Gaming in the New Market Environment

Gaming in the New Market Environment Edited by Matti Viren Gaming in the New Market Environment This page intention...

Author: Matti Viren

17 downloads 967 Views 926KB 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

Gaming in the New Market Environment Edited by

Matti Viren

Gaming in the New Market Environment

This page intentionally left blank

Gaming in the New Market Environment Edited by Matti Viren

Selection and editorial matter © Matti Viren 2008 Chapters © their authors 2008 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 authors have asserted their rights to be identiﬁed as the authors of this work in accordance with the Copyright, Designs and Patents Act 1988. First published 2008 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: 978–0–230–50050–1 hardback ISBN 10: 0–230–50050–1 hardback This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. Logging, pulping and manufacturing processes are expected to conform to the environmental regulations of the country of origin. A catalogue record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Gaming in the new market environment / edited by Matti Viren. p. cm. Includes bibliographical references and index. ISBN 0–230–50050–1 (alk. paper) 1. Gambling industry. 2. Gambling. 3. Lotteries. I. Viren, Matti. HV6710.G38 2008 338.4′7795–dc22 10 17

9 16

8 15

7 14

6 13

2008014546 5 12

4 11

3 10

2 09

1 08

Printed and bound in Great Britain by CPI Antony Rowe, Chippenham and Eastbourne

Contents List of Tables

vi

List of Figures

vii

Acknowledgements

viii

List of Contributors

ix

Chapter 1

Introduction

1

Chapter 2

When Welfare Economics and Gambling Studies Collide

23

Chapter 3

Demand Issues in the Market for Lotto and Similar Games

54

Chapter 4

Lottery Design Lessons from the Dismal Science

75

Chapter 5

Lotteries as a Source of Revenue

99

Chapter 6

Problem Gambling and European Lotteries

126

Chapter 7

The Economics of Scale and Scope in the Lottery Industry

160

Chapter 8

The Political Economy of Regulating Gambling

184

References

209

Index

221

v

List of Tables 3.1 3.2 3.3 3.4 4.1 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 6.1 7.1 7.2 7.3 7.4 7.5 7.6 7.7 8.1 8.2 8.3 8.4

Summary of demand elasticities Regression results, UK Saturday Lotto sales (millions) UK Saturday Lotto sales regression Spending on Lotto and other models of gambling Estimated parameters of log sales equations Gross tax rates, EU lotteries Gambling revenues (£ million) and gross tax rates, UK Excise rates – UK National Lottery, alcohol, tobacco and petrol products Tax revenues, with and without the lottery Share of total income spent on ‘vices’, by income decile Share of spending on ‘vices’, by income decile Correlation coefﬁcients of expenditures Why do people play the National Lottery? What is ‘an excellent way to spend’ lottery money? A summary of the most common forms of ofﬂine commercial gambling Panel data estimates of the cost function for European lotteries Cross-section estimates from the European data Cost function estimates with the US data Change in cost shares over time in the US Marginal effect of sales on costs Further estimates of scale economies Some demand function estimates with the US data The variables and the main bookkeeping relations Names in Danish and our translation Categories and main results of Danish gambling dependency studies Gambling prices for various products in Denmark

vi

60 65 67 72 89 100 102 103 104 110 112 112 121 122 128 169 170 171 172 172 172 177 192 197 198 199

List of Figures 2.1 2.2 2.3 2.4 3.1 3.2 4.1 4.2 4.3 4.4 4.5 4.6 5.1 5.2 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10 8.1 8.2 8.3 8.4 8.5 8.6

Consumer surplus Consumer surplus for two types of gamblers – recreational and pathological Efﬁciency of production and consumption in a world with legalised gambling Shifts in the production possibility curve Sales of Saturday Lotto tickets in the United Kingdom Probability distribution of winnings in a typical UK Lotto draw Lotto’s peculiar economies of scale Expected value in rollover and regular draws Two-game lottery market UK sales (£m per week) Effects of varying n on sales and revenue Effects of varying τ on sales and revenue Tax Lorenz curves – lottery, gambling, tobacco and alcohol Good causes and lottery spending Optimal payout ratio with different demand elasticities and marginal costs Operating costs in Europe 1995–2005 Personnel costs in Europe 1995–2005 Operating costs in the US in 2005 Factors affecting operating costs in the US Two alternative marginal costs curves Marginal costs for US companies 1992–2005 Effect of prizes on sales in the US Operating costs/sales from European and US lotteries Operating costs for certain private gaming companies, 2005 Demand for gambling by a normal and an addicted gambler Demand for gambling by an addicted gambler The effect of a tax increase on normal gamblers A stylised theory of the gambling problem Adding the possibility of excess production The distributions of unit costs for the typical market ﬁrm and SOE vii

26 30 41 43 57 62 82 82 85 88 91 92 111 119 162 165 166 167 168 174 175 178 179 180 187 188 189 194 196 202

Acknowledgements The preparation of this book would not have been possible without the support of the Finnish Veikkaus. The CEO, Risto Nieminen, and director, Jari Vähänen, deserve special thanks for supporting the original initiative and encouraging work at latter stages. Ms Paivi Ahtola has also helped a lot in all practical matters, especially in organising the authors’ workshop along with the EL congress in Budapest in May 2007. Marja Heikkinen has kindly checked the English of several chapters and Päivi Nietosvaara has edited the ﬁnal version of the manuscript. Several persons have provided useful comments and other help during the process of writing different chapters of the book. Some of these people are mentioned in the context of individual chapters but the authors would like to express their gratitude to all of them. Also all the European lotteries that responded to a survey on the consequences of the changing market environment deserve warm thanks. Finally, the assistance of Alec Dubber of Palgrave and Shirley Tan of EXPO Holdings Sdn Bhd was paramount for the rapid publication of the book. Helsinki, September 2007 Matti Viren

viii

List of Contributors Lisa Farrel, School of Economics and Geary Institute, University College Dublin, Ireland. David Forrest, Centre for the Study of Gambling, University of Salford, UK. Mark Grifﬁths, International Gaming Research Unit, Psychology, Nottingham Trent University, Nottingham, UK. Martin Paldam, Department of Economics, University of Aarhus, Denmark. Sarah Smith, Centre for Market and Public Organisation, University of Bristol, UK. Matti Viren, Department of Economics, University of Turku, Finland. Ian Walker, Department of Economics, University of Warwick, Coventry, UK.

ix

This page intentionally left blank

1 Introduction Matti Viren

1.1

Background

The idea behind this book is the observation that gaming markets are changing rapidly. This change manifests itself in institutions, the nature of games and in the volume and pricing of gaming.1 Formally, the institutional set-up has not changed very much during the past ten years. However, under the surface a great deal has happened. The most important engine of change, of course, has been the internet which has enabled cross-border gambling in an unprecedented scale. Although we do not exactly know the volume of cross-border sales, we can use the estimates of a recent survey conducted across European lotteries.2 According to this survey, cross-border sales constitute an average of 7.4 per cent of the gaming companies’ total sales. It is not a strikingly big ﬁgure but when we look at the individual games, we are faced with certain very high percentages. For example, while crossborder sales in lotto are around 1 per cent, their share in sports betting amounts to up to 39 per cent. From the point of view of existing companies and governments, the situation is alarming already because of major losses of revenue. Governments are also concerned about tax revenue: private bookmakers do not pay taxes for their cross-border sales, which in turn further deteriorates the governments’ ﬁscal position. State-owned Lotto companies have legal monopolies in almost all European countries. This applies to lotto all over Europe and, in many countries, to all gambling activities (including slot machines). Along with the state companies, there are a number of private bookmakers, which obviously would like to expand their activities to markets now governed by public monopolies. For this purpose they keep challenging the markets by selling their products via the internet and by questioning 1

2 Gaming in the New Market Environment

the justiﬁcation of exclusive rights in courts of law. In fact, dozens of court cases are currently pending. Thus far, the exclusive rights have not been found illegal. Yet, the current system has been found to be at variance with other legislation and principles of competition in a number of countries (cf., e.g., Littler and Fijnaut, 2006). More importantly, the European Commission has shown strong interest towards the opening of the gambling sector to competition. Consequently, it has made several inquiries on the functioning of the gambling markets in Europe, the most notable of which was made by the Swiss Institute of Comparative Law (the so-called Swiss Institute Report, 2006). Thus, it is quite obvious that cross-border sales will continue to grow in the future unless strict regulations, such as the restrictions imposed in the US on the transfer of money via the banking system, are enforced. Moreover, there is growing pressure on the current institutional set-up dealing with sport-betting, in particular. This does not mean that the only thing that could happen is the simple liberalisation of gambling markets. It is possible that the tax system changes so that a heavy excise tax (along with the VAT) is imposed on all types of gambling, which would effectively take out the proﬁts from gambling. At present, tax rates vary considerably from one country to another (among the EL members, taxes may equal the lotteries’ net proﬁts). This is not important in the current state monopoly system because the distinction between tax revenue and company proﬁts is somewhat artiﬁcial from the government point of view. In a more competitive system, this would no more be true and familiar problems of tax competition and tax harmonisation would become relevant. Because lottery services can be produced in basically any country, there would be large incentives for tax competition, which would limit the governments’ chances of compensating the reduced proﬁts from state lottery companies by taxes. Another complicated tax issue deals with the taxation of internet gaming which is right now a grey area in the European Union (Jeanneret-Druckman, 2004). Along with the lines of VAT taxation cross-border internet gambling ought to be taxed in the customer country, yet taxes are not collected in practice. Whatever will happen, it is clear that we are moving towards more competitive markets and prices will fall (accordingly, prizes and the payout ratios will increase).Thus, while the average payout ratio of the European companies is now at 53.3 per cent as compared to 48.5 per cent ten years ago, all companies expect it to increase substantially in the next ten years. If all restrictions were abolished, the payout ratios

Introduction 3

could reach 70 per cent according to the current European state lotteries.3 This would obviously increase the volume of sales. Using the subjective estimate of the companies as a benchmark we could say that such a change in the payout ratios would increase the overall demand by almost one-third. Quite clearly, such changes would show in all indicators: production, employment, advertising and, maybe most importantly, problem gambling. What would be the nature and magnitude of these effects is a subject to the analyses and reviews that are presented in this book. Even though we are focusing strongly on the future, this book is not a forecast. Rather than trying to say what the situation is in 2017 we wish to put forward alternatives which are available to decision-makers and consequences of these alternatives, including the alternative that no signiﬁcant changes in the current institutional system would take place. The nature of gaming will change even without institutional change. This can be clearly seen in the increase of multi-jurisdictional lotteries. The Nordic countries (Denmark, Finland, Iceland, Norway, Sweden and also Estonia) have the Viking lotto, several European lotteries (Austria, Belgium, France, Ireland, Luxembourg, Portugal, Spain, Switzerland and the UK with a total population base of 208 million people) are involved in the EuroMillions, and in the US, as many as 29 states participate in Powerball. The motivation for this trend is clear: consumers prefer very large jackpots and they are only achievable by putting together the clientele of several small lotteries. The idea that we might have a fully global lotto one day does not sound terribly unrealistic any more. In fact, some suggestions have already been put forward. Along with globalisation, the structure and nature of games keep changing. To illustrate previous developments, we could take the example of Finland. In 1940, when the Finnish lottery Veikkaus was founded, it started with just one game (Football Pools). In 2006, there were already 12 lotto, keno and sports betting games and, in addition, 12 instant games. Over the years, at least the same number of games have come and gone and many of the remaining games have changed their form – frequently into games playable over the internet (Veikkaus’ Annual Report, 2006). When considering these alternatives, the deep dividing question concerns the role of government regulations and the role of state monopolies in the production of gambling services. Thus, we could ask whether there is a case for curtailing the demand for lotto and gambling. A partly related question concerns the role of state monopolies: should we preserve the monopoly or not? In the latter case of no state

4 Gaming in the New Market Environment

monopoly, we have to consider whether we should have a licence system (one or several licences), or whether we should implement a completely unregulated system. These questions are difﬁcult because they concern the role of government and the scope to which governments can go in securing their ﬁscal needs and protecting consumers without jeopardising free entrepreneurship and free competition. To illustrate the difﬁculty of deciding what form the lottery markets should take in the future, let us look at the history of lotteries in the United States. In the United States lotteries have a long history starting before the independence. Initially, there were both public and private lotteries with different licensing systems in different states. In the early 19th century lotteries had their ‘golden age’, but since 1830 growing criticism and popular opposition led to the curtailment of the lotteries’ operations and sales. After the Civil War, the situation culminated in that the government (different states with the help of the Federal Government) abolished gambling altogether for almost a century (for details, see Clotfelter and Cook, 1991 and Ezell, 1960). Basically, the prohibition could be explained by the general view that gambling creates more harm than beneﬁt for consumers. To some extent at least, this presumption was caused by misdeeds of criminal-minded lotteries that received wide publicity. Finally, in 1964, New Hampshire legalised lottery activities again, and was gradually followed by most US States. Currently, only eight US states prohibit legalised lotteries.4 In Europe, the attitudes have not been as strongly divided, although they have varied across different countries (cf., e.g., Douglas, 1995). It is only in the case of casinos that there has been extensive resistance in almost all European countries. In the rest of the world, gambling has a somewhat shorter history and it has only recently become a major industry that has beneﬁted from relatively liberal government policies. Right now, the general attitude towards gambling is fairly positive. Relatively few consumers and organisations would be ready to prohibit all sorts of gambling. This, in turn, can be interpreted as an indication that the favourable welfare effects of gaming proceeds are generally well recognised. Even so, the measurement and role of the welfare effects are still worth thorough analysis. Through demand analysis, we can derive a large amount of useful and objective information on these effects. It is interesting to notice that the acceptance of gambling manifests itself in long cycles. This can be seen especially in the US, where the attitudes have moved from one extreme to another. It is hard to say

Introduction 5

how much of this is due to religion or, more generally, a certain way of thinking. Sauer (2001) argues that ﬁscal factors are decisive in this respect. When ﬁscal needs have been high – due to wars or otherwise increased government expenditures there has been more willingness to allow gambling to alleviate the pressure of taxation. In Sauer’s view, the recent growth of lotteries can be attributed to the growth of government expenditure worldwide (e.g., due to the construction of the welfare state). Fiscal pressures have paved a way for new sources of revenue, and gambling is an obvious possibility for new income source. It is difﬁcult to say exactly how important ﬁscal matters are but it is clear that we cannot isolate ‘attitudes’ and ‘ways of thinking’ from the development of the economy. Certainly, there exist differences between e.g., federalist and non-federalist countries because state-level authorities have had less alternatives in generating additional revenues. This may show up in attitudes towards casinos, age-limits and so on. Naturally, the demand for gambling has its own dynamics: when gambling grows extensively, the dark side becomes more prominent and opposition emerges. By contrast, in a heavily regulated system, virtually no problems seem to exist and there is no demand for an anti-gambling movement. In the 1950s and 1960s, state gaming monopolies seemed the obvious way of arranging gaming activities.5 Governments could delegate the production of services to the monopoly company and curtail the possible harmful effects by monopolistic pricing. At that time, the harmful effects probably mattered much less than the ﬁscal effects. As ﬁscal agencies, government monopolies were ideal. There was no need to design a tax system for various products. Instead, it was left to the monopoly company to ﬁnd the optimal pricing and production scheme. Revenues from gaming monopoly (monopolies) are deﬁnitely important for governments, although the ﬁgures do not seem very high. Thus, in the US, lottery revenue represented roughly 2.3 per cent of all own-source general revenue or lottery states (Hansen, 2004). According to Eurostat, taxes on gambling, lotteries and betting correspond to about 0.4 per cent of all taxes and social security contributions to general government. We have to be careful with these numbers because they are not comparable and because the quality of the European ﬁgures at least is dubious.6 Moreover, we have to keep in mind that a possible squeeze of lottery revenues would not hurt all levels of government on the way. Let us take one example. In Finland, the share of taxes on gambling, lotteries and betting out of the total taxes and social security contributions to general government was 1.3 per cent in 2006. However, if we compare these revenues to central government income tax

6 Gaming in the New Market Environment

revenues, the share is as high as 7.6 per cent. Thus, from the central government’s point of view, lottery revenues are important because if they have to be compensated with the main tax instruments (income tax or VAT), the average rates should be increased by more than one percentage point. Although all lottery revenue does not go to government (literally to the account of the treasury) the revenue is equivalent to other government revenue from the government’s point of view in most cases. Thus, if state lottery proceeds were not available, the governments should probably provide similar levels of funding to the lottery beneﬁciaries (when the government is not the beneﬁciary directly). This would require higher taxes, basically entailing a trade-off between the ‘lottery tax’ and, say, the income tax. Comparing these two alternatives is not easy because it is not completely clear to what extent ‘lottery tax’ is comparable with income tax (rather it is comparable to some commodity taxes). The only thing that is clear is that the two ‘taxes’ have quite different distributional effects: the ‘lottery tax’ is known to be highly regressive, whereas the income tax is usually progressive. Thus, moving away from the ‘lottery tax’ would be good for distributional reasons at ﬁrst sight. Yet, we have to keep in mind that the ‘lottery tax’ does not necessarily disappear if we abolished state monopolies but rather it changes its form to (private) lottery companies’ proﬁts. At any rate, this trade-off is a real problem because taxation is not only an issue of distribution. From the economics point of view, the deadweight losses are probably more important. In the case of income taxation, we have long thought to be relatively unimportant (assuming that labour supply is almost inelastic). However, the general views are quite different at present. Thus, in public ﬁnances we have experienced a ‘ﬂight from high income taxes’ to alternative sources of ﬁnance. Unfortunately, or fortunately, in the modern world, we have very few options for alternative sources of ﬁnance: tax competition sets strict limits to capital taxation and commodity taxation. On top of that, we should keep in mind that at least in income taxation, both trade unions and employer’s associations have great inﬂuence on decisionmaking. There will never be equally powerful gamblers’ associations to lobby in London or Brussels. Needless to say, the ﬁscal problem varies from country to country depending on the earmarking system and budgeting practice. In the case where the lottery money goes to sports and arts, it looks like there is no direct obligation on the government to compensate for the loss of money. We should remember, however, that sports organisations and

Introduction 7

the arts lobby have always been powerful and skilful lobbyists for government expenditures and that they would probably receive as much as now from the government if there were no lottery funds available. Although the revenues from gambling constitute an important fraction of governments’ resources there is a big policy problem with this source of funds: the importance of lottery revenue does not justify monopolies as such. The fact that a government monopoly is highly proﬁtable in a certain market does not provide the government with a justiﬁcation to set up a monopoly and prevent competition. The government must have a proper justiﬁcation for preventing market outcome. Basically, this is the position adopted by the European Court of Justice and several national courts. If the ﬁscal necessity does not constitute the case for government lottery, what could then be the proper justiﬁcation? The primary explanation appears to be the need to curtail the harmful side effects of gambling, by which we usually mean problem gambling. The idea is simple: with monopolistic pricing, the volume of sales stays relatively small and if problem gambling is related to the level of sales, problem gambling should also remain well under control. The previous argument makes sense; yet it can be challenged by at least two arguments. First, problem gambling does not appear to be really widespread and second, the reduction of demand could be achieved by means other than state monopoly, i.e., by taxes and legal restrictions. In this respect, government monopoly could only be better if it behaved in a way different from that of private operators, e.g., by showing greater self-restraint in creating new games, restricting sales to high risk-groups and using less (effective) advertisement to stimulate demand. These are issues which probably vary enormously from country to country depending on the role of ﬁscal authorities (Ministry of Finance) in the administration of the companies, the system followed in appointments (does the board consist of politicians, business people, academics, etc.), and what is the general business culture in public enterprises like. In the US, state lotteries seem to clearly resemble each other, but in Europe there is no uniform model for them. Take for instance, Finland and Italy and compare their market structure, licensing policy, and legislation! The role of problem gambling is quite delicate as it is suggested later in this book. It is generally argued that problem gambling has been well under control thus far, and that it only represents a problem for a small minority, say 1 per cent of the adult population, which appears a typical estimate in most jurisdictions. But we have to be careful in quantifying the signiﬁcance of problem gambling. Small numbers of

8 Gaming in the New Market Environment

participation may not necessarily mean small economic impact. The consequences of problem gambling can be experienced by a much larger set of people than the problem gamblers themselves, and the damage on the lives of the sufferers and their families can be much larger than our numbers would suggest. We could imagine that problem gambling could go to extreme forms, such as theft and fraud. Then, its true incidence would no more be marginal. From the point of view of the present analysis, the key issue is not really the current state of problem gambling but the connection between market change and the growth of problem gambling. The fact that problem gambling has been under control does not necessarily mean that problems will not occur shall markets change. Increase in sales and, in particular, the development of more intensive games could change the problem gambling environment entirely.7 We should also notice that even if problem gamblers might not lose money, they (and their social environment) would still suffer from problem gambling. With less binding budget constraints, they could devote their entire lives to gambling leaving aside schooling, social contacts and work, which sounds problematic even though the gambler’s utility would be maximised. Thus, even if increased competition lowered prices, cutting the existing problem gamblers’ monetary losses, assuming that their demand is relatively inelastic, problem gambling would not automatically become less problematic as a whole and may even become worse. Besides the desire to limit problem gambling, we have basically two further rationalisations by which the gambling monopoly can be justiﬁed: (1) the gambling markets may not stay competitive and (2) there are several public policy reasons for monopoly. If the gambling market does not become competitive, consumers will not beneﬁt from lower prices (increased consumer surplus). It is this fact that is often used as to motivate government intervention in public economics. If consumers were not awarded greater surplus, public monopoly would not be any worse than the market solution. Our argument stems basically from the fact that costs are predominantly ﬁxed in gambling, especially in lotteries. Thus, marginal costs are small and apparently falling.8 Such a cost structure clearly favours monopolies (creating an environment for what could be called a natural monopoly). A natural monopoly would have a permanent cost advantage in comparison with smaller ﬁrms, driving smaller companies out of the market and stopping new companies from entering the market. Thus, we would not end up with a competitive market

Introduction 9

solution in the long run, but rather with a single monopoly or monopolistic competition depending on the role of costs when sales become very large. Some caveats should, however, be considered here. First of all, we know relatively little about the cost conditions of gaming companies. Moreover, all empirical analyses make use of data for relatively small companies. In fact, in the US, the biggest company (New York State Lottery) represents just 12 per cent of the total sales in US, whereas the biggest European company (Lottomatica) represents 15 per cent of the European sales. Hence, we know little about the cost conditions of companies that would cover the total US and/or European markets. Another caveat deals with the nature of games. The cost conditions for setting up new games or selling them may be quite different in the future, which obviously affects the shapes of relevant cost curves. This is obvious if we compare, e.g., the costs incurred by internet sales with traditional retailer commissions. It might be argued that a natural monopoly would not be a problem because the government could control it by various regulations, as has been the case with various energy, water and transportation monopolies customarily (for a comprehensive literature review on the regulation of natural monopolies, see e.g., Depoorter, 2000). Controlling a natural monopoly is somewhat controversial since it is the most efﬁcient way of production. Yet, the adverse distributional effects usually make governments set upper limits to prices. Correct pricing is a very delicate matter because the government should basically give subsidies to the company to obtain the most efﬁcient production structure. In the case of full-scale market liberalisation a more serious institutional problem would, however, arise because all the companies would be inevitably global. Consequently, we would need a global organisation to control the supply side. Right now, there is no obvious organisational solution for the problem although regional institutions like the EU Commission, might try to do that.9 Turning to public policy problems, there are several issues to be considered. The ﬁrst issue concerns fraud and other misbehaviour. Such phenomena would most probably be relevant in the case of a seemingly competitive system with many small providers of gambling services. Those having difﬁculties in surviving would be tempted to use illegal, or morally dubious means of surviving in the market. They might also collaborate with the criminal-minded producers of false results.10 Small ﬁrms might be founded solely for the purpose of crime, money laundering, in particular. Money laundering is a big issue in modern

10 Gaming in the New Market Environment

societies, since it is lies behind all economic crime and misbehaviour. On the other hand, gambling is ideal for money laundering because of its big volumes and frequent transactions. Even now, payout ratios are sufﬁciently high to exceed the critical levels required to support money laundering (see, e.g., Fabre G., 2003). The problem is generally recognised and various legal and administrative steps have been done to prevent it but very few cases have, after all, been discovered and brought to justice (Skala, 2004). Finally, on the list of public policy issues, there are regulatory problems caused by advance payments, presentation of correct results etc.11 All of the items on this list suggest that, under all institutional arrangements, a completely unregulated system could not be allowed. Instead, a certain licence system should be imposed, similar to the one used in banking. In a non-monopoly system, more funding should probably be required to regulatory activities to the extent of setting up a separate agency for lottery supervision. While the above-mentioned issues may provide justiﬁcation for government gambling monopolies, there are also arguments that favour the abolishment of those monopolies. On the top of the list is probably the need to increase competition, decrease government intervention and, in addition, the desire to achieve a more efﬁcient production of lottery services. A more competitive system would, ceteris paribus, lower the ‘lottery tax’ and increase the demand for lotto. It would enhance welfare even if it increased problem gambling. If markets did not stay competitive but turned into private monopolies, we would not gain very much from liberalising the markets. Rather, we would be dealing with certain adverse distributional issues. A relevant question would then be: who will gain from the government’s monopoly lottery versus who will gain from the private lottery monopoly. Public vs. private ownership would probably – irrespective of market structure – affect the efﬁciency of production. There are two possible reasons for this. First, even if we had a monopoly for reasons other than legal considerations in private production, there would always be a contestable market option. Unlike the case of legal monopoly, a private monopolist would have to be prepared for the entry of a possible new competitor. Thus, it could not set its prices too high even if there were no apparent competition. Secondly, a private monopoly has to be responsive to its shareholders and therefore try to increase its proﬁts by, e.g., reducing costs. Basically, public monopolies have the same requirement but they have been typically less keen on reducing

Introduction 11

costs and increasing efﬁciency. What is said about reducing costs basically applies to innovations and investments, as well. If managers are not under constant pressure to increase proﬁts, less effort is devoted to improvements. Naturally, it is generally assumed that the ‘public ownership slack’ also shows in public lottery companies. Unfortunately, there is no empirical analysis which would allow us to conclude whether the presumption is right or wrong. Existing data facilities allow only very rough comparisons. Even though these comparisons are not very informative casual evidence from lottery companies suggests that the differences are not very large and at least some improvement in the efﬁciency at public lottery companies has taken place. Thus, in spite of a rapid growth of sales, employment has not increased but rather decreased. There can be various reasons for this but it seems likely obvious that increased interest in the public companies’ efﬁciency could be one reason behind the change. We also have to pay attention to the fact that state monopolies, although they have a legal monopoly, are de facto competing with private gambling companies. Thus, gambling markets in Europe are more competitive than contestable in an abstract sense. But for efﬁciency comparisons, we would deﬁnitely need empirical analysis to answer the question of whether important efﬁciency differences exist, both between state monopolies and between private and public companies. The problem is that even the efﬁciency issues have two sides: a more efﬁcient company can sell more, which, from the problem gambling point of view, could be detrimental. This reﬂects the current schizophrenic situation in the European gambling markets. Increased cross-border trade has made markets competitive and state lottery monopolies face difﬁcult choices. If they start competing (in the sense of defending their market share) by lowering prices, increasing advertising and introducing more games then private competitors and possibly even the regulators would accuse the monopolies for behaving like competitive ﬁrms and ignoring the harmful side-effects of gambling. In such a case, their existence would be hard to justify. If, by contrast, state monopolies do not react in any way to cross-border competition, they lose their market shares, and the markets will start functioning like any other market. Again it would be difﬁcult to justify the state monopoly if it has no role in market behaviour and no ﬁscal importance. How should the companies, or their regulators, behave in such an environment? Right now, nobody seems to know. Obviously, some solution

12 Gaming in the New Market Environment

has to be reached, and transparent and enforceable rules have to be imposed on the market. It is also evident that, in the future, we will have to reconsider ﬁscal issues from various viewpoints. It is probable that some tax harmonisation will take place. It would be highly timely for Europe where both the rates and the institutions vary considerably. As we have pointed out earlier, some European countries have a formal lottery tax while others rely on direct transfers of proﬁts from state lottery companies. Some (mainly Eastern European) countries also tax winnings, whereas they are tax free in most other countries. The taxation of winnings from other countries (private companies & internet) is even more of an open issue. In the future, one might expect that the European countries would move to a system of formal gambling taxation. The taxation would, of course, be realised as source taxes so that winnings would be tax free. But what type of taxes should be imposed? This issue is handled in several articles in this volume (especially in Chapter 5). The problem is that ‘optimal’ tax considerations seem to indicate that the taxes should not be excessively high. Yet, there are political economy considerations that might lead to relatively high rates. Governments would probably be eager to impose very high tax rates on all gambling, not only to reduce problem gambling but to discourage private gaming companies from entering the market. Gambling is an area where various special interest groups and lobbies have a lot of inﬂuence. We think it is unrealistic to expect that taxation would be organised on a purely theoretical basis. Yet, bearing in mind the above issues, certain basic principles of welfare economics should be acknowledged. Here we will not answer the question of how the gambling markets should be arranged, even though we provide a large amount of material which helps to take a stance on this matter. Thus, individual chapters of the book focus on the key issues of gambling and provide necessary information to answer the questions: how much will gambling (lotteries) increase if competition increases?; what is the contribution of the present games and changes in the existing games?; what are the welfare and ﬁscal effects of these changes?, what happens to the market structure if various restrictions on cross-border gaming are abolished?; how will problem gaming respond to changes in the sales volume and gaming menu?; and, what public policy concerns the change will bring? We are well aware that all relevant issues are not covered. Take, for instance, the interplay between lottery companies, governments and various beneﬁciary organisations in the area of public policy, or the

Introduction 13

problems of crime and fraud. Despite the partial coverage, we believe that the key issues are relatively well represented here and they will facilitate more intelligent discussion of the choices that could, or have to be made in the future. What follows next is a short summary of the individual chapters. At the end, some concluding remarks are presented.

1.2

Summary of different sections

Just to give an idea of what the book is about, we try to summarise here the individual chapters. Needless to say, the summaries are no substitutes to the articles. In the ﬁrst article (Chapter 2) ‘When Welfare Economics and Gambling Studies Collide’, Lisa Farrell starts with a basic issue: how does gambling affect welfare. To start, she notes that the betting and gaming sector attracts a lot of interest from outside parties. Those within the sector view their products as leisure activities competing in a tough market place, whilst groups outside of the industry question the (ethical) nature of these products and their contribution to society. There are few industries (tobacco and alcohol included) that not only have to face the daily struggle to earn a proﬁt but also have ﬁght to justify their very existence. It is not always easy to unravel some of these often emotive issues and examine the link between gambling and welfare. Lisa Farrell shows that in a rapidly expanding market place it is important to be able to assess the impact of the gambling industry on society in order to inform the policy debate concerning the relaxation of betting and gaming legislation and the licensing of new products. Her chapter explores the theoretical foundations of gambling impact studies. Such studies aim to weigh the costs associated with the betting and gaming industry against the beneﬁts in order to provide a ﬁgure which reﬂects the net monetary effect of the industry on society. Most studies ﬁnd that the net impact is negative, i.e., that the costs are greater than the beneﬁts. However, it can be shown that this result is due to the fact that most studies ignore the direct consumption value derived from the purchase of these products (as well as a number of other theoretically inconsistent assumptions). Farrell deﬁnes the total community costs/beneﬁts of any activity as: Total beneﬁts = private beneﬁts + social beneﬁts and Total costs = private costs + social costs. This can be rearranged as: Net impact = total beneﬁts – total costs = (Private Beneﬁts – Private costs) + (Social Beneﬁt – Social Costs). In the literature to date it is common to include private costs but

14 Gaming in the New Market Environment

to neglect to include the private beneﬁts associated with the consumption value of the product. Other studies only look at social costs and beneﬁts and therefore totally disregard consumer preferences. In essence, most studies are methodologically ﬂawed and this renders them entirely non-comparable. To address these shortcomings Farrell’s chapter reverts back to welfare economics and looks at the theoretical deﬁnitions of private costs/ beneﬁts and social costs/beneﬁts in order to aid applied researchers. Returning to ﬁrst principles provides a useful insight into the existing studies and allows us to suggest improvements to the current methodologies utilised by applied economists when conducting impact studies. However, even with ﬁrm theoretical foundations difﬁculties still exist in trying to apply theoretical models and concepts to complexity at real world situations. In the second article ‘Demand Issues in the Market for Lotto and Similar Games’ (Chapter 3), David Forrest seeks to review attempts by economists and statisticians to model the demand for online lottery games such as lotto. This idea behind this article is the notion that knowing the demand patterns allows us both to make predictions about future developments and to assess the current market structure behaviour of supplies of gambling services. The article starts with an overview of published research and recent results obtained in a number of current lottery-related projects. The purpose is to draw out the extent to which the modelling exercises reported can illuminate the debate on the major trends and issues in the world of lotteries. The dominating issue in demand studies is the sensitivity of demand with respect to prices. Forrest illustrates the importance of the concept of price elasticity in assessing the correctness of the pricing policy by existing lotteries. He focuses not only on the price effects but also looks at the effect of other products on the demand for individual games. This phenomenon – usually called cannibalisation – is important in the modern gaming environment where new games appear continuously and the prices of existing games are manipulated even more often. If the cross-price effects are ignored we are faced with clearly misleading results in terms of the overall price sensitivity of lottery products. Forrest points out that lotteries operate in a fast changing and increasingly competitive gaming sector. Worldwide, potential players enjoy new opportunities to gamble via the internet, whether on lotteries, betting or forms of gaming (such as poker) to which they may not

Introduction 15

previously have had any access at all. The availability of cheaper versions of their own products, and of new potential substitutes, poses a threat to the established operators in Europe and elsewhere and makes the maintenance of their traditionally high take-out rates problematic. The problem of cannibalisation may not only be related to different lottery products but also to other gambling activities, most notably casinos. Forrest summarises some recent evidence from Australia, which seems to suggest that casinos do not exert great impact on the overall level of lottery expenditure except where the network of venues gives easy access to the most of population. Most demand analyses have thus far been done with traditional demand equations which focus on the expected prices (by ‘prices’ we refer to the amounts that individuals pay for lotto – in the aggregate, it is the inverse of the share of prizes out of total sales). Forrest reviews some second generation demand equations taking into account the speciﬁc risk features of the lotteries. He concludes that it is not only the expected price that is important but also the distribution of prizes. Thus far, there has been relatively little evidence on these features. Yet, Forrest convincingly argues that the new models provide a better basis for analysing demand patterns, at the same time offering the suppliers of lottery services new guidance on how to set the prizes. Ian Walker’s article ‘Lottery Design Lessons from Dismal Science’ (Chapter 4) also deals with demand issues. Basically, it explores the use of estimates on how lottery sales vary, in order to understand how changes in design might affect sales. The essence of Walker’s analysis is based on exploiting exogenous variation in the shape of the prize distribution to explain sales variation over time. In particular, he uses the rollover-driven variation in the prize distribution as a source of exogenous variation to test the hypothesis that players are motivated, in part, by the skewness of the prize distribution. While this idea has a long history, it has rarely been tested convincingly in the existing literature. Lotto games have strongly left skewed prize distributions – typically most players lose their stakes and only a tiny proportion receive large prizes. However, when a rollover occurs, the size of the largest prizes on offer usually rise and this causes the degree of skewness in the prize distribution to change. While the dataset that is used in Walker’s analyses does not feature any changes in the design parameters of the game, we can exploit the fact that there is a deterministic relationship between the mean, variance and skewness of the prize distribution (the distribution’s ﬁrst three moments), as well as both the design parameters

16 Gaming in the New Market Environment

and the level of the prize pool rolled over from the previous draw. Since we know how the moments depend on both of these factors, and since we can estimate how sales depend on the moments, we can combine these to infer how sales depend on the design parameters even when there has been no variation in those parameters across the sample. This modelling strategy complements that pursued by other researchers who attempt to analyse the proximate determinants of sales (see, for example, Forrest (2008) and references therein). Both strands of related literature make arbitrary functional form assumptions. In one case the focus is on the role of the prize distribution and the assumption is that this can be captured through a ‘linear’ function of the moments of the prize distribution. One advantage of this more structural approach that emphasises the effect of the moments of the prize distribution on sales is that it is explicitly couched within a theoretical framework that has been widely used, in other markets, to analyse brand choices. It is common for lottery markets to feature a portfolio of games each with particular characteristics so that each game can be thought of as a brand. This theory helps us to understand the relationship between sales of different games in the same market and, in particular, enables us to make inferences about how to design games. Sarah Smiths’s article ‘Lotteries as a Source of Revenue’ (Chapter 5) focuses on the highly controversial ﬁscal aspects of lotteries. Smith notices that lotteries run by states around the world are typically characterised by two features. The ﬁrst is a high tax rate. While a lottery is not itself a tax, as it is sometimes called, the fact is that almost the entire take-out rate (the price of a ticket minus the expected value of the prize) goes to the government as revenue. Smith shows that the tax rates on most lotteries are high – above those on other forms of gambling and those on alcohol and tobacco. In the new competitive gaming environment there is likely to be pressure on governments to reduce lottery taxes. Assuming that the current tax rates are at the revenue-maximising level, which the evidence suggest that they broadly are, cutting the tax rate is the appropriate response if demand becomes more price-sensitive, in order to continue to meet the goal of revenue-maximisation. There is no obvious economic efﬁciency argument for the fact that taxes on lotteries are so high. Taxes on vices, such as smoking or drinking are typically justiﬁed by these activities’ big negative externalities. But the evidence on problem gambling suggests that the negative externality argument is unlikely to justify such a high tax rate in the case of lotteries. Moreover, the fact that demand for lotteries is fairly

Introduction 17

sensitive to price suggests that the consumer welfare loss associated with taxing the lottery is likely to be fairly high, and the so-called Ramsey rule would imply that the overall distortionary cost could be reduced by raising taxes on goods with relatively less sensitive demand, subject to distributional considerations. Reducing lottery taxes would also be likely to improve the overall fairness of the tax system, although the impact would be very small since spending on lotteries is a small part of total income. Smith also looks at the issue of hypothecation, and asks what reduced earmarked revenues might mean for the good causes and for the lotteries themselves. Smith points out that if lottery taxes were cut it would mean a loss of earmarked revenue for the good causes. The evidence suggests that this would have little or no effect on the demand for the lottery – for most people it is the jackpot that matters, not the good causes. However, the loss would obviously be strongly opposed by the good causes themselves. Most studies (from the US) ﬁnd that the good causes do beneﬁt from earmarked revenues, albeit not by the full amount. While the good causes may have been seen as a useful bit of window dressing when lotteries were introduced, they are likely to turn into a political headache if lottery taxes – and the good causes funding – have to be reduced. What is perhaps most interesting in Smith’s analysis is the comparison of welfare and ﬁscal effects of the different taxes, including lottery tax. If we take government expenditures as given we may ask how good or bad lottery taxation really is. Although we cannot necessarily produce deﬁnite numbers, we can establish a proper framework for comparison and predictions. Mark Grifﬁths’ article ‘Problem Gambling and European Lotteries’ (Chapter 6) takes us a bit outside of the economic jargon and focuses on the complex issue of problem gambling mainly from the point of view of psychology. Although most people gamble occasionally for fun and pleasure, Mark Grifﬁths highlights that gambling brings with it inherent risks of personal and social harm in the form of problem gambling. Problem gambling can compromise, disrupt or damage family, employment, personal or recreational pursuits. More speciﬁcally, it can negatively affect signiﬁcant areas of a person’s life, including their physical and mental health, employment, ﬁnances and interpersonal relationships. His chapter examines the extent to which lotteries can cause or exacerbate problem gambling. In doing this, Grifﬁths overviews problem gambling by examining the deﬁnitions of problem gambling and the social context in which people gamble. He also examines

18 Gaming in the New Market Environment

in detail the pathological features of problem gambling; brieﬂy considers some of the main psychometric screening tools that have been (and are currently) used in the ﬁeld; and discusses the consequences and co-morbidities of problem gambling. Grifﬁths argues that the type of gambling also impacts on the development of gambling problems. This has particular relevance for the lottery sector. It also has consequences for understanding the risk factors involved in the disorder, as well as the demographic proﬁle of those individuals who are most susceptible. For instance, certain features of games are strongly associated with problem gambling. These include games that have a high event frequency (i.e., that are fast and allow for continual staking), that involve an element of skill or perceived skill, and that create ‘near misses’ – i.e., the illusion of having almost won (Grifﬁths, 1999). Size of jackpot and stakes, probability of winning (or perceived probability of winning), and the possibility of using credit to play are also associated with higher levels of problematic play. Games that meet these criteria include electronic gaming machines (EGMs) and casino table games. In general, lottery products are not associated with problem gambling with the exception of those that have the potential for continuous gambling (e.g., video lottery terminals (VLTs), scratchcards, and some instant win games on the internet). Further, Grifﬁths overviews other speciﬁc areas in relation to lottery products, including youth gambling and remote gambling (with particular reference to internet gambling). The main form of problem gambling among adolescents has been the playing of slot machines although, as Grifﬁths’ chapter highlights, a number of studies have also indicated that scratchcards can be problematic to a small minority. Grifﬁths goes on to argue that gambling is a multifaceted rather than a single phenomenon. Consequently, many factors may come into play in various ways and at different levels of analysis (e.g., biological, social or psychological). Central to the latest thinking is that no single level of analysis is considered sufﬁcient to explain either the aetiology or the maintenance of gambling behaviour. Moreover, this view asserts that all research is context-bound and should be analysed from a combined, or biopsychosocial, perspective. Grifﬁths also argues that the situational and structural characteristics of the gambling activity are important in the development of problem gambling and that many lottery products have structural characteristics that are unlikely to facilitate problem gambling behaviour. Grifﬁths concludes his chapter by exploring some of the policy implications for lotteries, including the type of research that needs to

Introduction 19

be carried out, legislative policy implications and policy initiatives in the areas of education, prevention and treatment. Matti Viren’s article ‘The Economies of Scale and Scope in the Lottery Industry’ (Chapter 7) takes us to the supply of gambling. More concretely, Viren focuses on empirical evidence on scale economies in the lottery industry. The analyses deal with both European and US state lotteries for the period of the past 15 years. The article is strongly empirical, although certain conceptual and measurement issues are also discussed. Most of the article is devoted to the estimation of the cost functions from different panel data sets. Thus, the record of individual countries is largely bypassed. The analyses, which make use of a large number of data, more than 500 observations, show quite clearly that there is evidence of both economies of scale and scope. In other words, average costs decrease along with the volume production and the number of products. Practically the same result applies to Europe and the US. The paper has potentially important policy implications because the existence of economics of scale (and scope) indicates that a competitive market solution would not probably prevail in the long run, at least in the case where all legal restrictions to market entry and market operations were abolished. Because the biggest company would always have the lowest unit costs it could take over the whole market. The existence of economics of scope suggests that, in the lottery industry, multi-product suppliers would have an advantage over providers of a single lottery product. Thus, it would not be completely unrealistic to expect that we would end up with a very large global multi-product gambling monopoly. Although the results of the empirical analysis seem quite clear, there is still some uncertainty in terms of predictions of costs for very large (global) gambling companies. Thus, it is possible that the long-term solution would not literally be a natural monopoly, but rather monopolistic competition between relatively few large market players. Even then, prices (and consumer welfare) could be anything but an ideal and perfect competition case. Whether regulation could help is an issue brieﬂy touched upon in the article. Viren’s article also presents some evidence on demand behaviour at the aggregate (country/company) level. The results seem to indicate that the aggregate demand is relatively sensitive to the output menu. Moreover, the sensitivity of demand seems to be related to the size of the market in the sense that sensitiveness increases along with market size. Martin Paldam’s article ‘The Political Economy of Regulated Gambling’ (Chapter 8) deals with the public policy issues that are related to gambling. The idea is to apply some basic concepts and ideas from the

20 Gaming in the New Market Environment

public policy research agenda to the analysis of gambling. Paldam points out that the gambling sector is regulated in order to collect taxes and control the gambling problem of addiction. The regulation may be done without or through a state-owned gambling enterprise with monopoly over some or all of the range of gambling products. Paldam introduces a macro model to identify the problem as excess net gambling expenditure and a dead weight loss. The model is calibrated using Danish data, to give estimates of the aggregate cost to society of the gambling problem. Paldam notices that the rents from public gambling monopolies are often distributed in a way that creates stakeholders. The role of these stakeholders represents itself an interesting research issue. Paldam’s view is that those afﬂicted by the gambling problem are few and unorganised. Politicians are under lots of pressure to spend, and the stakeholders are normally powerful groups. Consequently, the goal of tax collection easily dominates over the goal of reducing the gambling problem. Paldam also considers efﬁciency issues: what are the implications of ownership structure for efﬁciency of production? From policy point of view this issue is, of course, highly signiﬁcant. Yet, unfortunately, we have very few hard empirical facts from the gambling world on this speciﬁc point.

1.3

Final remarks

The present book does not cover all areas of the economics of gambling. Especially the area of public policy is so wide and includes so many different issues that it simply cannot be covered in one volume. Because some of these issues have not been analysed in depth in the existing literature, it is even more difﬁcult to make sensible assessments. Issues like ‘gambling and crime’ have been subject to some research but the outcome has typically been limited to the observation that gambling and crime are connected somehow. The exact nature of this connection has not been identiﬁed, nor has it been established empirically. We have deliberately bypassed all the legal issues of gambling. This has been done even if we are well aware that these issues are both important and timely. We have wanted to concentrate on a broad economic perspective avoiding, at the same time, the highly controversial legal and political questions, which will require different sort of legal and political expertise. Whilst dealing with the recent advances of research, this book also covers certain deﬁciencies in the analysis and data. We have had par-

Introduction 21

ticular problems with data. For the demand analysis, there are relatively few data sets, especially outside the UK and the US. There is little information usable for public policy type analyses, such as the ﬁscal issues and cost conditions. For instance, interesting issues of efﬁciency (efﬁciency comparisons) cannot be tackled with the existing information. The lack of data effectively discourages all research attempts on the part of the academic community, thus creating an obstacle for useful cooperation between researchers, on the one hand, and providers and regulators of gambling services, on the other hand. Whether for lack of data or other reasons, there has been almost no research on gambling in continental Europe. This is lamentable, since our market structure and institutions are drastically different from those of the Anglo-Saxon countries, where most research results come from. The book has touched several controversial issues of gambling most of which are related to the role of government in regulation, taxation and production of gambling services. In this respect, this book does not arrive at a given simple result or conclusion. Nor does it provide any policy recommendation. Rather, it shows the level of complexity of the issues we are dealing with. We cannot, for instance, simply say ‘just look at the elasticities, stupid’. Rather we have to control a wide range of factors and issues before we can make afﬁrmative conclusions in any direction. Needless to say, this requires a lot of additional research. It would also be valuable that both the general public and the decisionmakers would know much more about ‘the economics and gambling’ and the issues behind the current controversies. Notes 1 To give some idea of the magnitude of the change, we can mention here that in the European lotteries, sales increased by an average of 4.5 per cent in real terms per annum during 1995–2005, whereas the average payout ratios increased by 4 percentage points (from 48.5 per cent to 53.5 per cent). In the US, the corresponding ﬁgures for growth in sales and payout ratios for the period between 1992 and 2005 were 3.3 per cent and 6.2 per cent). In Europe, rapid growth resulted mainly from new state lotteries’ fast recovery from a very low initial level of sales. In the US, sales have mainly increased due to new state lotteries. In the existing lotteries, the sales between 1992 and 2005 only increased 1 per cent on the average. In 2006, sales in the US increased again considerably, more than 5 per cent in real terms. In 2006, the total sales of European lotteries amounted to EUR 74,850 million (USD 93,912 million), whereas the sales of US state lotteries totalled EUR 45,777 million (USD 57,436 million). The US ﬁgure roughly corresponds to 0.43 per cent of the GDP and 0.69 per cent of personal disposable income (cf. NASPL, 2007, and La Fleur’s, 2007).

22 Gaming in the New Market Environment 2 The survey was conducted by the Finnish National Lottery Veikkaus in April 2006. The questionnaire was sent to all of the EL (European State Lotteries and Toto Association) member companies of which nearly 50 per cent responded. 3 These ﬁgures are obviously subject to the possible sampling errors in the EL questionnaire. The average payout rates may not be very informative because they vary considerably between games even now (and different companies have different gaming menus). This was reﬂected in the abovementioned survey: the payout ratios ranged from 30 to 95 per cent. 4 Between 1894 and 1964, no legal government sponsored lotteries operated in the United States. Even other forms of gambling were severely restricted. Thus, only three states permitted horse racing in 1900. In 1931 Nevada legalised gambling casinos, slot machines and horse racing (but not lotteries, although now it seems likely that the constitutional ban for a state lottery will be abolished). 5 Some lotteries are even older than that. For instance, the Swedish lottery (Penninglotteriet) was founded in 1897, whereas the Finnish Lottery (Veikkaus) was founded in 1940. In Denmark, the history of the Danish State Lottery goes back to 1753. 6 Unfortunately, we do not have complete data on governments’ gambling revenues. Although institutions like the Eurostat compile data on government revenues, the numbers are generally not comparable. Sometimes the numbers may only include direct lottery taxes and leave out the proﬁts which are contributed to government (see the GOV_A_TAX_AG Table in the public ﬁnances statistics, Eurostat, 2007). In the case of state lotteries’ proﬁts it is sometimes hard to say how to treat the direct contributions to good causes that the lottery companies make. If the government de facto determines the beneﬁciaries, the contributions are hardly distinguishable from government transfers. Finally, one has to keep in mind that, in some countries, winnings constitute taxable income, whereas in most countries they are tax-free (on the average, the taxation of winnings only constitutes a couple of basis points of the total government revenue). 7 When the respondents in the EL survey were asked to assess the elasticity of problem gambling to sales the mean value was roughly one. 8 This could be seen in the survey to European lotteries. Up to 70 per cent of the respondents said that if sales increased, say by 10 per cent, costs would increase less. Similarly, as we asked how much the retailers costs would increase as a response to an eventual increase in sales, two-thirds of the respondents said that they would increase less than the sales. 9 Legal cross-Atlantic disputes in the Microsoft case suggest that would not run smoothly, either. 10 Bribe might be paid both to ﬁx the results or to encourage the use of illegal means to achieve better results (e.g., doping). 11 These are issues which, at least in the history of American lotteries, were important and turned the public opinion against all types of gambling during the second half of the 19th century. Various cases are quite eloquently described in Clotfelter and Cook 1991 and Ezell (1960).

2 When Welfare Economics and Gambling Studies Collide Lisa Farrell

2.1

Introduction

The betting and gaming sector attracts a lot of interest from outside parties. Those within the sector view their products as leisure activities competing in a tough market place, whilst groups outside of the industry question the (ethical) nature of these products and their contribution to society. There are few industries (tobacco and alcohol included) which not only have to face the daily struggle to earn a proﬁt but also have to ﬁght to justify their very existence. In this chapter we will try to unravel some of these often emotive issues and examine the link between gambling and welfare. We will critically appraise different measures of welfare and existing studies that attempt to measure welfare effects. We will consider the consequences/impacts of the expanding gaming market on individual and social welfare. It is not however, our intention to provide a full appraisal of the beneﬁts and costs of the betting and gaming sector, the focus here is to provide a better understanding of the methodologies that practitioners use to address welfare issues, so that we can more clearly evaluate the results of such studies in order to make sensible, evidence-based, policy decisions. It is hoped that the chapter provides a valuable introduction to the important theoretical concepts on which the existing empirical studies are based. We shall present only ﬁrst principles of welfare economics in order to make the material accessible to a board audience. The betting and gaming industry consists of a wide range of very different products each with its’ own set of consumers, although some intersection across these sets exists (gambling activities are not mutually exclusive and the difference characteristics of each game results in variation in the groups of consumers who participate – reﬂecting the 23

24 Gaming in the New Market Environment

heterogeneity in consumer tastes and preferences). The market sector comprises the following commercial activities: broadly these are i) wagering (both on and off course), ii) gaming in the form of casinos and electronic gaming machines (EGM’s) and iii) numbers games such as keno, lotto and scratch cards. To undertake a complete review of this sector is not the focus of this chapter, however there some notable attempts to undertake such an activity. See, for example, the study conducted by the Australian Productivity Commission (1999) which provides an extensive and complete picture of an entire countries betting and gaming market. This review was in part possible due to the vastly superior data collected on the sector within Australia.1 Here we will predominantly refer to academic papers and independent reports on the welfare effects of gambling in order to present objective research. There are numerous reports published by interest groups but the difﬁculty of determining the independence of these studies means that most have not been included in this review. There are some general trends across countries in terms of the betting and gaming industry that are important to mention. Most countries have undergone considerable deregulation in the last decade and all countries face the challenge of trying to ‘police’ the growing online gaming market. The betting and gaming industry is at the forefront of technology and is characterised by extensive product innovation which means it often operates within (and exploits) legislative gaps. Historically, most countries imposed a stringent regulatory framework on the industry but a more liberal society and a realisation that valuable and signiﬁcant tax revenues can be extracted from gaming activities has lead to an increasing pattern of deregulation. Trends in the literature follow this pattern of deregulation. As the limits of operation within the market have changed, researchers have naturally asked what the effects of these changes are and from this perspective the literature makes a lot of sense. For example, the issue of the merits of Indian Reverse gaming is unique to the US market2 and the resurgence of literature on numbers games such as lotto by UK academics3 is a direct result of the launch of the new National Lottery game there. So the literature focuses on questions of national importance in the context of the domestic market, after all academic research, in part, evolves to meet the needs of policy-makers. Simply to look at the literature in its entirety and ask if the body or evidence mostly in favour of against the existence of the industry is hennas crime which decontextualises the studies from their historical and cultural setting. Each country has its own social, political and cultural environment meshed

When Welfare Economics and Gambling Studies Collide 25

with unique legislative and historical backgrounds, that means at best the question of the industries net welfare effects has to be asked for each country in turn and even within countries there are numerous state variations that can not be ignored.4 That said, it is not uncommon to ﬁnd government and consultancy reports that do exactly this. Whilst a shortage of studies (often explained by a paucity of data) means that often we do have to look to international studies to understand the impacts of legislative changes the warning still remains, that whilst the signs of the important coefﬁcients may be indicative and helpful, let the policy-maker beware of recommendations based in the size of estimated qualitative effects based on international experience. Here, we will be proactive and ask, ‘What are the limitations of our methodologies and the available data that gives rise to this general lack of evidence for policy-makers?’ We will look at the methods researchers have utilised in order to address questions relating to the welfare effects of the gambling industry and consider how these theories sit within the theoretical framework of welfare economics. It is our intention to provide a structure within which to think about welfare effects in order to be better able to evaluate existing and future studies. Whilst we will not provide an exhaustive review of all the literature we will look at typical examples. In essence, we will consider what we can learn from existing studies and what are the limitations of these studies which inhibit their generalisation.

2.2

Individual welfare effects

Welfare economics has a long tradition in economics. In measuring welfare the distinction made is between individual welfare and social welfare. Individual welfare considers the welfare of individuals (as opposed to communities) under the belief that individuals are the best judge of their welfare. Two essential assumptions are made when measuring individual welfare and these are: ﬁrstly, individuals prefer more to less and secondly, that individuals preferences are revealed through their observed behaviour. This implies that if there is a demand for a good then the provision of this good must be welfare enhancing. If this was not the case then individuals would simply switch their resources to the consumption of commodities which afford them greater utility (satisfaction). In the context of the gambling market this implies that so long as there is a demand for gambles then the provision of products to fulﬁl this demand generates welfare. Moreover, it also implies that markets should expand to the point where there is no demand for

26 Gaming in the New Market Environment

new products. The introduction of a new gambling product will increase consumer welfare because utility-maximising individuals voluntarily purchase the good. That is, assuming income remains constant, individuals must be substituting expenditure away from goods which yield lower utility. The introduction of a tax on the good will erode some of this beneﬁt, but the overall effect is still welfare improving for the player. Whether this is socially desirable is another question and one which we will address in due course, but from the individuals perspective there must be a welfare gain or there would be no demand for the product. Consumer surplus is a measure of the aggregate welfare of the group of individuals who consume a given commodity. It measures the difference between what consumers are willing to pay for a good and the market price of the good that is what they actually pay. It can be calculated by looking at the area bounded by the demand curve and the equilibrium price in the market (the market price). This corresponds to area PAB in Figure 2.1. Where P is the market price and Q is the quantity demanded. Gambling products are usually taxed so it is important to consider the impact the tax has on consumer surplus. Let’s assume that some proportion of the tax t, is passed on to the consumer through higher prices. It is evident from Figure 2.1 that this erodes away a proportion Price A

C

P+ t

B

P

Q Figure 2.1

Consumer surplus

Quantity

When Welfare Economics and Gambling Studies Collide 27

of the consumer surplus – decreasing individual welfare. The consumer surplus is now represented by the area (P + t)CA. In the extreme case when all the tax is absorbed by the producer then the price remains unchanged and the introduction of the tax has had a neutral impact on the level of consumer surplus. Brinner and Clotfelter (1975) and Cook and Clotfelter (1987) both make this point in the context of lottery taxes. One further point to note is that the consumer surplus story is one of partial equilibrium, with more than one market and other methods of taxation. It is welfare improving to raise revenue by taxes that introduce the smallest efﬁciency loss, which may or may not be taxes on gambling. We know of no optimal taxation studies to consider gambling taxes. However, Farrell (1997) does look at the dependence of lottery taxes, income tax and unearned income tax. The comparative statics analysis within suggests that there is considerable interdependence between direct taxation and the revenue that can be raised from gambling taxation. Changes in the tax regime have a signiﬁcant impact on risk seeking behaviour and so gambling taxes should not be perceived as lucrative new sources of revenue. The most recent literature on the taxation of gambling centres on the issue of what should be taxed, stakes (a commodity tax) or expenditure (an ad valorem tax), see Smith (2000). Paton et al (2001) show that an ad valorem tax is superior in terms of individual welfare (an ad valorem tax results in lower prices and greater turnover) and efﬁciency (more revenue is generated for a given level of price and quantity). However, it is sufﬁcient for our purposes to simply appreciate that the introduction of a tax results in a fall in consumer surplus and so decreases individual welfare. It is important to note that the welfare impacts of taxation are further complicated by the redistributional effects of the tax and this point will be discussed in detail later. There have been few empirical studies that have attempted to measure the size of the consumer surplus for gambling products. In the case of lotteries Brinner and Clotfelter (1975) and Cook and Clotfelter (1987) both acknowledge that a welfare gain exists, but they concentrate their analysis on the tax incidence (issues relating to the distribution of the tax burden across the population will be discussed in Section 2.3.2). There are, however, two key studies that we will focus on here. The ﬁrst is Farrell and Walker (1999) who illustrate the methodology for the case of lotto and the second is the Productivity Commission (1999) who take this methodology and rise to the challenge of estimating the consumer surplus associated with the entire Australian gambling market.

28 Gaming in the New Market Environment

Farrell and Walker (1999) directly measure the welfare effects from the availability of the market for lottery tickets following the launch of the UK National Lottery by measuring the associated consumer surplus.5 This was a natural research question at the time the study was conducted given that there was no lottery market in existence in the UK prior to the National Lottery’s launch (with the exception of small charity rafﬂes/lotteries). The analysis is slightly complicated by the fact that lottery tickets can only be purchased in integer amounts and that the real price of a lottery ticket is not the face value of the ticket but the expected value of the ticket calculated as, one minus the expected return. We shall call this the effective price. However, despite these empirical complications the basic principle remains the same. In the case of lottery tickets the supply curve is perfectly elastic – a horizontal line crossing the x-axis at the effective price of a ticket for any given draw. Note that in the case of lottery tickets the effective price is lower in a rollover week as the expected return is higher and this essential price variation allows us to measure the price elasticity of demand and hence to determine the slope of the demand curve. The results of the analysis suggest that in a typical year (with an average number of rollovers) the welfare gains from the game are in the order of £1 billion (equivalent to a 0.5 per cent reduction in income tax) suggesting considerable welfare gains! The relative size of the welfare gain is a reﬂection of the popularity of the game, 63 per cent of the population played on a regular basis at the time of the analysis (rising to 72 per cent in a rollover draw) although they purchased on average only 1.5 tickets per draw (rising to 2.3 tickets in a rollover draw).6 In principle it would be possible to conduct a similar analysis for all the products offered by the betting and gaming sector in order to ﬁnd the total welfare impact. However, in order to perform this analysis we need to know the response of sales to changes in price (the price elasticity of demand) of all products in order to estimate the demand curves (given that consumer surplus is a measure of the area under the demand curve and the equilibrium price). A general paucity of data makes this exercise very difﬁcult in practice. To calculate price elasticities we need to observe price variation and there is the added complication of understanding what constitutes the price of a gamble. It has become standard practice to deﬁne price as the amount paid out by the consumer minus the takeout (player losses), where the takeout is the proportion of any bet (sales revenues) that is not returned as winnings (the takeout covers supplier costs, proﬁts and taxes). This deﬁnition of price is called the effective price in the Farrell and Walker example just dis-

When Welfare Economics and Gambling Studies Collide 29

cussed. Under this deﬁnition price changes usually only occur when there are changes in the tax rates for the product which result in a change in the takeout (given that, the minimum amount that must be returned to players as prize money is either deﬁned under legislation or an item in the license contract issued to the operator). Since tax rate changes are not common events in practice (and tend to be nonrandom events) there are few studies which have estimated price elasticities. Without knowledge of price elasticities we cannot determine the slope of the demand curve and it is therefore impossible to measure consumer surplus.7 Lotteries are an exception where the price changes occur naturally and more importantly randomly through the design of the game, namely the occurrence of rollovers. Indeed, the vast majority of the studies that have estimated price elasticities in gambling markets have looked at lotteries. There are a few studies that have taken on the challenge of estimating the consumer surplus associated with the entire gambling industry and one of these was conducted by the Australian Productivity Commission (1999). They suggest that the net beneﬁt to the Australian consumer of the betting and gaming market is in the order of 4.4 to 6 billion Australian dollars. It should, however, be noted that in order to conduct the analysis a number of necessary modelling assumptions have been made. The most important of these is the assumption that elasticities for gambling activities are constant across national boundaries, across types of gambling products and through time. To obtain estimates of the required elasticities they look at all the estimates currently available in the literature and select preferred estimates. The problem with this approach is that the elasticities vary by type of product yet the Productivity Commission assumes them to be constant across products and further they vary by country. For example, estimates for the price elasticity of lotto are much higher for the USA than they are for the UK and New Zealand. Cloﬂelter and Cook (1990) report an elasticity of –2.55 for the USA, whereas Farrell and Walker (1998) report estimates of –1.55 to –2.6 for the UK and, BERL (1997) ﬁnd an elasticity of –1.054 for New Zealand. It is hard to know how these ﬁndings can be informative of the elasticity for Australia where there are no estimates in the literature. Furthermore, some of the studies which estimate price elasticities for gambling products date from the 1970’s when the range of products available to the gambler was very different from what it is today (following the recent trend in the liberalisation and deregulation) and this will have impacted on price sensitivity. This exercise is further complicated by the issue of

30 Gaming in the New Market Environment

short run versus long run elasticities. Initial responses to price changes may be different from long run equilibrium responses. However our understanding of the dynamics of gambling participation is limited by the fact that there is a paucity of panel data sets available from which we can analyse behaviour through time. A further complication in the measurement of consumer surplus arises due to the presence of pathological gamblers in the market place. Basic economic principals tell us that addictive goods have higher price elasticities and in the context of the betting and gaming industry, this suggests that addicted consumers will be less sensitive to price changes than non-addicted consumers; implying that there is potentially more than one demand curve for each gambling product. For recreational gamblers there is a whole market of leisure activities they can engage in and so they are sensitive to price changes (resulting in a less steep demand curve relative to pathological gamblers), problem gamblers, on the other hand, face a smaller market to satisfy their demand for gambles and therefore are less price sensitive (resulting in a steeper demand curve). In terms of measuring consumer surplus this means that we have to sum the surplus for recreational gamblers and pathological gamblers together. Figure 2.2 illustrates the associated consumer surpluses. The consumer surplus (in the absence of any tax) for recreational gamblers is the area PAB and the consumer surplus for problem gamblers is the area PEF. Again it is noted that the presence of tax reduces the consumer surpluses. Recreational Gamblers

Pathological Gamblers

Price

Price

E

A P+t P

C

P+t B

G F

P Dr

Dp Qr

Quantity

Qr

Quantity

Figure 2.2 Consumer surplus for two types of gamblers – recreational and pathological

When Welfare Economics and Gambling Studies Collide 31

It is often assumed that the presence of pathological gamblers can be ignored in empirical studies, given that their number is small relative to recreational gamblers. However, in terms of levels of expenditure, problem gamblers account for considerable volumes of revenues. In Australia, it is estimated that approximately 2 per cent of the population are problem gamblers and contribute a startling 33 per cent of expenditure revenue (in 1997–98).8 Interestingly, this expenditure does vary by product suggesting that there are characteristics of the gamblers themselves that make some products potentially more addictive than others, for example, problem gamblers account for 33.1 per cent of the expenditure in wagering markets but just 5.7 per cent of expenditure on lottery tickets. Therefore, splitting the playing population in this way may be very important in terms of the accuracy of the estimated total consumer surplus. In acknowledging the different demand curves the Productivity Commission report states that the consumption of problem gamblers over and above that of recreational gamblers represents ‘poor value-formoney’ and so they discount this section of the consumer surplus for pathological gamblers. Whilst this seems logical at ﬁrst glance, it actually violates the theoretical axioms on which the concept of consumer surplus is based. Utility-maximising individuals will cease consumption once marginal utility turns to zero; so simply subtracting away the component of consumer surplus derived from excessive play infringes the very deﬁnition of consumer surplus. If consumers are rational they will have accounted for all the associated impacts with their chosen level of play and therefore ought to receive the full consumer surplus derived from these activities. If they were not enjoying the surplus they would switch their expenditure to other goods. In markets where consumers can be differentiated according to the slope of their demand curves the economic argument usually follows that greater proﬁts can be obtained through product differentiation and hence charging each group a different price. For example, people travelling at peak-times on trains (with inelastic demand) usually pay a higher price for their journey than those travelling off-peak (with elastic demand). It would however be hard to justify the practice of charging addicted gamblers higher prices. That said, most suppliers of gambling products offer a range of different games which differ in accordance to their takeout rate (and hence their effective price), one could think of this as price discrimination if one abstracts from the other characteristics which also differ across gambling products. To summarise, whilst measuring consumer surplus for the betting and gaming sector is in theory simply a matter of calculating the appropriate

32 Gaming in the New Market Environment

area under the demand curve for each product and adding the resultant consumer surpluses, empirical practicalities make it a less than exact science in practice. Understanding price in gambling markets has long been an issue of debate and calculating price elasticities in order to determine the slope of the demand curve is complicated by the fact that price variations are hard to come by and elasticities are generally estimated from non-random price changes.9 In essence, the exercise becomes at best an educated guess, where the outcome is highly dependent on the treatment of surpluses accruing to pathological gamblers. Most studies ignore problem gamblers by looking at a single demand curve (usually due to an inability to identify problem gamblers from recreational gamblers when looking at expenditure or sales data). Some studies acknowledge that they might have differing demand schedules, and the productivity commission report noted here even goes as far as to discount the size of the consumer surplus under the problem gamblers demand schedule. Whilst economists are not new to the realisation that the real world is more complex than theory it is the added complication of problem gamblers in the market place that makes the economic models more difﬁcult to utilise in this industrial sector. That said, the results do suggest that the welfare gain to individuals from the sector is large. Moreover, it suggests that as the market continues to expand the consumer surplus will grow if the games are liked by consumers (that is the games match their tastes for risk). This leads us to an interesting question, ‘how can economists deal with pathological gamblers in their models and estimations?’ 2.2.1

Rationalising addiction

The individual approach to welfare economics is built on the assumption that consumers are the best judge of their own welfare, but if we allow consumers to exhibit addictive behaviour then in principle we violate this assumption. Furthermore, there may be negative costs imposed on other members of society. Indeed, there are numerous clinical psychology and sociology studies that point to the existence of a small group of individuals who consume gambling products excessively, who experience a loss of control whilst engaged in these activities (often leading to ﬁnancial implications) and later regret their actions (resulting in emotional stress and relationship conﬂict).10 Here we have adopted the term pathological gamblers to represent this group, although we accept that there are varying degrees of gambling addiction which members of this group experience. Were it not for the existence of this group the gambling industry would be free from much of the controversy that

When Welfare Economics and Gambling Studies Collide 33

surrounds its existence. That said, it would be hard given the body of evidence to deny their existence and hence in measuring the welfare associated with the industry we need to ﬁnd a way to model the consequences of addiction. For this reason there is a vast body of literature dedicated to understanding addiction as a medical afﬂiction and in aiding recovery. This literature is not however the focus of this section, see Korn (2000) for a discussion of these issues. Here, we focus exclusively on how economists deal with the presence of addiction when trying to measure the associated costs-beneﬁts of the gambling industry. The focus in the economics literature has been on extending rationality in order that utility maximisation remains intact as the central determinant of demand. Lets begin with the hypothesis that individuals may experience some reduction in utility associated with excessive levels of play (we will consider possible impacts of pathological gamblers participation on society overall in Section 2.3). Most models of gambling adopt the expected utility framework (Von Neuman and Morgernstern (1944)); the standard approach for analysing decision-making in the face of risk.11 This model states that the utility gained from consuming a risky product, such as a gamble, can be measured by the expected utility associated with the activity. The expected utility being the probability weighted average of the utilities associated with each of the possible outcomes. In the case of most gambles the expected utility is less than the utility associated with retaining the stake with certainty. So you are better off keeping your money rather than participating in the gamble. Hence the model predicts that people will not participate. In the case of lotto only around 45 per cent of the stake is returned as prizes and this has earned lotteries the title of the ‘suckers bet’. Many casino games offer 80 per cent or more: but also demand greater skill/ability from the players. So why do people gamble? This apparent contradiction may arise because there is some additional non-pecuniary beneﬁt from participating in the activity, gambling is enjoyable! Players enjoy the opportunity to join in this aspect of popular culture and to discuss with friends and family losses and successes. Participation gives them the ability to brag about near misses or perhaps they love the thrill of being at the race track or in the casino. In the case of lotteries there is often an associated donation to charity and one could postulate that players receive a ‘warm glow’ from the knowledge that they have made an (indirect) charitable donation. The standard framework that has been adopted in the literature is to place an additive positive non-pecuniary utility component to the

34 Gaming in the New Market Environment

expected utility. The ﬁrst formal representation of this model in the literature can be found in Simon (1998) who looks at the ‘dream function’ for lotteries12 and this model was later tested empirically by Johnson et al (1999) for the case of betting on horse races and Forrest et al (2002) for lottery players. Here we propose the next logical extension to this approach, to account for the potential disutility associated with excessive play, would be to include a non-pecuniary additive negative component to the individuals expected utility function. The expected utility for a gamble can then be speciﬁed as below, EUit = α + βXit + γEVt + δFUNit + λCOSTSit + εit

(2.1)

Where i denotes the individual and t denotes time, Xit is a vector of personal characteristics thought to be associated with demand such as age, gender, income and so on. EV is the expected value of the gamble, FUN is the non-pecuniary positive utility associated with play and COSTS is both the pecuniary and the non-pecuniary disutility associated with play incurred by the individual – the private costs of play. The non-pecuniary factors associated with play can be thought of as psychic factors such as emotional stress and regret, for example, whereas the pecuniary costs could be those associate with the cost of servicing gambling debts and so on. In this framework demand is a function of both the ﬁnancial investment in the gamble, the fun of playing and any associated private costs of playing. Empirically, one would expect costs to be positively related to the degree of addiction and to be equal to zero for recreational gamblers. The implications of the above framework is that individuals are choosing levels of consumption which have already factored in the costs of play and hence when calculating consumer surplus there is no rationale for treating the surplus associated with excessive play any different from that associated with recreational play. This is more theoretically pleasing than simply removing from consumer surplus calculations the surplus associate with play generated by pathological gamblers or by discounting the utility derived from excessive play in some way. One criticism of the ‘expected utility plus fun’ model is the ad hoc nature of the non-pecuniary fun component and the same criticism can be applied here when we make the costs of participation explicit in the individual’s utility maximisation problem.13 One might even think of interpersonal utility functions as an extension of this model where the utility of family and friends enter as arguments in the individual’s utility function. In this model a gambler would consider the impact

When Welfare Economics and Gambling Studies Collide 35

their behaviour has on those people whose utility they care about and choose their expenditure accordingly.14 Within the mainstream economics the literature there are few studies of addiction since traditional economic theory assumes that preferences are exogenous. There is a strand of literature on habit formation that has evolved from empirical demand system analysis that allows past consumption of a good to inﬂuence current consumption of the good, so that preferences are interdependent through time. This methodology has been used to explain repeated purchases but is not generally thought of as a model of addiction. Explicit models stem from the work of Becker and Murphy (1988) who present and test empirically a theoretical model of addiction. The novelty of their approach is that the individual can behave rationally over her addiction.15 Becker and Murphy’s theoretical model identiﬁes two forms of addiction – myopic and rational addiction. In the case of myopic addiction individuals consider their past consumption levels when deciding on current consumption. This is essentially the same as a single period, linear, lagged consumption, habit formation model. In the case of rational addiction individuals take account of both past consumption levels and future consumption levels in determining current consumption levels. In this scenario individuals are forward looking, they look at expected future prices and take these as an indicator of future consumption levels. If future prices are expected to fall individuals will expect to consume more in the future and hence increase consumption today in order to build up their stock of addiction and so achieve the maximum utility from anticipated future consumption. In essence, individuals behave rationally over their addiction. In the absence of addiction an expected fall in prices would decrease current consumption as consumers intertemporally substitute their expenditure. Farrell et al (1999) show the presence of myopic addiction in the case of lottery players but were unable to test for rational addiction given the absence of anticipated future price variation. Indeed, we know of no studies that test for rational addiction amongst gamblers.16 An important point to note is that the emphasis in the economics literature on rationalising problem gambling implies that the individual understands the potential costs associated with their play and makes consumption decisions taking these into account and hence enjoy the consumer surplus associated with their expenditure decisions. It suggests that welfare should not be adjusted to take account of the harm problem gamblers do to themselves. Indeed, it suggests they adjust their expenditures to account for the private costs of their addiction

36 Gaming in the New Market Environment

and hence society need not be concerned. Moreover, there is no requirement that we should adjust standard economic measures of welfare for this group of consumers. Psychologists may disagree with this statement on the basis that many problem gamblers report that they experience a ‘loss of control’ whist gambling (see Korn (2000)) and are therefore not factoring in the associated private costs of their expenditure at this point. Economists might respond to this by suggesting that they suffer from bounded rationality17 and fail to consider or understand the full consequences of their expenditure. Collins and Lapsley (2003) make a similar point but refer to individuals as having a lack of information and not being fully informed about the probability of becoming addicted and the consequences of addiction – an asymmetric information problem. However, even accepting this explanation, one still faces the challenge of understanding why some individuals experience bounded rationality (or asymmetric information) whilst others, recreational gamblers, do not (or have full information). This line of reasoning takes us back to a personality proﬁle search for addiction characteristics; where one attempts to identify the individual heterogeneity that can explain why some gamblers become addicts and others do not. Economists tend to ﬁnd this approach too ad hoc, and very difﬁcult to formulate in a behavioural model. An interesting departure from the assumption of total rationality is suggested by Donoghue and Rabin (1999) who proposed a model where individuals are allowed to suffer from self-control problems leading to time-inconsistent, present biased preferences. They show that a lack of self-control alone is not sufﬁcient to explain harmful addictive over consumption, one must also assume a degree of unawareness of the potential poor self-control problems that may arise in the future as a result of current consumption. This is a powerful model in that it helps us to understand the transition from recreational play to problem gambling and understanding this transition is especially important in the context of an expanding market. As a ﬁnal caveat to this section it should be noted that the emphasis in the demand literature on the individual’s decision-making behaviour distracts from the games themselves. There may be characteristics of the gambles which raise the probability of players becoming addicted, and so changes in game design may increase/decrease the number of pathological gamblers in society. This is an important distinction, if addiction is an individual speciﬁc afﬂiction then the industry as a whole might offer support to sufferers but can not itself be brought to blame for their condition, if alternatively the industry designs games

When Welfare Economics and Gambling Studies Collide 37

to foster and promote addiction then it might have more to answer to. There are a lot of studies undertaken by psychologists which investigate why some games appear to be more addictive than others. This literature talks about characteristics such as the length of play, the frequency with which play can occur and many other aspects of the gambles, see for example Grifﬁths (1995). Whilst we will not review this literature here we do note that it is clear that not all games are equal in their appeal to pathological gamblers. Lesieur (1998) shows that the expenditure shares of pathological gamblers vary immensely between different gambling products. His results suggest that pathological gamblers have a taste for gambles that occur with high frequency such as wagering and electronic gaming machines, and are less likely to purchase lottery tickets. Interestingly, there are notable gender differences with female problem gamblers having a higher spend on bingo/keno (high frequency numbers games) and male problem gamblers having a higher spend on wagering activities, see for example, Wayne et al (1996). Economists on the whole have not paid much attention to the games themselves as the discipline seeks to explain behaviour so the unit of analysis is usually the individual and the characteristics of the games do not vary across individuals. There are, however, a few studies that have sought to understand the impact on demand of different game designs but these in the main have concentrated on lotto (a game for which the evidence suggests that the percentage of addicted players is low). Let’s quickly note a couple of these. Farrell et al (2000) look at the roll of conscious selection (the idea that lotto players do not pick their numbers at random), in determining sales. They ﬁnd that players do pick the numbers they play in a nonuniform way and this leads to more rollovers than would otherwise occur and results in players getting locked into the game. Rollovers stimulate sales and have a knock on effect of lifting sales in following draws. If player always play the same numbers or have a system for picking their numbers then they can’t risk not playing in any given draw in case their numbers come up. Walker and Young (2001) look at the skewness of the prize distribution as a factor stimulating demand; they argue that lotto players have a taste for long odds. It is important to note that these studies look at the affect of the game characteristics on sales and do not distinguish between sales generated by pathological gamblers relative to those generated by recreational gamblers. Indeed, it is important to make the distinction between characteristics that enforce loyalty and habitual behaviour and those that encourage excessive play. In the case of lotteries if players feel they

38 Gaming in the New Market Environment

have ownership over a set of numbers and are hence locked into buying a ticket for each draw they are exhibiting persistence but not excessive play (there is no beneﬁt to buying more than one ticket for any given set of numbers). One could argue that by promoting conscious selection the operators of a lottery are doing little more than encouraging persistence in expenditure not dissimilar to that encouraged by supermarket reward schemes. There is nothing in these design aspects of the game that suggest that players are not behaving rationally when they participate. Habit and persistence is an accepted feature of applied demand analysis. Thus there is no justiﬁcation for adjusting consumer surpluses for particular games based on design features that foster persistence in play. In summary, this section has shown little support for the argument that in measuring welfare effects through consumer surplus for the betting and gaming industry one needs to adjust the measure to account for the presence of pathological gamblers. Whilst the presence of pathological gamblers is not denied, the economics literature explains addiction whilst preserving rationality and hence if agents are assumed to be rational then they must have factored into their consumption decisions all the relevant private costs and hence receive the full surplus associated with their expenditure. One could, however, argue that whilst the individual might be assumed to consider all the relevant costs borne on himself he has no incentive to consider costs that his consumption might impose on society overall. So how do we measure the welfare impacts of pathological gamblers on society?

2.3

Social welfare effects

Welfare economics looks at the welfare of society as a whole i.e., it is the summation of the individual’s welfares. The literature is divided into two parts: the ﬁrst is the efﬁciency of the production of wealth – how much output can be produced from a given input (i.e., how large is the pie?) and the second is the equity – the distribution of wealth across members of society (i.e., how is the pie shared out?). It is therefore logical to divide our discussion along these lines also. 2.3.1 2.3.1.1

Efﬁciency Deﬁning social costs

Researchers often assume that as well as there being private costs, borne by the individual, associated with excessive gambling participation, there are also costs (of some form) that affect society as a whole. There is a vast literature in this area. Economists have been called to evaluate

When Welfare Economics and Gambling Studies Collide 39

many parts of the betting and gaming industry through cost-beneﬁt exercises to determine if these activities are of net beneﬁt to society. Most of the cost-beneﬁt analysis that has been conducted consider either, the total gambling sector or casino gambling (as a result of their recent growth in the market). There are a number of government studies that try to evaluate the entire industry in order to inform the policy debate regarding the legalisation of gambling activities. This literature is predominantly USA and Canadian in origin: see, for example, Thompson (1996) among others.18 The literature is at best confused and at worse misleading, and the crux of the problem lies in understanding (deﬁning and measuring) what constitutes a social cost. There is also a secondary issue to be addressed, the ﬁrst question being which costs should we label as social costs and the second question is one of how to measure some of the more ‘intangible’ costs such as those associated with emotional stress.19 To bypass this issue some studies use the terminology of economic and psychic costs and state that psychic costs are non-pecuniary in nature and this renders them outside the business of economists. This seems like an excellent way to avoid tackling an interesting (if somewhat difﬁcult) question. Eadington (2003) notes these issues and also provides a critique of the approach made by the Productivity Commission (1999) when deﬁning and measuring social costs in Australia. However, the correct method of quantifying non-monetary costs is not the focus here. We will concentrate our efforts on deﬁnitional aspects of what constitutes a social cost.20 Whilst it would be possible to take each study in turn and ask if the authors ‘got it right’ in deciding what are and what aren’t social costs this would be a long and laborious task and the answer would almost certainty be ‘no’ to some aspect of every study. The reader is referred to Walker and Barnett (1999) for a thorough discussion of the literature in this context. For the interested reader with a need to digest numbers, here are a few of the estimated ﬁgures taken from the literature – make of them what you will! With respect to the entire gambling industry the Productivity Commission estimate the costs for Australia to be between AU$1.8 billion and AU$6.1 billion. Considering they measured the total consumer surplus to be AU$4.4–6 billion this implies a small (potentially negative) net beneﬁt! With respect to casino gaming, Gazel (1998) reports net ﬁgures (beneﬁts minus costs) of US$188.96 million for Wisconsin in 1995 and US$286.78 million for Illinois for 1995. These few examples alone illustrate the variation across studies. See Grinols and Mustard (2001) for the results from a host of studies on the net impact of casinos. Here we will not attempt to unravel the empirical

40 Gaming in the New Market Environment

studies in order to identify the factors which account for this variation. It is sufﬁcient to say that a lack of agreement in what constitutes a social cost is a primary factor. Therefore, it seems more fruitful to ask how far the theoretical literature can help in clearing up some of the confusion associated with the determination of the social costs and beneﬁt associated with gambling. Examples of social costs that have been used in the literature are psychic costs, ﬁnancial costs, treatment costs and productivity costs, among others. Most authors use the term social cost to deﬁne any affect or impact experienced by members in society as a result of the actions of those who participate in gambling activities i.e., any third party impacts. These are termed externalities in the economics literature, but not all externalities are social costs according to the welfare economics deﬁnition of a social cost. The crux of the matter is the usage economic terms without adhering to the strict deﬁnition of that term. What the empirical literature terms social costs are an impact as a result of gambling activities (usually excessive gambling activities) which are experienced by members of society other then the individual herself. However, utilising this deﬁnition the literature has not been able to determine a single set of accepted variables that need to be included in a cost-beneﬁt study of the gambling industry. Moreover, this is not the deﬁnition of a social cost as proposed by economic theory and hence the confusion. Walker and Barnett (1999) discuss these issues in detail and we will summarise some of their arguments below. Social welfare is deﬁned as the sum of the individual’s wealth in the society and welfare economics deﬁnes social costs as the extent to which an activity reduces the efﬁcient use of resources and so reduces total wealth. According to this deﬁnition redistributions of wealth are not social costs. There may be important equity issues associated with wealth redistribution (this point will be discussed in the next section) but they do not represent social costs as they are deﬁned in the theoretical welfare economics literature. Walker (2003) summarises this point nicely when he states ‘(a) social cost must be something other than mere expenditures by a person, or negative consequences to an individual’ (p. 169). The problem is that applied researchers have attributed a variety of different costs to social costs that are contrary to the welfare economics deﬁnition. Moreover, they have summed over social costs and redistributions for one person to another (transfer payments) to give nonsensical aggregates from the perspective of welfare economics.

When Welfare Economics and Gambling Studies Collide 41

Walker (2003) illustrates the welfare economics deﬁnition of a social cost through the use of the production possibility curve and social indifference curves i.e., efﬁciency of production and consumption. Figure 2.3 provides a useful starting point for understanding the economic theory. Consider a simple two good economy: good X denotes legalised gambling and good Y represents all other goods available in society. The production possibility frontier shows all the points of efﬁciency in the production of goods X and Y. The social indifference curve represents societies tastes and preferences for the two goods. Its shape reﬂects the law of diminishing marginal returns: the more you have of good X relative to good Y, the less of Y you are willing to trade for more X and vice versa. Assuming that individuals have some taste for participating in gambles (that is there is positive demand) then social welfare is maximised at point B in the diagram (sometimes called the bliss point). This is where the production possibility frontier is tangent to the highest possible social indifference curve (SIC*). In terms of social costs any cost that moves us to a point inside the production possibility curve reduces social welfare through a reduction in net wealth and meets the theoretical requirements for a social cost.

GOOD Y

B A SIC* SIC

PPF

GOOD X Figure 2.3 gambling

Efﬁciency of production and consumption in a world with legalised

42 Gaming in the New Market Environment

A movement along the production possibility curve to say point A moves us onto a lower social indifference curve (SIC) and is therefore inefﬁcient in consumption. Collins and Lapsley (2003) provide a list of criterion that must be satisﬁed if a cost is to be classiﬁed as private rather than social. Failure to meet this criterion implies that a cost is a social cost in that it reduces net wealth. • Gamblers must be fully informed • Gamblers must be rational, and • Gamblers must be prepared to bear the total cost of their gambling. Whilst this deﬁnition is helpful in a conceptual sense it is debatable if it adds clariﬁcation in a practical sense. As the authors themselves point out in order to utilise it we need to be sure what constitutes rational behaviour. To illustrate the precision of this deﬁnition lets take the well healed (but nevertheless informative) example of the crime of theft. Assume a pathological gambler steals a laptop to fund his gambling activities. The theft itself is simply a transfer from the victim to the perpetrator and therefore the cost of the laptop should not be considered a social cost. However, if the victim becomes security conscious as a result of the theft and purchases new locks and alarms then these costs do represent resources being taken away from other productive activities which they would have been utilised in, in the absence of the crime. Court costs for the theft also represent real resource costs. The literature is full of this kind of illustrative example and highlights how careful one needs to be in deciding on what constitutes a social cost if one is to adhere to the deﬁnition proposed by welfare economics. In summary, the literature to date is still debating the correct deﬁnition of a social cost to use in cost-beneﬁt studies of legalised gambling. The strict economic deﬁnition derived from welfare economics suggests that many impact studies include costs that do not satisfy this deﬁnition (also implying that many studies may have over estimated the social costs of pathological gamblers). Further the variation across studies in terms of what is and isn’t accounted for in the calculation means the studies are not comparable. A standard methodology is clearly needed and the deﬁnition debate looks set to continue. Despite the methodological literature, we know of no empirical studies that have used the welfare economics deﬁnition as its basis for the deter-

When Welfare Economics and Gambling Studies Collide 43

mination of social costs. Whilst some look to economic theory for clear deﬁnitions others, see for example Henriksson (2001), urge for a more interdisciplinary approach that might lead to a new paradigm for researchers to adopt in their empirical modelling. Henriksson argues that much of the confusion comes from the fact that researchers for different disciplines are using the same language to measure conceptually different things. It is therefore only through better dialog between the interested researchers across economics, psychology and sociology that the confusion will be eliminated. However, one suspects that, as the theoretical literature advances we shall ﬁnd out that even with the help of rigid deﬁnitions the real world can still be confusing! 2.3.1.2

Gambling and economic growth

Supporters of the expansion of the betting and gaming industry often claim that the industry leads to economic growth and on these grounds it is a net beneﬁt to society. Economic growth is clearly welfare enhancing as it shifts the production possibility frontier outwards reﬂecting the increase in net wealth. There are two ways the production possibility frontier can shift; the ﬁrst is as a result of an increase in technology and the second is through an increase in the factors of production

GOOD Y

PPF1

PPF2 PPF3

GOOD X Figure 2.4

Shifts in the production possibility curve

44 Gaming in the New Market Environment

(land, labour and capital). Let us illustrate these concepts in the context of the gambling industry with the help of Figure 2.4. The initial state of the economy is depicted by the production possibility frontier labelled PPF1, where Good X represents gambles and Good Y represents all other available commodities. If there is a technological advancement in the gambling sector that allows more output to be produced from a given level of inputs then the production possibility frontier shifts outwards to PPF2. The shape of the new curve shows the change in the opportunity cost between the production of gambling products and the production of other commodities in society. Alternatively, if there is an increase in any of the factors of production (land, labour or capital) and assuming the factor of production is equally efﬁcient in the production of both commodities then the production possibility frontier will shift out to PPF3. The majority of papers which claim that gambling is welfare enhancing due to the fact that it generates economic growth suggest that the mechanism for growth is through the theory of export led growth. This theory suggests that by exporting goods outside of the domestic economy (state or jurisdiction) an inﬂow of income is generated. In the case of the above analysis this additional income can be used to increase the capital used for production (i.e., to purchase new equipment and machinery) and so represents an increase in one of the factors of production. In most empirical studies economic growth (and the resultant employment creation) has been regarded as the principle social beneﬁt associated with the gambling market. It is the potential impact on growth that has led many state governments in economically struggling areas to adopt state sponsored gambling in an attempt to boost local prosperity. This trend has of course been the topic of much debate. Since the 1990’s policy-makers have looked to the legalisation of casino gaming as a means to generate growth. State ﬁnances also beneﬁt from the tax revenues collected. Walker and Jackson (1998) empirically tested the theory that legalised gambling leads to economic growth at the state level in the USA and ﬁnd evidence that growth in the gambling industry promotes economic growth in the state.21 Gazel (1998) states that the impact is more general than just an export effect he lists a host of impacts from employment effects to casinos being a stimulus for construction in the local area. However, Eadington (2001) warns that this growth and development strategy can only be successful if the gaming is licensed strategically. He highlights the fact that, growth only occurs if income is earned from visitors to the region and hence it is important to pro-

When Welfare Economics and Gambling Studies Collide 45

mote resort style gambling opportunities that will attract tourists. A general increase in gambling within the region through the licensing of many small venues will only attract domestic dollars and hence have no effect on net wealth. Some authors have argued that the growth story has been over sold and that there are losses to be considered from the ‘cannibalisation’ of the existing market. Displacement effects refer to revenues lost as consumers divert their consumption from products within the existing market towards the consumption of new products. This is often referred to as the cannibalisation of the market for existing products. Given that consumers have a ﬁxed budget, in order to consume a new product, expenditure has to be directed away from existing products. It is likely that the majority of this lost expenditure will be within the existing betting and gaming sector – but it is worth noting that some expenditure may also be diverted from the wider leisure and entertainment market. There have been a number of attempts to quantify the extent of displaced expenditures within the gambling sector. We will mention just a few here that relate to the impact of casinos on traditional forms of betting and gaming since casinos have been seen by many as the instrument for growth through gambling. Anders and Siegel (1998) and Siegel and Anders (2001) focused on the impact of Indian gaming on lottery revenue in Arizona. They estimate that a 10 per cent increase in the number of slot machines22 in casinos located on reservations is associated with a 3.7 per cent decline in lotto sales. Farrell and Forrest (2006) also look at this issue in the context of the Australian gambling market and ﬁnd evidence that casino gaming displaces expenditure from lottery games. These are just some of the examples of the studies that exist in this area, for a more complete review of the empirical literature see Paton et al (2003). However, such displacement does not imply economic costs or inefﬁciencies. The basic argument is that if one industry expands it must be at the expense of another industry i.e., resources are moved from the production on good Y towards the production of good X. Assuming this demand is voluntary then it must reﬂect the consumers tastes for gambling products. Looking back at Figure 2.1: consider the economy is initially at point A and then due to a taste for gambles resources are diverted away from the production of other goods towards the production of gambles then we move to point B which is clearly on a higher social indifference curve. This is based on the assumption that the supply of gambling is not restricted – if there are supply restrictions then the production possibility curve will be truncated at

46 Gaming in the New Market Environment

the supply limit and depending on the position of the social indifference curve a corner solution may arise. It should also be noted that there may be competitive pressure placed on other suppliers in the leisure sector. Anecdotal illustrations talk about the local restaurant that is forced to close as a result of competition from the new casino, but likewise, from an economist’s perspective, this just represents a transfer of resources following changing consumer tastes and preferences. It is just the natural outcome of a competitive market place for leisure goods and services. In summary, whilst there is some literature relating casinos to growth in economically depressed regions it is unclear how long this growth is sustainable as the opportunities to gamble increase. Arguments concerning the cannibalisation of the existing gambling and leisure markets are essentially red herrings. Whilst there is evidence that individuals substitute expenditure from existing activities towards new opportunities to participate in gambles this simply represents their changing preferences in a wider and more varied market place. If resource allocation follows these changes in consumer taste this is purely an efﬁcient market response and should not be seen as an economic cost. 2.3.1.3

The importance of the counterfactual

As a ﬁnal caveat to the above discussion of social costs; it is important to note that most studies ignore the counterfactual. The counterfactual condition is the basis of the comparison or control group in medicine, natural and social sciences. The experimental or treatment group demonstrates ‘if X is present, then Y is present’ (i.e., if gambling exists in society then social costs are incurred). The control group allows the testing of the idea that if X does not occur, neither will Y (i.e., if gambling is prohibited, then these social costs will not be incurred). Through this basis it is possible to establish causality, and hence control groups are used as one of the important conditions of empirical testing.23 Most of the existing studies relating to the social costs of gambling report ﬁgures along the lines of X dollars a year or X dollars per capita, but they do not report the counterfactual against which these ﬁgures are presented. Collins and Lapsley (2003) suggest 4 possible counterfactuals: i) no past or present gambling, ii) no past or present problem gambling, iii) all gambling banned and, iv) some forms of gambling banned whilst others remain legal. Each of these implies a different estimate of the social cost of gambling. In the majority of cases the implicitly assumed counterfactual (although not explicitly stated) is that of gambling prohibition under the assumption that prohibition is

When Welfare Economics and Gambling Studies Collide 47

costless to society. But if this is the counterfactual being utilised one has to ask what are the social costs of prohibition and is there a resulting net beneﬁt when these costs are considered? In essence, it is important to ask ‘what would have happened in the absence of a gambling industry?’ The answer to this question lies in part in knowing if addiction is a primary disease or secondary disorder and in part in knowing what are the alternative costs associated with prohibition. Consideration of pathological gambling as a primary or secondary disorder is an important step towards determining causality. Just because excessive gambling participation is correlated with social costs does not mean that gambling is the root cause of these costs. Correlation does not imply causality. We need to know if in the absence of a gambling industry the proportion of the population deﬁned as pathological gamblers would have no outlet for their addiction and hence be free from their afﬂiction or if they would simply divert their expenditure and time to other activities which may have greater (or perhaps smaller) social costs. Shaffer and Korn (2002) provide some evidence on this point. They note that there is evidence of co-morbidity amongst pathological gamblers, that is, other disorders may coexist. However, it is still hard to tell if disorder X causes disorder Y or vice versa. Whilst problem gambling and the associated mental disorders may be statistically related, they too do not have a clear cause and effect relationship. They also note that pathological gambling along with its associated comorbidity is a multidimensional disorder and is best thought of as a syndrome.24 Finally, they note that, pathological gambling is not diagnosed if the behaviour can be better described as a manic episode. Whilst this evidence might sway one to think of pathological gambling as a secondary disorder the medical profession is far from being able to conclusively make this statement. This leaves us in a difﬁcult position when trying to consider the outcome of the counterfactual of gambling prohibition on the alternative behavioural outcome. It is also worth noting that prohibition is not without its own set of associated costs. An important point to note is that the removal of the activity from the market place would mean the loss of the consumer surpluses discussed above. Secondly, due to the large potential proﬁtability of the industry there is an incentive for rent seeking behaviour to occur (see Walker and Barnett (1999) for a full discussion of these ideas). That is, interested parties may devote large amounts of resources to lobbying governments to legislate in their favour. Further, supporters of the ban on gambling will need to continually devote resources to lobbying for the continuance of the ban: all these activities require

48 Gaming in the New Market Environment

resources that could otherwise be used in a productive manor. Such behaviour means that resources are not being utilised in the most efﬁcient way and so society moves to a position inside the production possibility frontier and hence a social cost is created. Finally, the potential for illegal gambling activities means that a society exercising a prohibition on gambling needs to divert resource towards policing and enforcing the ban. The potential for illicit gambling (historically associated with criminal activities) suggests further impacts that would need to be considered if the counterfactual was to be properly formulated. 2.3.2

Equity

The ﬁnal aspect of welfare to be considered in this review is that of equity. Equity is the concept of fairness or justice in economics, particularly in terms of taxation and welfare economics. We will only brieﬂy discuss the relevant issues here in the interest of completeness but issues of taxation will be discussed in detail in later chapters of this book. One common argument in the literature is that the recent wave of liberalisation of gambling legislation is a reﬂection of the potential for jurisdictions to generate tax revenues from these activities. It is argued that gambling activities can be taxed successfully and at high rates with little public objection. The reason for this is twofold. Firstly, many people view the taxation on a ‘sinful’ product as just and therefore acceptable. This concept stems from the history of high taxation on tobacco and alcohol which were both taxed quite heavily long before health issues where taken into account. Secondly, there is no obligation to participate and high taxation is perceived as a necessary price to pay for participation in products of this nature and a mild deterrent from engaging too frequently. Of course there are limits on the levels of tax that can be set. If the rate is too high it provides an incentive for illegal participation. Sceptics have said that gambling taxes are voter friendly new source of tax revenues but the study noted earlier by Farrell (1997) suggests this logic is ﬂoored by a lack of consideration of the relationship between gambling taxes and the wider tax system. There may be some cases where institutional factors also play an important role. For example, in Australia gambling taxes are collected by the state whereas income tax is collected by the federal government and hence this factor leads to the attraction to gambling revenues by state legislators despite their interdependence in the wider tax system. In terms of the production possibility frontier any redistribution caused through taxation simply moves us along the curve, having the poten-

When Welfare Economics and Gambling Studies Collide 49

tial to move society away from the bliss point to some other position on the production possibility frontier and hence leads to inefﬁciency in consumption. That is, assuming there are no costs of enforcement or collection which would divert resources away from production and hence move us to a point inside the production possibility frontier. By deﬁnition these costs would represent the social costs associated with the tax. So taxation is simply a redistribution and as such not a social cost. Nevertheless one might be concerned about the fairness of this redistribution of wealth within society. If gambling is an activity concentrated amongst the poorest in society then they will carry the burden of the tax incidence. Regressive taxes reduce the tax incidence of people with higher incomes and shift the incidence disproportionately to those with smaller incomes. To understand if gambling taxes are regressive it is essential to know the pattern of expenditure across the income distribution. It is feasible that gambling may become more attractive to low income groups if it perceived as a means to increase ones wealth. However, it is not clear that the rich have no incentive to participate; a high stakes card game might be a perfect way for them to satisfy their taste for risk. Income elasticities of demand tell us how gambling expenditure responds to changes in income and hence are the ﬁrst step in understanding the incidence of play across the income distribution. However estimates of the income elasticities of demand for gambles are hard to ﬁnd, although some do exist. For example Haig and Reece (1985) report income elasticities of between 0.6 and 1.0 for horse racing in the US, Mason et al (1989) ﬁnd elasticities of 0.3–0.8 for Las Vegas gamblers and Farrell and Walker (1999) report ﬁndings of 0.1–0.4 for lottery play. As expected, gambles associated with higher levels of addiction have higher income elasticities. However the elasticities are all positive and between zero and one suggesting that gambles are normal goods. With respect to equity this is an important result as it shows us that as income increases spending on gambling increases but at a less than proportional rate i.e., individuals at the lower end of the income distribution spend a higher proportion of their income on gambles than those at the higher end of the income distribution and hence carry the greater tax burden. It is important to note that these studies only report income elasticities calculated at mean levels of income and expenditure but it is likely that the income elasticity for problem gamblers will be much larger than that of recreational gamblers. Unfortunately the expenditure surveys from which these estimates are calculated do not contain enough information

50 Gaming in the New Market Environment

for the researcher to distinguish between problem and recreational gamblers. To investigate the tax incidence issues further Suits (1977b) considers the issue of regressitivy of gambling taxes using data from the US National Survey of Gambling Behaviour 1974. He calculates a regressivity index, to compare the regressivity of lottery taxes with other gambling taxes. Lotteries are found to be among the more regressive of the gambling taxes and casino taxes are found to be progressive.25 It should be noted, however, that this study takes place before the recent rapid growth of casino gaming with large areas devoted to electronic gaming machines. It should also be noted that there are no studies of tax regressivity that account for the presence of pathological gamblers. Given that pathological gambling usually leads to ﬁnancial difﬁculties it is likely that the regressivity of gambling taxes is downward biased by this group relative to the measure one observe in their absence. It would be interesting to know what the degree of regressivity of gambling taxes looks like for the group of recreational gamblers alone. But as well as the tax incidence we need to think about how the tax dollar is spent. Gambling taxes have been accused of imposing a ‘reverse Robin Hood effect’ on society – they take from the poor and give to the rich. Cooper and Cohn (1994) discuss in detail the budgetary incidence of the revenues generated from a lottery tax which may heighten or reduce the regressivity of the tax. The central argument is that if the revenues raised fund activities primarily participated in by the wealthy then this heightens the regressivity of the tax. This hypothesis has been investigated in the case of lotteries as the tax on lottery play tends to be hypothecated and therefore it is easier to assess how the taxed is spent. Feehan and Forrest (2007) considers this question of the UK National Lottery whose tax revenues go to a number of ‘good causes’. He found that more afﬂuent areas of the country receive larger transfers and are also known to purchase fewer tickets. In summary, the evidence suggests that the taxation of gambling is regressive. However there is a need for updated studies following the recent expansion of the market for gambles and its integration into the wider leisure market. In the case of lottery taxes this regressivity appears to be heightened by the nature of the activities that tax revenues support. However, where gambling taxes accrue to general government spending this question is much harder to address as speciﬁc revenue ﬂows are difﬁcult to determine. Finally, taxation represents a transfer of wealth within society and is therefore only of interest from a social

When Welfare Economics and Gambling Studies Collide 51

welfare perspective in terms of equity (assuming that it is efﬁcient to collect).

2.4

Conclusion

This chapter has reviewed the literature on the welfare effects of gambling, in terms of both individual and social welfare. We have looked at the theoretical backgrounds of the empirical literature and asked if welfare economics can help clarify areas of confusion amongst practitioners. Whilst we have by no means surveyed all the literature we have drawn from typical studies in order to give the reader a ﬂavour of the analysis that has been conducted regarding the study of the welfare impacts of gambling. Our primary conclusion is that there is huge potential for good theoretical work on appropriate methodologies which would allow applied researchers to deliver methodologically sound and comparable studies. Most of the current literature is best viewed as case speciﬁc studies and reviews that have attempted to draw collective evidence on welfare effects invariably conclude that i) empirical studies are sparse and ii) methodologies are diverse. Welfare economics is grounded on the assumption that consumers are rational in making consumption decisions. Under this assumption the beneﬁts from gambling activities can be measured through the associated consumer surplus generated by the industry. The few studies that have measured consumer surplus suggest that it is quite large. However, the presence of pathological gamblers has caused some to doubt the validity of the utility maximisation framework in the context of gambling markets. Yet this theory, with a few tweaks, is surprisingly capable of explaining gambling whilst preserving rationality. Moreover, one might argue that utility maximisation is (in the very least) appropriate for approximately 98 per cent of participants in the market place (the recreational gamblers) and in this light it seems strange that there is pressure for the most fundamental economic model of individual behaviour to be dismissed so quickly. And if one is to dismiss utility maximisation in the case of the gambling industry then one has to ask if this should be the case for other industries too, such as, those of tobacco and alcohol, and if so: where do we stop? The power of utility maximisation and rationality is that it is universal across products and individuals. The danger of its abandonment is that things become ad hoc with different behavioural models for different products and different types of consumers. It remains to be seen if such an alternative paradigm can yield better predictions and it is worth remembering that the purpose of models of economic

52 Gaming in the New Market Environment

behaviour is not to mirror all aspects of the real world, but to simplify them, whist still allowing the policy-maker to accurately simulate behavioural responses to regime changes. If we are to entertain the possibility of an entirely new paradigm then the growth of health economics as a subdiscipline offers an exciting opportunity to combine the vast clinical knowledge on pathological gambling with models of economic behaviour. By not adhering to the welfare measures of costs and beneﬁts proposed by welfare economics this paper suggests that most studies underestimate the beneﬁts (mostly by ignoring private beneﬁts of consumption value measured by the consumer surplus) and over estimate the costs (by wrongly including transfers between individuals as economic costs): resulting in signiﬁcantly lower net welfare than economic theory would suggest. We conclude that as yet practitioners have not produced studies based on sound theoretical foundations for us to be able to draw conclusive quantitative evidence: despite the desperate need for such evidence to guide policy-makers at this crucial point in time when the market place for gambling products is expanding rapidly. Notes 1 Readers are referred to the Tasmanian Gaming Commission, who maintain and update state level gambling data for Australia. 2 See, for example, Anders (2003). 3 See, for example, work by Farrell, Forrest and Walker: amongst others. 4 See ‘Australian Gambling Comparative History and Analysis’ (1999) prepared by the Australian Institute for Gambling Research, University of Western Sydney for a comprehensive discussion of the importance of historical factors in shaping the current Australian gaming market. 5 The UK National Lottery is a standard (6/49) lotto game and was launched in 1994. 6 Source: National Opinion Polls data collected on behalf of the Ofﬁce of the National Lottery. 7 The reader is refereed to Table C.1 in the Productivity Commission review for a summary of the existing literature which has estimated price elasticities for betting and gaming products. The vast majority of these studies look at lotteries. 8 Figures obtained from the National Gambling Survey conducted by the Productivity Commission. The American Psychiatric Association estimates the ﬁgure to be 1–3 per cent in the USA. 9 If the price change is non-random the investigator cannot be sure if the changes in demand that are observed are true price effects or if they are a response to the factors that brought about the price change. 10 There are many clinical and psychological deﬁnitions of problem gamblers and this description includes those aspects which appear with greatest frequency in the literature although we accept that not all problem gamblers experience the same symptoms in relation to their addiction.

When Welfare Economics and Gambling Studies Collide 53 11 There are some non-standard expected utility gambling models but these are few and the majority of the theoretical literature exists within the expected utility paradigm. 12 The dream refers to the fact that the motivation for the purchase of lottery tickets is to enjoy the dream of winning and to fantasise about how winnings would be spent. Interestingly, this was not a new concept to the marketers of lottery products but no formal model had previously been speciﬁed in the economics literature. 13 See Hartley and Farrell (2002) for a further discussion of this and a general discussion of modelling gambles within the expected utility framework. 14 Interpersonal utility functions are often used to explain the motives for leaving bequests. 15 This model has been mostly utilised and tested for the consumption of physically addictive goods, for example Becker, Grossman and Murphy (1988) consider the demand for cigarettes and Walter and Sloan (1995) consider alcohol consumption, however it is perfectly applicable to activities like betting and gaming. 16 The theory of rational addiction has been tested for physically addictive goods where tax rate changes where anticipated which caused price rises. Becker, Grossman and Murphy (1988) look at the demand for cigarettes and Waters and Sloan (1995) consider alcohol consumption. 17 See, for example, Kahneman (2003). 18 See Walker and Barnett (1999) Table 1 for other examples of studies that look at the costs of gambling in total and those that look at the impacts of casino gaming. 19 See Collins and Lapsley (2003) Table 1 for a break down of social costs into tangible and intangible costs. 20 I do make some apology to the interested reader for once again side stepping the measurement issue. 21 This is one of the few papers to utilise states without gambling opportunities as a control group against which to measure impacts – that is, the authors consider the counterfactual. 22 Tribes are not obliged to ﬁle turnover and other ﬁnancial data and the number of slots therefore had to serve as a proxy for dollar expenditure on slots. 23 Panel data studies provide some opportunity to allow for the counterfactual. The statistical methodology makes use of observations (usually States) where gambling is prohibited as the base against which to compare the impacts in other States where gambling is permitted. 24 The description of a syndrome usually includes a number of essential characteristics: which when concurrent lead to the diagnosis of the condition. Frequently these are classiﬁed as a combination of typical major symptoms and signs – essential to the diagnosis – together with minor ﬁndings, some or all of which may be absent. 25 Farrell (1993) looks at tax Lorenz curves and Suits S-index for the UK and ﬁnds that the taxation of most types of gambling was regressive (except for the case of horse race betting where proportionality was found). However, this study was prior to the launch of the UK National Lottery which considerably changed patterns of gambling and hence these results may no longer hold.

3 Demand Issues in the Market for Lotto and Similar Games David Forrest

3.1

Introduction

This chapter seeks to review attempts by economists and statisticians to model the demand for online lottery games such as lotto. There is an overview of published research and further illustrations are drawn from results obtained with my co-authors in a number of current lottery-related projects. Throughout I attempt to draw out the extent to which the modelling exercises reported can illuminate debate on major trends and issues in the world of lotteries. Lotteries operate in a fast changing and increasingly competitive gaming sector. Worldwide, potential players enjoy new opportunities to gamble via the internet, whether on lotteries, betting or forms of gaming (such as poker) to which they may not previously have had any access at all. The availability of cheaper versions of their own products, and of new potential substitutes, poses a threat to established state-owned operators in Europe and elsewhere and makes it problematic that their traditionally high take-out rates can be maintained. Within Europe, state-sanctioned monopolists also face legal challenges based on claims that the freedom to trade across national boundaries within the European Union should apply equally to gambling as to other services. Thus, for example, bookmakers based in the UK have attempted to assert their right to trade in Italy and Greece and have challenged the legality of the policy in The Netherlands that internet service providers may not permit access to gaming sites located outside the country. Many such cases remain outstanding at the European Court of Justice and their outcomes will determine the body of case law on which the future shape of gambling in the European Union will depend. Early 54

Demand Issues in the Market for Lotto and Similar Games 55

judgements, in the Schindler and Gambelli cases, recognised that member states may restrict the availability of gambling services for social reasons, but only if rigorous conditions are met. Barriers to trade within the Union could not be based merely on an intention to protect existing tax revenue or proﬁts. Any restrictions that were maintained had to be non-discriminatory, proportionate to the objectives and consistent with other policies. The conditions set down have been widely interpreted as signalling that state-owned enterprises providing lotteries, betting and other forms of gaming are unlikely in the future to be allowed to enjoy the territorial monopolies under which they have operated until now.1 Perhaps the most striking development in global gambling in the recent past has been the relentless increase in the number of jurisdictions permitting machine gaming either within or outside casinos. This trend has been strongest in North America and Asia where there appears to be a ‘domino effect’: as more countries allow casinos, there is pressure on neighbouring states to do the same, to capture the expenditure of their own citizens who would otherwise gamble across the border. Thus far, Europe has lagged the rest of the developed world in casino investment. There are no large casinos on the international model such as are found in Las Vegas, Macau, Sydney or Cape Town. But the United Kingdom proposes to open a ‘super casino’ in 2009 and The Netherlands debates a Vegas style casino in Maastricht (close to borders with Belgium and Germany). If, in due course, Europe follows America and Asia in the direction of developing a new style casino sector, this is likely to intensify even further the competitiveness of the market in which the ofﬁcial lottery agencies operate. So, on several fronts, European lotteries face intensiﬁcation of competition. This poses a number of issues for them and for the societies in which they operate. In general terms, increased competition drives down prices.2 For example, horse betting is popular in both France and the United Kingdom. In France, bettors can wager only through the state-sanctioned monopoly Pari Mutuel-Urbain, where take-out is more than 27 per cent; in the United Kingdom, betting is available at a number of rival operators and they typically retain only 13 per cent of stakes. In any jurisdiction where new entry is permitted, there seems little reason to suppose other than that the current national monopolies will see the price the betting market will bear driven down to the levels observed in Britain. Any attempts by governments to tax bets to compensate for loss of proﬁts from state-owned betting operations are likely to be frustrated if their citizens are permitted to bet offshore via the internet.

56 Gaming in the New Market Environment

For lottery operations, falling prices for other gambling services will raise questions such as: how readily do lotto players switch to alternative modes of gambling when relative prices change? If they offer improved value for money on their own products, will committed lotto players buy more tickets? Would changing the prize structure of the games they offer help protect their markets? if state operators combine forces to offer multistate games with bigger jackpots, will this help maintain market share in gambling or merely cannibalise existing national games? For society as a whole, erosion of prices across the range of gambling services also raises questions. How much more of household budgets will be drawn into gambling? And will new gambling opportunities just reallocate spending within the gambling sector or will they signiﬁcantly raise gaming spend (with concomitant implications for the incidence of problem and pathological gambling)? Demand modelling can potentially answer many or most of these questions. This chapter focuses on the demand for lotto and similar games. The literature reviewed offers many insights but also has deﬁciencies. Amongst these is that models were developed that successfully tracked lottery sales and they offered answers to questions on the appropriate level of take-out; but the leading studies look back to the period before competition in, and globalisation of, gambling markets and there is an urgent need for new research employing contemporary data. Further, the academic literature is strongly orientated towards the English speaking countries where the ‘economics of gambling’ has become an established and respectable branch of the discipline. In most of Europe there are very few practitioners and this is reﬂected in a paucity of evidence on the behaviour of relevant gaming markets. There is a compelling case for European lottery operators to commission econometric research pertinent to issues raised in this chapter. Of course many operators in Europe have wider gaming interests than lotteries; but because the academic literature has come mainly from the United States and the United Kingdom, the focus of research has been very much on lotto (scratchcards, another offering of state agencies in the English speaking countries, have been largely ignored by investigators, probably because data are not so readily available and because the terms available to players tend often to have been invariant over time, frustrating any hope of drawing inferences concerning player preferences over game characteristics). Throughout America and in Britain, developments in lotto have followed a broadly similar trajectory with sales of the original game suffering secular decline and agencies responding by introducing new

Demand Issues in the Market for Lotto and Similar Games 57

games, grouping together in lottery blocks to offer multi-state lotto with more exciting jackpots and, more recently in American states, accepting a lower take-out. Demand modelling may offer a route to evaluating the success of such responses to falling sales and also offer a clue on the extent to which new gaming products from the private sector are likely further to erode the market. Figure 3.1 presents the time series of Saturday Lotto sales for the case of the United Kingdom National Lottery. It is typical of the data series that researchers have attempted to model. Two features stand out. First, there is a pronounced downward trend in sales from about 1997 (for example, average sales fell from approximately 60m in 1997 to just 38m in 2006, with real sales revenue falling even more drastically once inﬂation is taken into account). Second, there is considerable variation week-to-week as consumers respond to jackpots augmented by rollovers or by bonus money added to the prize pool when the operator declares a ‘superdraw’. These two features have been incorporated into the standard speciﬁcation employed by authors ﬁtting statistical models q = f(ev, trend, controls) Here variations in sales (q) are modelled as being accounted for by variations in the expected value (ev) of holding a ticket,3 by the passage of time as represented by ‘trend’ and by a selection of other variables

140,000,000

120,000,000

100,000,000

Sales

80,000,000

60,000,000

40,000,000

20,000,000

0 1994

Figure 3.1

1995

1997

1998

1999

2001 Date

2002

2004

Sales of Saturday Lotto tickets in the United Kingdom

2005

2006

58 Gaming in the New Market Environment

included as controls. The estimation method, as pioneered by Gulley and Scott (1993), is typically two-stage least squares, to allow for the fact that, while an increase in ev (caused, for example, by rollover money boosting the prize fund) is presumed to affect sales, sales in turn also change expected value (as there are more entries, the probability of there being no jackpot winner diminishes and ev is increased). In the two-stage least squares exercise, the amount rolled over from the previous draw is usually employed as a convenient instrument for ev. This permits statistically efﬁcient estimation of the coefﬁcient on ev. The coefﬁcient estimate shows how strongly demand responds when the value-for-money offered by a ticket is improved. The focus in the literature has been on this issue of how responsive demand is to value. We shall return to the question below and ask what has been learned and whether the results can be trusted. But one might note in passing that it would be potentially more useful to the industry to understand why trend is relentlessly negative. Authors often cite that the trend term in their equations captures increasing ‘boredom and disillusion’ with the game; but there is likely to be more to it than that. Lotteries are a ‘network good’ such that the attractiveness (to any one individual) of the proposition on offer depends on how many other people are playing because this determines how exciting the level of the jackpot is. If there is an exit from the existing player base, this reduces the jackpot, lowers the appeal of the game to remaining players and provokes further exit. This sets up the possibility of a downward spiral that could eventually lead to the collapse of the game if the jackpot were to pass below some critical level necessary to make any signiﬁcant number of people play at all. Unfortunately, models generally only allow for negative trend, they do not explain it. Because risks to operators posed by negative trend are serious, it would be desirable to reﬁne models by building into them dynamics that would capture the possibility of a vicious circle of declining sales. This has not been done as yet. But by choosing appropriate controls for inclusion in the speciﬁcation, existing models can at least be used to address issues such as how introduction of new lottery games and new alternative forms of gaming might have modiﬁed the severity of effects from trend.

3.2

Take-out

The standard Gulley-Scott model, set out above, was designed explicitly to help assessment of whether changes in take-out rate would be

Demand Issues in the Market for Lotto and Similar Games 59

capable of raising greater amounts of net revenue for government. For example, if more from each ticket were paid into the prize fund, players might buy many more tickets and, even accounting for greater payouts, the state would then be better off. But will the response of players in fact be sufﬁciently great? The model proposes that we can ﬁnd out from econometric analysis of how players responded in the past to the draw-to-draw variation in value for money that occurs when rollovers and promotional draws result in some draws having prize pools that are more generous than usual. What the Gulley-Scott model yields is a demand curve that shows predicted sales at all possible values of ‘effective price’ (effective price is the cost of purchasing a ticket minus the ev of winnings). This has policy relevance to operators because it shows what would happen if they changed take-out, for example a 5 cents reduction in take-out would lower effective price by (close to) 5 cents and the demand curve predicts by how much sales would increase. The demand curve is invariably found to be downward sloping, reﬂecting that consumers indeed respond to better value (more prize money, lower take-out) by, collectively, purchasing more tickets. Just how readily they respond is captured by own-price elasticity of demand. For example, if, at a speciﬁed take-out, elasticity were –2, this would mean that, on the basis of the demand curve estimated, a change of –1 per cent in take-out would induce a change in number of tickets sold of +2 per cent. In this case, operator revenue net of prizes (which equals take-out multiplied by number of tickets sold) would increase. Reducing the take-out would then be advantageous to the operator because a sufﬁciently greater number of tickets would be sold to more than compensate for the increase in the amount from each one that was paid into the prize fund. If elasticity were found to be –2 at the current take-out, take-out would have been wrongly set and should be reduced. Similarly if elasticity were estimated as –0.5 (inelastic rather than elastic), this would show players to be relatively unresponsive to take-out which could then proﬁtably be increased (a 1 per cent increase in take-out would be met by only a 0.5 per cent fall in ticket sales, allowing net revenue to rise). The test for whether current take-out is already consistent with net revenue maximisation is that elasticity should equal –1. In this circumstance, a small change in take-out would be met by an equiproportionate change in number of tickets sold, so that net revenue would remain the same. This would demonstrate that net revenue was already being maximised.

60 Gaming in the New Market Environment

Gulley and Scott applied their pioneering model to data on sales of lottery games in Ohio, Kentucky and Massachusetts. In the ﬁrst two of these states, they found no evidence of mispricing; and in Massachusetts, take-out was also broadly ‘correct’ though there were two lotto games and there appeared to be some scope to gain extra revenue by increasing take-out on one and reducing it on the other. Generally, their study and subsequent ones in the United States, found no evidence that the high take-outs on lotto games could be lowered and still maintain net revenue. A number of studies for the United Kingdom have applied the Gulley-Scott model to estimate elasticity for the national lotto game. Their results are summarised in Table 3.1. The earliest, that of Farrell and Walker (1999), reported elastic demand. But this ﬁnding was not replicated in subsequent studies where elasticity of demand was always estimated close to –1 (and never statistically signiﬁcantly different from –1 at the 95 per cent level of conﬁdence). Thus the early FarrellWalker result appears to have been misleading. The explanation is that they studied a data period where ﬁndings were unduly inﬂuenced by two particular draws with very high jackpots that occurred within a month of each other, about one year after the start of the game. These draws yielded huge publicity because the prizes offered were unprecedented in the United Kingdom. The resulting ‘lotto frenzy’ pushed sales to very high levels that were never to be repeated on future occasions when the jackpot was unusually large. Subsequent authors had longer data series to analyse and, with these, elasticity estimates settled at close to –1, indicating that legislation setting the take-out rate had got the price about ‘right’. Lotteries everywhere could take this as evidence that high take-outs relative to other forms of gaming were justiﬁed, to the extent that the policy goal was to maximise proceeds Table 3.1

Summary of demand elasticities Period

Draws included

Observations

Estimate of elasticity

Farrell and Walker (1999)

Nov 1994– Feb 1997

Saturday

116

–1.55

Forrest et al (2000)

Nov 1994– Oct 1997

Saturday/ Wednesday

188

–1.03

Forrest et al (2002)

Feb 1997– June 1999

Wednesday Saturday

127 127

–1.04 –0.88

Demand Issues in the Market for Lotto and Similar Games 61

for government, presumably because willingness to pay was higher for a long-odds/ high prize gaming product. Of course, it cannot be taken for granted that similar studies applied to contemporary data would yield similar results given changes in the market such as lower popularity of lotto and greater competition from new forms of gaming. With consumers enjoying more choice in gaming opportunities, the lotto market might, over time, become more sensitive to take-out. On the other hand, if large numbers quit lotto because they become bored with the game and prefer new forms of gambling, then those that remain may represent a hard core of players who are relatively insensitive to value for money. Whether these effects are present and which is dominant can be settled only empirically. Ongoing research on demand in the market is therefore likely to be valuable.

3.3

New generation demand models

However, mere replication of the earlier studies would now be regarded as unsatisfactory. The Gulley-Scott model has been subject to increasing scepticism. It has two principal ﬂaws. The ﬁrst ﬂaw is that it assumes that consumer response to one-off changes in expected value, associated with rollovers and promotional draws, will be replicated if changes are applied to every draw. In fact, response might well be less because greater sales in a rollover are likely to be partly attributable to inter-temporal substitution, such as some players saving all their expenditure for special draws. This would imply that model estimates of sensitivity of sales to take-out are biased upwards. The second ﬂaw is that we observe only player responses to variations in the expected value that occur in rollovers and promotional draws and these are inevitably accompanied by a change in prize structure because all the bonus funds accrue to the top-tier prize. It is implausible that consumers are indifferent to prize structure4 and therefore the model is wrong in attributing changes in sales only to changes in the value-for-money on offer. Consumers might in fact behave differently if faced with similar improvements in expected value accomplished by a lowering of take-out that spread more generous payments across all prize pools. This implies that model estimates are biased in an unknown direction. The ﬁrst ﬂaw has not been addressed in the literature. In principle, it could be if the model were applied to data from a jurisdiction that had changed take-out permanently. But there has been some attempt to deal with the second.

62 Gaming in the New Market Environment

While the Gulley-Scott model was an elegant innovation econometrically, its representation of player preferences may not have been adequate. Buying a lottery ticket can be interpreted as investing in a ﬁnancial asset that has an uncertain return. In the British case, there is an approximately .98 probability of winning nothing and an approximately .017 probability of winning a ﬁxed £10 prize. The probabilities of winning other amounts are very small, with the shape of the distribution reﬂecting the amounts in the various prize pools and the possible number of winners between which these pools may be shared. Figure 3.2, devised by Ian McHale, presents a stylised (not to scale) representation of the complex probability distribution that UK lotto players acquire when they purchase a ticket for a non-rollover draw. As noted above, there are two substantial ‘spikes’, corresponding to the ﬁxed prizes of zero and £10. The rest of the distribution is continuous because the amount paid to a winner depends not only on how many numbers on a ticket are ‘correct’ but also on how many other winners at that prize level there are and on how much of the prize fund has already been used up on ﬁxed £10 prizes. Local maxima in the probability distribution relate to the different prize pools (four numbers correct, ﬁve numbers correct, ﬁve numbers and the bonus ball correct, all six numbers correct).

£0 winners NOT TO SCALE

Probability

£10 winners

4 ball winners 5 ball winners

5 plus bonus ball winners

Jackpot winners

Winnings

Figure 3.2

Probability distribution of winnings in a typical UK Lotto draw

Demand Issues in the Market for Lotto and Similar Games 63

Rollovers provide an opportunity to observe how players respond to a new probability distribution of pay-offs when extra money is paid into the six ball (jackpot) prize fund. Rollovers have the effect of translating the rightmost segment of the probability distribution further to the right. The response of sales in such a case is assumed in the Gulley and Scott model to be purely a response to variation in expected value but players may in fact be inﬂuenced by other changes that a rollover presents (for example, possibilities of a higher absolute pay-out than is possible in an ordinary draw) and that are represented in the probability distribution of pay-offs. By modelling sales as dependent not only on expected value but on other measures that capture the complex probability distribution sold in the form of a lotto ticket, it is potentially possible that one could then examine more reliably what would happen if different take-outs and prize structures were introduced. Walker and Young (2001) proposed capturing the essential features of the probability distribution by three statistical measures rather than one as in the Gulley-Scott model. They calculated the variance and skewness, as well as the mean (expected value), of the distribution offered in each draw and included all three measures in the speciﬁcation of their lotto sales model. This speciﬁcation implies that gamblers are interested not just in overall value for money (the proportion of stakes that is expected to be returned in prize money) but also in issues such as how likely they are to win and the spread of prizes across pools. To an extent, mean, variance and skewness capture the terms of the game in all these dimensions. A study of returns to betting at American racetracks by Golec and Tamarkin (1998) had reported evidence consistent with gamblers having a positive preference for value for money (mean), a negative preference for risk (variance) and a positive preference for skewness (strong enough to allow odds on risky ‘outside’ bets to deliver lower expected value than bets on favourites and yet still ﬁnd takers). This pattern of preferences does not appear to characterise all betting markets (Forrest and McHale, 2007) but seemed to hold amongst British lotto players according to Walker and Young (2001), who reported estimated sales equation coefﬁcients on expected value, variance and skewness that were positive, negative and positive respectively. Obstacles to successful employment of the approach to demand modelling initiated by Walker and Young include that the calculations for variance and skewness are by no means straightforward and that correlation between variance and skewness is high. It is also difﬁcult to

64 Gaming in the New Market Environment

allow for the endogeneity issue that expectations of all three moments would be expected to inﬂuence sales by modifying consumer behaviour but sales then inﬂuence the actual values of the three moments in a deterministic sense (indeed Walker and Young estimated with ordinary least squares, thus ignoring the issue altogether). Notwithstanding these problems, I illustrate here what ﬁndings look like and imply when mean, variance and skewness (the standardised third moment about the mean) are included in a model to account for a historic time series of lottery sales (UK Saturday Lotto sales from launch in 1994 to early 2007). The model, the result of work currently in progress with my colleague Ian McHale, has three stages. The ﬁnal stage estimates the sales equation q = f(ev, variance, skewness, trend, controls) Given that ev, variance and skewness are all endogenous, ‘expected’ rather than actual values are included when this equation is estimated. Expected values of ev and variance are generated in stages 1 and 2. In stage 1, sales were regressed on a set of strictly exogenous variables: trend, controls, amount rolled over from the preceding draw, amount announced to be used to top up the prizes in certain superdraws, and a series of dummy variables representing various sizes of minimum jackpot guaranteed in different superdraws.5 In stage 2, we used ﬁtted values of sales from stage 1 as an input in the calculation of corresponding measures of mean, variance and skewness for each draw. These values were then employed in estimation of the equation of interest (stage 3). Results appear in Table 3.2. The ﬁndings contrast with those of Walker and Young to the extent that, while consumers respond positively to expected value, they appear from this estimation exercise to have a negative preference for skewness. That skewness here is statistically signiﬁcant as a determinant of sales is certainly suggestive of the limitations of the expected value model because this demonstrates that players are indeed inﬂuenced by the structure, as well as by the total value, of prizes. However, because all the variation in the three moments in the data set is driven by rollovers, there is high correlation between variance and skewness and this makes it problematic to infer aggregate player preferences from coefﬁcient estimates. The problem of colinearity does not however prevent the results from being employed for forecasting the level of sales if take-out and prize structure were changed in a speciﬁed way. In a standard, nonrollover United Kingdom Lotto draw, 45 per cent of sales revenue is

Demand Issues in the Market for Lotto and Similar Games 65 Table 3.2

Regression results, UK Saturday Lotto sales (millions) coefﬁcient

Constant Mean Variance

117.14 91.66 .000002

t-statistic 8.16 6.68 0.86

Skewness

–.024

–8.05

Trend

–.026

–15.69

ﬁrst lag of sales

.19

9.48

second lag of sales

.18

9.11

preceding Wednesday sales

.12

4.02

–2.96

–2.74

Dummy: Wednesday Lotto available Dummy: Thunderball available

–1.01

–2.44

Dummy: Princess Diana’s funeral

–9.86

–4.06

Dummy: ﬁrst two double rollovers

20.95

9.30

Number of observations

637

paid into the prize fund. Suppose this were increased to 50 per cent, i.e., take-out was reduced by 5 pence per ticket. According to the equation, the effect this would have on sales would depend on how the extra pay-out were distributed across the prize funds: different decisions on how to use the enhanced prize money would modify variance and skewness in different ways. The simplest way to implement a decrease in take-out of 5 pence would be to leave the ﬁxed prize for three correct balls at £10 and then allocate remaining prize money between the higher prize funds in the same ratios as now. For a standard Saturday draw (with controls set to plausible values and trend to a value corresponding to 2007), we calculated values for mean, variance and skewness for existing arrangements and for this alternative, lower takeout regime. Using the stage 3 equation estimate, we forecast sales of 34 million with a 55 per cent take-out and 41 million with a 50 per cent take-out. Lowering the take-out would in this case raise ‘revenue net of prizes’ for the draw by a non-trivial 9.6 per cent. The exercise reported illustrates the mean-variance-skewness approach might prove capable of being used to provide more soundly based guidance than hitherto. However, even with reﬁnement, it would still suffer from the fundamental problem that, if there have been no prior

66 Gaming in the New Market Environment

changes in game structure, all the variation in the properties of the lottery ticket (mean, variance and skewness) that are observed in the data originates with rollovers and occasional promotional draws; the data reveal the response of sales when different terms are offered on the ticket for one draw only. It remains a leap of faith to suppose that players would respond similarly to changes in terms that were introduced on a permanent basis. This sounds negative. But prospects for future model building based on an expected value-variance-skewness approach are nevertheless good because the lottery market is changing. In Britain the Saturday Lotto game still dominates other products offered by the operator but, as elsewhere, its relative importance has declined as a broader portfolio of games has emerged. The new games, by design, offer different packages of risk and skewness from Lotto because they are intended to appeal to segments of the market with particular tastes in these dimensions. Some of the new games themselves feature rollovers, generated independently of those of Lotto, providing considerable week-to-week variation in the package of choices as deﬁned by mean, variance and skewness of returns from the respective games. Further variation in the choices available to players occurs as the operator experiments with games offering unusual formats or prize structures. With all this variation present in the data sets, modelling how relative and absolute sales of each product respond week-to-week is likely to be fruitful in terms of informing policy on take-out and on the package of games that would generate most revenue. Another likely fruitful research strategy would be to focus research on jurisdictions where there have been signiﬁcant changes in arrangements for the main lotto game on offer, for example Florida and Texas where the game was made harder to win. Of course, such changes tend to be introduced when a game is in crisis and care would be needed in attributing any increase in sales to a change in the mean-varianceskewness offer rather than to recovery from a downturn secured by a relaunch and associated marketing.

3.4

Cannibalisation

Thus far the tone of the chapter has been that, to date, existing demand models have not proven capable of generating robust ﬁndings that would settle issues such as whether the take-out rate or the structure of Lotto should be changed. That is not to say that the models have no insights to offer. If expected value/ variance/ skewness are viewed as in

Demand Issues in the Market for Lotto and Similar Games 67

fact control rather than focus variables, some interesting insights may be gained from results on other covariates included in the model. For example, if operators tend to face a downward trend in sales of their lead game, they must replace lost revenue by innovating with new games. Their innovation will be successful only to the extent that new offerings do not simply cannibalise the sales of the core product. The results in Table 3.2 suggest that the introduction of Wednesday Lotto in 1997 took nearly 3 million from Saturday Lotto sales while the addition of Thunderball to the portfolio in 1999 has drawn about 1 million from Saturday Lotto. Both ﬁgures indicate that sales of the new game were only to a limited extent at the expense of the original game. When new games are introduced, it is often in recognition that player preferences are not homogenous and there may be segments of the market that would respond to a game with a radically different meanvariance-skewness proﬁle than Lotto. Forrest and Alagic (2007) trace the history of one game in Britain that was certainly different. Lotto Extra was offered as part of the United Kingdom National Lottery’s portfolio of games between 2000 and 2006. Played on Saturdays and Wednesdays, it had the same 6/49 format as Lotto but all of the prize fund was allocated to the ‘all six balls correct’ prize pool. At the disappointing levels of sales achieved for this unusual product, rollovers were the rule. Indeed, at one time 104 consecutive Wednesdays passed without Table 3.3

UK Saturday Lotto sales regression coefﬁcient

Constant Trend expected loss, Lotto

15.00 –.001

–22.08

.19

6.27

previous Wednesday sales

.002

number of observation r-squared

2.53

–.09

–7.31

.13

5.21

–.078

3.05

dummy: EuroMillions available expected loss, EuroMillions

27.25 –18.59

–1.29

log of previous Saturday sales

dummy: Superdraw

t-statistic

401 .915

t-statistics calculated with robust standard errors, dependent variable is the log of sales, two-stage least squares estimates.

68 Gaming in the New Market Environment

a winner. The result was that sometimes large jackpots accumulated (there was no cap on the number of rollovers) such that the expected value of a ticket reached nearly £1.40 (compared with a purchase cost of £1). Even when the expected return was positive and so large, sales barely reached 1 million, indicating little interest in a product that offered such extreme variance and skewness. Such low sales, even when expected return was actually positive, demonstrate very clearly of course that it is inadequate to model behaviour as driven solely by expected value. Variance and skewness also potentially matter.

3.5

EuroMillions

A special case of new game occurs when operators combine to offer multijurisdictional versions of Lotto. This has been the strongest trend in lotteries worldwide over the last ten years. All but one of the 42 American states offering lotteries are now afﬁliated to a lottery block and the last hold-out state in Australia has joined the others in replacing its domestic Saturday game with participation in a cross-country version. In Europe, the Nordic countries have long combined and in 2004 operators in the United Kingdom, Spain and France joined forces to offer EuroMillions. Six other national lottery organisations afﬁliated subsequently and this has allowed jackpots of up to nearly €200 million to emerge to provoke outbreaks of ‘lotto frenzy’ across the block. In its application to renew its licence to run the United Kingdom National Lottery, Camelot proposes forming a worldwide alliance that promises yet greater prizes. Multi-state games have the obvious advantage of combining populations to permit greater exploitation of the peculiar scale economies of lotto. This was illustrated when New South Wales became the ﬁnal addition to the Australian lottery block. Sales in all the existing member states grew because the extra population provided by New South Wales made the potential jackpot so much greater. Of course, smaller states have more to gain from cooperation as the development of their game is particularly constrained by the smallness of the jackpots their populations can support. A more subtle advantage for them is that barriers to entry are raised when existing games grow in scale. A new operator will ﬁnd it more risky to enter the market if it has to offer higher initial jackpot guarantees to persuade enough early players to buy and so make the market for its product self-sustaining. The relationship between existing games and a new multi-state game will likely vary between small and large jurisdictions. Therefore, research on the impact of EuroMillions on the sale of local lottery products

Demand Issues in the Market for Lotto and Similar Games 69

should be extended to cover a variety of countries. Here I note only preliminary results obtained for the British case. In the United Kingdom, lottery sales and jackpots were already high and the beneﬁts to the operator are therefore likely to be less than in small states like Portugal where EuroMillions permitted the lottery agency to offer the population dramatically high prizes for the ﬁrst time. For the United Kingdom, EuroMillions was to be a supplement to national Lotto rather than a replacement. But what would be the relationship of Lotto to EuroMillions? Two issues arise. First, will sales of EuroMillions lower the size of the Lotto market by causing some player expenditure permanently to switch to the new game? Second, will a particularly attractive EuroMillions draw eat into Lotto sales that week? Matheson and Grote (2006) ask similar questions concerning the relationship between state-speciﬁc and multi-state games in the United States. They ﬁnd that typically local game sales fell when a state joined one of the two cross-country lottery blocks though the displacement was only partial so that membership was still worthwhile in terms of revenue from the expanded portfolio of products. But they also found that in the new regime, where local and national games were offered side by side, the two emerged as complements, i.e. in weeks when, say, Powerball offered an enhanced jackpot, this induced greater sales of Powerball but also greater sales of local games. A similar ﬁnding emerges for the United Kingdom. Ian McHale and I estimated a Gulley-Scott type demand function for Saturday Lotto to assess the impact of the new EuroMillions game. To provide as clean an environment as possible for testing, we restricted our period of analysis so that it began with the introduction of Thunderball in 1999. Thus, EuroMillions was the only major product introduced during the period. We included a dummy variable set equal to one from the week that EuroMillions was launched and a price variable, the expected loss from purchasing a EuroMillions ticket that week. The combined results (Table 3.3) suggest that sales of Saturday Lotto were enhanced in the period when EuroMillions was available and that this effect was magniﬁed when EuroMillions became more attractive because of rollover. Of course, it is possible that the time trend has been speciﬁed too crudely (i.e., the unknown counter-factual could be that the rate of decline of Lotto sales may have slowed even without EuroMillions). But the results make it implausible that cannibalisation was a problem and certainly suggest that, once EuroMillions was in place, it proved complementary to Lotto: when rollovers made the multi-state game more attractive, this actually proved of beneﬁt to Lotto sales. The timing of the European draw

70 Gaming in the New Market Environment

may have been crucial to this outcome. If people join the queue at the lottery booth on Friday, the next Lotto draw is close enough for it to be worthwhile to buy a Lotto ticket as well. Even for those who planned to buy on Saturday anyway, the earlier purchase beneﬁts the operator because it eliminates the possibility that the customer will for some random reason fail to make it to the booth next day. The results suggest that the game has been well designed for the United Kingdom market and that it is possible to develop new games that actually help the old.

3.6

Relationship of lotteries to other modes of gambling

The strand of research discussed above focuses on the offerings of lottery agencies themselves. It shows that there is a potential for the development of new games, speciﬁcally multi-state lotto, to help sustain current levels of revenue as competition in the gaming market intensiﬁes. But how serious is the threat to lotteries from other sectors? A distinction here is drawn between the impact on lotteries as the price of current rival modes, such as betting, is driven downwards and the impact as new forms of gaming become available, for example electronic gaming machines in casinos or casino style gaming on the internet. Both these threats are likely to be of increasing concern to the operators. In many jurisdictions within the European Union, the opening of the betting market to bookmakers from other member states looks likely as a result of the principles laid down by the Court of Justice in the Gambelli case. William Hill, a major British bookmaker, has, for example, recently applied for licences to operate in Greece. Such incursions are likely to drive down the price of betting. Will this take custom from lotteries? Again, Europe as a whole has been slow to permit modern casino style gaming but liberalisation is on the agenda in many countries, for example with mega-casinos proposed in Manchester and Maastricht. Would increased accessibility to high prize casino style gaming machine (in or outside casinos) initiate decline in lottery play? And of course everywhere Internet gaming poses a competitive challenge with no guarantee that it will be possible in countries like France and the Netherlands to sustain the policy of restricting access to foreign gaming sites. Will the Internet be the death of conventional lotteries?

3.7

Cross price elasticity

I ﬁrst address the issue of whether, for example, improved value to consumers in betting will lead to that sector attracting expenditure pre-

Demand Issues in the Market for Lotto and Similar Games 71

viously spent on lotteries. The review of the scientiﬁc literature presented in Swiss Institute of Comparative Law (2006) certainly illustrates that bettors in America and Britain are conscious of value for money in the product. Across nineteen studies, the average estimate of own price elasticity of demand for betting was –1.76, i.e., a 1 per cent drop in take-out was typically associated with a near 2 per cent increase in betting volume. Typical of consumer response was the substantial boom in United Kingdom betting which followed the abolition of betting duty in 2001 (Paton et al, 2004). With so many studies generating a ﬁrm consensus as to how elastic demand is, it would be reasonable to predict that falling take-outs in Europe, at least in countries with a strong betting culture, will stimulate betting sectors hitherto constrained by the high take-outs associated with monopoly control of the market. Unfortunately there are far fewer studies which investigate how much of the extra betting will be new gambling and how much will represent expenditure drawn from lotteries (and other modes). This is an issue of cross-price elasticity of demand. A condition favouring strong effects from price of one product to amount sold of another is that the two share a common pool of potential customers. In the case of gambling, there is evidence that markets do in fact overlap heavily. In Forrest and Gulley (2006), we looked at the spending diaries of 6,634 United Kingdom households which took part in the government’s Family Expenditure Survey. Data distinguished whether and how much each household spent on Lotto, scratchcards, the Irish Lottery, other lotteries (e.g., run by charities), bookmaker betting and bingo. Table 3.4 shows correlations between participation in Lotto and each other form of gambling and in every single case the coefﬁcient is positive and highly statistically signiﬁcant. Table 3.4 also shows, for households who engaged in each of a pair of activities, the correlation between level of spending on lotto and level of spending on the other mode. Again correlation is always positive and in most cases highly statistically signiﬁcant. Thus many households to whom the Lottery appeals also display a taste for gambling elsewhere. Consequently it is reasonable to suppose that spending by current lottery players may be displaced to some extent if existing modes of gambling become better value. As noted, there are few studies that attempt to measure these cross effects between modes. Paton et al (2004) use monthly betting tax revenue data from the United Kingdom to assess the impact on betting of variations in the effective price of lotto (averaged across the draws in each month). They estimated a cross-price elasticity of +0.35, i.e., the

72 Gaming in the New Market Environment Table 3.4

Spending on Lotto and other models of gambling

Correlation between participation in lotto and participation in – football pools

.095 (.000)

other lotteries

.046 (.000)

Irish Lottery

.070 (.000)

scratchcards

.137 (.000)

bingo

.109 (.000)

bookmaker betting

.152 (.000)

Correlation between level of spending on lotto and level of spending on – football pools

.238 (.004)

other lotteries

.577 (.103)

Irish Lottery

.209 (.265)

scratchcards

.234 (.000)

bingo

.325 (.000)

bookmaker betting

.143 (.000)

p-values are in parentheses Source: UK family expenditure survey (2001), 6634 households.

activities were substitutes, a 1 per cent fall in the effective price of a lottery ticket reduced betting by 0.35 per cent. In Forrest, Gulley and Simmons (2008), we addressed the same issue with the advantage of much less aggregated data provided by the industry. We were able to model daily betting volumes, on horses, dogs, football and numbers games, at a major United Kingdom bookmaker over a period of more than ﬁve years. For all sectors except dogs, we identiﬁed a relationship between turnover and the size of jackpot on offer in the next scheduled draw of Lotto. Thus not only do the populations of lottery players and bettors overlap but they are also ready to reallocate funds between the bookmaker shop and the lottery booth. The degree of cross-price elasticity is, however, as in Paton et al, fairly low. The modelling literature therefore suggests strongly that competition in the betting sector will improve value for money for consumers and lead to substantially more betting activity. However, the limited investigation of cross effects between modes suggests that only a limited amount of this extra betting will be at the expense of lottery games.

Demand Issues in the Market for Lotto and Similar Games 73

3.8

Displacement of lottery sales by new modes of gambling

A possibly greater threat to lotteries may emerge as customers encounter entirely new forms of gambling. Internet gaming is one focus of concern for the industry but also for social reasons: lotteries are regarded as particularly ‘soft’ form of gambling whereas internet gaming is likely to be a more serious source of problem gambling given that it presents opportunities for repeat and fast paced play. As yet, econometric modelling has little to say about displacement of Lottery spending by rival internet gambling options. Developments are too recent and data still scarce. But there is strong evidence that liberalisation of conventional gaming, another manifestation of increasing competition, has the capacity to damage state lottery operations. Much of the evidence is from the United States where several studies demonstrate that lottery revenue is threatened when high prize slot machines become more accessible (Elliott and Navin, 2002, Fink and Rork, 2003, Walker and Jackson, 2008). Further, the impact is large.6 In Farrell and Forrest (2008), we checked whether Australian experience was the same as American. Australia provides a convenient laboratory in which to examine the impact of legalising machine gaming. A small number of what Australians call ‘casinos’ (what others might term mega-casinos) have appeared across the country, roughly one per state. Seven of the eight states and territories have also legalised ‘gaming halls’ which can house typically 100–200 casino-style, high or unlimited jackpot gaming machines. In some states these are very thick on the ground, with provision as dense as, for example, that of bookmaker ofﬁces in British cities. Because legalisation of such venues occurred at different dates in different states (and not at all in Western Australia), there is sufﬁcient variation in the data to permit statistically precise estimates of the degree of cannibalisation of online lottery games by the new facilities. Further Australian data distinguish by state between turnover and player losses in casinos on the one hand and gaming halls on the other. In our preferred speciﬁcation we regress turnover of a state’s online lottery games on state casino turnover, state gaming hall turnover and a set of control variables. We employ ﬁxed effects instrumental variables estimation with a panel of eight states and territories. Our ﬁnding is that the growth of ‘casinos’ had no effect on lottery sales. Recall though that Australia uses the word ‘casino’ to refer to a very small number of destination facilities, visits to which are necessarily infrequent for most

74 Gaming in the New Market Environment

of the population. The implication for Europe is that a single large isolated casino in Manchester or Maastricht would not be expected to harm the lottery market. However, the ﬁnding is very different for the gaming halls which provide local service in city centres and suburbs in most Australian states. Here we ﬁnd that every extra dollar placed in a slot machine results in lottery spending falling by 1.8 cents. Such is the volume achieved in the gaming hall sector that this implies that, at recently reported levels of activity in gaming halls, individual state lotteries have lost 20–30 per cent of their online games revenue directly as a result of the emergence and growth of machine gaming. The Australian evidence conﬁrms, but is yet more compelling, than the American. It tells us more because it is possible to distinguish clearly between effects from isolated mega-casinos and effects from machine gaming in neighbourhood locations. It is the more accessible form of casino-style gaming (with high jackpots) that provides the threat to lotteries. Perhaps few European jurisdictions would contemplate providing such a network of, effectively, neighbourhood casinos as is now in place in New South Wales or Victoria. But internet gaming can provide even more accessible opportunities to lose money on casino style games and it has yet to be learned how serious this competition might be to state lotteries. Notes 1 A comprehensive review of previous cases and several incisive commentaries on evolving European case law are provided in Littler and Fijnaut (2006). 2 In economic analysis, the price of a gambling service is commonly identiﬁed with ‘take out’, i.e. the proportion of stake retained by the operator. This varies widely, from as high as 50 per cent for many lotto games to as low as 2 per cent in the most competitive machine gaming markets in Nevada. 3 Several articles present models expressed in terms of ‘effective price’ but this is merely the difference between the price of a ticket and its expected value. 4 Indeed, Forrest et al (2002) demonstrated that there was misspeciﬁcation, for example, if size of jackpot pool were excluded from the sales equation. 5 In a few early superdraws, the operator announced a bonus payment with which the jackpot was to be topped-up. The amounts are treated as exogenous. These top-ups are therefore akin to rollovers. In a large majority of superdraws, however, the operator has advertised a minimum level of jackpot. This has always been binding in the sense that extra money was indeed paid into the prize fund but the amount added was endogenous since less will need to be paid the more successful such a promotion is in boosting number of tickets sold. We therefore accounted for these draws with dummy variables for various levels of guarantee rather than actual payment into the prize fund. 6 An extensive review of American evidence is provided in Walker (2007).

4 Lottery Design Lessons from the Dismal Science Ian Walker1

4.1

Introduction

Understanding gambling is a problem for economists. The economist’s traditional toolbox is not well equipped to understand why risk-averse individuals would participate in gambles that are actuarially unfair bets. However, gambling is a pervasive feature of most economies and a number of simple departures from traditional economic thinking have been suggested in the literature to attempt to reconcile gambling with the idea that individuals are risk averse. One recent phenomenon in betting markets has been the growth of lottery products. The scale and scope of lotteries has expanded considerably over recent decades and there has been considerable controversy over their increasing use by governments as sources of revenue. In most economies lottery operators are charged with raising revenue that can then be used either to support general expenditure or is earmarked for particular purposes. Indeed, operators are generally expected, within bounds, to maximise the revenue that they raise. We will take it for granted that such gambles are not (or rather, are not just) ﬁnancial assets which have no intrinsic value in themselves. Rather we view a lottery ticket as a product within which is embedded a number of characteristics some of which consumers like and some they don’t – but that, at least for some individuals, the attractive characteristics outweigh the unattractive ones. Some of these characteristics are ﬁxed and while these might contribute to sales, on average, because they are ﬁxed they cannot explain variation in sales over time. For example, the number of lottery sales outlets is usually relatively static over time. Thus, here we conﬁne our attention to the effect of those characteristics that vary over time to explain the variation in sales over 75

76 Gaming in the New Market Environment

time. In this we exploit a peculiar feature of many lottery games – that they are pari-mutuel bets where the number of winners in any particular draw is a random variable. This leads to a probability that there will be no winner in any draw (a rollover). Lottery games that have this feature are usually referred to as lotto and are designed so that when a rollover occurs the prize pool is transferred to the next draw. In this chapter we explore how the sales of lottery tickets are affected by the shape of the prize distribution that is ultimately determined by the way in which the lottery game has been designed. By shape, we mean the distribution of the prize fund across the winners of different prize pools. The size of the overall fund is determined by the take-out rate, the proportion of sales revenue that is not put into the prize fund (and so is used to pay for the costs or is given to the government or its agent), and the level of sales. Most lotto games feature a jackpot prize pool that is reserved for winners of the hardest to win jackpot prize (usually matching all n of the balls drawn), another for those that fulﬁl the easier task of matching n-1 balls, etc. Each prize pool will consist of a given proportion of the overall prize fund. Thus, in lotto games there is random variation in the number of winners of each prize pool from draw to draw. And, in particular, the number of jackpot winners varies so that there will sometimes be no jackpot winners in a given draw The rules of lotto then usually dictate that the unclaimed jackpot pool gets transferred to the following draw and added to that jackpot pool – a phenomenon known as a rollover. Casual empiricism shows that rollovers are associated with higher sales in the subsequent draw because they increase its jackpot prize pool.2 Thus, the distribution of possible winnings is determined by: game design parameters, such as the take-out rate and the given proportions of the overall funding going to each prize pool; random ﬂuctuation in the number of winners in each prize pool; and the size of any rollover amount from the previous draw. The design parameters are known to players and are ﬁxed; the rollover amount varies from draw to draw but is known one draw in advance; the number of winners of each prize pool depends on the number of players and the numbers that are drawn in that draw (if popular numbers are drawn there will be many winners) which are not known in advance. In practice there are many complications that have to be faced, but the essence of our analysis is based on exploiting the variation in the shape of the prize distribution to explain sales variation over time. In particular, we examine how variation in the prize distribution affects sales to test the hypothesis that players are motivated, in part, by the

Lottery Design Lessons from the Dismal Science 77

skewness of the prize distribution. Lotto prize distributions are highly left skewed because almost all players lose their stakes, with a small proportion winning small prizes, and a very small proportion winning very large prizes. But when a rollover occurs the size of the largest prize rises and so the degree of left skewness falls because the largest prize suddenly becomes larger. In addition to testing the proposition that skewness motivates gambling, our analysis has a practical purpose: we use this estimated model to try to make counterfactual inferences about how changes in the design of the lottery game that generates our data might affect the level of sales. While the dataset we use does not contain any changes in the design parameters of the game we exploit the fact that there is a deterministic relationship between the mean, variance and skewness of the prize distribution (the prize distribution’s ﬁrst three moments) and both the design parameters and the level of the rollover from the previous draw. Since we know how the moments depend on both of these factors, and since we can estimate how sales depend on the moments, we can combine these to infer how sales depend on the design parameters even though there has been no variation in those parameters across the sample. This paper updates and extends earlier research in Walker and Young (2001). The major empirical difﬁculty with that earlier work is that it failed to resolve satisfactorily the fact that, although sales will depend on the moments of the prize distribution, the moments of the prize distribution themselves depend on sales. Thus the relationship between sales and the moments of the prize distribution is a simultaneous one – so the major empirical difﬁculty we face is to estimate the extent to which variations in the moments cause variations in sales. Moreover, the moments are not known until the draw is closed and sales can be tallied. Thus, it seems plausible that individuals form expectations of the moments based on sales in the current draw and that these expectations are driven, in part, by any observed rollover from the previous draw. A large rollover will lead potential players to expect larger sales in this draw so that the expectation of the expected value will be larger, the expected variance will be larger, and the expected level of skewness will be less left skewed, than would be the case with no rollover. The usual way of resolving simultaneity in the relationship between two variables is to exploit factors that affect one relationship but not the other. A possible candidate factor is the size of the rollover – this affects sales in the next draw via the effect on the moments of the prize distribution.

78 Gaming in the New Market Environment

However, sales in one draw may be correlated with sales in the next, even in the absence of rollovers. For example, this correlation might arise because a temporary factor that boosts sales in one draw may bring a few new individuals into the game who then gain (or regain) the habit of playing and so subsequent sales are affected. This serial correlation in sales implies that we cannot use sales in a previous draw. Unfortunately this rules out exploiting the rollover size because this is a given fraction of the sales in the previous draw. Instead we exploit a peculiar feature of the game whereby the proportion of winners in each prize draw is a random variable which depends on the winning numbers that are actually drawn. Some numbers are more popular than others (see Farrell et al (2000) for estimates of the extent of this phenomenon – which is sometime referred to as conscious selection). In draws where unpopular numbers are drawn there are few winners of prizes below the jackpot level (as well as few winners of the jackpot). One purpose of these smaller prizes in lotto design is that many small winners might encourage players to think that winning big is more likely than it really is. Moreover, many players participate through consortia – groups of workplace colleagues or family members who agree (explicitly or not) to share their winnings. When such consortia win small prizes they are thought to often reinvest their winnings, rather than take the trouble of dividing the winnings into very small amounts per consortium member. Thus, for a given level of sales in a particular draw, the number of small prize winners (which is a random variable, conditional on overall sales) does affect the level of sales in the subsequent draw. This will be true irrespective of any serial correlation is sales. Our modelling strategy is to use this property to isolate the effect of moments on sales. Once we allow for this endogeneity of the moments of the prize distribution, we ﬁnd quite conclusive evidence that skewness is an important factor in driving lottery sales. A complementary modelling strategy has been pursued by other researchers who attempt to analyse the proximate determinants of sales (see, for example, Forrest (2008) and references therein). The rationale for this line of research is to provide estimates of the effect of actual changes in the marketplace. Thus this approach has, for example, been used to analyse the effects of de-regulation of other forms of gambling on lottery sales; and has been used to estimate the effect on the sales of one product of the introduction of another to address the extent to which one game cannibalises sales from another. This complementary literature has some advantages: not least, it is much more straight-

Lottery Design Lessons from the Dismal Science 79

forward. In principle, such a more straightforward regression exercise could estimate the effect of actual game design changes on sales. In practice, such game design changes are few and far between and, moreover, are not likely to be exogenous. Observed changes in game design are likely to be a response by the operator to an opportunity to raise sales – for example, if sales are ﬂagging, for some reason, the operator may feel that it is appropriate to change the design of the game to attempt to arrest the decline in sales. Thus, observed game design changes (and even the introduction of new games) are unlikely to be exogenous and estimation methods ought to address this problem. In practice, such studies do not consider this causality issue. Moreover, both strands of this literature make arbitrary functional form assumptions. In one case, the focus is on the role of the prize distribution and the assumption is that this can be captured through a linear function of the moments of the prize distribution. In the other case, the speciﬁcation is more pragmatic and sales are assumed to depend on the jackpot size, or the expected jackpot winnings that depend, in turn, on the size of the rollover. In practice, it is difﬁcult to argue that time series variation in sales is really due to the variation in the skewness in the prize distribution rather than arising from some complicated non-linear function of the rollover size. It is especially difﬁcult to argue for a speciﬁc model if there is limited variation in the size of rollovers across draws. In practice, there is simply insufﬁcient variation in the jackpot size across draws to be able to argue for one speciﬁcation or another.3 One advantage of the more structural approach here, that emphasises the effect of the moments of the prize distribution on sales, is that it is explicitly couched within a theoretical framework that has been widely used, in other markets, to analyse brand choices. It is common for lottery markets to feature a portfolio of games, each with particular characteristics so that each game can be thought of as a brand. The theory is due to Lancaster (1971) and has famously been applied to UK egg sales in pioneering work by Gorman (1980) and, more recently, to understanding the market for cellphones in the US by Hausman (1997) and minivans by Petrin (2002). This theory helps us understand the relationship between sales of different brands in the same market and, in particular, makes it possible for us to make inferences about how to design games. The rest of the paper is organised as follows. Section 4.2 explains lotto games and how to calculate their odds which is fundamental to understanding lotto design. Section 4.3 builds on this to work out the

80 Gaming in the New Market Environment

expected value of a lottery ticket, and Section 4.4 extends this to include higher moments of the prize distribution. Section 4.5 introduces the Lancastrian approach in the context of lottery games and explains how one might recover the effects of game design effects from estimates of a model of sales which is driven by variations in the moments of the prize distribution induced by rollovers. Section 4.6 explains the econometric methodology that provides unbiased estimates of the effects of exogenous variation in the moments. Section 4.7 presents some tentative simulations of how sales would be affected by variation in the major parameters in the design of the game. Section 4.8 concludes with some directions for future research.

4.2

The odds of winning a (pari-mutuel) lottery

Lotteries have traditionally featured selection mechanisms that have numbered tickets drawn at random from an urn. While the details differ from game to game, modern incarnations of these mechanism usually feature N numbered balls bouncing in a transparent, often rotating, container from which n are drawn (usually without replacement). Individuals buy tickets by marking n numbers on a printed matrix of N numbers on an entry form. The form is scanned electronically and a ticket is printed out and given to the customer as a record. Each combination of n numbers has the same chance of winning given by p = N!/n! (N – n)! For example in the commonly used n = 6, N = 49 design p = 1/13,983,816. Thus, if sales are 40 million then we would expect, on average, the number of winners to be approximately 3. The probability that a given ticket does not win is 1 – p. If two tickets are sold, then the probability that neither are winning tickets is (1 – p)2, and so on for 3 tickets, 4 etc. Thus the probability that there are no winners (that is, that there is a rollover) in draw t when sales are St is given by (1 – p)St. Thus, suppose sales are around 40 (20) million and p is 1 in 14 million then the probability of a rollover is approximately 6 per cent (25 per cent).4 Rollovers are doubly important if there is serial correlation in play. Rollovers cause sales to rise in the next draw and the serial correlation then causes sales to rise in the draw after that (and after that …), albeit by to a lesser extent, in all subsequent draws. Sequential rollovers could, in principle, cause sales to ratchet upwards. However, in practice the tedium of lotto games causes sales to exhibit a long run secular decline. Thus rollovers, of the appropriate frequency and size, can offset the tendency for sales to decline. However, more rollovers are not necess-

Lottery Design Lessons from the Dismal Science 81

arily better. If rollovers have high frequency then players will come to expect them and will engage in intertemporal substitution – reserving their lottery spend until a rollover (or several) has already occurred before playing. The optimal frequency of rollovers will depend on the relative strengths of the trend decline, the intertemporal elasticity of substitution, and the degree of serial correlation.

4.3

The expected value calculation

In rafﬂes the expected value of a ticket falls with sales since the prizes are ﬁxed. At low levels of sales tickets are valuable, and as sales rise their value falls. The operator faces the risk that sales revenue will not exceed the cost of prizes and the players face the risk that they have to compete with many other players for the prizes. In lotto games the design is pari-mutuel and players compete for shares of the prize pool(s) as opposed to ﬁxed prizes. Thus, as sales rise the prize pool expands and the expected winnings (that is, losses) remain the same. However, the possibility of a rollover implies that the (top) prize pool might not be won at all in this draw and this depresses the expected value of participating in the current draw. Thus, sales have an additional effect on the expected value of participating in any given draw – higher sales cause the rollover probability to fall which raises the expected value. Thus the expected value of a ticket in a given draw depends on the size of the prize fund: which is the proportion of sales that is not taken out as tax and operator costs (that is, (1 – τ)St where τ is the take-out rate) plus any rollover from the previous draw. But this is multiplied by the probability that there is at least one winner, that is, 1 – (1 – p)St. Further details of the algebra of the calculation of the expected value are in the Appendix but Figures 4.1 and 4.2 capture the important intuition. Figure 4.1 shows how the expected value or mean return, m1t, of a lottery ticket that costs $1 for common types of design in a regular (non-rollover) draw. In the ﬁgure the take-out rate τ is set at 0.55 which is a typical value. The shape of this ﬁgure has given rise to what has been called lotto’s ‘peculiar economies of scale’5 since it shows that the game gets cheaper to play (in the sense that the expected value gets higher) the higher is sales. This is because the higher is sales the smaller is the chance of a rollover occurring because more of the possible combinations are sold. This makes the return higher in the current draw because rollovers take money from the current draw and add it to the next draw; and your ticket in this draw gives you a possible claim on prizes in this draw but not the next. So the higher the chances that

82 Gaming in the New Market Environment 0.46 0.44

Mean return

0.42 0.4

6/49

0.38

6/42 6/45

0.36

6/44 0.34

6/53

0.32 0.3 20 Figure 4.1

30

40

50

60 70 Sales (£m)

80

90

100

Lotto’s peculiar economies of scale S=S(V)

rollover

1– τ

Expected value

no rollover

Sales Figure 4.2

Expected value in rollover and regular draws

a jackpot rolls-over, the less a ticket for the current draw is worth. Indeed, from an individual player’s perspective, it doesn’t matter if the money is rolled over or given to another player in this draw, only that he or she does not win.

Lottery Design Lessons from the Dismal Science 83

Note that at very large levels of sales all game designs have the same mean return which simply equals the 1 – τ, because the chance of a rollover is very small when ticket sales are very large since most possible combinations will be sold. Notice also that, at any given level of sales, easier games offer better value in regular draws since the rollover chance is smaller. When a rollover of size Jt–1 has taken place, the expected value function shifts upwards by an amount equal to Jt–1/St (because everyone has the same chances of getting the rolled over amount) and this diminishes as St rises. Thus, the function shifts upwards more at low sales than at high sales and the shifted function will, in general, have a single maximum. Figure 4.2 shows the shape of the m1t function for regular and rollover cases. Note that, the effects of rollover on expected value, and thence on sales, will be small when sales are high, and when Jt–1 is small because πn, the share of the total price pool for those who match all n balls (i.e. jackpot winners), is small or because τ is high. A double rollover, when the jackpot had not been won for two earlier draws, would shift the m1t even higher upwards. In principle, with multiple rollovers the expected value could exceed the cost of the ticket – although we would expect sales to rise with the expected value and consequently we would be unlikely to observe tickets being worth more then their cost.

4.4

Higher moments of the prize distribution

It seems likely that sales are affected by higher moments of the prize distribution – in particular, by the variance and skewness of the prize distribution. Indeed, if only the ﬁrst moment mattered then we would ﬁnd it difﬁcult to reconcile gambling with aversion to risk. Individuals who are risk averse would not be expected to participate in lotto games that offered expected values that were less than the cost of participation. There is accumulating evidence that gambling is affected by the skewness in the prize distribution.6 That is, lotteries offer the possibility of a huge change in lifestyle. Lottery prize distributions are highly left skewed – the vast majority of players lose a small amount (the stake), a small number win low prizes, and a very small number win large prizes. However, the existing studies neglect the fact that the observed moments in prize distributions are themselves a function of the bets that are placed. For example, as sales rise in lotto the expected value rises and this makes participating more attractive. Moreover, lottery operators have chosen the design of the game to deliver a particular

84 Gaming in the New Market Environment

vector of moments, taking into account the likely size of the market. If the market is expected to be small then we might expect the operator to choose a different design (one that is easier to win) than if the market were large. Similarly, as well as the expected value, the higher moments of the prize distribution will be affected by sales. Thus, the mechanical relationship between sales and the moments of the prize distribution may contaminate the behavioural response of sales to variation in the moments that are due to chance. The distribution of winnings implied by the n/N design typically has a large spike at zero, since most players lose altogether, and successive peaks corresponding to matching more of the n winning numbers until a ﬁnal peak occurs for matching the n-ball and winning (a share in) the jackpot prize pool. The distribution at these further local maxima associated with more difﬁcult-to-win prizes arises because the amount won depends on the number of people who also win a share in each prize pool. Thus, instead of a spike, there is a small peak with a (local) maximum in the distribution for each prize type, which corresponds to the most probable number of winners for that type, but around this is a distribution that arises because there may be fewer winners each getting a larger share of the pool or more winners than expected each getting a smaller share. Successive peaks, corresponding to the mean winnings of bigger prizes are lower (as the chance of winning is smaller) and narrower (because the variance in the number of prize winners is lower for the more difﬁcult to win prizes). The overall distribution is thus left skewed (by the large majority of losers) but a rollover decreases the left skew since it increases the size of the jackpot pool. The formulae for the variance and skewness are complicated functions of the rollover size and the lotto design parameters and can be seen in the Appendix to Walker and Young (2001).7

4.5

Lancastrian approach

The characteristics approach to demand due to Lancaster (1971) is useful for thinking about sales of closely related products. The ﬁrst application was a famous paper by Gorman (1991) that modelled egg sales in the UK. More recently, the approach has been used for thinking about brands of goods (see Hausman (1997) and Putrin (2002)). The essence of the approach is that, in contrast to conventional microeconomic theory, goods are not valued for their goodness but rather for the bundle of characteristics that they contain. So goods that are good substitutes for each other are likely to contain similar combinations of character-

Lottery Design Lessons from the Dismal Science 85

istics. In the context of the lottery market the products sold effectively differ in their prize distributions and a convenient way of summarising these distributions is via their ﬁrst three moments. The ﬁrst moment, the expected value m1t, is effectively the price variable since the expected cost of participating is the face price (usually one unit of currency) minus the expected value. On average each lottery ticket costs 1 – mlt and the larger is the expected value of a ticket the better the bet is. Apart from this mean return to a ticket, a ticket is characterised by design parameters that imply speciﬁc variance (m2t) which is a characteristic that is disliked, and speciﬁc skewness (m3t) which is liked. Suppose a market is characterised by two lottery games, labelled W and S, both with the same take-out rate for convenience. The position is illustrated in Figure 4.3 where each product is portrayed as a ray in characteristics space. The length of the rays are given by multiples of 1 – mlt to indicate their cost. Both games have speciﬁc combinations of –m2t (variance is disliked so it is measured negatively in the ﬁgure) and m3t (skewness) with the W game being less skewed, at given variance, than the S game. Each ray indicates a ticket type and movements

W’ –m2t

W E’

E

S

m3t O Figure 4.3

Two-game lottery market

86 Gaming in the New Market Environment

along the ray away from the origin indicate higher levels of sales of that game. We could envisage the population having a distribution of preferences between –m2t and m3t with many preferring the heavily skewed S game but, nonetheless, some preferring the W game. Aggregate sales of the two products can be described by a position along the line WS and imagine that the position E is chosen where the market has a slightly higher proportion of S-type tickets than W-type tickets (as seen by their distances along the respective rays from the origin). Now consider what happens when the W game experiences a design change that results in a price fall so the line OW gets longer since more W-type tickets can be afforded. And suppose that the skewness in W’s prize distribution, at any level of variance, gets larger so OW gets less steep. Thus, W’ has become a more attractive bet and the frontier for the aggregate market changes to W’S. Now that W has become both cheaper and more similar in the combination of characteristics that it offers to S we would expect E to move to a position like E’ where many more W-type tickets are bought and somewhat fewer S-type. Note that if only S existed, sales would be OS. If W is now introduced then sales of S slump, but this is more than compensated for by the extra W-sales. Thus, diversiﬁcation of the game portfolio can increase aggregate sales because two products better cater for the distribution of preferences for the characteristics in the population.8

4.6

Econometric methodology, data, and estimates

One way of summarising the complications of how all the various aspects of game design impacts on sales is through the mean, variance and skewness of the prize distribution.9 For example, one might allow St = S(m1t, m2t, m3t, Xt) + εt by choosing some parametric form for S(.).10 The term εt captures the effects of all unobservable factors (and may be correlated with the same factors in previous draws – serial correlation) on sales; while Xt is a vector of observable factors that, independently of the parameters of the game, affect sales. Most other research has focussed on the role of such Xt’s in determining sales.11 One modelling approach would be to substitute the formulae that show how the moments are related to the game design variables into S(.) to obtain a reduced form model of sales whereby St = F(n, N, τ, π, R t Jt–1) + ut. This F(.) is a complicated highly non-linear function and it is not practical to estimate the structural model coefﬁcients from this reduced form. A special case of F(.) arises when all of the game design para-

Lottery Design Lessons from the Dismal Science 87

meters are ﬁxed in the time series being modelled, so that only the rollovers determine sales variation across time, although this will still be highly non-linear.12 A fundamental issue is that while Rt–1Jt–1 affects St in F(.) it also depends on St–1, because lagged sales affect the rollover probability and also determine the size of the previous jackpot. Thus, the presence of serial correlation is sufﬁcient to undermine the assumption that rollovers occur randomly since it implies that St depends on St–1 directly, as well as indirectly via any rollover. However, conscious selection is a strong phenomenon in lotto markets. Operators encourage it since having ownership over the numbers that one bets on makes playing more persistent – individuals are more likely to play in every draw if they feel that they are playing their own numbers. It is thought that, by allowing players to choose their own numbers, this has a large impact on sales. If players consciously select their numbers when they bet then there is likely to be larger variation in the number of prize winners in each prize pool than if there were no conscious selection. In particular, the number of 3-ball matches is likely to vary considerably from draw to draw. With sales of 40 million in 6/49 we would expect the number of 3-ball winners to be approximately 700 thousands with a standard deviation of approximately 100 thousand.13 3-ball winners usually win a modest prize14 and it is common for players to re-invest these modest levels of winnings in more lottery tickets in the next draw.15 Moreover, a substantial minority of sales is accounted for by syndicates16 who are thought to reinvest such small dividends to save the trouble of distributing them to syndicate members. Thus, in practice the number of 3-ball winners affects subsequent sales.17 Thus, we use an approach that uses the proportion of 3-ball winners in the t – 2 draw to explain the variation in R t–1 Jt–1 and then use the predicted value of this instead of the actual value. The assumption here is that 3-ball winners in draw t – 2 affect sales in t – 1, and hence the probability and size of any rollover, but it is assumed not to directly affect sales in t. This seems like a reasonable assumption. The data is the sales and rollover information, winning number and number of winners of each prize pool that Camelot, the UK operator, is obliged to publish for every draw. The data is sales in the Saturday and Wednesday online lotto games in the UK from the ﬁrst draw, though the introduction of the Wednesday game was introduced in early 1997, to draw 310 in mid-1999 just before the start of the Thunderball game was introduced.18 This spans the period when the portfolio of games consisted of the Saturday lotto game, then the Wednesday and Saturday

88 Gaming in the New Market Environment

6/49 lotto games, then scratchcards, which were introduced in early 1998,19 as well as the two lotto draws. Figure 4.4 shows the history of sales over this period and for some time subsequent to this. The estimation period covered a time when there was rapid growth in sales fuelled, in part, by several large rollovers which are indicated by the sharp spikes in sales; followed by the introduction of the Wednesday lotto draw; and then the period, after the introduction of scratchcards, which saw a rapid decline in scratchcard sales, a slow decline in Saturday sales, and an even slower decline in Wednesday sales.20 We assume that the parametric form for Saturday and Wednesday sales, Ss(.) and Sw(.), are log-linear. That is, the log sales in each draw is assumed to be a linear function of the log of the moments.21 We include only the moments relating to the current draw in each equation.22 Our vector X includes data on: daytime maximum temperature and rainfall (the averages over the week of sale) prior to the draw, averaged over all met stations; seasonal (month) control variables; a cubic trend; and the sales of scratchcards in the week of sale). Further control variables were included in experiments and checks for robustness but even when they were statistically signiﬁcant they failed to change the coefﬁcients on the moments. Indeed, the estimated coefﬁcients hardly change at all even when all of the control variables (except the cubic trend) are excluded.

160

Scratch Weds Extra Sat Extra Sat Regular Weds Regular

140

£ millions/week

120 100 80 60 40 20 0 1

Figure 4.4

27

53

79

105

131

157

UK sales (£m per week)

183

209 235 Draws

261

287

313

339

365

391

Lottery Design Lessons from the Dismal Science 89

An important innovation in this research is that our estimation method treats the dataset as an unbalanced panel (of two games) – it is unbalanced because the Wednesday game was not introduced until 15 months after the establishment of the Saturday draw. The speciﬁcation is a logical one but explicitly excludes any dynamic effects apart from via the residuals. This is quite restrictive in this context but relaxing this gives rise to signiﬁcant estimation difﬁculties and we reserve more complex speciﬁcations for future research.23 The results are presented in Table 4.1. We compare results with OLS (where the m’s are treated as exogenous but we still allow for serial correlation). The standard errors are robust to heteroskedasticity and a variety of tests of speciﬁcation, parameter stability and forecasting power, were employed.24 The OLS results show that: sales are a statistically signiﬁcant increasing function of the mean of the prize distribution – so better bets are more attractive ones; sales are a statistically signiﬁcant but decreasing function of the variance in the prize distribution – so riskier bets are less attractive; and sales are a statistically signiﬁcant and increasing function of the skewness of the prize distribution – so players appear to exhibit a preference for skewness.25 The GMM estimates are also similar although the Wednesday results, in this case, show somewhat higher sensitivity to all three moments which is consistent with endogeneity being a bigger problem in the Wednesday game because at lower level

Table 4.1

Estimated parameters of log sales equations

OLS using actual jackpot to determine m’s Wednesday

Saturday

GMM using t-2 3-ball winners Wednesday

Saturday

m1

0.344 (0.090)

0.211 (0.076)

0.451 (0.110)

0.471 (0.193)

m2

–0.042 (0.009)

–0.072 (0.007)

–0.050 (0.018)

–0.102 (0.040)

m3

0.062 (0.024)

0.111 (0.023)

0.194 (0.083)

0.162 (0.043)

ρ

w

0.057 (0.019)

0.003 (0.022)

0.051 (0.020)

0.011 (0.023)

ρ

s

0.123 (0.020)

–0.001 (0.010)

R2

–0.001 (0.009) 0.932

0.800

0.931

0.297 (0.088) 0.813

Note: All equations contain control variables listed in the text. First stage results for the estimates show that the moments are signiﬁcantly affected by the number of 3-ball winners and are available on request.

90 Gaming in the New Market Environment

of sales, in 6/49, the moments are more sensitive to the level of sales and to the jackpot size than is the case at a higher level of sales. Although the qualitative results support our original vision – that sales depend positively on the mean of the prize distribution, negatively on its variance, and positively on its skewness, the quantitative interpretation of the results are less straightforward. To give a feel for what the results mean in quantitative terms imagine a (large) rollover which adds £5 million to the Saturday jackpot, where Saturday sales are approximately 40 million. This bonus would increase the expected value by approximately 25 per cent, given the UK lotto design parameters, and would cause a modest rise in variance, and a large increase in skewness (of about 10 per cent). Thus, according to the GMM estimates in the ﬁnal column of Table 4.1, this would raise sales by about 12 per cent though the effect on the expected value and a further 2 per cent because of its effect on skewness, offset by an effect of more variance of less than 1 per cent. There would be a jump in sales of about 13 per cent. But note that the serial correlation would raise sales the following Saturday by about 4 per cent over what would have happened (with a small effect on Wednesday too). Cumulating the effects over time, and across both games, the effect of a £5 million rollover would be to raise sales by approximately £8 million.26 Clearly our estimates rely on the functional form restriction that rollovers affects sales only through their effect on the ﬁrst three moments for the prize distribution and that this relationship is log-linear. However, it is difﬁcult to test this speciﬁcation against alternatives because rollovers have rather limited variance so that tests of functional form would have little power. Thus, in the absence of other evidence, and with support from studies of gambling on other datasets, we feel that our interpretation of the estimates of the way in which sales vary with rollover size is the best available, and we exploit the estimates to simulate the effect of game design induced changes in moments below.

4.7

Game design simulations

Making inferences from the observed estimated relationships between sales and the moments requires that we solve the estimated equations for sales as a function of the design parameters. This is complicated because these equations are highly non-linear and do not admit an analytical solution.27 Throughout we use the estimates from the GMM estimation procedure – note that these estimates tended to imply that sales were more sensitive to the moments. The simulation strategy was

Lottery Design Lessons from the Dismal Science 91

to draw an initial value for εt for each game, solve the model for initial sales for given values of the design parameters, then use this solution to generate a rollover probability and a level of the rollover conditional on one occurring in the next draw. Whether or not a rollover occurred was chosen at random from distribution with mean given by the predicted rollover probability. This then generated a subsequent sales prediction and the rollover probability and conditional sales were solved, etc. This exercise was repeated for 250 draws and the average sales computed. This process was looped a further 100 times, drawing a new initial εt on each occasion. The ﬁgures report the average sales from these 100 simulated 250-draw histories. Our simulations are entirely illustrative. To simplify the calculations the π vector is ﬁxed so that πn = 1 and all π’s for lower prize pools were set to zero. The lack of lower level prizes implies that we are analysing hypothetical games with a very high variance and high skewness. We concentrate here on the effects of τ and of n on revenue (N is ﬁxed at 49). Figures 4.5 and 4.6 show the effects of varying τ and of n where we have added together the sales ﬁgures from Wednesday and Saturday games to compute the predicted sales and revenue (R) for an average week over the 250 draw period. In Figure 4.7 τ = 0.5 while in Figure 4.8 N=49. Figure 4.5 suggests that a harder to win game, such a 6/53 would raise higher revenue – almost 20 per cent higher than 6/49. It also suggests 70 Sales

60

R 50 40 30 20 10 0 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Figure 4.5

Effects of varying n on sales and revenue

92 Gaming in the New Market Environment 120 Sales 100

R

80

60

40

20

0 1 Figure 4.6

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

Effects of varying τ on sales and revenue

that making the game very hard to win, beyond 53, would cause sales and revenue to fall steeply. Figure 4.5 shows the effects of the take-out rate – in this case revenue is fairly insensitive to the take-out over a wide range although the revenue maximising rate is approximately 0.4.28

4.8

Conclusion

Our analysis has considered some of the important questions relevant to running a lottery. Our methodology for analysing the implications of game design is analytically rigorous and yet it reﬂects the informal received wisdoms that dominate industry debate. Thus, it probably captures many of important features of the realities of the game but provides a degree of abstraction from reality to allow counterfactual changes to be analysed in a formal and quantitative way. However, our results need to be qualiﬁed. The simulations assume that sales respond to variations in mean, variance and skewness from design changes in the same way as they respond to these variables when rollovers occur. However, it is plausible that people may respond differently to these two types of changes. Firstly, changes induced by occasional rollovers allows for the possibility of substitution between draws,

Lottery Design Lessons from the Dismal Science 93

but this possibility does not exist for changes coming through the game design rather than rollovers. This suggests that ticket sales are higher when changes come from rollovers than from game design. Rollovers are rather like sales promotions – they induce people to change their behaviour quite differently to a temporary difference in the offer than they would to a permanent change. This failure to come to grips with intertemporal substitution is an important shortcoming on the present research. It would be difﬁcult to overcome this in the UK data because, so far, there have been no clean and simple game changes that would allow us to challenge the implicit assumption that sales respond to temporary changes in moments in the same way as to permanent ones. However, some preliminary work using Irish time series data, which contains several simple game design changes, suggests that, despite the large incentives to engage in this, intertemporal substitution does not seem to be statistically important.29 There are several avenues for further research. The econometric analysis needs to be extended to incorporate newer games if we are going to be able to exploit a longer run of data. There are some practical difﬁculties in doing this because their introduction is potentially endogenous and because the newer games have typically involved rather low levels of sales. However, Forrest (2008) provides results that suggest that this extension is potentially important. Smith (2008) highlights the role that the hypothecation of the lottery revenue may play in determining sales. This is an interesting suggestion and could be incorporated into a time series analysis such as the present one providing there is sufﬁcient variation in how the hypothecation is done. In the UK there have only been limited changes. US data offers better prospects for this kind of exploration because there have been changes in the hypothecation across states and across time. This essentially time series work could usefully be complemented by a microeconometric analysis of the effects of income using cross-section data in order to compute the regressivity of the take-out. Smith (2008) clearly shows that the share of income spent on lotteries declines steeply across the income distribution. Moreover, she shows that very high income shares seem to be conﬁned to quite low income households. Such a microeconometric analysis could also incorporate the effects of rollovers if panel data were available. Ultimately such an analysis could enable researchers to simulate the effects of game design on the size distribution of individual levels of play, as well as on aggregate sales, and hence offer the prospect of designing out the problem gambling alluded to in Grifﬁths (2008), and Paldam (2008).

94 Gaming in the New Market Environment

Notes 1 This paper builds on earlier research that was funded by ESRC under research grant R000236821. This work was supported, in part, by a research studentship jointly funded by ESRC and the UK National Lottery Commissioners. However, the views expressed here are those of the author alone. 2 In horserace betting the idea that punters favour longshots unduly is a related phenomenon. 3 Kearney (2002) is more convincing in this respect because her analysis is based on US data where the game design and population varies quite markedly across states as well as across time. 4 The ﬁgures correspond approximately to UK ﬁgures for the Saturday and Wednesday lotto games. In practice, we observe many more rollovers in lottery games than would be implied by the observed level of sales and the discrepancy is due to the extent to which their are similarities in the way players choose their numbers – a phenomenon known as conscious selection. This can be accommodated into the analysis by allowing the rollover probability to be (1 – p)αSt with α ⭐ 1 such that α = 1 implies no conscious selection (i.e., completely random choice) and α = 0 implies that all payers have chosen exactly the same combination of numbers. Casual inspection of UK data suggest that α ⬇ 05. 5 See Cook and Clotfelter (1993). 6 See Golec and Tamarkin (1998) for US racetrack betting, and Garrett and Sobel (1999), and Kearney (2002) for US lotteries. 7 That paper makes clear that there is a strong correlation between each of the moments of the prize distribution. In particular, locally at least, they each respond in similar ways to a variation in sales. What is required to break this multicollinearity is data that contains sufﬁcient variation in sales. 8 Note that m 2t and m3t are not, in general, independent of each other – it may be difﬁcult to have more of one without having more of the other also. While, it will not be possible to have any combination of the two one may be able to expand the set of possible combinations through more complex design – such as introducing a bonus ball. 9 Note that each moment depends in a known and deterministic way on n, N, τ, π (the vector of π’s), and on St and Jt–1. That is, mkt = mkt(n, N, τ, π, St, Rt–1Jt–1), for k = 1,2,3 and so it is clear that, while sales depend on the moments, the values of the moments also depend on sales. Thus, when estimating the causal effects of the moments on sales account needs to be taken of their endogeneity. Forrest (2008) also presents estimates of such an S(.) function for the UK Saturday lotto game. That paper uses simple regression methods that ignore the endogeneity of the moments. 10 For a simple example, Farrell, Morgenroth and Walker (1999) assumed that S(.) was a linear function of m1t and did not depend on higher moments at all. 11 Notice also that the speciﬁcation treats rollover-driven temporary variation in the moments as equivalent to design-induced permanent variation in moments. Forrest is, rightly, critical of this and the same point is made in Walker and Young (2000). Nonetheless, it is useful, for the moment, to

Lottery Design Lessons from the Dismal Science 95

12

13

14 15

16

17

18 19 20

21

22

think of S(.) as capturing the structural determinants of sales – that is the variation in sales that is due to variation in the moments that arise for any reason. Moreover, F(.) is only the reduced form of S(.) if εt exhibits no serial correlation. If it does, then the rollover size would be determined by previous sales and on the level of sales two draws previously. The reduced form would then have to substitute RtJt–1, Rt–1Jt–2, etc. out of the model recursively. Moreover, a further problem arises because total sales, and hence the moments, are not known when individuals are making purchase decisions so players have to estimate the moments from the information that is know prior to the draw – and this information set will include the value of Rt–1Jt–1. In the two games analysed here one has sales of the order of 40m and yet the standard deviation of the number of 3-ball winners in the data used here is more than 200 thousand. In the UK game the 3-ball prize is not pari-mutuel. Rather it is a ﬁxed prize of £10. In most games, retailers are empowered to pay out the small prizes. This, in itself, will encourage players to spend their winnings on more lottery tickets, for the next draw, when they collect their winnings. Indeed Guryan and Kearney (2005) identify a lucky store phenomenon that arises because of habit persistence in players’ expenditures. These are consortia of players who agree to contribute to the stakes each week and share the winnings. They are commonly organised at the workplace. A simple correlation in our data for our two draws suggest that an additional 100 thousand 3-ball winners in the large game draw at t will increase sales in t+1 by a statistically signiﬁcant 178 thousand in the large game, while the same for the small draw raises subsequent sales by 128 thousand. We reserve the data beyond draw 310 as a hold-out sample that allows us to conduct some forecasting tests. The results were very similar when the dataset was extended to later periods when further portfolio diversiﬁcation took place. One of the peculiar features of the two games in the UK is that they were linked via their rollovers. That is, if there are no winners then the jackpot from Wednesday (Saturday) is added to the jackpot on the following Saturday (Wednesday). When computing the moments we allowed for conscious selection by calibrating α for each game so that the level of sales were consistent with the correct number of rollovers across the estimation period. We use a log speciﬁcation to avoid negative predictions in simulation. We allowed for serial correlation and estimation was conducted using seemingly unrelated regression methods to allow for the correlation in the residuals across draws. That is we allow the residual from the sales on a Wednesday (Saturday) to affect both the coming Saturday’s sales as well as the following Wednesday’s. That is, we estimate Swt = βwXwt + γw1m1t + γw2m2t + γw3m3t + εwt where εwt = ρwwεwt–1 + ρwsεwt–1 Swt = βsXst + γs1m1t + γs2m2t + γs3m3t + εst where εst = ρswεwt–1 + ρssεst–1

96 Gaming in the New Market Environment 23 Estimation was conducted using GMM provided by the xtabond facility in STATA9. 24 A number of speciﬁcation tests fail in the OLS and OLS using operator jackpot forecast suggesting that the endogeneity of the moments is an important factor. Tests of their endogeneity strongly suggest that this is a problem – especially in the Wednesday game. This is to be expected since when sales are small the moments are considerably more sensitive to sales variation and jackpot size than when sales are low. In the GMM model we ﬁnd that the AR test is passed, indicating absence of higher order autocorrelation than the ﬁrst order allowed here although this is sensitive to the inclusion of the cross-correlation between the two games. The ARCH test for heteroscedasticity also passes which is surprising given how the variance of sales seems larger in rollover weeks compared to regular weeks. The normality test passes although this failed when we tried to include earlier data in the analysis-possibly because of the large outliers associated with early double rollovers. The parameter constancy tests for both models (which were carried out on the remaining observations up to draw 395) now failed. The RESET test, however fails (albeit marginally) for Wednesday suggesting that there is some further speciﬁcation problem. It seems likely that this is associated with the use of a ﬁxed degree of conscious selection across draws. This implies that rollover frequency, conditional on sales, should be constant across draws. In fact there is some suggestion that rollover frequency is falling. Moreover, it seems likely that the degree of conscious selection is lower in rollover draws because players often wish to purchase more then their usual number of tickets and may be more likely to use the random number generator built-in to the retailers’ tills for this purpose. 25 Similar results were obtained in Forrest (2008) for the Saturday draw data. 26 The operator might be tempted to lace the jackpot in order to improve sales. However, even with the large estimates sensitivity to the jackpot implied by these results this is unlikely to be proﬁtable for it. This is for two reasons. First, the operator retains only 55 per cent of the additional sales (i.e., a little over £4 million). Secondly, the assumption has been that this £5 million arises from a randomly occurring rollover – if the operator were to try to make such bonuses occur at a regular basis players would become to expect them and the model would cease to be a valid description of sales behaviour. In fact, in many cases operators do have some discretion over adding bonuses to the jackpot pool (in the UK this is referred to as a Superdraw). In practice they are used sparingly, to offset the effects of a temporary drought of rollovers. 27 The ‘ﬁndroot’ command in Mathematica, which solves non-polynomial expressions using the Jenkins-Traub algorithm was used and applied to simpliﬁed expressions for the moments. In practice this proved very time consuming which limited the range of simulations that could be conducted to those with πn = 1. 28 Unfortunately, it has not been possible to compute standard errors around these forecasts so it is difﬁcult to say how precise these forecasts are likely to be.

Lottery Design Lessons from the Dismal Science 97 29 This, albeit provisional, ﬁnding is consistent with there being a strong presentbias in individual preferences. That is, individual decisions may be time inconsistent because they discount the immediate future more than the more distant future. Farrell (2008) explores addiction in the context of the welfare economics of problem gambling. Making a formal connection between the behavioural economics of gambling and welfare would be an important contribution that empirical research such as this could play into.

Appendix The Expected Value Formula

The expected value of a ticket is the ﬁrst moment of the prize distribution – which is the product of the probability of it being won and the size of the prize if it is won. That is, it is given by m1t = (1 – (1 – p)St)[(1 – τ)St + Rt–1Jt–1] where Jt–1 is the jackpot from the previous draw (which equals (1–α)St–1 in the case where the only prize is the jackpot) and Rt–1 is an indicator for whether the previous draw was a rollover or not and is assumed to be known at the time that draw t tickets are purchased.30 The ﬁrst term in parentheses is the probability of no rollover (i.e., the probability that the jackpot is won) while the term in square brackets is the expected size of prize. The level of m1t is unaffected by the nature of the prize distribution31 – in the absence of a rollover it depends only on the takeout rate and sales revenue irrespective of how that fund of revenue is distributed across prize pools. In practice, it is only ever the jackpot prize pool that has no winner and is rolled over – the easier to win prize pools are usually sufﬁciently easy to win that there are always many winners. Thus, when a rollover from the previous draw occurs it is only the jackpot n-ball prize pool that is rolled over. Thus, in practice, Jt–1 = πn(1–α)St–1 where πn is the proportion of the overall prize pool going to the n-ball matches (that is, shared by those who have chosen all n of the winning numbers). Notes 30 In most lotto games it is possible to commit to buying a ticket before it is known whether the previous draw is a rollover or not. For example, one can place an order to buy a ticket for every future draw. In practice, only a small proportion of tickets are bought in this way. 31 In practice, lotto games are designed so that the prize fund is spread, in predetermined shares, between those that match all n balls (the jackpot fund), those that match n–1 balls, n–2 balls, etc. Let these shares be deﬁned by π = (πn, πn–1, πn–2…). which are usually set such that the expected prize in the n-ball pool exceeds that in the n-1 pool, etc. Some designs are more complex and feature additional bonus balls. 98

5 Lotteries as a Source of Revenue Sarah Smith1

5.1

Introduction

Lotteries run by states2 around the world are typically characterised by two features. The ﬁrst is a high tax rate. While a lottery is not itself a tax, as it is sometimes called, the fact is that almost all of the take-out rate (the price of a ticket minus the expected value of the prize) goes as revenue to the government.3 For most lotteries, the tax rate is above that for other forms of gambling and higher, even, than ‘sin’ taxes on alcohol and tobacco. The second feature is that most states formally hypothecate lottery revenue to one or a number of ‘good causes’. These range from very narrow purposes, such as a speciﬁc sports stadium, to wider designated good causes, commonly including sports, arts, heritage and charities. In a more competitive gaming environment this source of revenue is potentially under threat if lottery sales fall or if governments consider reducing tax rates to try to boost sales. The aim of this chapter is to examine the taxation of lotteries, judged by a range of criteria including cost and convenience, efﬁciency and equity – and to compare them to other taxes that may have to be raised to meet the revenue shortfall. It asks whether the typical tax rates on lotteries meet the stated goal of revenue maximisation, and what increased competition is likely to mean for the revenue-maximising tax rate. Finally it looks at the issue of hypothecation, and asks what reduced earmarked revenues might mean for the good causes and for the lotteries themselves. Most of the empirical work that is referred to here concerns lotteries run by US states. In addition, the chapter uses the UK National Lottery as a case study – it shares many features in common with other state lotteries and is the largest lottery in the world in terms of sales. 99

100 Gaming in the New Market Environment

5.2

Lotteries as a source of revenue

Many state lotteries have been set up with the explicit purpose of raising money. In the UK, for example, the duty of the National Lottery Commission, the regulator of the National Lottery is ‘to do [its] best to secure that the proceeds of the National Lottery are as great as possible’, subject to duties to ‘ensure that the National Lottery… is run with all due propriety’ and ‘that the interests of every participant in the Lottery are protected’.4 Many US states have similar objectives with Florida’s being typical of many: ‘[The mission of the Florida Lottery is to] maximize revenues in a manner consonant with the dignity of the state and the welfare of its citizens’ (see Garrett, 2001a, b). The design of many lotteries reﬂects their aim of revenue-maximisation. Most have been set up as monopolies run, or authorised, by the state. Competition from outside the state is also often limited, with (for EU lotteries) restrictions on sales – and advertising – by foreign competitors, although in the future, the growth of online lottery-playing may begin to affect states’ ability to control revenues, an issue we shall return to later. Tax rates on lotteries are typically high. There are at least two ways of measuring the tax rate on a lottery. The ﬁrst, summarised for EU Table 5.1

Gross tax rates, EU lotteries

Country

Gross tax rate

Country

Gross tax rate

Austria

32%

Italy

55%

Belgium

28%

Latvia

20%

Cyprus

37%

Lithuania

13%

Czech Republic

19%

Luxembourg

19%

Denmark

47%

Netherlands

13%

Estonia

13%

Poland

53%

Finland

47%

Portugal

33%

France

28%

Slovakia

22%

Germany

41%

Slovenia

19%

Greece

62%

Spain

20%

Hungary

4%

Sweden

57%

Ireland

41%

United Kingdom

40%

EU25 (excl. MT)

37%

Lotteries as a Source of Revenue 101

countries in Table 5.1, is to measure tax5 as a proportion of gross expenditures or turnover. The (unweighted) average tax rate across the EU countries is 37 per cent, while the comparable ﬁgure for the US is 33 per cent, down from 41 per cent in 1985 (Clotfelter et al, 1999). Table 5.2 compares revenue and the gross tax rate on the UK National Lottery with those on other forms of gambling. The Lottery is by far the biggest single source of revenue, contributing more than all other forms of gambling combined – and more than four times the next biggest single source (general betting duty). This doesn’t reﬂect the fact that more money is staked on the National Lottery – in the latest year (2005–06) £44 billion was staked on general betting compared to £5 billion on the Lottery – but instead the far higher gross tax rate. On other forms of betting, the general trend has been towards lower gross tax rates, motivated in large part by increased competition from offshore betting (in the case of general betting duty) and from the National Lottery in the case of pools betting. It remains to be seen whether increased competition will result in a lower tax rate on lotteries. In comparing taxes on lotteries with those on other goods, it is more usual to deﬁne the tax rate as a percentage of net expenditures since the true ‘price’ of playing is equal to one minus the expected value of the prize. In this case, with a take-out rate of 0.5, the tax rate on the National Lottery is equal to 80 per cent. As shown in Table 5.3, this is higher than ‘sin’ taxes on alcohol, tobacco and petrol in the UK, which are themselves relatively high compared to the EU average. Yet, the actual amount of revenue received from lotteries (and other forms of gambling) is relatively small for many countries. For the UK, which has the largest lottery in the world in terms of sales, National Lottery receipts (including money for the good causes) are well below those from alcohol (beer, wine and spirits combined), tobacco and petrol and, as shown in Table 5.2, have remained fairly constant in nominal terms over time. As a proportion of total tax revenue, receipts from the National Lottery (including the money for the good causes) have fallen over time, from 1.0 per cent in 1995/96 to 0.5 per cent in 2005/06. Including revenues from other forms of gambling increases the total share, but again revenues have been fairly constant in nominal terms and the overall share has fallen from 1.4 per cent in 1995/96 to 0.7 per cent in 2005/06. Comparing tax receipts across other EU countries is complicated by the different ownership rules. But given the much larger per capita expenditure on gambling in the UK, even higher tax rates would be needed to generate revenues of a similar order of (absolute) magnitude (see Europe Economics, 2004).

102

Table 5.2 Financial year

Gambling revenues (£ million) and gross tax rates, UK Betting duty

Pools betting duty

Gaming

Amusement Machines

Rev

Rev

Bingo Rev

Rate

National Lottery

Rev

Rate

Rev

Rate

Rev

Rate

1996–7

453

6.79%

127

27.36%

79

128

96

9.64%

1,891

40%

1997–8

462

6.71%

97

27.83%

92

131

102

9.71%

2,204

40%

1998–9

480

6.72%

70

27.18%

91

157

105

10.44%

2,084

40%

1999–00

492

6.75%

38

18.61%

107

160

107

9.97%

2,027

40%

2000–1

487

6.87%

30

17.33%

129

153

114

10.18%

2,017

40%

2001–2

433

4.27%

26

17.85%

130

155

116

9.93%

1,933

40%

2002–3

304

1.62%

16

13.03%

151

149

122

9.98%

1,820

40%

2003–4

383

1.18%

13

11.98%

138

144

111

7.71%

1,835

40%

2004–5

443

0.98%

12

11.04%

157

157

81

4.55%

1,904

40%

2005–6

433

0.97%

11

12.17%

143

154

82

4.47%

2,004

40%

Notes: Rates are calculated as tax receipts divided by total stakes except in the case of the National Lottery. Information taken from HM Revenue and Customs (2007a,b). Data for total stakes is not available for gaming or amusement machines. The pools competitions are based on predicting the outcome of football matches.

Lotteries as a Source of Revenue 103 Table 5.3 products

Excise rates – UK National Lottery, alcohol, tobacco and petrol

Tax rate as % of sale price National Lottery

80%

Total revenue (£ million) 2,004

Spirits

63%

2,309

Wine

53%

2,308

Beer

49%

3,076

Cigarettes

76%

7,473

Petrol

67%

23,438

Notes: The tax rate on alcohol, tobacco and petrol includes both the ad valorem Value-Added Tax and the ﬁxed-rate excise duty. Since excise duties apply to quantity rather than price, tax rates will depend on assumed pre-tax prices. These are taken from HM Revenue and Customs (2007) for alcohol and tobacco and Wood Mackenzie OPAL for petrol. Receipts are for excise duties only.

In general, lotteries run by state governments tend to be relatively more important as a source of revenue, because of the typically narrower tax base than for national governments. For the US states, lotteries account for between 0.41 per cent (New Mexico) and 4.07 per cent (Georgia) of total state revenues (Kearney, 2005). Among the highest revenue shares from gambling are to be found in Australia. Worthington (2001) reports that in the case of New South Wales, taxes, fees and ﬁnes on gambling contributed 10.3 per cent of state taxes in 1996/7. However, these revenue numbers overstate the extent to which the introduction of a lottery would generate additional revenue for the government since lottery spending substitutes away from other taxable goods.6 Table 5.4 illustrates the potential revenue effects of a lottery. In these examples, we assume a gross lottery tax rate of 0.4 out of every £1 spent, a general sales tax rate on other non-lottery goods and services of 0.2 out of every £1 spent, additional ‘sin taxes’ on alcohol and tobacco of 0.5 out of every £1 spent and an income tax rate of 0.25. The expected value of the prize is equal to 0.5. These stylised examples are intended to illustrate the range of potential net revenue effect of the lottery – which may even be negative – and the importance of factors such as what Lottery spending substitutes for, whether the winnings are taxed and what the winnings are spent on. In practice, the revenue effects are likely to be further complicated by

104 Gaming in the New Market Environment Table 5.4

Tax revenues, with and without the lottery

£1 spent on the Lottery

Tax revenue lost

Tax revenue gained

Net revenue effect

Lottery spending substitutes for spending on other goods and services; winnings are taxed and post-tax winnings are spent on other goods and services

0.2

0.4 + (0.5 × 0.25) + (0.5 × 0.75 × 0.2) = 0.6

+0.4

Lottery spending substitutes for spending on other goods and services; winnings are not taxed and are spent on other goods and services

0.2

0.4 + (0.5 × 0.2) = 0.5

+0.3

Lottery spending substitutes for alcohol and tobacco; winnings are not taxed and are spent on other goods and services

0.7

0.4 + (0.5 × 0.2) = 0.5

–0.2

the distributional impact of the lottery. In the examples in Table 5.4, it was assumed that spending patterns were relatively unaffected by winning the lottery, but clearly with the size of the prizes on offer, that is unlikely to be the case. Indeed, the opportunity to change your life is precisely why many people play the lottery in the ﬁrst place. In this case, calculating the impact of the lottery on tax revenues may need to take into account where and how the winnings are invested (housing, equity, etc). Several studies in the US have attempted to measure the effect of lotteries on tax revenues from other sources. Borg et al (1993) looked at the relationship between total sales tax revenue and lottery sales over the period 1953–87 for nine states that operated lotteries on a continuous basis, ﬁnding a negative and signiﬁcant effect for six out of the nine states. Their results indicate that the lottery has a positive effect on overall revenues, but a smaller one than looking at lottery revenues would suggest. Using more recent data (1967–99) and controlling for self-selection into being a lottery state, Fink et al (2004) ﬁnd that the lottery has a negative effect on tax revenues overall. This is driven by a decline in sales and excise revenue, and is in spite of offsetting income

Lotteries as a Source of Revenue 105

and corporate tax increases. They argue that the increases in income and corporate tax revenues reﬂect a political as well as a behavioural response to the introduction of a lottery that needs to be factored into the overall impact on revenue. These two studies both rely on time-series variation in both lottery sales and tax revenues to identify any effect. One potential problem with this approach is that there may be other time-varying factors which affect both lottery sales and other tax revenues and which cannot be properly controlled for. Kearney (2005) provides further evidence of the (behavioural) spending response to the introduction of a lottery using household-level data from the US Consumer Expenditure Survey. She uses data from 1982–98 and exploits the fact that lotteries were introduced in 21 states over this period, but at different points in time, to identify the difference between changes in spending patterns in states that introduce lotteries compared to changes in spending in states that did not. The effects are not precisely estimated, but she ﬁnds that there is a decline in spending spread across a number of categories, including necessities such as food, rent, mortgages and other bills. This is particularly the case for low-income households. There is also a signiﬁcant reduction in the likelihood of buying alcohol and tobacco. Lottery spending does not appear to be ﬁnanced by a reduction in other forms of gambling – if anything, spending on other forms of gambling increases. From the point of view of revenuemaximisation, Kearney’s results are ‘good’ news for policy-makers – spending on lotteries appears to substitute away from goods which are not heavily taxed, implying only a small loss in (non-lottery) revenue. However, they highlight one of the main concerns with the lottery as a source of revenue, which is it impact on low-income households and the fact that a tax on the lottery is regressive, something we return to in Section 5.3.

5.3

Are high tax rates on lotteries ‘optimal’?

Since the time of Adam Smith, economists have attempted to deﬁne characteristics by which to judge taxes. There is fairly broad agreement on the following three desirable criteria (see Heady, 1993, for further discussion): • cost and convenience – that taxes should be cheap and easy to collect (both for the government and for taxpayers); • efﬁciency – that taxes should minimise disincentive effects; and

106 Gaming in the New Market Environment

• equity – that taxes should be fair, although the deﬁnition of fairness varies. So, how does a tax on the lottery perform when assessed against these criteria? Section 5.4 discusses equity – here we focus on cost and convenience and efﬁciency. 5.3.1

Cost and convenience

Lotteries tend to be relatively attractive as a source of revenue in terms of cost and convenience of collection. The costs of collecting income tax in the UK (including compliance costs for taxpayers) have been estimated at between 4–6 per cent of total revenue.7 The administrative costs of running a lottery are of a similar size (5 per cent of receipts in the UK), but these reﬂect the cost of delivering a service to consumers and not the cost to the government (and lottery-provider) of collecting the revenue, which is a tiny fraction of this. Dealing with a single lottery provider is clearly one way of keeping tax-collection costs to a minimum. It has been argued in the US that lotteries provide an unreliable source of revenue owing to large year-to-year variation. However, much of the measured variability arises from fast growth which, as Clotfelter and Cook (1989) point out, is not the sort of variability that politicians are likely to lose much sleep over. Moreover, Szakmary and Szakmary (1995) argue that lotteries have a potential stabilising role in a portfolio of taxes because the variability of lottery revenue is uncorrelated with that of revenue from other sources. Mikesell and Zorn (1990), for example, show that lottery revenues vary with such things as the age of the lottery and competition from lotteries in neighbouring states, factors that do not affect revenues from other taxes, while lottery spending has been found to be relatively unaffected by unemployment (Vasche, 1985), suggesting that it is likely to be less variable over the economic cycle than many other taxes. The revenue ﬁgures for the UK shown in Table 5.2 suggest that it is the lack of variability – and in particular the lack of growth – in revenue from the National Lottery which is a potentially greater issue. 5.3.2

Efﬁciency

High taxes on alcohol, tobacco and petrol are often justiﬁed on the grounds of economic efﬁciency as Pigouvian taxes, i.e., the additional taxes required to reﬂect the negative externalities that their consumption imposes on others (through e.g., the cost of healthcare for smokers,

Lotteries as a Source of Revenue 107

the cost of policing to deal with drunken disorder and the cost of pollution) and reduce consumption to an optimally lower level.8 There is little reason for believing that high taxes on the lottery are justiﬁable on the same externality grounds – much of the evidence presented in the chapter on problem gambling in this volume suggests that the social costs of playing the lottery are relatively small. Moreover, the fact that lotteries are typically heavily promoted by the states running them – and that the adverts are deliberately designed to encourage play, rather than simply providing information on the odds of winning or the potential dangers of excessive gambling – is inconsistent with the use of taxation as a tool to reduce consumption. These arguments suggest that taxes on lotteries could be lowered without undue concern about excessively high problem gambling. In the absence of signiﬁcant negative externalities, the work of Ramsey (1927) has provided important and inﬂuential insights into the design of optimal indirect taxes. He addressed the problem of designing sales taxes to raise a given amount of revenue at the least possible distortionary cost. The resulting ‘Ramsey rule’ is that taxes should be set to yield equal proportional reductions in the consumption of each good. Assuming that the demand for each good is independent of the prices of other goods, this equates to the ‘inverse elasticity rule’, that goods with more price-inelastic demands should be taxed more heavily. The basic intuition is simple – since the distortion from imposing indirect taxes arises as a result of individuals reducing their consumption, then in order to minimise distortion, taxes should be applied to goods where individuals will least change their behaviour. Incorporating distributional considerations, Diamond and Mirrlees (1971), showed that goods which are consumed relatively more by the poor should experience a lower-than-average proportional reduction – the exact degree depending on the coefﬁcient of inequality aversion. This produces a clear trade-off between efﬁciency and equity considerations since in practice many goods with low price elasticities also have low income elasticities. However, if redistribution can be achieved via other means (e.g., transfer payments) then, so long as all goods are normal, the government would be better achieving its distributional aims via transfer payments than from reductions in sales taxes, which will beneﬁt the rich more (see Atkinson and Stiglitz, 1976). As discussed further in Section 5.5, most estimates of the price elasticity of demand for lotteries suggest that demand is elastic (greater than one in absolute value), which would, according to the Ramsey rule, imply that taxes should be relatively lower than those on other

108 Gaming in the New Market Environment

goods. From a demand function, it is possible to estimate the welfare loss from the tax rate on the National Lottery and the associated reduction in demand. Walker (1998) estimates the following demand function: Weekly Sales = 4.7 + 135 × (Expected Ticket Value). At the current payout rate of 0.5, the estimated demand is 72.2 million (sales per week) while, if the payout rate were to increase to 0.9 (i.e., in the absence of any tax), then estimated demand would rise to 126.2 million. The implied loss in consumer surplus associated with the tax on the National Lottery is therefore equal to £10.8 million per week, or 37 per cent of the amount of revenue raised. It is important to emphasise, however, that this is a very rough estimate and that there are a number of issues around the demand elasticity estimate, which are discussed further in the next section. Moreover, this obviously overstates the extent to which cutting the tax rate on lotteries would increase consumer welfare since the revenue would have to be raised somewhere else, imposing other distortionary costs. However, the implication of the Ramsey rule is that (subject to distributional considerations) the overall welfare cost could be reduced by lowering the lottery tax and raising taxes on goods with less elastic demand because the distortionary behavioural response would be smaller. What about making up the revenue by an increase in income tax? For the UK, the amount of money raised from the National Lottery (approx £2 billion a year) would require a 0.5 per cent increase in the basic rate of income tax.9 Traditionally, estimates of the deadweight cost of income taxes focused on reductions in labour supply arising as a result of increased taxation. The fact that most studies ﬁnd labour supply to be fairly inelastic, at least for men, implied that the welfare cost of income tax was quite low.10 However, more recent work has recognised that the standard labour supply behavioural response is only one component of the total welfare cost of increasing income taxes, which should also take into account changes to the form of compensation (increasing tax-exempt forms of remuneration), unmeasured effort and compliance. Rather than looking just at labour supply, taxable income responses have been used to derive a more inclusive assessment of the deadweight cost of income taxes. At one extreme, the inﬂuential US study by Feldstein (1999) suggested that the deadweight cost of income tax could be as high as $2 for every $1 raised. Of course, the responsiveness to tax changes varies with income and is typically greater for higher rate taxpayers, suggesting that the cost of an increase in the basic rate of tax is likely to be nowhere near as high

Lotteries as a Source of Revenue 109

as Feldstein’s ﬁgures suggest. Nevertheless, the overall conclusion from the recent literature is that the deadweight cost of increased income taxation is likely to be at least of a similar order of magnitude to the estimated welfare loss associated with the high tax rate on the National Lottery.

5.4

Just how (un) fair is a tax on the lottery?

Tax from lotteries is sometimes referred to as a ‘painless tax’. There is a perception that it is voluntary, on the basis that it is only paid by those who willingly choose to play the lottery11 and, for politicians, raising incomes or sales taxes to replace lost revenue from the lottery is likely to impose considerable political pain. However, the one aspect of lottery taxes which may be seen as inﬂicting (ﬁnancial and political) pain is the fact that they disproportionately impact on the poor. Numerous studies have looked at the distributional impact of taxes on lotteries – see inter alia, Clotfelter and Cook (1989), Scott and Garen (1993), Stranahan and Borg (1998), Walker (1998) and Worthington (2001). The common ﬁnding, across time and across different countries, is that a tax on lotteries is regressive, i.e., the burden of the tax as a proportion of income is greater for poor households than it is for richer households. This is shown for the National Lottery – as well as for other ‘vices’, including other gambling, alcohol and tobacco – using recent data for the UK in Table 5.5.12 The tax is regressive not because playing the National Lottery is more likely among low income households; younger households (aged 25-plus) and older households (aged 65-plus) are less likely to play the Lottery and typically have lowerthan-average incomes. Nor is it regressive because low-income households who play the Lottery spend more than high income households. But, as household income rises, spending on the National Lottery rises by less, causing the percentage of income spent on the National Lottery to fall – from an average of 1.67 per cent of income among the poorest 10 per cent of households to 0.13 per cent among the richest 10 per cent of households. Suits (1977a) deﬁned a methodology for summarising and comparing the regressivity of different taxes, based on tax Lorenz curves, shown in Figure 5.1. These relate the cumulative proportion of the tax burden13 to the cumulative proportion of total income, across the income distribution. In practice, we deﬁne the curve over ten income deciles. If a tax were proportional, its Lorenz curve would lie along the diagonal. For a progressive tax, the Lorenz curve lies below this line

110 Gaming in the New Market Environment Table 5.5

Share of total income spent on ‘vices’, by income decile

Household income

Share of income on the National Lottery

Share of income on other gambling

Share of income on alcohol

Share of income on tobacco

Poorest 10%

1.67%

1.06%

6.67%

5.50%

Decile 2

0.82%

0.73%

3.51%

2.66%

Decile 3

0.83%

0.48%

3.43%

1.93%

Decile 4

0.70%

0.56%

3.22%

1.65%

Decile 5

0.56%

0.30%

3.20%

1.31%

Decile 6

0.48%

0.28%

2.98%

1.21%

Decile 7

0.40%

0.30%

2.69%

0.94%

Decile 8

0.36%

0.15%

2.53%

0.90%

Decile 9

0.28%

0.23%

2.26%

0.52%

Richest 10%

0.13%

0.08%

1.78%

0.29%

Average

0.58%

0.39%

3.11%

1.55%

Notes: Author’s calculations using data from (the database of) Food and Expenditure Survey 2004–05. Data from the Food and Expenditure Survey used in this paper were made available through the UK Data Archive (Http://www.data-archive.ac.uk/). Neither the original collectors of the data or the Archive bear any responsibility for the analyses or interpretations of the data that are presented here. Households with heads aged less than 18 or over 75 are dropped from the analysis. The FES is a diary-based expenditure survey in which respondents are asked to write down everything they purchase in a two-week period, averaged to estimate weekly spending ﬁgures. Assuming a total of 25 million households, the survey under-estimates total spending on these goods. But if the degree of under-reporting is not related to household income the analysis of regressivity will still be valid. Other gambling includes football pools, other lotteries, Irish Lottery, bingo (including admission) and spending at bookmakers.

and for a regressive tax, such as those on the lottery, other gambling, alcohol and tobacco, the Lorenz curve lies above. Figure 5.1 shows that the tax on the National Lottery is more regressive than taxes on alcohol, but very similar to taxes on tobacco and other gambling, although the difference is very small. The ‘Suits index’ is a summary measure of tax progressivity/regressivity, deﬁned by (1 – L/K), where L is the area under the Lorenz curve and K is the area under the diagonal.14 A negative number denotes a regressive tax and the larger in absolute terms, the more regressive the tax is. The index is equal to –0.35 in the case of the Lottery, –0.36 for other gambling, –0.38 in the case of tobacco and –0.14 for alcohol.15

80 60 40 20 0

cumulative percentage of spending

100

Lotteries as a Source of Revenue 111

0

20

40 60 cumulative percentage of income lottery alcohol

Figure 5.1

80

100

gambling tobacco

Tax Lorenz curves – lottery, gambling, tobacco and alcohol

These ﬁgures indicate that taxes on lotteries are among the most regressive of all taxes. Lowering the tax on the lottery and raising income or sales taxes16 is therefore likely to improve the overall fairness of the tax system. However, the fact that lottery spending comprises, on average, only a small component of total income means that changing the tax rate on the National Lottery is unlikely to have a major impact on the overall fairness of the tax system – the distribution of income taxes will have a far greater effect. However, as argued by Clotfelter and Cook (1989), looking at average spending within income deciles may fail to capture the full distributional impact because of the concentration of lottery purchases within income groups – Clotfelter et al (1999) report that in the US the top 5 per cent of players are responsible for more than half of all spending and the ﬁgures are similar for the UK. To illustrate the scale of the burden of high lottery taxes for certain households, Columns (1), (2) and (3) of Table 5.6 show the average proportion of household income spent on the National Lottery by, respectively, all Lottery players within each income decile, and the 10 per cent and 5 per cent heaviest players within each income decile. Within the bottom income decile, Lottery players spend more than 4 per cent of their income on the National Lottery, while the most active

112 Gaming in the New Market Environment Table 5.6

Share of spending on ‘vices’, by income decile

Household income

(1) Share of income on the National Lottery, Lottery players

(2) Share of income on the National Lottery, Heaviest 10% players in each decile

(3) Share of income on the National Lottery, Heaviest 5% players in each decile

(4) Share of income on ‘vices’ – those who play the Lottery, gamble, smoke and drink

Poorest 10%

4.20%

15.62%

19.61%

30.75%

Decile 2

1.96%

5.71%

6.79%

28.56%

Decile 3

1.81%

5.56%

6.78%

18.71%

Decile 4

1.40%

4.78%

6.42%

16.19%

Decile 5

1.13%

3.64%

4.52%

16.85%

Decile 6

0.90%

2.81%

3.45%

11.98%

Decile 7

0.71%

2.38%

3.09%

10.97%

Decile 8

0.65%

1.97%

2.47%

10.82%

Decile 9

0.52%

1.93%

1.57%

8.78%

Richest 10%

0.30%

1.05%

1.21%

6.41%

Average

1.18%

3.06%

4.32%

13.38%

See the notes to Table 5.5.

10 per cent of players spend more than 15 per cent of their household income on the Lottery and the most active 5 per cent spent nearly oneﬁfth of their income on the Lottery. Another consideration in the distributional impact of taxes on lotteries is the correlation between playing the lottery and other (heavily

Table 5.7

Correlation coefﬁcients of expenditures Play the Lottery

Other gambling

Other gambling

0.2046**

Drinking

0.1214**

0.1233**

Smoking

0.0989**

0.0325*

** denotes statistically signiﬁcant at the 1% level, * at the 5% level.

Drinking

0.0704**

Lotteries as a Source of Revenue 113

taxed) activities, particularly other gambling, smoking and drinking – shown in Table 5.7. Nearly one in ten of households who spend money on the National Lottery also spend money on all of the other three activities17 and for these households, the expenditure shares on the four activities are reported in Table 5.5, column (3). This shows that among the poorest 10 per cent of households, those who play the Lottery, gamble, drink and smoke spend 31 per cent of their income on these activities. Note that these considerations about the concentration of spending do not change the conclusion about the overall regressivity of the tax, but they may affect the conclusion about the implications for the burden of the tax on particular households. Where lottery revenues are genuinely earmarked for particular purposes, the overall distributional impact of a lottery tax should take account of who beneﬁts from the spending, as well as who pays the tax since a reduction in tax could mean a loss of beneﬁts as well as a reduction in tax burden. A couple of studies have considered earmarked spending on merit-based education scholarships in Florida (Borg and Stranahan, 2004) and Georgia (Rubinstein and Scaﬁdi, 2002), ﬁnding that, if anything, the distributional impact of the lottery is more regressive when taking account of spending, since lower-income households tend to beneﬁt less from these programmes. Feehan and Forrest (2007) reach a similar conclusion looking at the relationship between the distribution of National Lottery grants across UK local authorities and the local authority socio-economic characteristics. They ﬁnd that the amount of grants received is positively related to the proportion of residents in social class I and II. They suggest that the grant-application process and the need to raise matching funds may tend to disadvantage the less well-off in terms of receiving Lottery funding. More fundamentally, at least three of the areas initially allocated good causes funding from the National Lottery (sport, the arts, heritage) are likely to disproportionately beneﬁt those with relatively higher incomes. As outlined in an early government report on the potential for a National Lottery, the funding is intended for ‘good causes of the kind that are desirable rather than essential’ (Home Ofﬁce report, 1978). If the good causes were seen as essential, they would be likely to be already funded out of general tax revenue, and earmarked funding would be seen as too unreliable a source of funding. So, earmarked lottery spending seems almost inevitably to be dedicated to the kind of ‘desirable’ public goods that are of less beneﬁt to the less well-off.18

114 Gaming in the New Market Environment

5.5

Revenue maximisation and the new gaming environment

Assuming that the goal of the government is to set the rate of lottery taxation to maximise revenue (irrespective of the impact on consumer welfare), an obvious question is whether the current rate is revenuemaximising. For the government, there is a trade-off – further increases in the tax rate would yield more tax revenue from each lottery ticket sold, but tax revenue would be lost from a possible fall in demand. The overall effect on tax revenue depends on the balance of the two factors, which in turn depends on the sensitivity of demand to the price of the lottery. It has long been recognised that, so long as demand for a good is not completely inelastic, there is a non-linear relationship between tax rates and tax revenues. This was stated explicitly by Dupuit (1844): If a tax is gradually increased from zero up to a point where it becomes prohibitive, its yield is at ﬁrst nil, then increases by small stages until it reveals a maximum, after which it gradually declines until it becomes zero again. Since, for a lottery, tax revenues are simply a proportion of total revenues, the government will maximise tax receipts when total lottery revenues are maximised. In turn, the point of revenue maximisation occurs where the elasticity of demand with respect to price is equal to –1.19 So, if empirical estimates of the price elasticity of demand for a lottery are at, or close to, –1, then the tax rate is revenue-maximising; above this rate (in absolute value) and the tax rate is too high; below this and raising the tax rate will cause revenue to increase. Finding genuinely exogenous variation in the price of lotteries that permits identiﬁcation of the responsiveness of demand is not straightforward. Take-out rates vary across US states and across EU countries, but of course, there may be other (unobservable) regional or country factors which jointly determine both price and demand. Changes in the underlying parameters of the game design that affect price are relatively few and far between and, of course, may be linked to falling sales. The approach taken to estimating the elasticity of demand for the National Lottery in the UK (see Walker, 1998 and Forrest et al, 2000) has been to exploit weekly variations in the price that are linked to roll-overs.20 During a roll-over week, the expected value of the prize increases and the price of a ticket falls. Weekly sales data clearly show that demand is

Lotteries as a Source of Revenue 115

higher in roll-over weeks. Assuming that this increase in demand is driven solely by the reduction in price, then this can be used to identify and quantify demand responsiveness. Using aggregate sales data, Forrest et al (2000) estimate demand elasticity to be –1.03. Using microdata Walker (1998) obtains a very similar estimate of –1.07. If accurate, these estimates suggest that the current tax rate is close to, if slightly above, its revenue-maximising level. However, a crucial issue for policy-makers is the extent to which they can extrapolate from these estimates to the demand responses to changing the parameters of the game. In this case, there are some potential issues arising from the methodology used to obtain the elasticity estimate, which need to be borne in mind. First, the changes in price brought about through roll-overs are nonmarginal. Walker (1998) reports that the expected value of the prize changed from an average of 0.45 during regular draw weeks to 0.58 in a roll-over week, to 0.65 in the case of a double roll-over. If the government is thinking about ﬁne-tuning the price to its revenue-maximising level, there is an element of uncertainty over whether the response in demand to small changes in price would be similar. Second, an important identifying assumption is that roll-overs affect demand only through their effect on price. There is assumed to be no independent ‘rollover effect’ on demand. However, as well as the size of the prize on offer, at least part of the reason why people take part in the lottery draw is for fun and this may in turn be related to the number of people taking part. It is plausible that there is some additional excitement generated by the exceptional nature of a rollover – the more people take part, the greater the buzz there is surrounding the weekly lottery draw (over and above the effect of the larger jackpot), and the greater the fun associated with buying a ticket. In this case, the estimated responsiveness of demand will also capture these positive network externality effects. Third, there is an underlying assumption that people correctly perceive the relative value of a rollover ticket compared to a regular weekly ticket. Walker (1998) presents evidence from a 1995 UK Consumers’ Association survey which shows that two-thirds of consumers misperceive the odds of winning (although there is no systematic bias across the population as a whole). What matters for the estimated demand elasticity is that individuals’ misperceptions are systematic across the two types of draw. If people playing the lottery incorrectly estimate the change in price that occurs when there is a rollover, the demand elasticity will be imprecisely measured.

116 Gaming in the New Market Environment

Fourth, if these results are to be extrapolated to changes in the takeout rate, there is an underlying assumption that demand for lottery tickets is unaffected by the amount of money going to the good causes. The changes in price and demand associated with roll-overs do not affect the design of the lottery and the amount of money allocated to the good causes. A permanent reduction in the take-out rate, on the other hand, would reduce the amount of money going to the good causes and, in principle, this may affect demand, although the evidence presented in the next section suggests that any effect is likely to be negligible. Finally, there is the impact of the new gaming environment on the price sensitivity of demand. The effect of increased competitive pressure from other forms of gambling and/or online lotteries will be to increase the price-responsiveness of demand, and this will reduce the optimal, revenue-maximising tax rate, at the same time as reducing the overall amount of revenue that the government can raise from lotteries.21 This line of argument explains why taxes have been reduced on other forms of betting to counter the effect of increased competition from offshore betting in the UK, and why taxes on US state lotteries have been falling. Similar pressures are likely to affect taxes on lotteries in Europe. The studies that have estimated price elasticities typically predate the new competitive environment – understanding the extent to which demand is becoming more sensitive to price – and the implications of this for the revenue-maximising tax rate – in the new competitive environment is clearly crucial.

5.6

Are lotteries good for the ‘good causes’, and are the ‘good causes’ good for lotteries?

A feature of most state lotteries is that the revenues are earmarked for particular good causes. Only ten out of the 38 US states that operate a lottery include lottery revenues as part of general tax revenue. Sixteen states earmark revenues for spending on education, including meritbased college scholarship schemes. Other recipients range from the relatively narrow – police and ﬁremen’s pensions in Indiana, for example and the Mariners’ stadium in Washington, to the broad, including ‘parks and recreation’ and ‘economic development’. On average across lotteries in the EU, around half of total revenue was allocated to good causes, with the main beneﬁciaries being sport, culture, education and youth, charities, heritage and science and technology (see London Economics, 2006). In the UK, the current main recipient of money for

Lotteries as a Source of Revenue 117

the good causes is the Big Lottery Fund which receives half the money and allocates it to projects involving health, education, environment and charitable causes. The remaining half is divided between sports, the arts and heritage. If the tax rate on lotteries is reduced, this will affect the revenues for these earmarked good causes. The two questions we address here are: • Does hypothecation have a real effect on government expenditure patterns, and will the good causes suffer if earmarked revenues dry up? • Does hypothecation have any effect on lottery participation and will reduced earmarking affect demand for the lottery? There is an historical association between lotteries and charitable purposes. In the past, lotteries were used by the government to fund good works, such as building the British Museum in the UK and Harvard and Columbia Universities in the US. And individual charities have often used lotteries to raise funds. One possibility is that governments have exploited this link to earmark revenues for good causes, purely as a window-dressing exercise. Their aim in doing so is to engender broad support for the lottery and to silence any opposition to the introduction of lotteries on moral grounds. Hypothecating the money to good causes, and introducing a principle of additionality, as in the UK, is a way of meeting the criticism that the lottery is being introduced purely to raise money; it may also dampen any pressure to cut other taxes. More strategically, targeting lottery revenues at particular good causes creates a vested interest that will lobby hard to keep the lottery in the future.22 As with any hypothecated tax, the good causes funding is subject to the criticism that it is an inefﬁcient way to fund public services. The level of spending will be determined by the amount of revenues on a year-to-year basis, rather than being set at an optimally efﬁcient level. This is particularly the case with ‘strong hypothecation’ (see Wilkinson, 1994) where the level of expenditures is completely determined by dedicated revenues. The potential variability in revenues, and the fact that the drivers of revenue are not necessarily linked to the drivers of spending need, make it unlikely that the money will be hypothecated to essential public services. The other common criticism is that hypothecation is purely a PR exercise and that revenue from the lottery will simply displace other revenue. This is perhaps even more likely to be the case with weak

118 Gaming in the New Market Environment

hypothecation, where the lottery revenues form only a small part of total spending. As in the UK, the monies can be ring-fenced and collected and spent by non-governmental agencies. But even then, there is no guarantee that total spending on the good causes will actually increase since the government may simply use the earmarked revenues to reduce the level of spending they otherwise would have committed. Establishing the true effect of the lottery is very hard in the absence of any veriﬁable way of determining the counterfactual. US studies by Spindler (1995), Novarro (2002) and Evans and Zhang (2007) ﬁnd that earmarking did increase the total amount of state spending on education (compared to states that did not earmark), but by less than the full amount of lottery money raised. In the UK, the effect on spending patterns depends on a fairly complex interaction between the central funding agencies, the central government and the local governments who are often allocated grants. Even if the central government doesn’t change pre-lottery spending levels, there may still be offsetting effects at the local level. Bailey and Connolly (1997) show how the use of matching grants rather than lump sums, as in the UK, makes it more likely that there will be a change in spending patterns because of the implied change in price of the good causes.23 However, spending on other goods and services will also increase because of the income effect. So, are lotteries good for good causes, and will the good causes suffer if lottery revenues dry up? At the very least, allocating lottery revenue to designated good causes, particularly via independent non-governmental agencies, is likely to change the distribution of recipients, compared to funding the same causes out of general taxation. This re-distribution creates very obvious lottery winners, and those who currently received the earmarked funds will strongly oppose any reduction in high, hypothecated lottery taxes. The evidence suggests that if lottery taxes were to be cut, the amount these good causes would actually lose would be less than the full amount of currently earmarked revenues, although this is not a message that politicians will ﬁnd easy to deliver. Reducing lottery taxes and the associated earmarked spending is likely to be far from politically painless. The second issue is whether the good causes are good for lotteries and whether cutting the amount of money going to the good causes may damage popular support. Lotteries which earmark part of their revenue for good causes represent a unique bundle of a lottery ticket linked with a charitable donation. Morgan (2000) presents a model in which such a bundle is used to increase contributions to good causes,

Lotteries as a Source of Revenue 119

compared to voluntary donations. The argument is that the negative externality associated with the lottery overcomes the free-rider problem associated with voluntary donations. Morgan and Sefton (2000) provide laboratory-setting evidence that introducing a lottery increases donations compared to pure voluntary subscriptions, that bigger prizes are associated with more donations and that changes in the desirability of the public good have a signiﬁcant effect on betting behaviour. However, the main result applies to ﬁxed-prize lotteries, rather than the pari-mutual type of lotteries run by most states. Moreover, the comparison for state lotteries is not whether earmarking some of the revenues increases charity revenue compared to voluntary donations. Instead, assuming that the goal is revenue-maximisation, the relevant question is whether earmarking raises participation. If consumers do appear to derive some beneﬁt (‘warm glow’) from earmarking, this might also offset at least some of the loss in consumer welfare associated with the high tax rate. Good causes and lottery spending

400

Figure 5.2

200

IT

AU

FR

100

Spending per capita

300

SP

LU NE

ES LI

0

IR GE

DE

POR

SW CY UK

SLO

HU

0

BE

FI

CZ SL LA

.2 .4 Proportion on good causes

GR POL

.6

AU = Austria, BE = Belgium, CY = Cyprus, CZ = Czech Republic, DE = Denmark, ES = Estonia, FI = Finland, FR = France, GE = Germany, GR = Greece, HU = Hungary, IR = Ireland, IT = Italy, LA = Latvia, LI = Lithuania, LU = Luxembourg, NE = Netherlands, POL = Poland, POR = Portugal, SLO = Slovakia, SL = Slovenia, SP = Spain, SW = Sweden, UK = United Kingdom. Source: Data taken from London Economics (2006)

120 Gaming in the New Market Environment

There is very little evidence on the impact of earmarking on participation in lotteries. Vasche (1985) and Morgan (2000) show that US states that earmark lottery revenues for good causes have higher average per capita lottery expenditure than states that do not, but of course, this relationship does not prove causality. As shown in Figure 5.2, there is also a positive relationship across EU states between the proportion of ticket price allocated to good causes and per capita spending. The Spanish lottery is something of an outlier, but taking this observation out, a regression of per capita spending on the proportion spent on good causes (and total spending on gambling) yields a positive and statistically signiﬁcant coefﬁcient. Again, however, this proves very little. Morgan (2000) argues that the fact that the good causes feature in lottery advertising is evidence of an important linkage. He ﬁnds evidence that approximately 4 per cent of television lottery advertising content consists of information about the public beneﬁt of lotteries. Clotfelter et al (1999) analyse the key themes in the marketing plans of 22 of the state lotteries in the US. The most common themes are about winning the jackpot (56 per cent of plans have a theme relating to the size of the prize and 46 per cent to previous winners) and about the fun of playing (56 per cent), but 28 per cent of plans include the beneﬁts to the state of lottery dollars and 28 per cent have sports themes. This suggests that advertisers keen to promote state lotteries think it beneﬁcial to talk about the good causes, although this may be as much about positioning the lottery strategically, as about selling it to consumers. Survey evidence on the link between the National Lottery and the good causes in the UK is available from The British Social Attitudes Survey, and summarised in Table 5.8. When people are asked directly what most explains why they play the lottery, winning the jackpot is the main factor for three-quarters of players. Giving to the good causes is the most important factor for only 2 per cent. As shown in Table 5.8, this proportion varies by age (it is higher among older players than among younger) and by income (it is lowest among the middle-income group). Clearly, the good causes are not the main driver of lottery participation, but they still may be a factor. Respondents are also asked which is the least important of the three reasons and, on this basis, we can divide the group who play primarily for the jackpot, into those who consume a bundle either of (1) ‘jackpot + fun’ or of (2) ‘jackpot + donation’. The responses suggest that more people think of it as (1) than (2), but only slightly – 53 per cent compared to 47 per cent. Other evidence from the British Social Attitudes Survey is less convincing on the importance of the good causes. There is some confusion

Lotteries as a Source of Revenue 121 Table 5.8

Why do people play the National Lottery? To win the jackpot

To have some fun

To give to the good causes

All

77.1%

20.7%

2.2%

Household income < £10,000 a year £10,000–£41,000 > £41,000

76.3% 77.1% 71.9%

20.7% 21.1% 22.8%

3.0% 1.2% 5.3%

Age 18–34 35–54 55+

81.2% 79.0% 69.9%

18.5% 19.0% 25.6%

0.3% 2.0% 4.6%

Lottery participation Once a week Two or three times a month Once a month or less

80.0% 74.5% 70.4%

18.1% 22.4% 22.4%

2.0% 3.1% 2.3%

Note: Author’s calculations using data from the British Social Attitudes Survey, 1997 database. The question asks: ‘What most explains why you take part in the lottery’. Respondents are asked to pick one of the three answers. Answers are restricted to those who play directly, rather than as part of a consortium. This reduces the proportion of who play to have some fun.

about who beneﬁts from the good causes money. One quarter of respondents thought that some money went to a special fund for business and industry and 15 per cent thought that some money went to a special fund for the National Health Survey, neither of which received any money at the time of the survey. Also, as shown in Table 5.9, there appears to be little support for good causes that beneﬁt from Lottery money. Of the four causes that got the most popular support, one (medical research) received no Lottery money, while three beneﬁted only indirectly via the charities fund (children, homeless people and the disabled). Three of the good causes that received Lottery funding – sports, the arts and heritage – have very little popular support. They get relatively more support from those with higher incomes, but even among this group, other causes are more popular. Since the introduction of the Lottery (and after this survey was carried out), the government has changed the distribution of good causes funding, reducing the amount allocated to sports, the arts and heritage and introducing the Big Lottery Fund which distributes half the Lottery good causes money to health, education, the environment and charitable causes.

122 Gaming in the New Market Environment Table 5.9

What is ‘an excellent way to spend’ lottery money? All

Income < £10,000

£10,000– £41,000

Income > £41,000

Helping to protect children in need in Britain

39.9% (1)

46.8% (1)

37.3% (1)

32.7% (3)

Medical research in Britain

37.3% (2)

41.0% (3)

36.1% (2)

33.7% (2)

Helping disabled people in Britain

36.6% (3)

42.3% (2)

32.4% (3)

38.8% (1)

Helping homeless people in Britain

25.6% (4)

31.5% (4)

22.6% (4)

30.6% (4)

Helping to protect the environment

15.3% (5)

15.0% (6)

14.5% (5)

18.6% (5)

Helping to prevent cruelty to animals in Britain

15.1% (6)

21.8% (5)

11.6% (6)

13.3% (7)

Providing sports facilities in Britain

10.8% (7)

7.7% (8)

10.2% (7)

14.3% (6)

Helping starving people in poor countries

10.7% (8)

13.2% (7)

9.1% (8)

11.2% (8)

Helping ex-prisoners to ﬁnd homes and jobs

3.2% (9)

2.9% (10)

3.2% (9)

3.1% (11)

Helping to restore historic buildings in Britain

3.1% (10)

3.8% (9)

2.0% (10)

6.1% (10)

Supporting art galleries, theatres and orchestras in Britain

2.1% (11)

1.0% (11)

1.2% (11)

9.2% (9)

Note: Authors’ calculations using data from the British Social Attitudes Survey, 1996 database. Respondents are asked to pick one of ﬁve answers, ranging from an ‘excellent way to spend the money’ to ‘should not be spent on this at all’. The percentages in the table refer to the percentage who agreed it was an excellent way to spend the money. The numbers in the brackets are the order of priority, based on the percentages.

However, if the preferences shown in Table 5.9 are to be believed, the overall balance of Lottery good causes funding is still a long way from reﬂecting popular priorities.

Lotteries as a Source of Revenue 123

When presented with a hypothetical scenario where none of the money went to good causes, 21 per cent said this would make a lot of difference to whether they took part and a further 28 per cent said it would make some difference. However, these hypothetical questions are hard to interpret, while the other, harder, evidence suggests that reducing the proportion of Lottery money going to the good causes would have very little impact on demand.24

5.7

Conclusions

In the new competitive gaming environment there is likely to be pressure on governments to reduce lottery taxes. That is the right thing to do in order to meet the goal of revenue-maximisation if demand becomes more price-sensitive as a result of increased competition. Although they raise nowhere near the amount of revenue as income or sales taxes, lottery taxes are a useful source of revenue for governments – they are a painless tax to the extent that they are cheap and easy to collect and politically less difﬁcult than increases in sales or income taxes. There is no obvious economic efﬁciency argument for the current high tax rates on lotteries compared to other goods. High taxes on vices, such as smoking or drinking are typically justiﬁed by these activities’ big negative externalities. But the evidence on problem gambling suggests that the negative externality argument is unlikely to justify such a high tax rate in the case of lotteries. The fact that demand for lotteries appears to be quite sensitive to price suggests that the consumer welfare loss associated with taxing the lottery is likely to be fairly high, and the Ramsey rule would imply that the overall distortionary cost could be reduced by raising taxes on goods with relatively less sensitive demand, subject to distributional considerations. Reducing lottery taxes would also be likely to improve the overall fairness of the tax system, although the impact would be very small since spending on lotteries is a small part of total income. If lottery taxes were cut it would mean a loss of earmarked revenue for the good causes. The evidence suggests that this would have little or no effect on the demand for the lottery – for most people it is the jackpot that matters, not the good causes. However, the loss would obviously be strongly opposed by the good causes themselves. Although most studies (from the US) ﬁnd that the good causes do not beneﬁt by the full amount of the earmarked revenues, there does appear to have been a positive effect of hypothecation on education spending. While

124 Gaming in the New Market Environment

the good causes may have been seen as a useful bit of window dressing when lotteries were introduced, they are likely to turn into a political headache if lottery taxes – and the good causes funding – have to be reduced.

Notes 1 [email protected] 2 Throughout this chapter, ‘states’ will be used to refer to both state governments and national governments. Where US states are meant, this will be made clear. 3 The tax may either be explicit, or represent the proﬁts from a state-run enterprise. In the case of a monopoly, the two are equivalent. 4 Section 4, National Lottery etc Act, 1993, amended. 5 If revenues are earmarked for good causes, these are included as tax. 6 In setting a 12 per cent explicit tax rate on the National Lottery, the stated aim of the UK government was that the introduction of the Lottery should be revenue neutral. 7 Kay and King (1990). 8 This is only a theoretical argument – whether the actual level of these taxes is justiﬁed by the size of these externalities is far from clear. See e.g., Viscusi (1994) for a discussion. 9 The 2007 Budget pre-announced a 2 pence cut in the basic rate, costed at £8.090 billion in 2008–09 and £9.640 billion in 2009–10. 10 Harberger estimated the deadweight loss to be around 2.5 per cent of income tax revenue. 11 Although the argument that a tax is only paid by those who choose to engage in the activity is true of any tax with the exception of a poll tax. 12 The ﬁgures refer to total spending rather than tax. Where the tax is a proportion of total spending, as in the case of the National Lottery, the pattern of tax regressivity will be the same as for spending, although the burden of the tax will be smaller than these ﬁgures suggest. For taxes on alcohol and tobacco which are based on quantity, the degree of regressivity in the distribution of tax is likely to be greater than the spending ﬁgures suggest if poorer households typically buy goods of lower quality and lower (pre-tax) price. 13 In practice we deﬁne the Lorenz curves over total spending rather than tax. See footnote 10 for further discussion. 10

14 L is calculated as ∑ 0.5[T(yi) – T(yi–1)](yi – yi–1), where T(yi) is the cumulative i=1

percentage of tax accounted for by income decile i and yi is the cumulative percentage of income. K is equal to 5000. 15 Clotfelter and Cook (1989) report similar tax incidence ﬁgures of –0.32 – –0.48 for US state lotteries, –0.38 for tobacco and –0.21 for alcohol. The ﬁgure for federal income tax is +0.2. 16 The regressivity of sales taxes is typically reduced by excluding certain goods, such as food in the UK, which form a large part of the budget of poorer households. 17 Among the heaviest 10 per cent of players, this ﬁgure rises to 12 per cent.

Lotteries as a Source of Revenue 125 18 The Labour government changed the good causes in 1998, introducing a New Opportunities Fund for innovative projects in health, education and the environment. Arguably such projects may have been of greater beneﬁt to the less well-off, but there was a criticism that funding was being diverted into areas traditionally funded by the exchequer. 19 This is true where the government imposes a tax as a percentage of revenue. Where a lottery is run as a state monopoly, the government will maximise its revenue by maximising proﬁts. This in turn involves setting marginal cost equal to marginal revenue – both of which will be affected by levels of demand. Viren (2008) discusses marginal cost issues in more detail. 20 Clearly, the expected prize value varies from week to week because of variations in sales, but this is endogenous. 21 Crawford and Tanner (1995) discuss how the introduction of the Single Market might affect the optimal tax rate on alcohol and tobacco; similar arguments apply to increased competition for the lottery. 22 In the UK, many charities initially opposed the National Lottery because of concern that it would lead to a reduction in voluntary donations and proceeds from their own lottery games. However, once it became clear that this was not the case, they became strong supporters of the Lottery because of the good causes money given to charities. 23 There may also be a ‘ﬂypaper effect’ – i.e., at least some of the money sticks where the central government (or its designated agencies) sends it. 24 The Monday Lottery was introduced in the UK as a competitor to the National Lottery which aimed to give players more control over where the good causes money could go (‘to a charity of their choice’). The failure of this lottery to compete on anywhere near the same level as the National Lottery may indicate the lack of popular support for a charity lottery, although, as a new entrant, the Monday Lottery was also heavily disadvantaged in a number of other ways relative to the incumbent.

6 Problem Gambling and European Lotteries Mark Grifﬁths1

6.1

Background

Although most people gamble occasionally for fun and pleasure, gambling brings with it inherent risks of personal and social harm. Problem gambling can negatively affect signiﬁcant areas of a person’s life, including their physical and mental health, employment, ﬁnances and interpersonal relationships (e.g., family members, ﬁnancial dependents) (Grifﬁths, 2004). There are signiﬁcant co-morbidities with problem gambling, including depression, alcoholism, and obsessive-compulsive behaviours. These co-morbidities may exacerbate, or be exacerbated by, problem gambling. Availability of opportunities to gamble and the incidence of problem gambling within a community are known to be linked (Grifﬁths, 2003a, Abbott and Volberg, 2007). This chapter examines to what extent lotteries can cause or exacerbate problem gambling. This chapter makes particular reference to the British Gambling Prevalence Survey (BGPS; Sproston, Erens and Orford, 2000), but makes reference to other European studies where relevant (e.g., if other countries have markedly different ﬁndings).

6.2

Problem gambling: What do we know?

Deﬁnition of gambling: Gambling is a diverse concept that incorporates a range of activities undertaken in a variety of settings. It includes differing sets of behaviours and perceptions among participants and observers (Abbott and Volberg, 1999). Predominantly, gambling has an economic meaning and usually refers to risking (or wagering) money or valuables on the outcome of a game, contest, or other event in the hope of winning additional money or material goods. The activity varies on several 126

Problem Gambling and European Lotteries 127

dimensions, including what is being wagered, how much is being wagered, the expected outcome, and the predictability of the event. For some things such as lotteries, most slot machines and bingo, the results are random and unpredictable. For other things, such as sports betting and horse racing, there is some predictability to the outcome and the use of skills and knowledge (e.g., recent form, environmental factors) can give a person an advantage over other gamblers. Some of the most common types of ofﬂine commercial forms of gambling are summarised in Table 6.1. As can be seen from Table 6.1, gambling is commonly engaged at a variety of environments including those dedicated primarily to gambling (e.g., betting shops, casinos, bingo halls, amusement arcades), those where gambling is peripheral to other activities (e.g., social clubs, pubs, sports venues), and those environments where gambling is just one of many things that can be done (e.g., supermarkets, post ofﬁces or petrol stations). In addition, most types of gambling can now be engaged in remotely via the internet, interactive television and/or mobile phone. This includes playing roulette or slot machines at an online casino, the buying of lottery tickets using a mobile phone or the betting on a horse race using interactive television. In these remote types of gambling, players use their credit cards, debit cards or other electronic forms of money to deposit funds in order to gamble (Grifﬁths, 2005a). Issues surrounding remote gambling will be examined later in this chapter. Deﬁnition of terms: The term ‘problem gambling’ has been used by many researchers, bodies, and organisations, to describe gambling that compromises, disrupts or damages family, employment, personal or recreational pursuits (Budd Report, 2001, Sproston et al, 2000, Grifﬁths, 2004). The two most widely used screening instruments worldwide are the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSM-IV) for pathological gambling (American Psychiatric Association, 1994), and the South Oaks Gambling Screen (SOGS) (Lesieur and Blume, 1987) (see Appendices 1 and 2). There is some disagreement in the literature as to the terminology used, as well as the most appropriate screens to diagnose and measure the phenomenon. Researchers internationally are beginning to reach a consensus over a view of problem gambling that moves away from earlier, clinical often heavily DSM-based deﬁnitions. For instance, early conceptions of ‘pathological gambling’ were of a discrete ‘disease entity’ comprising a chronic, progressive mental illness, which only complete abstinence could hope to manage. More recent thinking regards problem

128

Table 6.1

A summary of the most common forms of ofﬂine commercial gambling

Type of gambling

Brief description

Lotto

Lottery game where players pick (say) six out of 49 numbers to be drawn bi-weekly for the chance to win a large prize. Tickets can be bought in a wide variety of outlets including supermarkets, newsagents or petrol stations.

Bingo

A game of chance where randomly selected numbers are drawn and players match those numbers to those appearing on pre-bought cards. The ﬁrst person to have a card where the drawn numbers form a speciﬁed pattern is the winner. Usually played in bingo halls but can be played in amusement arcades and other settings (e.g., church hall).

Card games (e.g., poker, bridge, blackjack)

Gambling while playing card games either privately (e.g., with friends) or in commercial settings (e.g., land-based casino) in an attempt to win money.

Sports betting

Wagering of money for example on horse races, greyhound races or football matches. Usually in a betting shop in an attempt to win money.

Non-sports betting

Wagering of money on a non-sporting event (such as who will be evicted from the ‘Big Brother’ house) usually done in a betting shop in an attempt to win money.

Scratchcards

Instant win games where players typically try to match a number of winning symbols to win prizes. These can be bought in the same types of outlet as the National Lottery.

Roulette

Game in which players try to predict where a spinning ball will land on a 36-numbered wheel. This game can be played with a real roulette wheel (e.g., in a casino) or on an electronic gaming machine (e.g., in a betting shop).

Table 6.1

A summary of the most common forms of ofﬂine commercial gambling – continued

Type of gambling

Brief description

Slot machines (e.g., These are stand-alone electronic gaming machines that come in a variety of guises. These include many fruit machines, different types of ‘slot machine’ (typically played in amusement arcades, family leisure centres, casinos, ﬁxed odds etc) and ﬁxed odds betting terminals (FOBTs) typically played in betting shops. betting terminals) Football pools

Weekly game in which players try to predict which football games will end in a score draw for the chance of winning a big prize. Game is typically played via door-to-door agents.

Spread betting

Relatively new form of gambling where players try to predict the ‘spread’ of a particular sporting activity such as the number of runs scored in a cricket match or the exact time of the ﬁrst goal in a football match in an attempt to win money. Players use a spread betting agency (a type of specialised bookmaker).

a) Most of these forms of gambling can now be done via other gambling channels including the internet, interactive television and/or mobile phone. b) There are other types of gambling such as dice (casino-based ‘craps’), keno (a fast draw lottery games) and video lottery terminal machine. c) Technically, activities such as speculation on the stock market or day trading are types of gambling but these are not typically viewed as commercial forms of gambling and they are not taxed in the same way.

129

130 Gaming in the New Market Environment

gambling as behaviour that exists on a continuum, with extreme, pathological presentation at one end, very minor problems at the other, and a range of more or less disruptive behaviours in between. Moreover, this behaviour is something that is mutable. Research suggests it can change over time as individuals move in and out of problematic status and is often subject to natural remission (Hayer, Grifﬁths and Meyer, 2005). Put more simply, gamblers can often move back to non-problematic recreational playing after spells of even quite serious problems. This conception ﬁts in with an emphasis on more general public health, with a focus on the social, personal and physical ‘harms’ that gambling problems can create among various sectors of the population, rather than a more narrow focus on the psychological and/or psychiatric problems of a minority of ‘pathological’ individuals. Such a focus tends also to widen the net to encompass a range of problematic behaviours that can affect much larger sections of the population. The screening tools that are currently used to diagnose the existence and severity of problem gambling reﬂect this change of focus. There have been criticisms of both the DSM-IV and the SOGS. In part, these criticisms stem from an acknowledgment that both screens were designed for use in clinical settings, and not among the general population, within which large numbers of individuals with varying degrees of problems reside. Other alternative screening instruments have been developed, and these are increasingly being used internationally (Abbott, Volberg, Bellringer and Reith, 2004). One such screening tool is the Problem Gambling Severity Index (PGSI), which was developed in Canada and has been used in that country, the USA, Australia and the UK. A ‘harm based’ conception of problem gambling has implications for policy and treatment. Given that the most severe cases of pathological gambling are one of the most difﬁcult disorders to treat (Volberg, 1996), and given that, at various points in their lives, hundreds of thousands of people in the general population may experience some degree of gambling-related harms, it becomes important to provide intervention strategies that can prevent this potentially larger group developing more serious problems. To this end, public health education and awareness-raising initiatives come to the fore, and these are recognised internationally as the most cost-effective way of dealing with problem gambling in the long term (Shaffer, Hall and Vander Bilt, 1999, Abbott et al, 2004, National Gambling Impact Study Commission, 1999). Such strategies have been successfully deployed in countries such as Australia, New Zealand and Canada.

Problem Gambling and European Lotteries 131

There is a multitude of terms used to refer to individuals who experience difﬁculties related to their gambling. These reﬂect the differing aims and emphases among various stakeholders concerned with treating patients, studying the phenomenon, and inﬂuencing public policy in relation to gambling legislation. Besides ‘problem’ gambling, terms include (but are not limited to) ‘pathological’, ‘addictive’, ‘excessive’, ‘dependent’, ‘compulsive’, ‘impulsive’ ‘disordered’, and ‘at-risk’ (Grifﬁths and Delfabbro, 2001, Grifﬁths, 2006). Terms are also employed to reﬂect more precisely the differing severities of addiction. For example, ‘moderate’ can refer to a lesser level of problem, and ‘serious problem gambling’ for the more severe end of the spectrum. Although there is no absolute agreement, commonly ‘problem gambling’ is used as a general term to indicate all of the patterns of disruptive or damaging gambling behaviour. This chapter follows this precedent, employing the use of the term ‘problem gambling’ to refer to the broad spectrum of gambling-related problems. Problem gambling must be distinguished from social gambling and professional gambling. Social gambling typically occurs with friends or colleagues and lasts for a limited period of time, with predetermined acceptable losses. There are also those who gamble alone in a non-problematic way without any social component. In professional gambling, risks are limited and discipline is central. Some individuals can experience problems associated with their gambling, such as loss of control and shortterm chasing behaviour (whereby the individual attempts to recoup their losses) that do not meet the full criteria for pathological gambling (American Psychiatric Association, 1994). Social Context: Research into gambling practices, the prevalence of problem gambling, and the socio-demographic variables associated with gambling and problem gambling, has not been considered part of mainstream health research agendas until quite recently. The BGPS (Sproston et al, 2000) revealed gambling to be a popular activity in Britain. In the year covered by the survey, gambling was engaged in by almost three-quarters of the population (72 per cent), with the most popular gambling activity being the National Lottery Draw (i.e., Lotto). Two-thirds of the population bought a National Lottery ticket in the year covered by the survey (65 per cent), while the next most popular gambling activity was the purchase of scratchcards (22 per cent), followed by playing slot machines (14 per cent), horse race gambling (13 per cent), football pools (9 per cent) and bingo (7 per cent). For a large number of people (39 per cent of those who purchased national Lottery tickets), the National Lottery Lotto game was the only gambling

132 Gaming in the New Market Environment

activity they participated in. In the few other countries that have carried out national prevalence surveys, similar results have been found, with lottery products being the most popular (e.g., Sweden; Orford, Sproston, Erens et al, 2003). The BGPS also found that men were more likely than women to gamble (76 per cent of men and 68 per cent of women gambled in the year covered by the survey), and tended to stake more money on gambling activities. The gambling activities men and women participate in were also varied. Men were more likely to play football pools and slot machines, bet on horse and dog races, and to make private bets with friends, while women were more likely to play bingo, and tended to participate in a lesser number of gambling activities overall (Sproston et al, 2000). Examination of prevalence and socio-demographic variables associated with problem gambling undertaken in the BGPS revealed that between 0.6 per cent and 0.8 per cent (275,000 to 370,000 people) of the population aged 16 and over were problem gamblers (Sproston et al, 2000). In comparison to other countries (such as Australia, the United States, New Zealand and Spain which have problem gambling rates of 2.3, 1.1, 1.2 and 1.4 per cent respectively), the number of problem gamblers in Britain is – based on the 2000 prevalence survey – relatively low (Sproston et al, 2000). Only one other country (i.e., Sweden) has a lower rate for problem gambling (Orford et al, 2003). Proﬁling: The BGPS revealed that there were a number of sociodemographic factors statistically associated with problem gambling. These included being male, having a parent who was or who has been a problem gambler, being separated or divorced and having a low income. Low income is one of the most consistent factors associated with problem gambling worldwide. This may be both a cause and an effect. Being on a low income may be a reason to gamble in the ﬁrst place (i.e., to try and win money). Additionally, gambling may lead to low income as a result of consistent losing. In Britain, people in the lowest income categories are three times more likely to be classed a problem gambler than average (Sproston et al, 2000). Although many people on low incomes may not spend more on gambling, in absolute terms, than those on higher wages, they do spend a much greater proportion of their incomes than these groups. The links with general ‘disadvantage’ should also be noted. Research shows that those who experience unemployment, poor health, housing and low educational qualiﬁcations have signiﬁcantly higher rates of problem gambling than the general population (Grifﬁths and Delfabbro, 2001, Grifﬁths, 2006).

Problem Gambling and European Lotteries 133

The American Psychiatric Association (1994) claims that approximately one-third of problem gamblers are women. In the USA this loosely corroborates the results of the BGPS that showed that approximately 1.3 per cent of men and 0.5 per cent of women in Britain could be classiﬁed as problem gamblers (Sproston et al, 2000). Results of the BGPS also showed that the prevalence of problem gambling decreased with age. For instance, the prevalence of problem gambling was 1.7 per cent among people aged between 16 and 24 years, but only 0.1 per cent among the oldest age group. Further, the prevalence was highest among men and women aged between 16 and 24 years (2.3 per cent and 1.1 per cent respectively). The types of games played also impact on the development of gambling problems. This has particular relevance for the lottery sector. It also has consequences for understanding the risk factors involved in the disorder, as well as the demographic proﬁle of those individuals who are most susceptible. For instance, certain features of games are strongly associated with problem gambling. These include games that have a high event frequency (i.e., that are fast and allow for continual staking), that involve an element of skill or perceived skill, and that create ‘near misses’ (i.e., the illusion of having almost won) (Grifﬁths, 1999). Size of jackpot and stakes, probability of winning (or perceived probability of winning), and the possibility of using credit to play are also associated with higher levels of problematic play (Parke and Grifﬁths, 2006). Games that meet these criteria include electronic gaming machines (EGMs) and casino table games. In general, lottery products are not associated with problem gambling with the exception of those that have the potential for continuous gambling (e.g., video lottery terminals (VLTs), scratchcards, and some instant win games on the Internet). According to the BGPS, the most problematic type of gambling in Britain is associated with games in a casino (8.7 per cent of people who gambled on this activity in the past year were problem gamblers according to the SOGS, and 5.6 per cent according to the DSM-IV). Groups most likely to experience problems with casino-based gambling were single, unemployed males, aged under 30 years. Other subgroups include slightly older single males, aged over 40 years, often retired, who are also more likely to be of Chinese ethnicity (Sproston et al, 2000) and adolescent males who have problems particularly with slot machines (Grifﬁths, 1995, 2002a). The problem of adolescent gambling will be examined in more detail below. The BGPS also indicated that other types of gambling activity were engaged in by problem gamblers. These included betting on events

134 Gaming in the New Market Environment

with a bookmaker (SOGS 8.1 per cent; DSM-IV 5.8 per cent), and betting on dog races (SOGS 7.2 per cent; DSM-IV 3.7 per cent). Problem gamblers were less likely to participate in the National Lottery Draw (1.2 per cent of people who gambled on this activity in the past year were problem gamblers according to the SOGS; 0.7 according to the DSM-IV), or playing lottery scratchcards (SOGS 1.7 per cent; DSM-IV 1.5 per cent). In addition, problem gambling prevalence was associated with the number of gambling activities undertaken, with the prevalence of problem gambling tending to increase with the number of gambling activities participated in. As noted above, for a large number of people, the National Lottery Draw (i.e., Lotto) was the only gambling activity they engage in, and problem gambling prevalence among people who limit their gambling to activities such as Lotto and lottery scratchcards was very low at 0.1 per cent. As might be expected, prolem gambling was associated with higher expenditure on gambling activities. Internationally, as in almost every other country worldwide, the greatest problems are, to a very considerable degree, associated with non-casino EGMs such as arcade ‘slot machines’ (Grifﬁths, 1999, Parke and Grifﬁths, 2006). It has been found that as EGMs spread, they tend to displace almost every other type of gambling as well as the problems that are associated with them. EGMs are the fastest-growing sector of the gaming economy, currently accounting for some 70 per cent of revenue. Australia’s particularly high rates of problem gambling are almost entirely accounted for by its high density of these non-casino machines. It is likely that Britain’s relatively lower rates of problems associated with EGMs is explained by its current legislative environment, which limits the numbers of machines in what are relatively regulated venues. All this indicates that attention should be focused on EGMs as a source of risk (but also has implications for VLTs which are structurally very similar – and in some cases almost identical – to slot machines). The spread of EGMs also impacts on the demographic groups who experience problems with gambling. Until very recently, such problems were predominantly found in males, but as EGMs proliferate, women are increasingly presenting in greater numbers, so that in some countries (e.g., the USA), the numbers are almost equal. This trend has been described as a ‘feminisation’ of problem gambling (Volberg, 2001). These types of games appear to be particularly attractive to recent migrants, who are also at high risk of developing gambling problems. It has been suggested that ﬁrst generation migrants may not be sufﬁciently

Problem Gambling and European Lotteries 135

socially, culturally or even ﬁnancially adapted to their new environment to protect them from the potential risks of excessive gambling (Productivity Commission, 1999, Shaffer, LaBrie and LaPlante, 2004). Many are therefore vulnerable to the development of problems. Variations in gambling preferences are thought to result from both differences in accessibility and motivation. Older people tend to choose activities that minimise the need for complex decision-making or concentration (e.g., bingo, slot machines), whereas gender differences have been attributed to a number of factors, including variations in sex-role socialisation, cultural differences and theories of motivation (Grifﬁths, 2006). Variations in motivation are also frequently observed among people who participate in the same gambling activity. For example, slot machine players may gamble to win money, for enjoyment and excitement, to socialise and to escape negative feelings (Grifﬁths, 1995). Some people gamble for one reason only, whereas others gamble for a variety of reasons. A further complexity is that people’s motivations for gambling have a strong temporal dimension; that is, they do not remain stable over time. As people progress from social to regular and ﬁnally to excessive gambling, there are often signiﬁcant changes in their reasons for gambling. Whereas a person might have initially gambled to obtain enjoyment, excitement and socialisation, the progression to problem gambling is almost always accompanied by an increased preoccupation with winning money and chasing losses.

6.3

Youth gambling

Adolescent gambling is mainly a cause for concern in the UK rather than other European countries, and is related to other delinquent behaviours. For instance, in one study of over 4,500 adolescents, lottery gambling was highly correlated with other potentially addictive activities such as illicit drug taking and alcohol abuse (Grifﬁths and Sutherland, 1998). Another study by Yeoman and Grifﬁths (1996) demonstrated that around 4 per cent of all juvenile crimes in one UK city was slot machinerelated based on over 1,850 arrests in a one-year period. It has also been noted that adolescents may be more susceptible to problem gambling than adults. For instance, in the UK, a number of studies have consistently highlighted a ﬁgure of up to 5 to 6 per cent of problem gamblers among adolescent slot machine gamblers (see Grifﬁths, 2002a, 2003b, for an overview of these studies) with a very small proportion including those who gamble on lottery scratchcards. This ﬁgure is at least two to three times higher than that identiﬁed in adult populations. On this

136 Gaming in the New Market Environment

evidence, young people are clearly more vulnerable to the negative consequences of gambling than adults. A typical ﬁnding of many adolescent gambling studies has been that problem gambling appears to be a primarily male phenomenon. It also appears that adults may to some extent be fostering adolescent gambling. For example, a strong correlation has been found between adolescent gambling and parental gambling (Wood and Grifﬁths, 1998, Wood, 2004). This is particularly worrying because a number of studies have shown that individuals who gamble as adolescents, are then more likely to become problem gamblers as adults (Grifﬁths, 2003b). Similarly, many studies have indicated a strong link between adult problem gamblers and later problem gambling among their children (Grifﬁths, 2003b). Other factors that have been linked with adolescent problem gambling include working class youth culture, delinquency, alcohol and substance abuse, poor school performance, theft and truancy (e.g., Grifﬁths, 1995, Yeoman and Grifﬁths, 1996, Grifﬁths and Sutherland, 1998). The main form of problem gambling among adolescents has been the playing of slot machines although a number of studies have also highlight lottery scratchcards to be problematic to a small minority. There is little doubt that slot machines are potentially ‘addictive’ and there is now a large body of research worldwide supporting this. Most research on slot machine gambling in youth has been undertaken in the UK where they are legally available to children of any age. The most recent wave of the UK tracking study carried out by MORI and the International Gaming Research Unit (2006) found that slot machines were the most popular form of adolescent gambling with 54 per cent of their sample of 8,017 adolescent participants. A more thorough examination of the literature summarising over 30 UK studies (Grifﬁths, 2003b) indicates that: • At least two-thirds of adolescents play slot machines at some point during adolescence • One-third of adolescents will have played slot machines in the last month • That 10 per cent to 20 per cent of adolescents are regular slot machine players (playing at least once a week) (17 per cent in the latest 2006 MORI/IGRU survey) • That between 3 per cent and 6 per cent of adolescents are probable pathological gamblers and/or have severe gambling-related difﬁculties (3.5 per cent down from 4.9 per cent in the latest 2006 MORI/IGRU survey).

Problem Gambling and European Lotteries 137

All studies have reported that boys play on slot machines more than girls and that as slot machine playing becomes more regular it is more likely to be a predominantly male activity. Research has also indicated that very few female adolescents have gambling problems on slot machines. Research suggests that irregular (‘social’) gamblers play for different reasons than the excessive (‘pathological’) gamblers. Social gamblers usually play for fun and entertainment (as a form of play), because their friends or parents do (i.e., it is a social activity), for the possibility of winning money, because it provides a challenge, because of ease of availability and there is little else to do, and/or for excitement (the ‘buzz’). Pathological gamblers appear to play for other reasons such as mood modiﬁcation and as a means of escape. As already highlighted, young males seem to be particularly susceptible to slot machine addiction with a small but signiﬁcant minority of adolescents in the UK experiencing problems with their slot machine playing at any one time. Like other potentially addictive behaviours, slot machine addiction causes the individual to engage in negative behaviours. This includes truanting in order to play the machines, stealing to fund machine playing, getting into trouble with teachers and/or parents over their machine playing, borrowing or the using of lunch money to play the machines, poor schoolwork, and in some cases aggressive behaviour (Grifﬁths, 2003b). These behaviours are not much different from those experienced by other types of adolescent problem gambling. In addition, slot machine addicts also display bona ﬁde signs of addiction including withdrawal effects, tolerance, mood modiﬁcation, conﬂict and relapse. It is clear that for some adolescents, gambling can cause many negative detrimental effects in their life. Education can be severely affected and they may have a criminal record as most problem gamblers have to resort to illegal behaviour to feed their addiction. Gambling is an adult activity and the government should consider legislation that restricts gambling to adults only.

6.4

Pathological features

Though many people engage in gambling as a form of recreation and enjoyment, or even as a means to gain an income, for some, gambling is associated with difﬁculties of varying severity and duration. Some regular gamblers persist in gambling even after repeated losses and develop signiﬁcant, debilitating problems that typically result in harm to others close to them and in the wider community (Abbott and Volberg, 1999).

138 Gaming in the New Market Environment

In 1980, pathological gambling was recognised as a mental disorder in the third edition of the Diagnostic and Statistical Manual (DSM-III) under the section ‘Disorders of Impulse Control’ along with other illnesses such as kleptomania and pyromania (American Psychiatric Association, 1980). Adopting a medical model of pathological gambling in this way displaced the old image that the gambler was a sinner or a criminal. In diagnosing the pathological gambler, the DSM-III stated that the individual was chronically and progressively unable to resist impulses to gamble and that gambling compromised, disrupted or damaged family, personal, and vocational pursuits. The behaviour increased under times of stress and associated features included lying to obtain money, committing crimes (e.g., forgery, embezzlement or fraud), and concealment from others of the extent of the individual’s gambling activities. In addition, the DSM-III stated that to be a pathological gambler, the gambling must not be due to antisocial personality disorder. These criteria were criticised for (i) a middle class bias, i.e., the criminal offences like embezzlement, income tax evasion were ‘middle class’ offences, (ii) lack of recognition that many compulsive gamblers are self-employed and (iii) exclusion of individuals with antisocial personality disorder (Lesieur, 1988). Lesieur recommended the same custom be followed for pathological gamblers as for substance abusers and alcoholics in the past (i.e., allow for simultaneous diagnosis with no exclusions). The new criteria (DSM-III-R, American Psychiatric Association, 1987) were subsequently changed taking on board the criticisms and modelled extensively on substance abuse disorders due to the growing acceptance of gambling as a bona ﬁde addictive behaviour. In 1989 however, Rosenthal conducted an analysis of the use of the DSM-III-R criteria by treatment professionals. It was reported that there was some dissatisfaction with the new criteria and that there was some preference for a compromise between the DSM-III and the DSM-III-R. As a consequence, the criteria were changed for DSM-IV. The updated DSM-IV consists of 10 diagnostic criteria (see Appendix 1). A ‘problem gambler’ is diagnosed when three or more of criteria A1–A10 are met, and a score of ﬁve or more indicates a ‘probable pathological gambler.’ The diagnosis is not made if the gambling behaviour is better accounted for by a manic episode (criterion B) (American Psychiatric Association, 1994). Problems with gambling may also occur in individuals with antisocial personality disorder and it is possible for an individual to be diagnosed with both pathological gambling and manic episode gambling behaviour if criteria for both disorders are met (American Psychiatric Association, 1994).

Problem Gambling and European Lotteries 139

According to the American Psychiatric Association (1994) DSM-IV ‘Pathological gambling typically begins in early adolescence in males and later in life in females. Although a few individuals are ‘hooked’ with their very ﬁrst bet, for most the course is more insidious. There may be years of social gambling followed by an abrupt onset that may be precipitated by greater exposure to gambling or by a stressor. The gambling pattern may be regular or episodic, and the course of the disorder is typically chronic. There is generally a progression in the frequency of gambling, the amount wagered, and the preoccupation with gambling and obtaining money with which to gamble. The urge to gamble and gambling activity generally increase during periods of stress or depression’ (p. 617). SOGS is based on the DSM-III criteria for pathological gambling and is at present the most widely used screen instrument for problem gambling used internationally. It consists of 20 questions on gambling behaviour from which a total score (ranging from 0 to 20) of positive responses is calculated. A score of three to four indicates a ‘problem gambler’ and ﬁve or more indicates a ‘probable pathological gambler’ (see Appendix 2).

6.5

Internet and remote gambling

A recent report published by the Department of Culture, Media and Sport (2006) noted that online gambling had more than doubled in the UK since 2001. Worldwide there are around 2,300 sites with a large number of these located in just a few particular countries. For instance, around 1,000 sites are based in Antigua and Costa Rica alone. The UK has about 70 betting and lottery sites but as yet no gaming sites (e.g., online casinos featuring poker, blackjack, roulette, etc.). The ﬁndings reported that there were approximately one million regular online gamblers in Britain alone making up nearly one-third of Europe’s 3.3 million regular online gamblers. It was also reported that women were becoming increasingly important in the remote gambling market. For instance, during the 2006 World Cup, it was estimated that about 30 per cent of those visiting key UK based betting websites were women. The report also reported that Europe’s regular online gamblers staked approximately £3.5 billion pounds a year at around an average of £1,000 each. It was also predicted that mobile phone gambling was also likely to grow, further increasing accessibility to remote gambling. The introduction of the internet and other remote gambling developments (such as mobile phone gambling, interactive television gambling)

140 Gaming in the New Market Environment

has the potential to lead to problematic gambling behaviour and is likely to be an issue over the next decade. Remote gambling presents what could be the biggest cultural shift in gambling and one of the biggest challenges concerning the psychosocial impact of gambling. To date, there has been little empirical research examining remote gambling in the UK. The one and only prevalence survey was published in 2001 (from data collected in 1999) when internet gambling was almost non-existent (Grifﬁths, 2001b). Many gamblers however, are technologically proﬁcient and use the internet and mobile phones regularly. To date, knowledge and understanding of how the internet, mobile phones and interactive television affect gambling behaviour is sparse. Globally speaking, proliferation of internet access is still an emerging trend and it will take some time before the effects on gambling behaviour surface (on both adults and young people). However, there is strong foundation to speculate on the potential hazards of remote gambling. These include the use of virtual cash, unlimited accessibility, and the solitary nature of gambling on the internet as potential risk factors for problem gambling development (Grifﬁths and Parke, 2002, Grifﬁths, 2003c, 2005, Grifﬁths, Parke, Wood and Parke, 2006). There is no conclusive evidence that internet gambling is associated with problem gambling although very recent studies using self-selected samples suggest that the prevalence of problem gambling among internet gamblers is relatively high (Grifﬁths and Barnes, 2008, Wood, Grifﬁths and Parke, 2007). What is clear, however, is that online gambling has strong potential to facilitate, or even encourage, problematic gambling behaviour (Grifﬁths, 2003c). Firstly, the 24-hour availability of Internet gambling (and other remote forms) allows a person to potentially gamble non-stop (Grifﬁths, 1999). The privacy and anonymity offered by internet gambling enables problem gamblers to continue gambling without being ‘checked’ by gambling venue staff concerned about behaviour or amount of time spent gambling (Grifﬁths et al, 2006). Friends and family may also be oblivious to the amount of time an individual spends gambling online. In addition, the use of electronic cash may serve to distance a gambler from how much money he or she is spending, in a similar way that chips and tokens used in other gambling situations may allow a gambler to ‘suspend judgement’ with regard to money spent (Grifﬁths and Parke, 2002). There are a number of factors that make online activities, such as internet gambling, potentially seductive and/or addictive including anonymity, convenience, escape, accessibility, event frequency, inter-

Problem Gambling and European Lotteries 141

activity, short-term comfort, excitement and loss of inhibitions (Grifﬁths, 2003c, Grifﬁths et al, 2006). Further, there are many other speciﬁc developments that look likely to facilitate uptake of remote gambling services including (i) sophisticated gaming software, (ii) integrated e-cash systems (including multi-currency), (iii) multilingual sites, (iv) increased realism (e.g., ‘real’ gambling via webcams), (v) live remote wagering (for both gambling alone and gambling with others), and (vi) improving customer care systems (Grifﬁths, 2003c). To a gambling addict, the internet could potentially be a very dangerous medium. For instance, it has been speculated that structural characteristics of the software itself might promote addictive tendencies. Structural characteristics promote interactivity and to some extent deﬁne alternative realities to the user and allow them feelings of anonymity – features that may be very psychologically rewarding to some individuals. There is no doubt that internet usage among the general population will continue to increase over the next few years. Despite evidence that both gambling and the internet can be potentially addictive, there is no evidence (to date) that internet gambling is ‘doubly addictive’ particularly as the internet appears to be just a medium to engage in the behaviour of choice. What the internet may do is facilitate social gamblers who use the internet (rather than Internet users per se) to gamble more excessively than they would have done ofﬂine (Grifﬁths, 2003c, Grifﬁths et al, 2006). In addition, a recent survey of British internet gambling sites showed very low levels of social responsibility (Smeaton and Grifﬁths, 2004). Technological advance in the form of remote gambling is providing ‘convenience gambling’. Theoretically, people can gamble all day, every day of the year. This will have implications for the social impact of internet gambling. There are a number of social issues concerning internet gambling. Some of the major concerns are brieﬂy described below and adapted from Grifﬁths and Parke (2002). Gate-keeping and protection of the vulnerable: There are many groups of vulnerable individuals (e.g., young people, problem gamblers, drug/alcohol abusers, the learning impaired) who in ofﬂine gambling would be prevented from gambling by responsible members of the gaming industry. Remote gambling operators however, provide little in the way of ‘gatekeeping’. In cyberspace, how can you be sure that young people do not have access to internet gambling by using a parent’s credit card? How can you be sure that a young person does not have access to internet gambling while they are under the inﬂuence of alcohol or other intoxicating substances? How can you prevent a young problem

142 Gaming in the New Market Environment

gambler who may have been barred from one internet gambling site, simply clicking to the next internet gambling link? Electronic cash: For most gamblers, it is very likely that the psychological value of electronic cash (e-cash) will be less than ‘real’ cash (and similar to the use of chips or tokens in other gambling situations). Gambling with e-cash may lead to a ‘suspension of judgment’. The ‘suspension of judgment’ refers to a structural characteristic that temporarily disrupts the gambler’s ﬁnancial value system and potentially stimulates further gambling. This is well known by both those in commerce (i.e., people typically spend more on credit and debit cards because it is easier to spend money using plastic) and by the gaming industry. This is the reason that ‘chips’ are used in casinos and why tokens are used on some slot machines. In essence, chips and tokens ‘disguise’ the money’s true value (i.e., decrease the psychological value of the money to be gambled). Tokens and chips are often re-gambled without hesitation as the psychological value is much less than the real value. Increased odds of winning in practice modes: One of the most common ways that gamblers can be facilitated to gamble online is when they try out games in the ‘demo’, ‘practice’ or ‘free play’ mode. Further, there are no restrictions preventing children and young people playing (and learning how to gamble) on these practice and demonstration modes. Recent research (Sevigny et al, 2005) showed that it was signiﬁcantly more commonplace to win while ‘gambling’ on the ﬁrst few goes on a ‘demo’ or ‘free play’ game. They also reported that it was commonplace for gamblers to have extended winning streaks during prolonged periods while playing in the ‘demo’ modes. Obviously, once gamblers start to play for real with real money, the odds of winning are considerably reduced. This has some serious implications for young people’s potential gambling behaviour. Online customer tracking: Perhaps the most worrying concerns over remote gambling is the way operators can collect other sorts of data about the gambler. Remote gamblers can provide tracking data that can be used to compile customer proﬁles. When signing up for remote gambling services, players supply lots of information including name, address, telephone number, date of birth, and gender. Remote gambling service providers will know a player’s favourite game and the amounts that they have wagered. Basically they can track the playing patterns of any gambler. They will know more about the gambler’s playing behaviour than the gamblers themselves. They will be able to send the gambler offers and redemption vouchers, complimentary accounts, etc. The industry claims all of these things are introduced to enhance customer exper-

Problem Gambling and European Lotteries 143

ience. More unscrupulous operators however, will be able to entice known problem gamblers back on to their premises with tailored freebies (such as the inducement of ‘free’ bets in the case of remote gambling). Given the brief outline above, remote gambling could easily become a medium for problematic gambling behaviour. Even if numbers of problem remote gamblers are small (and they by no means necessarily are), remote gambling remains a matter of concern. Remote gambling is a relatively new phenomenon and is likely to continue expanding in the near future. It is therefore crucial that the new legislation does nothing to facilitate the creation or escalation of problems in relation to remote gambling. The recent decision by the US to ban internet gambling by making it illegal to pay with debit and credit cards is likely to drive the problem internet gambling ‘underground’ and result in even less protection for vulnerable gamblers. New innovative ways of paying electronically for internet gambling will emerge and the prohibitive stance taken by the US is likely to have little long-lasting protective effect.

6.6

Consequences and co-morbidities

Problem gambling is often co-morbid with other behavioural and psychological disorders, which can exacerbate, or be exacerbated by, problem gambling. Some of the psychological difﬁculties a problem gambler may experience include anxiety, depression, guilt, suicidal ideation and actual suicide attempts (Daghestani et al, 1996, Grifﬁths, 2004). Problem gamblers may also suffer irrational distortions in their thinking (e.g., denial, superstitions, overconﬁdence, or a sense of power or control) (Grifﬁths, 1994a). Increased rates of attention-deﬁcit hyperactivity disorder (ADHD), substance abuse or dependence, antisocial, narcissistic, and borderline personality disorders have also been reported in pathological gamblers (APA, 1994, Grifﬁths, 1994b). There is also some evidence that co-morbidities may differ among demographic subgroups and gambling types. For instance, young male slot machine gamblers are more likely to abuse solvents (Grifﬁths, 1994c). There is frequently a link with alcohol or drugs as a way of coping with anxiety or depression caused by gambling problems, and, conversely, alcohol may trigger the desire to gamble (Grifﬁths, Parke and Wood, 2002). According to the DSM IV, pathological gamblers tend to be highly competitive, energetic, restless, easily bored, and believe money is the cause of, and solution to, all their problems (see also Parke,

144 Gaming in the New Market Environment

Grifﬁths and Irwing, 2004). According to the American Psychiatric Association, pathological gamblers may also be overly concerned with the approval of others and may be extravagantly generous. Further, when not gambling, they may be workaholics or ‘binge’ workers who wait until they are up against deadlines before really working hard. Pathological gamblers may also be prone to stress-related physical illnesses including insomnia, hypertension, heart disease, stomach problems (e.g., peptic ulcer disease) and migraine (Daghestani et al, 1996, Abbot and Volberg, 2000, Grifﬁths, Scarfe and Bellringer, 1999, Grifﬁths, 2004). Like other addictive behaviours, while engaged in gambling, the body produces increased levels of endorphins (the body’s own morphine like substance), and other ‘feel good’ chemicals like noradrenaline and seretonin (Grifﬁths, 2006). Many of these physical negative effects may stem from the body’s own neuro-adaptation processes. Health-related problems due to problem gambling can also result from withdrawal effects. Rosenthal and Lesieur (1992) found that at least 65 per cent of problem gamblers reported at least one physical side-effect during withdrawal including insomnia, headaches, upset stomach, loss of appetite, physical weakness, heart racing, muscle aches, breathing difﬁculty and/or chills. Their results were also compared to the withdrawal effects from a substance-dependent control group. They concluded that problem gamblers experienced more physical withdrawal effects when attempting to stop than the substance-dependent group. Interpersonal problems suffered by problem gamblers include conﬂict with family, friends and colleagues, and breakdown of relationships, often culminating in separation or divorce (Grifﬁths, 2004, 2006). The children of problem gamblers also suffer a range of problems, and tend to do less well at school (Jacobs, Marston, Singer et al, 1989, Lesieur and Rothschild, 1989). School- and work-related problems include poor work performance, abuse of leave time and job loss (Grifﬁths, 2002b). Financial consequences include reliance on family and friends, substantial debt, unpaid creditors and bankruptcy (Grifﬁths, 2006). Finally, there may be legal problems as a result of criminal behaviour undertaken to obtain money to gamble or pay gambling debts (Grifﬁths, 2005b, 2006). The families of problem gamblers can also experience substantial physical and psychological difﬁculties (Grifﬁths and Delfabbro, 2001, Grifﬁths, 2006).

6.7

Structural characteristics in gambling

Gambling is a multifaceted rather than unitary phenomenon. Consequently, many factors may come into play in various ways and at dif-

Problem Gambling and European Lotteries 145

ferent levels of analysis (e.g., biological, social or psychological). Theories may be complementary rather than mutually exclusive, which suggests that limitations of individual theories might be overcome through the combination of ideas from different perspectives. This has often been discussed in terms of recommendations for an ‘eclectic’ approach to gambling or a distinction between proximal and distal inﬂuences upon gambling (Walker, 1992). For the most part however, such discussions have been descriptive rather than analytical, and so far, few attempts have been made to explain why an adherence to a singular perspective is untenable. Put very simply, there are many different factors involved in how and why people develop gambling problems. Central to the latest thinking is that no single level of analysis is considered sufﬁcient to explain either the aetiology or maintenance of gambling behaviour. Moreover, this view asserts that all research is context-bound and should be analysed from a combined, or biopsychosocial, perspective (Grifﬁths, 2005c). Variations in the motivations and characteristics of gamblers and in gambling activities themselves mean that ﬁndings obtained in one context are unlikely to be relevant or valid in another. Another factor central to understanding gambling behaviour is the structure of gambling activities. Grifﬁths (1993, 1995, 1999) has shown that gambling activities vary considerably in their structural characteristics, such as the probability of winning, the amount of gambler involvement, the use of the near wins, the amount of skill that can be applied, the length of the interval between stake and outcome and the magnitude of potential winnings. Structural variations are also observed within certain classes of activities such as slot machines, where differences in reinforcement frequency, colours, sound effects and machines’ features can inﬂuence the proﬁtability and attractiveness of machines signiﬁcantly (Grifﬁths and Parke, 2003, Parke and Grifﬁths, 2007, in press). Each of these structural features may (and almost certainly does) have implications for gamblers’ motivations and the potential ‘addictiveness’ of gambling activities. For example, skilful activities that offer players the opportunity to use complex systems, study the odds and apply skill and concentration appeal to many gamblers because their actions can inﬂuence the outcomes. Such characteristics attract people who enjoy a challenge when gambling. They may also contribute to excessive gambling if people overestimate the effectiveness of their gambling systems and strategies. Chantal and Vallerand (1996) have argued that people who gamble on these activities (e.g., racing punters) tend to be more intrinsically

146 Gaming in the New Market Environment

motivated than lottery gamblers in that they gamble for selfdetermination (i.e., to display their competence and to improve their performance). People who gamble on chance activities, such as lotteries, usually do so for external reasons (i.e., to win money or escape from problems). This ﬁnding was conﬁrmed by Loughnan, Pierce and Sagris (1997) in their clinical survey of problem gamblers. Here, racing punters emphasised the importance of skill and control considerably more than slot machine players. Although many slot machine players also overestimate the amount of skill involved in their gambling, other motivational factors (such as the desire to escape worries or to relax) tend to predominate. Thus, excessive gambling on slot machines may be more likely to result from people becoming conditioned to the tranquilising effect brought about by playing rather than just the pursuit of money. Another vital structural characteristic of gambling is the continuity of the activity; namely, the length of the interval between stake and outcome. In nearly all studies, it has been found that continuous activities (e.g., racing, slot machines, casino games) with a more rapid playrate are more likely to be associated with gambling problems (Grifﬁths, 1999). The ability to make repeated stakes in short time intervals increases the amount of money that can be lost and also increases the likelihood that gamblers will be unable to control spending. Such problems are rarely observed in non-continuous activities, such as weekly or bi-weekly lotteries, in which gambling is undertaken less frequently and where outcomes are often unknown for days. Consequently, it is important to recognise that the overall social and economic impact of expansion of the gambling industry will be considerably greater if the expanded activities are continuous rather than non-continuous.

6.8

Situational characteristics in gambling

Other factors central to understanding gambling behaviour are the situational characteristics of gambling activities. These are the factors that often facilitate and encourage people to gamble in the ﬁrst place (Grifﬁths and Parke, 2003). Situational characteristics are primarily features of the environment (e.g., accessibility factors such as location of the gambling venue, the number of venues in a speciﬁed area, possible membership requirements, etc.) but can also include internal features of the venue itself (décor, heating, lighting, colour, background music, ﬂoor layout, refreshment facilities, etc.) or facilitating factors that may inﬂuence gambling in the ﬁrst place (e.g., advertising, free travel and/or accom-

Problem Gambling and European Lotteries 147

modation to the gambling venue, free bets or gambles on particular games, etc.) or inﬂuence continued gambling (e.g., the placing of a cash dispenser on the casino ﬂoor, free food and/or alcoholic drinks while gambling, etc.) (Grifﬁths and Parke, 2003, Abbott and Volberg, 2007). These variables may be important in both the initial decision to gamble and the maintenance of the behaviour. Although many of these situational characteristics are thought to inﬂuence vulnerable gamblers, there has been very little empirical research into these factors and more research is needed before any deﬁnitive conclusions can be made about the direct or indirect inﬂuence on gambling behaviour and whether vulnerable individuals are any more likely to be inﬂuenced by these particular types of marketing ploys. The introduction of super-casinos into the UK will almost certainly see an increase in these types of situational marketing strategies and should also provide an opportunity to research and monitor the potential psychosocial impact.

6.9

Policy implications

In this ﬁnal section, policy implications for problem gambling are examined with particular emphasis on lotteries. One of the most obvious questions to be answered in this context is how governments can most effectively prevent problem gambling. Media commentators often call for such actions as lowering the jackpots, restricting the frequency/opening hours, restricting the amount money that can be gambled at the outset, and imposing age limits. Some of these options are examined in more detail below particularly those where there appears to be some evidence. However, there are some of these that are likely to have a mixed effect and decisions have to be taken on the number of people that the measure is likely to affect. For instance, lowering the jackpots will signiﬁcantly reduce the amount of people who may gamble in the ﬁrst place – particularly on lottery type activities. However, as we have seen, problem and habitual gambling often relies on the unpredictability of the rewards rather than factors like stake size or jackpot size. Having said that, a large jackpot size on activities like playing slot machines or video lottery terminals may facilitate chasing behaviour (a known risk factor in problem gambling). In the absence of empirical research and/or a cost-beneﬁt analysis, policy recommendations in the area of jackpot size are very difﬁcult to make. Another complicating factor is that gambling behaviour is not homogenous and there is a different psychology of gambling to almost all activities. Furthermore, there are some forms of gambling that may

148 Gaming in the New Market Environment

actually be proﬁtable to gamblers but cause huge problems because of the impact in other areas of their life. For instance, Wood, Grifﬁths and Parke (2007) in a study of online poker players argued that a new breed of problem gambler was emerging. These were gamblers who actually won more money than they lost but may be spending disproportionate amounts of their day gambling that ended up compromising their relationships, work, and social life. For these problem players, the loss is not monetary but time. There is also the issue of the changing market environment. Over the last few years there has been two major shifts across the gambling sector. Firstly, there has been an increase in remote forms of gambling. Almost all sectors that can be played remotely are being played remotely. Secondly, games are getting faster and more instant. The lottery sector has seen a shift from weekly and bi-weekly games to more high frequency, instant win games through scratchcards, video lottery terminals and online instant win games. What might be the most important feature in the change in the market environment? The combined effect of this needs to be monitored and is likely to increase problem gambling rates unless social responsibility safeguards are put in place. The simple rule of thumb is that more money, more games, more options (in both time and space) all increase gambling participation. Although gambling is clearly of policy interest it has not been traditionally viewed as a public health matter (Grifﬁths, 1996, Korn, 2000). Furthermore, research into the health, social and economic impacts of gambling are still at an early stage. There are many speciﬁc reasons why gambling should be viewed as a public health and social policy issue – particularly given the massive expansion of gambling opportunities across the world. The following provides some recommendations to consider relating to policy initiatives. Research: Understanding problem gambling is seriously hindered by a lack of high quality data. It is important to expand the research base on the causes, progression, distribution and treatment of gambling problems in all areas including the lottery sector. One way to begin tackling the problem could be to link up with overseas networks and researchers in order to pool knowledge and expertise. In short there should be: • Regular surveys of problem gambling services, including helplines and formal treatment providers, and evaluations of the effectiveness and efﬁcacy of these services. • Research into the efﬁcacy of various approaches to the treatment of gambling addiction needs to be undertaken.

Problem Gambling and European Lotteries 149

• Research into the association of Internet gambling and problem gambling. • Research into the impacts of gambling, including health, family, workplace, ﬁnancial and legal impacts. • Longitudinal research into problem gambling, treatment, and the impact of gambling legislation on prevalence of problem gambling. In particular, why some people develop problems and, just as importantly, why the majority do not develop problems. Legislation – Limit the opportunities and accessibility to gamble: There is little doubt that opportunities and accessibility to gamble will increase as a result of increasing deregulation across Europe. Underpinning this recommendation is psychological research into the ‘availability hypothesis’ (Orford, 2002). What has generally been demonstrated from research evidence in other countries is that where accessibility of gambling is increased there is an increase not only in the number of regular gamblers but also an increase in the number of problem gamblers (Grifﬁths, 1999) and supports the availability hypothesis. This obviously means that not everyone is susceptible to developing gambling addictions but it does mean that at a societal (rather than individual) level, the more gambling opportunities, the more problems. Therefore, number of outlets and opportunities could be capped (such as putting a cap on the size and number of casinos nationally). Particular psychological concern must be given to gambling in new media (e.g., Internet, interactive television, and mobile phone gambling) that may affect individuals in different ways. Legislation – Raise the minimum age of all forms of commercial gambling to 18 years: A public-health approach to gambling-related harm adopts a broader conception of the causes of gambling-related problems. Traditional approaches tend to focus on the characteristics that pre-dispose some gamblers to develop problems, whereas a public health approach focuses on the characteristics of the environment that encourages excessive gambling (e.g., advertising, time restrictions etc.). The single most important measure would be to raise the legal age of gambling. This would signiﬁcantly reduce the age at which children start to gamble and would also help gaming operators and shopkeepers prevent underage gambling. Research has consistently shown that the younger a person starts to gamble, the more likely they are to develop problems (Grifﬁths, 2002a). Furthermore, gambling, like other addictions involving alcohol and illicit drug use, are ‘disorders of youthful onset’ (Teeson, Degenhardt and Hall, 2002). At present, many young adolescents as young as 11 and 12 years of age can pass for being sixteen.

150 Gaming in the New Market Environment

An age rise to 18 years would stop a lot of the very young adolescents gambling in the ﬁrst place. Young problem gamblers who think that they can make their living with gambling neglect education and other productive investments and this may become a burden to the state later in life. Education – Raise awareness about gambling among health practitioners and the general public: There is an urgent need to enhance awareness within the medical and health professions, and the general public about gambling-related problems (Grifﬁths and Wood, 2000, Korn, 2000). The lack of popular and political support for policies that increase price or reduce availability has encouraged other approaches such as public education. Problem gambling is very much the ‘hidden’ addiction. Unlike (say) alcoholism, there is no slurred speech and no stumbling into work. Furthermore, overt signs of problems often do not occur until late in the pathological gambler’s career. When it is considered that problem gambling can be an addiction that can destroy families and have medical consequences, it becomes clear that health professionals and the public should be aware of the effects. General practitioners routinely ask patients about smoking and drinking but gambling is something that is not generally discussed (Setness, 1997). Problem gambling may be perceived as a somewhat ‘grey’ area in the ﬁeld of health and it is therefore is very easy to deny that health professionals should be playing a role. Prevention: Set up both general and targeted gambling prevention initiatives: There is growing prevention and intervention initiatives. According to Korn (2002), the goals of gambling intervention are to (i) prevent gambling-related problems, (ii) promote informed, balanced attitudes and choices, and (iii) protect vulnerable groups. The guiding principles for action on gambling are therefore prevention, health promotion, harm reduction, and personal and social responsibility. Throughout the world there are many actions and initiatives that are carried out as preventative measures in relation to gambling. The most common examples of these include: general awareness raising (e.g., public education campaigns through advertisements on television, radio, newspapers); targeted prevention (e.g., targeted education programmes and campaigns for particular vulnerable populations such as senior citizens, adolescents, ethnic minorities, etc.); awareness raising within gambling establishments (e.g., brochures and leaﬂets describing problem gambling, indicative warning signs, where help for problems can be sought etc.); training materials (e.g., training videos about problem gambling shown in schools, job centres etc.); training of gambling industry personnel (e.g., training managers of gambling establishments, and those who

Problem Gambling and European Lotteries 151

actually have interaction with gamblers such as croupiers); and internet prevention (e.g., the development, maintenance and linking of problem gambling websites). Education and prevention programmes should also be targeted at children and adolescents along with other potentially addictive and harmful behaviours (e.g., smoking, drinking, and drug taking). More speciﬁcally, gambling operators and service providers should: • Supply information on gambling addiction, treatment and services to patrons. • Support development of centralised training for gambling venue staff to ensure uniform standards and accreditation. • Pay to fund research, prevention, intervention, and treatment programmes. The cliché that ‘prevention is better than cure’ is probably accurate in the case of problem gambling. Although there is some success in treating problem gamblers, it is probably more cost-effective to prevent people developing problems in the ﬁrst place, which is why education and prevention are so important. However, such programmes need to be evaluated to make sure that they themselves are cost-effective. Treatment – Introduce gambling support and treatment initiatives: In addition to the preventative measures outlined above, there are many support initiatives that could also be introduced. Although we do not at present know the rate of return from investing in the treatment of problem gambling, treatment still needs to be offered for those needing help. Such initiatives include: • The running of problem gambling helplines as a referral service. • The running of telephone counselling for problem gamblers and those close to them. • The running of web-based chat rooms and online counselling for problem gamblers and those close to them. • The funding of outpatient treatment. • The funding of in-patient and residential treatment. • Training for problem gambling counsellors (volunteers or professionals; face-to-face, telephone and/or online). • Certiﬁcation of problem gambling counsellors. The intervention options for the treatment of problem gambling include, but are not limited to: counselling, psychotherapy, cognitive-behavioural

152 Gaming in the New Market Environment

therapy (CBT), advisory services, residential care, pharmacotherapy, and/or combinations of these (i.e., multi-modal treatment) (see Grifﬁths, 1996, Grifﬁths and MacDonald, 1999, Grifﬁths and Delfabbro, 2001, Grifﬁths, Bellringer, Farrell-Roberts and Freestone, 2001, Hayer et al, 2005). There is also a very recent move towards using the internet as a route for guidance, counselling and treatment (see Grifﬁths and Cooper, 2003, Grifﬁths, 2005d, Wood and Grifﬁths, 2007). Treatment and support is provided from a range of different people (with and without formal medical qualiﬁcations), including specialist addiction nurses, counsellors, medics, psychologists, and psychiatrists. There are also websites and helplines to access information (e.g., GamCare) or discuss gambling problems anonymously (e.g., GamAid), and local support groups where problem gamblers can meet other people with similar experiences (e.g., Gamblers Anonymous). Support is also available for friends and family members of problem gamblers (e.g., Gam Anon). Many private and charitable organisations throughout Europe provide support and advice for people with gambling problems. Some focus exclusively on the help, counselling and treatment of gambling addiction, while others also work to address common addictive behaviours such as alcohol and drug abuse. The method and style of treatment varies between providers and can range from comprehensive holistic approaches to the treatment of gambling addiction (e.g., encouraging ﬁtness, nutrition, alternative therapies and religious counselling), to an abstinence-based approach. Many providers also encourage patients (and sometimes friends and families) to join support groups (e.g., Gamblers Anonymous and GamAnon), while others offer conﬁdential one-to-one counselling and advice. Most are non-proﬁt-making charities to which patients can self-refer and receive free treatment. Independent providers that offer residential treatment to gambling addicts are more likely to charge for their services. Some provide both in-patient treatment and day-patient services, and a decision as to the suitability of a particular intervention is made upon admission. Due to the lack of relevant evaluative research, the efﬁcacy of various forms of treatment intervention is almost impossible to address. Much of the documentation collected by treatment agencies is incomplete or collected in ways that makes comparisons and assessments of efﬁcacy difﬁcult to make. As Abbott et al (2004) have noted, with such a weak knowledge base, little is known about which forms of treatment for problem gambling in the UK are most effective, how they might be

Problem Gambling and European Lotteries 153

improved or who might beneﬁt from them. However, their review did note that individuals who seek help for gambling problems tend to be overwhelmingly male, aged between 18 to 45 years, and whose problems are primarily with on- and off-course betting, and slot machine use. The gaming industry has typically viewed pathological gambling as a rare mental disorder that is predominantly physically and/or psychologically determined. It supports recent ﬁndings that suggest many problem gamblers have transient problems that often self-corrects. However, some gambling providers have taken the initiative to address the issue of gambling addiction within their businesses. Secondary prevention efforts by the gaming industry have included the development and implementation of employee training programmes, mandatory and voluntary exclusion programmes and gambling venue partnerships with practitioners and government agencies to provide information and improved access to formal treatment services. Implementation of secondary prevention efforts by the gaming industry, such as employee training programmes and exclusion programmes, have not always been of the highest quality and compliance has often been uneven. In addition, observations from abroad appear to demonstrate that efforts by the gaming industry to address gambling addiction tend to compete with heavily ﬁnanced gaming industry advertising campaigns that may work directly to counteract their effectiveness (Grifﬁths, 2005e). Social policy – Embed problem gambling in public health policy: It is clear that increased research into problem gambling is being taken seriously by many countries across the world. This needs to be embedded into public health policy and practice (Shaffer and Korn, 2002). Such measures include: • Adoption of strategic goals for gambling to provide a focus for public health action and accountability. These goals include preventing gambling-related problems among individuals and groups at risk for gambling addiction; promoting balanced and informed attitudes, behaviours, and policies toward gambling and gamblers by both individuals and communities; and protecting vulnerable groups from gambling-related harm. • Endorsement of public health principles consisting of three primary principles that can guide and inform decision-making to reduce gambling-related problems. These are ensuring that prevention is a community priority, with the appropriate allocation of resources to

154 Gaming in the New Market Environment

primary, secondary and tertiary prevention initiatives; incorporating a mental health promotion approach that builds community capacity, incorporates a holistic view of mental health, and addresses the needs and aspirations of gamblers, individuals at risk of gambling problems, or those affected by them; and fostering personal and social responsibility for gambling policies and practices. • Adoption of harm reduction strategies directed at minimising the adverse health, social, and economic consequences of gambling behaviour for individuals, families, and communities. These initiatives should include healthy-gambling guidelines for the general public (similar to low-risk drinking guidelines); vehicles for the early identiﬁcation of gambling problems; non-judgemental moderation and abstinence goals for problem gamblers, and surveillance and reporting systems to monitor trends in gambling-related participation and the incidence and burden of gambling-related illnesses. Notes 1 mark.grifﬁ[email protected]

Appendix 1

DSM-IV Diagnostic criteria for Pathological Gambling A. Persistent and recurrent maladaptive gambling behaviour as indicated by ﬁve (or more) of the following: (1) is preoccupied with gambling (e.g., preoccupied with reliving past gambling experiences, handicapping or planning next venture, or thinking of ways to get money with which to gamble) (2) needs to gamble with increasing amounts of money in order to achieve the desired excitement (3) has repeated unsuccessful efforts to control, cut back, or stop gambling (4) is restless or irritable when trying to cut down or stop gambling (5) gambles as a way of escaping from problems or of relieving a dysphoric mood (e.g., feelings of helplessness, guilt, anxiety, depression) (6) after losing money gambling, often returns another day to get even (‘chasing’ one’s losses) (7) lies to family members, therapist, or others to conceal extent of involvement with gambling (8) has committed illegal acts such as forgery, fraud, theft, or embezzlement to ﬁnance gambling (9) has jeopardised or lost a signiﬁcant relationship, job, or educational or career opportunity because of gambling (10) relies on others to provide money to relieve a desperate ﬁnancial situation caused by gambling. B. The gambling behaviour is not better accounted for by a manic episode. Source: American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV), 1994, 615–16 155

Appendix 2

South Oaks Gambling Screen 1. Please indicate which of the following types of gambling you have done in your lifetime. For each type, mark one answer: ‘not at all’, ‘less than once a week’, or ‘once a week or more’. Not at all ___ ___

Less than once a week ___ ___

Once a week or more ___ ___

___

___

___

___

___

___

___

___

___

___

___

___

___ ___

___ ___

___ ___

___

___

___

___

___

___

a. played cards for money b. bet on horses, dogs or other animals (in off-track betting, at the track or with a bookie) c. bet on sports (parley cards, with a bookie, or at jai alai) d. played dice games (including craps, over and under, or other dice games) for money e. went to casino (legal or otherwise) f. played the numbers or bet on lotteries g. played bingo h. played the stock and/or commodities market i. played slot machines, poker machines or other gambling machines j. bowled, shot pool, played golf or played some other game of skill for money 156

Appendix 2 157

2. What is the largest amount of money you have ever gambled with any one day? ___ never have gambled ___ more than $100 up to $1,000 ___ $10 or less ___ more than $1,000 up to $10,000 ___ more than $10 up to $100 ___ more than $10,000 3. Do (did) your parents have a gambling problem? ___ both my father and mother gamble (or gambled) too much ___ my father gambles (or gambled) too much ___ my mother gambles (or gambled) too much ___ neither gambles (or gambled) too much 4. When you gamble, how often do you go back another day to win back money you have lost? ___ never ___ some of the time (less than half the time) I lost ___ most of the time I lost ___ every time I lost 5. Have you ever claimed to be winning money gambling but weren’t really? In fact, you lost? ___ never (or never gamble) ___ yes, less than half the time I lost ___ yes, most of the time 6. Do you feel you have ever had a problem with gambling? ___ no ___ yes, in the past, but not now ___ yes Yes No 7. Did you ever gamble more than you intended? ___ ___ 8. Have people criticised your gambling? ___ ___ 9. Have you ever felt guilty about the way you gamble or what happens when you gamble? ___ ___ 10. Have you ever felt like you would like to stop gambling but didn’t think you could? ___ ___ 11. Have you ever hidden betting slips, lottery tickets, gambling money, or other signs of gambling from your spouse, children, or other important people in you life? ___ ___ 12. Have you ever argued with people you like over how you handle money? ___ ___

158 Gaming in the New Market Environment

13. (If you answered ‘yes’ to question 12): Have money arguments ever centered on your gambling? 14. Have you ever borrowed from someone and notpaid them back as a result of your gambling? 15. Have you ever lost time from work (or school) due to gambling? 16. If you borrowed money to gamble or to pay gambling debts, where did you borrow from? (Check ‘yes’ or ‘no’ for each) a. from household money b. from your spouse c. from other relatives or in-laws d. from banks, loan companies or credit unions e. from credit cards f. from loan sharks (Shylocks) g. your cashed in stocks, bonds or other securities h. you sold personal or family property i. you borrowed on your checking account (passed bad checks) j. you have (had) a credit line with a bookie k. you have (had) a credit line with a casino

___

___

___

___

___

___

___ ___ ___ ___ ___ ___ ___ ___

___ ___ ___ ___ ___ ___ ___ ___

___ ___ ___

___ ___ ___

Scores are determined by adding up the number of questions that show an ‘at risk’ response, indicated as follows. If you answer the questions above with one of the following answers, mark that the space next to that question: Questions 1–3 are not counted. ___ Question 4: most of the time I lost, or every time I lost ___ Question 5: yes, less than half the time I lose, or yes, most of the time ___ Question 6: yes, in the past, but not now, or yes ___ Question 7: yes ___ Question 8: yes ___ Question 9: yes ___ Question 10: yes ___ Question 11: yes Question 12 is not counted ___ Question 13: yes ___ Question 14: yes ___ Question 15: yes

Appendix 2 159

___ Question 16a: yes ___ Question 16b: yes ___ Question 16c: yes ___ Question 16d: yes ___ Question 16e: yes ___ Question 16f: yes ___ Question 16g: yes ___ Question 16h: yes ___ Question 16i: yes Questions 16j and 16k are not counted Total = ________ (20 questions are counted) **3 or 4 = potential pathological gambler (problem gambler) **5 or more = probable pathological gambler

7 The Economics of Scale and Scope in the Lottery Industry Matti Viren1

7.1

Introduction

This paper deals with the market structure of the lottery industry. In explicit terms, our intention is to ﬁnd out whether there are important economies of scale and scope in this industry. The reason why we want to examine this is the question of what will happen in the future when the gambling markets will most likely become more competitive. Increased competition may represent just a continuation of the recent trends of crossborder gaming and de-facto competition on the national markets, but it is also possible that the legal framework, which has set rather tight limits to both domestic and crossborder competition, will change. Right now, both the US and European lottery markets are characterised by the prevalence of state monopolies. In some countries all gambling activities – not only lotteries – belong to a state monopoly, which means that market entry is prohibited by law, whereas in other countries (e.g., the UK), bookmaking and casinos are part of private market economy. In practice, state monopolies do not have complete monopoly power in their jurisdictions. There are many reasons for this, but one of them that deserves to be cited in particular is the internet. What happened if the legal restrictions were abolished? Would we have a perfectly competitive system in the gaming industry? These questions are difﬁcult to answer because we should know, for instance, what happens with the current state monopolies, what type of a tax system would be imposed on the market players and what side conditions the possible licensing would include.2 However, above all, we should know the cost structure of the gambling companies: are there economies or diseconomies of scale in the production of lottery products? Equally, we should know whether there are important economies 160

The Economics of Scale and Scope in the Lottery Industry 161

of scope in the production of lottery products. This is important when considering the effects of different licensing systems: if, for instance, companies were allowed to provide just one game instead of being able to offer the whole gaming menu. We have some indications that most of the costs of these companies are ﬁxed and the marginal costs are very small, suggesting that we could end up with a natural monopoly which would take over the whole lottery market and eliminate both state lotteries and private competitors. In this respect, the future lottery markets would resemble certain IT and communication industries, where it seems that at least small companies have difﬁculties in surviving in the market.3 Deﬁnition of concepts In the case of production of a single homogenous good, economics of scale is sufﬁcient although not a necessary condition for natural monopoly. If instead, there are multiple products, natural monopoly may arise also because of economics of scope. In general, we require that the so-called subadditivity condition holds with total costs. In other words, a natural monopoly is a ﬁrm that can produce two or more products within the same ﬁrm than in two or more ﬁrms with a lower cost. By economies of scale we mean that a ﬁrm’s average costs decline as its output expands. Denoting costs by C and output by q, we have d(C/q)/dq <0. By economies of scope we mean that it is more economical to produce several, products in one ﬁrm rather than in multiple ﬁrms. That in turn implies in the case two products q1 and q2: C(q1,q2) < C(q1,0) + C(0,q2). In the case multiproduct ﬁrms, consider a vector of two products qi = (q1i,q2i). Then we say that a cost function is subadditive if C(∑ q1i, ∑ q2i) < ∑ C(qi) for all possible, say N vectors of the products with the attribute that ∑q1i = q1 and ∑q2i = q2. For more details and illustrations of these deﬁnitions, see e.g. Train (1995) and Joskow (2007). If the (operating) marginal costs go down to zero when output reaches a very high level, the case is trivial. Yet, the costs do not need to be as low as zero to lead us into the natural monopoly solution. This can be illustrated with a simple example of monopoly pricing. Take an example (Figure 7.1) where we have constant elasticity of demand and where marginal costs are a constant proportion of sales. When we compute the proﬁt maximising solution for a monopoly we ﬁnd that the optimal payout ratio – demand elasticity nexus moves along with the level of

162 Gaming in the New Market Environment Effect of operational costs on the payout ratio 0,7

0,6

0,5

0,4

0,3

0,2 MC = 0% MC = 10%

0,1

MC = 20% MC = 30%

0 0,0

0,2

0,4

0,6

0,8

1,0 1,2 Demand elasticity

1,4

1,6

1,8

Figure 7.1 Optimal payout ratio with different demand elasticities and marginal costs

costs. Quite clearly, higher costs lower the payout ratio and, thus, the volume of demand. Even proﬁts will decrease. Thus, all in all, higher marginal costs (will) reduce the scale of operations. The nature of the costs is important not only because it may ultimately dictate the market structure but also because it affects the volume of trade, which is a question of wide public interest4 as such. In order to ﬁnd out whether this is the case, we focus on existing state lotteries both in the US and Europe and examine their cost development during the last few decades. Obviously, we should scrutinise all gambling companies but – for data reasons – we have to limit the present analysis to an analysis of a set of public or semi-public companies. Because the companies are relatively similar, comparative analyses also make sense, even if we cannot fully control the output menu. We do have some data on private gambling companies (bookmakers) that we can illustrate but, because of a very small sample size, we cannot carry out a systematic empirical analysis with these data.5 The European data have been collected from a survey which was made in April/May 2006 among the EL (European State Lotteries and Toto Association) members. The US data come from NASPL data base and from various International Gaming and Wagering Business and La Fleur’s World Lottery Almanac periodicals. The European survey also includes some soft data on growth prospects and an assessment of key elasticities.

The Economics of Scale and Scope in the Lottery Industry 163

The remainder of the paper is organised as follows: In the next section we brieﬂy review the analytical framework. In Section 7.3 we present the empirical results, both for cost and demand functions, and in Section 7.4 we offer a set of concluding remarks.

7.2

Analytical framework

What we are about to do here is simply analyse the cost conditions of the lotto companies. The analysis boils down to estimating the following, relatively simple cost function log(Cit) = a0i + a1 log(Sit) + a2 log(Xit) + uit

(7.1)

where C denotes either (i) the total (total) operating costs (TC), (ii) operating costs minus commissions paid to retailers (OC), or (iii) the commissions paid to retailers (AC). With the European data we have the following cost concepts: TC = total operating costs and PC = personal costs. Obviously, TC = OC + AC. S denotes total sales and X (a vector of) possible control variables. These include GDP per capita (as an indicator of the cost level or productivity), n = the number of games offered, t0 the start date of the lottery company’s operations and dens = population density as a possible determinant of retailer costs. uit denotes the error term which may differ over countries and states. To estimate costs functions, we ought to have data on input prices. Unfortunately these data are not easily available. Thinking about the contents of operating costs, we may distinguish three items: labour costs, advertising costs and IT costs. All of them are related to wage rates, but not in a simple way. In the case of the US data, we ought to use state level wage rate data, which we unfortunately do not have. However, we do have per capita income data, which we use instead. Similar data (GDP per capita) are also used for Europe. The subscripts i and t denote individual countries (in the US, states) and time periods (years), respectively. Almost all equations are estimated from cross-country time-series panel data so that the key elasticities are constrained to be equal over countries and time. Thus, we pay a lot of attention to the coefﬁcient of log(S), i.e., a1. In case (2.1) it simply gives the measure of scale economies, SCE (see Christensen and Greene, 1976) that is 1 – ∂log(C)/∂log(S) = 1 – a1. Moreover, using SCE we can compute the relationship between (short-run) marginal costs MC =∂C/∂S = {log(C)/∂log(S)}(C/S) and average costs AC = MC/(1 – SCE). If the parameter a1 is less than one we may conclude that the average costs decrease along with the growth of sales. With the European data we have a separate cross-section data set in which costs are expressed as cost

164 Gaming in the New Market Environment

shares.6 In that case, we simply regress the cost share against the total sales and examine whether the coefﬁcient of the sales variables is negative. In addition to this simple speciﬁcation we also estimate a translog-type cost function where the second power of log sales is introduced on the right-hand-side so that the estimating speciﬁcation takes the form log(Cit) = a’0i + a’1 log(Sit) + a’1l log(Sit)2 + a’2 log(Xit) + u’it

(7.2)

See Berndt (1991) for a general reference. Thus, in the basic speciﬁcation, we disregard the input price terms and the corresponding cross-terms with output. Yet, we do test the effect of the proxy variable, the per capita personal income. In our view, however, this variable is so akin to measurement errors that no formal tests of homotheticity are carried out. As for the economics of scope, our testing procedure is, as it is with homogeneity/homotheticity, a short cut. This is because we do not have complete output data on different products. Subsequently, we only have data on the number of games provided plus an indicator of the type of game provided (e.g., whether instant games are provided or not). Thus, in the estimating cost equation, we test the signiﬁcance of the additional number of games variable (given output and its squared term) to see whether, at a certain level of output, an increase in the output menu increases or decreases the total operating costs. For data reasons, this analysis only applies to the US. To have an idea of the absolute level of the marginal costs, we could also estimate a simple linear equation of the type Cit = b0i + b1Sit + b2(Xit) + eit

(7.3)

where marginal costs are, ceteris paribus, a constant fraction of sales. Thus, we could interpret b0 as the ﬁxed costs and b1 as the marginal costs. Although we concentrate on cost behaviour we also brieﬂy scrutinise the demand relationships that can be detected from the data. Since we only have annual aggregate (state level) data we cannot do very much to detect the basic determinants of demand. In particular, this is true in terms of price elasticity, which is the key parameter from a policy point of view. At any rate, the estimating equation takes the form of the following double log equation log

( )

( ( ))

Sit = d0i + d1 log 1 – POPit

Pit Sit

+ d2 log

( )

Yit + d3 log(Xit) + uit POPit (7.4)

where P denotes prizes, Y income, POP population and X other determinants of lottery sales. X includes indicators for education, race,

The Economics of Scale and Scope in the Lottery Industry 165

urbanisation, lottery in the neighbouring state and the number of games provided the lottery. In what follows, we present the empirical ﬁndings in the subsequent Section 7.3, and then discuss the implications of our ﬁndings in the concluding Section 7.4.

7.3 7.3.1

Empirical results Cost curves

The data are illustrated with Figures 7.2–7.5 below. These ﬁgures display scatter plots where costs (more precisely operating costs in relation to sales) are compared with the volume of sales. The European data are illustrated in Figures 7.2 and 7.3, whereas the US data are shown in Figure 7.4. With the US data, we also show how costs are related to the number of years since the start date, the number of games offered, population density and income level (Figure 7.5). Finally, we illustrate the relationship between sales and prizes, i.e., the payout ratio P/S (Figure 7.8).7 40

operating costs/sales, %

30

20

10

0 8

Figure 7.2

10

12

14 16 total sales (log)

Operating costs in Europe 1995–2005

18

20

22

166 Gaming in the New Market Environment 25

operating costs/sales, %

20

15

10

5

0 8 Figure 7.3

10

12

14 16 total sales (log)

18

20

22

Personnel costs in Europe 1995–2005

All estimates are derived from the panel so that the ﬁxed effects least squares estimator represent some sort of benchmark case. The equations have been estimated also with the random effects (RE) speciﬁcation. The benchmark model has also been estimated with the GLS and LAD (Least Absolute Deviations) estimator.8 Finally, the Instrumental Variable (IV) estimator has been used to account for obvious simultaneity problems. Now, let us turn to the estimation results. With the European data, parameter a1 is systematically below 1 and in most cases the difference between the parameter value and 1 is statistically signiﬁcant at the standard level of signiﬁcance of the t-ratio (Table 7.1). With the total operating costs, the estimate of a1 is relatively high but when we look at the personnel costs, the estimates are already much lower. Similarly, when move from level form to ﬁrst differences the estimates of a1 tend to decrease.9 Introducing the second power of the log sales variable (equation (7.2)) did not make a noticeable difference in results; in fact, it turned out that the t-ratio of the coefﬁcient a’11 was consistently statistically insigniﬁcant both for Europe

The Economics of Scale and Scope in the Lottery Industry 167 30

cost/sales, %

25

20

15

10

5 10

100

1000

10000

1000

10000

1000

10000

sales (log)

24

costs excl. com./sales, %

20 16 12 8 4 0 10

100 sales (log)

20

commission/sales, %

16

12

8

4

0 10

100 sales (log)

Figure 7.4

Operating costs in the US in 2005

168 Gaming in the New Market Environment 28 total expenses/sales, 2004, %

30

costs/sales, %

25

20

15

10

5 1960

24 20 16 12 8 4

1970

1980

1990

2000

2010

5

50

500

5000

Population density

30

30

25

25 costs/sales, %

costs/sales, %

first year of operation

20

15

10

20

15

10

5 2

3

4

5

6

7

8

5 24000

number of games

Figure 7.5

32000

40000

48000

income per capita

Factors affecting operating costs in the US

and the US. Therefore, we concentrate on reporting the results for the more parsimonious speciﬁcation (7.1). Just to illustrate to the results with (7.2), we report a summary of them in Table 7.6. With the European cross-section for 2005, the results follow the same pattern, although the estimates are not as precise (Table 7.2). The cross-section estimate suggests that the share of operating costs is negatively related to per capita GDP. If, indeed, the per capita GDP is used as an indicator of overall productivity, this sign of the coefﬁcient makes sense. We should, however, keep in mind that the per capita GDP could also reﬂect the level of wages which obviously should increase the cost share. The US data reﬂect the same basic features as the European data (Table 7.3). The only difference is that the results are (due to better

The Economics of Scale and Scope in the Lottery Industry 169 Table 7.1 Dependent variable

Panel data estimates of the cost function for European lotteries log(S)

R2

SEE

Estimator

log(TC)*

.562 (7.12)

0.998

0.141

LS, FE

log(TC)

.996 (87.12)

0.968

0.571

LS

log(TC)

.695 (12.63)

0.999

0.106

LS, FE

log(TC)

.781 (17.17)

0.923

0.113

GLS,CRE

Δlog(TC)

.889 (5.94)

0.631

0.104

LS

Δlog(TC)

.890 (5.94)

0.631

0.104

GLS,CRE

–.087 (0.62)

0.081

0.106

GLS

Δlog(TC/S) TC

.167 (78.46)

0..978

7995719

LS

log(PC)*

.370 (3.12)

0.996

0.262

LS, FE

log(PC)

.981 (24.01)

0.604

2.516

LS

log(PC)

.430 (3.12)

0.996

0.260

LS,FE

log(PC)

.498 (3.55)

0.457

0.263

GLS,CRE

Δlog(PC)

.362 (1.63)

0.081

0.187

LS

Δlog(PC)

.360 (1.68)

0.083

0.179

GLS,CRE

–.638 (2.86)

0.214

0.187

GLS

0.785

1779793

LS

Δlog(PC/S) PC

.010 (23.89)

* = the sample also includes Macedonia. Numbers inside parentheses are (corrected White) t-ratios. The number of observations is 141. When the left-hand side variable is expressed in differences (denoted by the difference operator Δ), also the dependent variable is expressed as Δlog(S). LS denotes the Least Squares panel estimates and GLS the Generalised (weighted) least squares estimates. FE indicates that ﬁxed (both time-series and cross-section) ﬁxed effects are used, whereas CRE indicates that the model includes cross-section random effects. TC denotes total operating costs and PC personnel costs.

data and larger number of data points) simply more precise. Therefore, no doubt is left about the magnitude of the scale variable. Its estimate is systematically below one and, in most cases, the deviation is statistically signiﬁcant.10 The only exception concerns the royalties (commissions) that are paid to retailers. They seem to more or less constant and only weakly related to the possible background variables. In fact, they have remained more or less constant over time, which makes them an exceptional cost category. Otherwise, costs (in relation to sales) have decreased along with an increase in sales. The costs also seem to decrease when the number of

170 Gaming in the New Market Environment Table 7.2

Cross-section estimates from the European data R2

SEE

0.049

0.420

–.265 (1.63)

0.128

0.412

–.045 (1.91)

–.325 (3.09)

0.044

0.431

log(P/Sales)

log(GDP)

.720 (0.26)

2.394 (4.56)

0.566

2.224

Dependent variable

log(S)

log(TC/S)

–.020 (1.70)

log(TC/S)

–.017 (1.21)

log(TC/S) (weighted)

log(S)

log(GDPpc)

The notation is the same as in Table 7.1. The number of observations is now 23.

games increases. Further, they decrease over time, suggesting that companies that have started just recently have, ceteris paribus, higher costs. This, of course, makes sense if we think about the various set-up costs in terms of personnel, advertising etc. We can also conclude that the costs are related to population density and income level. Again, the income level effect seems to be negative, which cannot be interpreted in any other way than assuming that the income level (also) reﬂects the level of overall productivity. The effects of the number of games, age of the company and population density are quite clear when the model does not include ﬁxed effects, entailing that both cross-section and time series variability are allowed to affect these variables. Introducing cross-section ﬁxed effects would account for most of the differences in the average costs, which is why we prefer not to use the ﬁxed effects speciﬁcation as the preferred model. The random effects model may make sense, although it does not pass muster very well in diagnostics tests. The important thing is, however, that the random effects model does not produce results that differ from those of the benchmark ﬁxed effects model. If the US results are scrutinised somewhat more carefully it can be seen that the commissions (which are) paid to retailers constitute a constant fraction of the total sales, reﬂecting the fact that the commissions contracts are made in terms of a given percentage of the total sales (which can also be seen from Figure 7.4). This does not mean that the true costs of providing retailer services are all variable and represent a constant fraction of sales. One might argue that most of the retailer costs are actually ﬁxed (to build the required facilities and to have the necessary educated manpower). The fact that contracts are

Table 7.3

Cost function estimates with the US data

Variables.

TC

TC

TC

TC

TC

TC

log(S)

.814 (120.6)

.816 778 (112.62) .(31.00)

TC

AC

OC

OC

OC

OC

OC

.854 (89.19)

.854 (138.6)

.826 .855 1.002 (34.75) (88.99) (209.7)

log(n)

–.154 (3.88)

–.063 (3.23)

–.174 (1.17)

–.155 (3.90)

–.314 (6.76)

.267 (3.23)

.270 (3.32)

log(2006-t0)

–.092 (3.25)

–.130 (8.74)

–.003 (0.09)

–.088 (2.95)

.038 (1.76)

–.180 (3.03)

–.175 (2.92)

log(dens)

–.040 (3.69)

–.040 (5.48)

–.029 (1.12)

–.035 (2.97)

.002 (0.85)

–.128 (4.22)

–.124 (4.18)

.691 (51.17)

.686 (41.16)

.613 (7.02)

.647 782 .783 (18.31) (32.63) (32.62)

–.076 (1.04)

log(PI/POP)

OC

–.048 (2.71)

R2

0.971

0.972

0.992

0.978

0.991

0.976

0.978

0.990

0.888

0.882

0.986

0.888

0.920

0.920

SEE

0.184

0.179

0.106

0.163

0.158

0.108

0.163

0.136

0.333

0.336

0.125

0.131

0.282

0.282

Estimator

LS

IV

LS

LS

GLS

GLS

LS

LS

LS

IV

LS

GLS

LS

LS

none

FE

none

none

CRE

none

none

none

none

FE

CRE

none

none

Panel setting none

The number of observations is 514 for TC and 377 for AC and OC. FE Indicates that the model includes ﬁxed (both period and cross-section) ﬁxed effects. CRE refers to cross-section random effects. The dependent variable is either log(TC), log(AC) or log(OC) as indicated on the ﬁrst line. In the IV estimation, the set of instruments includes, in addition to lagged endogenous variables, log(PI/POP), log(n), and Urban.

171

172 Gaming in the New Market Environment Table 7.4

Change in cost shares over time in the US constant

Δlog(S)

R2

SEE

Δ92-05log(TC/S)

–.078 (2.00)

–.292 (3.59)

0.268

0.193

Δ96-05log(OC/S)

–.141 (3.51)

–.310 (2.92)

0.127

0.254

Δ96-05log(AC/S)

.047 (0.90)

.268 (0.82)

0.093

0.261

The notation is the same as in Table 7.3. Δa–b denotes a difference between periods a and b.

Table 7.5

Marginal effect of sales on costs

Data

Coefﬁcient of S

R2/SEE

Europe; OC

.040 (13.88)

0.983 11.821

Europe; PC

.005 (7.92)

0.890 2.715

US, TC

.098 (19.97)

0.981 18.288

US; OC

.024 (1.01)

0.974 14.922

US, AC

.065 (12.62)

0.997 4.281

The estimating equation is equation (7.3). All equations are estimated using level-form data and the LS with cross-section ﬁxed effects.

Table 7.6

Further estimates of scale economies TC:E

PC:E

TC:US

AC:US

OC:US

log(S)

.912 (13.23)

.636 (2.21)

.747 (14.62)

1.224 (40.78)

.618 (6.50)

1

/2*log(S)2

.002 (0.95)

.009 (0.91)

.012 (1.39)

–.038 (7.17)

.012 (0.74)

R2

0.963

0.573

0.979

0.988

0.888

SEE

0.606

2.518

0.176

0.146

0.333

The estimating equation is equation (7.2). All equations are estimated with least squares without ﬁxed effects. ‘E’ denotes the European and ‘US’ the United States data.

The Economics of Scale and Scope in the Lottery Industry 173

made in the way they are made probably results from the fact that a ﬁxed fee contract would provide absolutely no incentive to the retailer to stimulate sales by better service. The estimation results suggest that if the retailer commissions component of costs depends on any variable, it is the number of games. This ﬁnding suggests that different games support each other. Lottery companies need not pay the same amount for all additional games that are introduced to the retailer network. Unfortunately we do not have complete data on the game menu of the different companies for each individual year. Thus, we only have a single set of values for 2000 and for every individual year. Moreover, we have a rough description of the game menu in the form of the data on number of games. In some cases two games may be only marginally different and in others, they may be completely different. Constructing a more sophisticated indicator is, however, beyond our data facilities. Estimation results suggest if the number of games increased, costs would, ceteris paribus, decrease. Because a slightly better ﬁt is obtained using a logarithmic transformation we may conclude that there are ‘decreasing returns’ in additional games. One could easily imagine that once a certain ‘optimal’ number of games is achieved, returns from additional games can even turn out negative. Even if the basic result of the/ analysis is clear, some caveats remain. The deﬁciencies in measuring the game menu show up in the results concerning companies’ own operating costs (OC). Using this concept of operating costs suggests that the effect of the number of games is negligible, or even positive. Whether this reﬂects genuine results is difﬁcult to say. In any case, the other results are basically similar to those concerning the total operating costs (TC). We have, in fact, also scrutinised the impact of different game categories. In other words, we found out which of the following games were offered: Lotto, Multi-state Lotto, Numbers Games, Video Lottery Terminals, Instant Games, Sports Betting, Powerball and Keno. If dummy variables are used to indicate whether each of these games/game categories are provided, we end up with the following result: Instant games, Keno and Powerball seem to have, ceteris paribus, a positive effect on costs, whereas Lotto and Numbers Games have a negative effect. The effect of the other games is zero, entailing that the introduction of these games will not change the predicted costs which are involved in the basic relationship (7.2).11 But what can we say about the size of the marginal costs? Basically, we could have at least two alternative shapes for the cost curve,

174 Gaming in the New Market Environment 1.0

0.8 1 0.6

2 0.4

0.2

0.0 250 Figure 7.6

500

750

1000

Two alternative marginal costs curves

illustrated in Figure 7.6. Marginal costs could go down to zero relatively quickly (ﬁrst alternative) or they could – after an initial rapid fall – diminish only gradually and practically never reach the zero level (alternative 2). The estimate of a1 gives some idea of the level of costs, but costs obviously change along with the volume of sales. As the coefﬁcients of the squared output terms a’11 can be set to zero, the marginal costs would be simply a1 × Cit/Sit. The corresponding values for the US panel data are reported in Figure 7.7 (they can be compared with, e.g., Viton, 1981). The outcome of these calculations is clear: marginal costs do seem to decrease continuously as the scale of operations increases. Thus, it would not seem that marginal costs would go down to zero very quickly, creating an indeterminate regime for large gambling companies, as hypothesised in alternative 2 of Figure 7.6.12 If the costs actually went down to zero with a relatively low level of sales, a door would be opened to monopolistic competition between the large players. This appears to be the current market situation with private gaming companies (cf. the case of Taiwan, illustrated in Wang et al, 2006). But as said, this is not what the data say.

The Economics of Scale and Scope in the Lottery Industry 175

Marginal operating costs

.20

.16

.12

.08

.04

.00 5

50

500

5000

500

5000

500

5000

Sales

Marginal costs for commissions

.10

.08

.06

.04

.02

.00 5

50 Sales

Marginal total operating costs

.30 .25 .20 .15 .10 .05 .00 5

50 Sales

Figure 7.7

Marginal costs for US companies 1992–2005

176 Gaming in the New Market Environment

If the estimates of the linear model (7.3) are computed instead of the log linear speciﬁcation (7.2), the immediate reaction is that the marginal costs are very small, especially if we disregard the commissions (which are) paid to retailers (which seem to be a constant fraction of sales almost universally), see Table 7.4. Consequently, we end up with an estimate of 2–4 per cent.13 An increase in sales of, e.g., 50 per cent would only increase the operating costs by one per cent or slightly more. In other words, the increase in costs would perhaps not really affect the optimal level of production.14 In the light of Figure 7.1, this means that even in a monopoly situation, the sales would grow very large. Some caveats regarding the cost estimates should perhaps be discussed here. The ﬁrst caveat deals with the sampling distribution. All the empirical analyses make use of data on relatively small companies. In fact, in the US data, the share of the biggest company (New York State Lottery) represents just 12 per cent of the total US sales, whereas the corresponding ﬁgure (for Lottomatica) in Europe is 15 per cent. Hence, we are left with considerable uncertainty concerning the cost conditions of companies that might have the scale of the total US and/or European markets. Another caveat has to do with the nature of the games. The cost conditions in setting up new games and selling them may be quite different in the future, which obviously affects the shapes of relevant cost curves. IT costs have decreased considerably and so have those incurred by internet sales. 7.3.2

Demand functions

Finally, let us turn to the results on the demand for lottery products. We have a (very rough) idea just by using the European data (Table 7.2). However, with these data, we cannot control the most important cost elements (the simultaneity bias, caused by the use of the actual (ex-post) payout ratio as an indicator of the expected price of a lottery ticket This can be done slightly better with the US data with which we can control the most important determinants of lottery demand (Table 7.7). Thus, a scrutiny of the results suggest that – in compliance with most of the earlier results – lottery demand in the US is positively related to income, education (low level of equation), race, and presence of lotto in the neighbouring states. In other words, we do not ﬁnd a positive spill-over effect coming from (the very few) non-lottery states. Rather, it looks like the areas (the Bible belt area) where lotteries are still prohibited represent an altogether low level of demand. As for the other variables, such as population density and the share of urban/rural population, we

The Economics of Scale and Scope in the Lottery Industry 177 Table 7.7

Some demand function estimates with the US data

Variables

1

2

3

4

log(1–(P/S))

–.810 (3.37)

–.898 (3.29)

–.386 –1.652 (1.83) (5.48)

log(PI/POP)

2.133 2.078 1.398 .943 (15.93) (15.05) (3.91) (6.38)

log(POP)

5

6

7

8

–1.710 –1.382 (5.56) (9.75)

–1.575 –1.756 (10.45) (12.23)

.731 (4.27)

.697 (9.00)

.421 (3.90)

.233 (2.06)

.098 (4.64)

.100 (9.88)

.082 (6.17)

.049 (4.03)

D

.425 .442 509 .531 .523 (10.78) (11.90) .(23.64) (18.67) (18.55)

Race

.022 (8.40)

.023 (8.81)

.021 .020 .018 (19.69) (15.10) (15.26)

Education

.032 (3.37)

.018 (1.63)

.008 (1.45)

Urban

.009 (1.45)

.003 (0.62)

.005 (3.47)

.009 (5.76)

log(n)

.328 (5.54)

R2

0.325

SEE

0.526

0.495

0.250 0.431

0.421

0.408

0.407

0.398

Estimator LS, Panel setting none

IV none

LS, FE

LS, none

GLS, none

GLS, none

GLS, none

0.333

0.863 0.549

LS, none

0.571

0.837

0.844

0.850

The estimating equation is equation (7.4). PI denotes personal income, POP population, Race the share of black people, Education the share of people who have not completed the 9th grade and Urban the share of urban population (as an opposite to rural population). D, in turn, denotes a dummy for a state which is surrounded by lottery states only. The dependent variable is the log of (real) per capita sales. In the IV estimation, the set of instruments includes, in addition to the lagged sales and prize variables, even log(n) and D.

ﬁnd them to involve relatively little explanatory power. Concerning the income effect, we ﬁnd slight evidence that elasticity is just above unity and that, as regards the income level and the number of people, the latter is slightly more important. This would seem to suggest that income distribution matters. Thus, we could argue that high-income people purchase (in relative terms) fewer lotto products than lowincome people. The most interesting issue is, of course, the coefﬁcient of the price term indicating how sensitive the market demand is in terms of

178 Gaming in the New Market Environment 75

prizes/sales 2005, %

70

65

60

55

50

45 0.01

0.10

1.00

10.00

sales 2005 (log) Figure 7.8

Effect of prizes on sales in the US

changes in the payout ratio (this is illustrated in Figure 7.8). Unfortunately, we cannot do much better here than with the European data in constructing a proper measure of the expected price of lotto in order to avoid measurement errors and simultaneity bias in terms of the sales variable.15 We estimate the model by the Instrumental Variable estimator, but the set instrument is not particularly exhaustive. Even so, the estimates of the coefﬁcient of the prize variable make sense and suggest that demand is sensitive to prices – as well as to the output menu. On the basis of these estimates, it is rather difﬁcult to say whether price elasticity stays above or below one. Whilst interpreting the price elasticity, we are faced with one complication: the elasticities do not appear constant according to company size (which could in a Cook and Clotfelter, 1993, sense translate into the absolute size of prizes). We have no data on the size distribution, which means that we cannot test the effect of the absolute sizes of the prizes on sales. Yet, we can test whether price elasticities are constant. The testing is based on a simple threshold model where the volume of sales is used as the threshold variable and s* the corresponding ﬁxed

The Economics of Scale and Scope in the Lottery Industry 179

threshold value. Least squares estimation produces the following results16 log

( )

( )

( )

S P P = 10.68 – 1.04log 1 – S ≤ s* – 1.68log 1 – POP S S + 1.28log

( )

S > s*

PI + .76log(n) POP

(7.5)

R2 = 0.51, SEE = 0.44, Wald F(1,510) = 193 (0.00), s* = 1060, t1 = 6.5, t2 = 3.7, t3 = 7.4, t4 = 5.6, t5 = 5.8 Thus, large companies seem to face more elastic demand. This might reﬂect the simple fact that small companies (states) are more isolated and the respective consumers have a smaller opportunity set in terms of consumption/leisure choices. It is also interesting to see that demand is sensitive to the menu of products. Thus, the number of games has a strong positive effect on aggregate sales. This is no surprise, because Figure 7.9

Operating costs/sales from European and US lotteries

Operating costs vs sales in 2000 20,00% 18,00% 16,00% 14,00% 12,00% 10,00% 8,00% 6,00% 4,00% 2,00% 0,00% 0

1000

2000

3000

4000

5000

6000

Data source: Raymond Bovaro (with the kind help Jean Jorgensen).

7000

8000

9000

180 Gaming in the New Market Environment Figure 7.10

Operating costs for certain private gaming companies, 2005

12 correlation = –0.81

operating costs/sales, %

10

8

6

4

2

0 100

1000 10000 SALES, €, log scale

40000

The data were kindly provided by Sami Ahtiainen and Annukka Ruuhela from the Finnish National Lottery Veikkaus.

launching new products is basically aimed at attracting new customers. Although new products always cause some cannibalisation on the existing products, the effect is hardly 100 per cent. Thus, the total sales of lottery products will certainly increase. This does not, of course, automatically imply that even proﬁts will increase. Since we cannot control the proﬁtability of each individual game here, we cannot conclude whether the number of games is optimal.

7.4

Concluding remarks

The above analyses have shown at least one important result: there are important economies of scale in lotteries – and possibly even in the whole gambling industry. In addition, economics of scope appears to be an important, if not dominating feature of the gambling industry. This, in turn, suggests that an entirely competitive market structure probably would not prevail in the long run. Rather, there would emerge tendencies towards a natural monopoly. This would, in turn, lead into a

The Economics of Scale and Scope in the Lottery Industry 181

loss of consumer surplus and turn out difﬁcult for public policy. Natural monopolies are often regulated to improve the consumers’ position, but it is generally recognised that there are various problem in regulation (they have been reviewed by, e.g., Depoorter (2000)). In gambling, the major problem would probably be related to the fact that the possible natural monopoly would very likely prove global and it would not be immediately obvious which organisation would assume the responsibility for the regulatory operations. If all companies were private, the development towards a natural monopoly would occur via acquisitions, mergers and bankruptcies. While the biggest company would take over the traditional lottery market, smaller companies could continue operations with more specialised games and supporting activities. In the case of the more cost intensive gaming activities (e.g., casinos), markets would probably not integrate completely, leading us into the prevalence of a spatial market structure.17 In case both public and private companies tried to stay in the market, the outcome would be far from clear – it is more a matter of political economy than industrial economics. Even if a (the biggest) company would come close to a natural monopoly situation, public gambling companies would not automatically exit the market. We could, in fact, face quite a complicated market structure and unpredictable pricing strategies. In such a case ‘price wars’ could be long-lasting and bitter. We can only speculate how the story will end. Our analysis has mainly focussed on the effect of company scale on costs. Clearly, this is not the only signiﬁcant issue. It could well prove that better efﬁciency in organisation, innovations, business culture etc. could compensate the scale/scope disadvantage. Therefore, it would be useful to examine this question, e.g., with the DEA type approach. A comparison of the efﬁciency of public and private companies could be a ﬁrst step towards solving this problem.18 Notes 1 Matti Viren, 20014 University of Turku, Finland. E-mail: matvir@utu.ﬁ, and Bank of Finland (Helsinki, Finland) and Public Choice Research Centre (Turku, Finland). I am grateful to Ville Korpela for research assistance. Useful comments from Ari Hyytinen and the participants of a faculty seminar in Århus, January 2007 are also gratefully acknowledged. 2 See, e.g., Clotfelter, Cook, Edell, and Moore (1999), Clotfelter and Cook (1991) and Douglas (1995) for US and UK (historical) experience on behaviour of lottery markets. 3 See Bohn et al (2001), Christensen and Greene (1976), Kwoka (2005), Sing (1987) and Sung and Gort (2000) as examples of analysis of potential natural

182 Gaming in the New Market Environment

4

5

6 7

8 9

10 11

12

13

14

15

16

monopolies in transportation, payment systems and power production. The results represent a mixed bag and do not strongly favour the existence of economies of scale for all levels of production. Almost all ﬁnd non-trivial economies of scale at low level of output but beyond that, the cost curves appear to be almost ﬂat. That is, because of the link between volume of gambling and the harmful effects of (excessive) gambling. See, e.g., Grifﬁths (2008) for empirical evidence on this link. The data from the 11 accessible private companies are illustrated in Figure 7.10 at the end of the paper. Quite clearly, these companies seem to follow exactly the same pattern as the lotto monopolies. Wang et al (2006) report some data from nine Taiwanese gaming ﬁrms which also seem to ﬁt into the same pattern. Thus, the biggest ﬁrm (market share 57 per cent) has by far the lowest average cost/sales ratio (4.8 per cent). We can also make use of the data that have been collected by Raymond Bovaro for both private and public companies (Figure 7.9). These data include more countries than the cross-country time series data. Both the European and the US data are real (all nominal values have been deﬂated by consumer prices). The European data have also been transformed to euros. The LAD results are not reported. This is because of space reasons and because estimates were practically identical with the least squares estimates. In the European data, there are some dubious observations. This is particularly true with Macedonia which is, in fact, disregarded in most of the analyses (see Table 7.1). Using the IV estimator suggests that deviation is not due to the simultaneity bias. Instant games are offered by all but one state lottery. Thus, the corresponding effect can simply reﬂect a sampling error. As for the effect, the following values for the dummies were obtained: Numbers Games –.06, Lotto –.14, Keno .12, Powerball .06, Multi-state Lotto .01, VLT .04 and Sports Betting .01. However, the t-values turned out to be high only in the case of Lotto and Keno. Even though we took the squared log sales terms into account, marginal costs would keep falling with the level of output that corresponds to the combined sales of the US lottery companies. These ﬁgures are in compliance with the data that we have on the individual companies’ cost structure, evolution of personnel, and the size and distribution of the marketing budget and IT expenses. Obviously, the market structure also depends on ﬁxed costs. Thus, it would seem that, irrespectively of the prospects of the large companies, small ﬁrms could not survive in the market unless they found a set of products special in terms of quality and technology. It should be recalled that we are considering the average payout ratio at the annual frequency, which makes the construction of a proper price variable quite demanding. The simple ex post data on payout ratios probably reﬂect the pricing regimes which basically determine demand behaviour. The value of the threshold parameter is almost identical to the sample average value of sales for the 41 companies over 1992–2005. Obviously, a lower threshold would make more sense.

The Economics of Scale and Scope in the Lottery Industry 183 17 When considering possibilities of preserving some local or game-speciﬁc monopolies one has to keep in mind that in this industry the role of patents, copyrights and trade-marks is presumably much smaller than in other industries and thus they give less protection against global competition. 18 Wang et al (2006) is an interesting attempt towards this direction.

8 The Political Economy of Regulating Gambling Illustrated with the Danish case

Martin Paldam1

8.1

Introduction: Three types of regulation with two purposes

Gambling2 is a range of service products that is regulated in most countries. The purpose of this article is to discuss the political economy of this regulation. Regulation is done for three reasons, of which the ﬁrst two are difﬁcult to accept for the economist: (i)

Gambling is often assessed morally as a vice (like liquor, tobacco and cars) that deserves especially high taxes. The moral assessment is often religiously based, and regulation goes back to the Middle Ages. (ii) Due to the long history, regulation has a strong path dependency – and it has created stakeholders. If for some reason regulation had to start all over, it would probably be done differently. (iii) The rational reason for the special treatment of gambling is that it creates the gambling problem of addiction for a small fraction of the population. The problem is an externality which should be regulated.

Externalities are often difﬁcult to assess empirically, and policy decisions on regulation need macro estimates. The ﬁrst half of the paper develops a method to assess the aggregate social costs of gambling for a country and actually assesses the numbers for one country (Denmark). This is done in sufﬁcient detail so as to allow others to use the model with data from their own country. The analysis consequently distinguishes between normal (recreational) and problem gambling, and assumes that the regulation on the gambling market today has two purposes: to collect taxes and to reduce the externality of the gambling problem. The two purposes are at odds: If gam184

The Political Economy of Regulating Gambling 185

bling is reduced, so is the revenue and vice versa. The outcome observed is a compromise that has been shaped by the political and economic forces operating in the ﬁeld. The political economy starts from the weak victim observation that the victims of the gambling problem are few and prefer to hide. This makes them unorganised, largely invisible and hence politically weak. On the other hand, the political decision-makers need taxes for many purposes, and the stakeholders, who use the tax revenue, tend to be powerful. Public choice theory consequently predicts that the tax purpose will capture the instrument. Clearly, it is a ﬁeld where the ideal of welfare maximisation and political realities may generate different outcomes. The externality is complex, and so is the regulation as it uses a mixture of three instruments: Gambling is taxed, and it is subjected to administrative regulation. In addition, some or the entire range of gambling products are normally reserved to a GSOE (i.e. Gambling State Owned Enterprise). The GSOE share on the gambling market differs widely across countries due to both history and choice: Some countries have a dominating GSOE, and others have privatised (most of) the sector. Regulation attempts to reduce the gambling problem by restricting the range of available gambling products to relatively benign ones and by making marketing of gambling products less aggressive. This may – together with high taxes – prevent gambling and hereby reduce the gambling problem. The alternative to prevention is treatment. As the gambling problem only affects a small fraction of the population, treatment may be more effective. This is a ﬁeld where a major cost beneﬁt study appears to be missing. Both as regards tax collection and gambling problem reduction, the regulation becomes softer and more informal when it is done through a GSOE than through a set of market ﬁrms. The GSOE is a monopoly that generates rent, and as the state is the owner, it is also the residual claimant, receiving the rent as tax revenue. Also, regulation of the GSOE may be done through its charter or by direct orders from the political decision-makers representing the owners. It is a tradition – going back to the start of regulation – to earmark the rents generated by the GSOE to a G-fund. That is, the rents are not paid into general revenue, but reserved to some special purpose outside the normal spending programmes. Most activities ﬁnanced by the G-fund might have been ﬁnanced anyhow, but the special rules for G-funds must generate a switching effect,3 which creates the stakeholders of the GSOE. Even if the switching is weak, the recipients will suspect that they may at most get the same under an alternative regime. Also,

186 Gaming in the New Market Environment

the ability to switch expenditures gives power to some decision-makers. So, two groups of stakeholders are created. Below we only consider the ﬁrst group of stakeholders: the beneﬁciaries of the G-fund. Section 8.2 discusses the microeconomics of the demand for gambling by normal and problem gamblers. The article turns to the macro level in order to discuss policy. To keep track of the effects, Section 8.3 presents a macro model – it is made to handle the data for any country. Section 8.4 gives a brief quantitative description of the Danish gambling sector, and shows how the data found can be used by the model of Section 8.3 to assess the costs of gambling for one country. Section 8.5 discusses the political economy of GSOEs, and ﬁnally Section 8.6 summarises the results.

8.2

Micro: The gambling demand of a normal and an addicted gambler

The micro analysis considers the demand for gambling for each of the two types of gamblers. Addiction is irrationality, which is difﬁcult to deal with in economics. Sections 8.2.1 to 8.2.3 rationalise the irrationality by assuming that the addicted gambler has a large and inelastic demand. Sections 8.2.4 and 8.2.5 sketch a more complex trigger/ spree mechanism. 8.2.1

Two demand curves

Figure 8.1 shows the two demand curves, Q = DN(λ) and Q = DP(λ), for a normal and a problem gambler. The DN-curve is a normal looking demand curve that gives a modest budget share for gambling, well within the means of the gambler. The evidence is that it has a price elasticity of about –2 (see Farrell, 2008). The problem gambler has the demand curve, DP, which is much higher and steeper than DN. The price elasticity is smaller (numerically). Perhaps it is – 21–, or even closer to zero. However, much of this curve is unobservable (the dotted part), as it quickly hits some x, where it crosses a gambling budget constraint, the m-curve, and becomes irrelevant. That is, the m-curve is where the product of the gambling price λ and quantity Q is equal to the most the gambler can afford, m. Thus, the m-curve is λQ = m, or Q = m/λ, which is a rectangular hyperbola

(2.1)

From point x, the demand comes to follow the max budget share upwards. This part of the demand curve is termed DP2. The demand

The Political Economy of Regulating Gambling 187 λ DP1(λ)

DP2(λ)

m

-c ur

m

=Q λ

ve

X

DN(λ)

Q Figure 8.1

Demand for gambling by a normal and an addicted gambler

curve, DP, thus has a kink in x, and becomes a constrained behaviour when prices are above x. Here the demand curve turns into a gambling constraint with the price elasticity of –1.4 8.2.2

The gambling constraint: The m-curve and its relation to income5

Figure 8.2 develops Figure 8.1, but it only looks at the compulsive gambler. Now we try to explain the m-curve. The essential point is that the m-curve is substantial relative to the income of the gambler. The distance is Z. It is the necessary expenditures – for the gambler and those dependent upon him – to everything else than gambling. The gambler is constrained by her income. She really wants to gamble more. If the price is λ1 the gambler can ‘afford’ to buy the quantity of gambling Q1, but she wants to buy the quantity Q1 + θ1 as shown. This puts pressure on reducing Z. In the same way if the price is λ2. Here the gambler wants to buy the quantity Q2 + θ2 that is higher than her income. Thus, θ may be termed the tension variable. It is likely that Z = Z(θ), so that the higher θ is, the more Z is reduced. And, of course, the more Z is squeezed, the more the family and friends of the gambler are tempted to turn away from the addict.

188 Gaming in the New Market Environment λ

DP(λ) Z

λ2

θ2 I nco

λ1

θ1

me

Z

m-cur ve m = Qλ Q2 Figure 8.2

8.2.3

Q1

Q

Demand for gambling by an addicted gambler

The effect of increasing the tax on gambling

Assume that the rate of revenue collection goes up either by increasing the monopoly power of the GSOE or the tax on the private enterprises goes up, corresponding to an increase in the cost of gambling from λ to (1 + t)λ, so that the tax increase is tλ. The revenue of the tax increase on normal gamblers gives the typical welfare calculations like the calculations for any other good – see Figure 8.3. As the good has a relatively elastic demand, it causes a large fall in consumption and thus a fall in the consumer surplus, A + B. It gives the revenue A, which is redistributed to some other purpose, which we for simplicity assume has the same utility. The revenue is A = tλQ2

(2.2)

The loss of welfare is B, which is roughly half of the square C, so we get B≈

C tλ = (Q1 – Q2) 2 2

(2.3)

For the non-constrained part of the demand curve for the addicted gambler, the story is the same except that the area B now becomes

The Political Economy of Regulating Gambling 189 λ

Tax increase: tλ (1+t)λ C

A

B

λ

DN(λ)

Q2 Figure 8.3

Q1

Q

The effect of a tax increase on normal gamblers

much smaller. Money is taken from the gambler and redistributed to some other purpose. Another story occurs for the constrained part of the demand curve for the addicted gambler. Now Q = m/λ. This allows us to calculate everything;6 but we only show the results corresponding to (2.2) and (2.3) A = tλQ2 =

B≈

tλ (Q1 – Q2) = 2

(

tλ t m= m (1 + t)λ 1+t

tλ tλ – λ (1 + t)λ

)

t2 m m t = =A (t + 1) 2 2 2

(2.2b)

(2.3b)7

Imagine that the tax increase is 10 per cent, so that t = 0.1. Then B is only 5 per cent of A. Thus we still have the fact that money is redistributed from the addicted gambler to everybody else with a small loss of consumer surplus. These results therefore show that an increase in the taxation on gambling causes a large loss of welfare for the normal gamblers and a small one for the addicted gamblers, who still spend all they can on gambling. Thus the gambling externality is not reduced. This is probably not what the political decision-makers imagine.

190 Gaming in the New Market Environment

8.2.4

Irrationality of addiction: The trigger/spree-mechanism

Compulsive gambling is often described in ﬁction, with Dostoyevsky (1866) as the most famous gambling novel. It is known as an accurate description of the author’s own addiction.8 Also, psychologists like Jørsel (2003) have given illuminating case studies suggesting that addicted gamblers at some stage in their addiction lose interest in and the ability to perform normal work. Part of the process is that addicted gamblers borrow money from everybody they know, and by not repaying turn friends and colleagues into enemies. This may lead to loss of job and a general social deroute. That is, excessive gamblers ﬁrst run down their social capital, and then they start borrowing on the grey market, soon getting into a real economic mess. Also, the need for extra income generated by the addiction may be directed toward antisocial behaviour. This is modelled as the DWL on Figures 8.4 and 8.5. Addiction is an irrationality corresponding to an extreme behaviour. Figure 8.2 gives a rational economic interpretation of addiction as a steep demand curve turning into a constrained one. This is probably realistic for the average of all addicted gamblers, but it is not a realistic description of each individual one. From the case description referred to it would perhaps be more accurate to imagine that the demand for gambling by the addict is highly irregular, like the one for alcohol by the alcoholic. Often, the addict refrains from indulging in his obsession for some time, but then something triggers the addiction, and once started, things get out of hand. Thus, the demand curve may shift between horizontal and vertical due to a trigger mechanism, which is difﬁcult to generalise. The addict is unable to bet only a little like normal people. This may be modelled, but the model is likely to be complex. Using the analogy of alcohol and drugs, we say that the addicted gambler does not gamble when she is sober, but once something triggers the mechanism of addiction, she goes on a gambling spree. 8.2.5

Controlling the gambling problem by reducing the number of triggers

One way to reduce the gambling problem is to identify and reduce the number of triggers. The main problem is that this is a ﬁeld

The Political Economy of Regulating Gambling 191

where knowledge is sporadic, and the triggering may have a large random element. However, 4 factors may be generalised: (r) Some gambling products are stronger triggers than others. (a) Advertisement may trigger sprees. (λ) It is likely that one person’s addiction may be triggered by other people gambling, and thus be a function of the amount of normal gambling that depends on the price l. (S) Treatment may reduce the number of problem gamblers. Section 8.3.1 argues that the number of problem gamblers, NP, may change as a function of these factors – this will be discussed as a macro effect ΔNP = q(r,a,NGN(λ),S), the partial derivatives are qr > 0, qa > 0,qλ < 0 and qS < 0

(2.4)

While the signs on the effects (the partial derivatives) are probably not controversial, the sizes of the effects are largely unknown. Grifﬁths (2008) argues that the effect, qr, of regulating the range, r, of gambling products is substantial, especially if the most trigger-prone products are singled out. Also, it appears likely that there is a sizable effect, qS, of treatment, S. The other two effects are more speculative, and I have found no numbers to allow even tentative quantitative assessments. While a tax increase has no effect on NGP = m, it may reduce the gambling problem by reducing normal gambling and hereby triggering fewer new problem gamblers as modelled in (2.4). This is probably not a strong effect. The argument for a GSOE is that it may provide fewer triggers by behaving more responsibly and less efﬁciently. Also, it is closer to the politicians, who have people’s welfare in mind. But then, politicians are also susceptible to stakeholders, and they want tax revenue.

8.3

Macro: The gambling problem

The present section presents a simple macro model of the gambling problem by use of a graph with a basic macro model. It uses the concepts of a dead weight loss, DWL, and excess gambling expenditures, NGP, to model the problem. The two key explanatory variables in the discussion below are the price of gambling, λ, and the GSOE share of the

192 Gaming in the New Market Environment Table 8.1

The variables and the main bookkeeping relations

a, r

Advertisement, and product range in the gambling sector

C

Value added of gambling industry. Second part of gambling costs

D

Demand for gambling, by a normal and a problem gambler: DN and DP = m

DWL Production loss due to problem gambling: DWL = β NGP, where β is a parameter G

Gross gambling expenditure: G = GN + GP = W + R + C = W + NG

⌫

Total costs of the gambling problem, estimated as: ⌫ = DWL + NGP

λ

Price of gambling: λ = λR + λC = T/G + C/G = NG/G

N

Population, divided into normal and problem gamblers: N = NN + NP

NG

Net gambling expenditure: NG = NGN + NGP = G – W = R + C, the cost to the gambler

NGN Normal gambling: NGN = NG – NGP = (1–α)NG NGP

Net spending on problem gambling: NGP = αNG, where a is a parameter

␮

The share of the GSOE in production, 0 ⭐ ␮ ⭐ 1

q

ΔNP, increase in the number of problem gamblers

S

Budget for problem treatment: qs = ∂q/∂S is treatment effectiveness

T

Revenue to public sector, where T = TM + TP. Main part of gambling costs

TM

Monopoly rent of GSOE (Gambling State Owned Enterprise). Paid as tax

TP

Gambling tax paid by market ﬁrms, over and above other taxes

W

Winnings returned to gamblers

industry, m. Things will be explained as we go along, but Table 8.1 may be useful to keep track of the variables. 8.3.1

From the micro level of Section 8.2 to the macro level

The population, N, consists of N = NN + NP, i.e., normal and problem gamblers. In the static calculation, the gross movements between the groups even out. From Section 8.2, we know that the average normal gambler spends Avr(DN(λ,a,r)). The aggregate normal net gambling is: NGN = NNAvr (DN(λ,…)). Section 8.3.2 replaces the average with a NG-curve that gives the ‘normal’ gambling expenditure of all gamblers. In the same way, we get that the average problem gambler spends m on her addiction. Thus, the aggregate net gambling of the addicted is

The Political Economy of Regulating Gambling 193

NGP = NPm. Section 8.3.2 replaces the average with a triangle between the NG-curve and NGN-curve as will be explained. This gives the aggregate net gambling expenditures based on averages NG = NGN + NGP = NNAvr(DN(λ,…)) + NPm

(3.1)

We have data for NG, and from sociological studies, we may have measures of the two N’s, so with data for either Avr(DN) or m everything can be estimated/assessed. However, there is the problem that the two groups may not be permanent. Section 8.2.5 argued that various factors, such as r, a and NGN(λ) may trigger dormant addictions. Thus, there may be a net movement of people to the problem group as modelled in equation (2.4) ΔNP = q(r,a,NGN(λ),S). The group-shift variable, q, turns the calculations dynamic. If the 5 variables in (3.1) can be calculated for more years, it is possible to calculate q and to estimate the effects on q of the other variables. This is certainly not possible for the Danish data we consider. 8.3.2

The core of a model – explaining Figure 8.4

Figure 8.4 is a macro model of the gambling problem. It considers a rectangle giving GDP measured as income: GDP is the number of people times their income. The ﬁgure depicts two key curves: The income curve and the NG-curve. The trick that generates the ﬁgure is that people are sorted by the net amount used for gambling, NG. The income curve: Figure 8.4 assumes that gambling is randomly distributed in the population, so that there is no relation between the amount of gambling and income. This makes the income-curve horizontal, except the downward bend to the extreme right that will be discussed. This assumption will be relaxed in Section 8.3.4. The NG-curve: By the sorting, the NG-curve must slope upward. NG is the ‘wattled’ area below the upward sloping NG-curve. The gambling problem only occurs for gamblers spending more than a certain amount (the G-limit) on gambling. Thus, the bold grey arrow of the G-limit divides the NG-area into normal gambling, NGN, and problem gambling, NGP. NG redistributes income (and hereby consumption) from the gamblers to the gambling industry, C, and the state, T. Here, the assumption of full employment (or steady state growth) is essential. We assume that instead of producing the service of gambling, the labour force would in its absence have produced the goods the gamblers

194 Gaming in the New Market Environment Figure 8.4

A stylised theory of the gambling problem

Income, expenditure DWL

Income-curve

GDP G-limit

NGP NGN

NG-curve 0%

25% 50% 75% Population sorted by gambling expenditures

100%

Note: The text concentrates on three areas: DWL and the two parts of NG: NGP + NGN. They are drawn so that the NG-area is approx. 5 per cent of GDP. The DWL and NGP both amount to approx. 2 per cent of GDP. This is exaggerated to make the areas visible. In fact, the full NG is only about 1.2 per cent of GDP in Denmark, or one third of the areas shown. About the same reduction is likely to apply to DWL and NGP.

would otherwise have consumed. For normal gambling, one may assume that gambling replaces other recreational goods. As discussed in Section 8.2.4, addiction causes problems for the gambler himself in the form of a social deroute. This is likely to lead to a loss of career opportunities and even job, so the income curve will eventually turn down, as shown. This is termed a Dead Weight Loss, DWL, as it is a potential production that is lost due to gambling. It is shown on the upper right hand side of the graph. The upper bold grey arrow shows where the DWL turns visible. It gives an alternative measurement of the G-limit, where the problem begins.9 For the addicted gambler, the excess gambling causes problems, both for himself, as caught by the DWL, and for others, mostly as a family externality. It is proportional to NGP, and we may simply use NGP as the measure. Both DWL and NGP are measured (or at least measurable) in simple economic terms. 8.3.3

The two problem areas DWL and NGP

My intuition is that the form of the two problem areas, DWL and NGP, are roughly similar. They certainly follow from each other. The DWL

The Political Economy of Regulating Gambling 195

occurs due to excess gambling, and once the social deroute of the gambler starts, gambling becomes more and more excessive. So I shall simply assume that DWL and NGP are proportional. Consequently we have two losses: DWL, dead weight loss, not measured, but guesstimated to be DWL = βNGP. NGP, excessive net loss of family income NGP = αNG, which is assessed using relation (3.1) The size of the gambling problem, Γ, thus becomes Γ = DWL + NBP = (1 + β)NBB = (1 + β)αNB

(3.2a)

As everything is kept simple, it allows a basic assessment of the size of the gambling problem from NG and estimates of the two parameters α and β. Here α can be calculated from the variables of relation (3.1). It is not accurate, but the order of magnitude is reasonably easy to assess. However, the evidence on the size of the DWL is weak. For Denmark Bonke (2007) and anecdotal evidence suggest it is small. I shall use the guess that β ≈ 0.5, which is probably too high, so that DWL is half NGP ⌫ ⬇ 1.5NBP = 1.5αNB 8.3.4

(3.2b)

Two complications: Income dependency and excess production

The income-curve on Figure 8.4 is horizontal, assuming that gambling is independent of income. This is in accordance with Bonke and Borregaard (2006); but Smith (2008) provides evidence that NG is falling relatively as income rises. This can be depicted as a downward sloping income line as drawn on Figure 8.5. Another possible complication is that the demand for income of the addicted gambler is such that it generates an excess production, EP, as drawn on Figure 8.5. Maybe the gambler does not want her family to suffer, and thus works extra hard to make money. This is actually what many of the respondents claim in Bonke (2007). Also, the addiction of Fyodor Dostoyevsky often caused his household to be close to economic ruin. It appears that he wrote particularly fast when bankruptcy was close. If there is an EP-area, it offsets some of the DWL. This is another reason not to take the DWL to be as large as the NGP-area. However, the assessment of the EP and the DWL-areas are based on little evidence.

196 Gaming in the New Market Environment Income, expenditure

EP Income-curve

DWL

GDP NGP NGN NG-curve 0%

Figure 8.5

25% 50% 75% Population sorted by gambling expenditures

100%

Adding the possibility of excess production

The model shown on Figures 8.4 and 8.5 can thus be amended to deal with several complications that appear in reality. Also, since it is a macro relation, a lot of the more random micro problems, such as the trigger/spree mechanism, are likely to aggregate out. Table 8.1 deﬁned the cost of gambling as the ratio between net and gross gambling λ=

NG (G – W) (T + C) = = = λR + λR G G G

(3.3)

We assume that the price inﬂuences demand in the usual way ∂B/∂λ < 0, with the price elasticity, see Section 8.2. This gives the same expression for the effect of taxes and other costs. It appears that gambling addiction in many cases can be cured like alcoholism once the addict recognises the problem and makes an effort not to bet on anything, see Jørsel (2003) and the references in Bonke (2007). Also it appears that a medical treatment may work. Hence, it is clear that qS > 0, and it is even likely that qS is substantial.

8.4

Macro: A case study of the Danish case

The model just presented should allow the reader to estimate the costs for society of gambling. The data from Denmark will be used to show how it is done in practice. If the reader believes the data, he may just

The Political Economy of Regulating Gambling 197

look at Table 8.1 and then jump to the conclusion in Section 8.4.4, where the total costs of gambling are calculated for Denmark. 8.4.1

The gambling sector

A recent study on privatisation and public ownership in the (old) EU countries (Kötherburger et al 2006) shows that Denmark is a country with few SOEs (State Owned Enterprises). Virtually all are in the network industries where natural monopolies have to be regulated anyhow. The Danish GSOEs are thus an anomaly,10 but a large liberalisation was implemented in the early 1990s, and the GSOEs now only produce about 40 per cent of the gambling. The two GSOE ﬁrms have an old history. The oldest is the State Lottery started by King Frederik V’s government to ﬁnance the main public orphanage in 1753. Now the surplus is paid into the general revenue as an ordinary tax. It has an annual turnover of about € 1/2 billion. The fraction of GDP for this lottery has been slowly falling, and it projects a somewhat ‘grandmotherly’ image. The second is Danish Gambling. It was started as a soccer pool to ﬁnance soccer in 1948 and has gradually grown and sprouted new Table 8.2

Names in Danish and our translation

Unit

Danish name

Ownership

Gambling Authority

Spillemyndigheden

Legal control – unit of Tax Ministry

Danish State Lottery, Ltd

Det Danske Klasselotteri A/S

100% state-owned

Danish Gambling, Ltd. Owns (1)–(3) (1) Danish Pool Betting,Ltd Has subsidiaries for Lotto, etc (2) Danish Totalisator, Ltd (3) Danish Slot Machines, Ltd

Danske Spil A/S

State owns 80% of sharesa Old name kept for main company

Private slot machine operators Six casinos, in major hotels

Taxed, but easy to get permit

Dansk Tipstjeneste A/S

Dantoto A/S Dansk Automatspil A/S

Operates 3,000 slot machines b Operates 22,000 slot machines b Permitted in 1991

Notes: a. The remaining 20 per cent of the shares are owned by sports organisations, which are NGOs. b. The Danish population is 5,300,000 so there are 5,300,000/25,000 = 210 Danes per machine.

198 Gaming in the New Market Environment

branches. It increasingly competes with new/foreign gambling products. One division runs approximately 12 per cent of the Danish slot machines in pubs and game parlours. The total turnover is about €11/4 billion. A little more than half is the monopoly rent, of which most goes into a special G-fund (Tipsfonden) which supports sports and other good causes. The private gambling sector runs 22,000 slot machines of which many are in gambling arcades. Also, there are six ‘high end’ casinos with roulettes, card games and dress code. Also some private clubs and NGOs are allowed to run bingo and various lotteries. Finally, illegal gambling should be mentioned. It has two parts: The illegal grey/black sub-sector has been (almost) wiped out by the liberalisation and will be disregarded.11 Gambling abroad using the internet appears to be increasing fast, but it is still a minor part of gambling. Approximately 85 per cent of all Danes have internet access from their homes. Thus, it is easy to bet abroad, and nobody has yet been punished for doing it.12 8.4.2

A study of problem gambling

The Danish Institute of Social Research has recently completed a study of the prevalence of compulsive gambling; see Bonke and Borregaard (2006) and Bonke (2007). It is made as similar recent studies in Norway and Sweden. That is, they use the standard US questionnaire which has been translated and adjusted to ﬁt in a Scandinavian setting.13 The study operates with three categories: Compulsive gamblers, problem gamblers and others. It is based on interviews with a representative sample of 8,000 respondents. The main results are given in Table 8.3. As is obvious from the numbers, compulsive gambling is a small problem in Denmark. The category (1) is only 0.1 per cent of the population,14 and category (2) applies to only one part of life so it is much less at any one time. Table 8.3 studies

Categories and main results of Danish gambling dependency

Category

Deﬁnition

Fraction

(1) Compulsive gamblers

Life dominated by gambling

0.1%

(2) Problem gamblers

Major factor in part of life

1.5%

(3) Others

Gambling is no problem

Based on Bonke and Borregaard (2006).

98.4%

The Political Economy of Regulating Gambling 199

This allows us to assess the fraction of the population having the problem at any time. The average Dane lives 78 years. Imagine that he has a gambling problem for ﬁve of these years. This amounts to about 6 per cent of his life, and as the population pyramid is almost rectangular also to 6 per cent of the problem gamblers. This is 0.06•1.5% ≈ 0.1% that should be added the 0.1 per cent having a permanent problem. Thus, at any time the problem afﬂicts 0.2 per cent of the population. The study also contains summaries of detailed interviews with 453 persons, where half are problem gamblers (from the ﬁrst study) and the other half is a control group with the same socio-economic characteristics. It appears that the social deroute due to excessive gambling comes fairly late, and that many people with a problem still manage to live an almost normal life. For example, the data show that the divorce rate for problem gamblers is only marginally higher than that for the control group. 8.4.3

The price and the treatment observation, and a look at the other Nordic countries

Table 8.4 gives the estimates of the prices (as deﬁned in Table 8.1) of various gambling products in Denmark estimated by the Danish Gambling Authority. When these prices are compared with the potential of the various products for triggering the gambling problem (according to Grifﬁths

Table 8.4

Gambling prices for various products in Denmark

Products of the GSOE Public revenue mainly administrated via G-fund Problem a

Product

Price, λ

Problem a

Football betting, 55% Lotto

Small/weekly

Slot machines

18%

High

Scratch card games

30–50%

Small/small size

High end casinos

2–10%

High

Kendo, Oddsett

30–40%

Medium

Internet poker

2–5%

High

Product

Price, λ

Products of the private companies Tax revenue paid to general tax revenue

Notes: (a) From an unpublished report of the Gambling Authority cited in Bonke (2007). (b) The assessments are based on Grifﬁths (2007). The price structure is fairly unchanged the last 20 years, thus it is not a reaction to the starting international competition in the sector.

200 Gaming in the New Market Environment

2008), it appears that they have a negative correlation – products with the highest triggering potential have the lowest tax. This is termed the price observation. One group of NGOs (partly) ﬁnanced by the G-fund is the (handful of) institutions treating problem gamblers. While about €1 billion is collected in taxes from gambling, only about €4–5 million is used to treat the addicted. This is termed the treatment observation. Both the price and the treatment observation suggest that the pious purpose of reducing the gambling problem is secondary to the purpose of tax collection. The political economy of the explanation of these priorities will be discussed in Section 8.5. The estimated fraction of the gambling problem is only between 50 per cent and 70 per cent of the corresponding numbers from Norway and Sweden, even though the turnover in the gambling industry is relatively high in Denmark. Interestingly, the pattern of consumption corresponds to that of alcohol consumption. Danes drink more than other Scandinavians, but Denmark has a lower incidence of alcoholics (and alcohol-related diseases) than Norway and Sweden. The beverage sector is fully privatised in Denmark, and the alcohol tax has been lowered almost to the EU-level. It is sometimes argued that Denmark is a marginally more laid back society than its neighbours – at least this is what many Danes think. 8.4.4

The calculation: The cost of gambling in Denmark

In 2005/06 the turnover of the GSOE was about €1.75 billion, at a time when GDP is about € 200 billion, so GSOE gambling is a little less than 1 per cent of GDP. When the market sector is added, we reach about B = €4 billion or 2 per cent of GDP. Net gambling NG = €21/2 billion or 1.2 per cent of GDP. If we assume that the income line is horizontal as on Figure 8.4, we get a crude estimate of the two NG-areas. Section 8.4.2 estimated that N = NN + NP = 99.8% + 0.2% respectively. If the average compulsive gambler uses m = 50% of her (net of tax) income, equation (3.1) gives 100% = 99.8 • x + 0.2 • 50%, so that x ≈ 90%. It means that α = 0.1 (3.1b) Thus, about 10 per cent of the net gambling expenditure NG is in the problem area NGP. This is not, of course, a precise estimate, but it gives an order of magnitude. Problem gambling thus accounts for 10% ± 5% of the gambling. Using 10 per cent as the estimate, relation (3.2b) gives ⌫ ⬇ 1.5NBP = 1.5 • 0.1NB = 0.15 • 1.2 = 0.18, in % of GDP

(3.2c)

The Political Economy of Regulating Gambling 201

As some of the quantities are in the high end, we conclude Γ = 0.15 ± 0.05% of GDP or € 300 million

(4.1)

This means that the DWL and the NG triangles on Figure 8.4 are drawn far too big (for visibility as explained in the note to the ﬁgure). The NG triangle should only constitute 1.2 per cent of GDP. Hence, both the DWL and the NGP areas are far too big as well. Expressed in percentage of GDP, we are dealing with a small problem, and it is small if, for example, compared to the costs of alcohol. However, there are other ways to look at the size of the problem: 0.2 per cent of the population is 10,000 people, so at any one moment, about 10,000 Danes are afﬂicted by compulsive gambling. For them it is surely a large problem,15 which should be treated.

8.5

Political economy: The gambling monopoly as an SOE

No study has compared the cost structure in the GSOEs with the cost structure of comparable market companies. However, a large literature studies the relation between costs in SOEs and market ﬁrms in other ﬁelds – it is both based on theory and many empirical studies. Consequently, we know what to expect in the GSOE sector. That is what we have to believe in the absence of detailed studies of the sector. 8.5.1

The market vs. the SOE, the literature in a nutshell16

All organisations have the ‘internal’ goal of cost maximisation. That is, the employees want higher salaries, better fringe beneﬁts, and organisational growth. A market ﬁrm is forced to turn cost maximisation into minimisation by two mechanisms (F1) Competition punishes inefﬁciency. (F2) The owner is the residual claimant and the ultimate decision-maker. He thus has both a strong interest in and the means to enforce cost minimisation. It is much more uncertain and debatable what SOEs maximise. SOEs always have some degree of monopoly and cannot go bankrupt as easily as market ﬁrms – so (F1) is weak. The owners are people at large, as represented by the politicians. They are not the residual claimants, so also (F2) is weak. Thus, the standard result when comparable private and public ﬁrms are compared is that the average cost

202 Gaming in the New Market Environment

Frequency

ACM ACSOE

Market

SOE

Unit cost Figure 8.6

The distributions of unit costs for the typical market ﬁrm and SOE

difference is in the order of 20–30 per cent in developed countries. However, there is a fairly large variation in the costs, notably in the SOEs. The cost distributions found in studies with many cases are shown on Figure 8.6. The market enforces a fairly narrow (and symmetrical) distribution of the costs of the private ﬁrms, where the standard deviation of the distribution is often even smaller than the average. Variations in the market structure matters a great deal to the distribution. However, for the SOEs the distribution is asymmetrical with a long upward tail. Thus, the medians of the two distributions are not as different as the average. Though the political decision-makers – who represent the owners – are not the residual claimants, they have other, namely political goals. So basically, the SOEs have to observe these goals. One way in which the goals are presented to the SOEs is through their charter. However, there is also a board with representatives for the political interests, and a top management, which typically consists of (senior) ex-civil servants, who are well trained in the complex game of combining loyalty and politics. Finally, there is the threat of liberalisation and privatisation that some governments and the EU try to push.17 In the absence of studies of the sector, we have to believe that the usual result applies: BSOEs are a bit less efﬁcient than the market ﬁrms would be if the sector was privatised. However, due to the externality (of the gambling problem) efﬁciency is not the only goal.

The Political Economy of Regulating Gambling 203

8.5.2

Which goal wins? The case of the GSOE

The GSOE has the contradictory goals of problem prevention and tax collection. Compromises have to be made, and different owner representatives may differ on which part of the compromise should take ﬁrst priority. In addition to the ideal of maximising social welfare, there is the pressure from the groups with the two interests. When the relative power of the groups is assessed, the tax goal is likely to win for two reasons: The ﬁrst reason works even if there is no switching.18 It starts from the weak victim observation: The victims of the problem are few, they tend not to advertise their problem, and as they do not suffer from visible disabilities, it is relatively easy for them to hide. They are certainly not organised and hence constitute a politically weak group. On the other hand there is daily pressure on the political decision-makers to increase public expenditures and therefore they need taxes. The second reason operates if the tax revenue is administrated via a special fund, as it normally is for GSOEs. This creates switching and stakeholder interests: In most countries, the stakeholders are sport and culture organisations. In some countries, it is also the social NGOs, which are often associated with the churches. Due to switching, the tax goal is supported by large and well organised – and hence powerful – pressure groups. Consider organised sport, which has the support of millions of fans, and culture which is always well connected to the press and the upper crust of society. It seems obvious that only the toughest and most idealistic policy-maker has any choice but to bow to the big guns. The collection of revenue must win and make the prevention of problems a weak second. This is well in accordance with both the price and the treatment observation in Section 8.4.3. Also, it appears to be a general result. 8.5.3

SOEs as tax instrument: A general rule?

The literature on SOEs as tax instruments probably only amounts to hundred papers looking at special cases: An important case is natural resources, where economic theory suggests that the resource rent is the ideal tax subject as tax prevents the rent from being wasted in excess costs expended to capture the rents.19 As regards exhaustible resources, such as oil, some countries have chosen to keep the national wealth under public ownership (e.g. Mexico and Norway); while others (e.g., the USA and Denmark) prefer private operators, which are then taxed.

204 Gaming in the New Market Environment

However, it appears that the tax authorities are quite successful in extracting the rent regardless of the institutions used. Another much studied case is marketing boards in LDCs, where the classical study is Bates (1981, 2005). It shows how an institution set up by colonial administrations to protect poor farmers in the LDC world from the monopsony forces (typically multinational ﬁrms) in the market normally turns into tax instruments as the boards are taken over by the new national governments after independence. For long, the study did not receive much attention, but from approx. 1990 it led to a set of studies of the dynamics of SOEs, which are used as tax instruments when possible. Consequently, we know that the story of the BSEOs is rather typical. Once the SOE has tax collection as one of its goals, it tends to be the cuckoo chick in the nest.20 8.5.4

The symbiotic relation of the GSOEs, stakeholders and politicians

What makes the GSOEs a special case is that their history has generated a political consensus agreement between the GSOEs, stakeholders and politicians. In Denmark, it goes so far that the main stakeholders – the sports organisations – have been given 20 per cent of the shares of the GSOE to cement the consensus on the policies of the GSOE. Other institutional arrangements exist in other countries, but switching arrangements are common, and they have provided the GSOEs with politically strong stakeholders. This, of course, somewhat constrains the behaviour of the GSOEs, but it does provide them with a pressure group that gives some protection against liberalisation threats. Political decision-makers are experts at balancing interests. It is the key to the art of politics. Obviously there are a number of interests to be balanced in this case: powerful stakeholders, normal recreational gamblers, and victims of the gambling problem. These interests are clearly contradictory, so compromises have to be made. One part of many good compromises is to keep things somewhat unclear. As argued in the introduction, the GSOE is an instrument that allows both taxation and regulation to be softer and more informal. Thus, compromises are easier in this case. The complex political games surrounding the GSOEs give them an image problem. Are they an almost market ﬁrm with proﬁts and efﬁciency as the goal, or are they almost a social institution devoted to the dual purpose of taking the sin out of a morally dubious type of entertainment and at the same time ﬁnancing good deeds?

The Political Economy of Regulating Gambling 205

The home page and advertisements of the Danish GSOE mentions both good roles, but it puts more emphasis on looking like any other successful business.21 Certainly, in Denmark the GSOE tries hard to project the image that it is a modern ﬁrm behaving like any other ﬁrm. This is well in line with the demand for public efﬁciency pushed by the Ministry of Finance. 8.5.5

The effect of market form

So far the argument has assumed that the market solution is in perfect competition. In standard economic theory, this is the ideal that gives the highest short-run efﬁciency if externalities are regulated. This is a somewhat unfair comparison, as it compares an SOE in a highly politicised environment with an economic ideal. Let us imagine that gambling in the EU is liberalised and eventually privatised. As shown by Viren (2008), the enterprises in the gambling sector have rather marked economics of scale. Especially as regards lotteries, it is easy to understand. The graphs look rather like the corresponding graphs one can make for the companies in the medical drugs market and in fact for most sectors with modern mass production. Hence, it is fairly safe to predict that if the sector is privatised the ﬁrm structure will converge a situation where a handful of multinationals dominate and are engaged in (ﬁerce) monopolistic competition. Some theory sees this ﬁrm structure as optimal for innovation and technical progress, and thus as preferable in a long-run perspective.22 Consequently, it is not clear that economics of scale is a problem in a longer-run perspective. However, if the ﬁrm structure converges to full private monopoly, it is surely an inferior solution that has to be regulated by the EU competition authority.

8.6

Conclusion: What have we found?

We can now summarise the analysis of the trade-offs in this difﬁcult case. The article has data from Denmark to give orders of magnitudes. I hope it is sufﬁciently clearly explained how the calculations are made so that readers with a different national background, more knowledge or different priors than the author can amend the numbers. The basis for the analysis is a simple graphical model of the gambling problem. It is reached by sorting the population by gambling expenditure. This allows us to draw GDP as a rectangle. The top line

206 Gaming in the New Market Environment

is an income curve, and near the bottom is an upward bending net gambling expenditure curve. At the right side of the rectangle, the loss for society appears as two triangles. One is the top of the gambling expenditure curve that is a peak of excessive gambling due to compulsive gambling. It gives a welfare loss that is mainly a family externality. The other is the dead weight loss due to the downward bend in the income curve as addiction causes people to drop out of normal work. My intuition is that the two triangles are mirror images of each other. The (weak) evidence suggests that the dead weight loss triangle should be scaled down relative to the excess gambling triangle. The paper makes an effort to describe how the two triangles can be assessed, and to do so it uses data from Denmark. Here the analysis is perhaps a bit too pedestrian, but it should allow the reader to repeat the exercise for another country, and to improve the estimate if more information becomes available. This gives the size of the gambling problem of 0.2 ± 0.05 in percentage of the GDP. Consequently the problem is small. At any point in time, it affects about 0.2 per cent of the population or about 10,000 people in Denmark. However, about 1.5 per cent of the population experiences the problem during a period of their life. It is certainly an externality that should be reduced. For most people, gambling is thus innocent entertainment. Some do not gamble at all, but most do so to add a little spice to the enjoyment of spectator sports. The utility it produces (the consumer surplus) is considerable, and it is inconceivable that it is larger than the costs borne by the minority. So the solution to reducing gambling in general in order to reduce the gambling problem does not increase welfare in society. This proposes that other regulation should be made to reduce the problem. This can be done by regulating the product mix and the marketing methods of either the private business or the Gambling State Owned Enterprises – the GSOEs – that exist in most countries. The GSOEs are a product of history. When tax collection was much harder than now, many countries created gambling monopolies earmarked to ﬁnance speciﬁc expenditures. This has reduced many problems in the past, but it has also generated stakeholders who are often well organised and powerful groups, while the victims of the gambling problem are few and unorganised and thus politically weak. This has caused the goal of tax collection to dominate the goal of controlling the problem.

The Political Economy of Regulating Gambling 207

It thus appears that it would be better for both the welfare in general and the reduction of the problem if taxes on gambling were fully paid into the general revenue. This would allow the restrictions on gambling to be more focused on prevention and treatment of the gambling problem than on tax collection.

Notes 1 School of Economics and Management (building 1322), University of Aarhus, 8000 Aarhus C, Denmark, Phone: +45 8942 1607 or 08. E-mail: [email protected]. http://www.martin.paldam.dk. I am grateful to Matti Viren for discussions of this fascinating subject, and for a long conversation with Michael Bay Jørsel, who heads the largest rehabilitation institution for compulsive gambling in the country, and for assistance from Gunnar Viby Mogensen. Also, I have beneﬁted much from a day of discussions at the EL2007 meeting in Budapest. 2 The term gambling is used to cover all sorts of commercial betting and lotteries. 3 It is deﬁned as the difference between the spending patterns in two cases: (i) The gambling rent is paid into the general revenue, and (ii) the rent is distributed by the G-board. 4 If λQ = m, then Q = Q(λ) = m/λ. And the price elasticity is ∂Q λ = ∂(m/λ) = λ – m λ2 = –1 ∂λ Q ∂λ m/λ λ2 m 5 Here income is taken to be net of income taxes. 6 A note with the complete calculations is available from the author. 7 This formula is an approximation. The exact formula is: CS = (ln(1+t)– t/(1+t))m. The approximation is very good for small t’s, as the fault stays below 1 per cent of CS till the tax rate reaches 40 per cent. 8 A biography of Dostoyevsky that covers the amazingly strong relation between his life and work is Kjetsaa (1987). Dostoyevsky’s gambling addiction was so strong that his economy was always on the brink of collapse. 9 It is interesting to read the questionnaires used to identify problem gamblers. Most items actually try to catch one of the two G-limits shown on Figure 8.4. 10 Countries use different combinations of regulatory instruments and institutions, partly due to history, but also because solutions follow traditions used for other addictive goods. Alcohol is heavily taxed in all Nordic countries (lowest in Denmark). Only Denmark does not combine the tax with an SOE liquor retail monopoly. 11 The largest attempt to measure the grey sector in Denmark is from the Rockwool Foundation (see e.g. Mogensen, 1994), which concentrates on the effort to escape taxation. Grey is deﬁned as a normal activity, which is hidden, so that it leaves no paper trail for the tax man to follow. The estimates are at about 3–5 per cent of GDP. Gambling is not singled out and only poorly covered.

208 Gaming in the New Market Environment 12 If the computer on which the gambling program runs is placed in a location where it is legal, there is little the (Danish) authorities can do. It is even legal (in Denmark) to advertise gambling abroad. Likewise, cruise trips on boats with casinos and journeys to tourist resorts with casinos cannot be prohibited. Thus, there are complex issues involved in drawing a line and making some sorts of gambling abroad illegal. 13 The deﬁnition of compulsive gambling is from the DSM IV of the American Psychiatric Association. 14 The estimate of compulsive gamblers is based on 0.1 per cent of 8,000 respondents, i.e., eight persons. This is a small number and likely to be Poisson distributed. Thus, it cannot be a precise estimate. However, the 1.5 per cent of problem gamblers are 120 persons. In the detailed re-interviews, twice as many are covered, so the study (Bonke, 2007) covers people with a small gambling problem only. 15 This gives an average gambling problem of €30,000 per addict. It includes DWL of €10,000, which is not produced. The net DWL problem is hence €1,670 per month. This only applies in the worst cases. 16 See here World Bank (1996), the surveys by Megginson and Netter (2001, 2003), Parker and Saal (2003), Kötherburger et al (2006) and Christoffersen et al (2007). A few studies, notably Willner (2003, 2006) ﬁnd no efﬁciency difference between SOEs and comparable private ﬁrms. 17 The push from the EU to liberalise has already generated amazing results – in Scandinavia the most visible is that it has forced the public airline, SAS, to cut costs to half over a period of less than ﬁve years. 18 That is, even if all public revenue collected from gambling is paid into the general tax fund. 19 For renewable resources, such as ﬁsh, all countries keep the enterprises private, and then tax or regulate use. The choice is rarely in accordance with the recommendations of economists, but appears to be heavily inﬂuenced by a mixture of path dependency and politics. 20 Tax capture often happens even of instruments that are not used through a SOE. An example is green taxes introduced for the pious primary purpose of protecting the environment, and with a secondary purpose of tax collection. It appears that after some time, the second purpose normally wins (see Daugbjerg and Svendsen, 2001, for a survey of the evidence). 21 One is reminded of the Danish cigarette industry, which is an efﬁcient export industry, paying a high tax, publishing a warning on each package and giving half its surplus to foundations ﬁnancing art and science. 22 This is the prevailing view in the theory of endogenous growth, notably in the research started by Romer (1986). It also plays a prominent part in modern trade theory, see Chapters 5 and 10 in Feenstra (2004).

References Abbott, M.W. and Volberg, R. A. (1999) ‘Gambling and problem gambling in the community: an international overview and critique’. Report number one of the New Zealand gaming survey. Wellington: Department of Internal Affairs. Abbott, M. W. and Volberg, R. A. (2000) ‘Taking the pulse on gambling and problem gambling in New Zealand: Phase One of the 1999 National Prevalence Survey’. Report number three of the New Zealand Gaming Survey. Wellington: Department of Internal Affairs. Abbott, M., Volberg, R., Bellringer, M. and Reith, G. (2004) ‘A review of research on aspects of problem gambling’. Auckland University of Technology, Gambling Research Centre. London: Responsibility in Gambling Trust. Abbott, M. and Volberg, R. (2007) ‘Situational factors that affect gambling behaviour’. In: Smith, G., Hodgins, D. and Williams, R. (Editors) Research and measurement issues in gambling studies, 251–78. New York: Elsevier. American Psychiatric Association (1980) ‘Diagnostic and statistical manual of mental disorders (3e)’. Washington, DC. American Psychiatric Association (1987) ‘Diagnostic and statistical manual of mental disorders (3e – revised)’. Washington, DC. American Psychiatric Association (1994) ‘Diagnostic and statistical manual of mental disorders (4th ed.)’. Washington, DC. Anders G. C. (2003) ‘Reconsidering the economic impact of Indian casino gambling’. In The Economics of Gambling, Edited by Leighton Vaughan Williams. New York and London: Routledge. Anders, G. C. and Siegel, D. (1998) ‘An Economic Analysis of Substitution Between Indian Casinos and the State Lottery’, Gaming Law Review, 2(6): 609–13. Annual Report of Veikkaus (2006). Helsinki, Veikkaus. Atkinson, A. and Stiglitz, J. (1976) ‘The Design of Tax Structure: Direct Versus Indirect taxation’. Journal of Public Economics, 6: 55–75. Australian Gambling Comparative History and Analysis (1999) ‘Project Report by the Victorian Casino and Gaming Authority’, Australian Institute for Gambling Research at University of Western Sydney Macarthur; Melbourne. Australian Productivity Commission (1999) Australia’s Gambling Industries. Inquiry Report, Report No. 10, Melbourne, Commonwealth of Australia. Bailey, S. and Connolly, S. (1997) ‘The National Lottery: a preliminary assessment of net additionality’. Scottish Journal of Political Economy, 44(1): 100–12. Bates, R.H. (1981, 2005) Markets and States in Tropical Africa: The Political Basis of Agricultural Policies. Berkeley: University of California Press. Becker, G. S. and Murphy, K. M. (1988) ‘A Theory of Rational Addiction’. Journal of Political Economy, 96: 675–700. Becker, G. S., Grossman, M. and Murphy, K. M. (1988) ‘An Empirical Analysis of Cigarette Addiction’. American Economic Review, 84: 396–418. 209

210 References BERL (1997) ‘Sensitivity analysis of gross win to price elasticity of demand’. In Responsible gambling: A Commentary by the New Zealand Lotteries Commission on the Department of Internal Affairs’ proposal for Gaming and Gambling, contained in Gaming a New Direction for New Zealand and its associated Impact Reports, New Zealand Lotteries Commission, Wellington. Berndt, E. (1991) Econometrics: Classic and Contemporary. Reading MA: Addison Wesley. Bohn, J., Hancock, D. and Bauer, P. (2001) ‘Estimates of scale and cost efﬁciency for Federal Reserve Currency Operations’. Federal Reserve Bank of Cleveland Economic Review, No. 4, 2–26. Bonke, J. and Borregaard, K. (2006) ‘Ludomani I Danmark I. Udbredelsen af pengespil og problemspillere’. (Compulsive gambling in Denmark. Frequency of gambling and problem gamblers) SFI 06:12, København. Bonke, J. (2007) ‘Ludomani I Danmark II. Faktorer af betydning for spilleproblemet’. (Compulsive gambling in Denmark. Factors inﬂuencing the gambling problem). SFI 07: 14, København. Borg, M., Mason, P. and Shapiro, S. (1993) ‘The cross effects of lottery taxes on alternative state tax revenue’. Public Finance Quarterly, 21: 123–40. Borg, M. and Stranahan, H. (2004) ‘Some futures are brighter than others: The net beneﬁts received by Florida Bright Futures recipients’. Public Finance Review, 32: 105–26. Brinner, R. C. and Clotfelter, C. T. (1975) ‘An Economic Appraisal of State Lotteries’. National Tax Journal, 28: 395–404. Budd Report (2001) Government commissioned report on the gambling industry. www.culture.gov.uk/global/publications/archive_2001/gamb_rev_report.htm Christensen, L. and Greene, W. (1976) ‘Economics of scale in US Electric power generation’. Journal of Political Economy, 84: 655–76. Christoffersen, H., Paldam, M. and Würtz, A. (2007) ‘Public versus private production and economics of scale’. Public Choice, 130: 311–28. Clotfelter, C. and Cook, P. (1989) Selling Hope: State Lotteries in America. Cambridge MA, Harvard University Press. Clotfelter, C. and Cook, P. (1990) ‘On the economics of state lotteries’. Journal of Economic Perspectives, 4(4): 105–19. Clotfelter, C. T. and Cook, P. J. (1991) Selling hope: State Lotteries in America (paperback). Cambridge, MA: Harvard University Press. Clotfelter, C., Cook, P., Edell, J. and Moore, M. (1999) ‘State lotteries at the turn of the century: Report to the National Gambling Impact Study Commission’. Washington DC, US Government Budget Ofﬁce. Collins, D. and Lapsley, H. (2003) ‘The Social Costs and Beneﬁts of Gambling: An Introduction to the Economic Issues’. Journal of Gambling Studies, 19(2): 123–48. Cook, P. J. and Clotfelter, C. T. (1987) ‘Implicit Taxation in Lottery Finance’. National Tax Journal, XL: 533–46. Cook, P. J. and Clotfelter, C. T. (1993) ‘The Peculiar Scale Economics of Lotto’. The American Economic Review, 83(3). Cooper, S. and Cohn, E. (1994). ‘Equity and Efﬁciency of State Lotteries: Review Essay’, Economics of Education Review, 13: 355–62. Daghestani, A. N., Elenz, E. and Crayton, J. W. (1996) ‘Pathological gambling in hospitalised substance abusing veterans’. Journal of Clinical Psychiatry, 57: 360–3.

References 211 Daugbjerg, C. and Svendsen, G. T. (2001) Green Taxation in Question: Politics and Economic Efﬁciency in Environmental Regulation. London: Palgrave Macmillan. Department of Culture, Media and Sport (2006) A literature review and survey of statistical sources on remote gambling. London: DCMS. Depoorter, B. (2000) ‘Regulation of Natural Monopoly’. In Bouckaert, B. and De Geest, G. (Editors) Encyclopaedia of Law and Economics, 498–531. London: Cheltenham, Edward Elgar. Diamond, P. A. and Mirrlees, J. A. (1971) ‘Optimal taxation and public production: I and II’. American Economic Review, 61, 8–27 and 261–78. Donoghue, T. and Rabin, M. (1999) ‘Addiction and Self Control’, in Addiction: Entries and Exits. Edited by Jon Elster, New York: Russel Sage Foundation. Dostoyevsky, F. (1866) The Gambler. (A. R. MacAndrew, Trans.). New York: Norton. Douglas, A. (1995) British Charitable Gambling 1956–1994. London: Athlone. Dupuit, J. (1844) ‘De la mesure de l’utilité des travaux publics’, Annales des Ponts et Chaussées, s. II, 2nd semester, 332–375; English translation ‘On the Measurement of the Utility of Public Works’, by R.H. Barback, International Economic Papers 1952 (2): 83–110. Eadington, W. R. (2001) ‘The spread of casinos and their role in tourism development’. In Contemporary Issues, In Tourism Development, Edited by Pearce, D. G. and Bulter, R. W. New York: Routledge. Eadington, W. R. (2003) ‘Measuring Costs from Permitting Gaming: Concepts and Categories in Evaluation Gambling’s Consequences’. Journal of Gambling Studies, 19(2): 185–213. Elliott, D. and Navin, J. (2002) ‘Has riverboat gambling reduced state lottery revenue?’ Public Finance Review, 30: 235–47. Europe Economics (2004) The case for a single European gambling market. London: Europe Economics. Eurostat (2007) http://epp.eurostat.ec.europa.eu/extraction/getExport/en/theme2/ gov/gov_a_tax_ag. Evans, W. and Zhang, P. (2007) ‘The Impact of Earmarked Lottery Revenue on K–12 Educational Expenditures’. Education, Finance and Policy, 2(1): 40–73. Ezell, J. (1960) Fortune’s Merry Wheel: The Lottery in America. Cambridge MA: Harvard University Press. Fabre, G. (2003) Criminal Prosperity. London: Routledge Curzon. Farrell, L. (1993) ‘The Microeconometrics of Gambling in the UK’. MA Thesis, University of Manchester. Farrell, L. (1997) ‘The Economics of Lotteries’. PhD Thesis, Keele University. Farrell, L. (2008) ‘When welfare economics and gambling studies collide’. Chapter 2 in this volume. Farrell, L. and Forrest, D. (2008) ‘Measuring Displacement Effects across Gaming Products: A Study of Australian Gambling Markets’. Applied Economics, 40(1): 53–62. Farrell, L., Hartley, R., Lanot, G. and Walker, I. (2000) ‘The Demand for Lotto: the Role of Conscious Selection’. Journal of Business and Economic Statistics, 18: 228–41. Farrell, L., Morgenroth, E. and Walker, I. (1999) ‘A Time Series Analysis of UK lottery Sales: The Long Run Price Elasticity’. Oxford Bulletin of Economics and Statistics, 61: 513–26.

212 References Farrell, L. and Walker, I. (1999) ‘The Welfare Effects of Lotto: Evidence from the UK’. Journal of Public Economics, 72: 99–120. Feehan, P. and Forrest, D. (2007) ‘Distribution of UK National Lottery Grants across local authority areas’. Applied Economics Letters, 14(5): 361–5. Feenstra, R. C. (2004) Advanced International Trade. Princeton NJ.: Princeton University Press. Feldstein, M. (1999) ‘Tax Avoidance and the Deadweight Loss of the Income Tax’. The Review of Economics and Statistics, 81(4): 674–80. Fink, S., Marco, A. and Rork, J. (2004) ‘Lotto nothing? The budgetary impact of state lotteries’. Applied Economics, 36(21): 2357–67. Fink, S. and Rork, J. (2003) ‘The importance of self selection in casino cannibalization of state lotteries’. Economics Bulletin, 8(10), 1–8. Forrest, D. (2008) ‘Demand issues in the market for lotto and similar games’. Chapter 3 in this volume. Forrest, D. and Alagic, D. (2007) ‘The history of a lottery game that was seldom won’. Journal of Gambling Business and Economics, 1: 57–68. Forrest, D. and Gulley, O. D. (2006) ‘Participation and level of play in the UK National Lottery: correlation with spending on other “vices”’. Manuscript under review by journal. Forrest, D., Gulley, O. D. and Simmons, R. (2000) ‘Elasticity of demand for UK National Lottery tickets’. National Tax Journal, 53: 853–63. Forrest, D., Gulley, O. D. and Simmons, R. (2008) ‘The relationship between betting and lottery play: a high frequency time series analysis’. Forthcoming in Economic Inquiry. Forrest, D. and McHale, I. (2007) ‘Anyone for tennis (betting)?’ European Journal of Finance, 13(8): 751–68. Forrest, D., Simmons, R. and Chesters, N. (2002) ‘Buying a Dream: Alternative Models of Demand for Lotto’. Economic Inquiry, 40(3): 485–96. Garrett, T. (2001a) ‘The Leviathan Lottery? Testing the Revenue Maximization Objective of State Lotteries as Evidence for Leviathan’. Public Choice, 109 (1–2): 101–17. Garret, T. A. (2001b) ‘An International Comparison and Analysis of Lotteries and the Distribution of Lottery Expenditures’. International Review of Applied Economics, 15(2): 213–27. Garrett, T. A. and Sobel, R. S. (1999) ‘Gamblers favor skewness, not risk: Further evidence from United States’ Lottery games’. Economics Letters, 63. Gazel, R. (1998) ‘The Economic Impacts of Casino Gambling at the State and Local Levels’. Annals of the American Academy of Political and Social Science, 556: 66–84. Chantal, Y. and Vallerand, R. J. (1996) ‘Skill versus luck: a motivational analysis of gambling involvement’. Journal of Gambling Studies, 12: 407–18. Crawford, I. and Tanner, S. (1995) ‘Bringing it all back home: Alcohol taxation and cross border shopping’. Fiscal Studies, 16(2): 94–114. Grifﬁths, M. D. (1993) ‘Fruit machine gambling: The importance of structural characteristics’. Journal of Gambling Studies, 9: 101–20. Grifﬁths, M. D. (1994a) ‘The role of cognitive bias and skill in fruit machine gambling’. British Journal of Psychology, 85: 351–69. Grifﬁths, M. D. (1994b) ‘An exploratory study of gambling cross addictions’. Journal of Gambling Studies, 10: 371–84.

References 213 Grifﬁths, M. D. (1994c) ‘Co existent fruit machine addiction and solvent abuse in adolescence: A cause for concern?’ Journal of Adolescence, 17: 491–8. Grifﬁths, M. D. (1995) Adolescent gambling. London: Routledge. Grifﬁths, M. D. (1996) ‘Pathological gambling and its treatment’. British Journal of Clinical Psychology, 35: 477–9. Grifﬁths, M. D. (1999) ‘Gambling technologies: Prospects for problem gambling’. Journal of Gambling Studies, 15: 265–83. Grifﬁths, M. D. (2001a) ‘Gambling – An emerging area of concern for health psychologists’. Journal of Health Psychology, 6: 477–9. Grifﬁths, M. D. (2001b) ‘Internet gambling: Preliminary results of the ﬁrst UK prevalence study’. Journal of Gambling Issues, 5. Located at: http://www.camh.net/ egambling/issue5/research/grifﬁths_article.html Grifﬁths, M. D. (2002a) Gambling and gaming addictions in adolescence. Leicester: British Psychological Society, Blackwells. Grifﬁths, M. D. (2002b) ‘Internet gambling in the workplace’. In Anandarajan, M. and Simmers, C. (Editors) Managing web usage in the workplace: A social, ethical and legal perspective. Hershey, Pennsylvania: Idea Publishing, 148–67. Grifﬁths, M. D. (2003a) ‘Problem gambling’. The Psychologist: Bulletin of the British Psychological Society, 16: 582–4. Grifﬁths, M. D. (2003b) ‘Adolescent gambling: Risk factors and implications for prevention, intervention, and treatment’. In Romer, D. (ed.) Reducing adolescent risk: Toward an integrated approach, 223–38. London: Sage. Grifﬁths, M. D. (2003c) ‘Internet gambling: Issues, concerns and recommendations’. CyberPsychology and Behavior, 6: 557–68. Grifﬁths, M. D. (2004) ‘Betting your life on it’. British Medical Journal, 329: 1055–6. Grifﬁths, M. D. (2005a) ‘Remote gambling: Psychosocial aspects’. In: Remote Gambling (Westminster eForum Seminar Series), 11–20. London: Westminster Forum Projects Ltd. Grifﬁths, M. D. (2005b) ‘Gambling and crime in the workplace: Will it be a growing problem?’ Justice of the Peace, 169: 188–9. Grifﬁths, M. D. (2005c) ‘A “components” model of addiction within a biopsychosocial framework’. Journal of Substance Use, 10: 191–7. Grifﬁths, M. D. (2005d) ‘Online therapy for addictive behaviours’. CyberPsychology and Behavior, 8: 555–61. Grifﬁths, M. D. (2005e) ‘Does advertising of gambling increase gambling addiction?’ International Journal of Mental Health and Addiction, 3: 15–25. Grifﬁths, M. D. (2006) ‘Pathological gambling’. In Plante, T. (ed.) Mental Disorders of the New Millennium (Volume 1: Behavioral Issues), 73–98. New York: Greenwood. Grifﬁths (2008) ‘Problem Gambling and European Lotteries’. Chapter 6 in this volume. Grifﬁths, M. D. and Barnes, A. (2008) ‘Internet gambling: An online empirical study among student gamblers’. Forthcoming in International Journal of Mental Health and Addiction. Grifﬁths, M. D., Bellringer, P., Farrell-Roberts, K. and Freestone, F. (2001) ‘Treating problem gamblers: A residential therapy approach’. Journal of Gambling Studies, 17: 161–9. Grifﬁths, M. D. and Cooper, G. (2003) ‘Online therapy: Implications for problem gamblers and clinicians’. British Journal of Guidance and Counselling, 13: 113–35.

214 References Grifﬁths, M. D. and Delfabbro, P. (2001) ‘The biopsychosocial approach to gambling: Contextual factors in research and clinical interventions’. Journal of Gambling Issue, 5: 1–33. Located at: http://www.camh.net/egambling/issue5/ feature/index.html. Grifﬁths, M. D. and MacDonald, H. F. (1999) ‘Counselling in the treatment of pathological gambling: An overview’. British Journal of Guidance and Counselling, 27: 179–90. Grifﬁths, M. D. and Parke, J. (2002) ‘The social impact of internet gambling’. Social Science Computer Review, 20: 312–20. Grifﬁths, M. D. and Parke, J. (2003) ‘The environmental psychology of gambling’. In Reith, G. (ed.) Gambling: Who wins? Who loses? 277–92. New York: Prometheus Books. Grifﬁths, M. D., Parke, J. and Wood, R. T. A. (2002) ‘Excessive gambling and substance abuse: Is there a relationship?’ Journal of Substance Abuse, 7: 187–90. Grifﬁths, M. D., Parke, A., Wood, R. T. A. and Parke, J. (2006) ‘Internet gambling: An overview of psychosocial impacts’. Gaming Research and Review Journal, 27: 27–39. Grifﬁths, M. D., Scarfe, A. and Bellringer, P. (1999) ‘The UK National telephone Helpline – Results on the ﬁrst year of operation’. Journal of Gambling Studies, 15: 83–90. Grifﬁths, M. D. and Sutherland, I. (1998) ‘Adolescent gambling and drug use’. Journal of Community and Applied Social Psychology, 8: 423–7. Grifﬁths, M. D. and Wood, R. T. A. (2000). ‘Risk factors in adolescence: The case of gambling, video game playing and the internet’. Journal of Gambling Studies, 16: 199–225. Grinols, E. and Mustard, D. B. (2001) ‘Business Proﬁtability versus Social Proﬁtability: Evaluating Industries with Externalities, The Case of Casinos’. Managerial and Decision Economics, 22: 143–62. Golec, J. and Tamarkin, M. (1998) ‘Bettors love skewness, not Risk, at the horse track’. Journal of Political Economy, 106(1): 205–15. Gorman, W. M. (1980) ‘A Possible Procedure for Analysing Quality Differentials in the Egg Market’. The Review of Economic Studies, 47: 843–56. Gulley, O. D. and Scott, F. A. (1993) ‘The demand for wagering on state operated lottery games’. National Tax Journal, 45: 13–22. Guryan, J. and Kearney, M. S. (2005) ‘Lucky Stores, Gambling, and Addiction: Empirical Evidence from State Lottery Sales’. NBER Working Paper 11287. Haig, B. and Reece, B. (1985) ‘Explaining and predicting expenditure on, and revenue from legal gambling’. In Gambling in Australia edited by Caldwell, G., Haig, B., Dickerson, M. and Sylvan, L., Sydney: Croom Helm. Hansen, A. (2004) ‘Lotteries and State Fiscal Policy’. Tax Foundation Background Paper No. 46. Hartley, R. and Farrell, L. (2002) ‘Can Friedman Savage Utility Functions Explain Gambling?’ American Economic Review, 92(3): 613–24. Hausman, J. (1997) ‘Cellular Telephones, New Products, and the CPI’. Journal of Business and Economic Statistics, 17: 188–94. Hayer, T., Grifﬁths, M. D. and Meyer, G. (2005) ‘The prevention and treatment of problem gambling in adolescence’. In: Gullotta, T. P. and Adams, G. (Editors) Handbook of adolescent behavioural problems: Evidence based approaches to prevention and treatment, 467–86. New York: Springer.

References 215 Heady, C. (1993) ‘Optimal Taxation as a Guide to Tax Policy: A Survey’. Fiscal Studies, 14(1): 15–41. Henriksson, L. E. (2001) ‘Gambling in Canada: Some Insights for Cost Beneﬁt Analysis’. Managerial and Decision Economics, 22: 113–23. HM Revenues and Customs (2007a) Alcohol Factsheet. London: Her Majesty’s Stationary Ofﬁce. HM Revenues and Customs (2007b) Tobacco Factsheet. London: Her Majesty’s Stationary Ofﬁce. Home Ofﬁce report (1978) Royal Commission on Gambling: Final Report. London: Her Majesty’s Stationary Ofﬁce. Jacobs, D. R., Marston, A. R. and Singer, A. R., Widamon, K., Little, T. and Verzades, J. (1989) ‘Children of problem gamblers’. Journal of Gambling Behaviour, 5: 261–7. Jeanneret Druckman, K. (2004) ‘Taxation of E Gaming: A Brave New World’. In Cross Border Gambling on the Internet. 351–72. Research Conducted by the Swiss Institute of Comparative Law. Geneva, Schulthess. Johnson, E. V., O’Brian, R. and Shin, H. S. (1999) ‘A Violation of Dominance and the Consumption Value of Gambling’. Journal of Behavioural Decision Making, 12(1): 12–36. Jørsel, M. B. (2003) Ludomani – ikke ﬂere indsatser, tak! (Ludomania – no more bets please!) København, Gyldendal. Joskow, P. (2007) ‘Regulation of Natural Monopolies’. In Handbook of Law and Economics 2 (Polinsky, A. and S. Shavell, editors), 1227–348. Amsterdam: North Holland. Kahneman, D. (2003), ‘Maps of bounded rationality: psychology for behavioural economics’, The American Economic Review, 93(5): 1449–75. Kay, J. and King, M. (1990) The British tax system. 5th edition, London: Oxford University Press. Kearney, M. S. (2002) ‘State lotteries and consumer behaviour’. NBER WP 9330. Kearney, M. (2005) ‘State lotteries and consumer behavior’. Journal of Public Economics, 89: 2269–99. Kjetsaa, G. (1987) Fyodor Dostoyevsky. A writer’s life. (Translated from Norwegian). New York: Elisabeth Sifton Books/ Viking. Korn, D. (2000) ‘Expansion of gambling in Canada, implications for health and social policy’. Canadian Medical Association Journal, 163(1): 61–4. Korn, D. (2002) ‘TeenNet Gambling Project’. Paper presented at Discovery 2002 Conference, Responsible Gambling Council, Niagara Falls. Kwoka, J. (2005) ‘Electric Power distribution: economics of scale, mergers and restructuring’. Applied Economics, 37: 2373–86. Kötherburger, M., Sinn, H. W. and Whalley, J. (Editors) (2006) Privatization Experiences in the European Union. MIT Press: CESifo Seminar series. LaFleur’s (2004) ‘World Lottery Almanac Fast Facts’. URL: http://www.laﬂeurs.com La Fleur’s (2007) La Fleur’s World Lottery Almanac. Lancaster, K. (1971) Consumer Demand: A New Approach. Columbia University Press. Leiseur, H. R. (1998) ‘Costs and Treatment of Pathological Gambling’. Annals of the American Academy of Political and Social Science, 556: 153–71. Lesieur, H. (1988) ‘Altering the DSM III Criteria for pathological gambling’. Journal of Gambling Behaviour, 4: 38–47.

216 References Lesieur, H. R. and Blume, S. B. (1987) ‘The South Oaks Gambling Screen (SOG): A new instrument for the identiﬁcation of pathological gamblers’. American Journal of Psychiatry, 144: 1184–8. Lesieur, H. R. and Rothschild, J. (1989) ‘Children of Gamblers Anonymous members’. Journal of Gambling Behaviour, 5: 269–81. Littler, A. and Fijnaut, C. (Editors) (2006) The Regulation of Gambling: European and National Perspectives. Leiden; Martinus Nijhoff. London Economics (2006) ‘The case for state lotteries’. A Report for the State Lottery and Toto Association. Loughnan, T., Pierce, M. and Sagris, A. (1997) ‘The Maroondah Assessment Proﬁle for Problem Gambling (G Map&trade): A new direction in problem gambling counseling’. Paper presented at the Tenth International Conference on Gambling and Risk Taking, June Montreal, Canada. Mason, P. M., Shapiro, S. L. and Borg, M. O. (1989) ‘Gambling Tax Incidence for Three Groups of Las Vegas Gamblers’. Applied Economics, 1267–77. Matheson, V. and Grote, K. (2006) ‘Duelling jackpots: are competing lotto games complements or substitutes?’ Atlantic Economic Journal, 34: 85–100. Megginson, W. L. and Netter, J. M. (2001) ‘From state to market: A survey of empirical studies on privatization’. Journal of Economic Literature, 39: 321–89. Megginson, W. L. and Netter, J. M. (2003) ‘History and methods of privatization’. Chapter 3 in Parker and Saal (2003). Mikesell Zorn (1990) ‘The effect of maturity and competition on state lottery markets’. Journal of policy analysis and management, 6(2): 251–3. Mogensen, G. V. (1994) The shadow economy in Denmark 1994: measurement and results. Danmarks Statistik, København: Rockwool Foundation Research Unit. Morgan, J. (2000) ‘Financing public goods by means of lotteries’. Review of Economic Studies, 67: 761–84. Morgan, J. and Sefton, M. (2000) ‘Funding public goods with lotteries: experimental evidence’. Review of Economic Studies, 67: 785–810. MORI/International Gaming Research Unit (2006) Under 16s and the National Lottery. London: National Lottery Commission. NASPL (2007) North American Association of State and Provincial Lotteries data base. North American Association of State and Provincial Lotteries. URL: http://www.naspl.org/ (17.10.2005). National lottery Act (1993) London: Her Majesty’s Stationary Ofﬁce. National Gambling Impact Study Commission (1999) Final report. Washington, DC: Government Printing Ofﬁce. Available at http://govinfo.library.unt.edu/ ngisc/index.html. Novarro, N. (2002) ‘Does earmarking matter? The case of state lottery proﬁts and educational spending’. Stanford SIEPR Discussion Paper No. 02–19. Orford, J. (2002) ‘A nation of gamblers?’ The Psychologist, 15: 278–9. Orford, J., Sproston, K., Erens, B., White, C. and Mitchell, L. (2003) Gambling and Problem Gambling in Britain. Hove: Brunner Routledge. Paldam, M. (2008) ‘The political economy of betting monopoly: Illustrated with the Danish case’. Chapter 8 in this volume. Parke, A., Grifﬁths, M. D. and Irwing, P. (2004) ‘Personality traits in pathological gambling: Sensation seeking, deferment of gratiﬁcation and competitiveness as risk factors’. Addiction Research and Theory, 12: 201–12.

References 217 Parke, J. and Grifﬁths, M. D. (2006) ‘The psychology of the fruit machine: The role of structural characteristics (revisited)’. International Journal of Mental Health and Addiction, 4: 151–79. Parke, J. and Grifﬁths, M. D. (2007) ‘The role of structural characteristics in gambling’. In Smith, G., Hodgins, D. and Williams, R. (Editors) Research and Measurement Issues in Gambling Studies, 211–43. New York: Elsevier. Parker, D. and Saal, D. (Editors) (2003) International Handbook on Privatization. Cheltenham: Edward Elgar. Paton, D., Siegel, D. S. and Vaughan Williams, L. (2001) ‘Gambling Taxation: A Comment’. The Australian Economic Review, 34: 437–40. Paton, D., Siegel, D. S. and Vaughan Williams, L. (2003) ‘The Demand for Gambling: A Review’. In The Economics of Gambling, Edited by Vaughan Williams, L. London: Routledge. Paton, D., Siegel, D. S. and Vaughan Williams, L. (2004) ‘A time series analysis of the demand for gambling in the United Kingdom’. National Tax Journal, 57: 847–61. Petrin, A. (2002) ‘Estimating the Consumer Surplus and Welfare Gains from the Introduction of Minivans’. Journal of Political Economy, 110: 705–29. Productivity Commission (1999) ‘Australia’s Gambling Industry’. Report No. 10, Canberra: AusInfo. Ramsey, F. P. (1927) ‘A contribution to the theory of taxation’. Economic Journal, 37: 47–61. Romer, P. M. (1986) ‘Increasing returns and long run growth’. Journal of Political Economy, 94: 1002–37. Rosenthal, R. J. (1989) ‘Compulsive gambling’. Paper presented at the California Society for the Treatment of Alcoholism and Other Drug Dependencies, November, San Diego. Rosenthal, R. J. and Lesieur, H. R. (1992) ‘Self reported withdrawal symptoms and pathological gambling’. American Journal of the Addictions 1: 150–4. Rubinstein, R. and Scaﬁdi, B. (2002) ‘Who pays and who beneﬁts? Examining the distributional consequences of the Georgia Lottery for education’. National Tax Journal, LV 2: 223–38. Sauer, R. (2001) ‘The Political Economy of Gambling Regulation’. Managerial and Decision Economics, 22: 5–15. Scott, F. and Garen, J. (1993) ‘Probability of purchase, amount of purchase and the demographic incidence of the Lottery tax’. Journal of Public Economics, 54: 121–43. Setness, P. A. (1997) ‘Pathological gambling: When do social issues become medical issues?’ Postgraduate Medicine, 102: 13–18. Sevigny, S., Cloutier, M., Pelletier, M. and Ladouceur, R. (2005) ‘Internet gambling: Misleading payout rates during the “demo” period’. Computers in Human Behaviour, 21: 153–8. Shaffer, H. J., Hall, M. N. and Vander Bilt, J. (1999) ‘Estimating the prevalence of disordered gambling behaviour in the United States and Canada: A research synthesis’. American Journal of Public Health, 89: 1369–76. Shaffer, H. J. and Korn, D. A. (2002) ‘Gambling and Related Mental Health Disorders: A Public Health Analysis’. Annual Review of Public Health, 23: 171–212. Shaffer, H. J., LaBrie, R. A. and LaPlante, D. (2004) ‘Laying the foundation for quantifying regional exposure to social phenomena: considering the case of legalized gambling as a public health toxin’. Psychology of Addictive Behaviors, 18: 40–8.

218 References Siegel, D. and Anders, G. (2001) ‘The impact of Indian casinos on state lotteries: a case study of Arizona’. Public Finance Review, 29: 139–47. Simon, J. (1998) ‘Dreams and Disillusionment: A Dynamic Model of Lottery Demand’. In Four essays and a note on the demand for lottery tickets and how lotto players choose their numbers. Florence: European University Institute. Sing, M. (1987) ‘Are combination gas and electric utilities multiproduct natural monopolies?’ The Review of Economics and Statistics, 69: 392–8. Skala, J. (2004) ‘Money Laundering and Internet Gambling: A Suspicious Afﬁnity’? In Cross Border Gambling on the Internet, 305–48. Research Conducted by the Swiss Institute of Comparative Law. Schulthess: Geneva. Smeaton, M. and Grifﬁths, M. D. (2004) ‘Internet gambling and social responsibility: An exploratory study’. CyberPsychology and Behavior, 7: 49–57. Smith, J. (2000) ‘Gambling taxation: Public equity in the gambling business’. Australian Economic Review, 33: 120–44. Smith, S. (2008) ‘Lotteries as a source of revenue’. Chapter 5 in this volume. Spindler, C. (1995) ‘The lottery and education, robbing Peter to pay Paul?’ Public Finance and Budgeting, 15(3): 54–61. Sproston, K., Erens, B. and Orford, J. (2000) Gambling Behaviour in Britain, Results from the British Gambling Prevalence Survey. London: National Centre For Social Research. Stranahan, H. and Borg, M. O. (1998) ‘Horizontal equity implications of the lottery tax’. National Tax Journal, 51: 71–88. Suits, D. (1977a) ‘Measurement of Tax Progressivity’. The American Economic Review, 67(4): 747–52. Suits, D. B. (1977b) ‘Gambling Taxes: Regressivity and Revenue Potential’. National Tax Journal, 30(1): 19–35. Suits, D. B. (1979) ‘The Elasticity of Demand for Gambling’. Quarterly Journal of Economics, 93(1): 155–62. Sung, N. and Gort, M. (2000) ‘Economics of scale and natural monopoly in the U.S. local telephone market’. The Review of Economics and Statistics, 82: 694–7. Szakmary, A. and Szakmary, C. (1995) ‘State lotteries as a source of revenue: A reexamination’. Southern Economic Journal, 16(4): 1167–81. Swiss Institute (2006) Study of gambling services in the internal market of the European Union: Study carried out by the Swiss Institute of Comparative Law (dated April 24, 2006). Teeson, M., Degenhardt, L. and Hall, W. (2002) Addictions. New York: Taylor & Francis. Thompson, W. N., Gazel, R. C. and Rickman, D. (1996) ‘Social and legal costs of compulsive gambling’. Gaming Law Review, 1: 81–9. Train, K. (1995) Optimal Regulation: The Economic Theory of Natural Monopoly. Cambridge MA: The MIT Press. Vasche, J. D. (1985) ‘Are taxes on lotteries too high?’ Journal of Policy Analysis and Management, 4: 269–71. Viren, M. (2008) ‘The economics of scale and scope in the lottery industry’. Chapter 7 in this volume. Viscusi, W. K. (1994) ‘Cigarette Taxation and the Social Consequences of Smoking’, NBER Working Paper No. W4891. Viton, P. (1981) ‘A translog cost function for urban bus transit’. Journal of Industrial Economics, 29: 287–304.

References 219 Volberg, R. A. (1996) ‘Prevalence studies of problem gambling in the United States’. Journal of Gambling Studies, 12: 111–28. Volberg, R. A. (2001) When the chips are down: Problem gambling in America. New York, NY: The Century Foundation. Von Neuman, J. and Morgernstern, O. (1944) Theory of Games and Economic Behaviour. New York: Princeton University Press. Walker, I. (1998) ‘The Economic Analysis of Lotteries’. Economic Policy, 13(27). (Oct., 1998), 357–401. Walker, D. M. (2003) ‘Methodological Issues in the Social Cost of Gambling Studies’. Journal of Gambling Studies, 19(2): 149–84. Walker, D. (2007) The Economics of Casino Gambling. New York: Springer. Walker, D. M. and Barnett, A. H. (1999) ‘The Social Costs of Gambling: An Economic Perspective’. Journal of Gambling Studies, 15(3): 181–212. Walker, D. M. and Jackson, J. D. (1998) ‘New Goods and Economic Growth: Evidence from Legalised Gambling’. Review of Regional Studies, 28: 47–69. Walker, D. and Jackson, J. D. (2008) ‘Do U.S. gambling industries cannibalize each other?’ Public Finance Review, forthcoming. Walker, I. and Young, J. (2001) ‘An economist’s guide to lottery design’. Economic Journal, 111: F700–22. Walker, M. B. (1992) The Psychology of Gambling. New York, NY: Permagon. Wang, T. C., Hung, K. T. and Jy, T. J. (2006) ‘Evaluating Game Industry Operational Efﬁciency with DEA Method’. Unpublished, I Shou University. Wayne, H., Smith, G. and Jacobs, D. (1996) ‘Adolescent gambling and problem gambling in Alberta’. Edmonton, Alberta, Canada, Alberta Alcohol and Drug Abuse Commission. Waters, T. M. and Sloan, F. A. (1995) ‘Why Do People Drink? Tests of the Rational Addiction Model’. Applied Economics, 72: 727–36. Wilkinson, M. (1994) ‘Paying for public spending: is there a role for earmarked taxes?’, Fiscal Studies, 15(4): 119–35. Willner, J. (2003) ‘Privatization: a skeptical analysis’. Chapter 4 in Parker and Saal (2003). Willner, J. (2006) ‘Privatization and public ownership in Finland’. Chapter 5 in Kötherburger et al, 2006. Wood, R. T. A. and Grifﬁths, M. D. (1998) ‘The acquisition, development and maintenance of lottery and scratchcard gambling in adolescence’. Journal of Adolescence, 21: 265–73. Wood, R. T. A. (2004) ‘Adolescent lottery and scratchcard players: Do their attitudes inﬂuence their gambling behaviour?’ Journal of Adolescence, 27: 467–75. Wood, R. T. A. and Grifﬁths, M. D. (2007) ‘Online guidance, advice, and support for problem gamblers and concerned relatives and friends: An evaluation of the Gam Aid pilot service’. British Journal of Guidance and Counselling, in press. Wood, R. T. A., Grifﬁths, M. D. and Parke, J. (2007) ‘The acquisition, development, and maintenance of online poker playing in a student sample’. CyberPsychology and Behaviour, 10: 354–61. World Bank (1996) ‘Bureaucrats as managers. Bureaucrats in Business: The Economics and Politics of Government Ownership’. World Bank Policy Research Reports. Oxford and New York: Oxford U.P.

220 References Worthington, A. (2001) ‘Implicit Finance in Gambling Expenditures: Australian Evidence on Socioeconomic and Demographic Tax Incidence’. Public Finance Review, 29: 326–42. Yeoman, T. and Grifﬁths, M. D. (1996) ‘Adolescent machine gambling and crime’. Journal of Adolescence, 19: 183–8.

Index addiction, 20, 47, 49, 52–3, 97, 131, 184, 186, 206, 207 irrationality of addiction, 190–6 machine addiction, 137 rationalising addiction, 32–8 research on, 148–53 adolescent gambling, 133, 135, 136 ad valorem, 103 ad valorem tax, 27 advance payments, 10 advertising, 11, 100, 120, 146, 149, 153, 163, 170 American Psychiatric Association, 52, 127, 131, 133, 138, 139, 144, 155, 208 ARCH test, 96 Australian gambling market, 27, 45 average costs, 19, 161, 163, 170 Becker and Murphy’s model, 35 bingo, 71–2, 102, 128, 156, 198 blackjack, 128, 139 bookmakers, 1, 54, 70, 110, 162 British Social Attitudes Survey, 120–2 cannibalisation, 14, 15, 45, 46, 66, 69, 73, 180 Camelot, 68, 87 casino, 15, 50, 53, 127–9, 133, 160, 181, 197–9 attitudes towards, 5 casinos and growth, 44–6 casinos and lotto, 70, 73, 74 casinos and problem gambling, 127–9, 133–4, 139, 142, 146–7, 149 history of, 4, 55, 197–9 casino games, 33, 146 charitable purposes, 117 charities, 71, 99, 116, 117, 121, 125, 152 churches, 203 commissions, 164

competition of lotteries, 160 compulsive gamblers, 138, 198, 208 compulsive gambling, 190, 198, 201 conscious selection, 37, 38, 78, 87, 94–6 consumer surplus, 8, 47, 51, 52, 181, 188–9, 206 consumer surplus and taxes, 108 measurement of, 26–32, 34–5, 38, 39 crime, 9, 20 cross-border sales, 1–2 cross-price elasticity of demand, 71 Danish Gambling, 186, 197–9 DEA, 181 dead weight loss, 20, 191, 194, 195, 206 demand for lottery products, 176 demand curve, 26, 28–32, 59, 186–90 demand elasticity, 60 DSM-III, 138–9 DSM-III-R, 138 DSM-IV, 127, 130, 133, 134, 138–9, 155 earmarking, 6, 117–20 economic growth, 43, 44 economics of gambling, 20, 56, 97 economies of scale see also scale economies, 161 economies of scope, 161 effective price, 28, 31, 59, 71, 72, 74 elastic demand, 40, 60, 71, 108, 179, 188 electronic cash, 140, 142 EuroMillions, 3, 67, 68, 69 European Court of Justice, 7, 54 European State Lotteries and Toto Association, 22, 162 equity, 38, 40, 48, 49, 51, 99, 104, 106, 107 excess gambling expenditures, 191

221

222 Index excessive gambling, 38, 40, 47, 107, 135, 145, 146, 149, 182, 199 excise duties, 103 expected utility framework, 33, 53 externality, 16, 107, 115, 119, 123, 184, 185, 189, 194, 202 football pools, 3, 62, 110, 129, 131, 132 gambling prevention initiatives, 150 gambling revenues, 22, 48, 102 gambling taxes, 27, 48–50 game design, 36, 37, 76, 79, 80, 83, 86, 90, 92–4, 114 good causes, 17, 22, 50, 99, 101, 113, 116–25, 198 GamAid, 152 Gam-Anon, 152 Gamblers Anonymous, 152 gambling legislation, 48, 131, 149 Gambling State Owned Enterprise, 185, 192, 206 GamCare, 152 game design, 36, 37, 76, 79, 80, 83, 86, 90, 92–4, 114 Gaming and Wagering Business, 162 gate-keeping, 141 gender, 34, 37, 135, 142 green taxes, 208 gross tax rate, 100, 101, 102 Gulley-Scott model, 58–63 harm based conception, 130 history of lotteries, 4, 22, 48, 67, 88 horse racing, 22, 49, 127 illegal gambling, 48, 198 income level, 165, 170, 177 Indian Reverse gaming, 24 instant games, 3, 164, 173, 182 Instrumental Variable estimator, 178 Internet, 1–3, 14, 18, 54–5, 70, 176, 198 and problem gambling, 127, 129, 133, 139–43, 149, 151–2 internet gaming, 2, 70, 73, 74 internet prevention, 151

intertemporal substitution, 81, 93 investment, 11, 34, 55, 150 jackpot, 17, 18, 68–9, 95, 96, 133, 147 jackpot and demand for lotto, 58, 60, 62–4, 72–4, 76, 89–90, 120–1 jackpot and prizes, 78–9, 82–4, 87, 98, 115 jackpots, 3, 56, 57, 60, 68, 69, 74, 147 jackpot level, 78 labour supply, 108 LAD estimator, 166 Least Squares estimator, 166 legalisation of gambling, 39 Lorenz curves, 53, 109, 111, 124 lotteries as source of revenue, 99 lottery advertising, 120 lottery spending, 73, 74, 103–6, 111, 113, 118, 119 lottery tax, 6, 10, 12, 16–17, 27, 50, 103, 108, 113, 123–4 lottery ticket, 28, 62, 66, 72, 75–6, 80–1, 85, 87, 95 Lotto, 14–15, 37, 74, 76–84, 87–8, 90, 94, 98, 128, 131, 134, 173, 176–8, 182 and consumer surplus, 27, 29 demand for, 54, 56–7, 60–72 in Denmark, 197, 199 expenditure, 45 Lotto companies, 1, 3, 197, 199 Lottomatica, 9, 176 marginal costs, 8, 161–4, 173–6, 182 marketing strategies, 147 mean-variance-skewness approach, 65 minimum age, 149 mobile phones, 140 moments, 15, 16, 64, 77–80, 83–6, 88–90, 93–6 Monday Lottery, 125 money laundering, 9, 10 multistate games, 56 National Lottery, 22, 24, 57, 67–8, 128, 131, 134 and good causes, 120, 121, 125 history, 50, 52, 53

Index 223 National lottery act, 124 revenues, 100–3, 106, 108–11, 113, 114 natural monopoly, 8–9, 19, 161, 180, 181 new generation demand models, 61 New York State Lottery, 9, 176 number of outlets, 149 numbers games, 24, 37, 72, 173, 182 off-track betting, 156 online customer tracking, 142 operating costs, 164 own price elasticity of demand, 59, 71 panel data, 89, 163 pari-mutual lottery, 80 participation, 8, 30, 33, 34, 47, 48, 68, 71–2, 148 cost of, 83 and earmarking, 119–21 participation of lotteries, 120 pathological features, 18, 137 pathological gamblers, 30–4, 36–8, 42, 47, 50–2, 136–9, 143, 144, 150 deﬁnition of, 127, 130, 131 pay-out, 63, 65 per capita sales, 177 political economy, 12, 19, 181, 184–6, 200–1 Powerball, 3, 69, 173, 182 price elasticity see also demand elasticity, 14, 28, 29, 70, 71, 107, 114, 164, 178, 186, 187, 196, 207 price of gambling, 191, 192 price structure, 199 prize distribution, 15, 16, 37, 76–80, 83–6, 89, 90, 94, 98 prize pools, 59, 61, 62, 76, 91, 98 problem gamblers see also pathological gamblers, 7–8, 49, 52, 144, 146, 191–2, 198–200 deﬁnition of, 132–7 and internet, 140–1, 143 legislation and prevention, 149–54 problem gamblers’ demand behaviour, 30–2, 35–7, 186

problem gambling, 7, 8, 35–6, 46–7, 107, 193, 198, 200 deﬁnition of, 126–7, 130–7, 192 and internet, 139–40 and other disorders, 143, 144 policy implications, 147–53 proﬁling, 132 progressive tax, 109 probabilities of winning, 62 promotions, 93 public ﬁnance, 6, 22 public ownership, 11 public policy, 8–10, 12, 19–21, 131, 181 Ramsey, 17, 107, 108, 123, 217 Ramsey rule, 17, 107, 108, 123 rational addiction, 35, 53, 209 regressive taxes, 49 regressivity, 50, 93, 109, 110, 113, 124 regulation, 10 regulatory operations, 181 remote gambling, 18, 127, 139–43, 211, 213 RESET test, 96 revenue maximisation, 16, 59, 99, 100, 114, 119, 123 revenues, 5–7, 17, 24, 28, 31, 44, 45, 48, 50, 123–4 comparisons, 100–6 and earmarking, 113 and good causes, 116–20 maximization of, 114 rollover, 15, 28, 29, 37, 57–9, 61, 63–9, 74, 76–84, 87, 88, 90–6, 98, 114–16 roulette, 127, 128, 139, 198 Saturday Lotto, 57, 64–7, 69, 87, 94 scale economies, 19, 68, 163, 172 scratchcards, 18, 56, 71, 72, 88, 128, 131, 133–6, 148 serial correlation, 78, 80–1, 86–7, 89, 90, 95 skewness, 15, 37, 63–8, 77–9, 83–6, 89–92, 212, 214 slot machines, 18, 22, 45, 73, 127, 129, 131–7, 142, 145–7, 156

224 Index social context, 17, 131 social cost, 13, 14, 38–43, 46–9, 53, 107, 184 speciﬁcation tests, 96 sports betting, 1, 3, 127, 128, 173, 182 spread betting, 129 stakeholders, 20, 131, 184–6, 191, 203, 204, 206 state ﬁnance, 44 subadditivity, 161 Superdraw, 57, 64, 67, 74, 96 syndicates, 87 take-out, 15–16, 54–61, 63–6, 71, 76, 81, 85, 92, 93, 99, 101, 114, 116 taxation, 2, 5–6, 12, 17, 27, 48–50, 99, 114, 118, 189 and welfare, 107–9 taxation of winnings, 12, 22 tax-collection costs, 106 tax incidence, 27, 49, 50, 124 tax on gambling, 188 tax rates, 2, 12, 16, 29, 99–103, 105, 114, 123 Thunderball, 65, 67, 69, 87

ticket price, 120 total sales, 1, 9, 15, 21, 95, 104, 163, 165, 166, 170, 180 trigger/spree-mechanism, 190 UK National Lottery see also National lottery, 28, 50, 52, 53, 94, 99, 101, 103 unbalanced panel, 89 Veikkaus, 3, 22, 180 Viking lotto, 3 weak victim observation, 185, 203 wealth redistribution, 40 weekly games, 148 welfare, 13–14, 33, 38, 40–4, 114, 185, 191, 206 and addiction, 32–3 measurement of, 23–8, 35–6 and taxation, 108–9, 188–9 and welfare, 48, 51 welfare economics, 12–14, 25, 32, 38, 40–2, 48, 51–2, 97 youth gambling, 18, 135