excludes jobs retained from scenario one). The use of employment created as a measure of programme impact is supported by evidence of the impact of tax incentives on (un)employment, presented in Chapter 4.

98 Incentives for Regional Development

3. Scenario three: Net benefits estimated from a proportion of the total employment (those created and retained) attributable to the tax incentive making two assumptions of elasticity. The assumptions were made to arrive at the various scenarios for two reasons: (i) The sensitivity or robustness of the B–C results to the various assumptions presents a range of multiple alternatives, according to assumptions or value judgment. (ii) Retained employment cannot be considered as being held by the unemployed, because they are, by definition, already held by someone.13 Alternatively, it is important to examine the benefits and costs when the jobs are held by the formerly unemployed. So when we take into account only jobs that are newly created, can the assumption of the jobs being held by unemployed be plausible.14 The assumption in scenario two reflects this. In addition, also note that retained employment is not a very reliable measure of programme impact. A pre-existing firm negotiates a tax incentive with officials and claims that its jobs would have otherwise moved. However, it is quite possible that the firm would have stayed where it was even in the absence of the programme. Moreover, data from the Ohio Department of Development on retained jobs (and their earnings) are not very reliable, especially for the earlier years (from 1984 to the early 1990s). So, in scenario two, I am able to overcome the data and other substantive limitations with retained employment. The net benefits, costs per job and benefit–cost ratios (all for local government investment) are computed at two levels: at the level of the enterprise zone and at the level of the firm within the zones, in all the scenarios. The justification, from a policy perspective, for computing B–C ratios for firms and zones is that they can indicate whether or not particular firms or particular zones should be targeted.

5.6.1

Scenario one

In this scenario, all of the jobs that were reported as being created and retained were taken into account in the computation of the benefit–cost ratios. The results are reported for 531 firms that did not relocate from within or outside the state in the 143 zones and those that received some tax incentives from their local governments.15 Table 5.5 shows the distribution of total employment (created/retained by firms in the zones), net benefits from jobs, costs and benefit–cost ratios per job at the level of the contracts that are negotiated with 531 firms in Ohio’s enterprise zones. Up to 1995, the 531 firms in the 143 zones had created and retained 104,840 jobs. Table 5.5 shows at the level of these firms a disaggregated distribution of employment, earnings, net benefits, costs and B–C ratio per job.

Table 5.5 Distribution of average net benefits and benefit–cost ratios for the (531) firms in Ohio’s enterprise zones, scenario 1

Total employmenta Annual deflated earnings per job ($) (1982–4 100) Net benefit per job ($)b Costs per job ($)c

B–C ratiod Unemployment rate (%)

Mean (min, max) for firms: 0th–25th percentile

Mean (min, max) for firms: 25th–50th percentile

Mean (min, max) for firms: 50th–75th percentile

Mean (min, max) for firms: 75th–100th percentile

Unweighted mean for all firms (std dev.) e

Weighted mean for all firms (weighted std dev.) f

12.07 (1, 23) 12,553.57 (10,148.45, 14,497.31) 11.69 (8,897.97, 3,508.82) 180.21 (21.99, 330.87) 2.5 (59.53, 2.71) 7.46 (3.05, 11.66)

40.32 (24, 63) 16,546.02 (14,497.71, 18,714.21) 5,819.81 (3,564.41, 8,161.69) 524.42 (336.14, 762.76) 5.85 (2.72, 9.36) 7.03 (2.05, 13.49)

95.51 (64, 152) 21,913.32 (18,764.75, 24,985.60) 11,185.96 (8,168.76, 14,369.24) 1,126.24 (764.53, 1,791.36) 15.22 (9.48, 24.14) 6.69 (2.36, 13.49)

646.91 (153, 11,000) 32,078.04 (25,100.70, 49,636.24) 20,624.94 (14,372.16, 36,913.09) 8,962.99 (1,795.12, 1,273,982.19) 80.25 (24.41, 1,448.75) 6.80 (2.85, 13.49)

197.44 (789.22) 20,821.28 (8,012.17)

197.44 (382.91) 20,766.91 (2,449.43)

9,458.97 (8,388.51)

9,407.17 (2,996.77)

5,085.54 (56,142.42)

2,691.40 (5,007.54)

24.73 (74.04)

24.73 (37.12) 6.99 (2.25)

6.34 (2.44)

Notes: The notes below are applicable to Tables 5.5–5.12. Min and max respectively refer to minimum and maximum, and std dev. refers to standard deviation in all the tables. Annual deflated earnings in all the tables refer to 1982–4 100 as the base year. a. b. c. d.

99

Total employment in this scenario includes jobs created and retained. Net benefit is defined as earnings per job – annual reservation wage. Costs per job include property tax abatements, other local incentives, and costs of infrastructure. BC ratio is defined as the ratio of net benefit to costs per job. Note that it is not average net benefit per job (row 3) divided by average cost per job (row 4). If I calculate BC ratios based on average net benefit and cost per job (rows 3 and 4 respectively), I would not reveal valuable information on the individual firm’s BC ratios which I have calculated as the ratio of individual firms’ net benefits to costs per job. The minimum and maximum values for the distribution of BC ratios should make clear the average values of the BC distribution, when we compare them to the net benefit and cost distribution. e. Unweighted means are the raw averages of the relevant variable for all the firms. f. Weighted means are weighted averages in which the weights are the number of firms in each category (firms in 0th–25th percentile, those in the 25th–50th, 50th–75th, and 75th–100th percentiles) of the relevant distributions. For the weighted means, the weights are multiplied by the means for firms in each category and sum of the weights and the means are divided by the total number of firms. A similar procedure is used to calculate weighted standard deviations for the various distributions.

100 Incentives for Regional Development

From Table 5.5, it is clear that there has been considerable variation in the employment created by the various firms in the state’s zones. The high-cost zones were those in which few jobs were created. With the variation, it may be noted that the benefits are greater than costs in the average contract negotiated in Ohio’s zones. The net benefits per job are approximately equal to earnings less the mean annual reservation wage (about $10,000 per job). Also observe the large variation in the costs incurred for every job created, ranging from less than $25 a job to about $1.27 million per job in Table 5.5. The highest costs (abatements and other incentives) were incurred for those firms that made the largest investments in personal and real property. Note the property tax abatement is an incentive to capital. Given that first-order effects dominate, as in the model (Chapter 2), and these capital-intensive firms did not create much employment, the costs per job are high. There was also considerable variation in the benefit–cost ratios (for local government investment) across zones. The benefit–cost ratio was less than one for about 10 per cent of the firms, indicating that costs were greater than benefits for these firms. The low benefit areas were those that awarded large abatements because of firms’ investments in capital (personal and tangible property), resulting in high costs to create employment. However, for firms in majority of the distribution, the (unweighted) B–C ratio was well above one, indicating that net benefits per job were substantially higher than costs per job in the case of these firms. In the data, at the aggregate level, there was little relationship between net benefits from employment and the rate of unemployment. In fact, there was a negative, but insignificant correlation between the net benefits from jobs created in the zones and the unemployment rate of the zone in which these firms are located, casting doubt on Bartik’s hypothesis about the relationship between net benefits from jobs and the unemployment rate. Table 5.6 shows the distribution of the employment, net benefits, costs and benefit–cost ratios by zone in this scenario. It shows that in 50 per cent of the zones the cost of creating a job was less than $1,000. In the average zone, the net benefit was above $10,000 per job, substantially higher than the costs per job. The difference between the earnings and the net benefit per job is equal to an amount in the range of the annual reservation wage reported earlier (about $10,000 per job). On average, the unweighted B–C ratio in this scenario at the zone level is about 27 (being 25 at the firm level, Table 5.5). The weighted B–C ratio (25.5) is also consistent with this, indicating that net benefits were 25 times greater than the costs of creating employment. I found that the correlation between the zone unemployment rate and net benefit per job was positive and stronger at the zone level, than that found at the firm level, being 0.09, with many lower B–C ratios found in the low-unemployment zones, and more consistent with Bartik’s hypothesis.

Table 5.6 Distribution of net benefits and benefit–cost ratios in Ohio’s (143) enterprise zones, scenario 1

Total employmenta Annual deflated earnings per job ($) Net benefit per job ($)b Costs per job ($)c B–C ratiod Unemployment rate (%)

Mean (min, max) for zones: 0th–25th percentile

Mean (min, max) for zones: 25th– 50th percentile

Mean (min, max) for zones: 50th– 75th percentile

Mean (min, max) for zones: 75th– 100th percentile

Unweighted mean for all zones (std dev.)e

Weighted mean for all zones (weighted std dev.)f

36.97 (1, 82)

153.75 (83, 231) 17,611.45 (15,310.28, 19,974.22) 7,048.75 (5,282.18, 8,950.70) 408.01 (268.69, 573.82) 8.11 (4.69, 11.55) 5.85 (2.36, 11.17)

335.83 (236, 533) 22,152.01 (20,020.53, 24,177.00) 12,002.45 (9,121.22, 14,443.00) 897.30 (581.38, 1,299.17) 22.18 (11.93, 35.66) 6.44 (2.85, 13.49)

2,453.83 (549, 26,360) 31,549.11 (24,534.56, 45,281.78) 20,454.63 (14,575.34, 36,121.29) 14,154.65 (1,383.08, 156,491.30) 72.34 (36.15, 224.84) 6.63 (2.88, 11.66)

733.15 (2,369.38) 21,089.35 (7,725.08)

733.14 (1,113.75) 21,089.35 (2,684.22)

10,666.51 (7,544.38)

10,487.25 (2,563.70)

4,927.96 (19,853.35) 26.95 (38.24)

3,860.62 (7,148.55) 25.55 (13.03)

6.34 (2.44)

6.51 (2.13)

13,074.35 (10,148.45, 15,154.95) 2,739.62 (2,120.76, 5,262.69) 179.80 (85.09, 267.59) 1.79 (3.17, 4.61) 7.13 (2.05, 10.71)

Notes: The notes below are applicable to Tables 5.5–5.12. Min and max respectively refer to minimum and maximum, and std dev. refers to standard deviation in all the tables. Annual deflated earnings in all the tables refer to 1982–4 100 as the base year. a. b. c. d.

101

Total employment in this scenario includes jobs created and retained. Net benefit is defined as earnings per job – annual reservation wage. Costs per job include property tax abatements, other local incentives, and costs of infrastructure. BC ratio is defined as the ratio of net benefit to costs per job. Note that it is not average net benefit per job (row 3) divided by average cost per job (row 4). If I calculate BC ratios based on average net benefit and cost per job (rows 3 and 4 respectively), I would not reveal valuable information on the individual firm’s BC ratios which I have calculated as the ratio of individual firms’ net benefits to costs per job. The minimum and maximum values for the distribution of BC ratios should make clear the average values of the BC distribution, when we compare them to the net benefit and cost distribution. e. Unweighted means are the raw averages of the relevant variable for all the firms. f. Weighted means are weighted averages in which the weights are the number of firms in each category (firms in 0th–25th percentile, those in the 25th–50th, 50th–75th, and 75th–100th percentiles) of the relevant distributions. For the weighted means, the weights are multiplied by the means for firms in each category and sum of the weights and the means are divided by the total number of firms. A similar procedure is used to calculate weighted standard deviations for the various distributions.

102 Incentives for Regional Development

5.6.2

Scenario two

In this scenario, the B–C ratios reported at the firm level and the zone level take into account only that proportion of employment that is newly created. The results in this scenario are based on jobs created by 575 firms in 148 zones.16 As explained earlier, retained jobs were excluded from the B–C analysis. Table 5.7 shows the distribution of B–C ratios at the firm level when calculated based on new jobs. As of 1997, the total employment created by firms in the zones was 33,896. The remaining 70,944 of the total of 104,840 jobs considered in scenario one were retained. The net benefit in this scenario was about $10,000 on average, and about $10,000 lower than the earnings per job, which was approximately equal to the annual mean reservation wage. The B–C ratios were less than one till roughly up to the 25th percentile, indicating that for more than 20 per cent of the firms, the costs of creating employment were greater than the benefits from local government investment. The firms with B–C ratios less than one were those that created few jobs (usually less than ten jobs) or were offered large abatements and were randomly distributed across low-unemployment and high-unemployment zones. I found that most of the firms with B–C ratios in higher percentiles of the distribution were located in the zones with high unemployment rates (had unemployment rates greater than 125 per cent of the state’s average for 1990). The correlation, however, between net benefits and unemployment rate of the zone (in which firm had located) was negative, though not statistically significant, being 0.1, similar to that found at the firm level, in scenario one. The same pattern of B–C ratios essentially repeated itself when zone-level performance was studied in scenario two, the results from which are reported in Table 5.8. Table 5.8 shows that the mean earnings per job was reasonable (around $20,000), concurring with the expectation of the Ohio Department of Development regarding the nature of jobs created by the firms typically locating in the zones. The high-cost areas were those in which firms made large investments in personal and real property and, as a result, large abatements were provided. Zones with B–C ratios less than one were those with low net benefits because of low earnings from employment. Thus, the lowest net benefits from employment were in zones in which the earnings per job were below average or when reservation wages were high because of the area’s low unemployment rate (confirmed by the negative, but statistically insignificant, sign on the unemployment rate, in Table 5.1). Lower net benefits were found in some high unemployment zones, with correlation between net benefits and unemployment rate at the zone level also being 0.1. 5.6.3

Scenario three17

Tables 5.9–5.12 show the results from the net benefits from employment and B–C ratios for respectively 198 (in 62 zones) and 91 firms (in 32 zones).18 The

Table 5.7 Distribution of net benefits and benefit–cost ratios for the (575) firms in Ohio’s enterprise zones, scenario 2

Total employmenta Annual deflated earnings per job ($) Net benefit per job ($)b

Costs per job ($)c BC ratiod Unemployment rate (%)

Mean (min, max) for firms: 0th–25th percentile

Mean (min, max) for firms: 25th–50th percentile

Mean (min, max) for firms: 50th–75th percentile

6.37 (1, 11) 12,409.71 (10,076.12, 14,554.11) 121.06 (10,138.02, 3,670.92) 462.65 (80.04, 754.12) 0.76 (46.43, 1.40) 7.57 (2.05, 11.66)

17.56 (12, 27) 16,472.92 (14,634.91, 18,596.79) 5,921.91 (3,692.26, 7,882.93) 1,086.88 (759.82, 1,501.95) 2.94 (1.49, 4.71) 6.76 (2.76, 11.66)

170.79 (69, 1023) 43.88 (28, 68) 21,491.57 31,477.84 (24,816.45, (18,600.32, 24,635.04) 48,436.18) 10,713.90 20,381.02 (7,921.71, (14,127.23, 14,097.20) 37,980.23) 2,173.33 14,062.94 (1,512.89, 3039.92) (3,041.45, 1,273,982.19f) 25.58 (11.59, 106.88) 7.82 (4.77, 11.47) 6.59 (2.36, 13.49) 6.76 (2.76, 11.66)

Mean (min, max) for firms: 75th–100th percentile

Unweighted mean for all firms (std dev.)e

Weighted mean for firms (weighted std dev.)f

58.95 (92.61) 20,534.26 (7,881.70)

58.95 (37.23) 20,534.26 (2,513.02)

9,331.02 (8,161.65)

9,281.10 (2,856.89)

6,623.86 (54,776.70) 9.05 (14.11) 6.34 (2.44)

4,415.92 (6,639.01) 8.88 (6.20) 6.87 (2.19)

Notes: The notes below are applicable to Tables 5.5–5.12. Min and max respectively refer to minimum and maximum, and std dev. refers to standard deviation in all the tables. Annual deflated earnings in all the tables refer to 1982–4 100 as the base year. a. b. c. d.

103

Total employment in this scenario includes jobs created and retained. Net benefit is defined as earnings per job – annual reservation wage. Costs per job include property tax abatements, other local incentives, and costs of infrastructure. BC ratio is defined as the ratio of net benefit to costs per job. Note that it is not average net benefit per job (row 3) divided by average cost per job (row 4). If I calculate BC ratios based on average net benefit and cost per job (rows 3 and 4 respectively), I would not reveal valuable information on the individual firm’s BC ratios which I have calculated as the ratio of individual firms’ net benefits to costs per job. The minimum and maximum values for the distribution of BC ratios should make clear the average values of the BC distribution, when we compare them to the net benefit and cost distribution. e. Unweighted means are the raw averages of the relevant variable for all the firms. f. Weighted means are weighted averages in which the weights are the number of firms in each category (firms in 0th–25th percentile, those in the 25th–50th, 50th–75th, and 75th–100th percentiles) of the relevant distributions. For the weighted means, the weights are multiplied by the means for firms in each category and sum of the weights and the means are divided by the total number of firms. A similar procedure is used to calculate weighted standard deviations for the various distributions.

Table 5.8 Distribution of net benefits and benefit–cost ratios for (148) Ohio’s enterprise zones, scenario 2

Net benefit per job ($)b

Costs per job ($)c B–C ratiod Unemployment rate (%)

Mean (min, max) for zones: 25th–50th percentile

Mean (min, max) for zones: 50th–75th percentile

Mean (min, max) for zones: 75th–100th percentile

Unweighted mean for all zones (std dev.)e

Weighted mean for zones (weighted std dev.)f

19.87 (1, 42) 13,337.93 (10,148.45, 15,311.77) 2,952.61 (755.87, 5,036.53) 440.94 (134.61, 662.61) 1.01 (0.59, 3.17) 7.30 (2.05, 11.17)

74.69 (43, 106) 17,706.85 (15,438.60, 19,707.11) 7,118.41 (5,194.37, 8,950.70) 915.05 (668.12, 1,233.86) 4.96 (3.20, 6.69) 5.90 (2.67, 10.30)

191.40 (107, 288) 21,468.59 (19,757.39, 23,618.26) 11,224.33 (8,964.85, 13,440.43) 1,723.47 (1,265.99, 2,428.51) 9.95 (6.77, 13.76) 5.99 (2.36, 11.66)

631.62 (290, 1,431) 29,282.95 (23,861.97, 47,345.77) 18,495.26 (13,702.25, 37,980.67) 19,970.69 (2,464.24, 182,954.77) 24.77 (13.90, 67.03) 6.65 (3.05, 13.49)

229.03 (280.56) 20,449.08 (6,574.58)

229.02 (93.06) 20,449.08 (2,326.66)

10,079.31 (6,643.42)

9,957.24 (2,172.44)

5,762.54 (20,148.18)

5,762.54 (9,452.37)

10.17 (10.96) 6.34 (2.44)

10.17 (4.16) 6.46 (2.11)

Notes: The notes below are applicable to Tables 5.5–5.12. Min and max respectively refer to minimum and maximum, and std dev. refers to standard deviation in all the tables. Annual deflated earnings in all the tables refer to 1982–4 100 as the base year. a. b. c. d.

Total employment in this scenario includes jobs created and retained. Net benefit is defined as earnings per job – annual reservation wage. Costs per job include property tax abatements, other local incentives, and costs of infrastructure. BC ratio is defined as the ratio of net benefit to costs per job. Note that it is not average net benefit per job (row 3) divided by average cost per job (row 4). If I calculate BC ratios based on average net benefit and cost per job (rows 3 and 4 respectively), I would not reveal valuable information on the individual firm’s BC ratios which I have calculated as the ratio of individual firms’ net benefits to costs per job. The minimum and maximum values for the distribution of BC ratios should make clear the average values of the BC distribution, when we compare them to the net benefit and cost distribution. e. Unweighted means are the raw averages of the relevant variable for all the firms. f. Weighted means are weighted averages in which the weights are the number of firms in each category (firms in 0th–25th percentile, those in the 25th–50th, 50th–75th, and 75th–100th percentiles) of the relevant distributions. For the weighted means, the weights are multiplied by the means for firms in each category and sum of the weights and the means are divided by the total number of firms. A similar procedure is used to calculate weighted standard deviations for the various distributions.

104

Total employmenta Annual deflated earnings per job ($)

Mean (min, max) for zones: 0th–25th percentile

Table 5.9 Distribution of net benefits and benefit–cost ratios for (198) firms in Ohio’s enterprise zones, scenario 3 (assumed elasticity 0.3)

Total employmenta Annual deflated earnings per job ($) Net benefit per job ($)b Costs per job ($)c BC ratiod Unemployment rate (%)

Mean (min, max) for firms: 0th–25th percentile

Mean (min, max) for firms: 25th–50th percentile

Mean (min, max) for firms: 50th–75th percentile

Mean (min, max) for firms: 75th–100th percentile

Unweighted mean for all firms (std dev.)e

Weighted mean for firms (weighted std dev.)f

3.53 (0.91, 6) 13,636.09 (10,177.08, 17,207.95) 2,075.81 (15,703.50, 6,272.03) 614.57 (16.22, 1,051.17) 0.18 (10.28, 1.73) 7.31 (3.05, 11.61)

9.29 (6.16, 12.51) 21,584.98 (17,274.79, 25,880.46) 10,959.53 (6,755.28, 15,741.49) 1,642.51 (1,073.13, 2,255.08) 3.49 (1.73, 5.60) 7.23 (2.88, 11.98)

18.33 (12.61, 25) 30,591.50 (25,919.16, 35,455.71) 20,199.52 (15,936.28, 24,809.38) 3,166.89 (2,264.07, 4,326.76) 9.41 (5.88, 15.21) 7.95 (3.70, 13.49)

148.18 (26, 1200) 42,440.13 (35,862.24, 49,737.04) 31,510.01 (25,017.49, 40,862.68) 16,051.59 (4,435.09, 1,273,982.19) 96.61 (15.50, 2,159.17) 6.61 (3.44, 13.49)

44.40 (124.41) 27,073.02 (11,198.61)

44.40 (56.47) 27,073.02 (2,898.34)

16,135.22 (11,518.54)

16,135.22 (3,538.53)

11,692.13 (91,494.86)

5,284.57 (7,456.57)

27.65 (159.40) 6.34 (2.44)

27.65 (79.54) 7.27 (2.18)

Notes: The notes below are applicable to Tables 5.5–5.12. Min and max respectively refer to minimum and maximum, and std dev. refers to standard deviation in all the tables. Annual deflated earnings in all the tables refer to 1982–4 100 as the base year. a. b. c. d.

105

Total employment in this scenario includes jobs created and retained. Net benefit is defined as earnings per job – annual reservation wage. Costs per job include property tax abatements, other local incentives, and costs of infrastructure. BC ratio is defined as the ratio of net benefit to costs per job. Note that it is not average net benefit per job (row 3) divided by average cost per job (row 4). If I calculate BC ratios based on average net benefit and cost per job (rows 3 and 4 respectively), I would not reveal valuable information on the individual firm’s BC ratios which I have calculated as the ratio of individual firms’ net benefits to costs per job. The minimum and maximum values for the distribution of BC ratios should make clear the average values of the BC distribution, when we compare them to the net benefit and cost distribution. e. Unweighted means are the raw averages of the relevant variable for all the firms. f. Weighted means are weighted averages in which the weights are the number of firms in each category (firms in 0th–25th percentile, those in the 25th–50th, 50th–75th, and 75th–100th percentiles) of the relevant distributions. For the weighted means, the weights are multiplied by the means for firms in each category and sum of the weights and the means are divided by the total number of firms. A similar procedure is used to calculate weighted standard deviations for the various distributions.

Total employmenta Annual deflated earnings per job ($) Net benefit per job ($)b Costs per job ($)c

BC ratiod Unemployment rate (%)

Mean (min, max) for firms: 0th–25th percentile

Mean (min, max) for firms: 5th–50th percentile

Mean (min, max) for firms: 50th–75th percentile

Mean (Min, max) for firms: 75th–100th percentile

Unweighted mean for all firms (std dev.)e

Weighted mean for firms (weighted std dev.)f

1.96 (0.37, 3.73) 13,962.83 (11,013.22, 16,497.22) 3,586.48 (125.77, 6,272.03) 750.35 (40.25, 1,028.13)

5.87 (4, 8.53) 20,016.23 (17,134.53, 24,146.00) 9,514.91 (6,899.56, 13,466.98) 2,254.77 (1,038.42, 3,587.38) 2.37 (1.40, 3.62) 7.20 (2.88, 11.61)

14.14 (8.89, 21) 30,263.82 (24,709.70, 35,165.22) 19,601.75 (13,585.95, 26,671.05) 5,516.82 (3,609.37, 9,016.16) 5.56 (3.69, 7.98) 7.44 (4.26, 11.98)

72.43 (21.5, 387) 41,796.84 (36,306.58, 49,292.28) 32,005.81 (26,974.90, 36,384.07) 27,686.49 (9,548.63, 152,773.40) 42.31 (8.62, 422.78) 6.54 (2.05, 13.49)

23.15 (48.62) 26,617.57 (11,100.76)

23.23 (21.40) 26,365.62 (2,636.23)

16,135.39 (11,002.71)

16,348.94 (2,731.12)

10,517.92 (23,761.22)

8,937.30 (8,125.57)

12.80 (45.36) 6.34 (2.44)

12.80 (22.11) 6.61 (2.02)

0.66 (0.03, 1.38) 7.08 (3.05, 9.35)

Notes: The notes below are applicable to Tables 5.5–5.12. Min and max respectively refer to minimum and maximum, and std dev. refers to standard deviation in all the tables. Annual deflated earnings in all the tables refer to 1982–4 100 as the base year. a. b. c. d.

Total employment in this scenario includes jobs created and retained. Net benefit is defined as earnings per job – annual reservation wage. Costs per job include property tax abatements, other local incentives, and costs of infrastructure. BC ratio is defined as the ratio of net benefit to costs per job. Note that it is not average net benefit per job (row 3) divided by average cost per job (row 4). If I calculate BC ratios based on average net benefit and cost per job (rows 3 and 4 respectively), I would not reveal valuable information on the individual firm’s BC ratios which I have calculated as the ratio of individual firms’ net benefits to costs per job. The minimum and maximum values for the distribution of BC ratios should make clear the average values of the BC distribution, when we compare them to the net benefit and cost distribution. e. Unweighted means are the raw averages of the relevant variable for all the firms. f. Weighted means are weighted averages in which the weights are the number of firms in each category (firms in 0th–25th percentile, those in the 25th–50th, 50th–75th, and 75th–100th percentiles) of the relevant distributions. For the weighted means, the weights are multiplied by the means for firms in each category and sum of the weights and the means are divided by the total number of firms. A similar procedure is used to calculate weighted standard deviations for the various distributions.

106

Table 5.10 Distribution of net benefits and benefit–cost ratios for (91) firms in Ohio’s enterprise zones, scenario 3 (assumed elasticity 0.1)

Evidence from Ohio’s Enterprise Zones 107

results in Table 5.9 are based on the assumption that the elasticity of employment with respect to taxes is 0.3 (the upper range for the elasticity, in Bartik’s summary of econometric studies) and they are reported at the firm level. Table 5.10 shows the results for the firm level assuming that the elasticity of employment with respect to taxes is 0.1 (the lower range in Bartik’s summary). Thus, the employment and net benefits that are reported in these tables refer to the portion of annual employment created and retained that is actually attributable to the tax incentive making two different assumptions of elasticity (0.3 and 0.1).19 Tables 5.11 and 5.12 report the results of benefit–cost analyses at the zone level for these two assumptions of elasticity. With an assumed elasticity of employment with respect to taxes equal to 0.3, of the 19,990 jobs that were reported by the firms to be created and retained, about 8,791 (44 per cent) are actually attributable to the tax incentive. With an assumed elasticity of 0.1, only about 1,894 (a smaller proportion, about 31 per cent) of the 5,974 jobs that were created are actually attributable to the tax incentive. This is consistent with what we expect. For firms in the lower end of the employment distribution, the number of jobs attributable to the tax incentive (dy) was always less when compared to the actual number of jobs created and retained.20 For firms in the upper part of the employment distribution, it was found that a majority of the employment they reported as being created and/or retained was due to the tax incentive, for either elasticity. Tables 5.9 and 5.10 provide a range for the benefit–cost ratio for local government investment at the firm level and Tables 5.11 and 5.12 provide a range for the ratio at the zone level, if we assumed elasticity in the range 0.3 to 0.1. Tables 5.9–5.10 show that, on average, the (weighted as well as unweighted) B–C ratio at the firm level (zone level) can be expected to be in the range 28 (14) (with an elasticity of 0.3) to 13 (10) (with an elasticity of 0.1). As we expect, with the assumption of elasticity in the lower range, the benefits from creating employment are lower when compared to costs. However, with either assumption of elasticity, it should be noted that for firms in the upper quintile of the B–C distribution, the net benefits from jobs are substantially greater than costs. A similar pattern repeated for zone-level performance (Tables 5.11–5.12), with the two assumed elasticities. At the zone level (unweighted) average B–C ratios were slightly lower than at the firm level. The correlation between net benefits from employment and zone unemployment rates with an assumed elasticity of 0.3 was negative, but small and insignificant, being about 0.03. With an elasticity assumption of 0.1, this correlation was positive and higher, being 0.12, confirming Bartik’s hypothesis regarding the positive relationship between net benefits and the unemployment rate. The zones in the top portion of the B–C ratio distribution were those with higher net benefits relative to lower costs of creating employment. The average unemployment

Total employmenta Annual deflated earnings per job ($) Net benefit per job ($)b Costs per job ($)c

BC ratiod Unemployment rate (%)

Mean (min, max) for zones: 0th–25th percentile

Mean (min, max) for zones: 25th–50th percentile

Mean (min, max) for zones: 50th–75th percentile

Mean (min, max) for zones: 75th–100th percentile

Unweighted mean for all zones (std dev.)e

Weighted mean for zones (weighted std dev.)f

7.56 (1.41,12.61) 15,508.67 (10,894.26, 19,022.55) 3,988.64 (4,990.01, 8,491.40) 672.67 (51.17, 1,120.73) 1.11 (4.15,2.98) 7.28 (3.05, 10.20)

26.77 (13.15,41.29) 22,630.06 (19,041.19, 27,154.26) 12,565.93 (8,506.76, 16,857.30) 1,354.99 (1,126.21, 1,735.04) 5.44 (3.00, 7.76) 7.74 (4.26,11.00)

95.81 (43.74,157.46) 29,532.43 (27,379.72, 32,714.84) 19,629.91 (17,688.55, 23,154.76) 2,535.31 (1,767.29, 3,394.43) 11.03 (7.93,15.55) 7.39 (4.76,11.17)

454.88 (162,1,570) 41,133.26 (33,360.46, 48,587.09) 30,809.09 (23,298.23, 39,143.82) 8,772.99 (3,618.97, 1,273,982.19) 38.63 (15.86, 150.33 ) 6.69 (3.49,13.49)

143.51 (268.11) 27,357.45 (10,212.04)

143.51 (112.24) 27,009.42 (2,574.45)

16,589.07 (10,458.32)

16,589.07 (3,262.76)

23,684.78 (161,479.00)

3,188.10 (2,394.90)

21.36 (63.55)

13.84 (10.61)

6.34 (2.44)

7.27 (2.14)

Notes: The notes below are applicable to Tables 5.5–5.12. Min and max respectively refer to minimum and maximum, and std dev. refers to standard deviation in all the tables. Annual deflated earnings in all the tables refer to 1982–4 100 as the base year. a. b. c. d.

Total employment in this scenario includes jobs created and retained. Net benefit is defined as earnings per job – annual reservation wage. Costs per job include property tax abatements, other local incentives, and costs of infrastructure. BC ratio is defined as the ratio of net benefit to costs per job. Note that it is not average net benefit per job (row 3) divided by average cost per job (row 4). If I calculate BC ratios based on average net benefit and cost per job (rows 3 and 4 respectively), I would not reveal valuable information on the individual firm’s BC ratios which I have calculated as the ratio of individual firms’ net benefits to costs per job. The minimum and maximum values for the distribution of BC ratios should make clear the average values of the BC distribution, when we compare them to the net benefit and cost distribution. e. Unweighted means are the raw averages of the relevant variable for all the firms. f. Weighted means are weighted averages in which the weights are the number of firms in each category (firms in 0th–25th percentile, those in the 25th–50th, 50th–75th, and 75th–100th percentiles) of the relevant distributions. For the weighted means, the weights are multiplied by the means for firms in each category and sum of the weights and the means are divided by the total number of firms. A similar procedure is used to calculate weighted standard deviations for the various distributions.

108

Table 5.11 Distribution of net benefits and benefit–cost ratios for Ohio’s (62) enterprise zones, scenario 3 (assumed elasticity 0.3)

Table 5.12 Distribution of net benefits and benefit–cost ratios for Ohio’s (32) enterprise zones, scenario 3 (assumed elasticity 0.1)

Total employmenta Annual deflated earnings per job ($) Net benefit per job ($)b Costs per job ($)c

BC ratiod Unemployment rate (%)

Mean (min,max) for zones: 0th–25th percentile

Mean (min,max) for zones: 25th–50th percentile

Mean (min,max) for zones: 50th–75th percentile

Mean (min,max) for zones: 75th–100th percentile

Unweighted mean for all zones (std dev.)e

Weighted mean for zones (weighted std dev.)f

4.50 (1.55, 7.42) 15,085.66 (11,013.22, 18,899.14) 5,148.80 (125.77, 8,692.74) 952.72 (199.66, 1,827.23) 0.56 (0.03,1.61) 5.40 (3.05,6.88)

11.34 (8,16.14) 21,350.98 (18,905.91, 25,450.02) 10,728.34 (8,800.55, 13,272.49) 2,435.89 (1,925.31, 3,175.07) 2.20 (1.88,2.39) 7.97 (5.24,9.35)

38.85 (27.97, 48.72) 28,404.33 (25,844.22, 32,716.81) 17,587.70 (14,066.98, 22,528.87) 5,605.98 (3,669.23, 10,585.77) 5.79 (3.69,8.76) 7.97 (4.76,11.00)

182.11 (64.01, 523) 39,968.96 (34,347.08, 43,895.87) 30,119.24 (25,039.41, 36,384.07) 34,076.44 (11,433.87, 97,277.84) 30.38 (9.48, 111.97 ) 6.87 (2.05,13.49)

59.20 (106.01) 26,202.48 (9,866.44)

59.20 (43.49)

15,896.02 (10,019.75)

15,896.02 (3,288.36)

12,742.80 (23,168.14)

10,015.86 (6,572.82)

9.73 (20.48) 6.34 (2.44)

9.73 (9.27)

26,202.48 (3,099.63)

7.05 (2.01)

Notes: The notes below are applicable to Tables 5.5–5.12. Min and max respectively refer to minimum and maximum, and std dev. refers to standard deviation in all the tables. Annual deflated earnings in all the tables refer to 1982–4 100 as the base year. a. b. c. d.

109

Total employment in this scenario includes jobs created and retained. Net benefit is defined as earnings per job – annual reservation wage. Costs per job include property tax abatements, other local incentives, and costs of infrastructure. BC ratio is defined as the ratio of net benefit to costs per job. Note that it is not average net benefit per job (row 3) divided by average cost per job (row 4). If I calculate BC ratios based on average net benefit and cost per job (rows 3 and 4 respectively), I would not reveal valuable information on the individual firm’s BC ratios which I have calculated as the ratio of individual firms’ net benefits to costs per job. The minimum and maximum values for the distribution of BC ratios should make clear the average values of the BC distribution, when we compare them to the net benefit and cost distribution. e. Unweighted means are the raw averages of the relevant variable for all the firms. f. Weighted means are weighted averages in which the weights are the number of firms in each category (firms in 0th–25th percentile, those in the 25th–50th, 50th–75th, and 75th–100th percentiles) of the relevant distributions. For the weighted means, the weights are multiplied by the means for firms in each category and sum of the weights and the means are divided by the total number of firms. A similar procedure is used to calculate weighted standard deviations for the various distributions.

110 Incentives for Regional Development

rate for the (two) zones with a B–C ratio greater than 100 (with an assumed elasticity of 0.3) was 9.9 per cent, higher than 120 per cent of the state’s average unemployment rate in 1990. With an assumed elasticity of 0.1, the only zone that had a B–C ratio greater than 100 (111.97) had high unemployment rate, being 8.13 per cent in 1990. 5.6.4

Summary of the B–C analysis

Table 5.13 summarizes the weighted and unweighted average B–C ratios for the various scenarios at the firm and the zone level. The summary in Table 5.13 conforms to our expectations. The most optimistic zone-level B–C ratios are in scenario one, which makes favourable assumptions regarding job creation and retention. These ratios become smaller when only those jobs that were attributable to the tax incentive are taken into account (especially with the lower assumed elasticity). B–C ratios are considerably lower when only jobs created were taken into account, also as expected.21 Thus, on average, the B–C ratio for firm-level and zone-level performance for local government investment indicated that the local net benefits from employment can be expected to be greater than the local costs of generating them. Certain caveats have to be noted immediately. The results from the B–C analyses refer only to local government investment, rather than to all government investment. Second, the summary in Table 5.13 is based on averages. This indicates that, while, on average, net benefits from jobs can be expected to be greater than the costs of generating them, it is not a good strategy for all zones to adopt tax incentives in order to create employment. The policy implications of the findings relate to programme design and focus on targeting of areas for zone designation and targeting of certain kinds of firms.

5.7

Policy implications

The results from the B–C analyses call for more selective designation criteria that can possibly result in a reduction of competition by reducing the number Table 5.13 Summary of average B–C ratios in various scenarios at firm level and zone level Firm-level

Scenario 1 (all employment taken into account) Scenario 2 (only jobs created taken into account) Scenario 3: elasticity 0.3 Scenario 3: elasticity 0.1

Zone-level

Unweighted

Weighted

Unweighted

Weighted

24.7

24.7

27.0

25.6

9.1

8.9

10.2

10.2

27.7 12.8

27.7 12.8

21.4 9.7

13.8 9.7

Evidence from Ohio’s Enterprise Zones 111 Table 5.14 Proportion of zones with B–C ratios 1 in various scenarios Scenario Scenario Scenario Scenario Scenario

Percentage of zones 1 2 3 (assumed elasticity: 0.3) 3 (assumed elasticity: 0.1)

90 75 80 78

of zones in this state. In this context, it must be enlightening to know that, as of 1997, Ohio was one of only three states in the United States (along with Louisiana and Arkansas) to have more than 75 enterprise zones in its territory. Recently the conventional state of Iowa in the Midwestern United States passed enterprise zone legislation, and as of September 2002, 332 EZs were certified in this state, as I highlight in Chapter 1. As of the end of 2003, Ohio had 391 enterprise zones. Based on the results found here, one implication is to decertify zones that perform poorly in terms of B–C ratios. Table 5.14 shows the proportion of zones that had unweighted B–C ratios greater than one in various scenarios.22 Table 5.14 shows that, as might be expected, the proportion of zones with B–C ratios greater than one is the highest in scenario one. In scenario two, only 75 per cent of zones have B–C ratios greater than one. In scenario three, the proportion of zones with B–C ratios greater than one is higher (being 80 per cent) with a higher assumed elasticity than with the lower assumed elasticity (where it is about 78 per cent), which is again consistent with what we would expect. What are the factors that distinguish zones with B–C ratios greater than one from those that have B–C ratios less than one, and what are the policy implications from these findings? I explain this at the end of this section. Based on the B–C ratios, it is possible to evaluate the efficiency implications of decertifying the poorest performing zones. Efficiency loss is defined as the extent to which costs exceed benefits in the zones (defined as poorperforming) in which the B–C ratio was less than one. Table 5.15 summarizes these efficiency losses in the poorest performing zones in various scenarios. When I studied total employment (created and retained) created by the firms (scenario one), I found that most (eight) of the (11) zones that had B–C ratios less than one were limited authority zones at the time, that is, were designated on the basis of their population, rather than on high unemployment. The cumulative efficiency loss in the form of property tax abatements and other incentives provided by the local governments of these (eight) limited authority zones amounts to $44 million (or about $5.5 million per zone).23 When we take into account only those jobs that are created (as in scenario two), the efficiency loss turns out to be of higher magnitude, as we would expect. More zones (21) turn out to be inefficient in this scenario than in

112 Incentives for Regional Development Table 5.15 Efficiency losses in various scenarios Scenario Scenario 1 (created and retained employment taken into account) Scenario 2 (only created employment taken into account) Scenario 3: assumed elasticity 0.3 Scenario 3: assumed elasticity 0.1

Efficiency losses (million $) 44.2

44.9 16.1 4.2

scenario one (8), with the assumption being more restrictive. The property tax abatements, provision of infrastructure and other incentives to firms in the 21 zones with B–C ratios less than one in this scenario amounts to $45 million (or about $2 million per zone). Understandably, many of the zones that perform poorly in scenario one are also the ones that perform poorly as well in this scenario. The average unemployment rate for the 21 zones that had negative B–C ratios is 6.92 per cent, only average unemployment by state standards (just about 120 per cent of the state of Ohio’s average unemployment rate of 5.7 per cent for 1990). In scenario three, the efficiency loss with the higher elasticity assumption is higher than with the low elasticity. With the assumed elasticity of employment with respect to taxes being 0.3, the cumulative efficiency loss is $16 million for six zones, the average unemployment rate for them being 6.73 per cent. If the elasticity were assumed to be 0.1, then the efficiency loss is $4 million for five zones (approximately $850,000 per zone) whose average unemployment rate was just 6 per cent for 1990. Overall, when I examine zones with B–C ratios less than one and their unemployment rates in various scenarios, I find that, consistent with Bartik’s hypothesis, they are mostly low-unemployment areas. I find that all of the firms and zones with negative B–C ratios and positive B–C ratios less than one are ones in which the earnings from jobs are low relative to reservation wages. This is because of the areas’ lower unemployment rate in relation to the state average (this follows from Table 5.1), leading to low net benefits and consequently low B–C ratios from employment. On the other hand, wherever firms that located in high-unemployment zones with low reservation wages create well-paid jobs, high net benefits and high B–C ratios are found. Some of the zones in which the net benefit from employment could have been high due for this reason, have lower B–C ratios because of the large abatements given to firms that located in them. I find, based on my analysis, that even when skilled or unskilled (respectively well- or poorly-paid) jobs are created in low-unemployment zones, the B–C ratios are low because of higher reservation wages in such low-unemployment zones. The first implication that emerges from this analysis is that it is beneficial for local government investment when skilled (that is, well-paid) jobs are

Evidence from Ohio’s Enterprise Zones 113

created in high-unemployment zones since they are lower reservation wage areas. This maximizes the net benefits from employment. The creation of well-paid jobs in high-unemployment areas can be induced by providing incentives (similar to current state-level incentives that exist relating to Comprehensive Employment Training Act (CETA) employees) to firms in the area to provide training to their low-skilled employees in order to help them obtain the necessary skills. Since the rate of unemployment tends to be higher in areas where the labour force is less skilled, in the absence of incentives, firms would have little incentive to offer skilled jobs to unemployed zone residents. Secondly, it is necessary for local governments to ensure strict compliance from firms in terms of job commitment so that they create the employment stated in the agreement. This implies a larger role for the tax incentive review council (TIRC) which has been constituted for this purpose. It is also necessary to place a ceiling on the amount of the abatement to be given to a single firm, based on the findings in this chapter. As of now, the legislation states the limits of the incentives only in terms of the percentage of assessed values of the property (see section 5.3 and Chapter 4). The location of firms that create low-skilled jobs in low-unemployment areas is not beneficial for either the state or for local government and hence should not be encouraged. As of now, retail operations are not eligible for tax incentives within zones and this is a step in the right direction, based on the results found from work in this chapter. Finally, it may be necessary to decertify any zones that perform poorly, depending on the assumption regarding employment. It may be noted from Table 5.15 that total efficiency losses are at their greatest in scenario two (a restrictive assumption involving only jobs newly created) and at their lowest in scenario three (with an assumed elasticity of employment with respect to taxes being 0.1). The analysis with scenario three (with an assumed elasticity of 0.3) provides a middle ground.24 These are a few zones that perform poorly in all the scenarios. The implication is that these zones can be decertified, thereby avoiding these efficiency losses. Alternatively, if one were to apply the model developed here only to new jobs created (when they are held by the local unemployed), the implication would be that we should decertify the zones that do not perform well in scenario two. Thus, it has to be recognized that the decertification implications for the programme depend upon the assumptions one makes. This final implication is consistent with what is reported in Chapter 4. While work in that chapter found support for the robust result that areas with tax incentives do observe a reduction in their unemployment rate, it concludes by discussing that there is less consensus regarding the duration for which the tax incentives should be offered. Based on the findings, that chapter suggests three to five years to be the optimum period for maximizing the effect of tax incentives on the unemployment rate, after which it is

114 Incentives for Regional Development

preferable that the area stop offering incentives. In terms of policy action, this translates into decertifying areas that have been tax incentive areas for longer than five years. It is easy to see why it is likely that unemployment problems return if incentives continue for long periods of time. With time, tax incentives encourage lobbying for such programmes and degenerate into employment redistribution games that increase the unemployment in those areas that lose firms. Also, we may note that, although decertification depends upon the assumptions we want to make, certain bottom-line results appear to be valid in every case. More selective designation criteria are to be used for zone designation – these could be distress criteria relating to high unemployment. This implies the decertification of zones that do not qualify on the basis of unemployment or other distress criteria. Appropriate incentive structures and monitoring mechanisms must be devised that reward labour-intensive firms (those with high labour–output ratios) that create skilled jobs along with on-the-job training. Since labour-intensive firms create employment, they must be made eligible for incentives to train labour to move up the value chain of producing a product or service. If these criteria are taken into account in the design of the programme and targeting, the EZ programme in Ohio can be capable of generating greater employment for its high-unemployment areas and reducing wasteful competition. Such enlightened programmes can provide a positive response to the policy debate and to the challenges that have been raised in the literature regarding such traditional tax incentive programmes. In the next part of the book, we will consider place-oriented regional development policies that focus primarily on infrastructure rather than tax incentives. The next couple of chapters present evidence from infrastructure incentives offered in the growth centres of India, as a contrast to the experience from tax incentives. The final chapter and part compares the lessons learnt from the two approaches to regional development.

Part III Infrastructure Incentives: Theory and Evidence

This page intentionally left blank

6 Impact of Growth Centres on Unemployment and Firm Location: Evidence from India

6.1

Introduction

In this part of the book, we move the focus away from incentives that are primarily financial, to physical incentives offered by state and local governments to attract firms. The most important of these physical incentives is infrastructure. In Part III of this book, consisting of this chapter and the next, I focus on infrastructure incentives. I present evidence from India’s growth centres (GCs), areas which offer incentives such as power, water, telecom, and banking to enable the states to attract industries. When we consider a developing country such as India, we find much greater regional disparities than those observed in developed countries. Because of this, the debate on convergence has occupied a great deal of attention in the development literature. Convergence implies that in a steady state, poorer regions can be expected to grow more rapidly than their richer counterparts. This occurs mainly as a result of the free flow of capital to the poorer regions (because of capital shortage, the rate of return to capital in such regions will be higher). Convergence also occurs because poorer regions need not reinvent the wheel: they can imitate the technological changes adopted by the richer regions. The debate on convergence in Indian states is divided, but the majority view has been that the disparities between the poorer and richer states and regions have widened enormously (Rao, Shand and Kalirajan, 1999; Ahluwalia, 2000; Srivastava, 2001; Singh and Srinivasan, 2002; Cashin and Sahay, 1996; Rao and Sen, 1997; Nagaraj, Varoudakis and Veganzones, 1998). This counter-intuitive observation is understandable because while the economic reforms initiated in India in the 1990s had a positive influence on economic growth and the richer states, the poorer states grew quite slowly. The extent of ‘trickle-down’ of growth in developing countries primarily dependent upon agriculture is determined by the employment elasticity of agricultural output, which has remained stagnant in India. Further, non-agricultural employment in rural 117

118 Incentives for Regional Development

India has not grown rapidly enough to absorb the expanding labour force. The infrastructure has remained a bottleneck and has hampered the ‘trickle-down’ of growth and employment opportunity. Thus, regional disparities in India have not diminished in recent years. The Government of India introduced the growth centres (GCs) programme in June 1988 to give impetus to industrialization processes in backward regions. Under the programme, 71 GCs were set up across the country that were supposed to be allotted to the various states on the basis of a set of combined criteria of area, population and the extent of industrial backwardness. These GCs, that continue to be in existence at the time of writing, provide basic industrial infrastructure like power, water, telecom, and banking to enable the states to attract industries. Central government funds for the programme are to be leveraged by the states for purposes of financing.

6.2 The importance of and the motivation for research A number of studies consider the effect of taxes on firm location decisions and the local economies, holding public services constant (see, for example, Bartik, 1991). The chapters in Part I of the book contain an exhaustive review of the literature. Here we will merely note that there are few studies dealing with the effect of non-tax incentives, but many with tax incentives (Steinnes, 1984; Papke, 1987, 1991; Tannenwald and Kendrick, 1995; Tannenwald, 1996, cited in Fisher and Peters, 1998). The work in this chapter is an attempt to fill this gap in the literature, and uses empirical evidence from India’s GCs programme, as a representation of the infrastructure incentives, that state and local governments can offer to firms. Consistent with the idea of place-oriented policies, growth centres in India are similar to the enterprise zones of the United States since they primarily offer infrastructure incentives for investors that locate there. As highlighted in Chapter 1, the GCs programme assumes particular significance in light of a recent decision to stop the war of tax incentives among the Indian states. Until recently, in a policy similar to that of their US counterparts, various states in India had been offering tax incentives to investors in order to persuade them to locate in their states, make investments and create employment. There are several instances of competition for firms among the Indian states in the post-liberalization (1991) period. Recently, the south Indian states, Karnataka and Andhra Pradesh, had been competing fiercely to secure the location of Microsoft’s newest India facility by attempting to showcase their pool of technical talent and skilled workforce, stable leadership and governance, and market potential. Furthermore, in terms of actual tax incentives, the past decade in India had witnessed several instances of generous abatements to firms.

India’s Growth Centres and Unemployment 119

In November 1999 a conference of state chief ministers and the Union Finance Minister, Government of India, decided to stop this tax war among the Indian states. This is noteworthy in view of the fact that even states in the United States have not taken this bold step, despite the point that a number of American economists and those in US policy circles continue to be worried about the effect of tax incentives on the fiscal health of states and their ability to provide public services (for instance see Netzer, 1991). The decision was taken by the Indian states because the offer of tax incentives, in addition to affecting the general fiscal health of the states, produced financial constraints on the states’ ability to provide infrastructure services, given that sales tax revenue accounts for nearly one-quarter of own source revenue for the majority of Indian states. Further, to alleviate the complex and cascading tax structure in the Indian states, a unified value added tax (VAT) is proposed to be introduced from April 2005.1

6.3

Objectives

The objectives of this chapter are to answer the following questions: 1. To examine empirically the effect of infrastructure incentives (GCs) on unemployment rate, as in the literature. 2. To study the effect of infrastructure, where GCs have been established, on firm location. In addition to the secondary data I obtained from DIPP regarding the functioning and effectiveness of GCs, secondary, district-level data from all Indian states (containing GCs) are obtained to answer the above questions to evaluate their effectiveness. In an attempt to answer the first question, I estimate the unemployment rate, using data from the Census of India, analogous to work undertaken in Chapter 4. To answer the second question, I estimate the impact of GCs on firm location, as depending on the existence of infrastructure. This chapter is organized as follows. The next section explains why the growth centres programme is relevant in the Indian context. This is followed by a description of relevant literature in the context of developing countries. The chapter then describes the growth centres programme, followed by an exposition of the theory and the empirical model that form the basis of the work. This is followed by a summary of the secondary data, and the results from two estimations: the estimation of the unemployment rate; and the estimation of the performance of the GCs. The final section summarizes the policy implications of the research. 6.3.1

Why growth centres?

There are several reasons why the GC programme is relevant for India’s industrial and urban policy. For example, the costs of infrastructure, specifically,

120 Incentives for Regional Development

real estate (see Sridhar, 2004), for some recent evidence regarding this phenomenon) in the major cities of the country make it difficult for Indian firms to win the price war. Faced with the transitions that occur with city growth, industry in India must relocate to smaller and medium-sized towns, as the GCs programme rightly emphasizes. If state and local governments invest in improving the physical infrastructure in small and medium-sized towns, they can compete as alternative locations for firms which are considering new investment or diversification decisions. This can reduce operational costs (such as rental or leasing costs) for firms that would find such costs to be much higher in metropolitan areas. In smaller and medium-sized towns, where the GCs are located, firms will be able to secure an adequately trained labour force at a relatively low cost. This is because the cost-of-living adjusted wage is itself lower in smaller and medium-sized towns, holding quality constant. Industries are keen to exploit this advantage. If industry were to locate in semi-urban and rural areas (the emphasis of the GC programme) this would change the areas’ economic base. As the Todaro model shows, rural–urban migration in most developing countries occurs for reasons of employment. Data compiled by India’s National Institute of Urban Affairs (NIUA) show that more than half of the men that migrated from rural to the urban areas of India in 1991 did so for employment opportunities. Relocation of industry to rural and semi-urban areas would help the rural poor and surplus labour to find alternative employment in their own localities. This acts as a brake on urban migration, thereby alleviating housing pressures and preventing the creation of slums. In addition, the relocation of industry would also help in the co-ordination of the government’s poverty alleviation programmes. So far, it has been found that recipients of various employment training programmes in India have been unable to find suitable employment in rural areas. The location of firms in semi-urban and rural areas would help in matching the demand for and supply of skills. Finally, the development of smaller and medium-sized towns might imply that they are self-contained communities, but eventually automotive ties have to develop between urban areas and their satellite towns. This implies that the development of an adequate road and highway system, which has been neglected for a long time in India, will finally receive the attention it needs. The GC approach is a test of the alternative to the tax war among the states. It is also a test of the attempt to foster convergence among Indian states. This chapter analyzes the effect of infrastructure incentives on firm location decisions and on the area’s unemployment rate. Note that Bartik’s (1991) work showed conclusively that tax incentives have an impact on firm location decisions, holding public services (infrastructure) constant. If there

India’s Growth Centres and Unemployment 121

were to be a moratorium on tax incentives, will infrastructure incentives become more important to firm location decisions? Can other countries learn from the Indian experience to say no to tax incentives? This chapter is an attempt to understand the role of infrastructure incentives in firm location decisions, and their impact upon the local unemployment rate of the local economies that adopt them. In order to do this, it uses evidence from India, and is partly analogous to Chapter 4 which reports similar results from tax incentives offered in Ohio, from the Midwestern United States. Further, as I highlighted in Chapter 1, in relation to enterprise zones and tax incentives in the United States, there is a lot of scepticism in Indian academic and policy circles regarding GCs and particularly as regards their effectiveness in attracting firms. Some commentators see the GC programme as having been a colossal failure in attracting firms, while others strongly believe in its effectiveness because of the infrastructure incentives available to firms that locate there. This chapter presents empirical evidence of the effectiveness of India’s GCs, based on secondary data from the Department of Industrial Policy and Promotion (DIPP), Ministry of Commerce, Government of India, and secondary, district-level data available from the Census of India. 6.3.2

Past literature

The regional development literature frequently speaks of ‘growth poles’. The core idea of the growth poles theory is that economic development, or growth, is not uniform across an entire region, but instead takes place around a specific pole. In the context of growth centres, the cities may be viewed as the growth poles. These poles are often characterized by a key industry around which related industries develop, through both direct and indirect effects. The expansion of this key industry implies the expansion of output, employment and related investments, as well as new technologies and new industrial sectors. Because of scale and agglomeration economies near the growth pole, regional development is unbalanced. Transportation, or infrastructure more generally, can play a significant role in such a process. The more dependent or related an activity is to transportation and other infrastructure, the more likely and strong this relationship. This literature argues that the emergence of a secondary growth pole is possible at a later stage, mainly if a secondary industrial sector emerges with its own linked industries. Boisier (1981) points out that regional development literature in the period 1975–9, in fact, makes no mention of growth poles. This may be said of literature in the recent past as well. In general, the literature on growth poles and growth centres is sparse. By contrast, there is a vast body of theoretical, empirical and policy literature that deals with firm location decisions and tax incentives from the USA. Literature dealing with infrastructure incentives is much more limited, especially as they relate to firm location, and this is particularly true in

122 Incentives for Regional Development

relation to developing countries. A majority of the literature studies the relative impact of tax incentives on firm location decisions, using data from the USA, and holding public services constant, as has been summarized in earlier chapters. The literature argues that if public services were not held constant, tax incentives are not important in firm location decisions. In order to compare the effectiveness of tax vis-à-vis infrastructure incentives, I shall summarize existing studies relating to tax incentives in India. Other developing countries have used the policy of growth centres with objectives which are similar to those pursued in India. Kimani and Taylor (1973), presenting the results of a joint venture between the geography department of Carleton University, Ottawa, Canada, and the University of Nairobi, Kenya, studied all aspects of Kenya’s Muranga district growth centre with a view to understanding how a strategy of growth centres can assist in rural development. This study found that in Kenya, ‘growth centres’ are created by policies designed to improve the capacity of rural areas to grow and to enhance their capability to hold their productive populations. They are meant to avoid serious problems arising from the concentration of people in large urban centres, very similar to those in India (see section 6.3). A study on sales tax incentives in the Indian context (Tulasidhar and Rao, 1986) shows a loss of both employment and output as a result of tax incentives, albeit in a partial equilibrium framework. Their study examines sales tax incentives in an Indian state, Madhya Pradesh, and places the revenue loss as high as 7 to 10 per cent of the sales tax revenue. Their analysis of a large number of medium and large-scale industries indicated that the sales tax incentive, whichever way it is designed, is not the most appropriate instrument to raise the level of investment or spread this to backward areas. A study by Shettar (1988) examines the impact of India’s growth-centred industrial policy on rural and agricultural development as it affects the provision of basic needs. This study calls for a reorientation of planning priorities towards creating a strong agricultural base for the rural majority. The study suggests that, in order to increase agricultural production, full development and utilization of the nation’s irrigation potential is necessary. Based on data from the Indian state, Madhya Pradesh, Rajaraman, Mukhopadhyay and Bhatia (1999) find that fiscal incentives have a statistically insignificant impact on medium- and large-scale investment in Madhya Pradesh. Conversely, the study finds that a well-developed energy infrastructure was an important factor attracting investment into the state during the 1980s, highlighting the importance of infrastructure in firm location decisions. Based on the literature, we can make several observations about the relative importance of tax incentives and public (primarily infrastructure) services. First, only when the level of public (infrastructure) services is held constant will the benefits to firms of tax incentives be significant. Bartik (1991), summarizing studies using data from the United States, found that, holding public services constant, if a local government in a state were to

India’s Growth Centres and Unemployment 123

reduce its taxes, in the long run, it could expect to see an increase in its business activity. These elasticities are for the USA and, in general, depend upon the degree to which the alternative locations considered are good substitutes. They could differ from country to country depending upon the degree to which different alternative locations are similar or not from the viewpoint of prospective businesses. The expectation is that in heterogeneous countries such as India, alternative locations are likely to be quite dissimilar in terms of attributes such as the availability of raw materials, labour force skills, transportation and other costs, and proximity to markets. So the tax elasticities might well be lower for India than Bartik’s study suggests. Secondly, it is not necessary that incentives to industry are harmful to the development of infrastructure, as is demonstrated by the proposal of the south Indian state Karnataka to replace financial incentives to the auto industry with a series of infrastructure improvements. Karnataka promised training institutions, schools, office complexes, housing, a globally well-knit telecom network, roads, dedicated power and water supply, all of which would be beneficial to the development of automobile manufacturing units, vendors and dealers. If other states in India follow this example, competition will likely enhance their infrastructure competitiveness and their image as good places to live and do business. Last but not least, as we have argued, the thesis of the work is that poorer states and regions are much more justified in offering infrastructure rather than tax incentives in order to attract industry and employment, similar to what Bartik (1991) argues with respect to supply-side incentives. This is especially relevant in the context of developing countries such as India. If distressed areas, as a result of providing infrastructure improvements, were to be successful in attracting firms to invest and create employment, greater social net benefits would accrue to the area.2 In developing countries such as India, there is much more reason for us to believe that, in the absence of social and family support networks, the unemployed will be willing to accept jobs at lower wages in distressed areas than would be the case in their richer counterparts. In India, since the 1950s, the disparities between the poorer states (BIMARU3 (Bihar, Madhya Pradesh, Rajasthan and Uttar Pradesh), as they are called), and the richer, southern and western states have widened enormously. Infrastructure incentives used to attract employment by distressed areas can become a tool to increase the net benefits from the employment generated in these areas, until such time the disparities between them and the richer states narrow. Further, we have to think about stimulating, rather than stifling, competition, in the provision of infrastructure by all the states. Such reforms are clearly necessary. From this viewpoint, geographically targeted programmes such as GCs are critical in influencing firm location decisions because they encourage competition among the states in the provision of infrastructure

124 Incentives for Regional Development

which are quite critical to firms. Further, can this have implications for American states which wish to compete in the provision of infrastructure incentives?4

6.4

Description of the growth centres programme

Growth centres in India were originally conceived as an attempt to move industrial development away from the urban centres, similar to the policy pursued in Kenya. Criteria for selection of GCs are described in a notification issued by the Ministry of Industry, Government of India, dated 8 December 1988. These criteria were to suit the size of cities, as follows: I

II III

GCs shall not be located: 1. Within 50 kilometres of the boundary of seven cities in the country with a population above 2,500,000; 2. Within 30 kilometres of the two cities with a population between 1,500,000 and 2,500,000; and 3. Within 15 kilometres of the boundary of the 12 cities in the country with a population between 750,000 and 1,500,000. The GCs should be located close to district, sub-divisional, block and sub-district headquarters or developing urban centres. GCs shall have access to basic facilities – proximity to railheads, national or state highways, water supply, power, telecommunications, educational and health facilities. If such facilities are not readily available, they should be developed as an urgent priority.

Currently, in theory, GCs are in place in 71 of the 560 districts of the country, although only around one-third (24) of these are completely functional. For the programme, central funds are leveraged based on the understanding that state funds will be released. The states use the central plus leveraged state funds under the programme to acquire lands in the areas chosen to develop them, and lease them to firms. The cost of land acquisition consists of the payment of compensation by the states to landowners. Development of land refers to the construction of infrastructure such as roads, water, sewer, power, and telecom services. Detailed summary descriptions of the GCs that have been establised, are given in section 6.8, where I estimate GC performance. GCs contribute to not more than 5 per cent of total (agricultural and non-agricultural) employment in the districts where they are located, thereby constituting a relatively small portion of the total economic activity in the districts. Nevertheless, intuition, theory, past literature and empirical evidence suggest that such programmes are likely to have positive effects in reducing the unemployment rate of areas that adopt them, consistent with the original objectives with which they were established in the country.

India’s Growth Centres and Unemployment 125

6.5

Theory and model

One of the objectives of this chapter is to examine the impact of infrastructure incentives on unemployment, to enable a better assessment of the programme, consistent with our overall objectives in the book, earlier theory and past literature (for instance Papke, 1994). Basic labour economic theory (Ehrenberg and Smith, 1994) shows that the labour market outcomes – the unemployment rate (the number of people employed or unemployed) and wages (price of labour) – are simultaneously determined, and are determined by forces of demand for and supply of labour. The demand for labour depends upon wages, technology, and capital. The supply of labour depends on wages (as demonstrated by the reservation wage property in the theoretical model), and non-wage characteristics such as hours of work, and flexibility. So we have: Q DL f (Wages, Technology, Capital)

(6.1)

Q LS g(Wages, Hours of Work, Flexibility)

(6.2)

where Q LD and Q LS respectively are the demand for and supply of labour. Note that the demand and supply equations (6.1) and (6.2) are both over-identified.5 To estimate unemployment rate (measure of Q DL Q LS in equilibrium), we may write the above equations in reduced form.6 The unemployment rate, in reduced form, is a function of wages.7 Pantuosco and Parker (1998) show the unemployment rate in reduced form as being dependent upon wages that are determined by various socio-demographic characteristics. It is reasonable for us to believe this because the number of those willing to work depends upon the wage rate. For instance, at very high wage rates (as in the instance of the software industry in India), those working at home (for instance, housewives with professional degrees) or dependents (students, for example) may also become willing to work. The model shows that along with these socio-demographic characteristics that determine wages in reduced form, infrastructure incentives (similar to tax incentives) (represented by the GC) also affect the rate of unemployment. In including the GC dummy as an explanatory variable, I adopt the standard approach used in the literature that estimates an econometric model to assess a programme (for instance, see Papke (1994)). Typically, in such models, the outcome variable of interest is estimated as a function of the variable that characterizes the programme to be evaluated. Consistent with this literature, I estimate the unemployment rate in reduced form as dependent on growth centre status and socio-demographic characteristics such as average age, proportion of minorities (in the Indian context, Scheduled Castes (SC) and Scheduled Tribes (ST)), proportion

126 Incentives for Regional Development

male, and literacy rate, that determine wages, using district-level data for India. Other variables that should be incorporated in the reduced form equation for the rate of unemployment include technology, capital, the hours of work and flexibility (see equations (6.1) and (6.2)).8 Since the data are for a single country, we do not expect the level of technology or the availability of capital to vary significantly enough for us to include them in the estimation. Further, the nature of jobs available in India is such that there is little variation in hours of work9 and flexibility (for instance, most of them tend to be eight-hour jobs, with little flexibility in timings).10 Hence I do not include measures of these variables in the estimation, but include only those that determine wages, in reduced form.11 The model estimating the unemployment rate of the ith district may be summarized as follows:12 Ui Dummy for GCi Duration of GCi Duration of GC squaredi Manufacturing employmenti Service employmenti Proportion SC and STi Proportion malei Literacy ratei Mean agei ei

(6.3)

As equation (6.3) shows, in addition to controlling for the GC dummy and its duration, I control for socio-demographic characteristics and the occupational composition of areas that determine the wage (in reduced form) at the district level, for all Indian states containing them. We expect sociodemographic characteristics such as average age, gender, minority, and literacy status to affect the unemployment rate, since all of these directly affect the individual’s labour force status. Further, different occupations, as reflected in the proportion of employment in manufacturing as opposed to service occupations, could have different unemployment rates as the result of different demand and supply conditions.13 These measures reflect the importance of these sectors in the area’s economic base, and it is necessary to control for them. I expect the proportion of manufacturing employment in the area to have some impact on the results, and the proportion in service employment to have negative impact on the unemployment rate. This is because of the increasing importance of services in India’s economy. The expectation from theoretical models that have been developed in the literature is that areas with targeted programmes (such as GCs) see a reduction in their unemployment rate. For instance, Ge (1995) and Seyfried (1990), along with Chapter 2 of this volume, contain theoretical models that show these impacts. Chapter 4 contains empirical evidence from Ohio regarding these effects. Further, Papke (1994) provides evidence about the employment impacts of Indiana’s EZ programme. Consistent with this literature, along with other variables that determine the unemployment

India’s Growth Centres and Unemployment 127

rate, I include a dummy for whether or not GC exists at the district level, for all Indian states.14 This methodology, while allowing us to control for all of the other variables that affect an area’s unemployment rate, enables us to consider the impact of the GC and infrastructure. This is also consistent with the original objective with which GCs were set up – to promote the industrialization of backward areas in India. Note that while in Chapter 4, the tax incentive programme and the unemployment rate are considered to be simultaneously determined, in this case, that problem does not arise. In fact, GCs were allocated to various districts in states on the basis of combined criteria consisting of area, population and industrial backwardness. Through several discussions, I confirmed that the unemployment rate is not a consideration for GC designation. Since, however, unemployment could be correlated with industrial backwardness, I use land area and population as instruments for the possibly endogenous GC dummy and also report results from 2SLS estimation. In addition to the effect of GCs on the unemployment rate, I control for the time period (duration, in months, as of December 2001)15 for which it has been in existence in the area (since the day of certification). One can imagine that GCs could reduce the unemployment rate of an area, but there could be some optimum period for which it is desirable.16 To facilitate such an understanding, the approach would be to include a variable that indicates how long the GC has been in existence in each district. The research in Chapter 4 suggests that anywhere from three to five years could be the optimum period for maximizing the effect of tax incentives on the unemployment rate, after which it is preferable that the area should stop offering incentives. Here, we test a similar hypothesis of an optimum period for infrastructure incentives. In terms of policy action, this translates into decertifying areas that have been GCs after a certain period of time. In addition to the duration of the GC, I include its squared term in the estimation. This is in order to check for any non-linearity in the impact of duration of GC on the unemployment rate.17 For instance, one may expect that the GC would initially be highly effective in reducing unemployment, but that its effect could gradually taper off, either because bureaucracies make way into the institutional structure, or simply because business and governmental interest in the programme wanes.

6.6

Description of variables and data

Data on unemployment rate were not readily available. However data, on population, main, marginal and non-workers were available by district from the 2001 Census of India. Main workers are classified as those who had worked for the major part of the year preceding the date of enumeration.18 Marginal workers are those who had worked for some time in the year preceding the enumeration, but had not worked for a major part of the year.19

128 Incentives for Regional Development

If an individual had not worked at all during the last year, he or she is treated as a non-worker by the Census. Non-workers in the Census of India include: (i) (ii) (iii) (iv) (v) (vi) (vii)

those attending to household duties at home; students; dependents; retired persons or renters; beggars; inmates of institutions; and other non-workers.

To be consistent with Census’ definition of non-workers, for the purposes of this chapter, non-workers are treated as those outside the labour force.20 Marginal workers are treated as those that were willing, but had not found full-time work. The unemployment rate is thus the ratio of these marginal workers to those in the labour force (main plus marginal workers). Other variables are easily calculated from the 2001 Census of India. The literacy rate for each district is the total number of literates divided by the population older than six. The proportion of males refers to the male population older than six years computed as a proportion of the population older than six years. The proportion SC and ST refers to the proportion of scheduled castes (SC) and scheduled tribes (ST) in the district. This variable is included because SC and STs are groups of minorities that traditionally have been socially repressed in India. If true, their presence would have a positive effect on the unemployment of districts that have higher populations of SC and ST. This is because SC and STs are known to have generally lower levels of education and skills. In the absence of affirmative action policies providing reservation to them in education and employment, areas containing higher proportions of them would invariably have higher unemployment.21 At the time that research for the book was completed, the Census of India 2001 had not yet published data on population in various age groups, SC and ST population, and employment by category. For the purposes of calculating the average age of population, the proportion of SC and ST, and the proportion in manufacturing and service occupations, I used the 1991 Census of India. I assume that the proportion of employment in manufacturing and services, the average age of the population in the districts and the proportion of SC and ST in the various districts, remained broadly constant during the decade.22 The average age is computed as the weighted average of the population in every age group (the weights) and the ages.23 The proportion employed in manufacturing (manufacturing and processing in household industry, and other than household industry workers) and those in services (this includes workers in trade and commerce, those in transport, storage and communications, and in other services) are all calculated as proportion of

India’s Growth Centres and Unemployment 129 Table 6.1 Description of data used in estimation of unemployment rate Variable

Mean

Unemployment rate, 2001 0.24 Work participation rate, 2001 0.41 Literacy rate, 2001 0.64 Proportion male, 2001 0.52 GC dummy, 2001 0.13 Duration of GC (months), 2001 10.97 Duration squared 1,083.76 Population, 2001 18,20,992 Land area, 2001 5,414.92 Average age, 1991 34.06 Proportion SC and ST, 1991 0.31 Proportion employment, manufacturing, 1991 0.08 Proportion employment, services, 1991 0.19

Minimum

Maximum

Standard deviation

0.05 0.24 0.30 0.46 0.00 0.00 0.00 31,362 9 31.56 0.00

0.49 0.64 0.97 0.58 1.00 129.00 16,641.00 96,38,473 38,428 36.46 0.98

0.09 0.07 0.13 0.02 0.33 31.07 3,417.71 13,41,518.11 4,221.75 0.77 0.21

0.00

0.37

0.06

0.06

0.72

0.10

Note: Number of observations 543.

total workers.24 Land area of districts is in square kilometres and population in actual numbers. Table 6.1 describes the relevant data for 543 districts in Indian states containing GCs for which all data were available, and that are included in the estimation of the unemployment rate.25 On average, the 2001 unemployment rate for Indian districts (for which data were available) was around 24 per cent. There are a few districts that had nearly 50 per cent unemployment rate, the maximum unemployment district, with no GC, in the western Indian state of Rajasthan, reinforcing the need for some kind of targeting. For the purposes of interest and comparison, I also report workforce participation rates. Workforce participation rate is computed as the ratio of main plus marginal workers (total labour force) to the total population. On average, the 2001 workforce participation rate for India was 41 per cent, lower than that recorded in advanced countries such as the United States and Australia (where it was roughly 51 per cent each) and New Zealand (where it was roughly 50 per cent, based on World Bank data for 2001). It may be noted from Table 6.1 that districts, on average, had young populations, as may be seen in their average age. Since we expect people in their mid-thirties to be actively involved in the labour force (seeking or changing jobs), the GC approach is important to study for its effects on the unemployment rate and on the workforce participation of these youth.26 On average, the literacy rate is 64 per cent for districts in states containing GCs, roughly consistent with the figure for the whole country. The lowest levels of literacy are found in an area that did not have a GC. The maximum

130 Incentives for Regional Development Table 6.2 Comparison of data for areas with and without GCs Variable

Land area, 2001 (in square, kilometres) Population, 2001 Unemployment rate, 2001 Work participation rate, 2001 Literacy rate, 2001 Proportion male, 2001 Average age, 1991 Proportion SC and ST, 1991 Proportion employment, manufacturing, 1991 Proportion employment, services, 1991 Duration of GC (in months)

Non-GC districts, N 475

Districts with GCs, N 68

Mean

Standard deviation

Mean

Standard deviation

5,346.05 17,93,267 0.24 0.41 0.64 0.52 34.06 0.31

4,144.75 13,79,021 0.09 0.07 0.13 0.02 0.77 0.21

6,431.50 19,60,689 0.24 0.40 0.66 0.51 34.10 0.28

6,691.55 10,62,291 0.08 0.07 0.13 0.02 0.79 0.19

0.08

0.07

0.08

0.05

0.19 N/A

0.10 N/A

0.21 87.59

0.09 31.58

N/A: Not applicable.

literacy area had a GC. Roughly, on average, little more than half of the population of the districts is male. On average, about one-third of the population in the districts belong to minority groups such as SC and STs. On average, the proportion of employment dependent upon services is much greater than that dependent on manufacturing, reinforcing India’s image as a service economy. Thirteen per cent of the districts contain GCs. The mean for the GC dummy shows this. Table 6.2 compares these data for districts with GCs and those without them.27 As we might expect, districts containing GCs are larger in terms of land area and population, consistent with broad criteria used for their designation. Aside from this, there is little difference across districts with and without GCs, except for the literacy rate and the proportion employed in services. On average, districts with GCs are more literate, and contain a higher level of service employment than those without them. Finally, the proportion of SC and ST is smaller in the GC districts than in their non-GC counterparts. Data on the duration of the GC (for districts with GCs) show that on average, GCs have been in place for nearly 88 months (or a little more than seven years).

6.7

Results from the estimation of the unemployment rate

Table 6.3 shows the results from OLS and 2SLS estimations of the unemployment rate (shown in equation (6.3)), based on districts in the country for which all data are available. OLS is applied to the reduced form equation,

India’s Growth Centres and Unemployment 131 Table 6.3 Results from unemployment rate) Variable

OLS

and

2SLS

estimations

(dependent

OLS estimates Coefficient (standard error)

Constant Literacy rate, 2001 Proportion male, 2001 GC dummy, 2001 Duration of GC (months), 2001 Duration squared Average age, 1991 Proportion SC and ST, 1991 Proportion employment, manufacturing, 1991 Proportion employment, services, 1991

0.75 0.11 0.72 0.05 0.00

Adjusted R2 F

0.23 19.46

(0.25) (0.03) (0.23) (0.10) (0.00)

0.00 (0.00) 0.00 (0.01) 0.01 (0.02)

variable:

2SLS estimates t

Coefficient (standard error)

t

2.95 4.98 (3.37) 1.48 3.24*** 0.24 (0.11) 2.16*** 3.18*** 3.44 (2.17) 1.58 0.51 0.08 (0.13) 0.67 0.81 0.0023 (0.0033) 0.73 1.01 0.08 0.30

0.20 (0.07)

2.91***

0.22 (0.05)

4.35***

0.00 0.08 0.10 0.13

(0.00) (0.06) (0.09) (0.10)

0.37 (0.13)

0.79 1.26 1.16 1.33 2.80***

Notes: *** Statistically significant at the 99 per cent level. Number of observations 543.

since it provides consistent estimates of the parameters. Estimation by twostage least squares (2SLS) is performed to account for any endogeneity of the growth centre dummy. Both of the estimations show that the literacy rate, and the proportion of employment in services, are statistically significant in explaining the rate of unemployment. As we should expect, the coefficient on literacy rate is negative. Higher literacy implies that the area’s workforce is more marketable and increases their employability. Specifically, for every one percentage point increase in the literacy rate of district, there is a small, but significant 0.11 (0.24) percentage point decrease in its unemployment rate in the OLS (2SLS) model. The proportion of employment in service occupations has a negative impact on the unemployment rate, showing that as the proportion dependent on these occupations increases, unemployment is likely to decrease to the extent of 0.22 percentage points (OLS model) or greater (0.37 percentage points, 2SLS model). This testifies to the dependence on services for employment. The sign on proportion male and proportion employed in manufacturing have significant impacts on the unemployment rate in the OLS estimation (both have negative impacts, as we expect), but lose their significance in the 2SLS version.

132 Incentives for Regional Development

The GC dummy, in contrast to the tax incentive dummy in Chapter 4, does not have a statistically significant impact in reducing the unemployment rate of districts that contain them in either estimation.28 It is possible that institutional rigidities are far too strong in these backward regions and that, therefore, any positive effect of the GCs is not strong enough to counter the institutional problems, or the inherent disadvantages encountered in the regions. Further, note that the GC is in place in only 13 per cent of the districts (compared to the existence of tax incentives in nearly 80 per cent of Ohio’s census block groups). Even where they are in place, the GCs were completely operational only in some of the 68 GCs. For instance, in the north Indian state, UP, although GCs were officially in place in six districts, they were functioning in only two of them (Gorakhpur and Jaunpur). The other proposed GCs have been delayed because of land acquisition and litigation problems over the level of compensation that is to be paid to landowners. Overall, taking into account districts in all of the states, functioning GCs may be said to have had little effect on the unemployment rate to date. Next, based on my field visits, I found that many areas would have grown the way they have even without the presence of a GC.29 For instance, Gorakhpur in the northern Indian state, Uttar Pradesh (UP), was an industrial area, with many firms having located there long before it became a GC. Similarly, Hassan GC, in the south Indian state, Karnataka, is located by the side of a state highway offering good access to firms locating there. These are areas with infrastructure already in place, much needed for firms to grow, and firm locations may not be necessarily attributed directly to the existence of the GC. Few areas with GC (take the instance of Satharia (in Jaunpur district, in the north Indian state, UP)), however, would not have grown the way in which they have without the GC. Satharia is in a remote location, about 47 kilometres from the Jaunpur district headquarters (is this ‘close’ to district headquarters? See criterion II for GC designation, section 6.3). Road access to Satharia is quite poor (winding roads in the midst of thick vegetation), which industries find quite disadvantageous, in terms of transport costs, time and logistics. It is reasonable to believe that only the establishment of the GC and the provision of infrastructure (power, telecom, paved roads) in this area has been able to attract industry. This discussion implies that designation criteria for GCs have to be distressbased. I shall have more to say about designation criteria in the section on policy implications. Finally, the insignificant impact of the GC dummy on the rate of unemployment could be the outcome of the large area that a district covers. On average, the land area (see Table 6.2) of a district containing a growth centre is 6,431 square kilometres. By contrast, the average land area of a growth centre is only four square kilometres, or only 0.06 per cent of the land area of the district!30

India’s Growth Centres and Unemployment 133

If data on the unemployment rate were available at a more disaggregated level, say block level, it is possible that we may find that GCs have a significant impact on the unemployment rate, as we found in Chapter 4. Even if the relevant block-level data were available, one question that remains is whether we can expect unemployment rates to vary substantially enough across blocks to enable estimation. In any case, depending upon data availability, estimating unemployment rate at India’s census block level is a potential extension to this work. I examined the correlation matrix of explanatory variables and found that no correlations were strong enough to suspect collinearity. Further, I performed formal tests of heteroscedasticity in the data, and found that it was small enough to be ignored.31 To see if GCs had separate effects on the unemployment rate across low- and high-unemployment areas, I performed OLS estimations of the sample separately for high- and low-unemployment areas.32 For example, if GCs were more successful in reducing the unemployment rate of high-unemployment areas, then they could be causing some hysteresis effects in those labour markets. When applied to the labour market, this means that a one-time job shock and the increase in the labour force participation rate as the result of increased employment increases households’ employability (due to training and acquisition of skills) in the long run. Such effects could be more pronounced in high-unemployment areas than in their low-unemployment counterparts. Separate estimations for high- and low-unemployment areas are summarized in Table 6.4. In the case of low-unemployment areas, all of the factors that have significant impacts are those that reduce their unemployment rate. These are the higher age of labour force, implying greater levels of work experience, a lower proportion of SC and STs, and better manufacturing and service bases. In high-unemployment areas, social factors such as literacy rate, and the proportion of men in the population, reduce the level of unemployment. While older people in the labour force have difficulty in finding employment in higherunemployment areas, probably owing to their poorer level of skills, they do not have difficulty finding jobs in low-unemployment areas. The GC dummy does not have any impact on the unemployment rate in either of the models. Overall, the model is a much better explanation of changes occurring in the labour force in low-unemployment than it is of changes in highunemployment areas.

6.8

Estimation for GCs

In addition to examining the effect of infrastructure incentives (GC) on the unemployment rate, I assess the performance of GCs, where they exist in the country. Here I study whether infrastructure affects firm location decisions, where GCs exist. Secondary data were available at the GC level from DIPP, Government of India. This database contained information on the date of

134 Incentives for Regional Development Table 6.4 OLS estimation of unemployment for low- and high-unemployment areas (dependent variable: unemployment rate) Variable

Constant Literacy rate, 2001 Proportion male, 2001 GC dummy, 2001 Duration of GC (months), 2001 Duration squared Average age, 1991 Proportion SC and ST, 1991 Proportion employment, manufacturing, 1991 Proportion employment, services, 1991 Adjusted R2 F Number of observations

High-unemployment areasa

Low-unemployment areasa

Coefficient (standard error)

Coefficient (standard error)

t

t

0.34 (0.27) 1.27 0.96 (0.17) 5.57*** 0.11 (0.04) 3.16*** 0.04 (0.02) 1.57 0.93 (0.24) 3.85*** 0.15 (0.16) 0.95 0.05 (0.10) 0.54 0.04 (0.07) 0.59 0.0013 (0.0026) 0.51 0.0017 (0.0020) 0.88 0.00 (0.00) 0.01 (0.005) 0.02 (0.02)

0.59 0.00 (0.00) 2.57*** 0.02 (0.003) 1.30 0.05 (0.01)

0.32 5.31*** 3.83***

0.07 (0.09)

0.75

0.20 (0.04)

4.81***

0.16 (0.03)

5.30***

0.16 (0.08) 0.14 5.40 244

1.90*

0.26 12.65 299

Note: a. High-unemployment areas are defined as those with greater than or equal to 24 per cent unemployment rate (this is the average for all Indian districts, 2001). Low-unemployment areas are defined as those with less than 24 per cent unemployment rate.

*** Statistically significant at 1 per cent level. * Statistically significant at 10 per cent level.

their approval, the approved project cost, the amounts of central and state releases,33 the final total expenditure, the land acquired, the number of plots developed and allocated, the number of firms established, the capital invested and the employment created by them. Table 6.5 details this data (obtained from DIPP) for the 68 GCs in the country as of 2001. The table shows that, on average, approximately 38 per cent of the total expenditure on the GCs is leveraged by funds from the central government, the total expenditure being less than the approved project cost in all cases, as one would expect. The average size of a plot on which an industrial unit sits in the GCs is roughly three acres, based on the land acquisition and developed plots data given in Table 6.5. On average, the number of plots allocated to firms (52) lags significantly behind the number developed (about 198), implying that the GCs still need to market themselves to businesses as good places in which to locate, invest and create jobs.34 On average, the number of units (firms) established in these GCs is even lower, averaging around 12. These numbers for cumulative firm location do seem disappointingly low. One strongly suspects that the programme has

India’s Growth Centres and Unemployment 135 Table 6.5 Description of data for GCs Variable

Approved project cost (US$)** Central release (US$) State release (US$) Total expenditure (US$) Land acquired (in acres)* Plots and sheds developed Plots and sheds allotted Number of units established Capital invested by units (US$) Employment Regional dummy

Average

Maximum

Minimum

Standard deviation

4,066,168.09

13,203,633.64

1,770,778.61

1,677,730.10

568,927.56 1,079,018.04

1,239,545.03 8,673,167.39

59,025.95 0.00

434,749.76 1,699,947.47

1,482,035.45

9,853,686.47

0.00

2,022,677.77

659.10

3,060.41

0.00

722.92

198.47

2,205.00

0.00

384.38

51.94

362.00

0.00

91.00

12.43

231.00

0.00

33.35

13,916,844.24 390 0.49

513,879,953.02 7191 1

0.00 0 0

64,770,111.62 1,189.74 0.5

Notes: Number of observations 68. * For this variable only the descriptive statistics are based on 63 observations since at the time I obtained the data, DIPP had not received reports regarding acquired land from five of the states. ** All monetary data in INR (Indian rupees) have been converted to US$ by using the exchange rate US$1 INR 44.35, which was the exchange rate reported by the Reserve Bank of India, on 13 December 2004. The 2004 US$ are converted to constant US$ (1982–4 100) by a mechanism described in endnote 35.

not been marketed appropriately to prospective businesses. Those that have marketed themselves (as in the case of Bawal GC, in Haryana) have been successful in attracting a large number of national and multinational firms. Further, only one-third of the GCs have secured the location of firms so far. Others are still at the stage of land acquisition and development, or are currently in the process of allocating them to firms. If we take into account only the GCs in which firm establishment activity had already taken place, on average, about 40 firms have located, with an average capital investment of $13.92 million.35 The average employment creation per GC (390) shows that the majority of firms to have located here are labour-intensive. Note that when we work on the averages, the average cost per job based on actual expenditures in the growth centres of India turns out to be $3,800. This is the same as that for Ohio’s zones in scenario one (Table 5.6) we find in Chapter 5. The cost per job in India’s growth centres is, in fact, lower than the weighted average cost per job of firms in most scenarios, incurred in the EZs of Ohio (see Tables 5.5–5.11).36

136 Incentives for Regional Development

To assess the impact of GCs where they exist, I estimate the number of firms as dependent upon the presence of GCs. I use the number of firms locating in the GC as a measure of their performance. This is reasonable since GCs were set up to promote the industrialization of backward areas in the country, and the assumption behind my methodology is that the greater the number of firms, the greater the extent of industrialization of the area. I estimate the number of firms locating in GCs as being dependent upon the number of plots developed and a dummy for the region in which the GC is located. I confirmed (from GC officials) that the number of developed sheds or plots represents the most important quantitative indicator of the presence of GCs. This is a reasonable assumption because developed plots do not represent the actual number of industrial units, but rather represent what GCs promise to industry – infrastructure and public services to enable their location. The infrastructure typically includes roads, electricity, telephone lines, and water and sewerage connections. This means that I could empirically validate whether firm location decisions are dependent upon the infrastructure offered by GCs. If the sign on the number of developed plots were to be statistically insignificant, it may be concluded that infrastructure is not a decisive factor in firm location decisions.37 In addition to the presence of GCs, I develop a regional dummy, based on the state in which the GCs are located. Holding all else constant, casual observation suggests that, irrespective of tax incentives or public services (which GCs provide), the most likely reason for firms to locate in prosperous areas is because of their presumed favourable business climate. This business climate could be a composite index consisting of the following characteristics: (i) The skills and work ethic of the labour force in the area; (ii) Local area income (indicative of demand conditions); (iii) Political factors, including law and order conditions and political leadership; (iv) Social factors such as communal harmony; and (v) Amenities such as temperate weather conditions (these affect the cost of doing business directly through fuel costs), schooling, and recreational facilities. In order to make a regional dummy for every GC, I assume that the abovementioned characteristics are most closely related to income. In the Indian context, this assumption is appropriate. For instance, some of the higherincome Indian states (such as Karnataka, Tamilnadu, Maharashtra, Punjab and Haryana) are also those that are better governed, have a more skilled labour force, and have a more developed infrastructure. For the purposes of making the regional dummy, I compute weighted rural income (the weights being the number of households in every income group),

India’s Growth Centres and Unemployment 137

based on data published by the National Council of Applied Economic Research, for all states. Whenever this weighted rural income for the state containing the growth centre is higher than the all-India weighted rural income, the regional dummy receives a value of one, implying that the GC is in a relatively prosperous area. When the weighted rural income for the state containing the GC is below the national average, the regional dummy receives a value of zero. The average value of the regional dummy in Table 6.5 shows that roughly half of the GCs are located in relatively prosperous states with higher rural incomes (in relation to the national average). Table 6.6 presents the results from this OLS estimation of the effectiveness of GCs. The value of adjusted R2 indicates that this model explains more than 60 per cent of firm locations in GCs. It shows that when controlled for the state in which the GC is located, the coefficient on the number of developed plots being positive, is a highly significant determinant of the decision of firms to locate in the GC. This implies that, among states having GCs, those having the financial resources to develop larger number of plots with infrastructure are the ones that would be successful in attracting firms. These results are consistent with the results from the qualitative discussions I had with several GC firms I visited in various GCs of the country. Chapter 7 describes these visits and the qualitative findings. In general, the prosperous states attract a larger number of firms, as may be seen in the sign of the coefficient on the regional dummy, but the effect is not found to be statistically significant. When combined, the results from estimation of the unemployment rate and that for the GCs show that, in contrast to tax incentives, infrastructure does not have a significant impact on the unemployment rate recorded in areas. However, where the GCs exist, infrastructure has a positive influence on firm location. One possible reason for these findings could be that infrastructure does influence firms to locate, but these firms may not always be the sort of labour-intensive firms that will significantly affect the unemployment rate in an area. Further, the GCs are Table 6.6 Estimation of GC performance (dependent variable: number of units established) Variable

Unstandardized coefficients

Constant Regional dummy Plots and sheds developed 2

Adjusted R F

Standard error

t

3.53 5.14

0.96 1.00

0.01

9.97

3.38 5.13 0.07*** 0.62 56.29

Note: Number of observations 68.

*** Statistically significant at 1 per cent level.

138 Incentives for Regional Development

not geographically extensive, especially in the distressed areas, and therefore constitute too small a proportion of total economic activity occurring in the districts to make a significant difference. Given the poor state of infrastructure, it is not a bad idea for states to invest in this in order to market the area as a better place to do business. This is a valid conclusion to draw even if one is not sure about the effectiveness of GCs. GC is a name we may give to the targeted development of infrastructure in the distressed pockets of regions.

6.9

Concluding remarks

With the designation of the most distressed areas, growth centres can become a tool to enhance their infrastructure competitiveness and locational advantages to firms, who would not otherwise locate in these distressed areas. With an increasing number of firm locations in the area, agglomeration economies occur, providing incentives to more firms (of a specific type) to locate, and enhance the economic base of the area, and thereby continually providing job opportunities to residents. Thus, over a period of time, GCs are likely to enjoy all of the benefits of development by catching up with other areas. When this is achieved and the area’s distress criteria indicate that it is no longer a high-unemployment or industrially backward area, its GC status should be re-examined. Thus, it is important to specify the time limit for which the areas would be designated as GCs. In the absence of a time limit, areas and industries could lobby for the continuation of the ‘backward’ status and incentives forever, not different from the lobby of small-scale industries in India that have been protected for over three decades now. The implications of this work for geographically targeted programmes that aim at convergence are for a time-bound programme which will sunset at the expiry of the period and is performance-based during the period that it is in existence. It should be performance-based both for the state and local government administering the programme and for the firms that make the commitments. In the next chapter, I will elaborate on the firm-level impacts, their implications, aggregate, qualitative GC-level discussions and their implications for specific GC designation criteria.

7 Firm Location Decisions and Their Impact on Local Economies: Evidence from India’s Growth Centres

7.1

Introduction and motivation

One theme that recurs throughout this book is the consideration of whether tax incentives have some impact on firm location decisions, when compared to the impact of infrastructure incentives. As we have discussed in earlier chapters, the GC approach is a test of the alternative to the tax war, and competition in providing infrastructure and public services, among the states. As stated earlier, there is considerable scepticism in policy circles regarding India’s GCs and their effectiveness in attracting firms. Commentators’ views have varied from a perception of the programme having been a colossal failure in attracting firms to one that strongly believes in their effectiveness because of the infrastructure incentives available to firms that locate there. Chapter 6 presented empirical evidence, based on secondary data, regarding the effect of GCs on the unemployment rate and their impact on firm location, where they exist in India. In this chapter, I attempt a qualitative assessment of the impact of GCs, based on primary data collected from field visits to several GCs, discussions with state governments, and visits to agencies administering the programme and to several firms located in the GCs. In this way, I account for factors and observations that might be missed in a purely quantitative approach which has been the focus of the previous chapters.

7.2

Objectives

Given the challenges that surround GCs, and, more generally, the challenges raised regarding the contribution of firms to communities where they locate, the objective of the research in this chapter is to qualitatively 139

140 Incentives for Regional Development

examine the following: 1. The impact of GCs (tax incentives vis-à-vis infrastructure services) on firm location decisions; 2. The effect of firms (located in the GCs) on local labour markets; 3. Firms’ export orientation; and 4. Firms’ social contribution to the local communities where they locate, to justify state or local spending on the programme. For various understandable reasons, the data required to answer these questions are not routinely collected by DIPP of the Ministry of Commerce, Government of India, that administers the programme. Where secondary data on GCs are available from DIPP, they do not provide enough information to answer the questions listed above, making it necessary to obtain primary information. To answer these questions qualitatively, I obtained primary data regarding infrastructure in GCs (to examine if they deserve to be designated as GCs) and had discussions with officials implementing the programme regarding their effectiveness and continuation. I also visited firms to explore the impact of GCs. In this chapter I describe my findings based on my visits to GCs in various parts of the country. This chapter is organized as follows. In the next section, I report the field visits I made to GCs located in various parts of the country to examine their distress and infrastructure, and to determine implications for their GC status. After this, I summarize visits to firms and GCs, and the policy implications of the work at the firm level and the GC level, their implementation, and performance appraisal. The final section contains some concluding remarks.

7.3

Field visits to GCs

As of 2001, there were 68 GCs spread across various parts of the country, although, as mentioned earlier, not all of them were fully functional. While the field visits were determined by funding constraints, I visited a sufficiently representative sample of GCs – a GC in the southern part of India (Hassan, Karnataka), one in the north – Haryana (Bawal), and both of the GCs in the northern state, UP, that are functioning and where firms have located.1 The location of these GCs is geographically dispersed in order to be a representative sample. I visited firms in each of the selected GCs (wherever possible and relevant, I visited firms in both the pre-GC and the post-GC period). I asked firms about their location decision, the characteristics of the labour force they had recruited (wage levels, whether or not the employees were local residents), the firms’ export profile and any community activity in which

India’s Growth Centres: The Impact of Firms 141

they were currently involved. I will discuss these effects later. First, I describe the primary data obtained from GCs, their infrastructure and the implications for their status. Then I explain the contribution of firms to local labour markets, their export orientation and their social activities in the local communities. I recorded the cost of land acquisition in all of the GCs visited, as this is a determinant of cost for the GCs, which eventually determines how many industries will locate there. Where land compensation is high and the corresponding land development costs are high, it means that the opportunity costs of industrial development are also higher. Land displacement and compensation to poor farmers are also frequent sources of local agitation. 7.3.1

GCs in Karnataka: Hassan GC

I visited the Hassan GC in May 2001, in order to obtain primary information about the effectiveness of that GC and to obtain the views of the firms that have located there. In Karnataka, the GC programme is administered by the Karnataka Industrial Areas Development Board (KIADB). At the time of the visit, the Hassan GC, the largest GC that I visited, covered an area of 1,662 acres, of which 652 acres were to be developed in the first phase, and the remainder in the second phase. At the time, 514 acres had been allotted to various industries to locate. Based on my discussions with KIADB officials, the average size of plot per firm was 7.8 acres, with the cost per acre being as high as $4,132 (Table 7.3).2 Observations of the area’s power, water (for industrial use), telecom and transport networks indicated that the promise of infrastructure provision had been met in this GC.3 I probed into distress criteria for the establishment of GCs in Karnataka. Designation of the three GCs in Karnataka (Raichur, Dharwar and Hassan) is curious if we take into account the fact that the state contains many more industrially backward and high-unemployment districts such as Kodagu and Koppal.4 Table 7.1 contains the distribution of industrial units (large, medium and small units) and employment by district for Karnataka, which I obtained from the District Industries Centre, Karnataka. Bangalore (Urban) dominates both, accounting for nearly 20 per cent of all firms located and 45 per cent of all employment created in the state.5 Figure 7.1 shows a district map of Karnataka. My discussions with KIADB, together with Table 7.1, show that, although KIADB has industrial areas in almost every district of the state, there are a number of areas which are industrially backward, and which therefore deserve GC status. If the objectives of establishing the GCs are to promote the industrialization of backward areas in the country, this is not being effectively achieved with existing criteria. I have more to say on this in the final section of this chapter.

142 Incentives for Regional Development Table 7.1 Registered industrial units and total employment in Karnataka’s districts, 1999–2000 Number

District

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

Kodagu Koppal Haveri Chamarajanagar Chickmagalur Bagalkote Chitradurga Uttar Kannad Gadag Udupi Davanagere Bijapur Mandya Bidar Hassan Gulbarga Raichur Shimoga Bellary Kolar Dakshin Kannad Dharwad Bangalore (R)** Tumkur Mysore Belgaum Bangalore (U)** Total

Number of firms* 161 165 200 241 259 312 319 339 342 361 384 397 437 467 468 486 492 513 533 643 803 913 966 1,029 1,218 1,241 3,247 16,936

% of total

Total employment*

% of total

0.95 0.97 1.18 1.42 1.53 1.84 1.88 2.00 2.02 2.13 2.27 2.34 2.58 2.76 2.76 2.87 2.91 3.03 3.15 3.80 4.74 5.39 5.70 6.08 7.19 7.33 19.17

1,187 769 752 1,478 4,627 1,430 12,978 7,264 1,337 3,200 1,299 8,539 2,501 6,854 4,568 19,400 12,707 13,260 17,341 30,603 9,907 30,493 21,511 11,257 38,436 27,769 239,899

0.22 0.14 0.14 0.28 0.87 0.27 2.44 1.37 0.25 0.60 0.24 1.61 0.47 1.29 0.86 3.65 2.39 2.50 3.26 5.76 1.86 5.74 4.05 2.12 7.23 5.23 45.15

100

531,366

100

Notes: * Includes small, medium and large industrial firms. ** Bangalore (R) refers to Bangalore (Rural) and Bangalore (U) refers to Bangalore (Urban). Source: Karnataka District Industries Centre.

7.3.2

GCs in Haryana: Bawal GC

In September 2001, I visited the Bawal GC (in Rewari district) with a view to understanding the designation process and the functioning of GCs in the northern Indian state of Haryana. GCs in Haryana are administered by the Haryana State Industrial Development Corporation (HSIDC). As of 2001, Bawal was the only GC designated in that state, although another was being proposed in the northern part of the state, in Ambala. Figure 7.2 shows a district map of Haryana, from the Census of India.

India’s Growth Centres: The Impact of Firms 143

KARNATAKA 2001

Bidar

Gulbarga Bijapur

Raichur

Bagalkot

Belgaum

Koppal Dharwad Gadag Bellary Uttara Kannada

Haveri

T-Tumkur

Davanagere Shimoga Udupi

T

Chitradurga ur

gal

a ikm

Tumkur

Kolar

Ch

Bangalore

Hassan Dakshina Kannada Kodagu

Mandya Bangalore Rural Mysore Chamarajanagar

Figure 7.1 District map of Karnataka, India Source: Census of India website, http://www.censusindia.net/

According to HSIDC officials, Rewari has only a small industrial base at present, with nearly 80 per cent of the employment being agricultural. According to Census 2001 data, this is one of the highest unemployment districts in the state (with a 39 per cent unemployment rate). This matches the development profile of a backward area that has the potential to develop, due to the existence of its large agricultural base. Recall from the development literature that growth is the process by which, with increasing agricultural productivity, labour is transferred from agriculture to other sectors. Further, note that agriculture provides many raw materials for industries – paper, beverages, rubber, textiles, and so on. By contrast, Ambala has one of the lowest unemployment rates in the state, roughly 15 per cent (which is, of course, only one measure of distress

144 Incentives for Regional Development

Panchkula HARYANA 2001

Ambala

Yamunanagar

Kurukshetra Kaithal Sirsa

Karnal

Fatehabad Panipat

Jind Hisar

Sonipat Rohtak

Bhiwani

Jhajjar

Gurgaon Mahendragarh

Rewari Faridabad

Figure 7.2 District map of Haryana, India Source: Census of India website, http://www.censusindia.net/

in an area), according to calculations from the 2001 Census of India. A district (located adjacent and west of Rewari, see Figure 7.2) (Mahendragarh) is the one that reported the highest unemployment, based on Census 2001 data (roughly 42 per cent, followed by Rewari with 39 per cent unemployment). It is not clear why Ambala, with its comparatively low unemployment, rather than the high unemployment Mahendragarh, was being designated as a GC. One reason could be ‘potential to develop’ which is a rather unclear assessment. Like Hassan GC in Karnataka, the Bawal GC area is located by the side of national highway eight (the Delhi–Jaipur highway). A well-established system of infrastructure including roads, sewerage, water supply and power is

India’s Growth Centres: The Impact of Firms 145

already in place. Further, better telecommunication and internet facilities, banking and educational services, including the establishment of technical institutions, were actively being considered. The Bawal GC, covering an area of 1,209 acres, was established in 1994, and various activities in the centre, including firm locations, began in 1997. Based on the discussions I held with HSIDC officials, the average cost of land acquisition is in the range $2,361–4,722 (in real terms) per acre, and the size of an industrial plot varies from 6.25 per cent of an acre to about 10 acres. Table 7.3 summarizes the cost of land data for all GCs visited. I will discuss this table in detail when I have summarized all of my visits to GCs. At the time of my visit, the HSIDC had not been offering tax incentives to industry in order to persuade them to locate in the GC.6 Usually, it was possible for the firm (if manufacturing) to start production roughly six months from the date of signing the letter of contract with the GC, since all infrastructure would be guaranteed to be in place by the end of this time period. The HSIDC thus appeared to depend upon the offer of non-financial incentives to firms to persuade them to locate there. Every one of the firms that had located in the GC was a manufacturing firm. One interesting feature is that the majority of these firms are national and multinational, although some of them were the result of local entrepreneurship. Due to the advantages that proximity to the National Capital Region (NCR) provides in the form of markets and inputs, the GC, in 2001, was able to attract investment by roughly 7 national and MNC firms (that increased to 23 in 2004), creating total employment around 500 (that increased to 900 in 2004). Further, as of 2001 the firms had made capital investments of about $85 million (in constant terms, with 1982–4 100) in the GC.7 Incidentally, a couple of firms had located in the area before the GC came into existence, which provided a test of the counterfactual and some responses as to why they located there (see Table 7.4). While they located there due to the influence of factors such as proximity to raw materials and highway, they represented local entrepreneurship that had subsequently benefited from the existence of GC infrastructure. As in the case of the other GCs, I visited several firms in the Bawal GC. I summarize the detailed responses of these firms to various questions in Tables 7.4–7.7. 7.3.3

GCs in Uttar Pradesh

There are seven GCs designated in the large northern state, Uttar Pradesh (UP), although only two were functioning as of August 2001 when the field visits were undertaken – Satharia (Jaunpur) and Shajanwa (Gorakhpur).8 The remaining five GCs were in the land acquisition stage at the time of fieldwork, when I visited. Among these, two were constantly in dispute because of land acquisition problems, and the location of the GCs had changed

146 Incentives for Regional Development

several times because of problems with compensation payable to landowners. One (in Khurja (Bulandshahar)) was still under litigation. Figure 7.3 shows a map of the districts of Uttar Pradesh. The UP State Industrial Development Corporation (UPSIDC), on paper, administers the currently non-functional GCs (Bijauli-Buzurg (Jhansi), Jamaur (Shahjahanpur), Pakbara (Moradabad), Dibiapur (Auraiya), and Khurja (Bulandshahar)).9 The Shajanwa GC (Gorakhpur) is administered by Gorakhpur Industrial Development Authority (GIDA) and Satharia (Jaunpur) GC, by SIDA (Satharia Industrial Development Authority). The two functional GCs in this north Indian state, Satharia and Shajanwa, were set up around 1989–90, although firm establishment activity only started about 1993. Compared to the distribution of industrialization in Karnataka, industrialization in UP is more uneven. Table 7.2 summarizes the

UTTAR PRADESH 2001

Saharanpur Muzaffarnagar

Bijnor

Baghpat Meerut JPN Ghaziabad

Rampur

Kheri

Budaun

Aligarh

Pilibhit

Bareilly

GBN Bulandshahr

Mathura

AN - Ambedkar Nagar GBN - Gautam Buddha Nagar JPN - Jyotiba Phule Nagar KN - Kanpur Nagar SRN - Sant Ravidas Nagar SKN - Sant Kabir Nagar

Moradabad

Etah Hathras Agra

Bahraich

Shahjahanpur

Farrukhabad Mainpuri Hardoi Firozabad Kannauj Etawah Auraiya

Sitapur

Maharajganj

Siddharthnagar

Gonda Barabanki Basti Unnao Lucknow Faizabad

Kanpur Dehat KN

Sultanpur

AN

Kushinagar Gorakhpur SKN

Deoria

Mau Azamgarh Ballia Pratapgarh Ghazipur Jaunpur Hamirpur Banda Kaushambi Mahoba SRN Varanasi Chitrakoot Allahabad Chandauli Mirzapur

Jalaun Jhansi

Shrawasti Balrampur

Rae Bareli Fatehpur

Lalitpur Sonbhadra

Figure 7.3 District map of Uttar Pradesh, India Source: Census of India website, http://www.censusindia.net/

India’s Growth Centres: The Impact of Firms 147

proportion of employment in manufacturing and services in the districts of UP. When we study Table 7.2, observe that, based on the average proportion of employment in manufacturing and services for the state, the designation of GCs in Dibiapur (Auraiya), Satharia (Jaunpur), are in areas without industrial bases, quite rightly, as this book has argued. As Table 7.2 shows, in 1991 the proportion of employment in manufacturing and services in these districts was far below the state’s average proportion of employment in manufacturing and services. Table 7.2 Summary of manufacturing and services employment, UP districts District

Agra Aligarh Allahabad Ambedkar Nagar Auraiya Azamgarh Baghpat Bahraich Ballia Balrampur Banda Barabanki Bareilly Basti Bijnor Budaun Bulandshahar Chandauli Chitrakoot Deoria Etah Etawah Faizabad Farrukhabad Fatehpur Firozabad Gautam Buddha Nagar Ghaziabad Ghazipur Gonda Gorakhpur Hamirpur

Employment in manufacturing (%)

Employment in services (%)

17.99 11.33 8.12 6.20 4.48 7.01 15.20 2.51 3.74 2.32 3.14 5.10 6.10 4.15 11.43 2.24 8.52 24.52 3.14 3.98 4.58 4.48 5.40 7.03 3.42 19.78 14.43 20.35 4.71 2.32 4.89 3.80

31.83 22.06 20.84 13.48 17.60 12.34 27.41 8.41 15.94 8.77 9.57 10.47 23.73 9.60 18.55 11.81 21.53 21.84 9.57 12.22 14.89 17.60 14.62 15.98 13.87 19.72 28.67 35.80 15.15 8.77 22.29 11.65 Continued

148 Table 7.2

Continued

District

Hardoi Hathras Jalaun Jaunpur Jhansi Jyotiba Phule Nagar Kannauj Kanpur Dehat Kanpur Nagar Kaushambi Kheri Kushinagar Lalitpur Lucknow Maharajganj Mahoba Mainpuri Mathura Mau Meerut Mirzapur Moradabad Muzaffarnagar Pilibhit Pratapgarh Rae Bareli Rampur Saharanpur Sant Kabir Nagar Sant Ravidas Nagar Shahjahanpur Shravasti Siddharthnagar Sitapur Sonbhadra Sultanpur Unnao Varanasi Average

Employment in manufacturing (%)

Employment in services (%)

2.63 10.02 3.61 7.16 9.23 11.57 7.03 3.83 24.22 8.12 2.07 3.98 3.39 10.62 1.54 3.80 3.30 8.71 16.38 15.20 16.81 11.57 10.21 5.01 4.20 4.89 9.15 10.42 2.89 24.52 4.23 2.51 1.63 4.08 7.07 4.06 4.82 24.52

11.27 23.32 16.16 15.10 25.40 19.09 15.98 11.88 56.68 20.84 10.09 12.22 11.82 45.97 9.18 11.65 16.08 24.59 13.85 27.41 14.22 19.09 19.11 13.77 12.12 12.69 16.42 22.56 8.50 21.84 15.08 8.41 7.41 11.20 12.63 10.92 12.19 21.84

7.93

17.22

Source: Calculations from 1991 Census of India data.

India’s Growth Centres: The Impact of Firms 149

In the cases of all of the other growth centres designated in this state, the districts either had a stronger manufacturing or service base for employment, when compared to the state average, as is evident from Table 7.2. Some of the existing and proposed GCs (in Pakbara (Moradabad), Shajanwa (Gorakhpur), Khurja (Bulandshahr), and Jamuar (Shahjahanpur)), are overlaid on top of areas already industrialized, whereas many unindustrialized areas (for instance, Ambedkar Nagar, Azamgarh, Bahraich, Balrampur, Banda, to mention only a few) continue to be neglected. Based on my discussions with officials, I find that the criteria on which these areas are set up as GCs relate to backwardness and their development potential (proximity/access to transport, and the availability of raw materials). According to them, similar to what was observed in the other states, no explicit distress criteria were used for GC designation. Approximately 500–900 acres of land per GC were allocated for the development of infrastructure in the state, with the average size of a plot (on which a firm sits) in the state being about 0.1 acre (see Table 7.3). Discussions revealed that the average compensation paid for land acquisition by the state was about INR 300,000 (or US $3,540, in real 1982–4 terms) per acre (lower than that in the south Indian state Karnataka), with the amount being higher in the western part of the state, where the land is more fertile.10 In sharp contrast to the relatively prosperous Haryana, despite the financial situation in UP (where revenue expenditure traditionally constitutes up to 80 per cent of total expenditure), sales tax incentives were being offered in UP in 2001. These were most likely those that were committed in the pre-1999 period. At the time that this research was being conducted, the highest tax concessions in UP were available to firms in category A districts, which are the ‘No Industry Districts’. Firms in C districts (the industrially advanced) were eligible for the lowest level of concessions.11 Various firms in UP were given tax incentives according to the type of district in which they are located, which is equitable, based on what has been argued in earlier chapters. These tax incentives, being incentives to capital, would theoretically exert all of the effects discussed in the analytical framework of Chapter 2. However, the offer of tax incentives might be detrimental to the long-run economy of the state, given that its social indicators are some of the lowest in the country. Given the finding, based on my discussions, that 70 per cent of investment in UP is concentrated in just a few districts, I explored the options for the remaining areas. Based on my discussions with UPSIDC officials, I deduced that traditional factors, such as the availability of raw material, the proximity and size of market and transportation costs, are important factors in influencing firm location, consistent with firm location theory (see Table 7.4). Whenever these factors are not favourable, tax or infrastructure incentives are necessary to persuade firms to locate. This is shown clearly in the case of Satharia GC, where access to transportation is poor and no raw materials

150 Incentives for Regional Development

are available. Only because of the GC incentives has the Satharia area developed into what it has become today. In fact, firm location in Satharia points to evidence regarding firm births rather than firm relocation, because of the availability of infrastructure. In contrast, the Shajanwa GC is close to the highway, as shown by the manufacturing and services base of the district (Gorakhpur) in which it is located (Table 7.2), and does not have the same accessibility problems as Satharia. In the wake of the abolition of tax incentives for firms, in the post-1999 period, there are two options for the development of the backward areas of the state: 1. Adoption of a service-sector-based approach to industrialization, which did not depend on tax incentives, but on customer base and infrastructure. GCs can become powerful engines of the growth of small and mediumsized towns containing such infrastructure. 2. A consumer-goods-based approach, which does not require the factors (the availability of raw materials, the size of markets) required for heavy manufacturing location. Examples are of consumer products such as bread and confectionery items. GCs in UP: Shajanwa (Gorakhpur) GC In March 2002, I visited the Shajanwa GC in the Gorakhpur district of UP, to obtain primary information regarding its effectiveness. The Shajanwa GC is administered by the Gorakhpur Industrial Development Authority (GIDA). The Shajanwa GC, covering an area of 800 acres (600 acres industrial and the remaining residential), was established in 1993, and various activities in the centre, including firm locations, started around the same time. Quite similar to the GCs in the other states, the GC area is located by the side of national highway 28. A well-established system of infrastructure, including roads, sewerage, water supply and power, is in place. Based on my discussions with GIDA officials, Gorakhpur was an industrialized area even before the GC came into existence. This was confirmed by the fact that the GC in Shajanwa was designated in 1993, whereas, even as early as 1991, the district’s service base was in fact greater than the state’s average (Table 7.2)! Shajanwa is one of the sites in the state that started from plain agricultural fields. Of the 521 acres that had been acquired by GIDA, 290 acres had been developed with the necessary infrastructure.12 Because of frequent local agitations against the compensation paid, it had proved impossible to acquire any more land. The average cost of land acquisition was in the range of INR 350,000 (or $4,132, in constant terms) per acre, the same rate as in Hassan GC. The size of an industrial plot varied from about 6.25 per cent of an acre to about three acres (the lower range being similar to Bawal GC, Haryana). Note that the costs of developing land everywhere include the construction of roads, drainage and culverts, electricity in poles, wiring networks and

India’s Growth Centres: The Impact of Firms 151

water management (drainage).13 According to discussions, developed plots/sheds are the most important quantitative indicator that represents the presence of the GC in the area since that represents infrastructure, without which most of the firms would not have located there.14 Recall that the empirical work in Chapter 6 also supports this observation. Consistent with what was decided by all of the Indian states’ chief ministers in November 1999, GIDA had stopped offering tax incentives to industry locating in the GC. Those that located prior to 1999, however, secured a maximum sales tax incentive of 250 per cent of their fixed capital investment (which is, again, an incentive to capital). This incentive was unavailable to firms locating in the GC in 2002, which creates an uneven playing field, since some firms continued to enjoy incentives and others did not. The view of the GC officials was that firms located there only because of the infrastructure that is available in the GC area that would be unavailable elsewhere. As evidence of this point, 50 plots had been allocated to various industrial units in that financial year alone. Designation of GCs is a political process in UP as in the other Indian states I visited, and is not based on any objective distress criteria such as levels of unemployment or poverty. Although the Gorakhpur area has traditionally been a low-unemployment one, its 2001 unemployment rate stood at 39 per cent, being well above the national average of 24 per cent. With this, the area had recently experienced out-migration to urban areas in the region. The GC appeared to be the suitable instrument to check migration from the area. Overall, the existence of a GC in the area had facilitated the development of the area in a more organized fashion, which had been leapfrogging. Besides, the presence of an integrated industrial township consisting of residential, industrial, commercial and institutional complexes had been a direct outcome of the presence of GC in the area. Finally, the GC acted as the facility provider for various land uses in the area. The issue about which there seemed to be concern was the time frame necessary to evaluate the programme. For instance, what does one do if it is realized that the GC has not attained its objectives even 20 years after its existence? Obviously it cannot recover all the expenditure it has made!15 This calls for performance appraisal of those firms locating in the GCs. By 2001, 73 firms, all of them involved in manufacturing, had located in the growth centre creating employment for about 1,500 in the area. One interesting feature was that all were local entrepreneurs who would not have started their ventures had the GC not been in place, in contrast to the Bawal GC (located near the NCR) where most of the firms were national or MNCs. There was one firm that had located in the Shajanwa GC area, long before the GC came into existence, and which was, incidentally, the largest employer in eastern UP (employing 1,400 persons), that qualitatively provided a test of the counterfactual (see Table 7.3). This family-based firm owned a jute

152 Incentives for Regional Development

mill with a sprawling, 55-acre campus where its factories were spread, and which also included housing for its employees. The firm had its own roads, its own water treatment plant and its own power generation capability. The firm claimed to be not using any benefits of being in the GC. While this does give the reader the impression that GC infrastructure is ineffective, remember that this is example of a large firm, by average standards. For small firms, the infrastructure offered by GCs was reported to be quite critical in their decision to locate there. Satharia GC (Jaunpur), UP In July 2002 I visited the Satharia GC in UP, administered by the Satharia Industrial Development Authority (SIDA). This GC is located between Jaunpur and Allahabad on state highway 36, as in the case of the Hassan GC in the south Indian state, Karnataka (not national highway, as in the case of the other growth centres). A fully developed system of infrastructure, including telecommunication, roads, housing, transport connectivity, medical and educational facilities, banks, post office, clean water supply, community and shopping centre, along with 24-hour dedicated industrial power, was in place in the GC. The Satharia GC covered an area of 508 acres, and contained 462 plots developed with infrastructure. The weighted average size of a plot is roughly 0.07 of an acre,16 quite small, when compared to that observed in the other GCs. The average land compensation is in the range of US$2,361–7,083 (in real terms) per acre, consistent with what I find in GCs of other states (see Table 7.2). The area was designated as a GC in 1993, along with the other GCs in the state. Prior to 1993, the area was indeed a No Industry District. Unemployment in the area ranges up to 70 per cent (compare this with the 35 per cent unemployment rate for the district of Jaunpur (in which the GC is located), calculated from 2001 Census data). The literacy rate for Mungra Badshahpur block in which the GC is located, is shockingly low, 37.8 per cent, which for the urban parts of the district is higher (61.2 per cent) than for the rural parts (40.8 per cent). The area is largely agricultural. Because of this, one expects to find agro-based and food-processing firms in the GC. There is, for instance, scope for the paper industry (because of the raw material available), but there were no such firms in the area. Discussion with SIDA officials indicated the frustration of local youth in perpetually trying for jobs before they found something suitable. While the profile of the area is consistent with a distressed area that has potential opportunities for development or attracting businesses, the GC appeared to have been a vehicle of growth – with a number of firms including an MNC and important national firm having located in the area.17 Most of these firms have substantially benefited the local population. A majority (85 per cent) of the firms represented local entrepreneurship, originating from within the state. Roughly 27 per cent of firms were from within the district and some were from neighbouring districts. In Table 7.5, I summarize

India’s Growth Centres: The Impact of Firms 153

the favourable effects firms have had on local labour markets where they locate. In the case of all growth centres, firms from outside the state usually sourced their skilled labour force from outside in which case there is little benefit to the local labour force. In order to address such situations, SIDA specifies to firms that they should provide employment at least to the local labour force that lost their land because of the GC. Beyond this, no other formal agreement between the firms and the local government/SIDA and/or community appeared to be in place regarding the amount of investment, output, nature of employment, wages, and time frame for achieving these outcomes. The area had been offering sales tax exemptions up to 250 per cent on fixed investment till December 2000, which had been discontinued since then, because of the stoppage of incentives agreed to by all state chief ministers in 1999. So, while a number of firms that located in the GC prior to 2000 had located for the sake of the tax incentive, they were continuing their operations there because of the guaranteed availability of all infrastructure.

7.4

Summary of GCs visited

Table 7.3 gives an overview of the GCs I visited and summarizes the size of the GCs, the number of plots that have been developed and allocated, the average size of a plot and the average compensation per acre, based on my discussions with various GC officials. While the area covered by the GC is the largest (in absolute terms) in Hassan GC, the Satharia GC has the largest number of plots allocated.18 This somewhat reinforces Bartik’s idea that competition does encourage growth in high-unemployment areas. While land compensation is in the range of US $2,361–7,083 per acre in all areas, Satharia pays the highest rate of compensation to landowners. The compensation to be paid for land, no doubt, is a determinant of the cost of land development for the GCs, which eventually determines how many industries will locate there. Yet industrial development is what all GCs strive for, as the discussions with various officials indicated. First, based on my visits to firms in various GCs, I will summarize the impact of GC, tax incentives, and infrastructure on firm location decisions. Secondly, I will summarize the impact of firm location on local labour markets, taking into account the nature of the firm (whether capital or labour-intensive), the proportion of jobs created by the firm held by local residents and the nature of jobs (measured in terms of wages paid). This provides tests of the analytical framework that was developed in Chapter 2 of the book, since tax incentives offered in the pre-99 period in India, like those offered in the US, appear to have been incentives to capital in general.19 Thirdly, I examine the export orientation of GC firms and summarize the firms’ suggestions for improving various aspects of the states’ industrial policy.

154 Incentives for Regional Development Table 7.3 Overview of GCs visited GC, State

Area covered by GC (in acres)

Number of plots developed (allotted)

Average size of plot

Average compensation per acre (in constant US $)

Bawal, Haryana

1,209

561 (208)

2,361–4,722 per acre

Hassan, Karnataka Shajanwa, UP

1,662 800

213 (65)* 290 (265)

508

462 (337)

Ranges from 0.06 of an acre to 10 acres 7.8 acres Ranges from 0.06 of an acre to 3 acres Ranges from 0.1 of an acre to 7 acres, average plot size being 0.07 of an acre

Satharia, UP

4,132 per acre 4,132 per acre

2,361–7,083 per acre

Note: * In Hassan GC, 1,662 acres have been developed out of which 514 acres (31 percent of developed area) have been allotted so far. Based on the average size of 7.8 acres per plot, I surmise that 213 plots are developed out of which 65 plots have been allotted. Source: Field visits and author’s calculations.

Finally, I study firm activity in providing any social services in the communities where they have located in order to determine whether firm locations are detrimental or beneficial to the communities in which they locate. These aspects have implications for the following: 1. Whether the state government is justified in spending its resources on the programme. 2. Given the waning bureaucratic interest in the programme, whether the programme is to be merged with another programme or deserves to enjoy stand-alone status. 3. Whether state/local industrial development authorities entrusted with the implementation of the programme should make a contract or, at least, an informal agreement with the firms before they locate, regarding firms’ commitments.

7.5 Impact of GC (infrastructure), and tax incentives on firm location Table 7.4 summarizes the effectiveness of GC infrastructure and of tax incentives in firm location decisions. I distinguish between those firms that located in the pre-GC period (to assess what influenced their decision to

155 Table 7.4 All firms located in GCs visited, and their location decisions Firm number

GC

Effectiveness of GC infrastructure

Effectiveness of tax incentives

Importance of traditional factors

Other locations considered

1

Hassan, Karnataka

Yes

No

Yes (proximity to raw materials)

Yes

2

Hassan, Karnataka

Yes

Yes

Yes (proximity to raw materials and skilled labour)

Local entrepreneurship

3

Hassan, Karnataka

No

Yes

Yes (proximity to raw materials)

Yes

4

Bawal, Haryana

No (pre-GC firm)

Yes

Not important

No

5

Bawal, Haryana

No (Pre-GC firm)

Yes

Yes (proximity to raw materials and access to highway)

Local entrepreneurship

6

Bawal, Haryana

Yes

Yes

Yes (proximity to market)

Yes

7

Bawal, Haryana

Yes

Yes

Yes (proximity to raw materials and skilled labour)

Yes

8

Bawal, Haryana

Yes

Not clear

Yes (cheap land, proximity to raw materials and skilled labour)

No

9

Bawal, Haryana

Yes

No (not effective without market)

Yes (proximity to market)

Yes

10

Shajanwa, UP

Yes

Not clear

Yes (proximity to raw materials and market)

Yes, local entrepreneurship

11

Shajanwa, UP

Yes

No

No

Local entrepreneurship (even without GC)

12

Shajanwa, UP

Yes

Yes

No

Local entrepreneurship (even without GC)

13

Shajanwa, UP

Yes

Yes

Yes (cheap land from GIDA)

Local entrepreneurship (even without GC)

14

Shajanwa, UP

No (Pre-GC)

No

No

Local entrepreneurship (even without GC)

Continued

156 Incentives for Regional Development Table 7.4

Continued

Firm number

GC

Effectiveness of GC infrastructure

Effectiveness of tax incentives

Importance of traditional factors

Other locations considered

15

Satharia, UP Satharia, UP

No (Pre-GC)

Yes

No

Yes

No

Yes (distribution network) Yes (agglomeration economies)

17

Satharia, UP

No (Pre-GC)

Yes

18

Satharia, UP

Yes (mainly power) Yes

16

Yes (proximity to good quality raw material) No

Yes, Local entrepreneurship (wouldn’t have located without GC) Yes

No, local entrepreneurship (wouldn’t have located without GC)

Source: Field visits.

locate there) and those firms that have located since the GC has come into existence, in order to qualitatively evaluate the effect of the GCs and the infrastructure available there. 7.5.1

Pre-GC firms

We may note from Table 7.4 that five of the 18 firms I visited (that is, 27 per cent) in all GCs, are pre-GC, which located there before the GC came into existence. As we would expect, GC infrastructure was not important to them, although firms that were grandfathered into the GC once it came into being, were receiving the benefits of the GC infrastructure that were unavailable to them earlier. Of the five pre-GC firms, the majority (four) of them expressed the view that tax incentives were the reason why they located there.20 Two of the five pre-GC firms were multinational firms whose decisions were not determined by the GC, and two others represented local entrepreneurship that would have located there anyway. The other pre-GC firm located in the GC area did so because of its proximity to raw materials and the availability of tax incentives at the time. For none of these firms was infrastructure an important factor in determining location. These responses are consistent with what we would expect of non-GC firms. 7.5.2

Post-GC firms

It is instructive to note from Table 7.4 that while tax incentives were effective for all pre-GC firms (except one), for only half of the post-GC firms were tax

India’s Growth Centres: The Impact of Firms 157

incentives effective. These post-GC firms located where they did because of traditional factors such as agglomeration economies, and the proximity to raw material sites and markets. More than one-third (35 per cent) of firms surveyed represented local entrepreneurship and would have located there even without GC status. Does this mean that we should discontinue GCs, since these areas would experience industrial development in any event? Although several of these firms represented local entrepreneurship, they found the GC infrastructure and/or tax incentives that were being offered at the time of their location to be effective in influencing their entrepreneurship. With the GC, and the guarantee of infrastructure provision, their entrepreneurship decision occurred sooner rather than being deferred to a later period. GCs do have a substantial amount of local entrepreneurship forthcoming. Discussions with firms in the various GCs indicated that the availability of all infrastructure (power, roads and telecom) at the same time is an important factor attracting local entrepreneurs. This was especially the case when compared to regions (for example, other Indian states) where infrastructure comes to be available only in phases (first power, then telecom, and finally roads). Observe from Table 7.4 that GC infrastructure is effective for all the firms that located in the GC after it came into existence, and played an important role in their decision to locate there. Given that most of these are manufacturing firms, it is reasonable to assume that infrastructure, including the availability of uninterrupted power, is critical to their production process. The firms’ responses showed that states and their industrial development agencies have to aggressively promote their backward areas by providing infrastructure, including road/highway access, power supply, telecom and Internet infrastructure, which would otherwise be unavailable to the firms. Without these services, there would be absolutely no incentive for firms to locate in these remote areas of the country.

7.6

The impact of GC firms on local labour markets

Table 7.5 summarizes the effects of firms that located after the GC came into existence. In this and in future tables, I will omit those firms that were established in the pre-GC period, since our primary interest in studying the impact on local labour markets is to consider firms that have located since the GC came into being. Remember that government’s resources are being spent on the programme, so favourable impacts, if any, have to be recorded in order to determine whether it is an appropriate use of public money. The 15 post-GC firms that had located in the various GCs visited, created 1,977 jobs – or an average of 131 jobs per firm. By contrast, the five pre-GC firms created a total of 2,852 jobs, with an average of roughly 570 jobs per firm (much higher than that for GC firms). Remember, however, that more than 80 per cent of the 2,850 jobs created by non-GC firms were accounted

158 Incentives for Regional Development Table 7.5 Impact of GC firms on local labour market GC

Firm

Capital or labourintensive

Employment

% Local labour force

Salary/wages

Hassan, Karnataka

1

Labour intensive

33

About 60

Minimum wages

Hassan, Karnataka

2

Labour intensive

7

100

Average wages

Hassan, Karnataka

3

Labour intensive

8

100

Minimum wages

Bawal, Haryana

4

Capital intensive

122

50

Thrice the normal

Bawal, Haryana

5

Capital intensive

70

70

Twice the minimum wage

Bawal, Haryana

6

Technology intensive

85

Probably small

Minimum wage

Bawal, Haryana

7

Labour intensive

300

100

Twice the minimum wage

Shajanwa, UP

8

Capital intensive

32

100

Minimum wage

Shajanwa, UP

9

Labour intensive

100–125

100

Half being well paid and half at minimum wage

Shajanwa, UP

10

Labour intensive

125 200 (Off-site)

Technical staff from outside, unskilled – 100 local residents

Minimum wages

Shajanwa, UP

11

Capital intensive

300

60%

Average wages

Satharia, UP

12

Capital intensive

131

90

According to skill level, minimum wage for bluecollar and higher than market wage for white-collar workers

Satharia, UP

13

Labour intensive

20

75

Average wages

Satharia, UP

14

Labour intensive

364

90 for workers and staff, 40 for management

Well above average, highest in the region

Satharia, UP

15

Capital intensive

70

80

Average wages

Source: Field visits.

for by two firms – a large firm representing local entrepreneurship and another, a pre-GC MNC firm.21 It is possible then that the offer of tax incentives to pre-GC firms created these favourable labour market effects. Observe from Table 7.5 that roughly half of the (15) GC firms are labour intensive in their production processes and that these firms account for 59 per cent (1,167 of the 1,977 jobs) created by the firms.22 In most cases, the local labour force is recruited for unskilled jobs, and technical and

India’s Growth Centres: The Impact of Firms 159

management staff is sourced from outside the region (for instance, from company headquarters). The capital-intensive firms are also the ones that pay higher average salaries.23 In the case of one firm, the officials reported that they deliberately make their processes less mechanized (the effect of this on their efficiency was not clear) in order to provide employment to a more local labour force. This firm also paid their employees twice the minimum wage. So it appears that firms (whether pre- or post-GC) have a favourable impact on the local labour market, in terms of recruiting a local labour force, and paying good wages, with some of them even being open to the training of the local labour force. So, from a labour market perspective, it is a good strategy to lure firms with the promise of infrastructure provision, since they create hysteresis effects, and increase long-run employability of the labour force.

7.7

The export orientation of firms

Because of the virtues of export promotion and their impacts on growth, in the 1960s a large number of developing countries adopted outward-looking strategies. India, being a latecomer to industrialization and having adopted a more or less closed external trade policy until 1991 (substitution of imports and pessimistic about its exports), has recently opened up its exports in Chinastyle export promotion zones in the country. At the time work for this research was completed, there was a proposal to merge the export zones (called the Special Economic Zones (SEZs)) with the growth centres. Because of this, I paid some attention to the export orientation of firms locating in the GCs. Table 7.6 summarizes the export orientation of the GC firms and detailed suggestions that were made by these firms to improve the state governments’ industrial policy. It may be noted from the table that at the time of the field visit only two of the 15 firms were exporting, with one exporting a large part of its output. Most of the firms I visited had either scale or financial constraints that prevented exports. They were primarily catering to the domestic Indian Market – or even to the local, regional markets. If a similar pattern exists across all firms in the GCs, then, merging of the existing GCs with the new Special Economic Zones (SEZ) programme being proposed by the Government of India may be irrelevant for a large number of firms. This does not imply, however, that such a merger is undesirable. Relevant suggestions made by firms for improving policy regarding their export orientation pertained to the improvement of road and telecom infrastructure in these areas, a reduction in the level of bureaucracy, greater financial support from state-level financial institutions, technological support from the government, and many others, as summarized in the table. The lack of export emphasis among these firms in the GCs does imply that they cannot be expected to contribute significantly to the growth of the country’s GDP (based on what the past literature has argued) through their

160 Incentives for Regional Development Table 7.6 Export orientation of GC firms and suggestions for government policy GC

Firm

Whether exporting or have plans to export

Suggestion for improving policy

Hassan, Karnataka

1

No

Decrease regulatory burden of pollution requirements

Hassan, Karnataka

2

No

Clearer guidelines for approvals and one-stop window

Hassan, Karnataka

3

No, financial constraints in expanding

Nothing specific

Bawal, Haryana

4

No

Lack of subsidies since 1999 a pinch

Bawal, Haryana

5

Yes, 5% of output, to increase over time

Continuation of infrastructure

Bawal, Haryana

6

Yes, 80%

Merging of GC with SEZ programme beneficial

Bawal, Haryana

7

No

Better service road infrastructure

Shajanwa, UP

8

Very little

Greater technological assistance, Better financial support from financial institutions

Shajanwa, UP

9

Not yet

Secondary effluent treatment plant needed

Shajanwa, UP

10

Not regular exporters as of now

Stoppage of subsidy has made them less competitive vis-à-vis products from Nepal

Shajanwa, UP

11

This unit doesn’t

Government needs to make area attractive for ancillary suppliers to locate there

Satharia, UP

12

Not clear

Improve road infrastructure and access to the area

Satharia, UP

13

No, meets only domestic demand

Bureaucracy in tax collection should end as it gives rise to bribery and corruption

Satharia, UP

14

This unit not exporting

Improvement in telecom infrastructure needed

Satharia, UP

15

No

Maintenance of infrastructure quality

Source: Field visits.

employment and output effects. However, the field visits showed that these firms contributed quite significantly to the labour markets of regional economies where they located, as demonstrated in the previous section. As I highlight in the final chapter and section, employment is a much bigger

India’s Growth Centres: The Impact of Firms 161

issue for developing countries, along with openness (represented by the proportion of exports and imports in the country’s GDP). So, naturally, in the case of regional economies, we are concerned only with the effects on regional employment and not with the effects on overall national growth that may be brought about by exports.

7.8

Corporate social responsibility

There is a need to study the social activities, if any, of firms locating in the GCs of India. This is because, as in the US, a lot of scepticism is expressed in India regarding the wisdom of wooing the private sector to locate in backward areas of the country, since, in general, we do not expect the private sector to be socially responsible. Various theories of corporate social responsibility (CSR) suggest that firms can be expected to engage in social or community-level

Table 7.7 Corporate social responsibility: evidence from GC firms GC

Firm

Whether engaged in some community service

Hassan, Karnataka Hassan, Karnataka Hassan, Karnataka Bawal, Haryana Bawal, Haryana Bawal, Haryana

1 2 3 4 5 6

Not clear Not clear Not clear No No No

Bawal, Haryana

7

Yes

Shajanwa, UP

8

Currently none

Shajanwa, UP Shajanwa, UP Shajanwa, UP

9 10 11

Satharia, UP

12

Currently none Yes Nothing except to invite schoolchildren to watch the making of their product occasionally Yes

Satharia, UP

13

Yes

Satharia, UP

14

Satharia, UP

15

No because of financial problems Not yet

Source: Field visits.

If yes, nature of service

Opened dispensary for village Future: provide training to local labour force Afforestation of the area

Providing free medical camps twice a year to a village; Distribution of winter blankets, textbooks and sporting goods to schools Donations to schools and construction of community halls

162 Incentives for Regional Development

activity only when it increases demand for their own goods and services, and when the cost of supplying the CSR good or service is lower (McWilliams and Siegel, 2001). Evidence from India’s corporate sector suggests that only large, profitable firms engage in such activity to any significant degree. Table 7.7 shows the evidence from GC firms regarding their social contribution to local communities where they located. It shows that 26 per cent (four out of 15 firms) had engaged in some form of community service. While it is not desirable for firms to spend a lot of resources on social service, especially during a period of recession, or when they are making losses, other options do exist. I suggested to firms that they be open to allow local unemployed youth to watch their various processes one or two days a week, and enable them to learn the skills of the jobs that interest them. This, while not necessarily providing them with employment (which, nevertheless, could be a possibility), could increase their employability and create hysteresis effects in the local labour markets. The social activities of firms that have located in the GCs shows conclusive evidence of their partnership in the development process in the area, wherever it was financially feasible for them. So the question that has to be asked is not whether or not the private sector should be involved in the social development of local communities where they locate, but how.

7.9

Firm-level implications

We know from industrial location theory that for manufacturing firms, factors such as the proximity to markets, the availability of inputs, and transportation costs are the most important in influencing their location decisions, and this is supported by the findings from field visits reported in this chapter. Only if such factors are similar within a state will the intra-state incentives (tax or infrastructure) that local governments offer be more likely to influence firm location decisions. In the case of services (banks, insurance, telecommunications), expected customer base and infrastructure are important and the approach proposed here allows discretion in the provision of infrastructure incentives for such firms. Based on visits to firms, the incentives to firms deciding to locate in the GCs should be negotiated on a case-by-case basis. Note that, as in the case of Ohio’s EZs, the very fact of negotiation on a case-by-case basis does not mean being arbitrary. Such an approach avoids the risk of an across-the-board subsidy that could be irrelevant in the decision-making process of any given firm. The financial losses to the government of the consequences of a firm locating in the area just for a token subsidy, can thereby be avoided. If a firm requires assistance to locate, it must be required to demonstrate the need for it and its effects on its performance and functioning in the GC. 7.9.1

Performance appraisal

In the post-1999 period, after the stoppage of tax incentives to firms, typically Indian states provide only infrastructure to firms that locate in GCs, except

India’s Growth Centres: The Impact of Firms 163

tax incentives that were committed before 1999. Based on discussions I have had with DIPP, and the various GCs that I visited, the practice thus far is that firms that locate in the GCs do not have to commit threshold levels of output, investment or employment in order to be eligible for the infrastructure or other incentives. State/local authorities entering into agreements with firms must be required to notify all affected stakeholders. The stakeholders could be landowners,24 residents of the area that experience increased traffic, congestion and/or pollution resulting from location of the business, and any residents that are likely to be affected by a reduction in public services (for instance, reduction in water or power) because of the provision of infrastructure to the firm. Actually, infrastructure should be made applicable to each firm only if it meets certain requirements as stated below, similar to the performance requirements that apply in American EZs: 1. It has hired new employees to fill positions at the facility. At least a certain proportion of these employees should be local residents25 and are either unemployed, or eligible for receipt of food grains under the Public Distribution System (PDS), which could indicate their poverty status, or are physically handicapped. Currently, in Karnataka, physically handicapped entrepreneurs that locate in GCs receive an extra subsidy of 5 per cent of the project costs. Discussions with firms in most GCs indicate that most of the unskilled jobs are offered to local residents. Those who lack the skills must be given training by the local industrial authorities. Some incentives could also be given to encourage firms to train the local labour force. Attempts can be made by them to match jobs and skills available. This makes the local resources spent on infrastructure given to firms in the GCs viable, because they will result in a reduction in local unemployment. This is valid despite the fact that most of the employees are offered only minimum wage jobs, because of the hysteresis effects the jobs are likely to have on workers’ skills and their long run employability. 2. The firm must not have closed a facility or reduced employment elsewhere in the district, or state, for the primary purpose of establishing, expanding, renovating or occupying a facility in the GC. Recall from the earlier chapters that because of the proliferation of EZs in states in the United States, firms save millions of dollars simply by moving across the border.26 But relocation is less likely to be a problem in India, where firms are not footloose.27 When a firm locates in an area and has access to markets and the various distribution networks, it is less likely to move out just because tax incentives are not available any more. In fact the uniform stoppage of tax incentives by all states throughout the country is likely to check firm relocation, even if it is prevalent.28

164 Incentives for Regional Development

Further, performance criteria should be specified for firms that locate in the GCs. Firms currently report their performance to the financial institutions that lend to them, but not to the state or local government or to the agencies implementing the programme. Firms must be required to report their commitments, including investment, employment, and wages to be paid to employees, to the respective state or local government or agency, for making use of the various infrastructural facilities and incentives, prior to their location in the GCs. Byrnes, Marvel and Sridhar (1999) find that firms in the enterprise zones of Ohio in the United States that promised a greater number of jobs in the contract, were more likely to receive more generous tax incentives. In India, taxes are specified by law, and cannot be negotiated, but commitments from firms prior to location and, later, performance appraisal, should be encouraged, in order to make them eligible for GC infrastructure and other in order incentives. A review mechanism must be set up to periodically evaluate the performance of firms in the GCs in order to ensure that those investment, output, and job creation commitments specified in the contract are indeed met by the firms. If there is any discrepancy between commitment and performance, the subsidies and incentives have to be re-examined and the appropriate clauses have to be noted in the contract. This will act as a disincentive to any erring firms. Review and performance appraisal are important tools in ensuring that the objectives of the GCs are achieved and they provide benchmarks to assess employment targets. They also enable a better appraisal of the programme from the government’s perspective, which is particularly important because state funds are spent, and they can also ensure that the appraisal does not add another layer of bureaucracy for administering the programme. One possible solution is to give this discretionary power of monitoring to the local industrial development authority. In order to facilitate this, it is desirable that such industrial authority officials are divested of other responsibilities so that they can concentrate all of their attention on addressing firms’ grievances. Currently in a number of Indian states, industrial development authorities are frequently employees that also have other state government functions, which reduces the effectiveness of their role as facilitators to firms in GCs. A similar experience has been recorded in many American EZs as well.

7.10 Growth-centre-level implications Based on the empirical work, my field visits, and discussions with the agencies involved in implementation of the programme, it is apparent that GCs are designated by the state governments based on a variety of criteria, which are unclear to the states, and to the industrial authorities implementing the programme. While official criteria for the establishment of GCs

India’s Growth Centres: The Impact of Firms 165

relate to the lack of urbanization or presence of infrastructure, it is quite possible for many undeserving areas to be granted GC status. This is because if an area is located at a certain distance from an urban area as specified by the policy, it can either have the infrastructure or not have it, but in both cases it could get designated as a GC, according to existing criteria. For instance, note criterion III for GC designation (Chapter 6). There can be several problems with existing criteria: The first problem would be that undeserving areas, or those that could attract industries without GCs, would be given sops at the cost of the state and central exchequer. Second, lack of clear criteria makes the designation process arbitrary, and gives rise for lobbying, as it has happened in the case of many states in India. Additional criteria that should be taken into account for the designation of GCs, to industrialize backward areas and to alleviate unemployment, based on the theory, empirical work from the US and field visits to GCs in India, are as follows: 1. High unemployment: This can be measured by the degree of unemployment in the state relative to the average unemployment rate throughout the nation. For instance, states having 125 per cent of the nation’s average unemployment rate during the most recent 12 months, could be determined to be high-unemployment areas. I have elaborated on the theoretical arguments in Chapter 1, as to why we may expect high-unemployment areas to be more deserving than others to attract industry and create employment. For the purposes of designation, first, states with high unemployment rates in relation to the national average can be identified. The selected states can be asked to identify their worst unemployment areas and backward industrial areas. This can encourage states to undertake mapping of the spatial distribution of informal sector (un)employment and industries, and can be feasibly used as criteria for the designation of GCs. Currently, these data are collected at all levels for the organized sector only. It is plausible that some informal sector data are compiled for the metropolitan areas. But no data are collected for informal sector in semi-urban or rural areas that would indeed be important for the growth centre programme. Industrial backwardness: The prevalence of closed industrial facilities (for instance, a minimum of 5 per cent of existing units) can be said to indicate the extent of industrial backwardness within an area. Again, this criterion would persuade the states to undertake spatial mapping of industrial backwardness and informal (unorganized) sector employment, encouraging them to probe why some areas are more industrially backward than others. If these were to be policy-relevant variables, then they could well be within the states’ ambit and offer scope for their improvement.29

166 Incentives for Regional Development

7.11 Concluding remarks The findings of this chapter show that the success of the GC programme depends upon the performance of all actors. It is performance-based both for the state/local government administering the programme and for the firms that make the commitments. This implies that the state and local governments should be on their guard to protect the interests of the consumers, residents and public of the community, and that of the exchequer. One way in which they can do this is to base GC designation upon explicit, distress criteria, some of which are indicated in this chapter, and to encourage competition among distressed areas in the offer of infrastructure. Further, local governments must sign agreements or formal contracts with firms locating in the GCs, in which firms commit to investment and jobs, and these contracts should include clauses as to the possible actions that could be taken against them if the commitments are not met. With this, performing firms need not be viewed as being those with mercenary objectives, but those that can be involved as partners in the development process, along with local governments.

Part IV Lessons Learned

This page intentionally left blank

8 Lessons Learned from the United States and India

8.1

Introduction

We started the book by attempting to answer several questions. First, using an analytical framework, we studied the impact of tax incentives on the economies that adopt them, adopting a general equilibrium framework from the standard literature on the subject. Knowing completely well that it is impossible to isolate the incidence of any tax cut solely to the local economy, we made an attempt to understand if such incentives merely redistribute employment, if a perspective is taken, that extends beyond the local economies adopting them. We found that such redistribution may not be zero-sum, even in a general equilibrium context, provided the redistribution of firms and jobs occurs towards the most distressed regions. These findings are based on the analytical framework and the EZ programmes of two large Midwestern states of the United States, which adopted EZs as soon as the concept was introduced in the USA, and have been controversial for various reasons. Regarding the benefits and costs of regional development, based on the empirical evidence, this work suggests that the net benefits from jobs are more than the costs of generating them, if certain conditions are met: (i) The jobs are created in high-unemployment areas; (ii) The jobs have to be well-paid, implying a certain level of skills; (iii) Careful targeting of labour-intensive firms that can create skilled jobs without the offer of generous tax incentives. Further, we examined the impact of tax (financial) and infrastructure (physical) incentives on the unemployment rate, offering evidence from the USA for tax incentives and from India’s GCs regarding infrastructure incentives. We found that tax incentives are highly successful in reducing levels of unemployment in the areas adopting them. Some reasons why this can occur are that tax incentives provide some incentive to firms to locate in 169

170 Incentives for Regional Development

distressed areas. The property tax abatement, being an incentive to capital, facilitates the use of capital in these areas to increase the productivity and marginal product of labour and increases their market wage. This facilitates the further employability of labour and reduces the unemployment rate of the area. However, there is an optimum time period for which tax incentives may be offered, beyond which their effects on unemployment become unfavourable. The work in this book finds that three to five years would be the optimum for maximizing the beneficial effects of tax incentives on unemployment. On the other hand, infrastructure (physical) incentives are not as successful in reducing the unemployment rate. This is especially the case if infrastructure incentives are offered only on a small scale in a few areas of large countries such as India. This is because the empirical work in this book also shows that wherever infrastructure incentives exist, they are a powerful determinant of the number of firms locating there, consistent with the findings in earlier literature. This implies that if physical, infrastructure incentives are offered by distressed areas, they can enable firms to become drivers of employment growth. Further, physical, infrastructure incentives have to be more geographically more extensive in distressed areas, so that disparities in infrastructure in large, developing countries such as India are not an impediment for what these areas may have to offer to firms and residents. Note that the implication of this observation is not to advocate an acrossthe-board proliferation of such incentives. As we have observed in earlier chapters, the effects of having too many tax incentive areas, especially in areas that could attract firms anyway, are harmful. There are several benefits, however, if physical, infrastructure incentives are provided by distressed areas. Some benefits are improvement in their state of infrastructure, which might be financed by the business themselves, enabling them to attract residents and other related businesses. The empirical work from various chapters supports this notion. This chapter is organized as follows. First, I summarize lessons for subnational units in the United States, based on the empirical work. Then, I focus on the relationship between terrorism and/or natural calamity and incentives since they are of relevance respectively for the USA post-9/11 and for developing countries, post-tsunami. I then summarize lessons for India. Here I dwell upon the impact infrastructure has on growth, poverty, and local government finances, in the context of developing countries. I summarize estimates of infrastructure gaps in India’s context, and explore whether the American experience is relevant for India to consider and if yes, what are these lessons. Here I discuss employment in the unorganized sector and implications for decentralization. I explore the ways in which the corporate sector can help in a country like India. The final section contains concluding remarks.

Lessons Learned from the USA and India 171

8.2

Lessons for the United States

Given these findings, are infrastructure incentives better than tax incentives? What does the experience of those Indian states that have called a stop to the tax war indicate for sub-national units in the United States? The evidence from India suggests that the Indian states have been courageous to call a stop to the tax war – the so-called ‘race to the bottom’. One reason why countries such as the United States have not called a stop to the tax war is that they continue to believe in laissez-faire and supply-side economics (for instance, Reaganomics) that upholds low tax rates. There are several factors that determine whether or not states offer tax incentives, and how generous they are with respect to tax incentives – their affluence, the infrastructure, the tax behaviour of neighbouring states and their own tax rates. While affluent states can afford tax incentives, states usually offer tax incentives to offset the negative effect of their poor infrastructure, the poor quality of their public services, the low tax rates of their neighbours, or their own high tax rates. In the United States, scepticism continues to surround the offer of tax incentives, since it hampers the ability of state and local governments to provide good quality public services, and it is unclear if tax incentives have influenced firm location decisions, despite Bartik’s econometric evidence. This might hold good even if tax incentives have favourable impacts on the unemployment rate of the regional economies adopting them. It is true that the American states are much more homogenous in terms of infrastructure than are the Indian regions. However, even in the USA, there are notable disparities in the levels of infrastructure one observes in richer states such as California and those offered by the poorer states such as West Virginia. While one reason for the difference is the fiscal capacity of these states, another reason could be their topography. While topography is not a policy variable, the stoppage of tax incentives directly impacts the fiscal ability of poorer states to offer better infrastructure. The richer, sun-belt states (such as California’s Silicon Valley) with better ‘business climate’ indices and rankings, continue to attract the largest number of firms. The lesson for the distressed areas of the USA, from Indian states, is that they should convince firms of the importance of infrastructure for manufacturing and services, and market their states and local economies accordingly. As indicated earlier, they will stand to benefit by incentivizing businesses to finance the infrastructure improvements in their areas, which they are otherwise not able to afford, or would have lose a significant portion of their tax base if they were to offer tax incentives. There is a considerable literature (for instance, Saxenian, 1994) that speaks of the differences in corporate culture across the Silicon Valley and Route 128 (in Boston) that have explained their successes. In this book, however, we have explored ways of recruiting firms, rather than explaining what makes them more successful. We have found

172 Incentives for Regional Development

that infrastructure incentives are a much more sustainable way of reducing the disparities between the richer and the poorer regions. 8.2.1

Terrorism and incentives

In the context of the United States, there is considerable concern that has developed regarding terrorism, and in the Asian countries such as India, regarding natural calamities (this book was completed when a devastating tsunami struck Asia in December 2004). This book would be incomplete if we did not consider the impact of terrorism or natural calamities on tax incentives and public services. In the wake of terrorism attacks, the costs of redevelopment have to be financed by future residents and businesses of the areas that have been the targets of terrorist attacks (for instance, New York in the 11 September 2001 attacks). In American local public finance, as must have been clear from Chapter 3 (recall the illustration in Table 3.3), property taxes are used to finance local public services so that a higher property tax rate generally means a higher level of public services. In the event of terrorist attacks or natural calamity, a higher property tax rate is implied, while only managing to hold the quality of public services constant. As an instance, the National Institute of Public Finance and Policy (NIPFP) in India was entrusted with a study by the Andaman and Nicobar Administration relating to public services, norms, and their financing, before the tsunami struck Asia and the A&N islands. The tsunami considerably complicates the analysis and the finances used to reconstruct the infrastructure such as roads, water and sewer lines, that were substantially destroyed in the islands. This, being example of only a study, could make such terrorist or natural calamity targets less attractive to current and future residents, that would vote for safer communities that offer their preferred combinations of the tax rate–public service package, for their household location decisions. What are the implications of terrorist attacks or natural calamities for business location decisions? Generally, it might be argued that, when all is said and done, terrorist targets (such as New York) will continue to remain attractive (or as unattractive) (as the financial capital of the USA) to firms. However, if the overall business climate consisting of amenities, public services, topography (indicating vulnerability to natural calamities such as earthquakes and tsunami) and crime, is examined by firms who are considering locating to the area, then past targets of terrorist attacks or natural calamity will stand to lose out. From this viewpoint, these regions could lose a significant amount of power in bargaining with firms. This implies that in the tax incentive war that continues to rage among various sub-national units in the United States to attract businesses, governments of such areas might have to offer larger tax abatements or more appropriate infrastructure, in order to persuade firms to locate there. Incidentally, the literature on

Lessons Learned from the USA and India 173

bargaining between local governments and firms shows that the higher the property tax rate, the higher would be the tax or other incentives that have to be offered, thereby reducing the effective tax rate and offset their negative impact for businesses. This implies a higher fiscal burden on such governments and school districts that share taxes in the area. Such terrorist attacks and natural calamities would significantly affect public finances, infrastructure and other public services, and thereby business location decisions towards those areas. Both in the developed countries such as the USA and in developing countries such as India, often, too much time, lack of work, and frustration lead to vandalism in a number of instances. In such instances, the corporate sector can play a positive role by creating employment opportunities by providing training to residents (not necessarily employees). This is positive from a firm’s perspective as better skills, jobs, and increasing incomes mean increased purchasing power in its market. If incentives were to be provided to firms to allow them to provide training to those prospective employees that might lack the skills, firms can be successful in reducing vandalism. Finally, firms may also find that their social awareness indirectly contributes to advertising their products or services. One may readily recognize that all these approaches and lessons are indeed market-based and suit laissez-faire economies such as the US. The corporate sector can also play a useful role in the context of developing countries such as India, which I elaborate in a subsequent section below.

8.3

Lessons for India

In developing countries such as India, there are several constraints which might prevent sub-national governments from offering financial incentives such as tax abatements: (i) The pressure for growth restricts the budgetary deficits that the centre and states can incur, because of the adverse impact that fiscal deficits have on growth and macroeconomic stability (Kochhar, 2004); (ii) Much of Indian states’ own source revenues depend upon sales taxes and they cannot afford to lose the autonomy by providing sales tax incentives.1 Further, there is a strong case why sub-national governments in countries such as India need to foster competition in the provision of infrastructure: (i) First, the state of infrastructure in countries such as India is generally poor. It is agreed that adequate infrastructure is not only necessary for increasing productivity, but also for raising the general quality of living. According to India’s Union Urban Development Ministry, however, 20 per cent of India’s urban households are denied access to safe drinking

174 Incentives for Regional Development

water, 58 per cent do not have safe sanitation, and more than 40 per cent of garbage generated is left uncollected for want of proper waste management.2 (ii) Even in the now developed countries such as the Republic of Korea, the government estimates that infrastructure shortages result in a GDP loss of as much as 16 per cent of its potential in the mid-1990s (Singh and Ta’i, 2000). It is estimated that losses from traffic jams in Bangkok range from US$272 million to US$1 billion a year. Because of these considerations, it is now generally accepted that an improvement in infrastructure competitiveness can actually foster growth. 8.3.1 The impact of infrastructure on growth and poverty in developing countries There is a vast body of recent literature that shows the favourable effects of infrastructure investment on poverty and, more generally, on growth, in the context of developing countries. Recent research shows that cellular phone services that have experienced leapfrogged growth in developing countries, have a positive impact on their national output, much greater than that contributed by landline penetration.3 Research by Sridhar and Sridhar (2004) shows that cell phones contribute positively to national output of developing economies, on average being 2.48 per cent, much higher than that contributed by landlines (1.62 per cent), when controlled for country-specific fixed effects.4 Hence developing countries have to facilitate investments in the rapidly growing infrastructure sectors such as telecommunications. Note, however, that increasing spending on any infrastructure sector implies that there is a direct relationship between services provided and the user charge, as in the case of the telecoms sector. The lesson based on telecommunications, for other infrastructure sectors in developing countries, is to demonstrate to consumers the need to pay for services, when a road is used, in the same way a payment is made for making phone calls. The user charge–service nexus is very much required to increase spending on those sectors, apart from budgetary allocation in these countries, which tends to be sparse, and is often the victim of ‘pork-barrel’ politics. One reason why the growth and reforms that have occurred at the macrolevel in India since the 1990s have not trickled down to rural areas is because the road infrastructure has remained poor and has impaired access to jobs where they exist. Research by Fan, Hazell and Thorat (2000) based on state-level panel data for India over the period 1970–93, suggests that if the government were to increase its investment in roads, the incidence of rural poverty will decrease. Specifically, this research shows that if investment were to be increased by INR 100 billion (at 1993 constant prices) (or US$1.18 billion (in constant 1982 prices)), the incidence of rural poverty will decrease by 0.65 per cent. Further, this research reports that for each

Lessons Learned from the USA and India 175

INR million increase in investment in roads, 124 poor people would be lifted above the poverty line. If an additional million INR is invested in research and development (R&D), 85 people are lifted above the poverty line because of the impact on productivity and incomes. Roads increase productivity and wages, and provide access to non-agricultural employment, much required for developing countries dependent on agriculture. Hence it is possible to believe that infrastructure investments in roads, telecommunications, power, water and other sectors will greatly enhance the ability of developing countries such as India to not only grow more rapidly but also reap the benefits of such growth. 8.3.2

Impact of infrastructure on local government finances

Stimulating competition in providing infrastructure services can bring about not only infrastructure reform in countries such as India, but can also have far-reaching impacts on the revenue bases of local economies. The literature shows that public and infrastructure services are likely to be capitalized in higher real estate values. Given that local government property taxation in India is now moving towards the more objective, unit area system, that assesses property on the basis of its objective characteristics, there is an opportunity for local governments to capture the capitalization in the form of higher property tax revenues.5 It is important to increase reliance on the property tax. This is because most state governments in India and around the world have now (or at least on the verge of) abolished octroi, which is widely known to be a highly distortionary tax. Octroi is a tax on business activity levied on the entry of goods into a municipal area for consumption, use or sale. Octroi, until now, had been an important source of revenue, contributing, in most cases, twothirds of own-source revenue for local governments in India.6 In India, as in the USA, zoning is one way of ensuring a quid pro quo relationship between taxes and public services rendered. To motivate the need for zoning, imagine that we have combined heterogenous property values together in an area, but that the owners all pay the same property tax rate. Note that while the smaller house owners pay the same tax rate, they pay a lower amount of tax in absolute terms, when compared to that paid by owners of higher property values. This becomes especially stark if the lower property value residents also have a larger household size, since they are likely to consume larger amounts of public services (such as water, sanitation, sewerage, roads and so on). This is a situation in which owners of higher property values subsidize such public services for owners of lower property values, who receive identical public services, with payment of a lower property tax. With zoning, uniform property tax rates apply within specified boundaries, so that properties of a certain value all pay the same property tax rate and are eligible for identical public services. The role and advantages of zoning in the provision of good quality, local infrastructure, and their financing

176 Incentives for Regional Development

from local taxes, in countries such as India should be clear. Certainly, this is a lesson for India to be learnt from the USA, no matter how ‘exclusionary’ the critics of zoning claim it to be. 8.3.3

Infrastructure estimates, financing and gaps

The investment requirements of infrastructure in developing countries such as India are also mind-boggling. Consider the level of investment required in roads just at the national level for India. Just for the golden quadrilateral (a project linking the four metro areas of the country), and the north–south–east–west corridor, the requirements are estimated to be US$20 billion (in constant terms, with 1982–4 100). Of this, $1.9 billion (in constant terms) is debt and needs to be serviced. The National Highways Authority of India requires more than US$177 million (in constant terms) every year until the debt is redeemed. The Indian railways, one of the largest rail networks in the world, need $1.4 billion (in constant terms) in order to enhance capacity and strengthen the rail network in the golden quadrilateral, connecting the four metros. For this, multilateral funding has been requested from the African Development Bank, the World Bank and the Japanese Bank for International Co-operation. Further, regarding power requirements, capacity addition for power generation alone is estimated to be 111,500 megawatts over ten years, 142,000 megawatts over 15 years, in figures published by the expert committee on the commercialization of infrastructure, appointed by the Government of India, which submitted its report in 1996 (Mohan, 1996). Similarly, the average additional annual port capacity required, based on projected traffic, is 35.35 million tons up to 2006, and the average total annual capacity required for projected traffic is 356.5 million tons (Expert Committee’s estimates). Further, urban infrastructure services (water, sewerage, sanitation, solid waste management, roads, street lighting) require roughly US$1.8 billion (deflated) each year, according to the estimates of the expert committee. Available funding from user charges, taxes and tariffs for these various services is about $10 billion (in real terms), which leaves the remainder to be financed from elsewhere. Based on estimates made by past committees, the infrastructure gap is such that roughly US$37 billion (in real terms) is required on an annual basis for meeting the needs in power, telecoms, roads (new and maintenance), ports and urban infrastructure. These investment and financing requirements, enormous in the context of developing countries, demonstrate that it is critical to involve the private sector in financing these services. In the USA, local governments have traditionally used impact fees to recover the costs of relevant local public infrastructure from the developers themselves. This was appropriate in the circumstances of several growth centres in India, where the firms themselves bore the cost of financing various public services to them, in phases, a lesson to be disseminated much more extensively. In India, public–private partnerships (PPPs) are gaining increasing acceptance for many infrastructure

Lessons Learned from the USA and India 177

projects; however, it is not clear whether they are being targeted at those distressed areas that are most in need. The intention of this book has been to direct such attempts towards distressed regions, by providing appropriate incentives for the private sector, which would not be in a position to do so otherwise.

8.4

Is the American experience relevant for India?

Although there is a perception that the experience of the Asian countries is more appropriate for India, the linkages created by incentives in US-style EZs have greater relevance for India than the China-style zones meant for export promotion alone. Export promotion, no doubt, is the hallmark of an open economy. However, we have more severe, both social and economic, problems in unemployment in developing countries like India, than lack of openness.7 If an important objective of the allocation of resources in developing economies is to rapidly increase national output and per capita income, the workforce participation rate has to increase substantially. India’s current workforce participation rate, being 41 per cent (based on Census 2001 data), is much lower than that in advanced countries such as the United States and Australia (where it was roughly 51 per cent each) and New Zealand (where it was roughly 50 per cent, based on data for 2001). There are several ways in which workforce participation can be increased as Sridhar (2004a) points out: 1. Increase job opportunities (for those in the labour force, that is, those looking for jobs, but not employed); 2. Provide incentives for increasing wages (for those outside the labour force); For instance, at high enough wages, even those outside the workforce (such as housewives, the handicapped, or those recently retired) may be encouraged to participate in the labour market, since the opportunity costs of leisure are high. Recall the theoretical model in Chapter 2 points out that one cause of unemployment is when market wages are lower compared to the reservation wages of unemployed that are supported by social networks. In India, with the demand for skills having been fuelled by the success of IT, and IT-enabled services such as Business Process Outsourcing (BPO), high market wages have become reality now. 3. Increase local entrepreneurship (for those in and out of the labour force). Growth centres have adopted all of these policies, as learnt from the empirical evidence presented in this book. Many advocate the use of public works programmes to create employment in the rural areas of developing countries, since they have Keynesian effects. While that might be a good idea in the short run, we know that public works

178 Incentives for Regional Development

programmes (new or maintenance works) cannot go on indefinitely, even if they are implemented with the usual time overruns which are common in developing countries. For this reason, jobs created as part of such programmes would also be short-lived. India’s United Progressive Alliance (UPA) government committed to legislating an employment guarantee in late 2004, which would ensure at least 100 days of public employment for one member of poor and lower-middle-class households. What would happen beyond the 100 days? While it is undeniable that the workers earn some skills and income that can have hysteresis effects on their long-run employability and skills, at the end of the project, they are back to square one and have to go seeking work again! Further, such programmes have considerable consequences for the government’s exchequer. Traditionally, another policy approach towards geographically disadvantaged areas in India has been to set up institutes of national importance such as the Indian Institutes of Technology and the Indian Institutes of Management. This is likely to boost the area’s regional advantage in developing a pool of skilled labour force, attracting good infrastructure, employment and professional opportunities. While this approach definitely supplements the area’s employment and the skills of the labour force, similar to what is proposed in this book, the cost is once again met by the government’s exchequer. There are a large number of regions in developing countries such as India that do not contain institutes of national learning, neither do they have infrastructure, a pre-existing industrial or economic base or growth centres. For these reasons, a more sustainable way of increasing non-agricultural employment in the rural areas of developing countries like India, is what the preceding section of this book has argued. This strategy is to attract firms to the rural and semi-urban areas with the provision of infrastructure such as roads, power, telecommunications, banking and good schooling which is comparable to that found in their urban counterparts. Without infrastructure incentives, why would a firm (national or multinational or even local entrepreneurship) locate in a remote corner of the country? Investment in attracting firms to rural areas is also a sustainable alternative to the problem of providing non-agricultural employment. This is because, at least in the Indian context, firms do not move once they have located in an area and have access to all infrastructure, in addition to inputs and markets. As noted earlier, this is not the case in countries such as United States where firms pack their bags and leave overnight if they perceive tax savings to be substantial, across state or metropolitan area borders. Furthermore, in India, firms that locate in semi-urban and rural areas are socially quite responsible: they donate blankets during winter, open free dispensaries for the poor, and many times construct schools and hospitals in these areas, as we explained in Chapter 7. Local economic development and place-oriented policies such as EZs or GCs stimulate employment in the area and revitalize the neighbourhood.

Lessons Learned from the USA and India 179

Place-oriented policies are especially important for a country like India where a substantial part of the population is immobile because of psychological or social ties to the location, costs of moving, lack of education or information regarding the availability of jobs. Place-oriented policies also check migration to the urban areas that creates slums. Recent literature (Sridhar, 2004b) also shows that jobs follow people and skills, and not the other way round. If this were to be true, then we have to pursue placeoriented policies aggresively, to attract skills, and jobs will follow. Based on the empirical work reported from the GCs of India, one lesson has been that the designation of growth centres and infrastructure networks is driven by political economy considerations rather than objective criteria. A fervent wish and a potential area of policy-relevant research is to study the demography and economic base of all regions in the country. In fact, there are tools such as shift-share analysis that are available to analyze such data. This technique enables one to disaggregate the parts of output or employment in an area that are attributable to national growth, industry composition, and, finally, local factors. This would enable an understanding of local strengths and weaknesses so that the states and the concerned local governments can leverage and improve upon areas of strengths. This would make the study of these areas, their needs, and their local economic strengths more objective than has been the case so far. Autonomy must be given to Indian states, as with their American EZ counterparts, to set up growth centres in (cities/municipalities/unincorporated) areas of high unemployment, poor infrastructure or low income, as part of a competitive designation process in which various areas of states can apply. For example, in Ohio’s enterprise zone programme in the United States, areas having 125 per cent the state’s average unemployment during the most recent 12 months, or those having 51 per cent of their population below 80 per cent of the area’s median income, can apply to be designated as (full-authority) enterprise zones. Similarly, as we learnt in Chapter 3, in Illinois EZs, areas that have 120 per cent of the state’s average unemployment rate are designated as EZs. Distress criteria such as these tend to be objective and check lobbying for zone designation by areas that do not need special incentives in order to attract investment. Currently, however, in India, as we have learnt from Chapters 6 and 7, growth centres are not designated on any objective criteria. If they were to be designated and marketed on the basis of the objective characteristics of areas, and were provided sunset provisions, they could be a more fruitful way of attracting firms – a strategy that would provide information to current and prospective firms about the area’s strengths and weaknesses. 8.4.1 Lessons for employment: employment in the organized vs the unorganized sector A substantial part of this book has dealt with labour market issues such as reservation wages, market wages and employment/unemployment. In this

180 Incentives for Regional Development

section, I summarize lessons for employment in India’s unorganized sector. This has implications for other similar developing countries, which have substantial proportions of their labour force working in the informal sector. The unemployment problems in developing countries including India are compounded because data on unemployment are available in the organized sector, but not in the informal sector. It is possible that sub-national units that report unemployment from their organized sector do overstate the magnitude of unemployment, because of employment in the unorganized or informal sectors. But in general, high unemployment, even if it is from a purely organized sector perspective, does reveal much about the performance of, and constraints in, the organized sector. It indicates that organized sector jobs require a certain level of skills, when compared to those available in the informal sector of developing countries. In either case, increasing employment has hysteresis effects on the labour market, as we have argued throughout, following Bartik. Since states need to ensure that low-income and high-unemployment areas become designated as growth centres, it is important for the central government in developing countries with a federal structure, to persuade their sub-national units to collect systematic data on employment in the unorganized sector. Data from the unorganized sector are more amenable to being collected by the states because of their local nature. Data from the unorganized sector are important for the following reasons: (i) Data from the unorganized sector are important for enabling one to understand the magnitude of total unemployment in any country, along with that in the organized sector. (ii) In countries with a federal structure, complete unemployment data are needed for use as a criterion for resource transfers to states, and to assess their revenue generation capacity. (iii) Finally, data collection efforts can be tailored to local requirements and lead to innovation. These are vital issues if we intend to provide autonomy to sub-national units in a federation, enabling them to promote growth centres, and other programmes of their own. In its mid-term appraisal of the tenth plan, in January 2005, the Government of India’s planning body, the Planning Commission, favoured providing social security to all workers in the informal and unorganized sectors. The empirical evidence presented from the American data set, PSID, in this book (for instance, Table 5.1), shows that the provision of benefits on a job increases the wage at which a person would be willing to accept a new job (the reservation wage). The lesson from this for India is that the provision of a safety net to workers in the unorganized sector has the effect of increasing their current real wages. This appears to be an improvement. However, if we consider this closely, we can appreciate its wider implications.

Lessons Learned from the USA and India 181

One of the long-run objectives of employment policy in most developing countries is to induce workers in the unorganized to move in to the organized sector. This is important for the reasons already identified. First, provision of benefits and a safety net are assured in the organized sector by the establishments employing them (the burden is not on the government, except where the government is the employer, which, in the case of India, is quite substantial, being close to 70 per cent of total employment). Secondly, workers are likely to have better working conditions in the organized sector (regular working hours, higher real wages, since they will be governed by minimum wage legislation). It is important to be able to identify the magnitude of unemployment in its entirety if we have to alleviate the problem. Currently, because of a large proportion of India’s labour force in the unorganized sector, one is not able to assess the magnitude of the problem due to the lack of monetization and data in the unorganized sector. So any safety net to workers in India’s unorganized sector would have the effect of slowing down reform by raising the real wages, their reservation wages and hampering mobility from the unorganized to the organized sector, which would be desirable for the reasons summarized above. For these reasons, it is important to collect data on individual reservation wages in India to understand the extent of the net benefits from jobs created. This is an area of future research, once the data become available. 8.4.2

Decentralization

Another valid lesson for India from the American experience is decentralization. That is, there is a need to examine the benefits and costs of transferring the growth centres programme to the states. There are several reasons for place-oriented programmes, such as growth centres, to be administered at the state level in India, as with American EZs: First, although intra-state disparities exist, the state is an independent political unit, having policies that affect its business climate. Recognize that certain states are more likely to influence firm location favourably than others, even though both may have growth centres offering various incentives. Data on growth centres also corroborate this. For example, on average, about 50 firms have (cumulatively) located (or have proposed to locate) in the growth centres of the south Indian state, Karnataka, as of 2004, when compared to none in the growth centres of states like Bihar and Assam. Industrial location theory indicates that, in the case of manufacturing firms, factors such as the availability of raw materials, transportation costs, access to and size of markets are more important factors than tax incentives in influencing firm location. In the case of services (banks, insurance, telecommunications), the expected customer base and the infrastructure (such as pre-existing backbone and access networks for telecommunications) are important. Growth centres allow discretion for provision of infrastructure incentives for such firms. We have to note that only within a state are

182 Incentives for Regional Development

these traditional and other factors likely to be approximately equal. For this reason, any incentives that state governments offer in the growth centres are more likely to influence firm location decisions. This represents a case for programmes such as the growth centres to be implemented at the state level in countries with a federal structure. Secondly, implementation and funding of the programme at the state level forces the states in a federation to be fiscally more prudent. The design of the growth centres programme by states could be more targeted and focus on their distressed, high-unemployment or low-income areas. Currently growth centres in India are not designated on the basis of any objective criteria. This gives room for discretion, arbitrariness and political lobbying in the granting of growth centre designation, which could waste a lot of the states’ resources. Finally, transferring the GC programme to the states gives them sufficient flexibility to design their own programmes that will suit local requirements. Zero-sum growth centres? Given the challenge that has been raised with all policies targeting branch plant recruitment, the suggestion that the GC programme should be administered and implemented at the state level does not imply that firm relocation across growth centres within a state, is likely to occur more frequently. This is because careful design of the programme will act as a check on the proliferation of growth centres providing incentives and hence the propensity of firms relocating from one growth centre to another for the sake of the incentive. If state governments are granted the autonomy they ask for, they have to induct some professionalism in their management of the programme. They can do this by adopting the following: First, it is necessary to pursue policies that highlight inherent competitive advantages present. For instance, the case of Indian states demonstrates that states like Kerala and Uttaranchal might fare well as tourist destinations, whereas Karnataka and Delhi could market themselves as having a large and well-skilled workforce. Secondly, the field visits provided important data regarding the application process by states in India. In a number of GCs, it was found that projects failed to take off, primarily due to problems relating to land acquisition. Based on these cases, and wherever it is not currently included, land acquisition costs (compensation costs) have to be included in the project assessment that deals with the viability of various projects, submitted to DIPP by the states as part of their application for requesting GCs. Further, in most of the states, there is a long time-lag between the submission of the necessary documents by the firm and the actual disbursal of the infrastructure incentive. It is desirable that there is a one-stop window to obtain all the necessary clearances for the industry to start production. Next, currently, pollution laws in some Indian states require entrepreneurs to obtain twice the amount of land that is actually required for effluent discharge and removal. Before production has started, this does not make

Lessons Learned from the USA and India 183

sense. It is, therefore, desirable that states facilitate the location of firms and initiation of production in GCs by simplifying administrative procedures. Finally, when infrastructure or public service incentives are offered to firms, states have to be sure to sign a contract that guarantees the jobs to be created, their nature, the time frame and their commitment to the community. With this, states can ensure that firms do not relocate to other states when they get better incentive packages. It is necessary to make a periodic assessment of the progress made with respect to the creation of productive assets, unemployment, infrastructure, and economic needs of the various regions, take stock, and direct funds accordingly. The devolution of the growth centres programme to the states, as with fiscal resources, therefore has to recognize that responsibility comes before freedom.

8.5

Can the corporate sector help?

In section 8.2.1, we dwelled briefly upon how the corporate sector can help socially by creating employment, in the context of developed countries. Here I summarize how such market-based approaches to reduce social tensions may be available in the context of heterogeneous countries such as India. With increasing private participation in the development process, we have to ask whether the corporate sector can contribute in ways apart from financing development and infrastructure in communities of developing countries. In a diverse country such as India, there are often social and communal tensions between religious groups. The corporate sector can facilitate the reduction of such tensions through their location decisions. The corporate sector can undertake a comprehensive analysis of the history of communal tensions in various parts of the country. It can publicize this and avoid locating in those areas that are characterized by social and/or communal tensions. This might provide a signal to the community that they have to be unified as a community if they want to encourage companies to locate to their area and create employment. To demonstrate this effect, imagine that one firm decides to locate in a community even if it is not safe communally, lured by its purchasing power, market potential, and tax and/or infrastructure incentives. Nevertheless, other firms decide to stay away from this communally unsafe community, because it is not conducive to their long-term interests. In such cases, the lone firm that is promoting the unsafe community for the sake of incentives may be in disrepute – or, at least, may have to re-think its decision to locate in this community. It may well be in the interests of all firms to form a lobby and collectively isolate communities that offer an environment conducive for communal tensions. The firms might stand well to gain, by showcasing that they are in favour of communal harmony. In such instances, depending upon the intensity of competition that prevails to attract investment, state governments that are risk-averse would be forced to take proactive measures to restore communal harmony so that its state becomes

184 Incentives for Regional Development

attractive to the private sector. If communal harmony does not work as an election agenda in democracies, certainly industry location would! The intention of this suggestion has not been to nullify the effect of infrastructure incentives on firm location decisions, but rather to demonstrate that market-based approaches exist that show how firms can contribute socially to the development process in those communities in which they locate.

8.6

Concluding remarks

Finally, when all is said and done, we have to recognize that it is laissez-faire that has worked in the USA, as in many other higher-income countries. The American EZs are the quintessential spirit of laissez-faire, recognizing, of course, that the state is needed to deregulate as much as to regulate. In India too, the economic liberalization of the 1990s has provided adequate support for laissez-faire and special economic zones (SEZs). In fact, there are some proposals to merge the SEZs and growth centres in India. As pointed out in Chapter 2, the thrust of such experimental programmes is that these areas are sufficiently deregulated in terms of labour laws and bureaucratic delays can be avoided, and their effectiveness put to the test. The question is whether the experimentation with geographically targeted SEZs and growth centres in small pockets will lead to a generally competitive atmosphere throughout the Indian states. Since the geographical area of these growth centres is very small, it is doubtful if they can stimulate development throughout the region. Further, it is a related question as to whether pockets in many truly distressed areas have been targeted so far. It is often said that firms would not locate in the truly distressed areas. The statute pertaining to the GCs in India further specifies that areas can be designated as GCs ‘as long as they have the potential’. Note that such judgements are quite subjective. Every area has a skill for producing some good or service. In fact, programmes such as SEZs or GCs can be a way of discovering the core competence of the area and targeting activity accordingly. What is desirable is that states be given the incentives to compete with each other not necessarily in terms of taxes or infrastructure, but as overall places to reside and do business. Only when Indian states become entrepreneurial in their quest for jobs, economic opportunity, public services, and offer political and social stability, and a corrupt-free participatory environment for public, can they create a climate conducive for prospective households and businesses. These policies, of course, amount to treating the states as entrepreneurs that are in the business of wooing consumers, much required for their viability. We have to wait and see if Indian sub-national governments can do this effectively. This book has laid out some ways in which this can be achieved.

Notes 1 Regional Development Incentives in the United States and India 1 The learning region offers a new perspective on the dynamics of change, which shape the economy. It is based on the premise that successful firms, as well as governments, are those that have control over and access to flows of information and knowledge of technologies, markets, organizational and managerial practices. There are many factors that together constitute learning regions: regional innovation policy, geographical clusters of collabourating firms, and research centres in the innovative potential of regions (see Boekema et al., 2000). 2 In other large federations such as Australia, a strong lobby has only emerged now in favour of decentralization (Bagchi, 2003). 3 This is noteworthy in view of the fact that even states in the United States have not taken any bold steps to stop the tax incentive war. 4 Convergence implies that, in a steady state, poorer regions/countries can be expected to grow more rapidly than their richer counterparts. Because of capital shortage, the rate of return to capital in poorer regions is higher, which encourages the flow of capital to these regions. A second reason why we expect convergence to occur is because poorer countries/regions need not reinvent the wheel and can imitate the technological changes adopted by the richer regions. Some literature classifies the phenomenon more finely and distinguishes between unconditional and conditional convergence and divergence (see Singh and Srinivasan, 2002). 5 They sample 75 zones throughout the country. Given considerable variability across state enterprise zone programmes, in their analysis, all states receive equal weighting. They undertake an analysis of 104 Ohio cities with and without zones. 6 Consumers’ surplus is the savings a consumer realizes in the market when the actual price of a good s/he purchases is lower than what s/he is actually willing to pay. 7 This is reasonable to assume for the following reasons: ● ●

Wages (adjusted for occupation) across areas are usually cost-of-living adjusted. Although rural–urban wage differentials exist, in the long run, they are equalized because of migration, as in Todaro’s model.

2 Impact of Tax Incentives on Economies: Analytical Framework 1 The reader may note that this chapter specifically analyzes the effect of tax incentives offered as part of place-oriented programmes. Further, note that in the book, evidence from tax incentives is provided only from the USA, although the general framework provided in this chapter is tested using data from India’s GCs as well, in Chapter 7. 2 This means that we can replace EZ in the model of this chapter with GC, as long as tax incentives are offered in GCs. Note that in India, in the pre-99 period, tax incentives were being offered in GCs. 185

186 Notes 3 Note that the EZ refers to any place-oriented programme that creates jobs through the offer of tax incentives. 4 Note that the capital market can be in equilibrium even when differing amounts of capital are used as may be implied by KZ/LZ KY/LY. 5 The section on disequilibrium in the labour market elaborates on why unemployment exists in the EZ. 6 Note that while this may not be a very appropriate characterization of unemployment in developing countries such as India, it is an appropriate feature of unemployment in developed countries such as the USA that have generous unemployment compensation programmes. In India, unemployment compensation does not exist as a policy. However, in developing countries such as India, social support networks (consisting of immediate and extended family and friends) decrease the costs of unemployment and they increase the reservation wages of the unemployed. 7 In the American context, they could be recipients of unemployment compensation or welfare payments such as Aid to Families with Dependent Children (AFDC), currently being phased out, with former President Clinton’s programmes ‘to end welfare as we know it’. The literature shows that policies such as unemployment insurance and AFDC have the effect of increasing reservation wages (Feldstein, 1978). In the context of countries such as India, the unemployed poor could be recipients of grains under the Public Distribution System meant for households below the poverty line. 8 Note that we sum over population rather than the labour force, as we usually do when calculating the unemployment rate. This is because we are also making an attempt to understand what factors cause an individual to be in or out of the labour force. This is determined by the difference between market and reservation wages. For instance, at very high wages, even persons outside the labour force (such as housewives) may become willing to work, and become part of the labour force. Hence population, not the labour force, is the appropriate measure to sum over. 9 Alternatively, rather than measuring the reservation wage as a single point that I actually measure in the empirical work, the reservation wage can be identified as a locus of points at various hours of work, consistent with neoclassical labour theory. This is because the reservation wage could be declining with the hypothetical hours of work that is offered at the ‘new job’. One could expect that the reservation wage increases with additional hours of work, to compensate for leisure lost. In the literature, there are few instances in which the hours of work as well as the reservation wage are taken into account in job search. Blau (1991) develops a model which predicts that a low-earnings job might be accepted if the hours were low also, and a high-earnings job might be rejected if the hours were also high. Empirically also he finds that workers are clearly not indifferent between alternative combinations of weekly hours and earnings. 10 The important point here is that if hours are flexible and chosen by the individual, they will continue to increase work hours until, at the margin, the reservation wage the market wage. In this case, the economic rent is zero. An alternative assumption is that hours of work are not flexible. Under this assumption, there is an economic rent. Given that the number of hours are fixed per week at 40 and the individual’s reservation wage is less than the wage, there is economic rent. Here I rely on the fixed hours hypothesis as a possibility, and so economic rents do accrue to individuals. This corresponds to the assumption of the reservation wage for a full-time job (40 hours a week) in the empirical work, where the

Notes 187 reservation wage is measured as the lowest wage one is willing to take home as pay for a full-time job (as in the Panel Study of Income Dynamics (PSID)). 11 One should not misconstrue this observation as leading to the finding that the reservation wage and unemployment rate are simultaneously determined in the model presented here. There is, however, some literature that treats this relationship as indeed being simultaneous (see Lancaster and Chesher, 1984; Kiefer and Neumann, 1979). In this model, the cause of unemployment for individuals in the highunemployment area is high reservation wages, but this is relative to the market wage. 12 We should note that we are comparing firms with high capital–labour ratios to ones with low ratios because the property tax abatement is an incentive to capital (see Chapter 4 for a description of tax incentives in Ohio for instance). 13 Note that dPK is endogenous because of the subsidy to capital. The price of labour (wages) is considered exogenous to the model. Note that although the model considers a subsidy to labour, this subsidy is considered in the model as a subsidy to the good produced by the EZ area firms (see section 2.5.3).

3 Competition Among American States: Evidence from Illinois Enterprise Zones 1 Bartik’s results showed that a shock that permanently raises a metropolitan area’s employment by one per cent reduced the area’s long-run unemployment rate by 7/100ths of one per cent (based on micro data results) or 6/100ths of one per cent (based on aggregate data results) (Bartik, 1991, p. 95). His estimates also suggested that long-run labour force participation rates (which, in turn, would have hysteresis effects) would increase by 14/100ths of one per cent, as a result of job growth. Although these effects may appear small, their persistence makes them important. 2 The City of Chicago is located in Cook County. 3 In Ohio’s programme, as I explain in Chapter 4, job creation is not tied to the issuance of building permits as it is in the case of Illinois. 4 The only study which throws some light on other tax costs in the state of Illinois is by Fisher (1992). This study used some hypothetical cases. It found that average investment tax credits (ITC) and job creation tax credits (JCTC) in the state of Illinois in 1990 would be $9,792 state-wide, with a maximum of $22,203 in a depressed area, for an average manufacturing firm making an investment of $1 million in a new facility. 5 The source of data on property tax abatements by enterprise zone in the state of Illinois was Property Tax Statistics, published by the Illinois Department of Revenue. 6 I am thankful to Peter Fisher at the University of Iowa for suggesting this illustration. 7 In Chapter 5, I provide some evidence from Ohio’s EZs of the costs of public services to firms, when compared to abatement costs and property tax revenue. 8 Tim Bartik suggested the possibility of these two scenarios and I am thankful to him. The source of data on jobs created by industry was the IL DCCA. I also obtained from the current Illinois DCEO, cumulative zone accomplishments for the period 1992–2004. There was, however, no information on the proportion of jobs in the zones created by new or expanding firms as against those created by relocating firms. 9 Note that Chapter 2 (equation 2.10) discusses net benefits bij wwr. Note, however, that E (earnings) of equation (3.1) may or may not be equal to w (wages).

188 Notes

10 11

12 13 14 15 16

17

18 19

Usually, wages refer to efficiency wages (being equal to the marginal product of labour), whereas earnings used as a proxy for wages here, includes other benefits from a job such as travelling, dearness and other allowances. NB (i) in equation (3.1) is the empirical counterpart of bij. Average earnings were used for jobs assigned to industry, and median earnings for jobs assigned to occupations. Since earnings (Ej) were expressed in annual terms and reservation wages in hourly terms, Hj and Wj were used to convert the estimated hourly reservation wages to weekly and annual terms, respectively. Only then could annual earnings be made net of annual reservation wages. Tim Bartik suggested this, in personal communications. I am grateful to him. I am thankful to Tim Bartik for suggesting this. In Ohio’s EZs, there was no need for this, since data on payroll were available from the Ohio Department of Development, as I discuss in Chapter 5. This survey is conducted by the Institute for Social Research, Survey Research Centre, at the University of Michigan in Ann Arbor. In Chapter 5, I use reservation wages, estimated from Sridhar (1998), from a timeseries cum cross-sectional sample over the period 1984–7 (when the reservation wage question was asked in the PSID) whose methodology makes corrections to econometric problems that can arise with selection of an unemployed sample. The estimated reservation wages were expressed in terms of dollars and cents per hour (for 1987). Since the abatements and earnings from jobs were expressed as dollar amounts for the year 1990, the hourly reservation wages were converted to weekly wages by using the weekly hours worked by industry, from Employment and Earnings, published by the US Bureau of Labor Statistics. In order to convert the 1987 weekly reservation wages to annual reservation wages for 1987, the Current Population Survey (CPS) was used. Since the CPS did not provide data on number of weeks worked by industry, cross-tabulations were made between the relevant variables – number of weeks worked and the detailed industry codes. Then a weighted average was calculated for average weeks worked by industry for 1987. These average weeks were used to convert the weekly reservation wages to annual reservation wages. The annual reservation wages (same for all jobs in a zone) for all jobs were subtracted from the annual earning (the earnings are obtained by industry for all jobs) in those jobs, to obtain total net benefits for each zone for 1992. The resulting figure represents net benefits from jobs in a zone, assuming that the jobs are ‘new’. The data did not permit us to distinguish between these types of workers. The enterprise zone contract is an agreement between the firm and the local government regarding various aspects of commitments to be made by firms – with respect to investment, jobs created, and jobs to be held by local unemployed and zone residents.

4 Impact of Tax Incentives on the Unemployment Rate: Evidence from Ohio 1 According to the Ohio Department of Development, three counties (Columbiana, Geauga and Holmes) were the only ones in the state that had not applied for zone designation as of 1997, even though they were potentially eligible, based on their distress criteria. Two of these counties have documented use of the Community

Notes 189

2

3

4 5 6

7 8

9

Reinvestment Area (CRA) programme under which part or all of these counties can offer tax incentives to firms. That leaves one county in the state that did not provide tax incentives to firms, as of 1997. For one thing, labour markets for college-educated workers are more likely to be regional or national than are labour markets for those with less education. For another, even if job opportunities exist beyond the local area, insufficient information prevents those with less education (and skills) from having access to those jobs. For instance, for a janitor in Pella, Iowa, to find out about employment opportunities in the eastern United States is like looking for the ‘proverbial needle in a haystack’ (Ehrenberg and Smith, 1994). Although such information is now possible to access with the growth of the Internet, they are applicable only to those with skills above a certain level. For instance, janitors might not know how to access the Internet. The assumption of this model is that non-zone areas are low-unemployment areas. Empirically, it is also easy to imagine that non-EZ areas (NEZ of the theoretical model) have a natural rate of unemployment (which is close to full employment). This assumption is consistent with EZ programmes in most of the states of the United States, which require high unemployment as a criterion for zone designation. However, in Ohio, where low-unemployment areas are also designated as zones, this is not the case (see the section on criteria for zone designation in Ohio). But this does not affect the model which predicts that tax abatements to labour markets in equilibrium (those with low unemployment, the limited authority zones in Ohio) result in distortions. Real property refers to any real estate, and buildings on it. Personal property refers to all tangible personal property, such as machinery, equipment, and inventory. Note that unemployment is fundamentally a disequilibrium phenomenon. At this point, it is not clear if a model is estimated in reduced form only when market clearing occurs (as when demand for and supply of labour meet). It is possible that special techniques are used in the labour literature, for estimating a disequilibrium model such as the one with unemployment (Quandt and Rosen, 1988). I am thankful to Indira Rajaraman for pointing to this in the context of Chapter 6. That is a potential extension to this work. Based on data available, I was able to control for the CRA in addition to the EZ. Note that high unemployment is a criterion for zone designation in Illinois EZs as well (see Chapter 3). However, there, the treatment effects (simultaneity) problem does not arise. Remember I do not estimate unemployment there as a function of a tax incentive dummy, but merely perform a benefit–cost analysis of that programme, after estimating reservation wages, analogous to Chapter 5, which performs a benefit–cost analysis of Ohio’s EZ programme. I looked at the literature to see not only how treatment (here, the tax incentive programme) affects the unemployment rate, but also how its duration (how long the treatment has been in effect) impacts this outcome variable of interest. I found the following from a literature search on treatment effects models with regard to how the duration of the treatment has been handled: i. Some literature (for example, Ashenfelter, 1978) considers the effect of the treatment by including explicit year dummies for several years previous to the treatment. This is one way of handling duration, being a before-and-after effect of the treatment (which in this case was the effect of a post-schooling training programme on participants’ earnings).

190 Notes ii.

iii.

10

11 12 13

14

15

Ham and LaLonde (1996) estimate an employment rate hazard model for participants in an employment training programme. In the hazard model estimation, the duration of employment, unemployment spell and their squared are used as independent variables along with other control variables. Korenman and Neumark (1990) employ the treatment effect dummy (which, in their case, is marital status of individuals) as well as the duration of the effect (number of years of marriage) on earnings.

Thus there appears to be some literature that treats the duration of the treatment as an independent variable. Others apparently treat this kind of model as a hazard model for which, however, a major extension of this research is needed. The duration for which an area is an EZ can vary when they either get designated or expire early. In the sample used for empirical work in this chapter, expired zones are excluded. So only factors determining early and late EZ designation determine the duration of tax incentive programme. See the section on the definition of variables to see how the duration variable is defined for zones and non-zones. But in the context of this static model, distinctions between short-run versus long-run responses cannot be made. In the section on results, I demonstrate how to calculate the impact of the tax incentive programme on the unemployment rate, using estimates on the tax incentive dummy and the duration variables. It is possible that the number of employed persons ‘at work’ could be understated, or, unemployment overstated, in self-reported data because persons who have irregular, casual, or unstructured jobs sometimes report themselves as not working. However, other persons such as those on long sick or maternity leave erroneously do report themselves as being ‘at work’, thereby overstating the number in the category. It is assumed that the combined impact of these errors in selfreporting on total employment (or unemployment) cancel out one another. The estimates in the Census data for each person or housing unit are obtained from an iterative ratio estimation procedure resulting in the assignment of a weight to each sample person or housing unit. The procedure used by the Census Bureau to assign weights is performed in geographically defined ‘weighting areas’ having a minimum sample of 400 persons. Within a weighting area, the procedure is applied to these groups: 17 household-type groups; groups with a sampling rate of one–two; groups with sampling rate less than one in two; householders/ non-householders; 180 aggregate age-sex race-Hispanic origin categories. The weighting is done in four stages as follows: The first step is to assign an initial weight to each sample person record. This weight is approximately equal to the inverse of the probability of selecting a person for the census sample. In stage II the stage I adjusted weights are adjusted further by the ratio of the complete Census count to the sum of the stage I weights for sample persons in each stage II group (groups with different sampling rates). In stage III, the stage II weights are adjusted by the ratio of the complete census count to the sum of stage II weights for sample persons in stage III group (householder/non-householder). At stage IV, the stage III weights are adjusted by the ratio of the complete census count to the sum of stage III weights for sample persons in each stage IV group (age/sex/race/hispanic origin).

Notes 191

16

17

18

19

20

21

The weighting procedure for housing units is essentially the same as that for persons; only vacant units are treated differently. The procedure for occupied housing units is done in 4 stages (same as for persons), but that for the vacant units is done in a single stage (with three categories: vacant for rent; vacant for sale; and other vacant). This weighing ensures representatives of the sample used in the Census. The digital map of Ohio’s EZs was obtained from the Ohio Department of Development, and a digital map of Ohio’s Census block groups was obtained from the Centre for Mapping at the Ohio State University. The overlaying of the two maps was done using GIS (Geographic Information Systems) software, ARCVIEW. There could be divergence between the prevalence of vacant commercial and/or industrial facilities and that of vacant housing units. In areas zoned for industrial use, where some units are vacant, for instance, there are no housing units at all. For these areas, this measure would indicate no vacant housing units, meaning that the area doesn’t need zone designation, even though it might. At the other extreme, in areas zoned for residential use, if a large number of vacant housing units exist, then this measure would indicate that the area needs rehabilitation or could be used for other purposes, deserving tax incentives. In block groups that contain a combination of housing and commercial facilities, I expect the prevalence of housing units to be a fairly good measure of the existence of vacant industrial or commercial facilities in the area. Thus, the chosen measure is a good proxy for its intended purpose in areas that contain mixed land uses. A gross measure of in-migration would be to use the percentage in migration of households or population over the period 1985–90 into the area, since the period is plausible for in migration to have occurred after the EZ programme came into being in the state in 1982. However what are used as measures of in-migration in the estimation are net measures that reflect population loss. I did not take into account CRAs certified after 1994 because I study the overlap of CRA and EZ areas as of 1990. Since data on unemployment rate were available for census block groups in 1990, I was interested in examining whether unemployment rate is affected by the presence of a tax incentive programme at that time. It might have been clearer if I had included a separate dummy for a CRA and another for an EZ. However, the reason why I combine the CRA with the EZ dummy is because I knew which (block group) areas contained only EZ (because the EZs were mapped into GIS). But I did not know which areas were only CRAs in terms of census block groups. Map of CRAs had not been prepared by the Ohio Department of Development at the time this research was completed. All I knew was this: when the EZ and CRA areas coexisted, when they overlapped and when they did not. This was based on my discussions with the 282 CRA administrators in the state of Ohio. It should be noted that zones that are once certified expire, if the project the community is interested in, requires zone designation only for a short period of time, or for other reasons. The fact that some areas were EZ at the beginning of the EZ period in the state (post-1982), but were not subsequently, could greatly complicate the analysis. This is because if, for example, say an area is an EZ from 1982 to 1989, but is so successful, that it is no longer an EZ. In such a case, if a regression of unemployment rate were to be done on the tax incentive status of areas (as it is done here), we would miss the true impact from 1982 to 1989. This would, however, be the case only if the expired zones were included as non-zone areas in the data set and their zone status expired before 1990. Then the duration variable

192 Notes

22 23

24 25 26 27

28

29

30 31

32

would show the wrong effect for them that non-tax incentive areas have low unemployment rates, although it was the tax incentive that reduced it. However, it should be noted that if the number of decertifications is small there would be no problem. In the sample used here, of the 322 zones in the state, only eight zones are decertified. I have excluded these zones in my sample. Only those EZs that existed as of June 1996 were mapped into Geographic Information Systems (GIS) by the Ohio DOD. So the exclusion of the expired zones did not pose a problem for the estimation. Part of the problem could also be that unemployment rate of an area is a continually evolving, dynamic phenomenon, whereas the status of the tax incentive programme is a single snapshot of a cross-section. Both the theoretical model that is developed in the book and the empirical model estimated here are static, as explained in Chapter 2. This variable is measured as percentage by multiplying the calculated rate by 100. It is not possible to include the proportion of employment in other industries (the residual category variable) in the estimation along with the other two variables because the three variables would add up to 100, creating a linear combination of variables in the system and collinearity. This category includes executive, administrative, managerial and professional specialty occupations. This refers to technicians and related support, sales, and administrative support, including clerical occupations. The same explanation as for industries applies here. Out of the 11,621 block groups, 176 block groups had missing data for at least one of the independent variables and so 11,445 census block groups were used in the estimation. It may be noted that estimation was performed only on those cases for which no data in any of the independent variables were missing. The City of Toledo EZ has 367 block groups, one of which had an unemployment rate of 83.08 per cent. There were 37 unemployed persons in this block group that had 43 persons in its labour force. One of these is in census tract 101, block group number two in Allen County, part of the Village of Bluffton EZ. The block group had 338 persons in the labour force and no unemployed persons, making its unemployment rate zero. Note that the small size of the labour force in the block groups discussed in this and the previous note does not imply unreasonably high or low unemployment rates. The average unemployment rate (in Table 4.1) demonstrates that the unemployment rate data are stable at the block group level. The equations were all estimated in the econometric software, LIMDEP. The impact of the EZ is calculated as the coefficient on the tax incentive dummy plus the impact of the duration of the tax incentive (the sum of coefficients on the duration variable and its squared) when it has been in existence for a year. Impacts for other years are calculated in a similar manner. In order to check whether the quadratic in duration is properly specified, I categorized the residual from this regression. I computed means for four groups according to values of duration (duration one year, duration between one and three years, three and five years, and five and seven years – seven years being the sample maximum for duration). When I did this, the following were the means of the residuals for the various groups respectively: 0.08, 0.26, 0.17, and 0.70. Thus I found the residual means varied randomly across the various groups. For this reason, there appeared to be no monotonicity to these means of the residuals classified across various duration groups. This shows that the duration

Notes 193 equation is properly specified. I am thankful to Don Haurin at the Ohio State University for pointing to this. 33 The effect of the EZ on the unemployment rate is calculated in a similar manner as in the previous estimation. It is calculated as the sum of the coefficient on the tax incentive dummy and that on the duration multiplied by the time period for which one is interested in knowing the impact.

5 Benefits and Costs of Regional Development: Evidence from Ohio’s Enterprise Zone Programme 1 Note from earlier chapters that the reservation wage is measured as the response to the question, ‘What is the lowest you would be willing to take home as pay in any job?’ 2 But for this national panel data set, I found that no secondary data exist on reservation wages, with the exception of the National Longitudinal Survey (NLS), published by the Centre for Human Resource Research at the Ohio State University, which is also national. NLS data are reported in panels for various age cohorts, making it difficult to generalize reservation wages for the labour force in all age groups. Secondary data on reservation wages are even more difficult to find exclusively for the Ohio labour force. So I used panel data from the PSID estimated in Sridhar (1998) to impute reservation wages for Ohio. 3 I did this with the help of GIS (Geographic Information Systems) software, ARCVIEW. 4 I estimated zone unemployment rates according to the procedure recommended by the Ohio Department of Development, as I explain in Chapter 4. According to this procedure, I first estimated the unemployment rate for all the counties based on the ratio of unemployed to the total labour force aggregated in the 1990 census at the county level. I then estimated a similar ratio for the block groups comprising the zones, obtaining a raw unemployment rate for every zone. I took the ratio of zone to county unemployment rates computed in this way. I then applied the zone to county unemployment rate ratio to the county unemployment rates published by the Ohio Bureau of Employment Services (OBES) to arrive at zones’ unemployment rate for 1990. Therefore, the characteristics, the reservation wages, net benefit and benefit–cost ratios that were estimated for the zones were for 1990. 5 Since the number of persons in each age range was reported in the Census, I calculated the mean age by substituting midpoints for intervals. I calculated a weighted average for age based on number of persons in each age group. I included only adults (that is, persons who are above 17 years of age). 6 It may be noted from Table 5.3 that at least one non-EZ census block group had an educated population, with its population having, on average, education up to college degree level. This is the one with a maximum of 16.23 grades (or college degree) completed, being in Franklin county, (the county in which Columbus is located), census tract 7820, block group number six. 7 The reader should compare Tables 5.2 and 5.3 for zone and non-zone areas, with its counterpart in Chapter 6 – Table 6.2, which compares similar characteristics for areas with and without growth centres in India. 8 These data were taken from Honigsberg (1981). 9 Note that in Chapter 3, for earnings from jobs, data from US Bureau of Labor Statistics are used. In the case of Ohio’s EZs, data on payroll was available for jobs, from the Ohio Department of Development.

194 Notes 10 In such instances, local government costs were legitimately counted as being zero. 11 However, note that costs are also incurred even when negotiations between the local government and firm do not result in a contract. Recall that in Ohio’s enterprise zone programme, tax abatements are not automatic upon a firm’s location in the zone, and they have to be negotiated with the local government. It is possible that if the affected School Board does not approve of the abatement to the firm, a contract does not result. Then it is up to the firm to decide whether or not it wishes to locate in the zone. Such negotiating (monetary or time) costs have not been included here as data exist only when negotiations result in a contract. I am thankful to Mary Marvel at the Ohio State University for pointing this out. 12 Note that in the assessment of IL’s programme in Chapter 3, no information, apart from property tax abatements, was available or used, as measures of programme costs. 13 Suppose incentives were given to a firm that threatened to relocate out of the zone otherwise, employment that would have been otherwise lost is referred to as retained employment. 14 A newly created job could be offered to someone that was previously employed elsewhere. However, with newly created jobs, the assumption that an unemployed person is likely to get it is more plausible, rather than in the case of retained jobs. 15 Of the 1,974 firms that located in the state’s 280 enterprise zones over the period 1984–95, there were some data missing for some firms, and, for some others, the data were inconsistent. Instances of inconsistent data were when the earnings were either reported as zero or as being too low or too high for some jobs. I excluded observations that had earnings less than $10,000 a year (around the minimum wage, that is $5.00 40 52), and greater than $50,000 a year (not very realistic for EZ jobs). I also excluded observations that had inconsistent earnings data for retained jobs (for instance, firms that reported earnings for retained jobs when no jobs were in fact retained). I also eliminated all firms that did not receive abatements. Because of these restrictions, I was left with 531 observations in the firm-level data and 143 observations in the zone-level data in this scenario. 16 Here I imposed the same restrictions for the data set of 1,974 firms as I did in scenario one (excluding observations with payroll per job less than $10,000 and greater than $50,000, those firms that did not receive tax abatements and those that relocated from within or outside the state). In this scenario, the number of observations were 575 (higher than in scenario one, where it is 531) because the restriction pertaining to retained jobs was not there. Remember this scenario takes into account only those jobs that are newly created. I was thus left with 575 observations in the firm-level data and 148 observations in the zone-level data. 17 I am thankful to Chuck Adams, now Professor (Emeritus) at the Ohio State University, for suggesting the possibility of this scenario. 18 Here I imposed the same restrictions as I did in scenario one (excluding observations with payroll per job less than $10,000 and greater than $50,000, those firms that did not receive tax abatements and those that relocated from within or outside the state). Apart from those data restrictions, the condition that reduced the number of firm observations in the scenario considerably was that of eliminating observations that had x (that is, total taxes that would have been paid in the absence of the abatement) equal to zero. This condition is necessary because x

Notes 195 appears in the denominator in the calculation of dy in the expression for elasticity: dy dx

x

y 0.1(0.3)

dx dy x 0.3(0.1) y The number of observations differs across the two elasticity assumptions because the restriction of deleting observations with payroll per job less than $10,000 and greater than $50,000 that is imposed takes into account only that employment which is attributable to the tax incentive with the two assumptions of elasticity. It should be clear that the employment attributable to the tax incentive will be lower with the lower elasticity assumption, than that with the higher elasticity assumption. Also note that the cost per job is undefined when the total employment attributable to the tax incentive is zero (since cost per job was defined as abatements/jobs). So I defined the cost per job to be missing if total employment was zero. If cost per job is missing, the B C ratio would also be missing and the number of valid observations reduced. Some firms created no jobs that were attributable to the tax incentive according to the lower elasticity assumption, and so a large number of observations were lost this way when elasticity 0.1 (rather than when it is 0.3). 19 The long-run (which may be defined as 10 years or more) elasticity of business activity (here, employment) due to a proportionate change in taxes, for intrametropolitan locations, according to Bartik (1991, 1992) is in the range of 1.0 to 3.0. For any given year, then, the elasticity is (divided by 10) dy dx

x

y 0.1(0.3),

where x refers to total taxes that would have been paid by firms in the absence of the EZ; y is the baseline employment at the site without EZ; dx is the change in taxes paid due to the abatement; and dy is the change in employment because of EZ. I substituted for the values of dx, x and y to obtain employment that was attributable to the tax incentive (dy), based on the two ranges of elasticity reported by (Bartik, 1991). If dy turned out to be greater than actual employment created and retained by the firm or zone, I made dy equal to actual total employment created and retained by the firm or zone. In cases where dy turned out to be less than or equal to actual employment, I used the smaller dy (instead of actual employment) to calculate earnings per job, net benefit per job, costs per job and B C ratios per job. 20 A calculated value of dy being equal to zero for a firm indicates that none of the jobs it created were attributable to the tax incentive (for instance, the case of 0.91 jobs a firm created, Table 5.9). Note that the specific instance of a firm’s jobs not being attributable to the tax incentive does not invalidate the general finding (as found in Chapter 4) that tax incentives reduce the unemployment rate of areas adopting it. 21 It should be noted that at the zone level, the (unweighted and weighted) B C ratios (13.8, with an assumed elasticity of 0.3 and 9.7, with an elasticity of 0.1) are lower in scenario three than in scenario one (25.6), as we would expect. At the firm level, the (unweighted and weighted) B C ratios are higher in scenario three

196 Notes (27.7 with the 0.3 elasticity) than in scenario one (24.7), and this may seem counter-intuitive. Some observations can be made regarding the firm-level B C ratios in scenario three. First, although scenario three is more restrictive, the number of firms is only 198 in scenario three whereas there are 531 firms in scenario one. Thus the sample in scenario three reduces to a subset of those firms in scenario one, whose employment creation is attributable to the tax incentive. Second, it should be noted that the B C ratio is defined as the ratio of net benefits to costs per job. The two scenarios themselves differ in the magnitude of the employment that is created. But what causes the higher ratios in scenario three compared to scenario one, is the magnitude of earnings and net benefit from employment created by the firms in scenario three. This is not inconsistent with the assumptions of the two scenarios. The firms in scenario three located in zones that had lower reservation wages and so net benefits from employment created by firms in scenario three were greater. Thus firm-level B C ratios in scenario three, although the scenario is based on a more restrictive assumption, are higher because of higher net benefit. This is true although the costs were more or less the same for firms in both scenarios one and three, and employment created is less in scenario three than in scenario one. 22 The proportion of zones with B C ratios greater than one could not be calculated with the weighted averages. This is because the weighted average is calculated (it is a single figure) for the entire distribution of zones taking into account zones in each percentile of the B C distribution (which is obtained by taking the ratio of net benefits from employment to the costs for every zone). 23 I calculate efficiency loss as the sum of the abatements, infrastructure costs and other incentives to firms in zones, that had B C ratios less than one in all scenarios. I divide the total efficiency loss by the number of zones in each category to arrive at efficiency loss per zone. 24 One could expect, for instance, that the efficiency losses in scenario one would be the least and that those in scenario two would be the highest because of their assumptions about job creation and retention. However, here efficiency loss is measured as the total amount of abatements and other incentives provided to firms in zones in which the B C ratios were less than one. The amount of abatements varies quite independently of jobs. I find that the abatement amount is mostly related to the amount of investment firms made in the zone, which of course could create differing amounts of employment depending upon the capital or labour intensity of the firm.

6 Impact of Growth Centres on Unemployment and Firm Location: Evidence from India 1 This may not be a uniform VAT, but could be only a sub-national VAT, given the heterogeneous tax structures across the Indian states. 2 Net benefit, or economic rent from a job, as defined in Chapter 1, is the extent to which an actual wage is higher than the wage at which a person is willing to accept a job (which is referred to as a person’s reservation wage). 3 Incidentally, BIMAR in Hindi means sick! 4 We defer an examination of this question until the final chapter. 5 Note that there are two variables (hours of work, measure of flexibility in the job) that are excluded from the demand equation (included in the model), whereas

Notes 197

6

7

8 9 10

11

12 13

14

15

there are two endogenous variables (wages and the quantity of labour demanded). Since the number of excluded exogenous variables in the demand equation is greater than the number of endogenous variables less one, by the order condition, the demand equation is over-identified. There are two exogenous variables excluded in the supply equation (included in the model – technology and capital) as well, and the number of endogenous variables is two. So by the order condition again, the supply equation is over-identified. I have checked, through the algebraic substitution of terms, the final reduced form equation for unemployment rate. Note that there is no need to estimate wages for addressing research objectives in this chapter. Also note that when I derive the reduced form equations, I assume linearity. This is based on assumption in the past literature on the subject (Pantuosco and Parker, 1998). Chapter 4 argues, based on neoclassical theory, that the individual’s unemployment status is a function of the difference between the actual wages offered and his and her reservation wage. This is aggregated in order to obtain the area’s unemployment rate. In this chapter, the framework from standard labour theory is used. The two approaches yield the same reduced form equation for wages. Also, note that unemployment is fundamentally a disequilibrium phenomenon. At this point, it is not clear if a model is estimated in reduced form only when market clearing occurs (as when demand for and supply of labour meet). Special techniques are used in the labour literature, for estimating a disequilibrium model such as the one with unemployment (Quandt and Rosen, 1988). I am thankful to Indira Rajaraman for pointing this out. That is a potential extension to this work. For estimating such a model, however, detailed data on wages that are required, are not available at the district level for India. I have checked this through algebraic substitution. It is only very recently that the concept of part-time work has slowly began to gain acceptance in India, as revealed by opinion polls. Telecommuting is even more rarely practised. For some preliminary evidence on telecommuting and flexible work patterns in India, see (Mitter, 2000; Irani et al., 2000). Two-stage least squares (2SLS) is the recommended procedure for estimating overidentified equations. If we were to use Indirect Least Squares to estimate such equations, remember that we will not get unique values of structural coefficients based on reduced form equations. I report results from OLS and 2SLS estimations. Note the similarity of this equation with that in Chapter 4, equation (4.2). Note that while the proportion of employment in manufacturing and services determine the unemployment rate of an area, there is no reverse causation from unemployment rate to the proportion in manufacturing and service occupations. We would expect the manufacturing and service base to be determined by exogenous factors such as natural resources available, the skills of the population, and the extent of integration with international markets. I have excluded from the estimation, districts in Indian states that did not have even a single GC, because these states could be systematically different from the rest of the sample. See endnote 25. The GC dummy is defined according to its status in 2001. To be consistent, I have defined the duration variable also as of December 2001. Also, note that whether or not a GC is completely operational on the day it is certified or designated is irrelevant. It is only the idea of marketing the area as a good place to do business, if not the actual incentives that could make a difference to prospective firms. This

198 Notes

16

17

18 19 20

21

22

23

24

25

idea supports the construction of the duration of GC variable since the day of its certification. Incidentally, the issue of optimality of the GC’s duration was highlighted by Gorakhpur Industrial Development Authority (GIDA) officials during our discussions about the GC’s effectiveness. Remember that the effect of GC, its duration and its squared, on the unemployment rate, over a period of 12 months (or one year), can be computed from their coefficients as follows: Coefficient on GC dummy (Coefficient on duration of GC 12) Coefficient on duration squared 122). If we wanted to know the effect of the GC over a period of two years, we substitute 24 (months) instead of 12 as above, and so forth. See Chapter 4. These workers are those that were engaged in any economically productive activity for 183 days, or six months or more during the year. These workers refer to those who worked for less than 183 days, or six months, during the year. This becomes tricky. If we agree that willingness to work itself depends on the wage rate, we have to accept that at very high wages, even those working at home or dependents may be willing to work! So if wages were to be high, taking into account only main workers at any given point in time could be an under-estimation of those in the labour force. I am thankful to Govinda Rao for pointing this out. In India, affirmative action policies do provide reservation to SC and ST and other minority groups in education and employment. Since the effect of these policies on their skills and employment is not empirically proven, however, it is an open question as to whether or not higher proportions of SC and ST cause their areas to have higher levels of unemployment. During the decade 1991–2001, several new districts were created in the country, and were formed out of existing districts. There were also three new states (Jharkhand carved out of Bihar, Chattisgarh carved out of Madhya Pradesh, and Uttaranchal out of Uttar Pradesh) created during the decade. For new districts and those in these new states, I have assumed that the data for the parent district (from which it was carved) hold good. Given the fact that these districts and those in the state have been carved only relatively recently (in 2000), the assumption is certainly reasonable to make. I calculated weighted average age after excluding persons below 15 and above 65, since we are concerned about the effect of age on the unemployment rate, only for those eligible to be in the workforce. Data on workers in manufacturing and services in the 1991 Census of India were broken down by the categories rural–urban and male–female, so I aggregated these categories in order to obtain total employment in the respective categories. In all, these 543 districts are in 27 (out of 35) Indian states and union territories that contain GCs. I did not include in the estimation states and union territories that did not contain GCs (these being Chandigarh, Uttaranchal, Delhi, Sikkim, Daman and Diu, Dadra and Nagar Haveli, Lakshadweep, Andaman and Nicobar Islands). Districts in these states could be systematically different from their counterparts in states containing GCs. The objective is to compare only districts that are equally likely to contain a GC, which would not be met if (districts in) states with not a single GC were to be included. Only one state with GCs (Jammu and Kashmir) has been excluded from the estimation because of the lack of 1991 data for districts in this state. The 1991 Census was not held in Jammu and Kashmir because of the unstable law and order situation in the state at the time. The Census of 2001 was not held in Kutch district,

Notes 199

26

27 28

29 30

31

Gujarat, because of a natural calamity. These and a few other districts in various states have been excluded from the estimation, due to the lack of complete data for all variables. The final estimation is based on a sample of 543 districts for which all data were available. It is to be noted that the highest average age in any district is 36 years. This does not mean that there are no persons in any district that are above this age, but rather that this is the maximum of the weighted average age calculated for all the districts, based on information regarding the number of people in each of the age groups (the weights) in the districts. The interested reader should compare this with Tables 5.2 and 5.3 in Chapter 5. Although my interest is to explain only those variables that are statistically significant, I probe into the statistically insignificant GC variable to explain it, as that is one of the primary objectives of research in the chapter. The signs on the duration (of GC) variable and its squared are interesting when combined with that for the GC dummy. While the negative sign on the GC dummy implies that areas with GCs can expect to see a reduction in their unemployment rate, sign on the duration variable indicates that the longer the GCs are in place, the higher would be their unemployment rate. This outcome, if it had been significant, points to the need for well-defined, predetermined sunset provisions of the programme when the objectives of industrialization have been met. This, if it had been significant, would have been consistent with the results reported in Chapter 4. The next chapter discusses, in detail, the field visits that form the basis of this finding. This finding is arrived at on the basis of data regarding the average size of a plot in the growth centre (roughly five acres, based on my field visits), and the average number of plots developed (with infrastructure), which is 200 (Table 6.5). Based on this data, the developed area of growth centres is, on average 1,000 acres, or four square kilometres. In most instances, wherever the growth centres were located in Urban Agglomerations (UAs), there was data available on land area of UAs to enable us to assess the proportion GC’s land area formed to total geographical area of the UAs. For instance, in the Hassan Urban Agglomeration (UA), the GC accounts for 25 per cent of the UA area. Note, however, that the unit of estimation is the district, not the UA, and more often than not, GCs are not situated in UAs, consistent with their objectives. I used White’s general test of heteroscedasticity to check for possibility of nonconstant variance of error term with the independent variables. I had to reject the null hypothesis of homoscedasticity. To detect specifically which exogenous variables were associated with the problem, I did a graphical plot of residuals against each of the variables. I found proportion of SC and ST population, proportion of manufacturing employment and those in services, to be causing increasing error variance, as in the assumption E(u2i 2Xi). I transformed all variables using 1 兹Pr oportionSCST *Pr oportionManufacturing *Pr oportionServices as the weight, and performed GLS (Generalized Least Squares) estimation of the transformed variables in the original model. I examined the ratio of the OLS variance to GLS variance, to examine how big of a problem heteroscedasticity is. The largest OLS error variance (of the proportion male variable) was only two times that of the GLS variance, leading me to conclude that the problem was small enough to be ignored.

200 Notes 32 I find, based on a Chow test on the full sample, that pooled estimation is not justified, and separate estimations have to be performed for sub-samples containing high- and low-unemployment areas. This is, in fact, apparent when we consider Table 6.4: the significant coefficients are different across the two samples, making the two estimations completely different regressions. The impact of the GC dummy is the same in both the estimations, being insignificant. 33 Currently, the programme is financed by funds from the centre leveraged by funding at the state level. There was a proposal to transfer the programme entirely to the states in 2003, but it is now not clear whether the programme is being scrapped (to downsize the government), or being transferred to the states, or will be continued in its present form. 34 The Bawal (Rewari) GC in the northern Indian state, Haryana, recently advertised itself in the Economic Times, India’s leading business newspaper, stating that it is the best global destination for businesses to invest! 35 The monetary data in Table 6.5 are in deflated US dollars (with the base being 1982–4 100). I have used the CPI-U since that covers nearly 87 per cent of the population. First the INR data were converted to US dollars by using the midDecember exchange rate reported by the Reserve Bank of India between the INR and the US$. Then the data in 2004 current US$ were converted to constant US$ (with 1982–4 100) by using the CPI-U deflator for November 2004, reported by the United States Bureau of Labor Statistics to be 191. The percentage change in the CPI-U from the base (1982–4) to the current year (2004) is thus an increase in inflation of 91 per cent over the period 1984–2004. This means that the 2004 US$ data is inflated by 1.91 times its value in 1984. The US current $ data are divided by 1.91 to arrive at constant US$ data for India’s growth centres. 36 Note that this comparison, per se, does not indicate anything about the cost-effectiveness of tax vis-à-vis infrastructure incentives as they are dependent on the generosity of tax incentives, required expenditure on infrastructure, and the capital or labour intensity of firms that locate in these areas. In fact, a related objective might be to study the cost-effectiveness of GCs where they exist. However, that is topic for a different paper and I deal with it elsewhere. 37 Note that a negative sign on the number of plots would indicate that infrastructure in fact deters firms from locating, which is not intuitive, as all firms need one kind of infrastructure or the other. The expectation is that the number of plots will be insignificant or positively affect firm locations.

7 Firm Location Decisions and Their Impact on Local Economies: Evidence from India’s Growth Centres 1 The Indian Institute of Management in Lucknow, India, funded these field visits as part of a seed money project. 2 Based on a map of Hassan GC, I found the distribution of plots allotted to various industries. The weighted average size of these plots was 7.8 acres, the weights being number of firms in each category (receiving plots of certain size). 3 A multi-fuel captive power plant of 200 megawatts (MW) was proposed to be developed by firms themselves. A hydroelectric power project of 24 MW had already been set up by a firm in Gorur, 23 kilometres away from the GC. Twoand-a-half million gallons of water per day for industrial use in the GC was being fed by the Hemavathy river, 18 kilometres from the centre. Finally, 29 kilometres of internal roads had been developed. The Hassan railway junction is just five

Notes 201

4

5

6

7 8 9

10

11

12 13

kilometres away from the GC. Broad gauge conversion of the Hassan–Mangalore (a port city in the western part of the state, see Figure 7.1) railway route was also in progress. An airport eight kilometers from the GC was under development. The telecom department had taken possession of an acre at the GC for catering exclusively to the firms set up there. I made several efforts to get in touch with senior officials in KIADB and the state government to enquire about the reasons why these districts were chosen to be designated as GCs. I was not successful in getting satisfactory responses – or, indeed, any response, for that matter. Note that Table 7.1 refers only to industrial units; services have been excluded. The strong suspicion, however, is that Bangalore (U) also dominates service firm locations. Further, districts that have large number of industrial units are those that tend also to attract service firms, because industrial firms are frequently the clients of service firms. So even if we were to take services into account, the picture presented in Table 7.1 might not be substantially different. The cost of developed land to industrialists in the GC is about US$7 (in real terms) per square metre. Typically, firms pay 10 per cent of plot cost as an application fee and 15 per cent of cost, for development. The remaining 75 per cent of the total cost of the land allotted to the firm is to be paid in half-yearly instalments for 2.5 years. If the unit comes into production within three years from the date of operation, a 20 per cent rebate is given on the cost of the plot. Thus, Haryana’s growth centres demonstrate how the cost of developing the required infrastructure need not always be borne by the government, but can be borne by the firms themselves. This means that the state government only has to incur the cost of acquiring land, confirming that these costs represent the opportunity costs of industrial development. The Bawal GC is about 2.5 hours’ drive west of New Delhi. In parentheses are the districts in which these GCs are located. This information was obtained from the UPSIDC, which is located in Kanpur, UP. I visited Kanpur in August 2001. The objectives of the visit were to obtain primary information regarding designation criteria, tax and non-tax or incentives offered to firms, reliability of data available from DIPP, availability of other data, and discuss the counterfactual, that is, what would have happened without GCs. This is apparently in contrast to the centrally located Indian state, Madhya Pradesh, where land is relatively cheap and hence acquisition is not a problem, or is, at least, manageable with a small compensation payment. For districts in category A, the monetary limit for exemptions was the highest, being 250 per cent of the fixed capital investment. For those in category B, this was 200 per cent of the fixed capital investment, and for category C districts, the relatively affluent ones (for instance, areas such as the capital region, Lucknow), the monetary limit was 175 per cent of the fixed capital investment, according to information from UPSIDC. Of the 290 acres of developed land, 60 per cent represented saleable land. The remainder was covered by roads, power stations and other infrastructure. The costs of these developments, along with a 15 per cent contingency fee and 15 per cent for overhead charges, were collected from entrepreneurs locating there, similar to the system being followed in the Bawal GC (Haryana). This reinforces the idea that the development of infrastructure can be borne by the firms themselves, similar to the concept of impact or development fees charged by US local governments to private developers.

202 Notes 14 This observation forms the basis of the growth centre estimation in Chapter 6, in which the number of locating firms is estimated as dependent upon the existence of the number of developed plots. 15 Recall from Chapter 4 that three to five years could be the optimum period for maximizing the effect of tax incentives on unemployment rate, after which it is preferable that the area stops offering incentives. 16 I had information from SIDA regarding the size of plots and the number of plots in every size category. I calculate a weighted average for the size of plots, with weights being the number of plots in each size category. 17 For reasons of confidentiality, I refrain from referring to the names of these firms. 18 Even when all GCs in the country are taken into account, Satharia is second in terms of plots allotted, second only to Bathinda GC, in Punjab, as of 2001. 19 Alternatively, examples of incentives for labour could be tax credits or exemptions to firms if local, or disadvantaged, or unemployed or poor residents, are recruited by them. 20 Recall that in the pre-GC (pre-93) period, tax incentives were being offered by the various Indian states. 21 Notice that a pre-GC firm does not necessarily imply that the GC infrastructure is ineffective. 22 In general, labour-intensive firms are those with a high labour–output ratio. I did not calculate this ratio in order to make the judgement regarding firms’ capital or labour intensity, but merely report what the firms reported regarding the capital or labour intensity of their output. 23 In contrast to labour-intensive firms, as defined in Chapter 2, capital-intensive firms are those with a high capital–output ratio. The payment of higher salaries is reasonable because of the higher level of skills involved. 24 For instance, in Satharia GC (UP), 321 landowners in the area had lost their land because of the acquisition of land for the GC. 25 Because of the way in which the unemployment rate is defined, note that only when local residents get jobs that the local unemployment rate decreases. This also justifies any local government expenditure on the programme, as in the earlier chapters. 26 Recall that Chapter 3 shows, using data from the EZs of Illinois, that even assuming relocation, as long as firm relocation occurs from a low- to a high-unemployment area, net benefits from employment are higher. 27 I have confirmed this based on discussions with firms that have located in various GCs and informal discussions I have held with employees of the Department of Company Affairs, Government of India. 28 This is not so in the United States where most of the areas have equal access to good transportation facilities and other infrastructure and tax incentives can potentially make a difference across otherwise similar locations. 29 Note that if a region that has a skilled pool of labour, a temperate climate, and its climate is responsible for its industrial backwardness, then policy can do little to help! Alternatively, if a skilled labour force is the constraint, policy can facilitate the building of the required skills, or the matching of jobs with skills available.

8

Lessons Learned from the United States and India

1 Sales tax incentives are relevant because the Indian states in the pre-2000 period had been actively offering sales tax incentives to firms locating in their states.

Notes 203 2 These data are for the urban areas in India. It is possible that analogous, if not worse, problems exist in the rural areas regarding which reliable data are not available. However, if urbanization is an inevitable occurrence of growth, it is important to address these infrastructure problems in the urban areas. If such problems cannot even be addressed in the urban areas, it would be much more difficult to address them in the rural areas! 3 Sridhar and Sridhar (2004) compute the compounded annual growth rate (CAGR) of cell phones for developing economies over the period 1996–2001, and find this to be 78 per cent compared to a mere 7 per cent growth for main telephone lines over the period 1990–2001. 4 This is valid despite their finding that cell phones are unable to offset the losses or declines in the economic output of these countries, which occur because of various exogenous shocks. 5 Many cities in India continue to use the archaic annual rental value (ARV) as the base for assessing property. For various reasons, the ARV method does not facilitate the property tax base as a buoyant source of revenue. First, ARV tends to be subjective regarding rental income that accrues from the property. Next, it does not capture the capitalization in property values that occurs with increasing infrastructure and amenities. Finally, since rent is determined by rent control in many Indian states, ARV freezes property tax revenues. Recently, a large number of India’s cities, including Delhi, have resorted to the unit area method of assessing property. The unit area method is more objective as assessment is based on characteristics of the property (as in models of hedonic prices), and promises to be a more buoyant source of revenue. 6 In India, the states of Maharashtra, Gujarat and Punjab continue to have the octroi. Punjab is actively considering abolishing it. 7 Lack of openness (import substitution and export pessimism), has, of course, been a major problem with India’s industrialization and development strategy, which has been one cause of its historically slow growth of its GDP at 3.5 per cent per annum (termed the Hindu rate of growth). (See Chapter 7 for a description of export orientation of firms located in the GCs of India.) While export promotion should, no doubt, be encouraged as a long-run strategy, immediate problems such as unemployment and poverty do need targeted attention in the short run, in the context of developing countries. Further, we know about the detrimental social effects of these economic problems, as we have discussed earlier.

References

Addison, John T. and W. Stanley Siebert, The Market for Labor: An Analytical Treatment (Santa Monica, CA: Goodyear Publishing Company Inc, 1979). Ahluwalia, Montek S., ‘Economic Performance in States in Post-Reforms Period’, Economic and Political Weekly, 6 May (2000) 1637–48. Alleman, J., C. Hunt, D. Michaels, M. Mueller, P. Rappaport and L. Taylor, Telecommunications and Economic Development: Empirical Evidence from South Africa, mimeo (Sydney: International Telecommunication Society, 2002). Retrieved from http://www. colorado.edu/engineering/alleman/print_files/soafrica_paper.pdf Ashenfelter, Orley, ‘Estimating the Effect of Training Programs on Earnings’, Review of Economics and Statistics, 60 (1) (1978) 47–57. Bagchi, Amaresh, ‘Fifty Years of Fiscal Federalism in India: An Appraisal’, National Institute of Public Finance and Policy Working Paper 2 (New Delhi, India: NIPFP, 2003). Bailey, Martin N., Charles Hulten and David Campbell, ‘Productivity Dynamics in Manufacturing Plants’, Brookings Papers on Economic Activity: MicroEconomics (1992) 187–249. Barron, J. and W. Mellow, ‘Changes in the Labor Force Status among the Unemployed’, Journal of Human Resources, 16 (1981) 427–41. Bartik, Timothy J., Who Benefits from State and Local Economic Development Policies? (Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, 1991). Basu, Susanto and John Fernald, ‘Returns to Scale in U.S. Production: Estimates and Implications’, Journal of Political Economy, 105 (2) (April 1997) 249–83. Blau, David M., ‘Search for Nonwage Job Characteristics: A Test of the Reservation Wage Hypothesis’, Journal of Labor Economics, 9 (2) (1991) 186–205. Boarnet, Marlon and William T. Bogart, ‘Enterprise Zones and Employment: Evidence from New Jersey’, Journal of Urban Economics, 40 (2) (1996) 198–215. Boekema, F., K. Morgan, S. Bakkers and R. Rutten (eds), Knowledge, Innovation and Economic Growth: The Theory and Practice of Learning Regions (Cheltenham: Edward Elgar, 2000). Boisier, S., ‘Growth Poles: Are They Dead?’, in B. Prantilla (ed.), National Development and Regional Policy (Nagoya, Japan: United Nations Centre for Regional Development, 1981). Brunori, David, State Tax Policy: A Political Perspective (Washington, DC: The Urban Institute Press, 2001). Brunori, David (ed.), The Future of State Taxation (Washington, DC: The Urban Institute Press, 1998). Butler, Stuart M., Enterprise Zones: Greenlining the Inner Cities (New York: Universe Books, 1981). Byrnes, Patricia, Mary Marvel and Kala Sridhar, ‘An Equilibrium Model of Tax Abatements: City and Firm Characteristics as Determinants of Abatement Generosity’, Urban Affairs Review, 34 (6) ( July 1999) 805–19. CAOAG (California Office of the Auditor General), A Review of Economic Activity in the State’s Enterprise Zones and Employment and Economic Incentive Areas, Report P-754 (Sacramento, CA: Office of the Auditor General, 1988). 204

References 205 Cashin, Paul, and Ratna Sahay, ‘Internal Migration, Centre-State Grants, and Economic Growth in the States of India’, International Monetary Fund Staff Papers 43 (1) (1996) 123–71. Chiang, Alpha, Fundamental Methods of Mathematical Economics (New York: McGraw Hill, 1984). Courant, Paul N., ‘How Would You Know a Good Economic Development Policy If You Tripped Over One? Hint: Don’t Just Count Jobs’, National Tax Journal, 47 (1994) 863–81. Dabney, Dan, ‘Do Enterprise Zone Incentives Affect Business Location Decisions?’, Economic Development Quarterly, 5 (4) (1991) 325–34. Dowall, David, ‘An Evaluation of California’s Enterprise Zone Programs’, Economic Development Quarterly, 10 (4) (1996) 352–68. Dreze, Jean and Amartya Sen (eds), Indian Development: Selected Regional Perspectives (Helsinki, Finland: The United Nations University, 1996). Ehrenberg, Ronald G. and Robert S. Smith, Modern Labor Economics (Glenview, IL: Scott, Foresman and Co., 1994). Eisinger, Peter, The Rise of the Entrepreneurial State (Madison, WI: The University of Wisconsin Press, 1988). Eisinger, Peter, ‘State Economic Development in the 1990s: Politics and Policy Learning’, Economic Development Quarterly, 9 (2) (1995) 146–58. Elling, Richard and Ann Sheldon, ‘Determinants of Enterprise Zone Success: A Four State Perspective’, in Roy E. Green (ed.), Enterprise Zones: New Directions in Economic Development (Newbury Park, CA: Sage Publications, 1991). Erickson, Rodney A. and Susan Friedman, Enterprise Zones: An Evaluation of State Government Policies, final report (University College, PA: Prepared for the US Department of Commerce, Economic Development Administration, 1989). Fan, Shenggen, Peter Hazell and Sukhadeo Thorat, ‘Government Spending, Growth and Poverty in Rural India’, American Journal of Agricultural Economics, 82 (4) (2000) 1038–51. Feldstein, Martin, ‘The Private and Social Costs of Unemployment’, American Economic Review, 68 (2) (1978) 155–8. Fisher, Peter S. ‘What’s Really New About the Economic Role of the American States in the 1980s?’, presented at the ACSP–AESOP Joint International Congress, Oxford, UK (1991). Fisher, Peter S. ‘State Economic Development Incentives, Unemployment, and the Redistribution of Jobs’, presented at the 34th Annual ACSP Conference, Columbus, Ohio (1992). Fisher, Peter S. and Alan H. Peters, Industrial Incentives: Competition among American States and Cities (Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, 1998). Fisher, Peter S. and Alan H. Peters, State Enterprise Zone Programs: Have They Worked? (Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, 2002). Florida, The Rise of the Creative Class and How It’s Transforming Work, Leisure, Community and Everyday Life (New York: Basic Books, 2002). Ge, Wei, ‘The Urban Enterprise Zone’, Journal of Regional Science, 35 (2) (1995) 217–31. Grasso, Patrick and Scott Crosse, ‘Enterprise Zones: Maryland Case Study’, in Roy E. Green (ed.), Enterprise Zones: New Directions in Economic Development (Newbury Park, CA: Sage Publications, 1991). Griliches, Zvi and V. Ringstad, Economies of Scale and the Form of the Production Function: An Econometric Study of Norwegian Manufacturing Establishments Data (Amsterdam: North-Holland, 1971).

206 References Ham, J.C. and R.J. LaLonde, ‘The Effect of Sample Selection and Initial Conditions in Duration Models: Evidence from Experimental Data on Training’, Econometrica, 64 (1) (1996) 175–205. Harberger, Arnold C., ‘The Incidence of the Corporation Income Tax’, The Journal of Political Economy, 70 (3) ( June 1962) 215–40. Haurin, Donald and Kala S. Sridhar, ‘The Impact of Local Unemployment Rates on Reservation Wages and the Duration of Search for a Job’, Applied Economics, 35 (13) (2003) 1469–76. Hill, Edward, ‘Tax Abatement War Within the State’, Draft manuscript, Levin College of Urban Affairs, Cleveland State University, 1994. Holzer, Harry J., ‘Reservation Wages and their Labor Market Effects for Black and White Male Youth’, Journal of Human Resources, 21 (1986) 157–77. Holzer, Harry J., ‘Job Search by Employed and Unemployed Youth’, Industrial and Labor Relations Review, 40 (4) ( July 1987) 601–11. Honigsberg, Peter Jan, The Unemployment Benefits Handbook (Reading, MA: AddisonWesley Publishing Company, 1981). Illinois DCCA, Enterprise Zone Annual Report (Springfield, IL: Illinois DCCA, 1990). Illinois DCEO, Enterprise Zone Descriptions and Accomplishments (Springfield, IL: Illinois DCEO, 2004). Illinois Department of Revenue, Illinois Property Tax Statistics 1990 (Chicago, IL: Illinois Department of Revenue, 1992). Immergluck, Daniel, ‘Barriers between Jobs and Residents in Low-Income Neighborhoods: Lessons for Empowerment Zone and Place-Based Employment Strategies’, paper presented at the 1997 Fall APPAM Research Conference, Washington, DC (November). Irani, A., Sujata Gothoskar and J.C. Sharma, ‘Potential and Prevalence of Teleworking in Mumbai’, Economic and Political Weekly, 24 June (2000) 2269–76. Jones, Stephen R.G., ‘Reservation Wages and the Cost of Unemployment’, Economica, 56 (1989) 225–46. Keefe, Suzanne O., ‘Job Creation in California’s EZs: A Comparison Using a Propensity Score Matching Model’, Journal of Urban Economics, 55 (2004) 131–50. Kenyon, Daphne A. and John Kincaid (eds), Competition among States and Local Governments: Efficiency and Equity in American Federalism (Washington, DC: The Urban Institute Press, 1991). Kiefer, Nicholas M. and George R. Neumann, ‘An Empirical Job-Search Model, with a Test of the Constant Reservation-Wage Hypothesis’, Journal of Political Economy, 87 (1979) 89–107. Kimani, S.M. and D.R.F. Taylor, Growth Centres and Rural Development in Kenya (Ottawa, Canada: International Development Research Centre, 1973). Kochhar, Kalpana, ‘India: Macroeconomic Implications of the Fiscal Imbalances’, Presentation at the IMF-NIPFP Conference on ‘Fiscal Policy in India’, 16–17 January 2004, New Delhi, India. Korenman, S. and D. Neumark, ‘Does Marriage Really Make Men More Productive?’, Journal of Human Resources, 26 (2) (1991) 282–307. Krauss, Melvyn B. and Harry Johnson, General Equilibrium Analysis: A Micro-Economic Text (London: Allen and Unwin, 1974). Lambert, Thomas and Paul Coomes, ‘An Evaluation of the Effectiveness of Louisville’s Enterprise Zone’, Economic Development Quarterly, 15 (2) (May 2001) 168–80. Lancaster, Tony, and Andrew D. Chesher, ‘Simultaneous Equations with Endogenous Hazards’, in G.R. Neumann and N.C. Westergaard-Nelson (eds), Studies in Labor Market Dynamics (Heidelberg: Springer-Verlag, 1984).

References 207 Landers, James, ‘A Hedonic Study of the Incentive Effects of Enterprise Zones in Ohio’, unpublished PhD, dissertation (Columbus, OH: The Ohio State University, 1996). Ledebur, Larry and Douglas Woodward, ‘Adding a Stick to the Carrot: Location Incentives with Clawbacks, Recisions and Recalibrations’, Economic Development Quarterly, 4 (3) (August 1990) 221–37. Levitan, S. and E. Miller, ‘Enterprise Zones Are No Solution for Our Blighted Areas’, Challenge, 35 (3) (1992) 4–8. Logan, William Jr. and A. Barron, ‘Florida’s Enterprise Zone Program: The Program After Sunset’, in Roy E. Green (ed.), Enterprise Zones: New Directions in Economic Development (Newbury Park, CA: Sage Publications, 1991). McDermott, Tim, Legislative Guide to EZs (Des Moines, IA: Legislative Service Bureau, 2002). McDonald, John, Comment on Kala Seetharam Sridhar’s (1996) ‘Tax Costs and Employment Benefits of Enterprise Zones’, Economic Development Quarterly, 11 (3) (1997) 222–4. McWilliams, Abigail and Donald Siegel, ‘Corporate Social Responsibility: A Theory of the Firm Perspective’, Academy of Management Review, 26 (1) (2001) 117–27. Mitter, S., ‘Teleworking and Teletrade in India: Combining Diverse Perspectives and Visions’, Economic and Political Weekly, 24 June (2000) 2241–52. Mohan, Rakesh, Report of the Expert Committee on Commercialization of Infrastructure (New Delhi: Government of India, 1996). Moroney, John R., ‘Cobb–Douglas Production Functions and Returns to Scale in U.S. Manufacturing’, Western Economic Journal, 6 (1) (December 1967) 39–51. Myles, Gareth, Public Economics (Cambridge, UK: Cambridge University Press, 1995). Nagaraj, Rayaprolu, Aristomene Varoudakis and Marie-Ange Veganzones, ‘Long-Run Growth Trends and Convergence across Indian States’ (1998) OECD Technical Paper No. 131. Netzer, Dick, ‘An Evaluation of Interjurisdictional Competition Through Economic Development Incentives’, in Daphne A. Kenyon and John Kincaid (eds), Competition among States and Local Governments: Efficiency and Equity in American Federalism (Washington, DC: The Urban Institute Press, 1991). Pantuosco, Lou and Darrell Parker, ‘Sources of Prevailing Differences in Unemployment Rates for Selected Regional Pairs’, Review of Regional Studies, 28 (2) (1998) 35–46. Papke, Leslie, ‘Subnational Taxation and Capital Mobility: Estimates of Tax–Price Elasticities’, National Tax Journal, 40 (2) (1987) 191–203. Papke, Leslie, ‘The Responsiveness of Industrial Activity to Interstate Tax Differentials: A Comparison of Elasticities’, in Henry W. Herzog, Jr. and Alan M. Schlottmann (eds), Industry Location and Public Policy (Knoxville: University of Tennessee Press, 1991). Papke, Leslie, ‘Tax Policy and Urban Development: Evidence from the Indiana Enterprise Zone Program’, Journal of Public Economics, 54 (1994) 37–49. Perrucci, Robert, Japanese Auto Transplants in the Heartland: Corporatism and Community (New York: Aldine de Gruyter, 1994). Peters, Alan H. ‘Clawbacks and the Administration of Economic Development Policy in the Midwest’, Economic Development Quarterly, 7 (4) (November 1993) 328–40. Pindyck, Robert S., and Daniel L. Rubinfeld, Econometric Models and Economic Forecasts (New York: McGraw-Hill Inc., 1991). Quandt, Richard and Harvey S. Rosen, The Conflict between Equilibrium and Disequilibrium Theories: The Case of the US Labor Market (Kalamazoo, MI: Upjohn Institute for Employment Research, 1988). Rajaraman, Indira, H. Mukhopadhyay and Namita Bhatia, Fiscal Industrial Incentives of the Government of Madhya Pradesh: Costs and Benefits, mimeo (New Delhi: NIPFP, 1999).

208 References Rao, M. Govinda, Ric Shand and K.P. Kalirajan, ‘Convergence of Income across Indian States: A Divergent View’, Economic and Political Weekly, 27 March–2 April (1999) 767–778. Rao, Govinda and Kunal Sen, ‘Internal Migration, Centre-State Grants, and Economic Growth in the States of India: A Comment on Cashin and Ratna Sahay’, International Monetary Fund Staff Papers, 44 (2) (1997) 283–8. Redfield, Kent and John McDonald, Enterprise Zones in Illinois (Springfield: Illinois Tax Foundation, 1991). Reese, Laura, ‘Municipal Fiscal Health and Tax Abatement Policy’, Economic Development Quarterly, 5 (1) (1991) 23–32. Rubin, Barry and Margaret Wilder, ‘Urban Enterprise Zones: Employment Impacts and Fiscal Incentives’, APA Journal (Autumn 1989) 418–31. Rubin, Barry and Kurt Zorn, ‘Sensible State and Local Economic Development’, Public Administration Review (1985) 333–9. Rubin, Marilyn, ‘Urban Enterprise Zones in New Jersey: Have They Made a Difference?’, in Roy E. Green (ed.), Enterprise Zones: New Directions in Economic Development (Newbury Park, CA: Sage Publications, 1991). Rubin, Marilyn and Regina Armstrong, The New Jersey Urban Enterprise Zone Program: An Evaluation (Wayne, NJ: Urbanomics, 1989). Saxenian, Anna Lee, Regional Advantage: Culture and Competition in Silicon Valley and Route 128 (Cambridge, MA: Harvard University Press, 1994). Seyfried, William, ‘The Impact of Enterprise Zones on Local Economies’, unpublished PhD dissertation (West Lafayette, IN: Purdue University, 1990). Shah, Anwar (ed.), Fiscal Incentives for Investment and Innovation (Oxford: Oxford University Press, 1995). Shettar, Shivangappa F. Kore, Disparities in Economic Development Learning from the ‘Growth Centre’ Experience of India’s Five Year Plan, unpublished Master’s thesis (Ottawa, Canada: Carleton University, 1988). Singh, Kulwant and Behnam Ta’i, ‘Introduction’, in Kulwant Singh and Behnam Ta’i (eds), Financing and Pricing of Urban Infrastructure (New Delhi: New Age International (P) Limited, Publishers, 2000). Singh, Nirvikar and T.N. Srinivasan, ‘Indian Federalism, Economic Reform and Globalization’, Stanford University Center for Research on Economic Development and Policy Reform (2002) Working Paper No. 150. Sridhar, Kala S., ‘Tax Costs and Employment Benefits of Enterprise Zones’, Economic Development Quarterly, 10 (1) (February 1996) 69–90. Sridhar, Kala S., ‘Urban Economic Development in America: Evidence from Enterprise Zones’, unpublished PhD dissertation (Columbus, OH: The Ohio State University, 1998). Sridhar, Kala S., ‘Tax Incentive Programs and Unemployment Rate’, Review of Regional Studies, 30 (3) (Winter 2000) 275–98. Sridhar, Kala S., ‘Benefits and Costs of Regional Development: Evidence from Ohio’s Enterprise Zone Program’, Journal of Regional Analysis and Policy, 31 (2) (2001) 1–32. Sridhar, Kala S., ‘Firm Location Decisions and Impact on Local Economies’, Economic and Political Weekly, 38 (39) (27 September 2003) 4121–30. Sridhar, Kala S., ‘Generating Employment: How will Schemes Be Financed?’, Economic and Political Weekly 39 (34) 2004a: 3781–2. Sridhar, Kala S., ‘Cities with Suburbs: Evidence from India’, National Institute of Public Finance and Policy Working Paper No. 23 (New Delhi, India: NIPFP, 2004b).

References 209 Sridhar, Kala S. and V. Sridhar, ‘Telecommunications Infrastructure and Economic Growth: Evidence from Developing Countries’, National Institute of Public Finance and Policy Working Paper No. 14 (New Delhi, India: NIPFP, 2004). Srivastava, D.K., ‘Intergovernmental Fiscal Transfers for Equitable in-Country Growth: Country Study: India’, paper presented at ADB–NIPFP Workshop on Intergovernmental Fiscal Transfers for Equitable in-Country Growth, 5–6 September 2001. Steinnes, Donald N., ‘Business Climate, Tax Incentives, and Regional Economic Development’, Growth and Change, 15 (2) (1984) 38–47. Tannenwald, Robert, ‘State Business Tax Climate: How Should it be Measured and How Important is it?’, New England Economic Review, January and February (1996) 23–38. Tannenwald, Robert and Christine Kendrick, ‘Taxes and Capital Spending: Some New Evidence’, in Frederic Stocker (ed.), 1994 Proceedings of the 87th Annual Conference on Taxation (Columbus, OH: National Tax Association, 1995). Theodossiou, Ioannis, Wage Inflation and the Two-Tier Labor Market (Aldershot: Ashgate Publishing Limited, 1992). Tobin, James, ‘Inflation and Unemployment’, American Economic Review, 62 (1972) 1–18. Tresch, Richard, Public Finance: A Normative Theory (Plano, TX: Business Publications, 1981). Tulasidhar, V.B. and M. Govinda Rao, ‘Cost and Efficacy of Fiscal Incentives – The Case of Sales Tax Subsidy’, Economic and Political Weekly, 21 (41) (1986) 1799–1806. US Bureau of the Census, Current Population Survey (Washington, DC: Government Printing Office, 1990). US Bureau of the Census, Census of Population and Housing (Washington, DC: Government Printing Office, 1990). US Bureau of the Census, Census of Population and Housing: Summary Tape File 3A (Washington, DC: GPO, 1991). US Bureau of Labor Statistics, Employment and Earnings (Washington, DC: Government Printing Office, 1990). US Department of HUD, State Designated Enterprise Zones: Ten Case Studies (Washington, DC: Government Printing Office, 1986). US Department of HUD, State Enterprise Zone Update 1992 (Washington, DC: Government Printing Office, 1992). US General Accounting Office (US GAO), Enterprise Zones: Lessons from the Maryland Experience (Washington, DC: Government Accounting Office, 1988). Wilder, Margaret, ‘Rhetoric Versus Reality’, Journal of the American Planning Association, 26 (4) (Autumn 1996) 473–491. Zuckerman, Dror, ‘On Preserving the Reservation Wage Property in a Continuous Job Search Model’, Journal of Economic Theory, 34 (1984) 175–9.

Index Addison, J., 26, 204, 206 AFDC, xv, 54, 186 agglomeration economies, 121, 157 Ahluwalia, M., 117, 204 Alleman, J., 5, 204 alternative location, 52, 53, 56, 57 amenities, 4, 172, 203 analytical framework, 10, 11, 13, 19, 20, 28, 88, 149, 153, 169 ANOVA, 43 Appalachia, 8 Armstrong, R., 5, 19, 88, 208 Ashenfelter, O., 189, 204 automobile manufacturing, 6, 123 Bagchi, A., xiv, 7, 185, 204 Bailey, M., 22, 204 Bakkers, S., 204 bargaining, 9, 65, 172, 173 Barron, A., 42, 67, 204, 207 Bartik, T., 3, 11, 12, 24, 40, 41, 48, 49, 53, 62, 82, 85, 88, 100, 107, 112, 118, 120, 122, 123, 153, 171, 180, 187, 188, 195, 204 Basu, S., 22, 204 Bhatia, N., 10, 122, 208 Blau, D., 186, 204 blight, 24, 66 Boarnet, M., 19, 42, 64, 88, 204 Boekema, F., 4, 185, 204 Bogart, W., 19, 42, 64, 88, 204 Boisier, S., 121, 204 branch plant, 3, 41 Brunori, D., 10, 204 building permit, 48, 91 business climate, 63, 136, 171, 172, 181 business incubators, 3 business location decisions, 48 Butler, S., 7, 22, 204 Byrnes, P., 9, 57, 65, 164, 204

capital mobility, xii, 11, 13, 36 capital intensity, 33, 100, 159, 202 Cashin, P., 117, 205, 208 causality, 71, 81 cell phones, 174, 203 CETA, xv, 113 Chesher, A., 187, 206 Chiang, A., 21, 205 City of Chicago, 4, 187 clawback, 5, 57 college education, 53, 54, 55 communal tensions, 183 commuting, 96, 97 comparative statics, 34 constant returns to scale, 21, 22 consumers’ surplus, 11 control groups, 86 convergence, 8, 117, 120, 138, 185 Cook County, 47, 187 Coomes, P., 42, 206 corporate planning, 6 corporate sector, ix, 162, 170, 173, 183 corporate social responsibility, 161 corporation income tax, 28 cost advantage, 21 cost-effectiveness, 60, 62, 63, 200 cost minimization, 20 counterfactual, 47, 48, 49, 86, 145, 151, 201 county of residence, 53, 55 Courant, P., 51, 205 CRA programme, 69, 76, 77, 82 Cramer’s rule, 37 Crosse, S., 42, 205 customer base, 150, 162, 181 Dabney, D., 43, 88, 205 DCCA, xv, 43, 187, 206 DCEO, xv, 44, 45, 47, 48, 60, 187, 206 decentralization, 170, 181, 185 decertification, 113, 114 decreasing returns, 22 demand elasticity, 29, 30

210

Index 211 demand for workers, 12 demand side, 3 demographic characteristics, 70, 74, 91, 93, 125, 126 developed plots, 134, 136, 202 disequilibrium, 20, 22, 37, 186, 189, 197 distress characteristics, 20 distributional, 51, 90 Dowall, D., 19, 42, 88, 205 Dreze, J., 8, 205 e-commerce, 10 econometric evidence, 171 econometric model, 70, 89 economic base, 6, 72, 79, 120, 126, 138, 178, 179 efficiency loss, 111 Ehrenberg, R., 24, 26, 67, 125, 189, 205 Eisinger, P., 3, 4, 57, 205 elasticity of substitution, 13, 31, 32, 35, 68 e-learning, 3 Elling, R., 19, 42, 49, 205 employment distribution, 107 employment elasticity, 14, 117 employment multiplier, 6 empowerment zones, 42 endogeneity, 71, 75, 81, 131 enterprise communities, xv, 42 enterprise zone agreement, 91 entrepreneurship, 4, 63, 184 Erickson, R., 19, 33, 66, 88, 97, 205 excess capacity, 34, 68 exogeneity, 36, 53, 70, 71, 80, 187, 197, 199, 203 export orientation, ix, 140, 141, 153, 159, 203 factor intensities, 13, 31, 35 factors of production, 21 Fan, S., 174, 205 federal structure, 6, 180, 182 federalism, 7, 204, 206, 207, 208 Feldstein, M., 23, 66, 186, 205 Fernald, J., 22, 204 field visits, 15, 132, 139, 140, 160, 162, 164, 165, 182, 199, 200 financial autonomy, 97 first-mover, 5 first-order effects, 33

fiscal deficits, 8, 173 fiscal effects, 49 Fisher, P., xiii, 9, 10, 14, 42, 118, 187, 205 Florida, xv, 4, 42, 205, 207 footloose, 5, 163 Friedman, S., 19, 33, 66, 88, 97, 205 full authority zones, 69, 90 Ge, W., 19, 66, 126, 205 gender, 54, 70, 78, 126 general equilibrium, 11, 13, 26, 28, 37, 88, 169 globalization, 10 golden quadrilateral, 176 goods market, 11, 29, 30, 34, 37, 91 Gothoskar, S., 206 Grasso, P., 42, 205 Great Britain, 7 Green, R., 205, 207, 208 Griliches, Z., 22, 205 gross earnings, 53, 58, 59, 61 growth poles, 121 Ham, J., 190, 206 Harberger, A., 11, 13, 28, 29, 31, 34, 35, 206 Haurin, D., xiii, 12, 27, 55, 193, 206 Hazell, P., 174, 205 Herzog, H., 207 heteroscedasticity, 133, 199 high-school degree, 54 high technology, 3 Hill, E., 65, 205, 206, 207 Holzer, H., 53, 67, 206 Honigsberg, P., 193, 206 house prices, 57 Howe, G., 7 Hunt, C., 204 hysteresis, 61, 85, 133, 159, 162, 163, 178, 180, 187 Illinois Department of Revenue, 187, 206 Immergluck, D., 33, 97, 206 impact fees, 176 imputation, 13, 56, 93, 94, 95 incidence, 13, 28, 35, 37, 169, 174 income effect, 30 Indian Institute of Management, 200 Indian Institutes of Technology, 178

212 Index industrial backwardness, 118, 127, 165, 202 industrial dispute, 73 industrial districts, 3 industrial location theory, 162 industrial recruitment, 4 industrial revenue bonds, 41, 64 informal sector, 165, 180 information technology, 5, 21 institutional structure, 127 instrumental variable, 75 internet, 145 investment incentives, 10 investment tax credit, 47 Irani, A., 197, 206 job interviews, 12 job search, 23, 26, 37, 91, 186 Johnson, H., 21, 206 Jones, S., 11, 13, 26, 27, 206 Kalirajan, K., 117, 208 Keefe, S., 6, 42, 206 Kendrick, C., 10, 118, 209 Kenyon, D., 10, 206, 207 Kiefer, N., 187, 206 Kimani, S., 122, 206 Kincaid, J., 10, 206, 207 knowledge, 4 Kochhar, K., 173, 206 Korenman, S., 190, 206 Krauss, M., 21, 206 labour intensity, 114, 135, 137, 153, 169, 202 labour productivity, 33, 36, 65 LaLonde, R., 190, 206 Lambert, T., 42, 206 Lancaster, T., 187, 206 land compensation, 141, 152 Landers, J., 19, 207 Latino, 33, 97 Ledebur, L., 57, 207 Levitan, S., 66, 207 local government investment, 97, 102, 110, 112 local labour force, 153, 158, 159, 161 Logan, W., 42, 207 managerialism, 4, 33, 77, 185, 192

manufacturing base, 4, 72, 77 manufacturing firm location, 6 marginal product of labour, 21, 66, 170 marital status, 53, 55, 93, 190 Marvel, M., xiii, 9, 57, 65, 164, 194, 204 McDermott, T., 4, 207 McDonald, J., 19, 48, 49, 91, 207, 208 McWilliams, A., 162, 207 Mellow, W., 67, 204 Michaels, D., 204 milieu of innovation, 3 military bases, 45, 46 Miller, E., 66, 207 minimum wage, 22, 158, 159, 163, 181, 194 Mitter, S., 197, 207 model of unemployment, viii, 23, 66, 70 Mohan, R., 176, 207 Morgan, K., 204 Moroney, J., 21, 207 Mueller, M., 204 Mukhopadhyay, H., 10, 122, 208 multiplier, 6, 97 Myles, G., 30, 34, 207 Nagaraj, R., 117, 207 National Longitudinal Survey, xvi, 67, 193 natural rate of unemployment, 34, 68, 189 neoclassical economics, 23, 66, 186, 197 Netzer, D., 119, 207 Neumann, G., 187, 206 Neumark, D., 190, 206 New Jersey, 5, 19, 42, 64, 65, 204, 208 new wave, 3 No Industry District, 152 non-financial, 3, 41, 145 non-tax incentives, 10, 118 occupational composition, 77 Octroi, 175 on-the-job training, 9, 33, 114 organizational skills, 21 organized sector, 12, 165, 180, 181 outcome variable, 49, 70, 125, 189 own-source revenue, 7 Panel Study of Income Dynamics, xvi, 43, 89, 187

Index 213 Pantuosco, L., 70, 125, 197, 207 Papke, L., 10, 19, 42, 70, 74, 82, 88, 118, 125, 126, 207 Parker, D., 70, 125, 197, 207 partial equilibrium, 10, 122 people orientation, 9 performance appraisal, 140, 151, 164 Perrucci, R., 57, 207 Peters, A., xiii, 9, 10, 14, 42, 57, 118, 205, 207 physical characteristics, 20 Pindyck, R., 83, 207 pirating, 14, 65, 87 plant-closing legislation, 57 political science, 10 positive sum, 63 Prantilla, B., 204 present discounted value, 26 previous wages, 53 price elasticities, 13, 35 price of capital, 22, 29, 34 private investment, 3, 8, 9, 12 probability density function, 23 production function, 30 production-possibility frontier, 22 programme design, 110 pro-growth coalitions, 71 propensity score, 42 Public Distribution System, xvi, 186 public sector jobs, 33, 97 public–private partnerships, 176 Quandt, R., 189, 197, 207 race to the bottom, 4, 9, 171 Rajaraman, I., 10, 122, 189, 197, 208 Rao, G., 10, 117, 122, 198, 208, 209 Rappaport, P., 204 Reagan, R., 7 Reaganomics, 171 Redfield, K., 19, 49, 91, 208 redistribution, 39 reduced form model, 68 Reese, L., 41, 208 reference week, 73, 74 relative prices, 30 renewal communities, xvi, 42 residential location, 57 revenue expenditure, 149 Ringstad, V., 22, 205 Rosen, H., 189, 197, 207

Route 128, 171, 208 Rubin, B., 5, 19, 21, 42, 43, 88, 208 Rubinfeld, D., 83, 207 rural poverty, 174 Rutten, R., 204 Sahay, R., 117, 205, 208 sales tax exemption, 47, 49 sample selection, x, 55, 89, 92 Saxenian, A., 171, 208 Schlottmann, A., 207 self-reports, 73 Sen, K., 8, 117, 205, 208 service employment, 72, 79, 80, 84, 126, 130 Seyfried, W., 19, 88, 126, 208 Shah, A., 10, 208 Shand, R., 117, 208 Sharma, K., xiii, 206 Sheldon, A., 19, 42, 49, 205 shift share, 42, 43 Siebert, W., 26, 204 Siegel, D., 162, 207 Silicon Valley, 171, 208 Singh, N., xiii, 117, 174, 185, 208 Smith, R., 24, 26, 67, 125, 189, 205 social security, 54, 66, 180 spatial mapping, 165 special economic zones, 7, 184 spin-offs, 5 Sridhar, K.S., iii, iv, 9, 12, 19, 27, 41, 55, 57, 65, 91, 92, 120, 164, 174, 177, 179, 188, 193, 203, 204, 206, 207, 208, 209 Srinivasan, T.N., 117, 185, 208 Srivastava, D.K., 117, 209 Steinnes, D., 10, 118, 209 substitution effect, 30 supply-side economics, 171 sustainable, 5, 9, 12, 172, 178 Ta’i, B., 174, 208 Tannenwald, R., 10, 118, 209 tax-incentive agreement, 69 taxable capacity, 76 tax-sharing arrangements, 57 Taylor, D., 122, 204, 206 telecommunications, 47, 124, 162, 174, 175, 178, 181 terrorism, 170, 172 testable hypothesis, 27, 40

214 Index Theodossiou, I., 24, 27, 209 theoretical model, 13, 37, 59, 65, 90, 94, 125, 177, 192 Thorat, S., 174, 205 TIRC, xvi, 113 Tobin, J., 34, 68, 209 topography, 171, 172 trade fairs, 3 traffic jams, 174 transport costs, 6, 57, 132 treatment effects, 14, 66, 71, 75, 80, 189 Tresch, R., 31, 209 tsunami, 170, 172 Tulasidhar, V., 10, 122, 209 2SLS, xi, xv, 75, 76, 80, 81, 83, 84, 127, 130, 131, 197 typology of cities, 21 unemployment benefits, 23, 44, 92 unemployment claims, 70, 74, 82 unemployment compensation, 23, 24, 53, 54, 66, 74, 186 unorganized sector, 170, 179, 180, 181

US Bureau of Census, 51 US Bureau of Labor Statistics, 51, 188 US Department of HUD, xvi, 33, 42, 50, 88, 97, 209 US GAO, 43, 209 utility-maximizing, 26 vacant housing, 75, 78, 79, 80, 191 Varoudakis, A., 117, 207 Veganzones, M., 117, 207 welfare payments, 23, 24, 54, 186 Wilder, M., 43, 66, 208, 209 Woodward, D., 57, 207 work experience, 73, 83, 95 workforce participation, 129, 177 zero-sum, 39, 40, 61, 65, 85, 87, 169 zone characteristics, 94, 95 zoning, 175, 176 Zorn, K., 21, 208 Zuckerman, D., 26, 209