Informal Employment in Emerging and Transition Economies 9781780527871, 9781780527864

Citation preview

INFORMAL EMPLOYMENT IN EMERGING AND TRANSITION ECONOMIES

RESEARCH IN LABOR ECONOMICS Series Editor: Solomon W. Polachek IZA Co-Editor: Konstantinos Tatsiramos Volume 23: Volume 24:

Volume 25: Volume 26: Volume 27:

Volume 28:

Volume 29:

Volume 30:

Volume 31:

Volume 32:

Volume 33:

Accounting for Worker Well-Being Edited by Solomon W. Polachek The Economics of Immigration and Social Diversity Edited by Solomon W. Polachek, Carmel Chiswick and Hillel Rapoport Micro-Simulation in Action Edited by Olivier Bargain Aspects of Worker Well-Being Edited by Solomon W. Polachek and Olivier Bargain Immigration: Trends, Consequences and Prospects for The United States Edited by Barry R. Chiswick Work Earnings and Other Aspects of the Employment Relation Edited by Solomon W. Polachek and Konstantinos Tatsiramos Ethnicity and Labor Market Outcomes Edited by Amelie F. Constant, Konstantinos Tatsiramos and Klaus F. Zimmermann Jobs, Training, and Worker Well-Being Edited by Solomon W. Polachek and Konstantinos Tatsiramos Child Labor and the Transition Between School and Work Edited by Randall K. Q. Akee, Eric V. Edmonds and Konstantinos Tatsiramos Who Loses in the Downturn? Economic Crisis, Employment and Income Distribution Edited by Herwig Immervoll, Andreas Peichl and Konstantinos Tatsiramos Research in Labor Economics Edited by Solomon W. Polachek and Konstantinos Tatsiramos

RESEARCH IN LABOR ECONOMICS VOLUME 34

INFORMAL EMPLOYMENT IN EMERGING AND TRANSITION ECONOMIES EDITED BY

HARTMUT LEHMANN Department of Economics, University of Bologna and IZA

KONSTANTINOS TATSIRAMOS Department of Economics, University of Leicester and IZA

United Kingdom – North America – Japan India – Malaysia – China

Emerald Group Publishing Limited Howard House, Wagon Lane, Bingley BD16 1WA, UK First edition 2012 Copyright r 2012 Emerald Group Publishing Limited Reprints and permission service Contact: [email protected] No part of this book may be reproduced, stored in a retrieval system, transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without either the prior written permission of the publisher or a licence permitting restricted copying issued in the UK by The Copyright Licensing Agency and in the USA by The Copyright Clearance Center. No responsibility is accepted for the accuracy of information contained in the text, illustrations or advertisements. The opinions expressed in these chapters are not necessarily those of the Editor or the publisher. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978-1-78052-786-4 ISSN: 0147-9121 (Series)

CONTENTS LIST OF CONTRIBUTORS

vii

EDITORIAL ADVISORY BOARD

ix

PREFACE

xi

CHAPTER 1 TAX EVASION, MINIMUM WAGE NONCOMPLIANCE, AND INFORMALITY Arnab K. Basu, Nancy H. Chau and Zahra Siddique CHAPTER 2 THE EFFECT OF TAXATION ON INFORMAL EMPLOYMENT: EVIDENCE FROM THE RUSSIAN FLAT TAX REFORM Fabia´n Slonimczyk

1

55

CHAPTER 3 WHO BENEFITS FROM REDUCING THE COST OF FORMALITY? QUANTILE REGRESSION DISCONTINUITY ANALYSIS Tommaso Gabrieli, Antonio F. Galvao Jr. and 101 Gabriel V. Montes-Rojas CHAPTER 4 DETECTING WAGE UNDER-REPORTING USING A DOUBLE-HURDLE MODEL + Bala´zs Reizer 135 Pe´ter Elek, Ja´nos Ko¨llo, and Pe´ter A. Szabo´ CHAPTER 5 DOES FORMAL WORK PAY? THE ROLE OF LABOR TAXATION AND SOCIAL BENEFIT DESIGN IN THE NEW EU MEMBER STATES Johannes Koettl and Michael Weber 167 v

vi

CONTENTS

CHAPTER 6 MIGRATION AS A SUBSTITUTE FOR INFORMAL ACTIVITIES: EVIDENCE FROM TAJIKISTAN Ilhom Abdulloev, Ira N. Gang and John Landon-Lane

205

CHAPTER 7 THE PERSISTENCE OF INFORMALITY: EVIDENCE FROM PANEL DATA Alpaslan Akay and Melanie Khamis

229

CHAPTER 8 JOB SEPARATIONS AND INFORMALITY IN THE RUSSIAN LABOR MARKET Hartmut Lehmann, Tiziano Razzolini and 257 Anzelika Zaiceva

LIST OF CONTRIBUTORS Ilhom Abdulloev

Department of Economics, Rutgers University, New Brunswick, NJ, USA

Alpaslan Akay

Institute for the Study of Labor (IZA), Bonn, Germany

Arnab K. Basu

Department of Economics, College of William and Mary, Williamsburg, VA, USA and IZA, Bonn, Germany

Nancy H. Chau

Charles H. Dyson School of Applied Economics and Management, Cornell University, Ithaca, NY, USA and IZA, Bonn, Germany

Pe´ter Elek

Department of Economics, Eo¨tvo¨s Lora´nd University, Budapest, Hungary

Tommaso Gabrieli

Henley Business School, University of Reading, Berkshire, UK

Antonio F. Galvao, Jr.

Department of Economics, University of Wisconsin-Milwaukee and University of Iowa

Ira N. Gang

Department of Economics, Rutgers University, New Brunswick, NJ, USA and IZA, Bonn, Germany

Melanie Khamis

Department of Economics, Wesleyan University, Middletown, CT, USA and IZA, Bonn, Germany

Johannes Koettl

Europe and Central Asia Region – Human Development Sector, World Bank, Washington, DC, USA and IZA, Bonn, Germany vii

viii

LIST OF CONTRIBUTORS

Ja´nos Ko¨llo+

Institute of Economics of the Hungarian Academy of Sciences, Budapest, Hungary and IZA, Bonn, Germany

John Landon-Lane

Department of Economics, Rutgers University, New Brunswick, NJ, USA

Hartmut Lehmann

Department of Economics, University of Bologna, Bologna, Italy and IZA, Bonn, Germany

Gabriel V. Montes-Rojas

Department of Economics, City University London, UK

Tiziano Razzolini

Department of Economics, University of Siena, Italy

Bala´zs Reizer

Central European University, Budapest, Hungary

Zahra Siddique

Institute for the Study of Labor (IZA), Bonn, Germany

Fabia´n Slonimczyk

International College of Economics and Finance, Higher School of Economics, Moscow

Pe´ter A. Szabo´

Reformed Presbyterian Church of Central and Eastern Europe, Budapest, Hungary

Michael Weber

Human Development Network – Social Protection and Labor, World Bank, Washington, DC, USA

Anzelika Zaiceva

Department of Economics, University of Modena and Reggio, Emilia, Modena, Italy and IZA, Bonn, Germany

EDITORIAL ADVISORY BOARD Orley C. Ashenfelter Princeton University

Daniel S. Hamermesh University of Texas

Francine D. Blau Cornell University

James J. Heckman University of Chicago

Richard Blundell University College London

Alan B. Krueger Princeton University

David Card University of California

Edward P. Lazear Stanford University

Ronald G. Ehrenberg Cornell University

Christopher A. Pissarides London School of Economics

Richard B. Freeman Harvard University

Klaus F. Zimmermann IZA and University of Bonn

ix

PREFACE Informality and informal employment are widespread and growing phenomena in all regions of the world, in particular in low and middle income economies. A large part of economic activity in these countries is not registered or under-declared and many workers enter employment relationships that do not provide any or only partial protection. Causes and consequences of informality in these regions have recently received growing attention, with a particular emphasis on the role of institutions. Several competing paradigms about informality and informal employment exist in the literature. The traditional dualistic view sees the informal segment as the inferior sector, the option of last resort. Due to barriers to entry, minimum wages, unions or other sources of segmentation, formal jobs are rationed. Workers in the informal sector are crowded out from the formal sector involuntarily, their wage being less than that in the formal sector. In contrast, the competitive view sees the formal and informal labor markets not segmented, but integrated. Voluntary choice regarding jobs and particular attributes of these jobs, such as flexible hours, working as a self-employed and being one’s own boss as a micro-entrepreneur, and not valuing social security benefits, can be the reasons for remaining in or moving to the informal sector. A third paradigm points to segmentation within the informal sector. Embedding theoretical and empirical analysis of informality and informal employment in low and middle-income countries into the literature helps us to better characterize the labor markets in these countries. Growing informality and informal employment are not only of academic interest, they are actually an important policy issue. There exist equity and efficiency considerations that point to a strong need to vigorously pursue policies that increase the shares of formal economic activity and employment. It is certainly inequitable if part of the workforce and some firms do not pay their taxes since this implies that those who are formal, whether workers or entrepreneurs, have to bear a disproportionate burden in the financing of public goods that are also of benefit to those being economically active without registration. If the informal part of the economy becomes more substantial this can also mean that governments have to raise taxes and contributions on the formal part and thus have to increase the costs of xi

xii

PREFACE

being formal, which in the final analysis can result in even more informality and a reduced tax base. Furthermore, often workers in informal jobs are severely exploited and are working under conditions that can be hazardous to their health. Turning to efficiency, most economists maintain that employment in the formal sector is associated with a greater use of physical capital that requires human capital acquisition on the part of the employed workers, while the informally employed often work with little or no physical capital. Since physical and human capital are very important ingredients of growth, an economy with a relatively large formal sector will, ceteris paribus, grow at a more rapid pace than an economy with a smaller formal sector. In the medium run, policies combating informality and informal employment are thus vital for raising income and welfare of low and middleincome countries. This volume considers informal employment in emerging and transition economies. It contains eight articles, seven of which were presented at the IZA/World Bank Workshop ‘‘Institutions and Informal Employment in Emerging and Transition Economies’’ in Bonn in June 2011. The different articles shed light on the incidence and persistence of informality and the role of institutions and government regulations. The articles offer insights into issues such as how labor and tax regulations determine the incidence of informality, whether reforms on tax and other regulations can reduce informal employment and who benefits more, to what extent informality occurs as a result of job separations, how persistent is informal employment, how informal employment can be detected and whether migration can be a substitute for informal employment. There are two different views of informality in the labor market as far as its voluntary nature is concerned. The first view considers informality as an involuntary state in a segmented labor market in which formal jobs are in high demand but short supply. The second view considers informality as a voluntary response by firms and workers to avoid labor tax and other regulations. In the first article, Arnab K. Basu, Nancy H. Chau and Zahra Siddique provide an analytical framework, which encompasses these two different perspectives of informality, to study the impacts of tax and minimum wage reforms on the incidence of informality, as measured by the extent of minimum wage non-compliance, the extent of tax evasion, and the size of the informal workforce. In their model, firms choose between operating in the formal or the informal sector taking into account the costs associated with being formal, which are related to tax regulations, the minimum wage and the cost of entry. One of their main finding is that taxes on profits and on personal income, the minimum wage and the strength of

Preface

xiii

tax enforcement are positively associated with informal sector employment. The effect of the tax and minimum wage policies on tax evasion and minimum wage noncompliance depends critically on the level of the minimum wage relative to worker productivity. A key aspect of their model is the distinction between the reported wage distribution and the actual wage distribution, which pertains to the true take home income of working individuals. Focusing on the latter, the authors evaluate the effectiveness of optimal wage and tax policies in alleviating poverty and maximizing tax revenues. Informal employment is positively associated with the level of taxation. The empirical question is about the extent to which tax reforms can relax the disincentives to operate in the formal sector. In the second article, Fabia´n Slonimczyk investigates empirically the effect of tax reforms on the incidence of informal employment in Russia using the Russian Longitudinal Monitoring Survey (RLMS), covering the period 1998–2009. Exploiting the Russian 2001 flat tax reform, which reduced the average tax rates for the personal income tax and the payroll tax affecting mostly individuals in higher income brackets, he estimates the effect of the reform using a differences-in-differences approach. The tax reform reduced significantly the incidence of informal employment. The largest reduction is observed on the prevalence of informal irregular activities and for the individuals in the top income brackets who benefited the most from the reform. These results suggest that tax reforms may have important behavioral responses, which go beyond the standard labor supply decision. Taxes may not only affect the incidence of informality but also firms’ performance. In the third article, Tommaso Gabrieli, Antonio F. Galvao, Jr., and Gabriel V. Montes-Rojas investigate the effect of tax reductions and simplifications schemes through Brazil’s SIMPLES program on the performance of micro-firms. Their main conjecture is that tax reductions can benefit only a segment of the firm population. Comparing eligible and non-eligible firms before and after the introduction of the program, and using data from the Brazilian Survey of the Urban Informal Sector, they find that increasing formality through tax reductions leads to positive effects on revenues. They also find that the program has an effect on 40–50% of the firm population, which are low ability firms, and that among the affected firms there is heterogeneity with low ability ones benefiting more. As suggested in the theoretical study by Arnab K. Basu, Nancy H. Chau, and Zahra Siddique in the first article of this volume, the level of the minimum wage is an important determinant of the extent of tax evasion and minimum wage noncompliance by the firms. Quantifying the extent of wage

xiv

PREFACE

underreporting is an empirical challenge. In the fourth article, Pe´ter Elek, + Bala´zs Reizer, and Pe´ter A. Szabo´ estimate the probability of Ja´nos Ko¨llo, wage underreporting using the Wage Survey of the National Employment Service in Hungary, which is an employer–employee dataset. Estimating a double hurdle model they find that employers paid cash supplements to around half of all minimum wage employees, which corresponds to a 150% gap between reported and actual wages. In order to test the relevance of their estimates, they exploit the introduction of a minimum social security contribution base equal to 200% of the minimum wage. They find that firms underreporting were more likely to raise the wage of their workers to this new level in order to avoid a tax audit, facing faster average wage growth and slower output growth. Not only taxes but also the benefit system in place in each country may create disincentives to engage in formal employment. In the fifth article, Johannes Koettl and Michael Weber investigate the role of labor taxation and social benefit design on the disincentives for formal work. They propose a new synthetic measure, the formalization tax rate, which takes into account not only the costs due to additional taxes one has to pay by engaging in the formal economy but also the losses from benefit withdrawal due to formalization. Focusing on several European Union New Member States, they find that the disincentives for formal work as measured by the formalization tax rate are especially high for low-wage earners and that the higher the disincentives the higher is the incidence of informal employment. Their analysis suggests that existing measures such as the tax wedge may not be sufficient in capturing disincentives for formal work. Labor market regulations and institutions can impose obstacles in the process of obtaining a formal job. Considering informality as an involuntary state it is usually viewed as an alternative to formal employment, especially for individuals who are less integrated in the labor market such as migrants. In the sixth article, Ilhom Abdulloev, Ira N. Gang, and John Landon-Lane adopt a different view on the relationship between migration and informality. Rather than being complements, they consider the role of migration as a way to reduce informality. This can be achieved through migrant’s earnings and remittances, which can help improve families’ finances and encourage their members to be less involved in informal employment. Using the Living Standards Measurement Survey from Tajikistan, a former Soviet republic, they investigate the relationship between external migration and local informal sector activity, where informality is defined by the discrepancy between reported household expenditure and reported household income. They find a negative

Preface

xv

relationship between informal activities and migration, since the gap between expenditure and income falls in the presence of migration. These results suggest that informality and migration may be considered as substitutes. Identifying the underlying causes of persistence of informality is very important from a policy perspective. The degree of persistence might depend on the actual experience of informality but might also be driven by underlying unobserved individual propensities to be an informal worker. In the seventh article, Alpaslan Akay and Melanie Khamis study the dynamics of informality using panel data from the Ukrainian Longitudinal Monitoring Survey for three years. Estimating a dynamic panel data probit model with endogenous initial conditions they find evidence of high persistence due to previous informality experience. Persistence in informality is higher for young singles and less educated males. These results suggest that policies aiming to reduce informality may have long-lasting effects. Understanding whether informality is the result of labor market segmentation in which goods jobs are in high demand but short supply, or not, requires sufficient information on the reasons driving the decision to obtain an informal job. In the eighth article, Hartmut Lehmann, Tiziano Razzolini, and Anzelika Zaiceva shed light on this issue by investigating the relationship between job separations and the incidence of informality. Using unique data from a displacement supplement to the RLMS in 2008 and from an informality supplement to the RLMS in 2009, they ask the question whether those who quit their job differ in their propensity to enter an informal job compared to those who are displaced. They find that displaced workers are more likely to end up in an informal job compared to their counterparts who quit, with the effect being larger for those with low human capital. They also provide evidence that informality can be persistent, so once a worker becomes informal it is difficult to return to a formal job. As with past volumes, we aim to focus on important issues and to maintain the highest levels of scholarship. We encourage readers who have prepared manuscripts that meet these stringent standards to submit them to Research in Labor Economics (RLE) via the IZA website (http:// www.iza.org/rle) for possible inclusion in future volumes. Hartmut Lehmann Konstantinos Tatsiramos Editors

CHAPTER 1 TAX EVASION, MINIMUM WAGE NONCOMPLIANCE, AND INFORMALITY Arnab K. Basu, Nancy H. Chau and Zahra Siddique ABSTRACT We study the impact of tax and minimum wage reforms on the incidence of informality. To gauge the incidence of informality, we use measures of the extent of tax evasion, the extent of minimum wage noncompliance, and the size of the informal workforce. Our approach allows us to examine (i) the distinction between determinants of firm-level reported wage distribution and actual wage distribution, (ii) the complementarity of tax and minimum wage enforcement, (iii) the impact that a minimum wage reform has on tax and minimum wage compliance, and (iv) the impact that a tax policy reform has on tax and minimum wage compliance. We conclude with the design of optimal minimum wage and tax policies (even in the complete absence of minimum wage enforcement). We do so based on two objectives derived from popular concerns associated with an unchecked expansion of informality: tax revenue maximization, and poverty alleviation among workers.

Informal Employment in Emerging and Transition Economies Research in Labor Economics, Volume 34, 1–53 Copyright r 2012 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1108/S0147-9121(2012)0000034004

1

2

ARNAB K. BASU ET AL.

Keywords: Tax evasion; minimum wage reform; flat tax reform; poverty; informality JEL classification: J3; J6; O17

INTRODUCTION What determines the incidence of informality? The answer clearly depends on how one defines informality. Within the development literature there are two different perspectives on how one defines informality, each focusing on causality going from government regulations to informality as a market response (Perry et al., 2007; Schneider & Enste, 2000). The first perspective takes informality as an involuntary state of employment in a segmented labor market. Good jobs in the formal sector, characterized by regulated wages and benefits, are in high demand but short supply (Chandra & Khan 1993; Fields, 1975; Harris & Todaro, 1970). Rooted in the labor market implications of informality, therefore, informality (according to this perspective) refers to workers, particularly the old and young, who would prefer a job with standard labor protections, but are unable to get one. (Perry et al., 2007, p. 21)

Associated with this perspective are measures of informality based on labor standards compliance. Such measures aim at capturing the coverage of workers by mandated labor protections such as the minimum wage (Saavedra & Chong, 1999). Table 1 provides some examples of such informality measures as: (i) the share of workers not covered by social security contributions (based on 2007 and 2008 EU-SILC), (ii) the share of workers not covered by an employment contract (based on the European Social Survey, 2008) and (iii) the share of workers not covered by social protection (OECD, 2009). From Table 1, there is substantial variation across countries in the degree of informality, irrespective of the specific measure used. If one focuses on measure (iii), which gives the share of workers not covered by social protection, then informality characterizes close to or more than 50% of workers in nearly all the countries for which this measure is available. A second perspective takes informality as a voluntary firm-level response to evade taxes and other costly regulations (de Paula & Scheinkman, 2010; De Soto, 1989; Djankov, La Porta, Lopez-de-Silanes, & Shleifer, 2002;

3

Tax Evasion, Minimum Wage Noncompliance, and Informality

Table 1.

Informality, Labor Standards Compliance.

Country

Alternative Measures of Informality I. Share of workers not covered by social security contributions 2007

Algeria Argentina Austria Belgium Bolivia Brazil Chad Chile Colombia Costa Rica Czech Republic Denmark Dominican Republic Ecuador Egypt, Arab Rep. El Salvador Estonia Finland France Germany Greece Guinea Haiti Honduras Hungary Iceland India Indonesia Ireland Israel Italy Iran Kenya Kyrgyz Republic Lebanon Luxembourg

35.4 38.8

40.8

2008

34.5 36.2

40.4

34.6 23.0 51.9

33.9 23.5 –

37.7

37.3

40.6 13.4

42.4 13.3

39.8

40.3

40.0

39.3

34.6

32.6

II. Share of workers without an employment contract

III. Noncoverage by social protection

2006

1995–1999 2000–2007 42.7 53.3

41.3 –

63.5 60 95.2 35.8 38.4 44.3

– 51.1 – – – –

47.6 53.5 55.2 56.6

– – 45.9 –

86.7 92.6 58.2

– – –

83.4 77.9

– –

– 71.6 – –

48.8 – 44.4 51.8

10 7

2 12

5 1 8 4 39

4 26

39 38

6

4

ARNAB K. BASU ET AL.

Table 1. (Continued ) Country

Alternative Measures of Informality I. Share of workers not covered by social security contributions 2007

Mali Mexico Moldova Morocco Netherlands Norway Pakistan Panama Paraguay Peru Philippines Poland Portugal Romania Russian Federation Slovak Republic Slovenia South Africa Spain Sweden Switzerland Syrian Arab Republic Thailand Turkey Tunisia United Kingdom Venezuela, R.B. West Bank and Gaza Yemen, Rep.

17.7 12.2

65.3 35.1

2008

21.6 13.2

II. Share of workers without an employment contract

III. Noncoverage by social protection

2006

1995–1999 2000–2007 94.1 59.4 – 44.8

81.8 50.1 21.5 67.1

64.6 37.6 65.5 – 72

– 49.4 – 67.9 –

5.4 –

22 8.6

–

50.6

42.9 51.5 30.9 47.1

30.7 – 33.2 35

46.9 – –

49.4 43.4 51.1

9 11

57.0 38.5

6 15

39.1 24.7

38.5 25.2

7 3 11

41.5 22.7

41.4 22.0

10 2 5

44 22

Note: Measure (I) is based on EU-SILC 2007 and 2008, reproduced from Schneider (2011), measure (II) is based on the European Social Survey 2008, reproduced from Schneider (2011), and measure (III) is taken from OECD (2009) tabulations from national labor force surveys where informality is defined as noncoverage by social protection.

Tax Evasion, Minimum Wage Noncompliance, and Informality

5

Friedman, Johnson, Kaufmann, & Zoido-Lobato´n, 2000; Loayza, Serve´n, & Oviedo, 2005; Schneider, 2005), when weak or nonexistent enforcement of taxes and other costly regulations make informal operations more lucrative than formal regulated operations. Rooted in the public finance implications of informality, therefore, informality (according to this perspective) refers to firms and individuals avoiding taxation or other mandated regulations because everybody else does, and because enforcement is weak and uneven. (Perry et al., 2007, p. 22)

Associated with this perspective are measures of informality based on tax compliance. Such measures may result from tax audits that define the magnitude of the informal economy as the difference between the income declared in tax returns and the income actually found after an audit (World Bank, 2011). Table 2 provides examples of such informality measures, including measures that capture different shades of informality even among formally registered firms: (i) the percentage of firms expressing that a typical firm reports less than 100% of sales for tax purposes, (ii) the percentage of firms competing against unregistered or informal firms, (iii) the percentage of firms formally registered when they started operations in the country and (iv) the average number of years firms operated without formal registration. From Table 2, there is (as in Table 1) substantial variation across countries in the degree of informality, irrespective of the specific measure used. For example, if one focuses on measure (i), then Liberia ranks particularly high with respondents putting the percentage of firms which report too few sales to escape taxes as more than 97% while for Jordan this fraction is just 13%. Moreover, measure (ii) shows a higher degree of informality for each country than does measure (iii). These two distinct perspectives on informality have also inspired a large subsequent literature on theoretical modeling, as well as empirical policy analyses. In terms of theoretical modeling, the seminal work of Fields (1975) and related subsequent studies address formal and informal labor market consequences of formal sector wage regulations. Early studies are in the tradition of dualistic labor market models (Doeringer & Piore, 1971, Fei & Ranis, 1964, Harris & Todaro, 1970; Stiglitz, 1974), while recent studies relax assumptions on competitive labor markets (Basu, Chau, & Kanbur, 2010), and contractual commitments (Basu, Chau, & Kanbur, 2011). Models of tax evasion incorporating both an above- and an under-ground sector (Jung, Snow, & Trandel, 1994; Kesselman, 1989) have also been developed. These models extend the traditional tax evasion frameworks of Allingham and Sandmo (1972) and Yitzhaki (1974) to a multisector setting and provide an analytical basis for

6

ARNAB K. BASU ET AL.

Table 2. Country (Year)

Informality, Enterprise Surveys. Alternative Measures of Informality

I. % of firms II. % of firms which report too competing with few sales unregistered/ informal firms Afghanistan (2008) Albania (2007) Algeria (2007) Angola (2010) Argentina (2010) Armenia (2009) Azerbaijan (2009) Bangladesh (2007) Belarus (2008) Benin (2009) Bhutan (2009) Bolivia (2010) Bosnia and Herzegovina (2009) Botswana (2010) Brazil (2009) Bulgaria (2009) Burkina Faso (2009) Burundi (2006) Cambodia (2007) Cameroon (2009) Cape Verde (2009) Chad (2009) Chile (2010) China (2003) Colombia (2010) Congo, Dem. Rep. (2010) Congo, Rep. (2009) Costa Rica (2010) Coˆte d’Ivoire (2009) Croatia (2007) Czech Republic (2009) Dominican Republic (2005) Ecuador (2010) Egypt, Arab Rep. (2008)

– – – – – – – – – – – – –

III. % of firms IV. Number of registered when years firm starting operated without operations registration

45.88 52.56 66.84 41.25 68.2 44.57 40.71 46.62 50.43 77.25 19.92 80.51 46.48

88.01 89.35 98.32 62.65 92.35 96.23 85.12 – 98.48 87.93 99.05 72.43 98.6

1.8 0.26 0.04 1.12 0.87 0.14 4.73 – 2.9 0.55 0.08 4.25 0.16

– – – 26.45 42.73 78.11 70.38 – – – 49.45 – –

54.64 55.02 54.11 74.95 60.3 – 90.11 44.54 89.8 55.84 – 70.93 89.96

93.94 95.78 98.48 77.74 – 87.45 82.06 81.25 77.13 96.05 – 94.27 61.86

0.16 0.48 0.08 1.01 – 0.69 0.63 1.96 1.23 0.2 – 0.47 1.87

90.02 – 68.06 – –

69.65 70.36 73.6 31.66 42.87

84.28 80.84 56.37 98.07 98.04

0.27 1.04 0.48 0.03 0.08

73.63 – 30

– 65.72 46.68

– 85.1 –

– 0.65 –

7

Tax Evasion, Minimum Wage Noncompliance, and Informality

Table 2. (Continued ) Country (Year)

Alternative Measures of Informality I. % of firms II. % of firms which report too competing with few sales unregistered/ informal firms

El Salvador (2010) Eritrea (2009) Estonia (2009) Ethiopia (2006) Fiji (2009) Gabon (2009) Gambia, the (2006) Georgia (2008) Germany (2005) Ghana (2007) Greece (2005) Guatemala (2010) Guinea (2006) Guinea-Bissau (2006) Guyana, CR (2004) Honduras (2010) Hungary (2009) India (2006) Indonesia (2009) Ireland (2005) Jamaica (2005) Jordan (2006) Kazakhstan (2009) Kenya (2007) Korea, Rep. (2005) Kosovo (2009) Kyrgyz Republic (2009) Lao PDR (2009) Latvia (2009) Lebanon (2009) Lesotho (2009) Liberia (2009) Lithuania (2009) Macedonia, FYR (2009) Madagascar (2009) Malawi (2009)

III. % of firms IV. Number of registered when years firm starting operated without operations registration

– – – 51.6 – 64.83 88.05 – – 59.2 53.19 – 95.37 68.19

65.24 28.16 26.32 – 39.6 75.96 60.34 52.23 – 69.13 – 69.84 62.8 53.7

75.67 100 97.39 – 93.48 63.73 – 99.55 – 66.44 – 89.98 – –

1.1 0 0.06 – 1.15 0.73 – 0.42 – 2.17 – 0.59 – –

74.36 – – 59.24 – 28.78 28.77 12.95 – 60.54 43.65 – –

– 63.17 49 – 65.09 – – – 36.89 – – 64.07 67.46

– 81.29 100 – 29.1 – – – 97.37 – – 89.23 95.94

– 0.86 0 – 2.43 – – – 0.04 – – 0.23 4.3

– – 34.98 – 97.32 – –

12.82 41.67 – 59.6 66.21 50.11 73.89

93.53 98.49 97.63 86.8 73.81 97.1 99.24

0.17 0.12 0.28 0.63 0.63 0.25 0.01

35.61 –

62.29 77.84

97.45 78.58

0.09 0.57

8

ARNAB K. BASU ET AL.

Table 2. (Continued ) Country (Year)

Alternative Measures of Informality I. % of firms II. % of firms which report too competing with few sales unregistered/ informal firms

Malaysia (2007) Mali (2010) Mauritania (2006) Mauritius (2009) Mexico (2006) Micronesia, Fed. Sts. (2009) Moldova (2009) Mongolia (2009) Montenegro (2009) Morocco (2007) Mozambique (2007) Namibia (2006) Nepal (2009) Nicaragua (2010) Niger (2009) Nigeria (2007) Oman (2003) Pakistan (2007) Panama (2010) Paraguay (2010) Peru (2010) Philippines (2009) Poland (2009) Portugal (2005) Romania (2009) Russian Federation (2009) Rwanda (2006) Samoa (2009) Senegal (2007) Serbia (2009) Sierra Leone (2009) Slovak Republic (2009) Slovenia (2009) South Africa (2007) Spain (2005) Sri Lanka (2004)

III. % of firms IV. Number of registered when years firm starting operated without operations registration

– – 82.5 36.25 57.65 –

– 75.38 65.16 50.95 69.67 41.11

52.97 79.22 – 84.17 94.08 96.87

0 0.51 – 1.54 0.1 0.16

– – – – 73.1 45.48 – – – 68 42.46 7.57 – – – – – 37.25 – –

46.46 43.04 27.26 47.68 75.45 33.09 49.37 62.61 85.98 60.07 – 12.45 51.4 75.28 68.59 37.52 32.73 – 35.14 32.23

97.86 90.11 95.47 86.02 85.86 – 93.96 74.03 90.51 – – – 99.74 98.71 82.63 97.53 99.25 – 98.71 94.68

0.11 2.49 0.21 0.21 0.6 – 0.33 2.02 0.54 – – – 0.01 0.25 0.73 0.53 0.02 – 0.11 0.77

28.9 – 21.63 – 81.92 –

47.07 63.56 74.11 53.62 80.34 40.34

– 88.36 78.86 95.02 89.22 100

– 0.58 0.95 0.19 0.86 0

– 40.3 18.33 41.97

27.44 45.32 – –

99.85 91.03 – –

0.01 0.26 – –

9

Tax Evasion, Minimum Wage Noncompliance, and Informality

Table 2. (Continued ) Country (Year)

Alternative Measures of Informality I. % of firms II. % of firms which report too competing with few sales unregistered/ informal firms

Swaziland (2006) Syrian Arab Republic (2009) Tajikistan (2008) Tanzania (2006) Thailand (2006) Timor-Leste (2009) Togo (2009) Tonga (2009) Turkey (2008) Uganda (2006) Ukraine (2008) Uruguay (2010) Uzbekistan (2008) Vanuatu (2009) Venezuela, R.B. (2010) Vietnam (2009) West Bank and Gaza (2006) Yemen, Rep. (2010) Zambia (2007)

III. % of firms IV. Number of registered when years firm starting operated without operations registration

74.57 57.06

39.72 52.57

– –

– –

– 71.03 – – – – – 74.49 – – – – –

35.31 66.85 – 66.4 80.49 86.77 52.36 73.11 48.48 67.66 39.17 39.91 29.76

92.74 – – 91.79 75.82 93.54 94.05 – 95.76 94.61 100 88.06 95.61

6.78 – – 0.67 1.15 0.25 0.42 – 0.92 0.64 4.09 1.26 0.13

– 25.68

55.56 –

87.46 –

1.38 –

82.33 –

37.16 73.84

81.69 96.18

0.82 0.23

Notes: The four measures give the degree of informality among businesses in 125 countries. The table is reproduced from results based on surveys of more than 120,000 firms by the World Bank (Enterprise Surveys, http://www.enterprisesurveys.org, methodology available from https://www.enterprisesurveys.org/Methodology/). Measure (I) is constructed from the percentage of firms expressing that a typical firm reports less than 100 percent of sales for tax purposes, measure (II) is constructed from the percentage of firms competing against unregistered or informal firms (Question: Does this establishment compete against unregistered or informal firms?), measure (III) is constructed from the percentage of firms formally registered when they started operations in the country (Question: Was this establishment formally registered when it began operations?) and measure (IV) is constructed from the average number of years firms operated without formal registration, computed only for the firms not formally registered when starting operations (Question: In what year did this establishment begin operations? Question: Was this establishment formally registered when it began operations? Question: In what year was this establishment formally registered?).

10

ARNAB K. BASU ET AL.

ongoing debates concerning the role of tax policy reforms on tax evasion among formal firms, as well as the underground informal economy. In terms of empirical policy analysis, there has been a recent surge in empirical research generating the following set of intriguing insights about how the informal economy operates in response to labor and tax regulations:  minimum wage noncompliance is widespread among formal sector employers, as well as informal sector employers (Ashenfelter & Smith, 1979; Gindling & Terrell, 1995, 2006; Lemos, 2004, 2006; Maloney & Nunez, 2004);  tax evasion is widespread among formal sector firms, as well as informal sector firms (Cobham, 2005; Cowell, 1990; Feige, 1989; Fuest & Riedel, 2009; Tanzi, 1980, 1999);  subminimum wages and informal employment can exhibit a diverse set of responses to minimum wage hikes (Baanante, 2004; Card & Krueger, 1995; Lemos, 2004; Strobl & Walsh, 2001);  tax policy reforms can have a significant impact on the incidence of informality in an economy (Gabrieli, Galvao, & Montes-Rojas, 2011; Jonasson, 2011; Koettl, 2011; Slonimczyk, 2011). Given the above, the point of departure of this chapter from existing literature is to provide a theoretical framework better equipped at capturing the realities that (i) both formal and informal firms may fail to comply with labor and tax regulations, while (ii) labor and tax regulations jointly determine the differing shades of informal firm like behavior through minimum wage noncompliance and tax evasion. We posit a framework wherein formal establishments are subject to three sets of costs not borne by informal firms. With respect to wage regulation, formal establishments are expected to pay workers at least a minimum wage. With respect to tax regulations, formal establishments are obliged to pay authorities a tax on profits earned, and to withhold taxes on wage income to be transferred directly to authorities. With respect to the cost of entry, formal firms incur a cost of registration. In contrast, informal establishments and informal workers are free from both the state mandated regulation on wages and the need to pay taxes on profits or wages while operating in an environment where entry and exit is cost-free. Firms must choose between operating in the formal or the informal sector, while workers seek wage protection in the formal sector (subject to wage income tax) using the informal wage as a reservation wage. Within this context, we examine three items of interest: (i) tax evasion via the misreporting of wage cost by formal firms, (ii) minimum wage noncompliance via the underpayment of the state mandated minimum wage among

Tax Evasion, Minimum Wage Noncompliance, and Informality

11

formal firms, and (iii) the incidence of informality as establishments choose to operate either in the formal or the informal sector. In doing so, the contributions of this chapter are the following. First, and to the best of our knowledge, this is the first attempt at examining minimum wage noncompliance by simultaneously scrutinizing the endogenous determinants of the reported wage cost of a formal employer and the actual takehome wage of a formal employee, in a setting where tax and minimum wage enforcement are both imperfect. A clear distinction between the two can offer important insights into empirical studies of wage distributions based either on firm-reported data on wage costs, or on labor force surveys from which records of actual take-home wage can be ascertained. In particular, it is by now well known that empirical studies of wage distributions in developing countries exhibit spikes at or around the minimum wage, despite lax enforcement. Our analysis provides the rationale for researchers to go one step deeper, to distinguish between reported wage distribution and actual wage distribution. For example, is the spike purely a reflection of a lip-service reporting of firm-level wage cost? Or is the true take-home wage distribution of formal employee also marked by a spike at or around the minimum wage despite lax enforcement? Second, the proposed model allows for a side-by-side comparison of tax compliance among formal employers, minimum wage compliance among formal employers, and the incidence of informality where neither tax regulations nor minimum wage regulations are fully respected. This is of particular importance for, in practice, tax enforcement in the formal sector is carried out via tax audits and random checks by tax authorities (Andreoni, Erard, & Feinstein, 1998), while minimum wage enforcement in the formal sector is carried out independently by labor inspectors (Weil, 2004). An understanding of how tax enforcement directly impacts tax evasion, alters minimum wage noncompliance, and changes the overall incidence of informality is clearly key to effective formation of tax and wage enforcement policies.1 Third, our model provides a setting in which the potential direct and cross-cutting impacts of wage and tax policies on various measures of informality can be better understood. For example, how does a minimum wage hike impact the extent of tax evasion by formal firms, and the number of firms that completely do away with the obligation to pay taxes as they exit the formal sector? In this regard, we present a list of testable hypotheses that complement a series of papers dedicated to understanding minimum wage noncompliance (Ashenfelter & Smith, 1979; Basu et al., 2010) without the additional complication of tax evasion. Next, how does a tax reform impact the extent of minimum wage compliance by formal firms, and the

12

ARNAB K. BASU ET AL.

number of firms that completely evade the obligation to pay a minimum wage by exiting to the informal sector? In this latter regard, we provide results that will substantially extend those of an earlier literature on tax evasion, for example, in which one or both of the following are absent: the minimum wage mandate, and the existence of an informal sector where tax obligations and minimum wage protection do not apply (Tonin, 2011; Yaniv, 1988, 1990). The model we posit allows us to identify six distinctive potential regimes of labor market equilibria and as many potential sets of comparative statics responses of labor market outcomes with respect to changes in minimum wage and tax policies, depending on the pairing of minimum wage and worker productivity, as well as the nature of the tax and minimum wage policies. These embody combinations of cases where in equilibrium (i) reported employer wage payment to government tax authorities exactly comply with, or exceed the official minimum wage, coupled with (ii) actual employer wage payment to workers that does not comply, complies exactly, or over-complies with the minimum wage law. In this chapter, we make a complete characterization of the six classes of labor market equilibria and the corresponding comparative statics responses. To pick one case in point, for minimum wages that are sufficiently high relative to worker productivity, strictly positive but imperfect tax enforcement, and no minimum wage enforcement, all employers report the exact payment of the minimum wage as required by law, but in truth evade taxes as the reported wage cost is an over-statement of wage that workers take home. Since tax evasion (or wage understatement) is synonymous with minimum wage noncompliance (or paying less than the reported minimum wage) here, the gap between the minimum wage and the actual wage is determined only by the strength of tax enforcement in place. Consequently, at constant tax enforcement, raising the minimum wage raises the reported wage as well as the actual wage dollar-for-dollar even in the complete absence of directly minimum wage enforcement, leaving the gap between the two constant. This suggests intriguingly that even when there is no minimum wage enforcement, the observed reported wage distribution exhibits a spike at the minimum wage, while the actual wage payment also exhibits a spike at a wage relative to the official minimum wage determined by the strength of tax enforcement that is in place. Making use of the rich array of possible equilibrium outcomes and comparative statics responses characterized here, we draw three broad set of conclusions illustrating the direct and cross-cutting effects of four policy instruments – the tax rate on profit, the tax rate on personal income, the

Tax Evasion, Minimum Wage Noncompliance, and Informality

13

minimum wage, and the strength of tax enforcement – respectively on informal employment, tax evasion, and minimum wage compliance. Specifically, we find that across all regimes of interest, and regardless of the size of the minimum wage relative to labor productivity, informal sector employment is weakly increasing in the four policy instruments. These follow directly from the weakly decreasing relationships between formal sector profit and respectively the two tax rates, the minimum wage, and the strength of tax enforcement, and the resulting weakly negative impact these policies have on job creation in the formal sector. Contrasting the uniformity in the direction of informal employment response to the four policies, we find that the effectiveness of the four policies in combatting tax evasion differ sharply, depending critically on the height of the minimum wage relative to worker productivity. Suppose that the minimum wage is sufficiently low that in fact employers report a wage cost that exceeds the official minimum wage. Tax evasion here purely reflects employers’ incentive to take full advantage of the gap between the profit tax and the personal income tax, adjusted for the strength of tax enforcement, to maximize posttax profits by under-reporting wage cost. Thus, raising the profit tax stimulates tax evasion, while raising the personal income tax, or the strength of tax enforcement discourages it. However, if the minimum wage is sufficiently high, we find that profit-maximizing employers no longer report over-compliance with the minimum wage, but rather they maximize profits by reporting the exact payment of the minimum wage. The distance between this reported wage and the actual take-home wage – the extent of tax evasion – naturally depends only on the strength of the tax enforcement in place. Thus, in sharp contrast to the case with low minimum wages, tax evasion here is fully independent of the two tax rates, and strictly decreasing in the strength of tax enforcement. The effectiveness of the four policies in combatting minimum wage noncompliance is likewise critically dependent on the height of the minimum wage. With a sufficiently high minimum wage, as discussed the reported wage coincides with the minimum wage. As such, tax evasion becomes synonymous with minimum wage noncompliance. Since tax evasion depends only on the strength of tax enforcement in this case as discussed above, the same is true of minimum wage noncompliance. If, however, the minimum wage is sufficiently low, employers report over-compliance with the minimum wage. This severs the direct link between tax evasion and minimum wage noncompliance. Indeed, while tax evasion depends critically on the profit tax rate and the strength of tax enforcement, minimum wage compliance is wholly independent of these considerations.

14

ARNAB K. BASU ET AL.

The model concludes with a formulation of two sets of optimal wage and tax policies, respectively with the stated objective of poverty minimization, and tax revenue maximization, taking as given the strength of minimum wage and tax enforcement. These two objectives echo popular concerns regarding the consequences of runaway expansion of the informal sector. We show that the existence of a poverty alleviating minimum wage depends critically, among other things, on the strength of the tax enforcement regime, when minimum wage enforcement is lax. Meanwhile, we show that a flat tax reform can be justified, requiring the harmonization of the tax on profits and tax on wage income, on the grounds of tax revenue maximization, so long as corresponding adjustments in the minimum wage can be made as well. The plan of this chapter is as follows: In the second section, the decision problem of a formal sector employer with respect to what wage to report to tax authorities, what actual wage to pay workers, and whether or not to exit to the informal sector will be established. In the third section, the implied determinants of equilibrium tax evasion, minimum wage noncompliance, and incidence of informality will be examined. In the fourth section, we provide an analysis of the optimal policy for poverty alleviation for a given poverty line, and finally in the fifth section, we provide an analysis of the optimal policy for tax revenue maximization. The sixth section concludes with suggestions for a number of directions for future research.

THE MODEL Workers  Each worker is We consider a labor market with a large pool of workers (N). distinguished by his skill level, aZ0, and the subset of skill type a workers is given exogenously by N(a). Two states of employment are available: informal sector employment and formal sector employment. The informal sector is a free-entry sector: any worker that desires a job can find one there, and any worker that desires to exit the sector can do so at will.2 By contrast, we denote the endogenous likelihood that a skill type a worker receives a formal sector job offer as a(a). Formal sector workers are obligated to pay personal income tax, but informal sector workers are not. The corresponding formal and informal sector take-home earnings, net of any personal income tax withheld by employers in the formal sector tf(a), will be denoted as of(a)  tf(a) and oi(a) respectively. Conditional on receiving a formal sector job offer, a skill type a worker compares of(a)tf(a) with the reservation benchmark oi(a), and

Tax Evasion, Minimum Wage Noncompliance, and Informality

15

selects the best of the two. Otherwise, without a formal job offer, the worker earns oi(a) in the informal sector. Formal Employers Let there be an endogenous M(a) number of employers in search of skill type a workers in the formal sector. A match between an employer and a skill type a worker in the formal sector generates net output a.3 Formal employers are furthermore subject to a minimum wage legislation and a tax on profits, unlike informal sector employers who are free from these obligations. For any formal sector employer with a contracted worker, expected profit is given by the value of net output a net of wage cost of(a), adjusted for any required tax withholdings of labor earnings tf(a), applicable taxes T(a) on wages and profits, and expected penalties associated with the discovery of tax evasion Ep(a), if any: pf ðaÞ ¼ a  ½of ðaÞ  tf ðaÞ  TðaÞ  EpðaÞ

(1)

Matching in the Formal Sector The meeting of formal sector job seekers and formal sector employers is characterized by the presence of match friction in the formal labor market. Specifically, let FðMðaÞ; NðaÞÞ ¼ yMðaÞb NðaÞ1b ; b 2 ½0; 1 denote a matching technology that gives the number of matches between employers and workers, given M(a) number of formal employers with a job offer no worse than an informal job, and N(a) number of formal sector job seekers. As stated, FðÞ is increasing in both arguments, and homogeneous of degree one in (M(a), N(a)). The associated likelihood that an individual worker seeking employment is matched with a formal sector offer is aðaÞ ¼

FðMðaÞ; NðaÞÞ NðaÞ

(2)

while the likelihood that an employer seeking a worker is matched with one is ae ðaÞ ¼

FðMðaÞ; NðaÞÞ MðaÞ

(3)

16

ARNAB K. BASU ET AL.

Becoming a Formal Employer Denote the cost associated with registering a formal sector firm plus any additional cost required to raise start-up capital, for example, as cf. Expected formal sector profit is thus: Epf ðaÞ ¼ ae ðaÞpf ðaÞ  cf

(4)

We now turn to employers in the informal sector. Many studies have argued that without formal registration, and without the ability to raise the start-up capital that a formally registered firm is able to, a productivity gap exists between formal and informal firms (Amaral & Quintin, 2006; Djankov, Lieberman, Mukherjee, & Nenova, 2003; Loayza, 1996; Straub 2005). We follow this line of argument and take the productivity of a worker in the informal sector to be only a fraction wio1 of his full productivity. With perfect competition and free-entry on both the supply and demand sides of the informal labor market, expected profits in the informal sector Epi(a) are driven to zero, Epi ðaÞ ¼ 0

(5)

while the informal wage is given by the marginal value of product of laborers there, oi(a) ¼ wia. Armed with Eqs. (1), (4), and (5), an employer makes a decision between operating in the formal and the informal sector by comparing Epf(a) and Epi(a). In equilibrium, employers are indifferent between the two sectors whenever Epf ðaÞ ¼ Epi ðaÞ3ae ðaÞpf ðaÞ  cf ¼ 0 By definition of the likelihood of a match with a worker ae(a) in Eq. (3), and the likelihood of formal job arrival facing workers a(a) in Eq. (2), ae ðaÞ ¼

  cf pf ðaÞ b=ð1bÞ ; and aðaÞ ¼ y1=ð1bÞ pf ðaÞ cf

(6)

It follows from Eq. (6) that in order to ascertain the likelihood of equilibrium formal sector employment, a(a), or that of equilibrium informal sector employment, 1a(a), information about after-tax profits of employers of skill type a workers, pf(a), is paramount. We turn to a detailed analysis of pf(a) now.

Tax Evasion, Minimum Wage Noncompliance, and Informality

17

Tax Evasion, Minimum Wage, and Expected Formal Profits The decision problem of a formal employer is twofold: choose a wage r(a) to be reported to tax authorities based on which the formal employer’s tax liability will be assessed, and an actual wage of(a) to be paid to workers that need not be equal to r(a). Whenever r(a)Wof(a), we say that there is over-reporting of wage cost. Otherwise, if r(a)oof(a), we say that wage cost is under-reported.4 Let too1 denote the personal income tax rate. Formal employers are required to withhold taxes on labor income and transfer them directly to tax authorities. Thus, an employer’s choice of reported wage cost r(a) will have a direct bearing on the amount of tax withholdings from labor income, for tax withholding is given simply by the personal income tax rate multiplied by the reported wage r(a): tf ðaÞ ¼ to rðaÞ

(7)

To ensure that formal sector employment is viable, in that the post tax income of a formal sector worker can exceed that of an informal sector worker, we work with parameter values such that a(1to) exceeds wia, or 1towiW0, requiring effectively that the post tax net output that each formal worker can generate exceeds the informal sector counterpart. Denote tpo1 as the profit tax rate. For a formal employer, the total amount of taxes due to tax authorities, calculated based on reported wage cost r(a), is given by tp times pretax reported profit (min{0, ar(a)}), plus personal income taxes withheld (tor(a)): TðaÞ ¼ tp minf0; a  rðaÞg þ to rðaÞ

(8)

where tp min{0, ar(a)}Z0 indicates a tax policy that does not provide subsidies to employers who reportedly earn negative pretax profits. Henceforth, let f(a) denote the likelihood of a tax audit. With probability 1f(a), the formal employer is not audited, and profit is simply given by net output a, net of wage cost adjusted for personal income taxes withheld (of ðaÞ  to rðaÞ), and net of profit tax due to authorities calculated based on reported wage cost (tp minf0; a  rðaÞg þ to rðaÞ): a  of ðaÞ  tp minf0; a  rðaÞg As may be expected, given the true wage cost of ðaÞ, formal employer profit rises with the reported wage rðaÞ since rðaÞ is inversely related to a formal employer’s calculated tax liability tp minf0; a  rðaÞg. With probability fðaÞ, the formal employer is audited. This leads to the discovery of the extent of misreported profits, if any. Let pðaÞ denote the

18

ARNAB K. BASU ET AL.

penalty associated with the discovery of misreported profits equaling j½a  of ðaÞ  ½a  rðaÞj ¼ jrðaÞ  of ðaÞj, and of misreported wage cost, also equal to jrðaÞ  of ðaÞj. Total penalty to be imposed on a tax-evading employer, if discovered, will given by p multiplied by the extent of misreporting:  pðaÞ ¼ pjrðaÞ  of ðaÞj p parameterizes the severity of the penalty associated with each dollar of misreported wages and profits jrðaÞ  of ðaÞj. Throughout we will assume that p  1, so employers must at least pay back to tax authorities that amount which is misreported when discovered. In the event of a tax audit, therefore, formal employer profit must now account for the penalty cost pðaÞ, and is given by: a  of ðaÞ  tp minf0; a  rðaÞg  pðaÞ   of ðaÞj ¼ a  of ðaÞ  tp minf0; a  rðaÞg  pjrðaÞ We assume in what follows that the likelihood of a tax audit fðaÞ is determined by the likelihood that a tax filing is red-flagged. We also assume that the criteria chosen by tax authorities to red-flag a tax filing are relevant, in the sense that the likelihood assigned to red-flag a tax filing rises with the extent of actual tax misreporting, jðrðaÞ  of ðaÞÞj. These are plausible assumptions. For example, the Internal Revenue Service in the United States formulates and implements a ‘‘discriminant function’’ on each tax return. The results inform the construction of a ‘‘DIF score’’, which is used to determine the likelihood of tax audit. According to Andreoni et al. (1998), over half of the tax audit selections in the United States are based at least in part on this score, and average tax assessments based on selections from the DIF score and other special examination initiatives are systematically higher than tax assessments generated by random audits.5 In view of these evidence, we assume henceforth that the likelihood of a tax audit is increasing in the extent of tax misreporting, and specifically,  fðaÞ  fjrðaÞ  of ðaÞj f parameterizes the frequency of tax audits, for given misreporting of true tax liability jrðaÞ  of ðaÞj.6 It follows that expected formal employer profit is: pf ðaÞ ¼ ð1  fðaÞÞ½a  of ðaÞ  tp minf0; a  rðaÞg  þ fðaÞ½a  of ðaÞ  tp minf0; a  rðaÞg  pðrðaÞ  of ðaÞÞ    of ðaÞÞ2 ¼ a  ½of ðaÞ þ tp minf0; a  rðaÞg  fpðrðaÞ

(9)

where the sum of ðaÞ þ tp minf0; a  rðaÞg gives the sum of wage cost net of taxes withheld and applicable taxes to be transferred to tax authorities

Tax Evasion, Minimum Wage Noncompliance, and Informality

19

½of ðaÞ  to rðaÞ þ ½tp minf0; a  rðaÞg þ to rðaÞ as shown in Eq. (1), and the  pðrðaÞ  expected penalty EpðaÞ in Eq. (1) is given by f  of ðaÞÞ2 . We are now in a position to formally examine the decision problem of a formal employer. In choosing of ðaÞ and rðaÞ to maximize profits pf ðaÞ, we note that any formal employer faces two types of constraints.

Participation Constraint The first constraint that formal employers must take note of accounts for workers’ option to seek employment in the informal sector at any time. Thus, to attract workers, a formal job offer must be no worse than an informal job offer. Equivalently, the take-home wage of a formal sector worker accounting appropriately for taxes withheld must be no less than the informal sector wage oi ðaÞ ¼ wi a from Eq. (5). Furthermore, since taxes withheld is given by the tax rate to times reported wage cost, the participation constraint is: of ðaÞ  tf ðaÞ  wi a; 3 of ðaÞ  to ðaÞrðaÞ  wi a

(10)

Minimum Wage Constraint The second constraint that formal employers must also take note of accounts for the minimum wage legislation. In our context where actual and reported wages can differ, the minimum wage constraint requires that the employer must report at least the minimum wage as their per worker wage cost: rðaÞ  w

(11)

A failure to do so is tantamount to a willful disregard of the minimum wage legislation, and is taken to trigger an immediate audit and fines, large enough to obligate compliance with the reporting constraint.7 Beyond this requirement that constrains employers’ reporting of wage cost, we focus in this chapter on a government that completely turns a blind eye to the possibility of minimum wage noncompliance, among employers that reportedly pay the minimum wage. We do so to scrutinize the potential link between the frequency of tax audits and minimum wage compliance even in the complete absence of minimum wage enforcement, and also to account for the reality that minimum wage enforcement is far less than perfect in both developed and developing countries (Ashenfelter & Smith, 1979; Basu, Chau, & Kanbur, 2010).8

20

ARNAB K. BASU ET AL.

A number of observations are in order before we proceed any further. First, an employer’s choice of of ðaÞ and rðaÞ determines first and foremost the extent of tax evasion per worker employed as given by the difference: jrðaÞ  of ðaÞj A priori, it is not clear at all whether tax evasion should come in the form of over-reporting or under-reporting of wages. For one thing, we know from the definition of pf ðaÞ in Eq. (9) that every dollar increase in reported wage cost raises employer profits by tp for given of ðaÞ. But meanwhile, raising rðaÞ decreases the take-home income of formal sector workers. If the participation constraint in Eq. (10) binds, every dollar increase in reported wage cost rðaÞ would require a corresponding increase in actual wage cost by at least to dollars. These suggest that two sets of considerations should be expected to be in play in the employer’s decision problem: (i) whether the participation constraint binds and (ii) whether a tax gap exists between the profit tp and the personal income tax rates to . Second, an employer’s choice of of ðaÞ and rðaÞ also determines the extent of minimum wage noncompliance. We gauge this noncompliance by evaluating the difference between the post tax wage income of a minimum   to Þ, and the actual wage income of the same worker net wage worker, wð1 of taxes withheld and calculated based on the reported wage, of ðaÞ  to rðaÞ:   to Þ  ½of ðaÞ  to rðaÞ wð1

(12)

If strictly positive, Eq. (12) gives the extent of minimum wage noncompliance. If strictly negative, there is over-compliance with the minimum wage legislation. Finally, if Eq. (12) is equal to zero, formal sector workers receive exactly the minimum wage net of any government mandated personal income tax evaluated based on the minimum wage as their takehome earning. There are three open questions here, each requiring an in-depth analysis: (i) Are there systematic differences between reported wages rðaÞ and actual take-home wages of ðaÞ  to rðaÞ across workers of differing skill levels? In other words, how should reported minimum wage compliance be expected to compare with actual minimum wage compliance? (ii) How does the extent of minimum wage noncompliance respond to a hike in the minimum wage when minimum wage enforcement is lax, but tax audit is carried out in regular frequency? (iii) What is the nature, if any, of the cross-cutting influence that changes in the tax rates may have on minimum wage noncompliance, or that changes in the minimum wage may have on the extent of tax evasion? As a third observation, note that economy-wide tax evasion and minimum wage compliance depend not just on what formal sector employers do, but

Tax Evasion, Minimum Wage Noncompliance, and Informality

21

on the incidence of informal sector employment as well, since informal employers do not pay taxes, nor do they comply with the minimum wage law. From Eq. (6), the incidence of informality, as measured by the fraction of informal sector workers among skill type a workers, is:  b=ð1bÞ 1=ð1bÞ pf ðaÞ (13) 1  aðaÞ ¼ 1  y cf Thus, in order to ascertain the effectiveness of a minimum wage policy w in raising labor earnings beyond wi a, and the effectiveness of the tax policy  pg  in collecting tax revenue, the possibility that any policy shock ftp ; to ; f can lead to a potential exodus of formal sector employers to the informal sector must be accounted for. With these three considerations in mind, the ensuing analysis will  on (i) tax evasion,  tp ; to ; f pÞ focus on the role of four policy measures ðw; (ii) minimum wage noncompliance, and (iii) incidence of informality as defined above. Later on in Section 4, the optimal choice of minimum wage and tax policies to fulfill two distinctive goals: to maximize tax revenue and to minimize the incidence of poverty, will be formally defined and examined.

TAX EVASION, MINIMUM WAGE NONCOMPLIANCE, AND INCIDENCE OF INFORMALITY We undertake here an analysis of tax evasion, minimum wage noncompliance, and the corresponding incidence of informality for the case where the minimum wage is less than the net output per worker in the formal  The complementary case of aow would of course trivially sector: a4w. generate a complete exodus of formal sector employers to the informal sector if the minimum wage law is enforced. It can be demonstrated that the qualitative findings in what follows remain unchanged even in the imperfect enforcement case, and the proof is available upon request. From Eq. (9), the decision problem of the employer is:   of Þ2 max a  of  tp minfa  r; 0g  f pðr of ;r

(14)

subject to the participation constraint of  wi a þ to r, and the minimum  Denote the solutions to Eq. (14) with an asterisk, wage constraint r  w. there are six possible regimes to consider: with strictly positive reported

22

ARNAB K. BASU ET AL.

profits a  r ðaÞ40, (I) only the participation constraint is strictly binding, (II) only the minimum wage constraint is strictly binding and (III) both constraints are binding.9 The remaining three possible cases apply when reported profit a  r ðaÞ is non-positive, with (IV) only the participation constraint is strictly binding, (V) only the minimum wage constraint is strictly binding and (VI) both constraints are binding.

Positive Reported Profits We begin with the case of positive reported profits, and consider each of the three regimes (I)–(III) alluded to above. Binding Participation Constraint For an employer who reports positive pretax profit, maxfa  r; 0g ¼ a  r, and as such from Eq. (14) his decision problem is:  pðr   of Þ2 max a  of  tp ða  rÞ  f of ;r

Maximizing the above subject to a binding participation constraint, of  to r ¼ wi a the profit-maximizing choices of a formal employer are:10 r ðaÞ ¼

wi a þ d wi a þ dto ; of ðaÞ ¼ ; of ðaÞ  r ðaÞto ¼ wi a 1  to 1  to

(15)

  to Þ. Intuitively, with the participation conwhere d  ðtp  to Þ=½2f pð1 straint strictly binding, formal sector workers take home their reservation wage of ðaÞ  to r ðaÞ ¼ wi a, and any increase in formal sector income tax to r ðaÞ must be compensated by a corresponding increase in pretax wage of ðaÞ. In addition, from Eq. (15), the formal employer over-reports wage cost r ðaÞ4of ðaÞ, or equivalently d40, if and only if the profit tax rate tp exceeds the personal income tax to , and under-reports wage cost r ðaÞoof ðaÞ otherwise when tp is less than to . Table 3 displays separately the profit tax rates and the lowest personal income tax rates in 120 countries (PricewaterhouseCoopers, 2011; World Bank, 2011).11 We take the lowest personal income tax rate as the relevant income class for minimum wage workers. As shown, in an overwhelming majority of these countries where tax data is available, the profit tax rate strictly exceeds the corresponding personal income tax rate. In view of this evidence, we will henceforth maintain the assumption that tp 4to , with the implication that formal employers will tend to

23

Tax Evasion, Minimum Wage Noncompliance, and Informality

Table 3. Taxation. Country

Albania Angola Antigua and Bermuda Argentina Armenia Australia Austria Azerbaijan Belarus Belgium Bolivia Bosnia and Herzegovina Botswana Brazil Bulgaria Cambodia Cameroon Canada Chile China Colombia Congo, Dem. Rep. Congo, Rep. Costa Rica Croatia Cyprus Czech Republic Denmark Dominican Republic Ecuador Egypt, Arab Rep. El Salvador Equatorial Guinea Estonia Fiji Finland France Gabon Georgia Germany Ghana Greece Guatemala

I. Tax Rate on Profit, tp

II. Tax Rate on Personal Income, to

40.6 53.2 41.5 108.2 40.7 47.9 55.5 40.9 80.4 57 80 23 19.5 69 29 22.5 49.1 29.2 25 63.5 78.7 339.7 65.5 55 32.5 23.2 48.8 29.2 40.7 35.3 42.6 35 59.5 49.6 39.3 44.6 65.8 43.5 15.3 48.2 32.7 47.2 40.9

0 0 0 9 10 0 0 14 12 25 13 10 0 0 10 0 11 15 0 5 0 3 1 0 12 0 15 37 0 0 0 0 0 21 0 6.5 0 0 20 0 0 0 15

24

ARNAB K. BASU ET AL.

Table 3. (Continued ) Country

Honduras Hong Kong SAR, China Hungary India Indonesia Iraq Israel Italy Jamaica Japan Jordan Kazakhstan Kenya Korea, Rep. Kyrgyz Republic Lao PDR Latvia Lebanon Lithuania Luxembourg Macedonia, FYR Madagascar Malawi Malaysia Mauritius Mexico Moldova Mongolia Montenegro Morocco Mozambique Namibia Netherlands New Zealand Nicaragua Nigeria Norway Oman Pakistan Panama Papua New Guinea Paraguay Peru

I. Tax Rate on Profit, tp

II. Tax Rate on Personal Income, to

48.3 24.1 53.3 63.3 37.3 28.4 31.7 68.6 50.1 48.6 31.2 29.6 49.7 29.8 57.2 33.7 38.5 30.2 38.7 21.1 10.6 37.7 25.1 33.7 24.1 50.5 30.9 23 26.6 41.7 34.3 9.6 40.5 34.3 63.2 32.2 41.6 21.6 31.6 50.1 42.3 35 40.2

0 2 16 0 5 3 10 23 0 5 7 10 10 6 10 0 25 2 0 0 10 0 0 0 15 1.92 7 10 9 0 10 0 33 10.5 0 5 0 0 0.75 0 22 8 0

25

Tax Evasion, Minimum Wage Noncompliance, and Informality

Table 3. (Continued ) Country

Philippines Poland Portugal Puerto Rico Romania Russian Federation Saudi Arabia Senegal Serbia Singapore Slovak Republic Slovenia South Africa Spain Sri Lanka St. Kitts and Nevis Swaziland Sweden Switzerland Syrian Arab Republic Tajikistan Tanzania Thailand Turkey Uganda Ukraine United Arab Emirates United Kingdom United States Uruguay Uzbekistan Venezuela, RB Vietnam Zimbabwe

I. Tax Rate on Profit, tp

II. Tax Rate on Personal Income, to

45.8 42.3 43.3 67.7 44.9 46.5 14.5 46 34 25.4 48.7 35.4 30.5 56.5 64.7 52.7 36.8 54.6 30.1 42.9 86 45.2 37.4 44.5 35.7 55.5 14.1 37.3 46.8 42 95.6 52.6 33.1 40.3

5 18 11.5 0 16 13 0 0 10 0 19 16 18 24 4 0 20 31 0 5 8 0 0 15 0 15 0 20 10 0 10 6 5 3

Notes: Tax rate on profit, tp, measures the amount of taxes and mandatory contributions payable by businesses after accounting for allowable deductions and exemptions as a share of commercial profits. Taxes withheld (such as personal income tax) or collected and remitted to tax authorities (such as value added taxes, sales taxes or goods and service taxes) are excluded. This variable is taken from the World Bank, Doing Business project (http://www.doingbusiness. org/), 2010 figures. Tax rate on personal income, to, is the lowest tax on personal income taken from PricewaterhouseCoopers International Limited (PwCIL), Worldwide Tax Summaries, current estimates (August 2011).

26

ARNAB K. BASU ET AL.

over-report wage costs. Thus, jr ðaÞ  of ðaÞj ¼ r ðaÞ  of ðaÞ. This finding is intuitive, and consistent with Yaniv (1990) which analyzes the case without a minimum wage and without an informal sector. In particular, since the profit tax rate exceeds the personal income tax rate, underreporting profits confers greater tax savings to the employer than underreporting wage cost, when the post tax take-home income of formal workers cannot fall below the reservation level wi a. Tax evasion, in this case, depends critically on the tax gap tp  to , in particular, from Eq. (15) r ðaÞ  of ðaÞ ¼ d ¼

t p  to   to Þ 2f pð1

the larger the tax gap, the larger will be the equilibrium tax evasion.  and the Furthermore, the frequency of tax auditing as parameterized by f,  both mitigate against penalty associated with discovered tax evasion p, employer incentives to evade taxes. Small changes in the minimum wage do not alter the tax evasion calculation based purely on the tax gap and tax enforcement, and have thus no impact on equilibrium tax evasion. From Eq. (15), minimum wage noncompliance is given by:   to Þ  wi a   to Þ  ðof  r ðaÞto Þ ¼ wð1 wð1 Clearly, formal employers of sufficiently high skilled workers, with   to Þ=wi , over-comply with the minimum wage legislation for a4wð1 otherwise an informal sector job will be more attractive to workers. All other employers offer workers just enough to make a formal job attractive, though not enough to comply with the minimum wage mandate. For those that do not comply with the minimum wage legislation, since workers take home exactly their reservation income wi a in the presence of a binding participation constraint, the extent of minimum wage noncompliance is exactly equal to the difference between the government mandated take  to Þ and the reservation wage wi a. It follows therefore that home wage wð1 an increase in the government mandated minimum post tax take-home wage   to Þ stimulates minimum wage noncompliance. Meanwhile, small wð1 changes in the profit tax tp have no impact on this minimum wage policy   to Þ  wi a, and thus have no induced mandate to increase wages by wð1 impact on minimum wage noncompliance.

Tax Evasion, Minimum Wage Noncompliance, and Informality

27

Now, substituting Eq. (15) into Eq. (13), and using Eq. (6), the incidence of informality 1  a ðaÞ is given by:   b=ð1bÞ 1  tp 1=ð1bÞ 2   =cf ðað1  to  wi ÞÞ þ fpd (16) 1y 1  to Routine differentiation reveals that 1  a ðaÞ is increasing in both tax  and p if rates, and in the frequency and penalty associated with tax audits f and only if the employer misreports the true wage cost, or if and only if tp ato . Again since workers are paid their reservation wage, formal  This is employer profits are in fact independent of the minimum wage w.   to Þ=wi who overtrue for employers of highly skilled workers a4wð1 comply with the minimum wage law, as well as those who do not comply   to Þ=wi Þ. ðaowð1 We now check the conditions under which our starting premises are true: that (i) the participation constraint is the only binding constraint, and that (ii) reported pretax profit a  r ðaÞ is strictly positive. From Eq. (15), the participation constraint is the only binding constraint if and only if formal  From employers report a wage greater than the minimum wage r ðaÞ4w. Eq. (15), this occurs if and only if the minimum wage is sufficiently small relative to the productivity of the workers: r ðaÞ ¼

wi a þ d wi a þ d  wo  4w3 1  to 1  to

(17)

This is illustrated in Fig. 1(a). To the right of the 45 degree line are all combinations of a and w such that worker productivity exceeds the minimum  wage. The upward sloping schedule labeled PC furthermore divides all ða; wÞ  that lie below the PC schedule, combinations into two groups. For those ða; wÞ the inequality in Eq. (17) holds, and as such, the participation constraint is the only binding constraint. Otherwise, for all other combinations that lie above the PC schedule, the inequality in Eq. (17) is violated as w is high enough now so that the minimum wage constraint starts to bind. Furthermore, and also from Eq. (15), reported pretax profit is strictly positive if and only if a4r ðaÞ, or equivalently, when worker productivity is sufficiently high: r ðaÞ ¼

wi a þ d d oa3a4 1  to 1  to  w i

(18)

28

ARNAB K. BASU ET AL.

Fig. 1a.

Binding Participation Constraint.

In Fig. 1(a), for all employer–worker matches involving a  d=ð1  to  wi Þ,  to the right of the upward sloping schedule labeled PC, formal and all ða; wÞ employers maximize profits by reporting positive profits, and by reporting a wage  To the left of a ¼ d=1  to  wi , cost that strictly exceeds the minimum wage w. worker productivity is too small and formal employer maximize profits by reporting non-positive profits. We will return to this case in the sequel. Summarizing, with high output per worker, a4d=ð1  to  wi Þ, formal employers report positive profits (from Eq. (18)). As long as there is a tax gap between profit and personal tax income tp  to , over-reporting of wage cost occurs (from Eq. (15)). The extent of this misreporting depends importantly on the size of the tax gap, the frequency of tax audits and the associated penalty. For some employers in this range, particularly those with sufficiently high skilled workers, this over-reporting of wage cost occurs as they over-comply with the minimum wage legislation, and pay each worker more than what the minimum wage legislation mandates. Other employers in the range continue to over-report wage cost, but do not comply with the minimum wage legislation. Even accounting for wage over-reporting, however, in this range with relatively high worker productivity and relatively low minimum  wage (woðw i a þ dÞ=ð1  to Þ), the participation constraint is binding, meaning that the post tax take-home earnings of formal sector workers is equal to the reservation wage wi a (from Eq. (17)). It follows, therefore, that in the absence of minimum wage enforcement, the minimum wage is ineffective in the sense that the income well-being of workers in this range is in fact fully invariant to the minimum wage, as well as the two tax rates tp and to .

29

Tax Evasion, Minimum Wage Noncompliance, and Informality

Binding Minimum Wage Constraint Now, let us consider the case in which the minimum wage constraint is the only binding constraint. Here, formal employers report paying the lowest  while at the same wage mandated by the minimum wage legislation at r ¼ w, time offer contracted workers a post tax take-home wage that strictly exceeds their reservation wage of  to r4wi a, consistent with a participation constraint that is not binding. With this combination of r and of , expected formal employer profit, when only the minimum wage constraint binds, is  pðr   of Þ2 ; max a  of  tp ða  rÞ  f of ;r

s:t:

r ¼ w

 pð  w  of Þ2  ¼ maxð1  tp Þa þ tp w  ½of þ f of

The solution to the employer’s decision problem when only the minimum wage constraint is binding is thus:12  of ðaÞ ¼ w  r ðaÞ ¼ w;

1 1   to Þ  ; of ðaÞ  r ðaÞto ¼ wð1   p 2fp 2f

(19)

As shown in Eq. (19), formal employers in this regime reports the minimum mandated wage w as the wage cost per worker. The actual pretax take-home wage of ðaÞ is strictly less than the reported wage of ðaÞ ¼  pÞo  w:  The associated tax evasion is: w  1=ð2f r ðaÞ  of ðaÞ ¼

1 2f p

 and the more severe Clearly, the higher the frequency of tax auditing f  the smaller will be the extent of tax evasion. Interestingly, the penalty p, while workers are paid strictly less than the minimum wage, their take-home subminimum wage co-moves with the mandated minimum dollar-for-dollar. Consequently, equilibrium tax evasion, or equivalently the difference  Furthermore, since small between r ðaÞ and of ðaÞ, is strictly invariant to w. changes in the two tax rates (tp and to ) bear no impact on the severity/ frequency of the penalty associated with wage under-reporting, these changes have no impact on the extent of tax evasion either. Turning now to minimum wage noncompliance, the difference between the mandated post tax wage and the actual post tax wage is:   to Þ  ðof ðaÞ  r ðaÞto Þ ¼ wð1

1 40  p 2f

30

ARNAB K. BASU ET AL.

Thus, minimum wage noncompliance is synonymous with tax evasion, in  pÞ40.  that both are exactly equal to 1=ð2f This implies that employers in this regime do not comply with the minimum wage legislation, and the posttax income of formal workers is less than what the minimum wage mandate would require. Perhaps more importantly, since the extent of wage overreporting depends on the strength of tax enforcement f p alone, the extent of minimum wage noncompliance is independent of the size of the minimum  p.  wage, but strictly decreasing in f Substituting Eq. (19) into Eq. (9), and using Eq. (6), the incidence of informality 1  a ðaÞ is: b=ð1bÞ  pÞ=c  þ 1=ð4f  1  y1=ð1bÞ ð½ð1  tp Þða  wÞ fÞ

Any policy measure that decreases formal employer profits will increase the incidence of informality. Routine differentiation reveals that these policies include increases in the minimum wage, increases in the profit tax  Note in rate to , and a strengthening of the tax enforcement regime f p. particular that since the participation constraint is not binding here, small changes in the personal income tax to do not impact formal employers’ ability to attract workers at constant of ðaÞ, and the incidence of informality is thus in fact independent of to . Let us now verify the starting premises in this regime: (i) the minimum wage constraint is the only binding constraint and (ii) reported pretax profit a  r ðaÞ is strictly positive. From Eq. (19), the minimum wage constraint is the only binding constraint if and only if of ðaÞ  r ðaÞto 4wi a, or if the minimum wage is sufficiently high relative to the productivity of the workers:  w4

 pÞ  wi a þ 1=ð2f 1  to

(20)

From Eq. (19), reported pretax profit is in fact always strictly positive in this regime upon substituting Eq. (19) into Eq. (14),  þ 1=ð4f pÞ40  pf ðaÞ ¼ ð1  tp Þða  wÞ

(21)

whenever worker productivity exceeds the minimum wage. This automatically rules out Regime V discussed at the beginning of this section, where the minimum wage is the only binding constraint in the presence of non-positive reported profit. Fig. 1(b) illustrates. As before in Fig. 1(a), to the right of the 45 degree line are all combinations of a and w such that worker productivity exceeds the minimum wage. The upward sloping schedule labeled MW divides all

Tax Evasion, Minimum Wage Noncompliance, and Informality

Fig. 1b.

31

Binding Minimum Wage Constraint.

 combinations into two groups. For minimum wage sufficiently high ða; wÞ and above the MW schedule, the inequality in Eq. (20) is satisfied and hence the minimum wage constraint is the only binding constraint. Otherwise, for relatively low minimum wage and below the MW schedule, the inequality in Eq. (20) is violated and the participation constraint begins to bind. Using  that lie above the MW Eq. (21), all employer–worker matches ða; wÞ schedule report positive profits. Summarizing, in sharp contrast to the earlier regime where the participation constraint is the only binding constraint when the minimum wage is low relative to the productivity of workers, the minimum wage constraint is the only binding constraint when the minimum wage is sufficiently high relative to    to Þ (from Eq. (20)). the productivity of workers w4ðw i a þ 1=ð2fpÞÞ=ð1 Also unlike the previous regime where tax evasion depends critically on the tax policy induced tax gap tp  to , while minimum wage noncompliance   to Þ  wi a, depends on the minimum wage policy mandated wage hike wð1 minimum wage noncompliance is in fact synonymous with tax evasion here in that they are both given by the same expression that depends only on the  pÞ  (from Eq. (19)). strength of the tax enforcement in place 1=ð2f Thus, the stricter the tax enforcement regime, the higher will be the actual wage that formal employers voluntarily pay, and tax enforcement becomes a means to discourage tax evasion, as well as a means to encourage minimum wage compliance. Indeed, for all employer–worker matches in this regime where the participation constraint is not binding, the formal  o , or sector wage accounting appropriately for the tax on wage income, wt  strictly exceeds the reservation wage wi a. Since the   to Þ  1=ð2f pÞ, wð1

32

ARNAB K. BASU ET AL.

 the formal take-home wage of ðaÞ improves with tax enforcement f p, formal–informal pay gap   to Þ  of ðaÞ  r ðaÞto  wi a ¼ wð1

1  wi a  p 2f

is thus strictly increasing in the strength of tax enforcement as well. This is an important observation, to be developed further in the section to follows: even in the complete absence of minimum wage enforcement, both the reported and the actual post tax take-home wage of formal sector worker strictly increases with a sufficiently high minimum wage satisfying Eq. (20). The hike in take-home  pÞ   to Þ  1=ð2f  in wages from the reservation level wi a in Eq. (5) to wð1 Eq. (19) is subject to the implicit discipline of the system of tax enforcement and  p.  The stricter the enforcement against tax evasion, the steeper audit in place f will be the hike in workers’ take-home wage relative to the reservation level.

BINDING PARTICIPATION AND MINIMUM WAGE CONSTRAINT Unlike in regimes I and II where the minimum wage is respectively sufficiently low, and sufficiently high for minimum wage in the intermediate range, w 2  pÞÞ=ð1   to ÞÞ; both the participation and the ððwi a þ dÞ=ð1  to Þ; ðwi a þ 1=ð2f minimum wage constraints are binding. The two binding constraints jointly and uniquely determine the reported and take-home wages in this regime:13  r ðaÞ ¼ w;

 o; of ðaÞ ¼ wi a þ wt

of ðaÞ  r ðaÞto ¼ wi a

(22)

With a binding participation constraint, formal workers receive an aftertax pay that is equal to their reservation wage wi a. With a binding minimum wage constraint, employers report a wage cost that is equal to the minimum wage. Jointly, these imply that formal employers compensate workers simply by paying the reservation wage wi a, and top off with an additional amount that compensates workers for the personal income tax that a formal sector   o , evaluated at the reported wage cost w. worker owes tax authorities wt The associated equilibrium tax evasion in this regime is   to Þ  wi a40 r ðaÞ  of ¼ wð1  pÞÞ=  where the inequality follows since w 2 ððwi a þ dÞ=ð1  to Þ; ðwi a þ 1=ð2f ð1  to ÞÞ. Thus, employers in this regime also over-report wage costs.

Tax Evasion, Minimum Wage Noncompliance, and Informality

33

In addition, with both constraints binding, equilibrium tax evasion is simply   to Þ  wi a. given by the minimum wage policy mandated wage hike wð1 Importantly, tax evasion is now fully independent of the frequency of tax  Small changes in the profit tax tp have surprisingly audit f or the penalty p. no local impact on tax evasion either. Effectively, in this range, the minimum wage is high enough so that reporting any wage higher than w invites too high an increase in expected penalty. Meanwhile, worker productivity is also high enough in this range, so that simply paying workers a wage that minimizes wage plus expected penalty costs, as in regime II above, is not sufficient to guarantee participation. Consequently, equilibrium tax evasion reflects these considerations as employers report exactly the minimum wage, while setting the true wage cost at a level that makes a formal job no worse than an informal job. In terms of minimum wage noncompliance, we note that minimum wage noncompliance is once again synonymous with tax evasion:   to Þ  wi a40   to Þ  ½of ðaÞ  r ðaÞto  ¼ wð1 wð1 though the extent of minimum wage noncompliance is now independent of the strength of tax enforcement, unlike regime II. Taken together, both equilibrium tax evasion and equilibrium minimum noncompliance strictly   to Þ. However, minimum increases with the mandated post tax income wð1 wage noncompliance can be expected to decrease with work productivity a, and thus the reservation wage wia. Now, the incidence of informality 1  a ðaÞ is:  b=ð1bÞ 1  y1=ð1bÞ pf ðaÞ=cf where  þ pf ðaÞ ¼ ð1  tp Þða  wÞ

 2 1  p 1  ½wð1   to Þ  wi a 40 f  p  p 4f 2f

(23)

It is straightforward to verify that pf ðaÞ is strictly decreasing in all three policies tp , to , and w in this range. Consequently, the incidence of informality is accordingly strictly increasing whenever either one of the two tax rates, or when the minimum wage rises. To close this section, let us check that reported pretax profit a  r ðaÞ is strictly positive in this regime as well. From Eq. (23) above, pf ðaÞ is always strictly positive. This automatically rules out Regime VI, where both constraints are binding in the presence of non-positive reported profit.

34

ARNAB K. BASU ET AL.

Fig. 1c.

Binding Participation and Minimum Wage Constraints.

 combinations that come under this regime. Fig. 1(c) illustrates the ða; wÞ  combinations to the right of the 45 degree line, and in As shown, all ða; wÞ between the PC and MW schedules belong this regime. To summarize, for intermediate levels of the minimum wage, both the participation constraint and the minimum wage constraint are binding. While employers continue to under-report wage cost from Eq. (22), the reported wage cost is bounded below by w here, and workers’ post tax take-home wage is bounded below by the reservation wage. Consequently, tax evasion and minimum wage noncompliance are once again synonymous, though this value is now equal to the extent to which the govern  to Þ exceeds workers’ reservation wage ment mandated post tax wage wð1 wi a. In sharp contrast to regime II, where employers also report the minimum wage as the wage cost, the strength of tax policy enforcement no longer play a role in determining employers’ compliance to tax and minimum wage laws (from Eq. (22)). Similarly, the profit tax rate has no impact on equilibrium tax evasion or minimum wage noncompliance (also from Eq. (22)).

Zero Reported Profits Since regimes V and VI have both been ruled out in the preceding paragraphs, a final regime to consider here involves regime IV. This has been shown to be a real possibility in equilibrium for the case where the

Tax Evasion, Minimum Wage Noncompliance, and Informality

35

participation constraint is the only binding constraint, and specifically when worker productivity is lower than a threshold: ao

d 1  t o  wi

as shown in Eq. (18).

Binding Participation Constraint When reported profit is non-positive r  a, and when the participation constraint is the only binding constraint, formal employers’ decision problem is as follows:   of Þ2 ; max a  of  f pðr of ;r

s:t: r  a; of  rto ¼ wi a

 pðrð1   to Þ  wi aÞ2 ; ¼ max a  wi a  rto  f r

s:t: r  a

As shown, expected profit is strictly decreasing in r. Intuitively, since formal employers are no longer subject to profit tax upon reporting non-positive profit, further raising the reported wage cost r will only raise the expected penalty associated with the misreporting of profit  f pðrð1  to Þ  wi aÞ2 , with otherwise no further beneficial impact on profits. The solutions to the employer’s problem in this case involves setting14 r ðaÞ ¼ a;

of ðaÞ ¼ aðwi þ to Þ;

of ðaÞ  r ðaÞto ¼ wi a

(24)

The associated tax evasion is given simply by r ðaÞ  of ðaÞ ¼ að1  to  wi Þ Since formal employers report zero profits r ðaÞ ¼ a, and make job offers by compensating workers for the reservation wage wia and any applicable personal income tax to r ðaÞ ¼ to a, the extent of tax evasion is independent of the strength of tax enforcement, or the size of the minimum wage. Meanwhile, the extent of minimum wage noncompliance is given by:   to Þ  wi a   to Þ  ðof  r ðaÞto Þ ¼ wð1 wð1 Based on reasoning that should by now be familiar, since the participation constraint is binding, employers make job offers that are just attractive enough to compensate workers for their forgone reservation earning wia. It follows that the extent of minimum wage noncompliance is given by the

36

ARNAB K. BASU ET AL.

  to Þ and the government mandated post tax take-home wage wð1 reservation wage itself. Thus, the strength of tax enforcement does not have any local impacts on minimum wage noncompliance. Finally, the incidence of informality 1  a ðaÞ is given by: 1  y1=ð1bÞ ðpf ðaÞ=cf Þb=ð1bÞ where pf ðaÞ ¼

 2 1  1  að1  to  wi Þ  pf 4pf 2pf

(25)

At zero reported profit, actual profit pf ðaÞ and thus the incidence of informality, 1  a ðaÞ, are both independent of the profit tax tp . However, increase in the personal income tax raises actual wage cost of , and decreases formal sector profits. Thus, the incidence of informality is strictly increasing in the personal income tax rate. Strengthening tax enforcement similarly lowers formal sector profit, and increases the incidence of informality. Synthesizing our findings up to this point, Tables 4 and 5 display the tax evasion and minimum wage noncompliance responses to the four policy  p.  and f  A number of features are particularly striking. measures tp , to , w, First, with the lone exception of the case where the personal income tax on minimum wage workers is greater than the profit tax rate, which we have demonstrated in Table 3 to be the exception rather than the rule empirically, we find that employers over-report wage costs. From the perspective of tax authorities, this over-reporting of wages manifests itself as tax evasion. From the perspective of labor inspectors, this over-reporting of wages gives rise to minimum wage noncompliance. However, tax evasion and minimum

Table 4. The Determinants of Tax Evasion jr ðaÞ  of ðaÞj in the Four Regimes, Assuming tpWto.

tp to w  p f

I  pð1   to ÞÞ d ¼ ðtp  to Þ=ð2f

II  pÞ  1=ð2f

III   to Þ  wi a wð1

IV að1  to  wi Þ

þ  0 

0 0 0 

0  þ 0

0  0 0

37

Tax Evasion, Minimum Wage Noncompliance, and Informality

Table 5.

tp to w f p

The Determinants of Minimum Wage Noncompliance   to Þ ½of ðaÞ  to r ðaÞ in the Four Regimes. wð1

I   to Þ  wi a wð1

II  pÞ  1=ð2f

III   to Þ  wi a wð1

IV   to Þ  wi a wð1

0  þ 0

0 0 0 

0  þ 0

0  0 0

wage noncompliance made possible by less than perfect enforcement raises expected formal employer profits. This puts checks on the incidence of informality, in which employers pay no taxes at all, and the minimum wage legislation is neither respect nor enforced. Second, in both regimes II and III where the minimum wage is sufficiently high, or at an intermediate range, we find that tax evasion and minimum wage noncompliance are synonymous. Thus while the four policies have regime-specific impacts depending on which of two constraints are binding and whether reported profit is positive, the effects of four policies on tax evasion, and on minimum wage noncompliance are identical. In other words, there are well-defined regimes of interest in which policies targeted toward mitigating tax evasion and minimum wage compliance can be complementary to each other. Third, it is clear from Tables 4 and 5 that a minimum wage hike can encourage tax evasion (regime III), while raising the personal income tax can tame minimum wage noncompliance (regimes I, III, and IV). Similarly, strengthening tax enforcement can discourage tax evasion (regimes I and II) and mitigate against minimum wage noncompliance (regime II). Thus, there are indeed cross-cutting impacts that a minimum wage policy can have on tax evasion, and that tax policies can have on minimum wage compliance. This apparent plethora of possible comparative statics responses suggests that the devil is in the details, and furthermore, that the nuanced responsiveness of tax evasion and minimum wage noncompliance to policy changes is likely to depend on the precise specification of the penalty scheme, the tax auditing technology, in addition to the height of the minimum wage, the two taxes, as well as the strength of the tax enforcement regime. All of this make for a very fruitful agenda for future research, and particularly on the design

38

ARNAB K. BASU ET AL.

of an optimal penalty scheme. This said, our analysis so far summarized in Tables 4 and 5 suggest a number of generalizations, and testable hypotheses:  Proposition 1. For all a  d=ð1  to  wi Þ, and minimum wage woa, the extent of tax evasion r  of is    

weakly weakly weakly weakly

increasing in the profit tax rate tp ; decreasing in the personal income tax rate to ;  and increasing in the minimum wage w;  p.  decreasing in the strength of tax enforcement f

The extent of minimum wage noncompliance is    

independent of the profit tax rate tp ; weakly decreasing in the personal income tax rate to ;  and weakly increasing in the minimum wage w;  p.  weakly decreasing in the strength of tax enforcement f

Table 6 summarizes responsiveness of the incidence of informality with respect to the four policy measures. Contrary to Proposition 1, the effectiveness of the four policies is more uniform:  Proposition 2. For all minimum wage woa, the incidence of informality is  and weakly increasing in the profit tax rate tp , to , and the minimum wage w,  p.  strictly increasing in the strength of tax enforcement f These results embody important policy implications, and some of these may not be apparent at first sight. Consider, for example, the consequences of a flat tax reform, which equates to and tp .15 Starting from an equilibrium with to otp , a flat tax reform can be accomplished either by raising to , or by decreasing tp . While our analysis of regime I suggest that closing the gap tp  to decreases individual formal sector employer’s incentives to evade taxes, Proposition 2 suggests that the overall tax consequences of a flat tax Table 6.

tp to w f p

The Determinants of the Incidence of Informality 1a(a) in the Four Regimes. I

II

III

IV

þ þ 0 þ

þ 0 þ þ

þ þ þ þ

0 þ 0 þ

Tax Evasion, Minimum Wage Noncompliance, and Informality

39

reform, accounting for the incidence of informality, will likely depend critically on whether the proposed flat tax requires an increase in to – which is expected to entice employers to turn to the informal sector, or a reduction in tp – which will have the exact opposite impact.

OPTIMAL POLICIES FOR POVERTY ALLEVIATION A key departure of this model from the existing literature is in its unique ability to identify and distinguish between two wage distributions that prevail among workers of differing skills. First, the reported wage distribution pertains to the wage distribution that can be ascertained from firm-level tax records for example. Second, the actual wage distribution pertains to the true take-home income of working individuals that may be ascertained from household/labor force surveys for example. Our objective in this section is to evaluate the effectiveness of policies in alleviating poverty, evaluated based on the actual, rather than the reported post tax take-home wage of individual workers. To do so, we first combine Figs. 1(a)–1(c) to yield Fig. 2, which summarizes in one diagram the juxtaposition of the four regimes (I–IV) examined so far. To recall,  in regime I, the reported wage exceeds the minimum wage, formal worker are paid their reservation income post tax, and employers report positive profits;

Fig. 2.

The Four Regimes.

40

ARNAB K. BASU ET AL.

 in regime II, the reported wage is equal to the minimum wage, formal workers are paid strictly higher than their reservation income post tax, and employers report positive profits;  in regime III, the reported wage is equal to the minimum wage, formal workers are paid their reservation income post tax, and employers report positive profits;  in regime IV, the reported wage exceeds the minimum wage, formal workers are paid their reservation income post tax, and employers report zero profit. We begin with a preliminary result, which can be verified simply by inspection of Fig.2: Proposition 3. I. For all a4w and tax enforcement sufficiently strict (f p sufficiently large) such that a4

1   to  wi Þ 2f pð1

the post tax take-home earnings of formal sector workers strictly exceed their reservation benchmark wi a if and only if the minimum wage exceeds the threshold: w 

 pÞ  wi a þ 1=ð2f 1  to  wi

 p is, the The stricter the tax enforcement regime, and thus the higher f lower this threshold will be. II. For all a4w and tax enforcement sufficiently lax, however, with ao

1  pð1   to  wi Þ 2f

 there does not exist a minimum wage woa that can raise the actual post tax income of formal sector workers beyond the reservation benchmark wi a. When tax enforcement is strict enough, the threshold minimum wage is  in regime given simply by the MW schedule in Fig. 2, which separates ða; wÞ II from that of I, III, and IV. As noted earlier in our discussion of the employer’s decision problem in regime II, what is particularly notable here is that a minimum wage is shown to raise the income of formal sector workers even though an official minimum wage enforcement mechanism is nonexistent. Rather, enforcement against tax evasion serves as the implicit discipline device, and prevents formal employers from over-stating their

Tax Evasion, Minimum Wage Noncompliance, and Informality

41

wage cost by too much, or equivalently, enforcement against tax evasion effectively discourages formal employers from paying workers too low a  wage relative to the reported minimum wage w.   to  wi ÞÞ in Fig. 2 divides the Now, the vertical dotted line at 1=ð2f pð1 range of possible workers’ skill levels into two regions separated by a critical  pð1   to  wi ÞÞ. To the right of a and hence for workers skill level a ¼ 1=ð2f with relatively high skills given the tax enforcement regime, a minimum wage which raises the post tax take-home earnings of formal sector workers beyond wi a can always be found for region II is nonempty for a  1=  pð1   to  wi ÞÞ. However, to the left of a and hence for workers with ð2f sufficiently low skills given the tax enforcement regime, there does not exist a minimum wage, however high or low, that raises the post tax take-home income of formal sector workers beyond the reservation level, since the set  in region II with aoa is empty. Strengthening tax enforcement of ða; wÞ through an increase in f p shifts this critical skill level a to the left, accommodating more formal workers whose income can be raised beyond wi a via an appropriately set minimum wage. In what follows, we examine the choice of a minimum wage and tax policy  tp ; to g for each skill level that minimizes poverty among package fw; workers of a given skill level a, taking as given the strength of the tax  p in place. To this end, let z40 denote the poverty enforcement measures f line. Consider all skill levels such that the benchmark reservation income is less than the poverty line wi aoz, or a4z=wi  az In the absence of poverty intervention through a minimum wage for example, all workers with aoaz are poor, while those with a  az are not. We consider the policy problem: min 1  a ðaÞ;

 p ;to w;t

s:t:

of ðaÞ  r ðaÞto  z

for aoaz . The objective of the policy maker is thus to minimize the incidence of informality, while at the same time guaranteeing formal sector workers an income level no less than the poverty line. Since aoaz , all informal workers are by definition poor. Stated differently, the objective of the policy maker is to minimize the poverty head count, when formal sector workers are guaranteed an income level no less than the poverty line. From Proposition 3, we know that in order to strictly raise the income of formal sector workers above the reservation level wi a, what is required is a minimum wage that is sufficiently high, as well as a tax enforcement

42

ARNAB K. BASU ET AL.

mechanism in place that is sufficiently strict, so that regime II applies. There are thus two distinctive possibilities: 1. if tax enforcement is sufficiently lax such that az

1    t o  wi Þ 2fpð1

it follows from Proposition 3 that there does not exist a minimum wage  woa that can lift any worker with aoaz out of poverty since regime II is empty; 2. if tax enforcement is strict enough so that az 4

1   2fpð1  to  wi Þ

there are thus two types of workers who will live under the poverty line in the absence of policy intervention. Those with relatively low skills   to  wi ÞÞ for whom regime II is empty, and those with ao1=ð2f pð1   to  wi ÞÞÞ; az , for whom regime II is relatively high skills a 2 ½ð1=ð2f pð1 nonempty. For the former group of lower skill workers, a minimum wage that lifts workers out of poverty continues to be nonexistent.  pð1   to  wi ÞÞ. Let us consider the Suppose therefore that az 41=ð2f  pð1   to  wi ÞÞ; az , group of workers with relatively high skills, a4½ð1=2f who would otherwise be living under the poverty line in the absence of  policy intervention. From Proposition 3, any minimum wage woa but greater than the threshold demarcated by the MW schedule in Fig. 2,   to  wi Þ, raises the post tax take-home income of ðwi a þ 1=ð2f pÞÞ=ð1 formal sector workers beyond wi a. From Eq. (19), the corresponding post-tax take-home wage is given by the government mandated post tax take-home  pÞ  in this   to Þ, net of the extent of wage under-reporting, 1=ð2f wage wð1  The higher the minimum   to Þ  1=ð2pfÞ. regime, or of ðaÞ  r ðaÞto ¼ wð1 wage, the higher will be the take-home wage of formal sector workers in this regime. It follows that if the post tax take-home wage is to exceed the poverty line, the minimum wage will need to be sufficiently high, since of ðaÞ  r ðaÞto  z3w 

 pÞ  z þ 1=ð2f   wðzÞ 1  to

(26)

 that lifts formal Eq. (26) above shows that the smallest minimum wage wðzÞ sector workers out of poverty is decreasing in the strength of the tax

Tax Evasion, Minimum Wage Noncompliance, and Informality

43

enforcement regime, and increasing in the personal income tax. Such a wage is feasible, in the sense that it does not exceed the net output a of a worker if and only if  ¼ wðzÞ

 z þ 1=ð2f pÞ oa 1  to

  It follows that for all workers in the range aowðzÞ, a minimum wage woa that lifts workers out of poverty does not exist. Otherwise, for workers in the   puts formal sector workers exactly at range a 2 ½wðzÞ; az , setting w ¼ wðzÞ the poverty line. Now, turning to the minimization of poverty head count, which requires minimizing the incidence of informality subject to Eq. (26) above, recall from Eqs. (6) and (21) that the incidence of informality in regime II is given by:  b=ð1bÞ  pÞ=c  þ 1=ð4f  1  y1=ð1bÞ ½ð1  tp Þða  wÞ f Clearly, the higher the minimum wage, the higher will be the incidence of informality implying that the constraint in Eq. (25) must be strictly binding.  Substituting wðzÞ into the equation above, it is straightforward to see that the incidence of informality is furthermore strictly increasing in tp as well as to . We have thus: Proposition 4. I. If tax enforcement is sufficiently lax, so that az 1=ð2f p ð1  to  wi ÞÞ, there does not exist a poverty alleviating minimum wage for all aoaz . II. If tax enforcement is sufficiently strict, so that az 41=ð2f p ð1  to  wi ÞÞ, a poverty alleviating minimum wage does not exist for low  pÞÞ=ð1   ðz þ 1=ð2f  skilled workers with aowðzÞ  to Þ. For higher skilled  workers, with a 2 ½wðzÞ; az , setting w ¼ z þ

1 ; tp ¼ 0; to ¼ 0  2pf

minimizes the poverty head count. To verify the final item noted in Proposition 4, suppose that the minimum  as indicated in the proposition. It follows wage is set below z þ ð1=2pfÞ from Eq. (26) that formal sector workers live below the poverty line. Since all informal sector workers live below the poverty line as well for aoaz , a  is completely  minimum wage policy w that stipulates woz þ ð1=2pfÞ ineffective in alleviating poverty.

44

ARNAB K. BASU ET AL.

Now suppose instead that the minimum wage is set above or equal to  It follows from Eq. (26) that all formal sector workers live at or z þ ð1=2pfÞ. above the poverty line, but all informal sector workers remain poor. To  minimize the incidence of poverty, the policy maker should set w exactly at wðzÞ  tp , and to . and tp ¼ to ¼ 0, since the incidence of informality decreases with w, Importantly, Proposition 3 demonstrates that a poverty minimizing minimum wage is set at the poverty line z, plus an additional expression that  p.  This remarkably simple depends only on the strength of tax enforcement f formula is independent of the productivity of the worker, or the informal sector wage. What Proposition 3 also points out is that this such an optimal minimum wage can be effective even when the minimum wage itself is not enforced, but that a sufficiently strict tax enforcement regime is in place.

OPTIMAL POLICIES FOR TAX COMPLIANCE  tp ; to g that maximizes We focus now on evaluating the policy package {w; total expected government revenue, including all tax revenue on profits and wage earnings, as well as any applicable fines and penalties collected, taking once again as given the strength of the tax enforcement regime in place. Since informal employers and workers do not pay taxes, total government tax revenue GðaÞ is given simply by the sum of per employer tax on formal profits tp minf0; a  r ðaÞg, tax on wage earnings to r ðaÞ, and the expected  pðr   ðaÞ  of ðaÞÞ2 , multiplied by the total fines and penalties collected: f number of formal sector employers, a ðaÞNðaÞ. From Eq. (14), and by definition of pf ðaÞ,   b=ð1bÞ   pf ðaÞ  pðr   ðaÞ  of ðaÞÞ2 tp minf0; a  r ðaÞg þ to r ðaÞ þ f cf   b=ð1bÞ   pf ðaÞ ¼ NðaÞy1=ð1bÞ a  of ðaÞ þ to r ðaÞ  pf ðaÞ ð27Þ cf

GðaÞ ¼ NðaÞy1=ð1bÞ

Intuitively, the true pretax profit (a  of ðaÞ þ to r ðaÞ) net of the post tax post-penalty profit per employer pf ðaÞ gives the total tax take per employer. Using the results in section ‘‘Positive Reported Profits,’’ government  combinations consistent with regimes revenue can be evaluated at ða; wÞ I–IV. Suppose to begin with that the minimum wage is sufficiently high, as in regime II, so that the participation constraint is not binding. From Eq. (19), the post tax take-home wage of a formally employed worker is

Tax Evasion, Minimum Wage Noncompliance, and Informality

45

 while formal sector expected profit   to Þ  1=ð2pfÞ, of ðaÞ  to r ðaÞ ¼ wð1 pf ðaÞ as displayed in Eq. (21), has been shown to be strictly decreasing in the  Substituting these expressions into Eq. (27) above, we minimum wage w. obtain:   b=ð1bÞ   pf ðaÞ 1  p  to Þ þ atp  wðt GðaÞ ¼ NðaÞy1=ð1bÞ cf 4pf It follows by inspection that since tp  to , total government revenue GðaÞ  combinations in regime II. Put is strictly decreasing in w for all ða; wÞ differently, any tax, and minimum wage policy combinations that put an employer in the interior of regime II can be ruled out from the set of government revenue maximizing policy. This is in sharp contrast to the set of poverty alleviating tax policies, which we have just shown to necessarily put employers in regime II (Proposition 4). Next, consider instead relatively low minimum wages such that regimes I, III, or IV apply, wherein the participation constraint is strictly binding, or, of ðaÞ  to r ðaÞ ¼ wi a, and formal sector expected profits pf ðaÞ are displayed in Eqs. (16), (23), and (25) respectively for regimes I, III, and IV. Thus:   b=ð1bÞ   pf ðaÞ að1  wi Þ  pf ðaÞ GðaÞ ¼ NðaÞy1=ð1bÞ cf Observe that changes in formal employer profits impact government revenue in two distinctive, and opposite directions. First, raising pf ðaÞ decreases the incidence of informality through a ðaÞ ¼ y1=ð1bÞ ðpf ðaÞ=cf Þb=ð1bÞ . Second, and going in opposite direction, a higher pf ðaÞ necessarily requires lowering the government revenue per employer að1  wi Þ  pf ðaÞ. On balance, total government revenue is maximized by choice of pf ðaÞ satisfying:t pf ðaÞ ¼ bað1  wi Þ so that profit per formal employer is a fraction b of the profit per formal employer in the absence of any taxes or minimum wage að1  wi Þ. One way to exactly achieve this level of profit is summarized in the following proposition: Proposition 5. A flat tax reform at minimal minimum wage protection maximizes government revenue GðaÞ. This is accomplished by setting w ¼ 0;

tp ¼ to ¼ ð1  bÞð1  wi Þ

46

ARNAB K. BASU ET AL.

A zero minimum wage (regime I, or IV) reflects the government’s desire to raise revenue, rather than to alleviate poverty. Next, by removing the tax gap, and thus by setting d ¼ 0, regime IV (where aod=ð1  to  wi Þ ¼ 0) vanishes, and as such all formal employers of workers with a40 report positive profits. With regime I remaining, setting d ¼ 0 also removes any incentives on the part of formal employers to evade taxes evaluated at w ¼ 0 since r ðaÞ  of ðaÞ ¼ d from Eq. (15) for regime I. To balance the role of the uniform tax rate tp ¼ to on tax revenue per formal employer, and the incidence of informality, the optimal tax rate is set at ð1  bÞð1  wi Þ with two considerations in mind.16 First, the higher the elasticity of employer entry on job creation b ¼ d log FðaÞ=d log MðaÞ, the lower the tax rate will be. Meanwhile, the higher the productivity of informal sector workers relative to formal sector workers wi , the lower the tax rate should be.

CONCLUSION In this chapter, we study the impacts of tax and minimum wage reforms on the incidence of informality, as measured by the extent of minimum wage noncompliance, the extent of tax evasion, and the size of the informal workforce. These measures are based on two distinctive perspectives on informality, associated with labor standard compliance, and tax compliance. The model we propose offers a list of empirically relevant observations and testable hypotheses, featuring (i) the endogenous distinction between firmlevel reported wage distribution and actual wage distribution, (ii) the complementarity of tax enforcement on minimum wage enforcement, (iii) the impact that minimum wage reform has on tax and minimum wage compliance, and (iv) the impact that tax policy reform has on tax and minimum wage compliance. The chapter concludes with a look at the design of optimal minimum wage and tax policies, given the objectives of tax revenue minimization, and poverty alleviation among workers. For example, we provide conditions under which a tax reform is consistent with the objective of tax revenue minimization. At the same time, we also offer an optimal minimum wage and tax policy formula consistent with the objective of poverty alleviation. In each case, we highlight the role of the strength of the tax enforcement regime on the optimal tax and minimum wage policy formula. This chapter is a first attempt at analyzing the role of minimum wage and tax policy reforms jointly on minimum wage and tax compliance.

Tax Evasion, Minimum Wage Noncompliance, and Informality

47

Many possible routes for future research remain. For example, while we highlight the role of tax enforcement in this chapter, ample opportunities for extension of the basic framework by introducing a combination of tax and minimum wage enforcement remain. In addition, in the tradition of the theory of tax enforcement, the risk attitudes of the employers naturally matter. The introduction of risk aversion, for example, into our basic model is a promising avenue for future work. Finally, in our optimal policy formulation, we have taken the strength of tax enforcement and minimum wage enforcement as given. An extension of the model to allow for endogenous tax and minimum wage enforcement regimes will also be of real interest in efforts to better understand the various nuances of tax and minimum wage policies on informality.

NOTES 1. The literature on the determinants, duration, and separation probability of formal employment, informal employment, and unemployment is a much better researched topic. On the duration, separation probabilities, and entry probabilities of formal and informal employment in Brazil, Argentina, and Mexico, see Bosch and Maloney (2010). Almeida and Carneiro (2011) study the impact of labor standards enforcement on mandated benefits on formal employment, informal employment, and unemployment in Brazil. Chong, Galdo, and Saavedra (2008) consider the role of labor legislations and worker productivity on the incidence of informality in Peru. Fugazza and Jacques (2004) theoretically examine the role of unemployment benefits, minimum wage, taxation, and audits on labor allocation in a matching model. In a search theoretic setup, Albrecht, Navarro, and Vronman (2009) examine the role of labor market policy on the incidence of informality and the distribution of wages across formal sector workers. 2. Though informal employment is modeled as wage employment here, the results of the our analysis will remain unchanged if informal self-employment is considered instead, when each self-employed informal sector worker receives a skill-specific take-home income of oi(a). 3. Net output a is taken to account for any other costs of production, associated with capital use for example, per worker hired. 4. We focus here exclusively on the possibility of tax evasion based on the misreporting of costs, and assume that revenue a is verifiable upon audit. Other methods of tax evasion clearly exist. For example, in a setting where a firm employs multiple workers, under-reporting the number of workers is another interesting possibility. We thank a referee for pointing this out, and leave the issue for future research. 5. In Danziger (2010) which studies the issue of minimum wage noncompliance, the probability of detection is taken to be strictly increasing in the number of workers hired, and independent of the extent of the minimum wage violation. In our setup

48

ARNAB K. BASU ET AL.

where each employer hires one worker, the question of the endogeneity of inspection likelihood does not arise. As another alternative, f(a) is a constant with random auditing. It can be easily verified that regardless of worker productivity, a corner solution applies – no employer ever evade taxes, or all employers evade taxes to the maximum extent possible. We thank two anonymous referees for pointing out these possibilities. 6. Given this assumption on f(a), we will check in what follows that upon substituting for the equilibrium employer choice of tax evasion jr ðaÞ  of ðaÞj, where an asterisk denotes equilibrium values, the implied equilibrium likelihood of a tax   ðaÞ  o ðaÞj is a fraction in the [0,1] interval. audit fðaÞ ¼ fjr f 7. Of course, reporting the payment of the minimum wage does not guarantee the actual payment of such a wage. Ashenfelter and Smith (1979), for example, touch on the strategy of issuing a bonafide paycheck. Employers issue a pay check for the amount requirement by the minimum wage, and in exchange, extract some value of the paycheck from workers. By contrast, Saget (2008) points out that there are countries in which payment of the minimum is optional rather than compulsory. This is the case in Indonesia for example where the minimum wage legislation lays out clearly exceptions that can be made for employers who are not able to pay the minimum wage. In our chapter, we focus our analysis the case where the payment of the minimum wage is compulsory by law, and nonpayment of the minimum wage is illegal. We thank an anonymous referee for pointing this out. 8. A full account of a formal employer’s decision problem in the presence of tax and minimum wage enforcement would require a full-length analysis that is beyond the scope of the current study. We leave this for future research. 9. It is straightforward to verify that for any (of, r) pair such that neither of  an alternative pairing ðo0f ; r0 Þ the two constraints bind, (of W wia þ tor) and r4w, can be found along a binding participation constraint that raises expected  pðr  pðr   of Þ2 oa  of  tp ða  r0 Þ  f  0  o0f Þ2 . A forprofit, a  of  tp ða  rÞ  f mal proof is relegated to the appendix. 10. In this, as well as in all of the subsequent maximization problems, it is straightforward to verify that the second order conditions are satisfied. In addition,   ðaÞ  from Eq. (15), the implied equilibrium likelihood of a tax audit f ðaÞ ¼ fðr   in regime I is in the interior of the [0,1] interval since the penalty of ðaÞÞ ¼ 1=ð2pÞ parameter p is assumed to take on values greater than or equal to unity. 11. World Bank (2011) data is from 2010 survey (latest data available for 2011) and PricewaterhouseCoopers (2011) is also the latest available (accessed 2011).   ðaÞ  o ðaÞj ¼ 12. The associated equilibrium likelihood of tax audit f ðaÞ ¼ fjr f  continues to lie in the interior of the [0,1] interval for p  1. 1=ð2pÞ   ðaÞ  o ðaÞj ¼ 13. The associated equilibrium likelihood of tax audit f ðaÞ ¼ fjr f  wð1   to ÞÞ; 1=ð2pÞÞ  lies in the interior of the [0,1]   to Þ  wi a 2 ððtp  to Þ=ð2pð1 f½  pÞÞ=ð1   to ÞÞ in this regime, and interval for w 2 ððwi a þ dÞ=ð1  to Þ; ðwi a þ 1=ð2f p  1.   ðaÞ  o ðaÞj ¼ 14. The associated equilibrium likelihood of tax audit f ðaÞ ¼ fjr f    to ÞÞ; 1=ð2pÞÞ  lies in the interior of the [0,1] fað1  to  wi Þ 2 ððtp  to Þ=ð2pð1  pÞÞ=ð1   to ÞÞ in this regime, and interval for w 2 ððwi a þ dÞ=ð1  to Þ; ðwi a þ 1=ð2f p  1. 15. We follow convention and refer a flat tax reform to a policy of constant marginal tax rate.

Tax Evasion, Minimum Wage Noncompliance, and Informality

49

16. It can also be verified that evaluated at this optimal tax rate ð1  bÞð1  wi Þ, the post tax net output of a formal sector worker exceeds that of an informal sector worker, consistent with our assumption that 1  to  wi 40 throughout this chapter.

ACKNOWLEDGMENT We are grateful to Hartmut Lehmann for encouraging us to write this chapter.

REFERENCES Albrecht, J., Navarro, L., & Vronman, S. (2009). The effects of labor market policies in an economy with an informal sector. Economic Journal, 119, 1105–1129. Almeida, R. K., & Carneiro, P. (2011). Enforcement of labor regulation and informality. IZA Discussion Paper No. 5902. IZA, Bonn. Allingham, M. G., & Sandmo, A. (1972). Income tax evasion: A theoretical analysis. Journal of Public Economics, 1(3–4), 323–338. Amaral, P. S., & Quintin, E. (2006). A competitive model of the informal sector. Journal of Monetary Economics, 53(7), 1541–1553. Andreoni, J., Erard, B., & Feinstein, J. (1998). Tax compliance. Journal of Economic Literature, 36(2), 818–860. Ashenfelter, O., & Smith, R. S. (1979). Compliance with the minimum wage law. Journal of Political Economy, 87(2), 333–350. Baanante, M. J. (2004). Minimum wage effects under endogenous compliance: Evidence from Peru. Econo´mica, 50(1–2), 85–123. Basu, A. K., Chau, N. H., & Kanbur, R. (2010). Turning a blind eye: Costly enforcement, credible commitment and minimum wage laws. Economic Journal, 120, 244–269. Basu, A. K., Chau, N. H., & Kanbur, R. (2011). Contractual dualism, market power and informality. CEPR Discussion Paper No. 8485. CEPR, London. Bosch, M., & Maloney, W. F. (2010). Comparative analysis of labor market dynamics using markov processes: An application to informality. Labour Economics, 17(4), 621–631. Card, D., & Krueger, A. B. (1995). Myth and measurement: The new economics of the minimum wage. Princeton, NJ: Princeton University Press. Chandra, V., & Khan, M. A. (1993). Foreign investment in the presence of an informal sector. Economica, 60, 79–103. Chong, A., Galdo, J., & Saavedra, J. (2008). Informality and productivity in the labor market in Peru. Journal of Economic Policy Reform, 11(4), 229–245. Cobham, A. (2005). Tax evasion, tax avoidance, and development finance. Working Paper No. 129. Queen Elisabeth House. University of Oxford, Oxford, UK. Cowell, F. (1990). Cheating the government: The economics of evasion. Cambridge, MA: MIT Press. Danziger, L. (2010). Endogenous monopsony and the perverse effect of the minimum wage in small firms. Labour Economics, 17, 224–229.

50

ARNAB K. BASU ET AL.

De Paula, A., & Scheinkman, J. (2010). Value added taxes, China effects and informality. American Economic Journal: Macroeconomics, 2, 195–221. de Soto, H. (1989). The other path. New York, NY: Basic Books. Djankov, S., La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (2002). The regulation of entry. Quarterly Journal of Economics, 117(1), 1–37. Djankov, S., Lieberman, I., Mukherjee, J., & Nenova, T. (2003). Going informal: Benefits and costs. In B. Belev (Ed.), The informal economy in the EU accession countries: Size, scope, trends and challenges to the process of EU enlargement (pp. 63–80). Sofia: CSD. Doeringer, P. B., & Piore, M. J. (1971). Internal labor markets and manpower analysis. Lexington, MA: Heath. Fei, J. C. H., & Ranis, G. (1964). Development of the labor surplus economy. Homewood, IL: Irwin. Feige, E. L. (1989). The underground economies: Tax evasion and information distortion. Cambridge: Cambridge University Press. Fields, G. (1975). Rural urban migration, urban unemployment and underemployment, and job search activity in LDCs. Journal of Development Economics, 2(2), 165–187. Friedman, E., Johnson, S., Kaufmann, D., & Zoido-Lobato´n, P. (2000). Dodging the grabbing hand: The determinants of unofficial activity in 69 countries. Journal of Public Economics, 76, 459–493. Fuest, C., & Riedel, N. (2009). Tax evasion, tax avoidance and tax expenditures in developing countries: A review of the literature. Report prepared for the UK Department for International Development. Oxford University Center for Business Taxation, Oxford, UK. Fugazza, M., & Jacques, J.-F. (2004). Labor market institutions, taxation and the underground economy. Journal of Public Economics, 88(1–2), 395–418. Gabrieli, T., Galvao Jr., A. F., & Montes-Rojas, G. V. (2011). Who benefits from reducing the cost of formality? Quantile regression discontinuity analysis. Paper presented at the IZA/ World Bank Workshop: Institutions and informal employment in emerging and transition economics, IZA, Bonn. Gindling, T., & Terrell, K. (1995). The nature of minimum wages and their effectiveness as a wage floor in costa rica, 1976–1991. World Development, 23, 1439–1458. Gindling, T., & Terrell, K. (2006). Minimum wages, globalization and poverty in Honduras. IZA Discussion Paper No. 2497. IZA, Bonn. Harris, J. E., & Todaro, M. P. (1970). Migration, unemployment and development: A two sectors analysis. American Economic Review, 60, 126–142. Jonasson, E. (2011). Government effectiveness and regional variation in informal employment. Paper presented at the IZA/World Bank Workshop: Institutions and informal employment in emerging and transition economics, IZA, Bonn. Jung, Y. H., Snow, A., & Trandel, G. A. (1994). Tax evasion and the size of the underground economy. Journal of Public Economics, 54, 391–402. Kesselman, J. R. (1989). Income tax evasion: An intersectoral analysis. Journal of Public Economics, 38(2), 137–182. Koettl, J. (2011). Does formal work pay? Paper presented at the IZA/World Bank Workshop: Institutions and informal employment in emerging and transition economics. IZA, Bonn. Lemos, S. (2004). The effects of the minimum wage in the formal and informal sectors in Brazil. IZA Discussion Paper No. 1089. IZA, Bonn.

Tax Evasion, Minimum Wage Noncompliance, and Informality

51

Lemos, S. (2006). Minimum wage effects in a developing country. Mimeo, University of Leicester, Leicester. Loayza, N. (1996). The economics of the informal sector: A simple model and some empirical evidence from Latin America. Carnegie-Rochester Conference Series on Public Policy, 45, 129–162. Loayza, N., Serve´n, L., & Oviedo, A. M. (2005). The impact of regulation on growth and informality: Cross-country evidence. World Bank Policy Research Working Paper WPS 3623. Washington, DC. Maloney, W., & Nunez, J. (2004). Measuring the impact of minimum wages: Evidence from Latin America. In J. Heckman & C. Page`s (Eds.), Law and employment. Lessons from Latin America and the Caribbean. Cambridge, MA: NBER. OECD. (2009). Is informal normal? Towards more and better jobs. Paris: OECD. Perry, G., Maloney, W., Arias, O., Pajnzylber, P., Mason, A., & Saavera-Chanduvi, J. (2007). Informality: Exit and exclusion. Washington, DC: The World Bank. PricewaterhouseCoopers. (2011). Worldwide tax summaries. Retrieved from http:// www.pwc.com/gx/en/worldwide-tax-summaries/index.jhtml. Accessed in August 2011. Saavedra, J., & Chong, A. (1999). Structural reforms, institutions and earnings: Evidence from the formal and informal sectors in Urban Peru. Journal of Development Studies, 35(4), 95–116. Saget, C. (2008). Fixing minimum wage levels in developing countries: Common failures and remedies. International Labour Review, 147(1), 25–42. Schneider, F. (2005). Shadow economies around the world: What do we really know? European Journal of Political Economy, 21(3), 598–642. Schneider, F. (2011). The shadow economy and shadow economy labor force: What do we (not) know? IZA Discussion Paper No. 5769. IZA, Bonn. Schneider, F., & Enste, D. (2000). Shadow economies: Size, causes, and consequences. Journal of Economic Literature, 38(1), 77–114. Slonimczyk, F. (2011). The Effect of taxation and informal employment: Evidence from the Russian flat tax reform. Paper presented at the IZA/World Bank Workshop: Institutions and informal employment in emerging and transition economics. IZA, Bonn. Stiglitz, J. E. (1974). Alternative theories of wage determination and unemployment in LDC’s: The labor turnover model. Quarterly Journal of Economics, 88(2), 194–227. Straub, S. (2005). Informal sector: The credit market channel. Journal of Development Economics, 78(2), 299–321. Strobl, E., & Walsh, F. (2001). Minimum wage and compliance: The case of trinidad and tobago. Economic Development and Cultural Change, 51(2), 427–450. Tanzi, V. (1980). Underground economy and tax evasion in the United States: Estimates and implications. Banca Nazionale del Lavoro Quarterly Review, 32, 427–453. Tanzi, V. (1999). Uses and abuses of estimates of the underground economy. Economic Journal, 109, F338–F347. Tonin, M. (2011). Minimum wage and tax evasion: Theory and evidence. IZA Discussion Paper No. 5660. IZA, Bonn. Weil, D. (2004). Compliance with the minimum wage: Can government make a difference. Working Paper HR 05. Harvard Center for Textile and Apparel Research. Harvard University, Cambridge, MA. World Bank. (2010). Doing business project. Retrieved from http://www.doingbusiness.org/ World Bank. (2011). Enterprise surveys. Retrieved from https://www.enterprisesurveys.org/

52

ARNAB K. BASU ET AL.

Yaniv, G. (1988). Withholding and non-withheld tax evasion. Journal of Public Economics, 35, 183–204. Yaniv, G. (1990). Tax evasion under differential taxation. Journal of Public Economics, 43, 327–337. Yitzhaki, S. (1974). A note on ‘income tax evasion: A theoretical analysis’. Journal of Public Economics, 3(2), 201–202.

Tax Evasion, Minimum Wage Noncompliance, and Informality

53

APPENDIX In this appendix, we verify that for any ðof ; rÞ pair such that neither of the  an alternative pairing ðo0f ; r0 Þ two constraints bind, of 4wi a þ to r and r4w, satisfying the a binding participation can be found that raises expected profit,  pðr   of Þ2 oa  of  tp ða  r0 Þ  f  0  o0f Þ2 . a  of  tp ða  rÞ  f pðr To see this, note that by adding and subtracting ðtp  to Þðr  of Þ=ð1  to Þ and rearranging terms, 1  tp   of Þ2 ¼ ðað1  to Þ  of þ rto Þ a  of  tp ða  rÞ  f pðr 1  to  2 tp  to þ f p  pð1   to Þ 2f  2 tp  to   fp  ðr  of Þ  pð1   to Þ 2f 1  tp ðað1  to Þ  of þ rto Þ 1  to 1  tp o ðað1  to  wi ÞÞ 1  to  0  o0f Þ2 ¼ a  o0f  tp ða  rÞ  f pðr where o0f ¼ wi a þ r0 to from (I), r0 ¼ ðwi a þ dÞ=ð1  to Þ, and d ¼ ðtp  to Þ=  pð1   to ÞÞ. ð2f

CHAPTER 2 THE EFFECT OF TAXATION ON INFORMAL EMPLOYMENT: EVIDENCE FROM THE RUSSIAN FLAT TAX REFORM Fabia´n Slonimczyk ABSTRACT The 2001 Russian tax reform reduced average tax rates for the personal income tax and the payroll or social tax. It also made the tax structure more regressive. Because individuals in the lower income bracket were for the most part not affected, it is possible to estimate the effects of the reform using a differences-in-differences approach. I study the effect of the reform on informal employment. Informality is defined using information on employment registration and self-employment. Applying parametric and semi-parametric techniques, I find evidence that the tax reform led to a significant reduction in the fraction of informal employees. Among the different forms of informality I study, the reform seems to have had the strongest effect on the prevalence of informal irregular activities. I also document stronger effects on individuals who benefited from the largest reductions in tax rates. The strong response to the tax reform is consistent with the emerging consensus in the literature on

Informal Employment in Emerging and Transition Economies Research in Labor Economics, Volume 34, 55–99 Copyright r 2012 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1108/S0147-9121(2012)0000034005

55

FABIA´N SLONIMCZYK

56

taxation that changes to the tax system lead to significant behavioral responses, although not necessarily in the form of a reduced labor supply. Keywords: Informal sector; entrepreneurship; tax reform; differencesin-differences; transition; Russia JEL classification: H24; J3; O17; P2

INTRODUCTION The high prevalence of informality is a well-known characteristic of labor markets in developing countries.1 Informal work is by nature heterogeneous. It includes self-employed individuals, as well as those working under them (often a few family members and friends). It also extends to those employed by larger organizations but who are not effectively covered by any of the institutions – such as the pension system and other social insurance – that protect formal employees. While for the most part the economic activities of informal workers are legal, they are often not taken into account in official statistics, and much of the income they generate goes untaxed. An important unresolved issue is to what extent informality in the labor market is responsive to the level of taxation. Trying to answer this question brings about some serious challenges. First and foremost, observable variation in taxation levels is generally endogenous and therefore inappropriate for estimating a causal effect. In this paper, I exploit a natural experiment in order to establish a causal link between the tax wedge – the overall tax burden imposed on employers and employees – and the incidence of informal employment. In 2001, Russia introduced a tax reform that drastically reduced taxation levels for upper income brackets. The pre-reform progressive personal income tax rates were replaced by a flat and low rate of 13%. Payroll taxes were also modified. Before the reform, employers had to make contributions – adding up to 38.5% of the gross salary – to four different social funds. Starting in 2001, these contributions were unified into a single social tax with a regressive scale. If lower levels of taxation causally affect informality, then such a comprehensive tax reform should have had a measurable impact. Because individuals in the lower income brackets were almost unaffected, the reform also created well-defined treatment and control groups. Thus, the effect of the tax reform on informality can be

The Effect of Taxation on Informal Employment

57

estimated following a differences-in-differences (DID) strategy. Intuitively, the DID estimator captures the post-reform average drop in the probability of participating in the informal sector experienced by the treatment group relative to the control group. Other specific institutional features of the tax reform make it a worthy object of study. First, the speed with which the reform was discussed and implemented made any anticipation effect extremely unlikely. The details of the changes to the tax code only started being discussed in the middle of the year 2000. Also, less than 50 days elapsed between the day the reform was announced to the day of its final presidential approval. Second, the flat tax schedule eliminated any incentives to report income just below the critical threshold separating treatment and control groups. Had the reform reduced taxes for the upper brackets while retaining ‘‘progressive’’ marginal tax rates, the treatment variable would be susceptible to systematic measurement error around the threshold. With a flat tax, however, income misreporting is still present but probably a much less serious problem for estimation purposes. While the tax reform provides an extraordinary opportunity to shed light on the research question, it is nevertheless not perfect. Treatment status is determined by the individual’s tax bracket, which in turn is bound to be correlated with informal employment. The paper addresses this issue in a number of ways. First, the treatment effect is estimated conditional on timevarying observable characteristics as well as time-invariant observables and unobservables. Second, I show that these estimates are robust to a wide range of reasonable modifications to the treatment definition. Third, I use two pre-reform years to study the ‘‘effect’’ of a placebo reform (and, reassuringly, find none). Fourth, I implement a weighted DID estimator that identifies the treatment effect by relying heavily on individuals whose income is closer to the bracket threshold and who therefore are less likely to differ systematically in time-varying unobservable characteristics. Finally, I also estimate the treatment effect using the matching DID estimator originally developed by Heckman, Ichimura, and Todd (1997). The research question also requires addressing a second challenge, namely measuring labor market informality to a reasonable degree of precision. My main data source is the Russian Longitudinal Monitoring Survey (RLMS). As is standard in the literature, informal status is determined on the basis of self-reported information on the type of production unit and registration of the employment relation. The wealth of data in the RLMS permits identifying informality in up to three distinct ‘‘jobs’’ or remunerated activities: the main job, a secondary job, and

58

FABIA´N SLONIMCZYK

irregular activities. In addition, the paper takes advantage of a special supplement on informal employment that was added to the main questionnaire of the RLMS in 2009. I use these additional data to cross-validate the definitions of informal employment. The main findings of the paper are that after controlling for observable characteristics and individual fixed effects, employed individuals who were affected by the tax reform were on average 2.5% less likely to be informal employees on the main job and 4% less likely to perform informal irregular activities.2 On the extensive margin, I find that individuals who were not employed right before the reform and found a job in its aftermath were also less likely to be informally employed. This paper builds upon the small number of existing studies of the Russian tax reform.3 Using a similar methodology to the one employed here, Ivanova, Keen, and Klemm (2005) and Gorodnichenko, MartinezVazquez, and Sabirianova-Peter (2009) documented positive effects of the reform on public revenue and tax compliance at the household level. Duncan and Sabirianova-Peter (2009) find that the reform resulted in very modest increases in male and female hours of work. Taken together, these studies and the present paper support the conclusion common in the modern literature on taxation, that the main form of response to changes in the tax system is not through labor supply but through other – equally important – margins of adjustment (Saez, Slemrod, & Giertz, forthcoming). This paper is also closely related to the burgeoning literature on the determinants of the size of the unofficial economy. The unofficial – also called shadow, hidden, or underground – economy refers to the production, whether legal or illegal, of goods and services for the market that escapes detection in the official estimates of GDP (Schneider & Enste, 2000). While the definition and the units of measurement of informal employment are different, in practice there is a strong overlap between the two concepts since a large proportion of informal work is probably not registered in official statistics and vice versa. One widely accepted interpretation is that underground economic activity is a response to excessive involvement of the State in the economy in the form of intrusive regulations and high levels of taxation. Among post-communist countries, there is evidence that only those which succeeded in limiting the political control of economic activity (at the same time as they improved the provision of key public goods necessary for the good functioning of markets) seem to have managed to keep the growth of the unofficial economy under control (Johnson, Kaufmann, Shleifer, Goldman, & Weitzman, 1997; Mcmillan & Woodruff, 2002).

The Effect of Taxation on Informal Employment

59

Modern Russia seems like a perfect illustration of the theory linking excessive government intervention and the shadow economy. Russian managers face higher effective tax rates, worse bureaucratic corruption, greater incidence of mafia protection, and have less faith in the court system than their peers in Slovakia, Poland and Romania, and that seems to go some way in explaining why Russia’s underground economy is relatively larger (Johnson, Kaufmann, McMillan, & Woodruff, 2000). Also, Russia inherited an unregulated sector from the Soviet times. Grossman (1977) coined the term ‘‘second economy’’ for the set of illegal and quasi-legal economic activities that individuals engaged in to put up with or exploit the severe rationing of goods and services under communism. Such activities encompassed the cultivation of small plots of land, simple stealing from state enterprises, speculation, illicit production at secondary occupations, and many others. In 1990, almost 15% of personal income of workers and employees had informal sources (Kim, 2003). In other words, the incipient Russian market economy inherited the ability to avoid regulation by the state when such regulation is too costly or otherwise excessive (Gerxhani, 2004; Guariglia and Kim, 2006). Using different methods and definitions, several studies have documented a rising share of the underground activity in Russia during the 1990s (Lacko, 2000).4 There are, however, reasons to believe that the statistical association between excessive regulation and a growing unofficial sector is not causal. Firms might decide to operate underground mainly in order to avoid predatory behavior by government officials rather than regulations per se (Johnson, Kaufmann, and Zoido-Lobaton, 1998). If that is the case, then it is not so much the letter of the law – for example, mandating high taxes – that influences informality but rather the discretional authority of administrative officials in the context of a corrupt administrative system. To the extent that informal employment is a good proxy for unofficial activities, my estimates of the effect of the tax reform can also be interpreted as a test of the theory that the size of the shadow economy is boosted by excessive regulation. The paper is organized as follows. The next section briefly discusses the data and the definition of informal employment. In the third section I present a descriptive analysis of informal workers in Russia using alternative data sources. The fourth section focuses on the structure of the tax reform and the definition of the treatment and control groups. The fifth section presents the main results for the intensive and extensive margins, as well as a number of robustness checks. The sixth section concludes.

FABIA´N SLONIMCZYK

60

INFORMALITY DEFINITION AND MEASUREMENT The main data source for this study is the RLMS. In this section I briefly describe the RLMS and discuss my working definition of informal employment.

Data and Variables The RLMS is a household panel survey based on the first national probability sample drawn in the Russian Federation.5 I use data from rounds VIII–XVIII covering the period 1998–2009. In a typical round, 10,000 individuals in 4,000 households are interviewed. These individuals reside in 32 oblasts (regions) and 7 federal districts of the Russian Federation. A series of questions about the household (referred to as the ‘‘family questionnaire’’) are answered by one household member selected as the reference person. In turn, each adult in the household is interviewed individually (the ‘‘adult questionnaire’’). The adult questionnaire includes questions regarding a primary and a secondary job. In addition, individuals are also asked whether they perform what I will refer to as ‘‘irregular remunerated activities.’’ The exact phrasing of this last questionnaire item is as follows: ‘‘Tell me, please: in the last 30 days did you engage in some additional kind of work for which you were paid or will be paid? Maybe you sewed someone a dress, gave someone a ride in a car, assisted someone with apartment or car repairs, purchased and delivered food, looked after a sick person, sold purchased food or goods in a market or on the street, or did something else that you were paid for?’’ The questionnaire structure is such that no one may answer questions on a secondary job unless they have a primary job. However, questions on the irregular activities are independent.6 Fig. 1 shows the employment and the unemployment rate, according to the RLMS and the standard labor force survey conducted by Rosstat. While the two data sources display some minor discrepancies,7 all series show that the period under analysis was – at least employment-wise – one of relative economic prosperity and stability. In order to gain further insight into the meaningfulness of my informality variables, I also make use of a special supplement of questions on informal work (INFSUP8) that was added to the RLMS adult interview in 2009 (round XVIII). The INFSUP questionnaire was administered to all employed individuals after the regular interviews had been completed.9

61

The Effect of Taxation on Informal Employment

60

50

40

30

20

10

0 1998

1999

2000

RLMS Emp

Fig. 1.

2001

2002

2003

RLMS Unemp

2004

2005

2006

LFS Emp

2007

2008

2009

LFS Unemp

Employment and Unemployment Rates. Note: RLMS, rounds VIII–XVIII and Rosstat labor force survey (1998–2009).

Definition of Informal Employment As has been clearly put in a recent book-length study by the World Bank: ‘‘The term informality means different things to different people, but almost always bad things: unprotected workers, excessive regulation, low productivity, unfair competition, evasion of the rule of law, underpayment or nonpayment of taxes, and work ‘underground’ or in the shadows’’ (Perry et al., 2007). The idea of the informal sector was originally adopted and popularized by economic anthropologist Keith Hart (1973) and a series of studies sponsored by the International Labour Office (ILO, 1972). Since the beginning, the concept was meant to comprise heterogenous labor practices including petty trading, self-employment of different sorts, own-account professionals, family workers, and other forms of non-standard (from a Western perspective) work prevalent in developing countries. Moreover, many of the initial bounds of the concept were eventually trespassed in one way or another. For example, while the informal sector was originally thought to be predominantly urban, it was quickly accepted that it should also include some forms of small-scale agricultural work. Despite these ambiguities – and partly thanks to them – the concept has proved useful to researchers with a wide range of interests.10

62

FABIA´N SLONIMCZYK

While the literature widely recognizes the blurry bounds of the concept, there are two most commonly used definitions of informality. On the one hand, the so called ‘‘productive’’ definition focuses on a number of characteristics of the production unit (Hussmanns, 2004). First, informal sector enterprises typically include only private unincorporated units, that is, enterprises not constituted as separate legal entities independently of their owners. Second, at least part of the goods or services they produce is meant for sale or barter. Lastly, their scale of operations is assumed to be very small. In fact, when better data is lacking, informal enterprises are often defined as those whose size in terms of employment is below a given threshold (typically less than five employees). On the other hand, the ‘‘legalistic’’ or social protection definition focuses on the status of workers in relation to labor law and the social safety net. It measures to what extent workers are effectively – as opposed to only de jure – protected by labor market institutions. Informal employment occurs in cases of noncompliance to the State in terms of labor regulations and the social security system. In this paper I use both legalistic and productive criteria to determine if an individual is informally employed. Table 1 shows a schematic representation of the different employment types and my working definition of informality in each case. Throughout the paper, I analyze informality at the main job, the secondary job and the remunerated irregular activities separately. At the main job, I start by distinguishing between entrepreneurs and employees. The former group is composed of those doing entrepreneurial activities who are either owners of firms or self-employed individuals who work on their own account with or without employees but not at a firm or organization.11 Following the productive definition, those not working at firms or organizations are considered informal. For those working at firms or organizations, the RLMS questionnaire includes an item that permits determining whether they are registered, that is, working officially.12 The Russian labor code mandates that all employees sign a written contract and deposit their ‘‘labor book’’ with the employer. Therefore, following the social protection criterion, I classify unregistered entrepreneurs and employees as informal. Some firms in Russia register their employees but declare a fictitious salary that is lower than the real amount in order to reduce the base of payroll and other taxes. The difference between the declared and the real salary is settled with an ‘‘envelope payment’’ at the end of the month. If such a practice were widespread, the registration criterion could err on the side of underestimating the extent of informality. Fortunately, the 2009 round of the RLMS included an item on envelope payments. As I show below,

The Effect of Taxation on Informal Employment

Table 1.

63

Working Definitions of Informal Work. Firm ownersa Entrepreneur

Formal Informal

Individual entrepreneurb

Informalc

For firm

Formal

Main job Informald

Employee

Employed

For individual entrepreneur Second job

Irregular activities

Informalc Formal Informale Formal Informalf

a

Firm owners work for a firm or organization that they own and where they perform entrepreneurial activities. Considered informal if unregistered. b Individual entrepreneurs do entrepreneurial activities independently (not within a firm or organization). c Registration information is not available for those not working within firms or organizations. d Employees are considered informal if they are not registered. e Informal in second job if unregistered or not working for a firm or an organization. f Irregular activities involve remunerated work like sewing a dress for someone or giving someone a ride in a car. Considered informal if not employed under official contract or agreement.

workers employed formally rarely admit to high levels of income underreporting. Thus, considering envelop payments would not significantly increase measured informality levels. While using the productive definition to classify all self-employed individuals and their employees as informal is standard practice, it would be reassuring if the social protection criterion could be applied as well. Unfortunately, a limitation of the RLMS data is that non-enterprise individuals are not asked about registration, so it is not possible to apply the legalistic definition to them.13 However, thanks to the INFSUP we have some good indication of the extent to which self-employed individuals comply with the regulations. As I show below, the level of compliance is quite low, so choosing between the legalistic and productive definition does not make a big difference for these workers. The supplementary questions also confirm that there is a high correlation between lack of registration and other forms of noncompliance with labor regulations.

FABIA´N SLONIMCZYK

64

In principle, the RLMS questionnaire contains enough detail to treat the main and the second job symmetrically. However, the number of observations would not be large enough for a meaningful statistical analysis of the resulting sub-categories. For example, only about 40 individuals per round do entrepreneurial activities in the second job. Therefore, a single category of informal work in the second job is considered, consisting of those unregistered14 plus those not working for a firm or organization. Finally, I consider remunerated irregular activities. Based on the productive definition, all employment of this kind could be classified as informal. However, since not much information is available regarding these activities I only consider them informal if the respondent gives a negative answer to the question: ‘‘Tell me, were you employed in this job officially, for example by an agreement, an official contract, or a license?’’15 This methodological decision is unlikely to affect results16 since almost 87% of irregular work is done without a contract. According to these definitions, Fig. 2 shows the evolution of informal employment participation rates over the period and provides some

16

12

9

12

6

8

3

4

0

0 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Entrep Inf Employee

Inf Entrep

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Sec Job

Irreg Act

Inf Sec

Inf Irreg

Fig. 2. Informality at Main Job, Second Job, and Irregular Activities. Note: RLMS, rounds VIII–XVIII (1998–2009). The left panel shows informality at the main job, disaggregated into employees and entrepreneurs. The panel on the right shows informality at the second job and remunerated irregular activities. The series for the main and the second job are defined as a percentage of those with a main job. For the irregular activities, the base are all those employed.

65

The Effect of Taxation on Informal Employment

preliminary evidence on the likely effects of the tax reform. First, the fraction of entrepreneurs in the main job – both formal and informal – has remained stable at around 4.5%. Second, informality among employees has risen almost uninterruptedly and toward the end of the period is well into the double digits. Eyeballing the time series suggests the tax reform might have caused a deceleration of the rate of growth of informal employment in the short run. However, no long run effect is apparent. Third, the percentage of second job holders of any kind has also not changed much during these 11 years. Informality in the second job is relatively uncommon. Finally, prima facie there seems to be a strong negative effect of the tax reform on the prevalence of irregular activities, informal or otherwise. This is important since, at least until the reform was implemented, irregular activities were the most common form of informal work in Russia. However, simple before-after comparisons are risky. Fig. 3 presents the evolution of real hourly wages for workers in the formal and informal sector. Real incomes were increasing over the period and, to the extent that a growing economy induces formalization, it might well be the case that the tax reform had very little to do with the downward trend in informal irregular activities after 2001. On the other hand, if what matters are relative

150

100

50

0 1998

2000 Inf Irreg Act Exclusive

2002 Inf Irreg Act

2004 Formal Employee

2006 Inf Employee

2008 Inf Entrep

Fig. 3. Real Hourly Earnings (in 2,000 Rubles). Note: RLMS, rounds VIII–XVIII (1998–2009). Real hourly earnings are monthly receipts divided by usual hours and deflated by the CPI. Informal irregular activities exclusive means that the individual held no other job.

FABIA´N SLONIMCZYK

66

rewards between the formal and informal sector, then it seems unlikely that the modest changes in relative wages shown in the figure might explain quantitatively large sectoral shifts.

DESCRIPTION OF INFORMAL EMPLOYMENT IN RUSSIA While my working definition of informality has many antecedents,17 it is also somewhat idiosyncratic to the extent that the questionnaire items of the RLMS are unique and that not much is known about the informal sector in modern Russia. There could be legitimate concerns regarding the correctness of the resulting measure of informality. Fortunately, some insight can be gained thanks to the wealth of information in the regular RLMS survey and in the INFSUP. In this section, I show that workers that are informal according to my definition have many of the characteristics found in other studies. I also present evidence that alternative definitions, while reasonable, would probably not affect the results. Demographics Table 2 provides demographic information on informal workers in Russia toward the end of 2009. The table confirms many of the empirical regularities observed in other countries. For example, informal employees tend to be low skilled. Only around 12% of them has a college degree and their level of schooling is below that of the average Russian worker. They are also relatively younger, predominantly male and less experienced. Workers performing informal irregular activities18 seem to show many of the same characteristics, although a larger share of them live in rural areas and belong to one of the many ethnic minorities. As in other countries, individual entrepreneurs in Russia are relatively well off. While they are less educated than average workers, their qualifications are not as low as those of informal employees. Entrepreneurs are also relatively more likely to marry and form a family. Individuals who participate in the informal sector through a secondary job also have higher than average incomes. In almost all other respects, however, they are difficult to be distinguished from the average worker. The 2009 round of the RLMS included an item on ‘‘envelope payments.’’19 Formal employees answered that 92% of their earnings were reported

67

The Effect of Taxation on Informal Employment

Table 2.

Female Age College degree Schooling (years) Experience Married Urban location Russian national Russian born Size HH

Background Characteristics of Informal Workers in Russia. All Employed

Informal Employee

Informal Entrepreneur

Informal Second Job

Informal Irregular Activities

0.54 39.5 0.27 12.3

0.49 36.4 0.12 11.5

0.42 40.1 0.23 12.1

0.56 38.9 0.28 12.5

0.45 38.6 0.15 11.4

14.3 0.51 0.77 0.87

9.2 0.42 0.76 0.86

14.4 0.66 0.80 0.77

14.8 0.48 0.88 0.86

11.3 0.42 0.63 0.81

0.91 3.4

0.88 3.5

0.82 3.6

0.87 3.0

0.92 3.4

11,043

18,661

7,142

7,043

32.0

62.9

NA

NA

11,132

18,878

17,024

12,470

815

204

158

583

‘‘After-tax’’ income This job 13,194 (rubles) % Reported 86.6 for tax All jobs 13,446 (rubles) Observations

7,192

Note: The data source is RLMS, round XVIII (2009). Employed workers are those with a job or who do remunerated irregular activities. Informal employees are those who work for a selfemployed individual or who work for a firm or an organization but are not registered. Informal entrepreneurs are either self-employed or owners of a firm who do entrepreneurial activities but are not registered. Informal second job includes both informal employees and informal entrepreneurs in their second job, regardless of the main job status. Informal irregular activities are other remunerated activities conducted without formal contracting.

to tax authorities. In turn, informal employees and individual entrepreneurs reported having payed taxes on a significantly lower fraction of earnings. While responses to such sensitive issues cannot be taken at face value, the high correlation between informality and declared tax avoidance is reassuring. Informal workers overwhelmingly belong to unskilled and service occupations, and work in the trade and construction industries.20 This is consistent with the idea that informal workers work in occupations/ industries with low barriers to entry – that is, requiring almost no start-up capital or specific knowledge.

FABIA´N SLONIMCZYK

68

Job Characteristics The standard RLMS survey offers some detailed information regarding the characteristics of the job,21 which I present in Table 3. Informal employees have relatively weak attachment to the job, as indicated by the low observed average tenure. Moreover, the probability of Table 3.

Job Characteristics for Informal Workers in Russia. All Employed

Informal Employee

Informal Entrepreneur

Informal Second Job

7.3 0.16 0.11 0.20

2.8 0.35 0.21 0.08

7.2 0.13 0.06 0.38

2.5a NA NA 0.10b

584.4

61.8

–

0.50 0.56

0.06 0.91

– –

0.20 0.70

0.59 0.07 0.01

0.09 0.13 0.03

– – –

0.40 0.19d 0.02d

Job benefitsc Paid vacation Paid sick leave Paid maternity leave Paid health care Paid trips to sanatoria Paid child care Assistance w/food Assistance w/transport Paid educational activities Assistance w/loans

0.90 0.87 0.79 0.24 0.28 0.05 0.12 0.12 0.25 0.05

0.17 0.11 0.07 0.01 0.01 0.01 0.04 0.03 0.02 0.01

– – – – – – – – – –

0.19 NA 0.17 0.05 0.03 0.01 0.03 0.01 0.04 0.00

Observations

7,192

815

204

158

Tenure (years) Changed jobs Changed occupation Has subordinates Firm characteristicsc Enterprise size (x of employees) State owns share Russian individual owns share Firm from Soviet times Firm owes money Firm pays in kind

76.2

Note: The main data source is RLMS, round XVIII (2009). Definitions are the same as for Table 2. a From round XVI (2007). b From round XVII (2008). c Only for those working for firms or other organizations. d From round XIV (2005).

The Effect of Taxation on Informal Employment

69

transition implies a mean job duration of only 1.5 years (E1/10.35). Informal second jobs seem to have short durations too. While we lack information regarding average duration of irregular activities, over 66% of workers answered affirmatively to a specific item asking whether these activities were ‘‘incidental’’. Interestingly, however, this is not the case with informal entrepreneurs, who have below average transition probabilities. According to the RLMS, almost 90% of non-enterprise individuals work alone or with a few family members.22 Consequently, a number of items in the adult questionnaire are only asked to individuals who work for firms.23 First, respondents are asked about the size of the firm. Table 3 confirms that informal employees work for firms that, while larger than a family enterprise, are still much smaller than average.24 This is also true of individuals who are informal in the second job. Second, there are questions regarding firm ownership and origin. The issue of informality and the shadow economy in Russia is often discussed in the context of the transition from the Soviet system (Johnson et al., 1997; McMillan & Woodruff, 2002). A familiar argument is that the incipient capitalist sector makes use of informal arrangements to escape confiscatory intrusions by the State. The data is consistent with this story. The involvement of the Russian State in the economy is very substantial. This is reflected not only in the relatively high prevalence of state ownership but also in the fact that almost 60% of employment in Russia is still supplied by enterprises that originate in Soviet times. Informal employment is, however, almost exclusively provided by new private firms. A third important set of questions touch on the issue of wage arrears. Faced with a negative shock, firms in Russia often choose to adjust via delaying the payment of wages (Gimpelson & Kapeliushnikov, 2011; Lehmann, Wadsworth, & Acquisti, 1999). Predictably, Table 3 shows that wage arrears and payments in kind happen relatively more frequently to informal employees. Finally, the RLMS asks enterprise workers regarding fringe benefits. Paid vacation, sick leave, and maternity leave are mandatory benefits according to the labor code and a large majority of employees claim to have them. However, many firms do not provide these benefits in practice. For example, only 66% of those employed had actually been on paid vacation in the previous 12 months, compared to the 90% who claim to have entitlement. In any case, the proportion of informal employees who are given the mandatory benefits is substantially lower than average.25 Non-mandatory benefits are infrequent in Russia, and almost non-existent for informal employees.

FABIA´N SLONIMCZYK

70

Table 4.

Informal Activities Last Year.

All Employed

Informal Employee

Informal Entrepreneur

Informal Second Job

Informal Irregular Activities

Worked extra job Raised cattle for sale Agriculture on own plot for sale Performed services for pay

0.09 0.04 0.04

0.08 0.03 0.02

0.08 0.04 0.03

0.96 0.03 0.04

0.33 0.14 0.14

0.08

0.08

0.06

0.11

0.61

Observations

7,192

815

204

158

583

Note: The data sources is RLMS round XVIII.

While informative, these questions provide insight into only a minority of informal jobs, that is, those which happen at firms or organizations. Nevertheless, the RLMS includes a series of questions on informal activities during the previous 12 months that are asked to everyone. A summary of these items is in Table 4. Two points are noteworthy. First, 9% of those employed reported having worked an informal second job in the previous year. Reassuringly, the agreement with informality in the second job according to my definition is almost perfect. Second, almost 40% of individuals who perform irregular activities live in rural areas. Not coincidentally, a significant proportion of them are involved in small-scale agriculture and husbandry. However, by far the largest share of these activities involves personal services: taxi rides, repair work, hair styling, tutoring, nursing, etc. Compliance with the Law Table 5 contains statistics based on answers to the INFSUP. An important cautionary note is that the INFSUP consisted of a stand-alone questionnaire that was administered to all individuals who had any form of employment. Respondents answered informality-related questions about two jobs (henceforth26 job-A and job-B). Unfortunately, these jobs are not certain to correspond to those of the standard adult questionnaire.27 I proceed as follows. I assume that the information about job-A corresponds to the main job if such a job is present. For individuals without a main job, I assume job-A must refer to (the main) remunerated irregular

71

The Effect of Taxation on Informal Employment

Table 5. Supplement for Employees

Compliance with the Law.

All Employed

Informal Employee

Informal Second Job

Informal Irregular Activities

Under oral agreement % Labor law compliance % Contract compliance % of income declared for SS

0.11 83.1

0.69 52.9

0.81a NA

0.96b 53.2b

86.1 87.6

64.3 31.2

NA NA

65.5b 10.5b

Observations

6,453

777

80

186

Supplement for Entrepreneurs

All Employed

Formal Entrepreneur

Informal Entrepreneur

Informal Irregular Activities

Unregistered % Labor law compliance % Contract compliance % Formal employees Contributes to SS fund

0.48 64.4

0.03 85.9

0.27 53.6

0.98b 21.3b

66.3 64.0 0.47

87.5 85.7 0.95

55.5 53.4 0.60

27.5b 8.3b 0.06b

Observations

397

64

194

126

Note: The data sources are RLMS round XVIII and the supplementary questionnaire on informality by the Centre of Labour Market Studies, Higher School of Economics, and the Labor Markets in Emerging and Transition Economies Research Program, IZA (2009). a Based on job-B answers by individuals who do not perform irregular activities. b Based on job-A answers by individuals who do not have a main job.

activity. In fact, all statistics on informal irregular activities are based on the latter group. Finally, I assume that job-B refers to the secondary job as long as the individual does not also perform irregular activities. This is the source of information on informal secondary jobs. A second issue is that the INFSUP asks a different set of questions regarding job-A depending on whether the individual is an entrepreneur or an employee. While for the most part individuals who identify themselves as entrepreneurs in the INFSUP are also classified as such based on the adult questionnaire, the correspondence is not perfect. I base the statistics only on individuals for whom the classifications coincide. A positive spillover is that the INFSUP provides us with some idea of the composition of remunerated irregular activities. A stunning 40% of these workers consider themselves entrepreneurs.

72

FABIA´N SLONIMCZYK

Working under an oral agreement is strictly forbidden under Russian labor law. The INFSUP asks all employees in job-A whether they have a written contract. This question is important for validating my working definition of informality, since the adult questionnaire only has registration information for enterprise workers in the main job. Remarkably, over 97% of those who work under an oral agreement according to the INFSUP are classified as either informal employees or individuals whose only source of income originates in informal irregular activities. The supplement also asks employees about the extent to which their employers comply with labor law and the specifics of the individual labor contract or agreement. These items are interesting because registration is only one of the many mandates of labor law. The average workplace has compliance levels well over 80%. Informal workers report significantly lower levels of compliance. These figures are consistent with the finding (Table 3) that absence of mandatory benefits and wage arrears are more frequent for informal employees. Finally, employees are also asked about the percentage of their earnings that is reported for social security purposes. In general, responses are very much in agreement with a similar item in the RLMS adult questionnaire (Table 2). Thanks to the INFSUP, however, we have information on those performing irregular activities. Predictably, tax compliance is extremely low for this kind of jobs. The questionnaire for entrepreneurs provides information regarding registration of business operations. In Russia, the self-employed can either register individually or as a company. While some form of registration is necessary to operate formally, it is unclear whether it is sufficient. Practically all of the few formal entrepreneurs in the RLMS sample are registered according to the INFSUP. Moreover, 64% of registrations are in the form of incorporated businesses. On the other extreme, individuals performing irregular activities are overwhelmingly unregistered. In between, a majority of those classified as informal entrepreneurs in the main job are registered, but only 17% of them have an incorporated business. Entrepreneurs are also asked a number of questions regarding their employees. On the one hand, in formal firms labor law and contract compliance is high, the share of informal work is low and contributions to social security are very frequent. Informal entrepreneurs, on the other hand, report much lower levels of compliance, specially in the irregular activities sector. Overall, the information in the regular adult questionnaire of the RLMS and the INFSUP confirm that my working definition of informality is

73

The Effect of Taxation on Informal Employment

meaningful and that informal workers in Russia share many of the characteristics documented in other countries.

THE TAX REFORM In January 2001, Russia introduced a radical reform of its tax system. The main components of the reform are shown in Table 6. A number of changes involved the personal income tax (PIT). Before 2001, the PIT had a progressive scale with marginal rates starting at 12% and reaching 30%. The new system fixed a flat and low rate of 13%. The reform touched other aspects of the PIT. The standard allowance was slightly increased, from 3,168 to 4,800 rubles but now could only be claimed by those earning less than 20,000 rubles. Also, the number of permissible deductions and other loopholes was greatly limited. Before the reform, employers were supposed to make separate contributions – adding up to 38.5% of the gross salary – to four independent social funds. The reform replaced this system with a unified social tax (ST) with a regressive scale. It also eliminated the 1% employee contribution to the social fund.

Table 6. Gross Yearly Income (RUR)

The Russian Tax Reform. Before 2000

PIT

Control

Treat1 Treat2 Treat3 Treat4

o3,168 3,168–4,800a 4,800–50,000

0 12 12

50,000–100,000 100,000–150,000 150,000–300,000 300,000–600,000 W600,000

20 20 30 30 30

a

After 2001

ST

PIT

Employee

Employer

1

38.5

1

38.5

0 0 13

13

ST Employee

Employer

0

35.6

0

35.6 20 20 10 2b

Note: The data source for the Personal Income Tax (PIT) and Social Tax (ST) is the Russian Tax Code, part 2 (2001–2002). a The tax allowance in 2001 was only available to those with income below 20,000 rubles. b Rate initially set to 5% and lowered to 2% in 2002.

FABIA´N SLONIMCZYK

74

Overall, the message of the reform was unambiguous. The government was offering a new deal to the Russian public: lower taxation levels and a more reasonable system. In exchange, it expected higher levels of compliance. The response from the public has been widely regarded as positive. Tax compliance improved significantly and government revenue increased despite the lower average tax rates (Gorodnichenko et al., 2009; Ivanova et al., 2005).

Identification of the Tax Reform Effect The combined effect of the PIT and ST reform can be seen in Fig. 4. The tax reform affected the costs and benefits of informality faced by all economic agents. However, some groups were more affected than others. Specifically, people earning less than 50,000 rubles per annum had a net tax reduction of only 1.4%. In comparison, those earning between 50 and 100 thousand rubles faced a reduction of 7.2%. Finally, it is clear from the graph that the greatest reductions in tax burden were received by those earning 100 thousand rubles or more. The design of the reform created a natural experiment that can be exploited to obtain a DID estimate of the effect of lower taxation levels on

50.2

43.0

37.2 35.8

27.5

20.9

14.7

50

100

300

600

Gross Yearly Income (thousands)

Before

Fig. 4.

After

Combined Tax Burden. Note: Russian Tax Code, part 2.

The Effect of Taxation on Informal Employment

75

informality. Individuals earning less than 50,000 rubles a year constitute a ‘‘control group’’ whose marginal tax rate remained practically unchanged. People with higher incomes faced lower tax rates and therefore are considered ‘‘treated’’. The DID identification strategy assumes that the evolution of participation in the informal sector for the control group can be used to estimate what would have happened to individuals in the treatment group had they not been treated. One potential source of concern for studies that exploit natural experiments like the Russian tax reform is anticipation effects. Individuals might try to play the system by changing their behavior before the reform is implemented. While this has been found to be a real issue in the case of capital gains tax reform (Saez et al., forthcoming), it is likely not to be very relevant for labor market behavior. In any case, the speed with which the Russian tax reform was discussed and implemented made any anticipation effect extremely unlikely. The details of the changes to the tax code only started being discussed in the middle of the year 2000. Also, less than 50 days elapsed between the day the reform was announced to the day of its final presidential approval. In practice, the determination of who belongs to the treatment group is complicated by the fact that people misreport income. In particular, a progressive tax system might provide specifically strong incentives to underreport to individuals whose actual income is just above the lower bound of a tax bracket. As formally argued by Gorodnichenko et al. (2009), the flat tax schedule in post-reform Russia means that this is not a reason of concern. Individuals just above the critical threshold of 50,000 rubles did not have special incentives to work less hours or misreport income. Moreover, because tax rates were generally lower and regressive after 2001, it is plausible that misreporting decreased in general. Other elements of the tax system that could potentially introduce biases are itemized deductions and other allowances. The former were almost completely eliminated in the reformed tax code. Moreover, only individuals earning up to 20,000 rubles per year may claim the standard allowance of 4,800 rubles. Finally, the reform also eliminated the exception of income taxes to military personnel. Based on this discussion, the treatment group will be defined based on post-reform reported income only. The RLMS adult questionnaire asks individuals to report their after-tax monthly earnings in each remunerated activity. These items include not only labor income but also ‘‘benefits, revenues, and profits’’, while

FABIA´N SLONIMCZYK

76

excluding pensions and other nontaxable transfers. In order to determine treatment status, I construct an aggregate income variable that adds the amounts received from all sources. In the absence of misreporting, individuals with after-tax monthly income above 3,625 rubles28 can be considered treated. If, however, income is under-reported, some individuals will be incorrectly included in the control group. Thus, the resulting DID estimate is a lower bound of the true effect of the reform on informality. One complication is that an individual’s income may be above the threshold only in some of the post-reform rounds. I consider anyone whose income is ever above the threshold as treated. I also run a series of robustness tests in which treatment is defined based on shorter periods. The control group is given by those untreated and employed in at least one postreform period. The latter definition means that employed individuals who do not report income are included in the control group. In practice this is a very small number of individuals and, as I show below, excluding them from the control group does not affect the results. In principle, individuals with post-reform annual income between 3,168 and 4,800 rubles could be considered treated. However, for this group the annual savings associated with the tax reform have an upper bound of 212 rubles.29 Since this amount is economically insignificant, my baseline specification includes these individuals in the control group. However, as shown below, the results are almost identical if this group is included in the treatment group instead. I report selected statistics on the control and treatment groups in Table A1 in Appendix. Over three-fourth of the sample is in the treatment group in my baseline definition. In short, the treatment group is younger and has less labor market experience, tends to be better educated, and is more likely to be married than the control group. The households of treated individuals are relatively more likely to be in urban areas, are slightly larger, and have more members who are female or young.

RESULTS As a first step into understanding the effect of the reform, I plot the informality time series for the treatment and control groups. The upper left panel of Fig. 5 shows that the reform probably affected informal employees. Before 2001, participation in this kind of informal work was approximately the same in both groups. However, their post-reform behavior was

77

The Effect of Taxation on Informal Employment

Main Job: Informal Employees

Main Job: Informal Entrepreneurs

24 4 20 3

16 12

2 8 1 4 0

0 1998

2000

2002

2004

2006

2008

1998

Informal Second Job

2000

2002

2004

2006

2008

Informal Irregular Activities

4 28 24

3

20 2

16 12

1

8 0

4 1998

2000

2002

2004

2006

2008 0

Control

Treated

1998

2000

2002

2004

2006

2008

Fig. 5. Informal Employment by Treatment. Note: RLMS, rounds VIII–XVIII (1998–2009). Treatment defined based on total after-tax monthly income in the postreform period.

very different. The prevalence of informal employees in the control group experienced a steady increase. The increase of informality among treated individuals was much less significant. While less conspicuous, this pattern is also present for informal entrepreneurs (upper right panel). Before the reform, informality was more prevalent among the treated. By 2009, the control group had a higher proportion of informals. The bottom left panel of the figure shows that the reform did not seem to affect informality in the second job.

78

FABIA´N SLONIMCZYK

Finally, the bottom right panel provides compelling graphic evidence that the tax cuts worked toward reducing informal irregular activities. Overall, Fig. 5 suggests that the tax reform was a success beyond the realm of tax compliance. The reduction in taxation levels seems to have pulled a large number of people into formal status.30 However, there is some chance that the visual evidence is not statistically significant. More importantly, as shown in Table A1, there are some marked observable differences between the treatment and control groups. The figures in the previous section do not control for any of these factors. It is possible that the visual evidence is an artifact of spurious correlation. In order to obtain statistical evidence on the effect of the reform and control for the possible confounding effect of observable characteristics, I estimate the following DID equation: INF it ¼ yt þ X it b þ Z i g þ cPostt þ mTreati þ aðTreati Postt Þ þ uit

(1)

where INFit is one of the informality-related dependent variables, yt are time dummies, Xit and Zi represent sets of time-varying and time-invariant individual characteristics respectively, Postt is a post-reform dummy, Treati is the treatment group indicator, and uit is the error term. The main object of interest is a, the DID parameter that measures the average change in the probability of informal status for the treatment group relative to the control group, conditional on all the observables. Table 7 presents OLS estimates of Eq. (1). The main identifying assumption of OLS-DID is that none of the unobservable characteristics that influence informality participation are correlated with treatment status. Attempting to correctly model the binary outcome variable would impose extra identification conditions, so I avoid it. I report Arellano (1987) standard errors that allow for heteroscedasticity and autocorrelation of arbitrary form.31 The results provide confirmation that the tax reform reduced the prevalence of informal employees. After controlling for all observable individual and household characteristics and for any macroeconomic shocks absorbed by the year dummies, the expected probability of informal status for the control group was 8% higher in the period after the reform. In contrast, informality grew 4% less among those facing lower levels of taxation. These estimates are both economically and statistically significant. The coefficients for the control variables have the expected signs. Informality is less likely among women, Russian nationals, and high-skill and married workers.

79

The Effect of Taxation on Informal Employment

Table 7.

The Effect of Tax Reform on Informality: DID OLS.

Household characteristics Number of members Number of female members Number of youth, 18 Number of elderly, 65 þ Urban location Individual characteristics Female Russian national Age Age2/100 Experience Experience2/100 Secondary schooling completed Vocational schooling completed Technical schooling completed College comp Graduate level comp Married DID estimates Post Treat

Informal Employee

Informal Entrepreneur

Informal Second Job

Informal Irregular Act

0.0006 (0.002) 0.0059 (0.003) 0.0090 (0.003) 0.0106 (0.004) 0.0043 (0.008)

0.0003 (0.002) 0.0027 (0.002) 0.0057 (0.002) 0.0003 (0.003) 0.0178 (0.006)

0.0054 (0.001) 0.0018 (0.001) 0.0058 (0.001) 0.0024 (0.002) 0.0045 (0.003)

0.0033 (0.002) 0.0039 (0.003) 0.0129 (0.003) 0.0050 (0.004) 0.0228 (0.009)

0.0189 (0.004) 0.0096 (0.004) 0.0025 (0.002) 0.0056 (0.003) 0.0117 (0.001) 0.0056 (0.003) 0.0026 (0.007) 0.0023 (0.010) 0.0349 (0.007) 0.0942 (0.007) 0.1270 (0.010) 0.0248 (0.003)

0.0161 (0.003) 0.0106 (0.004) 0.0087 (0.001) 0.0060 (0.002) 0.0043 (0.001) 0.0011 (0.002) 0.0053 (0.005) 0.0044 (0.006) 0.0021 (0.004) 0.0077 (0.005) 0.0198 (0.007) 0.0012 (0.002)

0.0001 (0.002) 0.0006 (0.002) 0.0012 (0.001) 0.0006 (0.001) 0.0017 (0.000) 0.0027 (0.001) 0.0007 (0.003) 0.0000 (0.005) 0.0011 (0.003) 0.0036 (0.004) 0.0224 (0.011) 0.0040 (0.002)

0.0555 (0.004) 0.0011 (0.004) 0.0008 (0.002) 0.0158 (0.003) 0.0111 (0.001) 0.0048 (0.002) 0.0190 (0.007) 0.0171 (0.010) 0.0484 (0.007) 0.0751 (0.007) 0.1244 (0.016) 0.0335 (0.003)

0.0026 (0.005) 0.0112 (0.004)

0.0089 (0.010) 0.0049 (0.009)

0.0774 (0.010) 0.0109 (0.007)

0.0017 (0.007) 0.0072 (0.005)

FABIA´N SLONIMCZYK

80

Table 7. (Continued ) Informal Employee Treat Post Region Dummiesa Year Dummiesa Constant Observations R2

0.0427 (0.009) YES YES 0.1475 (0.042) 44,452 0.061

Informal Entrepreneur

Informal Second Job

Informal Irregular Act

0.0060 (0.006)

0.0010 (0.004)

0.0722 (0.010)

YES YES 0.1544 (0.027) 44,452 0.022

YES YES 0.0649 (0.017) 44,452 0.012

YES YES 0.2299 (0.039) 53,769 0.115

Note: RLMS, rounds VIII–XVIII (1998–2009). Definitions are as in Table 2. Arellano (1987) robust standard errors in parentheses allow for heteroscedasticity and auto-correlation of arbitrary form. Omitted category is no educational degree. a Thirty-eight regional dummies, including Moscow and St. Petersburg, and nine year dummies were included but not reported.  po0.1, po0.05, po0.01.

The effect of the reform on informal irregular activities is estimated to be 7.2%. This is a very large effect considering that the overall share of workers in this category was just above 13% in 2000. As anticipated, the regression results also show that the effect on informal entrepreneurs and informality in the second job was neither economically nor statistically significant. I conclude that the reform did not have a strong impact on these groups. The reduction in the share of informal employment among wage and salary workers and those performing irregular activities could be due to omitted variable bias. Specifically, it could be the case that unobservable characteristics of people in the control group systematically differed from those of individuals that were treated. The panel structure of the RLMS can be used to control for individual heterogeneity by relying on withinindividual changes only. The key identifying assumption of the fixed effects model is that the effect of unobservables is constant over time. Formally, this is stated by assuming that the error term in Eq. (1) can be written as: uit ¼ ci þ eit, where ci is the constant individual heterogeneity and eit is an idiosyncratic error term with zero mean conditional on treatment, the other covariates, and the individual heterogeneity. As is well known, the price to be paid for the robustness of the fixed effects estimator is that none of the parameters of the time-constant regressors are identified.

81

The Effect of Taxation on Informal Employment

Table 8 presents the fixed effects estimation results for Eq. (1). The effect on informal employees is now estimated as 2.5%, while the effect on informal irregular activities is 4.0%. Both results are still statistically significant. Attenuation in the absolute size of fixed effects estimates occurs

Table 8.

The Effect of Tax Reform on Informality: DID FE. Informal Employee

Household characteristics Number of members Number of female members Number of youth, 18 Number of elderly, 65 þ Individual characteristics Age Age2/100 Experience Experience2/100 Secondary schooling completed Vocat schooling completed Tech schooling completed College comp Graduate level comp Married DID estimates Post

Informal Irregular Activities

Any Informal Employment

0.0010 (0.003) 0.0040 (0.005) 0.0003 (0.004) 0.0100 (0.006)

0.0088 (0.003) 0.0083 (0.005) 0.0112 (0.004) 0.0005 (0.006)

0.0121 (0.004) 0.0095 (0.007) 0.0105 (0.005) 0.0011 (0.008)

0.0091 (0.010) 0.0130 (0.004) 0.0025 (0.002) 0.0008 (0.003) 0.0053 (0.010) 0.0113 (0.011) 0.0132 (0.008) 0.0276 (0.011) 0.0321 (0.019) 0.0086 (0.004)

0.0135 (0.008) 0.0173 (0.004) 0.0048 (0.002) 0.0006 (0.003) 0.0066 (0.009) 0.0075 (0.010) 0.0214 (0.007) 0.0394 (0.011) 0.0704 (0.025) 0.0098 (0.004)

0.0062 (0.012) 0.0213 (0.005) 0.0061 (0.002) 0.0013 (0.004) 0.0037 (0.011) 0.0029 (0.013) 0.0174 (0.010) 0.0506 (0.015) 0.0649 (0.034) 0.0137 (0.005)

0.0495 (0.099)

0.0350 (0.075)

0.0315 (0.119)

FABIA´N SLONIMCZYK

82

Table 8. (Continued )

Treat Post Year dummiesa Constant Observations No. of individuals R2 overall

Informal Employee

Informal Irregular Activities

Any Informal Employment

0.0250 (0.010) YES 0.2799 (0.306) 44,452

0.0403 (0.010) YES 0.4481 (0.232) 53,769

0.0584 (0.014) YES 0.2996 (0.365) 47,718

11,263 0.04

12,411 0.03

11,969 0.01

Note: RLMS, rounds VIII–XVIII (1998–2009). Any informal employment includes informality at the main job, the second job or irregular activities. Other definitions are as in Table 2. Arellano (1987) robust standard errors in parentheses allow for heteroscedasticity and autocorrelation of arbitrary form. Omitted category is no educational degree. a Nine year dummies were included but not reported.  po0.1, po0.05, po0.01.

frequently, since within-individual variation is relatively more sensitive to measurement error (Griliches & Hausman, 1986). I interpret these results as indication that, while unobservable ability bias might be a factor influencing the OLS estimates, the tax reform caused a significant reduction in informality levels. Rather than reflecting a real reduction in overall informality, the results in this section could be illusory if the tax reform pushed individuals from one form of informal employment into others. To check against this possibility, I estimate the same equation for an index of overall informality. The estimates in the third column of Table 8 suggest that, if anything, the results for the detailed informality categories are conservative.

Robustness Checks In Table 9, I present estimates of the tax reform effect under alternative specifications.32 I also provide estimates for all irregular activities (contractual or otherwise) and for informal irregular activities as the exclusive source of earnings. In order to control for changes in characteristics at the regional level – such as local tax enforcement efforts, financial markets, etc. – I add to the

83

The Effect of Taxation on Informal Employment

Table 9. Informal Employee

0.0250 (0.010) Including interactions 0.0246 District Year (0.011) 0.0256 Control group (0.010) restricteda 0.0251 Treatment group (0.011) expandedb Treatment defined 0.0363 using income from (0.014) all sources 0.0183 Treatment defined (0.012) using 2001 labor income onlyc 0.0223 Treatment defined (0.011) using 2001–2004 labor incomec 0.0063 Treat Trendd (0.003) 0.0008 Placebo Reforme (0.012)

Baseline

Robustness Checks.

Informal Irregular Activities

Any Informal All Irregular Employment Activities

Informal Irregular Activity as Main Job

0.0403 (0.010) 0.0337 (0.010) 0.0408 (0.010) 0.0339 (0.010) 0.0219 (0.011)

0.0584 (0.014) 0.0467 (0.015) 0.0588 (0.014) 0.0518 (0.015) 0.0708 (0.019)

0.0421 (0.010) 0.0373 (0.011) 0.0427 (0.011) 0.0347 (0.011) 0.0339 (0.013)

0.0343 (0.009) 0.0295 (0.009) 0.0350 (0.009) 0.0276 (0.009) 0.0219 (0.011)

0.0455 (0.014)

0.0637 (0.019)

0.0514 0.0365 (0.015) (0.010)

0.0421 (0.010)

0.0517 (0.015)

0.0429 0.0346 (0.011) (0.009)

0.0148 (0.003) 0.0128 (0.015)

0.0187 (0.003) 0.0251 (0.019)

0.0159 0.0137 (0.003) (0.003) 0.0055 0.0074 (0.016) (0.010)

Note: RLMS, rounds VIII–XVIII (1998–2009). Arellano (1987) robust standard errors in parentheses allow for heteroscedasticity and auto-correlation of arbitrary form. ‘‘All irregular activities’’ includes those done under contract. ‘‘Informal irregular activities as main job’’ excludes individuals with any other form of remunerated work. a Excludes from control group individuals whose earnings were never reported in post-reform period. b The treatment group includes individuals with yearly income between 3,168 and 4,800 rubles in any post-reform round. c Excludes individuals who receive treatment in the late post-reform years (see main text for details). d Includes a post-reform time trend (2000 ¼ 1) instead of the post-reform dummy. e The placebo ‘‘reform’’ estimates are obtained by assuming that a similar change in the tax code happened between the years 1998 and 2000 (it did not). All other covariates are the same as in Table 8. po0.1, po0.05, po0.01.

84

FABIA´N SLONIMCZYK

equation interactions between the 39 districts and the year dummies. Including these additional controls does not affect the results significantly. I also try a number of modifications in the definitions of the treatment and control groups. First, I exclude from the analysis individuals whose labor income is never reported in the post-reform period, which under my baseline definition fell in the control group. Second, I include in the treatment group individuals whose yearly income was between 3,168 and 4,800 rubles in any post-reform round. These small modifications in the implementation of the DID strategy do not lead to any significant change in the estimates. Third, I define treatment based on an alternative income item in the adult questionnaire. This alternative includes income from all sources – some of them nontaxable – and is therefore not entirely appropriate to use for the determination of the treated.33 Nevertheless, it is reassuring to verify that the main results hold with this alternative definition. The table also presents estimates when treatment is defined based on income received during the early post-reform years only. As shown in Fig. 3 above, real wages were increasing over the post-reform period. As a result, in my baseline definition many individuals enter the treatment group late. It is possible that these ‘‘late comers’’ also had a higher propensity to become formal and are therefore driving the results. To guard against this possibility, I consider two re-definitions of treatment: (1) considering income from 2001 only and (2) considering income during the subperiod 2001–2004 only. In each case I exclude from the analysis all individuals who receive treatment (i.e., higher incomes) after the relevant subperiod. Overall, the results from these sensitivity tests are satisfactory. The estimates of the effect of the reform on informal irregular activities are larger in absolute value than in the baseline case and remain highly statistically significant. The estimates for informal employees are somewhat smaller and become statistically insignificant when treatment is defined based on 2001 income. I conclude ‘‘late comers’’ are not driving the main results. I further investigate the robustness to different definitions of treatment in the next subsection. Another robustness test involves obtaining an estimate of the effect of the reform on the time trend of informality in the post-reform period. This alternative specification implies a much larger overall effect. For example, by 2009 the reform is predicted to have reduced informal irregular activities by 1.5 8 ¼ 12%. In all these exercises, treatment is defined based on an individual’s income bracket. This raises the concern that what is driving the results is

The Effect of Taxation on Informal Employment

85

the fact that individuals in the control group are worse-off and therefore less likely to become formal. The final set of estimates in Table 9 corresponds to a placebo regression. I (wrongly) assume that a similar tax reform happened sometime between the years 1998 and 2000. The new ‘‘treatment’’ variable equals 1 if the individual is in the upper income brackets (W50,000 rubles) in the year 2000. If it were true that low-income individuals are less likely to become formal, we should find that the placebo reform had a negative and significant ‘‘effect’’ on informality. However, none of the estimates of the placebo reform are significantly different from zero and most have the wrong sign. It is possible to conclude that, conditional on the covariates, individuals in the lower tax brackets were not really less likely to become formal than upper-bracket individuals. It is still possible that things changed and lower-bracket individuals were less liable to become formal in the post-real-reform years. Alas, this is not testable.

Estimating Average Treatment Effect on the Treated (ATT) Using a Matching Estimator The fixed effects DID estimates seem to be robust to minor changes in specification. However, there are a number of assumptions underlying the estimating equation that are hard to relax within this parametric setting. First, the dependent variable in Eq. (1) is binary. While OLS consistently estimates the parameter of interest under the stated identifying assumptions, it has the undesirable property that the implied conditional probability of the dependent variable is linear. Second, the set of controls are assumed to enter the equation additively and under a specific (maybe incorrect) functional form. Finally, the fixed effects estimates do not restrict estimation to the region of common support of the independent variables between the treated and the control group. In this subsection I use the matching differences-in-differences (M-DID) estimator first introduced in Heckman et al. (1997) to estimate the effect of the reform year-by-year. This semi-parametric estimator has no implications regarding the specific form of the conditional probability function. It also allows me to check whether results are robust to changes in the functional form of the control function, as well as to restricting the estimation to the region of common support.34 I also use the estimator to investigate the sensitivity of the results to alternative definitions of the treatment group. The M-DID estimator is given by

FABIA´N SLONIMCZYK

86

b DID M

" # X 1 X ¼ ðINF i;t  INF i;2000 Þ  Wði; jÞðINF j;t  INF j;2000 Þ (2) N T;t i2T j2C

0

−0.05

−0.10

−0.15

−0.20

−0.25

where T and C are the sets of indexes for treated and control individuals respectively, and NT,t is the number of observed treated individuals in year t of the post-reform period (tA{2001,y, 2009}). Intuitively, the M-DID estimator compares changes in informality status between year t and the (pre-reform) year 2000 for each treated individual to similar changes for a set of appropriate control individuals. Which individuals are selected as controls for each treated individual depends on the weighting function W(i, j). The estimates presented here use nearestneighbor matching based on the propensity score.35 Figs. 6 and 7 presents M-DID estimates (and one-standard-deviation confidence intervals) for informal irregular activities and informal employees respectively. For each year in the post-reform period, the figures present estimates obtained when treatment is defined based on labor income

2001

2002

2003

2004

Treat 2001−2009

2005

2006

Treat 2001−2005

2007

2008

2009

Treat 2001 only

Fig. 6. Year-by-Year ATT for Informal Irregular Activities. Note: Reversed scale in y-axis. ATT obtained with a matching DID estimator for each post-reform year. Treatment defined based on labor income in years 2001–2009, 2001–2005, or 2001. The I-beams are one-standard-deviation confidence intervals.

87

0

−0.05

−0.10

−0.15

−0.20

−0.25

The Effect of Taxation on Informal Employment

2002

2003

2004

Treat 2001−2009

2005

2006

Treat 2001−2005

2007

2008

2009

Treat 2001 only

Fig. 7. Year-by-Year ATT for Informal Employees. Note: Reversed scale in y-axis. ATT obtained with a matching DID estimator for each post-reform year (except 2001 when no registration questions were asked to employees). Treatment defined based on labor income in years 2001–2009, 2001–2005, or 2001. The I-beams are onestandard-deviation confidence intervals.

received during the whole 2001–2009 period, when the period is reduced to 2001–2005, and when only income in 2001 is considered. Several points are worthy of note. First, the M-DID estimates of the impact of the reform using the baseline treatment definition are substantially higher in absolute value. The effect on irregular activities is estimated to have been 5.4% in 2001 and to have increased thereafter. By 2009, individuals affected by the reform were 16.6% less likely to perform informal irregular activities. For employees, the reform is estimated to have decreased informality by 5.5% in 2002. In this case, estimates do not trend upwards but neither do they wither away in time. Second, I check whether the time pattern of the effects is an artifact of the treatment definition. I estimated M-DID estimates under restricted treatment definitions. Restricting treatment to years 2001–2005 does not affect results. With the exception of a pronounced dip in the effect on irregular activities in 2008, all estimates lay within the one-standarddeviation band of estimates using the whole 2001–2009 period. Further restricting treatment definition to individuals treated in 2001 leads to a

FABIA´N SLONIMCZYK

88

uniform downward shift in the time series of estimates for informal irregular activities. Interestingly, the time pattern of effects is not affected, suggesting that the reform had significant long run effects. Summing up, the experiment in this subsection is interesting for three reasons. First, it shows that the tax reform led to a reduction in informality regardless of the time period considered either for defining treatment or for measuring the effect. Second, the effect of the reform is robust to a non-parametric specification of the control function and to restricting estimation to the region of common support. Finally, the time pattern of the effect was very different for informal irregular activities and informal employees.

Detailed Treatment Groups The tax reform affected individuals with annual earnings of 50,000 rubles or more. However, the effect was heterogeneous even within this group. In particular, as Fig. 4 shows, those in relatively higher tax brackets experienced a larger reduction in marginal tax rates. It is natural to expect that the effect of the reform was stronger for them. Following this intuition, I define four detailed treatment variables that distinguish among individuals with after-tax monthly earnings in the following intervals: 3,625–7,250, 7,250–10,875, 10,875–21,750, and 21,750 þ . I refer to these variables as Treat1 through Treat4, respectively.36 Naturally some individuals fall into different brackets in different periods. I operationalize the definition so that the groups are mutually exclusive.37 I then re-specify the DID equation as follows:

INF it ¼ yt þ X it b þ Zi g þ cPostt þ

4 X

mh Treathi

h¼1

þ

4 X

ah ðTreathi Postt Þ þ uit

ð3Þ

h¼1

As above, I assume that the error term has the constant unobserved effect structure, so I estimate Eq. (3) using fixed effects. In order to save space, Table 10 only reports the coefficients of interest.38 For informal employees, the estimates follow a simple pattern. The reform had the strongest effect in the highest income bracket. The effects on

89

The Effect of Taxation on Informal Employment

Table 10.

Post Treat1 Post Treat2 Post Treat3 Post Treat4 Post Observations No. of Individuals R2 overall

Detailed Treatment Groups: DID FE. Informal Employee

Informal Irregular Activities

Any Informal Employment

0.0494 (0.099) 0.0172 (0.012) 0.0235 (0.013) 0.0267 (0.011) 0.0388 (0.014)

0.0358 (0.075) 0.0209 (0.012) 0.0601 (0.013) 0.0501 (0.012) 0.0276 (0.015)

0.0298 (0.120) 0.0310 (0.017) 0.0768 (0.018) 0.0793 (0.016) 0.0390 (0.020)

44,452 11,263 0.04

53,769 12,411 0.03

47,718 11,969 0.01

Note: RLMS, rounds VIII–XVIII (1998–2009). Treat4 are individuals with after-tax monthly earnings above 21,750 rubles in any post-reform period. Treat3 are individuals with earnings above 10,875 rubles at least once but never above 21,750. Treat2 and Treat1 are similarly defined using 7,250 and 3,625 rubles as cutoffs. The control group includes all those untreated and employed in the post-reform period. Other definitions are as in Table 2. Arellano (1987) robust standard errors in parentheses allow for heteroscedasticity and auto-correlation of arbitrary form. Covariates are the same as in Table 8. All estimated coefficients have the same sign and level of significance and are available upon request.  po0.1, po0.05, po0.1.

the other groups were still negative but smaller in absolute value. Indeed, the estimate for Treat1 is not significant at the conventional levels. The pattern for informal irregular activities is nonlinear. The effect of the reform peaked among those in Treat2 and declined thereafter. One simple explanation could be that informal irregular activities are infrequent in the highest brackets. Moreover, it could be the case that informal activities are heterogeneous and that relatively wealthy individuals only perform the most profitable among them. Thus, it would take an ever larger reduction in taxes to lure these individuals out of the informal sector. An alternative explanation is that the reductions in the PIT and the ST had different effects on this kind of informal employment. As shown in Table 6, reform-wise the difference between Treat1 and Treat2 involved a reduction in the ST of over 15%, while the difference between treatment group 2 and groups 3–4 was mostly about the PIT.

FABIA´N SLONIMCZYK

90

Weighted Differences in Differences The analysis of the detailed treatment effects suggests that the effects of the reform may be heterogeneous. It also raises the concern that the reduction in informal sector participation was endogenously determined, and not a consequence of the reform. Even though we control for observable and (to some extent) unobservable characteristics that differ across groups, it could still be the case that individuals in higher brackets are somehow different in ways we fail to take into account. Because the reduction in tax rates occurred in discontinuous jumps at different income thresholds, it would in principle be possible to analyze the effect of the reform in a regression discontinuity framework. There are, however, not enough individuals in the RLMS to apply the RD method meaningfully. An alternative approach involves weighting observations by the distance of reported earnings from the threshold of 50,000 rubles.39 Specifically, the weighted DID estimand is: n X

oi ½INF it  yt  X it b  cPostt  aðTreati Postt Þ  uit 2

(4)

i¼1

where oi is the individual weight and I omit the time-constant regressors. The weights are a decreasing function of the distance of the individual’s post-reform income from the threshold at 50,000 rubles. Specifically, given reported monthly P income Yit, the weights are calculated as ðKððY it  3;625Þ=hÞÞ=ð ni¼1 KððY it  3;625Þ=hÞÞ, where K(  ) is a Gaussian kernel and h is the optimal bandwidth.40 I interpret the resulting weighted DID estimates as robustness check in the spirit of regression discontinuity, since individuals with incomes close to the threshold are probably relatively closer in terms of all unobservable characteristics. Table 11 reports the estimation results for Eq. (4) with individual fixed effects. The point estimates are fairly close to those in Table 8. The number of observations goes down because of a number of individuals who are assigned zero weights, which is the intended effect of the strategy. As a consequence, including the weights almost doubles the standard errors of the treatment interaction term. These results confirm that the reform caused a reduction in the prevalence of informal irregular activities. The estimate for informal employees is not statistically significant at the conventional levels. This could lead to the conclusion that the reform did not affect this category of informality. However, the lower levels of statistical significance affect all other regressors.41 For example, none of the

91

The Effect of Taxation on Informal Employment

Table 11.

Post Treat Post Observations R2 overall No. of Individuals

Weighted DID with FE.

Informal Employee

Informal Irregular Activities

Any Informal Employment

0.0658 (0.121) 0.0178 (0.019)

0.0245 (0.063) 0.0329 (0.019)

0.1852 (0.141) 0.0546 (0.027)

41,930 0.005 10,180

50,914 0.03 11,220

45,134 0.001 10,856

Note: RLMS, rounds VIII–XVIII (1998–2009). Treatment effect estimated by a weighted fixedeffects regression. Included covariates are the same as in Table 8. Arellano (1987) robust standard errors in parentheses allow for heteroscedasticity and auto-correlation of arbitrary form. po0.01, po0.05, po0.01.

education dummies was statistically different from zero at the conventional levels. However, having a degree from a higher-education institution appears to lead to a significantly lower probability of informal employment according to all other estimates (see, e.g., Table 8). For this reason, I attribute the lack of statistical significance to the relative inaccuracy of the instrument and not the absence of a real effect.

The Extensive Margin So far, all estimates of the effect of the reform on informality have implicitly relied on individual transitions in and out of informal employment. However, an alternative route through which the reform might have affected informality is by changing the probability of choosing a formal job for those who were unemployed before the reform and found employment in the post-reform period. In order to estimate the effect of the reform on the extensive margin, I restrict the sample to individuals who were unemployed in the year 2000 but found employment sometime during the post-reform period. The top panel of Table 12 reports estimation results for the baseline fixed effect specification. I interpret these estimates as the predicted change induced by the reform in the probability of informal employment, other things constant, and conditional42 on finding employment in the post-reform period.

FABIA´N SLONIMCZYK

92

Table 12.

A. Baseline Post Treat Post Observations No. of individuals R2 overall

Tax Reform Effect on the Extensive Margin. Informal Employee

Informal Irregular Activities

Any Informal Employment

0.2740 (0.093) 0.0146 (0.025)

0.4429 (0.058) 0.1433 (0.023)

0.5704 (0.114) 0.1355 (0.027)

21,224 7,339 0.027

24,924 8,080 0.016

22,899 7,709 0.054

0.1467 (0.023) 0.1387 (0.023) 0.0969 (0.023) 0.0948 (0.026) 0.1362 (0.027) 0.0212 (0.004)

0.1357 (0.028) 0.1310 (0.027) 0.1001 (0.028) 0.1242 (0.031) 0.1044 (0.033) 0.0197 (0.005)

B. Robustness tests Including District Year interactions 0.0111 (0.025) 0.0121 Control group restricteda (0.025) 0.0049 Treatment group expandedb (0.025) Treatment defined using income from 0.0314 (0.029) all sources 0.0284 Treatment defined using years (0.029) 2001–2004c 0.0007 Treat Trendd (0.004)

Note: RLMS, rounds VIII–XVIII (1998–2009). Sample restricted to those unemployed just before the reform and who were employed at least once in the post-reform period. The dependent variable is set to zero in round IX. Round VIII is excluded. a Excludes from control group individuals whose earnings were nev.er reported in post-reform period. b The treatment group includes individuals with yearly income between 3,168 and 4,800 rubles in any post-reform round. c Excludes individuals who receive treatment in the late post-reform years (see main text for details). d Includes a post-reform time trend (2,000 ¼ 1) instead of the post-reform dummy. All other covariates are the same as in Table 8. Arellano (1987) robust standard errors in parentheses allow for heteroscedasticity and auto-correlation of arbitrary form. po0.01, po0.05, po0.01.

The tax reform significantly reduced the probability of informal irregular activities for new jobs. Specifically, individuals in the treated group were over 14% less likely to choose this form of informal employment relative to the control group. The estimate for informal

The Effect of Taxation on Informal Employment

93

employment in the main job is negative but not statistically significant. Finally, the effect on the overall informality indicator was similar to that on irregular activities. In the bottom panel of the table I report some robustness tests (comparable to those in Table 9). In general, the results reinforce the message that the reform had a strong effect on the extensive margin for irregular activities, while for informal employees the effect was probably not significant.

CONCLUSIONS The modern Russian economy is notorious for the high level of uncertainty regarding regulations, the pervasiveness of corruption and tax evasion, and the relative powerlessness of the State to enforce the law. The economic and social costs of these institutional failures are probably large. Unfortunately, there are negative feedback effects that make the emergence of better institutions unlikely. Russia seems to be trapped in a low-level equilibrium of high informality and poor public goods provision by the State. In this context, the tax reform of 2001 appears as a very important experiment. The reform reduced average tax rates for the Personal Income Tax and the Social Tax and made the tax structure more regressive. Because individuals in the lower income bracket were for the most part not affected, it is possible to estimate the effects of the reform using a DID approach. In this paper I study the effect of the reform on individual participation in the informal sector. I find evidence that the reform led to a reduction in the fraction of informal employees. The reform seems to have had an even stronger negative effect on the prevalence of informal irregular activities. These effects – which I estimate to be in the order of 2.5% and 4.0%, respectively – are robust to a number of different specifications and small alterations in the definition of treatment status. A semi-parametric estimator gives even larger estimates of these effects. It also shows that the reform had a long-lasting impact. I also find that, predictably, the effects of the reform were relatively stronger in the top income brackets, where the reduction in marginal tax rates was more radical. Finally, at the extensive margin, the reform made it 14% less likely that someone entering the job market in the post-reform period would perform informal irregular activities. These results are consistent with the emerging consensus that changes to the tax system are bound to lead to strong behavioral responses, although

FABIA´N SLONIMCZYK

94

not necessarily in the form of a reduced labor supply. In particular, several studies have shown that income reporting and tax avoidance are sensitive to changes in marginal tax rates. The Russian flat tax reform provides strong evidence that labor market informality should be added to the list of possible margins of adjustment for individuals.

NOTES 1. In many Latin American countries, the share of informal employment exceeds 50% of the urban labor force (Gasparini & Tornarolli, 2007). Existing estimates for Sub-Saharan Africa and Asia are even higher (Ju¨tting, Parlevliet, & Xenogiani, 2008). 2. As I explain below, these estimates should be interpreted as lower bounds. I find no evidence that the tax reform affected informal main-job-entrepreneurs or informality in the second job. See Table 8. 3. The paper also contributes to the larger literature on ‘flat tax’ reforms (e.g., Keen, Kim, & Varsano, 2008). 4. These estimates put the size of hidden economy in the order of 40% of official Russian GDP. 5. The RLMS is conducted by the Higher School of Economics and the ‘‘Demoscope’’ team in Russia, together with Carolina Population Center, University of North Carolina at Chapel Hill. 6. In fact, 8.5% of those considered employed only work doing irregular activities. 7. The Rosstat labor force survey counts any form of work, including barter, as employment. It also asks employment-related questions regarding a reference week, while the RLMS asks about activities during the last month. 8. The INFSUP was designed and financed by the Centre for Labour Market Studies at the Higher School of Economics in Moscow and the Labor Markets in Emerging and Transition Economies Research Program at IZA in Germany. I thank Vladimir Gimpelson and Hartmut Lehmann for generously making these data available. 9. In rare opportunities, the INFSUP was administered on a later date than the regular questionnaire, although always by the same interviewer. 10. There are numerous reviews of the literature on the informal sector and informal employment. See, for example, Peattie (1987), Swaminathan (1991), and Ju¨tting et al. (2008). 11. This classification is based on four items of the adult questionnaire: (1) ‘‘do you work at an enterprise or organization? We mean any organization or enterprise where more than one person works, no matter if it is private or state-owned. For example, any establishment, factory, firm, collective farm, state farm, farming industry, store, army, government service, or other organization.’’ Enterprise workers are considered entrepreneurs if they answer positively to both (2) ‘‘Are you personally an owner or co-owner of the enterprise where you work?’’ and (3) ‘‘In your opinion, are you doing entrepreneurial work at this job?’’ The distinction between entrepreneurs and employees for non-enterprise individuals is based

The Effect of Taxation on Informal Employment

95

on: (4) ‘‘At this job are you y (a) involved in an employer’s or individual labor activity or (b) work for a private individual?’’ 12. The question is: ‘‘Tell me, please: are you employed in this job officially, in other words, by labor book, labor agreement, or contract?’’ This item was not included in round X (2001). 13. Russian law does not require self-employed individuals to create a corporation or special legal entity. They are instead allowed to operate under a special and simpler registration procedure. However, the obligation to sign a written contract and register employees applies to all employers without exception. 14. The registration question for the second job is identical to that in the primary job. It was also not included in round X. 15. This item is available in every round. 16. In Table 9, I show that this distinction does not affect the main results. Similarly, one could distinguish between those whose only remunerated work are irregular activities and those for whom irregular activities are supplementary. This distinction does not affect the results either. 17. For example, Lehmann and Pignatti (2007) use a similar definition for Ukraine. For a discussion of the relative merits of alternative definitions, see Swaminathan (1991) and Portes and Schauffler (1993). 18. Because I analyze informality in each of the three possible jobs separately, some individuals are counted under more than one category. 19. Specifically, after the regular earnings item for the main job, the questionnaire asked: ‘‘what percent of that money do you think was officially registered, i.e. taxes were paid?’’ No similar question was included for the other jobs. 20. Tables with the detailed distributions of occupation and industry for informal workers can be found in the working paper version (Slonimczyk, 2011). 21. For the second job, some questions were not included in round XVIII. I then report information from the most recent round when the item was available. See the notes at the bottom of the table for data sources. 22. This information comes from round XVII (2008). 23. The informal entrepreneur category is overwhelmingly populated (95%) by self-employed individuals, so I do not present these statistics for them. However, roughly 50% of informal employees work at enterprises. 24. The median size of informal employees’ firms is only 10 workers. The median for the whole sample is 50. 25. The proportion of informal employees who had actually been on paid vacation during the previous 12 months was only 8.6%. 26. I reserve the terms ‘‘main job’’ and ‘‘secondary job’’ to refer to the adult questionnaire-based categories. 27. The most significant concern arises for individuals who, according to the adult questionnaire, performed both a second job and irregular activities. 28. This threshold is obtained as follows: 3,625 ¼ (50,000/12) (10.13). 29. The maximum possible savings are given by (4,8003,168) 0.13E212. 30. More precisely, the tax reform seems to have led to a reduction in informal employment relative to a counterfactual estimate based on the control group. It is clear from the figure that the effect of the reform was not strong enough to revert the ascending trend of informality.

96

FABIA´N SLONIMCZYK

31. This is one of the recommended approaches for DID studies (Bertrand, Duflo, & Mullainathan, 2004). 32. To save space I omit all other covariates. 33. The question is: ‘‘What is the total amount of money that you received in the last 30 days? Please include everything: wages, retirement pensions, premiums, profits, material aid, incidental earnings, and other receipts, including foreign currency, but convert the currency into rubles.’’ 34. In practice imposing common support means that the M-DID estimator only uses information on control individuals if their observable characteristics are close to those of one or more treated individuals. In contrast, regression estimates use all control group observations. 35. Specifically, I used an average of the 10 nearest neighbors. I also experimented using a kernel matching procedure but while the results were almost identical, processing times were much longer. Therefore I stick to nearest neighbor matching. The propensity score was estimated using a logit model that included all timeconstant and time-varying controls (as in the OLS regression of Table 7). Matching was done on the index rather than the probability. Estimates were obtained using Leuven and Sianesi psmatch2 module for Stata. 36. These detailed treatment groups correspond with the following tax brackets: 50,000–100,000, 100,000–150,000, 150,000–300,000, and 300,000 þ (see also Table 6). 37. See notes to Table 10 for details. 38. The full set of results is available from the author upon request. 39. This approach was first suggested in Gorodnichenko et al. (2009). See also Duncan and Sabirianova-Peter (2009). pffiffiffi 40. h ¼ 0:9ðs= 5 nÞ and s is the smaller amount between the standard deviation of reported income and the inter-quartile range. 41. For brevity complete estimation results were not included. They are available from the author upon request. 42. Duncan and Sabirianova-Peter (2009) study the effect of the reform on the extensive margin of employment in general (independently of formal status). Using the same DID methodology, they find that the expected probability of finding a job in the post-reform period was significantly higher for individuals in the treated group. The estimated effect is between 0.09 and 0.14, depending on whether they use a male or female sample. However, these estimates require extrapolating earnings for individuals not employed throughout the post-reform period in order to assign them to treatment or control. This is not necessary for investigating the effect on informality, conditional on employment.

ACKNOWLEDGMENTS I would like to thank Vladimir Gimpelson, Rostislav Kapeliushnikov, Alexander Muravyev, Tiziano Razzolini, and seminar participants at the Centre for Labour Market Studies (HSE) and the IZA/World Bank Workshop ‘‘Institutions and Informal Employment in Emerging and Transition Countries’’ for helpful comments. I am grateful to the editors,

The Effect of Taxation on Informal Employment

97

Hartmut Lehmann and Konstantinos Tatsiramos, and two anonymous referees for their help in improving the manuscript. The support from the HSE Research Program and the MacArthur Foundation Grant ‘‘Labor Market Informality in Russia: Economic and Social Perspective’’ is acknowledged.

REFERENCES Arellano, M. (1987). Practitioners’ corner: Computing robust standard errors for within-groups estimators. Oxford Bulletin of Economics and Statistics, 49(4), 431–434. Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust differences-indifferences estimates? Quarterly Journal of Economics, 119(1), 249–275. Duncan, D., & Sabirianova-Peter, K. (2009). Does labor supply respond to a flat tax? Evidence from the Russian tax reform. IZA Discussion Paper No. 4257, IZA, Bonn. Gasparini, L., & Tornarolli, L. (2007, February). Labor informality in Latin America and the Caribbean: Patterns and trends from household survey microdata. CEDLAS Working Paper. Available at http://EconPapers.repec.org/RePEc:dls:wpaper:0046 Gerxhani, K. (2004). The informal sector in developed and less developed countries: A literature survey. Public Choice, 120(3), 267–300. Gimpelson, V., & Kapeliushnikov, R. (2011). Labor market adjustment: Is Russia different? IZA Discussion Paper No. 5588, IZA, Bonn. Gorodnichenko, Y., Martinez-Vazquez, J., & Sabirianova-Peter, K. (2009). Myth and reality of flat tax reform: Micro estimates of tax evasion response and welfare effects in Russia. Journal of Political Economy, 117(3), 504–554. Griliches, Z., & Hausman, J. A. (1986). Errors in variables in panel data. Journal of Econometrics, 31(1), 93–118. Grossman, G. (1977). The ‘‘second economy’’ of the USSR. Problems of Communism, 26(5), 25–40. Guariglia, A., & Kim, B. (2006). The dynamics of moonlighting in Russia: What is happening in the Russian informal economy? Economics of Transition, 14(1), 1–45. Hart, K. (1973). Informal income opportunities and urban employment in Ghana. The Journal of Modern African Studies, 11(1), 61–89. Heckman, J., Ichimura, H., & Todd, P. (1997). Matching as an econometric evaluation estimator: Evidence from evaluating a job training programme. The Review of Economic Studies, 64(4), 605–654. Hussmanns, R. (2004). Measuring the informal economy: From employment in the informal sector to informal employment. ILO Working Paper, ILO, Geneva. International Labour Office. (1972). Employment, income and equality: A strategy for increasing productive employment in Kenya. Geneva. Ivanova, A., Keen, M., & Klemm, A. (2005). The Russian ‘‘flat tax’’ reform. Economic Policy, 20(43), 397–444. Johnson, S., Kaufmann, D., McMillan, J., & Woodruff, C. (2000). Why do firms hide? Bribes and unofficial activity after communism. Journal of Public Economics, 76, 495–520.

98

FABIA´N SLONIMCZYK

Johnson, S., Kaufmann, D., Shleifer, A., Goldman, M., & Weitzman, M. (1997). The unofficial economy in transition. Brookings Papers on Economic Activity, (2), 159–239. Johnson, S., Kaufmann, D., & Zoido-Lobaton, P. (1998). Regulatory discretion and the unofficial economy. The American Economic Review, 88(2), 387–392. Ju¨tting, J., Parlevliet, J., & Xenogiani, T. (2008). Informal employment re-loaded. IDS Bulletin, 39(2), 28–36. Keen, M., Kim, Y., & Varsano, R. (2008). The ‘flat tax(es)’: Principles and experience. International Tax and Public Finance, 15(6), 712–751. Kim, B.-Y. (2003). Informal economy activities of soviet households: Size and dynamics. Journal of Comparative Economics, 31(3), 532–551. Lacko, M. (2000). Hidden economy – an unknown quantity? Comparative analysis of hidden economies in transition countries, 1989–95. Economics of Transition, 8(1), 117–149. Lehmann, H. & Pignatti, N. (2007). Informal employment relationships and labor market segmentation in transition economies: Evidence from Ukraine. IZA Discussion Paper No. 3269, IZA, Bonn. Lehmann, H., Wadsworth, J., & Acquisti, A. (1999). Grime and punishment: Job insecurity and wage arrears in the Russian Federation. Journal of Comparative Economics, 27(4), 595–617. Leuven, E. & Sianesi, B. (2010). PSMATCH2: Stata module to perform full mahalanobis and propensity score matching (Version 4.0.4). Available at http://fmwww.bc.edu/repec/ bocode/p/psmatch2.ado McMillan, J., & Woodruff, C. (2002). The central role of entrepreneurs in transition economies. The Journal of Economic Perspectives, 16(3), 153–170. Peattie, L. (1987). An Idea in good currency and how it grew: The informal sector. World Development, 15(7), 851–860. Perry, G., Maloney, W., Arias, O., Fajnzylber, P., Mason, A., & Saavedra-Chanduvi, J. (2007). Informality: Exit and exclusion. Washington, DC: The World Bank. Portes, A., & Schauffler, R. (1993). Competing perspectives on the Latin American informal sector. Population and Development Review, 19(1), 33–60. Saez, E., Slemrod, J., & Giertz, S. (forthcoming). The elasticity of taxable income with respect to marginal tax rates: A critical review. Journal of Economic Literature. Schneider, F., & Enste, D. (2000). Shadow economies: Size, causes, and consequences. Journal of Economic Literature, 38(1), 77–114. Slonimczyk, F. (2011). The effect of taxation on informal employment: Evidence from the Russian flat tax reform. HSE Working Paper Series ‘‘Labour Markets in Transition’’, No. 5. Available at http://www.hse.ru/org/hse/wp/wp3 Swaminathan, M. (1991). Understanding the ‘‘informal sector’’: A survey. World Institute for Development Economics Research of the United Nations University.

99

The Effect of Taxation on Informal Employment

APPENDIX Table A1.

Summary Statistics by Treatment. Control

Treated

All Employed

Female Age Secondary education comp College education comp Schooling (years) Labor market experience Married Urban location Russian national Russian born Size household No. of female household members No. of youth household members No. of elderly household members

0.61 42.29 0.76 0.12 11.07 20.12 0.47 0.63 0.63 0.92 3.32 1.77 0.72 0.29

0.52 37.18 0.87 0.23 12.16 16.26 0.59 0.78 0.73 0.92 3.54 1.86 0.84 0.18

0.54 38.21 0.85 0.21 11.94 17.04 0.57 0.75 0.71 0.92 3.50 1.84 0.81 0.20

Observations No. of individuals

17,404 3,545

68,475 11,487

85,879 15,032

Note: RLMS, rounds VIII–XVIII (1998–2009). An individual is considered treated if her aftertax monthly labor income from all sources is above 3,625 rubles in any post-reform round. The control group comprises the untreated individuals who were employed.

CHAPTER 3 WHO BENEFITS FROM REDUCING THE COST OF FORMALITY? QUANTILE REGRESSION DISCONTINUITY ANALYSIS Tommaso Gabrieli, Antonio F. Galvao Jr. and Gabriel V. Montes-Rojas ABSTRACT This chapter studies the effect of increasing formality via tax reduction and simplification schemes on micro-firm performance. We develop a simple theoretical model that yields two intuitive results. First, low- and high-ability entrepreneurs are unlikely to be affected by a tax reduction and therefore, the reduction has an impact only on a segment of the microfirm population. Second, the benefits to such reduction, as measured by profits and revenues, are increasing in the entrepreneur’s ability. Then, we estimate the effect of formality on the entire conditional distribution (quantiles) of revenues using the 1996 Brazilian SIMPLES program and a rich survey of formal and informal micro-firms. The econometric approach compares eligible and non-eligible firms, born before and after SIMPLES in a local interval about the introduction of SIMPLES. We develop an estimator that combines both quantile regression and the

Informal Employment in Emerging and Transition Economies Research in Labor Economics, Volume 34, 101–133 Copyright r 2012 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1108/S0147-9121(2012)0000034006

101

102

TOMMASO GABRIELI ET AL.

regression discontinuity design. The econometric results corroborate the positive effect of formality on micro-firms’ performance and produce a clear characterization of who benefits from these programs. Keywords: Formality; micro-firms; quantile regression; regression discontinuity JEL classification: J23; L25 ‘‘Lo que pasa es que aca´ si vos queres abrir un negocio te matan a papeles, y despue´s te controlan, y los impuestos te revientan.’’ [What happens here is that when you try to open a business they kill you on paperwork (red tape), then they control you, and taxes are unbearable.] Martı´ n Caparro´s, El Interior, a book on interviews and anecdotes from the poor countryside in Argentina.

INTRODUCTION Formality is broadly defined as participation in societal and governmental institutions, such as paying taxes, being registered with the authorities, etc. (see Gerxhani, 2004; Maloney, 2004, for a survey). Firms’ inability to become formal is thought to have deleterious effects on performance. As examples, formality offers the firm access to risk pooling mechanisms that may attract more educated paid workers and engage them in a longer relationship with the firm, which in turn makes training and capital goods acquisition more profitable; formality may be a requirement for access to formal credit markets or government provided business development services or as Paula and Scheinkman (2007, 2010) have argued, for subcontracting relations with formal firms. Moreover, to the extent that formality increases the ability of micro-entrepreneurs to establish property rights over their investments and reduces the risk of being fined by government inspectors, it creates incentives for operating out of fixed locations rather than in an ambulatory fashion (see de Soto, 1989). The high costs of complying with government regulations and institutions have often been seen as largely responsible for the presence of large informal sectors in developing countries. The perceived onerous cost of formality was tackled by several Latin American governments by introducing tax reductions and simplifications. Examples of such programs are the Monotributo1 in Argentina, SARE2 in Mexico, and the SIMPLES3 in Brazil. Available evidence shows that these programs had a positive effect on

Quantile Regression Discontinuity Analysis

103

formality. See Kaplan, Piedra, and Seira (2006) for SARE; and Monteiro and Assunc- a˜o (2006) and Fajnzylber, Maloney, and Montes-Rojas (2011) for SIMPLES. We contribute to this literature by answering three questions: First, what is the effect of formality on firm performance? Second, which firms benefit from tax reduction and simplification schemes? Third, is there heterogeneity on the effect of formality on firm performance? These questions have very important policy implications. In a Ricardian setting, tax reductions imply a redistribution of wealth, and therefore, it is important to quantify which firms are really benefiting from these programs. In particular, if tax reductions only benefit already well-off formal firms, then the program did not accomplish the task of broadening the scope of formality. We focus on the micro-firm sector, defined as own-account workers and firms with a maximum of five paid employees, that constitutes the majority of firms in developing countries.4 Within this sector three groups can be distinguished. First, high-ability entrepreneurs with substantial growth prospects may have self-selected into formality with the old (high) tax system, as the perceived benefits of being formal offset the cost of formality. Then, this segment benefits only from the tax reduction. Second, some micro-entrepreneurs are in the informal sector as a subsistence strategy as predicted in the Harris and Todaro (1970) dual labor market hypothesis (see Maloney, 1999, 2004; Mandelman & MontesRojas, 2009, for a discussion). These are low-ability entrepreneurs and they will not value future gains from becoming formal and, therefore, tax reductions will not affect them. Third, in between those segments there are micro-firms that may become formal only when the cost of formality is low enough. These micro-firms receive the gains from being formal but have to pay taxes as a result. We call this segment the target group and it corresponds to medium-ability entrepreneurs. These are the firms that should benefit from the tax reduction programs and change their formality status. We begin our analysis by developing a theoretical model motivated by the work of Rauch (1991) and Paula and Scheinkman (2007, 2010), with emphasis on the effect of a reduction in taxes. This model yields two intuitive results. First, low- and high-ability entrepreneurs are unlikely to be affected by a tax reduction policy reform and therefore, the reform has an impact only on a segment of the micro-firm population, defined by default as medium-ability entrepreneurs. Second, the benefits of such reform, as measured by profits and revenues, are increasing in the entrepreneur’s ability.

104

TOMMASO GABRIELI ET AL.

Empirically, our goal is to quantify the impact of formality on the conditional distribution (quantiles) of micro-firm’s revenues, and the size of the target group (i.e., which firms benefit from the tax reduction). Two problems arise in our empirical setup. First, formality is endogenous, and in particular, correlated with the unobserved entrepreneurial ability. Second, we might not be able to identify the effect of formality for all firms. To solve the first problem, the identification strategy makes use of the SIMPLES program in Brazil, that offers an exogenous change in legislation that can be used to control for self-selection and endogeneity. Thus, our chapter builds on the work of Monteiro and Assunc- a˜o (2006) and Fajnzylber et al. (2011) by analyzing the SIMPLES program in Brazil that offers an exogenous change in legislation that can be used to control for selfselection and endogeneity. We use the same unique dataset for micro-firms, the ECINF 1997. Moreover, following those authors we use a difference-indifferences approach with the age of the firm and with ineligible firms as a control group to identify the effect of formality on firm performance. Monteiro and Assunc- a˜o (2006) study the effect of SIMPLES on having a government issued license, which constitutes a necessary requirement for further formalization (such as paying taxes of social security), and they find an increase in formal licensing among retail firms of 13 percentage points, but no effect on eligible firms from other sectors (construction, manufacturing, transportation, and other services). In addition, using SIMPLES as an instrumental variable (IV) for formality, they show that the latter significantly increases access to credit, and alters the amount and composition of investment toward larger and longer-term projects. Fajnzylber et al. (2011) show that SIMPLES has only a local effect on licensing rates for firms born just after the introduction of the program. Using a regression discontinuity design (see Hahn, Todd, & van der Klaauw, 2001; van der Klaauw, 2002, for a discussion about regression discontinuity estimators), with weights given by time-in-business and its distance to the introduction of SIMPLES, they find a significant effect on licensing, tax registration, tax payments, and social security contributions. When more firms were taken into consideration, the statistical significance of these effects decreases monotonically with the sample average time–distance to the introduction of SIMPLES. We build on their analysis and extend it to a quantile regression (QR) discontinuity analysis. In order to address estimation of the distributional effects of formality, we make use of the heterogeneity in the conditional distribution of revenue applying QR techniques, which will prove an indispensable tool for the problem in question. QR methods offer the advantage of describing not only

Quantile Regression Discontinuity Analysis

105

averages of possible outcomes but also their entire distribution. Thus, QR techniques provide a systematic method to analyze differences in covariates effects (see Koenker, 2005; Koenker & Hallock, 2001), a framework for robust estimation and inference, and most importantly allow exploring a range of conditional quantiles exposing conditional heterogeneity. For the present problem, the micro-firm heterogeneity given by unobserved characteristics (entrepreneurial ability) can be analyzed along the single dimensional conditional quantiles of the firm revenues. Along this dimension, high quantiles correspond to high-ability entrepreneurs and low quantiles to low-ability entrepreneurs. Chesher (2005) studies identification under discrete variation and shows that the identifying intervals can be estimated using QR methods. Thus, as argued in Chesher (2005), the identification through QR strategy may work for some quantiles (in our case target entrepreneurs) but not for others (in our case the low- and highability entrepreneurs). We face a similar situation where the SIMPLES program can be used for identification only for medium-ability entrepreneurs but not for low- and high-ability ones. Our proposed estimation strategy thus combines the regression discontinuity approach and the QR framework. In this chapter, we employ the linear instrumental variables quantile regression (IVQR) estimator proposed by Chernozhukov and Hansen (2006, 2008) applied to estimate a fuzzy regression discontinuity design model. The model is semiparametric in the sense that the functional form of the conditional distribution of the response variable given the regressors is left unspecified. The use of IVQR in a regression discontinuity design appeared in Guiteras (2008) motivated by an empirical application to the returns to compulsory schooling, and PeredaFernandez (2010) estimating the effects of class size on scholastic achievement. Frolich and Melly (2008) propose a nonparametric identification of the quantile treatment effects in the regression discontinuity design and they propose an uniformly consistent estimator for the potential outcome distributions and for the function-valued effects of the policy. Frandsen (2008) introduces a procedure to nonparametrically estimate local quantile treatment effects in a regression discontinuity design with binary treatment. The rest of the chapter is organized as follows. The second section develops a theoretical model. The third section describes the ECINF microfirm survey. The fourth section describes the SIMPLES program and the identification strategy. The fifth section develops the QR discontinuity estimator. The sixth section presents the econometric results. The seventh section concludes.

106

TOMMASO GABRIELI ET AL.

TAXES AND THE INFORMAL SECTOR In this section, we present a simple model that generates a segmentation characterized by salaried workers, informal and formal micro-entrepreneurs. The model shows that an individual becomes an informal entrepreneur, rather than being a salaried worker, if her individual ability is higher than a certain threshold, and becomes a formal entrepreneur, rather than being an informal one, if her individual ability is higher than an even higher threshold. The higher is the cost of formality the higher is the threshold value of ability to become a formal entrepreneur. This simple model builds on the models of Rauch (1991) and Paula and Scheinkman (2007, 2010). The model will then be used to analyze the impact of SIMPLES on formality. We consider a continuum of agents, each denoted by i and characterized by entrepreneurial ability yi, which is distributed according to a probability density function g(  ). Agents choose between working for an existing firm and earning a wage of w independent of their ability,5 thus becoming a salaried worker, operating a firm in the informal sector or operating a firm in the formal sector. The last two options correspond to the entrepreneurial sector. An entrepreneur produces quantity yi of an homogeneous good using capital ki and labor li as inputs. In order to maintain tractability we consider a Cobb–Douglas technology yi ¼ yi kai l bi , with a,bW0 and a þ bo1.6 We normalize the price of the homogeneous good to 1. The unit costs of k and l are respectively r and w, where r and w are given. We distinguish between formal and informal entrepreneurs. A formal entrepreneur pays an ad valorem tax f. An informal entrepreneur cheats the system and pays no taxes, but if detected is out of business. We assume that the probability of detection p increases with the size of the firm and that p(k) ¼ 0 if krk and p(k) ¼ 1 if kWk, that is, an informal entrepreneur cannot employ more than k but is able to evade taxes.7 The profit functions for an entrepreneur of ability yi who chooses to be respectively informal or formal follow: pIi ¼ max fyi kai l bi  rki  wl i g l i ;ki k

pFi

¼ maxfð1  fÞyi kai l bi  rki  wl i g l i ;ki

ð1Þ

The maximization of Eq. (1) gives the optimal quantity of production factors which are respectively used by an informal and a formal entrepreneur, given her ability yi:

Quantile Regression Discontinuity Analysis

107

(

kIi l Ii kFi l Fi

) að1bÞ=ð1abÞ  b b=ð1abÞ ¼ min ; k ; r w (  ð1aÞ=ð1abÞ  i a 1=ð1bÞ )   by k 1=ð1abÞ a a=ð1abÞ b ¼ min yi ; , w r w  ð1bÞ=ð1abÞ  b b=ð1abÞ 1=ð1abÞ a ¼ ðð1  fÞyi Þ ; r w aa=ð1abÞ  b ð1aÞ=1abÞ ¼ ðð1  fÞyi Þ1=ð1abÞ r w 1=ð1abÞ yi

When is it optimal for an entrepreneur to become formal? In choosing whether to become formal or not micro-entrepreneurs trade-off the gains of employing more than k with the cost of paying the tax f. On one hand, formality decreases productivity as it decreases the marginal products of the factors of production and such effect shows that informality can work as a device to enhance flexibility and productivity. On the other hand, formality allows firms to grow bigger as it increases the production set. It is the extent of the trade-off between the two effects that determines which entrepreneurs find it optimal to become formal rather than remaining informal. As shown by Paula and Scheinkman (2007), the convexity of the profit functions Eq. (1) in y implies that there is a unique threshold level of ability above which entrepreneurs become formal. The following proposition formally establishes this result and finds an analytical expression for the threshold level of ability. The proof is given in the appendix. Proposition 1. There exists a threshold level of ability y such that an entrepreneur i will decide to be formal if and only if her ability yi is greater  y increases in f. than y. This result is driven by the fact that productivity increases in yi and therefore agents with higher yi can afford to trade-off a decrease (measured by f) in the marginal product of factors for an increase of the production set.8 Define an ability threshold y^ such the individual with ability y^ is indifferent between becoming a salaried worker or an informal entrepreneur, ^ Plugging the first-order conditions into Eq. (1) we find hence w ¼ pI ðyÞ. that y^ ¼ ð1  a  bÞðaþb1Þ ðr=aÞa ð1=bÞb w1a . Therefore, we have that: ^ then i is a salaried worker; if yi y, ^ y,  then i is an informal entrepreneur; if yi 2 ðy;  if yi 4y, then i is a formal entrepreneur.

108

TOMMASO GABRIELI ET AL.

Effect of a Policy Change If the salaried wage is fixed,9 the fact that y increases in f (Proposition 1) implies the following corollary. Corollary 1. The greater the tax f, the greater the cut-off level of ability y and the smaller the formal sector (and vice versa). It is interesting to note that those who gain the most out of a reduction in the cost of formalization from f to fu are the more able individuals. The following proposition shows this result, and proof is relegated to the appendix. As we will remark, this result is due to the convexity of the technology. Proposition 2. The greater the individual ability yi is the greater is the increase in the profit p(yi) and revenue yi(yi) for a decrease in the tax rate from f to fu. We illustrate the results from the propositions above using diagrams. Fig. 1 illustrates the informal entrepreneurs’ profit function (thick line) and

Fig. 1.

Ad-Valorem Tax. Profit Functions: Informal (Thick Line), Formal (Thin Line), Formal After Decrease in Tax (Dashed Line).

Quantile Regression Discontinuity Analysis

Fig. 2.

109

Lump-Sum Tax. Profit Functions: Informal (Thick Line), Formal (Thin Line), Formal After Decrease in Tax (Dashed Line).

the formal entrepreneurs’ profit and revenue function before and after a reduction in the tax (thin and dashed lines). From the figure it is possible to notice the results of Propositions 1 and 2.10 Moreover, from the figure, it is also evident that the result of Proposition 2 would not apply to a different 0 model in which pF(fu) is not always convex for y4y .11 The model can be extended to the case of a lump-sum tax. In this case, the profit function of a formal entrepreneur is the following: pFi ¼ maxfyi kai l bi  rki  wl i  fg, where f now represents a lump-sum tax. In l i ;ki

such case all the previous conclusions still hold. Fig. 2 illustrates the profit function plot for this case of a lump-sum tax change.12

DATA AND DESCRIPTIVE STATISTICS We employ the Brazilian Survey of the Urban Informal Sector (Pesquisa Economia Informal Urbana, ECINF) collected in October 1997 (11 months after the introduction of the SIMPLES) by the Brazilian Statistical Institute (IBGE, Instituto Brasileiro de Geografia e Estadı´ stica). This survey is a

110

TOMMASO GABRIELI ET AL.

cross-section representative of all the urban self-employed and micro-firm owners with at most five paid employees, excluding domestic workers. The stratified sampling design (in two stages) allows studying a population of units which are rare, heterogeneous and hard to detect in standard household surveys. Geographically, it covers all of the 26 Brazilian states, as well as the federal district, and also each of the 10 metropolitan areas (Bele´m, Fortaleza, Recife, Salvador, Belo Horizonte, Vito´ria, Rio de Janeiro, Sa˜o Paulo, Curitiba, and Porto Alegre) and the municipality of Goiaˆnia. In each of its two waves, ECINF interviewed roughly 50,000 households among which it found more than 40,000 individuals which reported owning a micro-enterprise. We analyze firms with a government issued license as our measure of formality. Only 23.2% of all micro-firms have a license which increases to 31.1% for micro-firms with at least one paid employee. Within the Brazilian micro-entrepreneur sector, the most frequent sectors of activity are retail trade (26% of micro-firms) and personal services (20%), followed by construction (15%), technical and professional services (11%) and manufacturing (11%). Respectively 8% and 7% of micro-firms belong to the sectors of hotels and restaurants, and transportation. Most firms are very small both in terms of revenues and employment: the average and median monthly revenues of Brazilian micro-firms were $US 1,083 and $US 600, respectively. We find that 87% of all Brazilian micro-firms have no paid employees, and 79% have no employees or partners at all, 10% of the surveyed micro-firms have one or two paid employees, and only 3% have between three and five paid workers. In those firms with at least one paid employee, roughly 22% of all workers are family members, almost twothirds of paid workers are non-registered (sem carteira assinada) and only 35% pay social security contributions. The ECINF asks whether respondents started their firms themselves or became owners at a later date. The survey then collects data on the number of years and months since respondents respectively started the firm or became owners-partners. We use this information to construct our time-inbusiness variable. For firms that were not started by their current owners, our time-in-business variable reflects the time since the current owner joined in as a partner, which is not necessarily the actual age of the firm. This problem, however, affects only 8% of firms (92% of respondents report having started their own firms) and it does not appear to have a significant impact on our main conclusions. Given that the IV strategy relies heavily on the validity of this measure we will also consider separately the subsample of micro-firms where the firm was started by the current owner.

Quantile Regression Discontinuity Analysis

111

THE SIMPLES PROGRAM AND IDENTIFICATION STRATEGY In November 1996, the Brazilian government implemented a new unanticipated simplified tax system for micro-firms and small firms, the SIMPLES. The new national system consolidated several federal taxes and social security contributions. Basically, the SIMPLES abridged procedures for the verification and payment of federal, state, and municipal taxes. At the federal level, the system allowed eligible firms to combine six different types of federal taxes and five different social security contributions into a one single monthly payment, varying from 3% to 5% of gross revenues for micro-enterprises, and from 5.4% to 7% of revenues for small firms. One important aspect of the new system is that it allowed substituting a fixed (and relatively low) percentage of total invoicing for the standard payroll contribution, which led to a substantial reduction in labor costs and hence created a strong incentive to hire new employees and/or legalize already existing labor relationships. The motivation behind these reductions in direct and indirect taxes was to enable small, unskilled labor-intensive firms to compete more effectively with larger enterprises, for which high tax burdens are more manageable due to scale economies. Moreover, while value added taxes collected at the state and municipal levels – the Imposto Sobre Circulac- a˜o de Mercadorias e Prestac- a˜o de Servic- os (ICMS) and the Imposto Sobre Servic- os (ISS) – were initially not included in SIMPLES, states and municipalities could enter into agreements with the federal government to transfer to the latter the collection of the corresponding taxes through an increase in the SIMPLES rates. As a result, SIMPLES permitted an overall reduction of up to eight percentage points in the tax burden faced by eligible firms MonteiroAssuncao06. SIMPLES, however, explicitly excluded from program eligibility all activities that by law require the employment of professionals in regulated occupations. Examples of ineligible activities include the manufacturing of chemical products, machinery and equipment, as well as education, health, accounting, insurance and financial services, among others.13 Given the previous model, firms’ output or revenues yi ¼ yi kai l bi can be re-expressed as a function of formality (which can be thought of as an indicator variable with 0 and 1 and labeled with d), and entrepreneurial ability yi: yi ¼ f ðd i ; yi Þ

112

TOMMASO GABRIELI ET AL.

As the previous section showed formality affects output through the quantity of capital as formal entrepreneurs can employ a quantity kiWk. Net of the effect of costs of formality f, an entrepreneur i would employ kiWk if and only if yiWy. Therefore, f ð1; yi Þ  f ð0; yi Þ40, yiWy (return to formality) and @f ð; Þ=@yi ¼ kai l bi 40 (return to ability).  and firms with As we have shown, there exists a cut-off value of ability, y, ability above that threshold will select into formality. SIMPLES can be conceived of as a reduction in the cost of formalization to f0 of (albeit across many margins: registration costs, labor costs, etc.) that will change 0 the cut-off value of ability from y to y (Corollary 1). Firms that change their 0  This also formality status because of SIMPLES are those with y 2 ðy ; y. implies that there will be a subset of firms who will not change their formality status: some will remain formal (best entrepreneurs), others will remain informal (worst entrepreneurs). The introduction of SIMPLES by unanticipated administrative decree can be seen as an exogenous policy change that significantly altered the incentives to become formal and hence is useful to solve the endogeneity problem. The theoretical model developed above predicts that only for a segment of firms we will be able to identify the effect of formality. The reason is that we will only observe an effect of SIMPLES on those firms 0  This is the group of firms that have a large enough y such with y 2 ðy ; y. that the SIMPLES tax reduction makes them to re-evaluate their formality status, but not so large as to make the change in f irrelevant to their formality decision. This segment contains firms that will become formal only after the reduction in taxes, and therefore we can identify b1 by using the regression discontinuity approach described above. Note that this does 0  formality has no effect on the not mean that for firms with yoy or yoy firm performance variable. Rather it means that we cannot identify the effect of formality for those firms. Monteiro and Assunc- a˜o (2006) argue that for relatively young firms (i.e., less than two years old) the time when the firm was started clearly differentiates firms that benefit from SIMPLES from those that did not. Although all firms could benefit from SIMPLES, firms born after SIMPLES show a much higher propensity to have a license than those born before. Overall this suggests a dual process for formalization: first, a firm’s decision to formalize is primarily taken at the time of its creation; second, the likelihood of becoming formal increases with time-in-business.14 The ECINF provides some evidence on this: only one out of four licensed business made no attempt at regularizing at the time of starting up compared to 4 out of 5 non-licensed business. Thus, the decision of whether

Quantile Regression Discontinuity Analysis

113

to operate formally or informally appears to be made in most cases at the time of start-up. This could be due either to costly and/or complex registration procedures, to high tax rates, or to a limited demand among very small businesses for the government services or the expanded access to markets that are associated with formality at any price. While the data do not allow us to distinguish among these different two possible explanations, 72% of the firms that do attempt to register report having no difficulties in the process. Monteiro and Assunc- a˜o (2006) exploit the first process, that is, the differential effect on licensing caused by the introduction of SIMPLES for firms born before and after it. Let AFTER be an indicator for whether a firm was created before or after the SIMPLES was implemented (such that AFTERi ¼ 1 if ti t and AFTERi ¼ 0 otherwise, where firms that have been in business for at most t months were created after SIMPLES) and ELIG an indicator for the eligibility status of the firm. Monteiro and Assunc- a˜o (2006) uses the interaction of eligible/non-eligible and before/after indicators, that is, AFTER ELIG as an IV difference-in-differences to measure the impact of formality on firm performance. Fig. 3 plots licensing rates for firms with different dates of creation (see section ‘‘Data and Descriptive Statistics’’ for a description of the database of micro-firms used). The first two graphs plot separately eligibles and noneligibles for all firms; the last two take only the sample of entrepreneurs that started as owners of the firm. The figures show that there is a significant jump in licensing rates for eligible firms, but no change for non-eligible firms. Moreover, the jump is observed only for firms born about the time of the introduction of SIMPLES. Then, as argued in Fajnzylber et al. (2011), the validity of AFTER ELIG as an IV for formality crucially depends on comparing firms that were born just after and before than t, that is, jti  tjo for e small enough. The regression discontinuity literature (see Hahn et al., 2001; van der Klaauw, 2002) argues that an unbiased estimate of the treatment impact can be obtained by giving heavier weights to observations arbitrarily close to a discontinuity. If, conditional on a set of exogenous covariates, we assume very similar distributions of unobservable characteristics of firms born immediately before and after SIMPLES implementation, the discontinuity that the introduction of SIMPLES produces in the factors determining formality can be exploited to provide unbiased estimates of the local average treatment effect of the program. Using this argument, Fajnzylber et al. (2011) show that the regression coefficient of AFTER ELIG is dependent on the weighting scheme. Following these authors we will implement a fuzzy

114

TOMMASO GABRIELI ET AL.

Fig. 3. Average Licensing Rates by Month of Firm Creation. Note: Average licensing rates by reported month of firm creation. Owners: Original owners of the micro-firm.

regression discontinuity design, where on a small enough interval about the introduction of SIMPLES, identification can be achieved by comparing firms born just before and just after the SIMPLES introduction. The validity of the estimates of the effect of formality on revenues relies on the validity of SIMPLES as an IV. In particular, if self-selection into treatment occurred this would produce biased estimates, and the direction of the bias would depend on the correlation between those that benefit from SIMPLES treatment and unobservables. The first concern is that some firms might have strategically delayed their creation after the introduction of SIMPLES, thus changing the composition of firms before and after. Monteiro and Assunc- a˜o (2006) show that SIMPLES did not produce any change in the number of starting firms as compared to similar months before (i.e., SIMPLES produced no ‘‘rush’’ to start a firm) and it only affected formality of eligible firms. Moreover, Fajnzylber et al. (2011) compare firms

Quantile Regression Discontinuity Analysis

115

born before and after together with eligible and non-eligible firms on several observable characteristics (education, age, gender, location) and find that there are no statistically significant differences. While this still does not rule out differences in unobservables, these characteristics are likely to be correlated with unobservables, and therefore they provide indirect evidence for the validity of SIMPLES as an IV. Finally, Monteiro and Assunc- a˜o (2006) show that the SIMPLES effect is not due to seasonal effects (they repeat their analysis one and two years later as if SIMPLES had been introduced in November 1995 and 1994, respectively, and they found no effect) which shows that there are no intrinsic differences between firms born before and after about the November cut-off in other years. The second concern is that SIMPLES might have changed the composition of eligible and non-eligible firms.15 First, low-skilled entrepreneurs might be pushed out by new entrants and then excluded from the survey (which is a retrospective survey, taken one year after SIMPLES, see section ‘‘Data and Descriptive Statistics’’). Although we cannot control for potential attrition bias and sample selection, sectoral transition studies (see Fajnzylber, Maloney, & Montes-Rojas, 2006; Maloney, 1999) suggests that micro-entrepreneurs will remain within the micro-firm sector and will not become salaried workers or unemployed, hence that the microentrepreneurs sector will not change its overall composition. Second, firms might have strategically changed their the industry or sector to become eligible. However, because the definition of non-eligibility mostly applies to regulated and professional occupations, for an entrepreneur to change from the non-eligible to the eligible sector would require a substantial change in the goods or services offered, a possibility which seems unlikely in the short run.16 To summarize, our identification strategy allow us to estimate the effect of 0  and born near the formality on firm performance for firms with y 2 ðy ; y introduction of SIMPLES, that is, jti  tjo for e small enough. This strategy requires the use of both QR (to model y) and regression discontinuity designs (to amplify the effect of SIMPLES at the time of its introduction).

QUANTILE REGRESSION DISCONTINUITY 0 In order to find the threshold values y and y we will consider the single dimensional conditional quantiles, indexed by t 2 ð0; 1Þ, of the firm’s revenues, y,

116

TOMMASO GABRIELI ET AL.

Qy ðtjd; x; jti  tjoÞ ¼ b1 ðtÞd i þ b2 ðtÞti þ b3 ðtÞxi ;

(2)

where i denotes the firm, d is a binary formality indicator (licensing), t denotes time-in-business and x is a set of exogenous covariates. If we assume that for all y1 y2 there exists 0ot1 t2 o1, then this conditional quantile 0  respectively. With function can be used to find t 0 and t that match y and y, the proposed identification we can estimate b1(t) for 0ot0 ot t o1. This case was discussed by Chesher (2003) where he argued about ‘‘the possibility of identification of a structural derivative evaluated at some quantile probabilities but not at others’’(p. 1411). It should be emphasized that b1(t) measures the difference in revenues due to the effect of licensing (i.e., being formal) and that the conditioning on a small interval about the introduction of SIMPLES, that is, jti  tjo, does not imply this effect occurred in a given interval in time. These differences are the result of potentially multiple simultaneous effects, such as hiring more labor, capital, access to credit, operating in a fixed location, etc.17 We only focus on the quantile heterogeneity in total revenues. As argued in the previous section we use z ¼ (AFTER ELIG) as a valid instrument for d. This identification condition is discussed in Monteiro and Assunc- a˜o (2006) and Fajnzylber et al. (2011). The IVQR estimation method may be viewed as an appropriate QR analog of the two-stage least squares (2SLS) that makes use of a valid exclusion restriction. More formally, and following Chernozhukov and Hansen (2006, 2008), from the availability of an IV, z, we consider estimators defined as: b^ 1 ðtÞ ¼ arg minb1 k^gðb1 ; tÞkA ;

(3)

where g^ ðb1 ; tÞ is obtained from arg minb2 ;b3 ;g

N X

oðjti  tjÞrt ðyi  b1 d i  b2 ti  b3 X i  gzi Þ;

(4)

i¼1

with o(  ) a weighting function that is monotonically decreasing in jti  tj, pffiffiffiffiffiffiffiffiffiffi rt ðÞ the t-QR check function, kxkA ¼ x0 Ax and A is a positive definite matrix.18 Differently to IV least squares, however, it does not have a first stage. The asymptotic properties of the estimator are described in Chernozhukov and Hansen (2006, 2008). In particular asymptotic normality holds,

Quantile Regression Discontinuity Analysis

117

pffiffiffi ^ d nðbðtÞ  bðtÞÞ ! Nð0; JðtÞ1 SðtÞJðtÞ1 Þ; where b ¼ ðb1 ; b2 ; b3 Þ0 , JðtÞ ¼ E½f ðtÞ ð0jd; t; x; zÞðt; X; zÞðd; t; xÞ0  with ðtÞ ¼ yi  b1 d i  b2 ti  b3 xi  gzi , f ðtÞ ðÞ the density function, and SðtÞ ¼ ðminðt; t0 Þ  tt0 ÞE½ðd; t; xÞðt; x; zÞ0 . We refer the reader to Chernozhukov and Hansen (2005, 2006) for a more detailed discussion on the assumptions used for identification and the asymptotic results of the IVQR estimator. One important assumption for identification of the IVQR is rank invariance. This implies that, conditional on all other variables, a common unobserved factor, such as unobserved ability, determines the ranking in the outcome conditional distribution of a given subject across treatment states.19 In our application, a firm considers a binary formality variable, d 2 f0; 1g. The potential outcome under each level is given by the firm’s earnings under the different licensing fyd ; d ¼ 0; 1g. We assume that the potential revenue outcomes, conditional on X ¼ (x,t), are given by Eq. (2), Qyd ðUjd; x; tÞ ¼ b1 ðUÞd þ b2 ðUÞt þ b3 ðUÞx, where rank UBU(0,1) indexes the unobserved heterogeneity, U(0,1) denotes the standard Uniform distribution, and Qyd ðUjd; x; tÞ is increasing in U. Thus, the distribution of potential outcome yd is characterized by the quantile functions Qyd ðUjd; x; tÞ. The rank variable U is assumed to be determined by entrepreneurial ability and other unobserved factors that do not vary with d. Moreover, in this model, the independence condition only requires that U is independent of the instruments z, conditional on X. Finally, the rank variable U (entrepreneurial ability) is assumed invariant to d, which ascribes an important role to conditioning on covariates X. Having a rich set of covariates makes rank invariance a more plausible approximation.

ECONOMETRIC RESULTS Our main goal is the estimation of Eq. (2), that is, the conditional quantiles of the logarithm of total revenues. In order to implement this we follow the strategy described in section ‘‘The SIMPLES Program and Identification Strategy,’’ where AFTER ELIG is used as an IV for having a license.20 We increase the power of the instrument by interacting it with gender and age of the entrepreneur. Moreover, we use the same weighting scheme as in Fajnzylber et al. (2011) with oðjti  tjÞ ¼ f ð0; jti  tjÞ, where f(0,s) is the normal density of a standard Gaussian random variable with mean 0 and standard deviation s.

118

TOMMASO GABRIELI ET AL.

Our measure of firm performance y is the logarithm of total monthly revenues. Unfortunately, we cannot apply the same analysis to profits, because this would need additional instruments for both capital and labor, which are endogenous and affected by SIMPLES. Moreover, there may be measurement errors in the cost of capital and imputation of the owner’s salary. These are potentially large in micro-firms surveys. Therefore, the return to formality is the ultimate effect on revenues arising from several channels: hiring both more labor and capital, higher productivity, more business opportunities, access to credit, etc. This effect may also include changes in the composition of clients as in Paula and Scheinkman (2007) model. As additional control variables x we use the AFTER, ELIG, gender (dummy for female), age and education of the entrepreneur (the latter as categorical dummies, base category: no formal education), number of members in the household, a set of dummy variables for the reasons to become an entrepreneur, time-in-business (interacted with AFTER and as a square polynomial), and dummy variables by industry and state. Tables 1 and 2 present the 2SLS and IVQR estimates of the conditional mean and quantiles (selected quantiles) of firm revenues for the selected weighting scheme described above for all and for those entrepreneurs that started as owners, respectively. Figs. 4 and 5 summarizes the effect of licensing on firm revenues. The figures show that the effect of licensing is not statistically significant for to0.10 and tW0.60 (tW0.50 for the sample of original owners). This 0 suggest that, in terms of the characterization proposed in this chapter, y ¼ 0:10 and that therefore, 10% of the sample corresponds to the entrepreneurs that did not benefit from SIMPLES because they opted out of formality even after the tax reduction. Moreover, y ¼ 0:50ð0:60Þ, and then the upper 50% (40%) of the sample were already considering that the cost of formality was not very high. For these segments, we cannot identify the effect of formality through the introduction of SIMPLES. Taking the complement of those groups, we define the target population given by 0.10rtr0.50 or 0.10rtr0.60 depending on the sample. Note that for this group the effect is roughly similar to the 2SLS estimate. Note, however, that the point estimates being non-statistically significant does not imply that the instruments are not working and that the effect of licensing cannot be identified. In fact, this cannot be a priori be distinguished from it being statistically equal to zero. The lack of a first stage does not allow us to use the OLS techniques for evaluating the IV performance. Therefore, we propose a new procedure based on the

119

Quantile Regression Discontinuity Analysis

Table 1.

Quantile Regression Discontinuity Analysis – All MicroFirms. IV LeastSquares Regression

License Female Age

3.40 (1.04) 0.546 (0.075) 0.0039 (0.020)

IV Quantile Regression t ¼ 0.1

t ¼ 0.25

t ¼ 0.5

t ¼ 0.75

t ¼ 0.9

2.03 3.48 1.90 4.92 2.60 (1.09) (0.66) (0.49) (6.93) (4.15) 0.676 0.292 0.587 0.474 0.538 (0.162) (0.200) (0.100) (0.111) (0.120) 0.0021 0.021 0.016 0.018 0.030 (0.0059) (0.006) (0.004) (0.015) (0.14)

Education categories (base: no formal education) 0.195 Primary incomplete 0.334 (0.090) (0.253) Primary complete 0.411 0.135 (0.119) (0.388) Secondary incomplete 0.735 0.562 (0.111) (0.313) Secondary complete 0.591 0.632 (0.196) (0.306) College incomplete 0.573 0.717 (0.301) (0.492)

0.425 (0.296) 0.555 (0.329) 1.15 (0.36) 0.633 (0.351) 0.764 (0.455)

0.672 (0.136) 0.918 (0.49) 1.16 (0.16) 1.21 (0.17) 1.41 (0.47)

0.988 (0.414) 1.19 (0.47) 1.37 (0.46) 1.39 (0.58) 1.75 (0.57)

1.24 (0.16) 1.52 (0.37) 1.66 (0.21) 1.90 (0.23) 2.08 (0.50)

Reasons to become entrepreneur (base: did not find a job) 0.968 0.103 0.513 1.136 1.64 Profitable business 0.402 (0.287) (0.441) (0.614) (0.441) (0.454) (0.65) Flexible hours 0.227 0.022 0.397 0.127 0.369 0.476 (0.132) (0.338) (0.496) (0.184) (0.386) (0.445) Be independent 0.127 0.350 0.048 0.409 0.390 0.472 (0.165) (0.286) (0.268) (0.118) (0.165) (0.322) Family tradition 0.230 0.526 0.030 0.494 0.334 0.689 (0.302) (1.225) (0.354) (0.214) (0.427) (1.304) 0.171 0.023 0.029 To help family income 0.204 0.469 0.152 (0.060) (0.211) (0.203) (0.110) (0.120) (0.156) Accumulated experience 0.330 0.530 0.447 0.422 0.407 0.909 (0.151) (0.230) (0.244) (0.158) (0.519) (0.912) Make good deal 0.090 0.070 0.061 0.409 0.558 0.395 (0.136) (0.470) (0.301) (0.153) (0.211) (0.405) As a secondary job 0.558 1.013 0.886 0.380 0.968 0.768 (0.178) (0.413) (0.495) (0.338) (0.431) (0.353)

Note: 6,741 observations. Standard errors in parenthesis. Instrumental variables: AFTER

ELIG interacted with gender and age of the entrepreneur. See text for additional details. Significant at the 10% level; Significant at the 5% level; Significant at the 1% level.

120

TOMMASO GABRIELI ET AL.

Table 2.

Quantile Regression Discontinuity Analysis – Owners. IV LeastSquares Regression

License Female Age

3.23 (0.97) 0.549 (0.077) 0.0043 (0.019)

IV Quantile Regression t ¼ 0.1

t ¼ 0.25

t ¼ 0.5

t ¼ 0.75

t ¼ 0.9

4.97 3.37 1.87 5.00 2.98 (1.61) (0.73) (0.82) (7.53) (2.65) 0.034 0.317 0.577 0.482 0.421 (0.382) (0.176) (0.095) (0.112) (0.135) 0.015 0.021 0.015 0.019 0.027 (0.012) (0.006) (0.004) (0.017) (0.11)

Education categories (base: no formal education) Primary incomplete 0.294 0.364 (0.095) (0.686) Primary complete 0.391 0.058 (0.121) (0.772) Secondary incomplete 0.718 0.307 (0.111) (0.883) Secondary complete 0.553 0.014 (0.201) (1.054) College incomplete 0.647 0.487 (0.278) (1.013)

0.291 (0.258) 0.480 (0.293) 1.05 (0.30) 0.570 (0.320) 0.728 (0.512)

Reasons to become entrepreneur (base: did not find a job) Profitable business 0.222 0.201 0.106 (0.300) (0.961) (0.747)  Flexible hours 0.387 0.853 0.325 (0.140) (0.690) (0.400) Be independent 0.182 0.257 0.089 (0.146) (0.433) (0.258) Family tradition 0.172 0.618 0.189 (0.262) (1.257) (0.342) To help family income 0.224 0.104 0.208 (0.058) (0.301) (0.205) 0.017 0.393 Accumulated experience 0.323 (0.148) (0.675) (0.246) Make good deal 0.084 0.452 0.050 (0.132) (0.437) (0.298) As a secondary job 0.657 1.58 1.03 (0.194) (0.64) (0.337)

0.606 (0.164) 0.863 (0.157) 1.09 (0.17) 1.14 (0.18) 1.52 (0.45)

0.968 (0.459) 1.17 (0.51) 1.42 (0.52) 1.36 (0.64) 1.88 (0.66)

1.17 (0.20) 1.37 (0.32) 1.66 (0.24) 1.74 (0.24) 2.04 (0.37)

0.685 0.863 1.71 (0.690) (0.742) (0.36) 0.177 0.369 0.770 (0.208) (0.366) (0.478) 0.445 0.384 0.367 (0.120) (0.158) (0.226) 0.688 0.486 1.00 (0.255) (0.387) (0.496) 0.210 0.062 0.063 (0.113) (0.132) (0.174) 0.426 0.395 0.944 (0.197) (0.555) (0.592) 0.448 0.526 0.370 (0.193) (0.203) (0.193) 0.478 1.00 0.569 (0.311) (0.411) (0.228)

Note: 6,300 observations. Standard errors in parenthesis. Instrumental variables: AFTER

ELIG interacted with gender and age of the entrepreneur. See text for additional details. Significant at the 10% level; Significant at the 5% level; Significant at the 1% level.

Quantile Regression Discontinuity Analysis

121

Fig. 4. Quantile Regression, All Micro-Firms. Note: Plot for 2SLS and IVQR estimates with their corresponding 95% confidence intervals. y-axis contains the coefficient estimates and x-axis the quantiles. The dashed horizontal line is the 2SLS estimate, and the dotted lines the corresponding confidence interval. the solid line is the IVQR estimate, and the shadow its corresponding confidence interval.

Chernozhukov and Hansen (2006, 2008) estimator. If the identification strategy using the IV works well, then g^ ðb1 ; tÞ, based on Eq. (3), should have a clear global minimum. If, however, the IV is not appropriate, it should not have a clear minimum. We thus plot several graphs of ð^gðtÞ; b1 Þ for different quantiles t and analyze them. Figs. 6 and 7 report these for both samples and tA{0.10,0.25,0.50,0.75,0.90}. From the graphs it can be noted that only for tA{0.25,0.50} the function is convex almost everywhere with a clear minimum, but it is less so for the remaining quantiles. This implies that the lack of significance in b^ 1 is associated with an IV that does not satisfy the Chernozhukov and Hansen (2006, 2008) identification criterion. The 2SLS point estimate is 3.40 (SE 1.04) for all firms and 3.23 (SE 0.97) for the owners subsample. Note that the subsample of firms whose current entrepreneur was the original owner has higher standard errors. These high and rather imprecise estimates are similar in magnitude to those in Monteiro

122

TOMMASO GABRIELI ET AL.

Fig. 5. Quantile Regression, Started Firm as Owner. Note: Plot for 2SLS and IVQR estimates with their corresponding 95% confidence intervals. y-axis contains the coefficient estimates and x-axis the quantiles. The dashed horizontal line is the 2SLS estimate, and the dotted lines the corresponding confidence interval. The solid line is the IVQR estimate, and the shadow its corresponding confidence interval.

and Assunc- a˜o (2006) and Fajnzylber et al. (2011). Moreover, although not reported, similar point estimates are obtained in levels if we compute the corresponding percentage increment. As a result the large log estimates appear because of the fact that firms have in fact low levels of revenues. Overall, they clearly point out that formality (licensing) has a positive effect on firms’ revenues. In fact, these high positive effects are observed for all quantiles, although as mentioned above the effect is statistically significant only for the target population. To examine the heterogeneity associated with the IVQR estimates we perform diagnosis tests using Kolmogorov–Smirnov tests.21 First, we test the hypothesis of a zero constant coefficient for the IVQR estimates across quantiles, that is, we test the hypothesis that H 0 : b1 ðtÞ ¼ 0. In order to implement the test, we estimate the model for tA[0.1,0.9], compute the Wald statistic for each particular quantile and take the maximum over the corresponding quantiles. The results for the test statistics are 27.83 and

123

Quantile Regression Discontinuity Analysis Plot of objective function (τ = 0.10)

5

Plot of objective function (τ = 0.25)

12

4.5 10

4 3.5

8

|γ|

|γ|

3 2.5

6

2 4

1.5 1

2

0.5 0

1

1.5

2

2.5 β1

3

3.5

0

4

2.5

3

3.5 β1

4

4.5

5

Plot of objective function (τ = 0.75)

Plot of objective function (τ = 0.50)

6

2

2.4 2.2

5

2

1.8

4

|γ|

|γ|

1.6 3

1.4 1.2

2

1

0.8

1

0.6 0

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

0.4

3

2

2.5

3

3.5

β1

4

4.5

5

β1 Plot of objective function (τ = 0.90)

3

2.5

|γ|

2

1.5

1

0.5

0 1

1.5

2

2.5

3

3.5

4

β1

Fig. 6. Validity of the IV. Note: Plot of the function k^gk – all micro-firms. y-axis contains the estimates of k^gk and x-axis b1. Selected quantiles t ¼ {0.10, 0.25, 0.50, 0.75, 0.90}.

124

TOMMASO GABRIELI ET AL. Plot of objective function (τ = 0.10)

6

Plot of objective function (τ = 0.25)

10 9

5 8 4

7

|γ|

|γ|

6 3

5 2

4 3

1 2 0

1 2

2.5

3

3.5

4

4.5

5

5.5

6

6.5

7

2

2.5

3

3.5

4

Plot of objective function (τ = 0.50)

7

4.5

5

5.5

6

6.5

7

β1

β1

Plot of objective function (τ = 0.75)

4.5 4

6

3.5

5

3 |γ|

|γ|

4 2.5

3 2 2

1.5

1

1

0

0.5 0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

2

2.5

3

3.5

β1

4

4.5

5

5.5

6

β1 Plot of objective function (τ = 0.90)

4

3.5

|γ|

3

2.5

2

1.5

1 1

1.5

2

2.5

3

3.5

4

4.5

5

5.5

6

β1

Fig. 7. Validity of the IV. Note: Plot of the function k^gk – owners. y-axis contains the estimates of k^gk and x-axis b1. Selected quantiles t ¼ {0.10, 0.25, 0.50, 0.75, 0.90}.

Quantile Regression Discontinuity Analysis

125

21.74 for the all micro-firms and owners samples, respectively. These results strongly reject the null hypothesis at the 1% level of significance (the critical values are: 12.69 at 1% level of significance, 9.31 at 5% level of significance, and 7.63 at 10% level of significance). Thus, there exists strong evidence to reject the hypothesis of zero or negative impact of licensing on log revenues. Second, we test the hypothesis of a constant given effect of SIMPLES on  where we set b as the 2SLS estimate. The revenues, that is, H 0 : b1 ðtÞ ¼ b, results for the tests statistics are 9.43 and 6.53 for all micro-firm and owners samples respectively, such that we reject the null at 5% level of significance for the first case. Thus, although the confidence interval of the IVQR contains the point estimate of 2SLS, for various intermediate quantiles, the evidence suggests that the effect of SIMPLES on revenues is heterogeneous. However, in the second sample the wide confidence intervals made the 2SLS estimate to remain inside the bands and we cannot reject the null hypothesis.  only over the selected Finally, we apply the latter test, H 0 : b1 ðtÞ ¼ b, quantiles where we have evidence of identification of the parameters of interest, that is, for tA[0.10,0.60] (tA[0.10,0.50] for the sample of original owners).22 In this case, the results for the test statistics are 11.08 and 7.57 for all micro-firms and owners subsamples, respectively, such that we reject the null at 5% level of significance for the first case, and at 10% for the second case. This shows that there is heterogeneity within the target group segment. In fact, we observe that the effect is actually decreasing on t for this range. This result contradicts that in Proposition 2 and could be due to the nonconvexities described in McKenzie and Woodruff (2006), where the return to capital is higher for low-capital firms. Overall, this suggests that, over the range of identified quantiles, the formality treatment has a bigger impact on low quantiles than in high quantiles. The study of the covariate effects is of independent interest too. The negative coefficient of female reflects the fact that women engage in less profitable activities, possibly due to household commitments or outright gender discrimination.23 There is no clear pattern across quantiles, which determines that the gender effect applies uniformly to all types of firms. Education is non-monotonic for the conditional mean model and for low quantiles. In those cases, incomplete secondary education has the highest effect in both subsamples. However, education becomes monotonically increasing for tZ0.5. This determines that for firms in the low conditional quantiles, higher education is not necessarily associated with higher revenues, but it is with outstanding firms. Finally, the reasons to become

126

TOMMASO GABRIELI ET AL.

entrepreneur show interesting variability across quantiles. Reasons such as ‘‘Accumulated experience’’, ‘‘Be independent’’, ‘‘Make a good deal’’ and ‘‘Profitable business’’ which may be associated with entrepreneurs with high ability are larger for high quantiles, while reasons for low-ability entrepreneurs (such as ‘‘To help family income’’) are larger for the low quantiles. We also implement the method of Frolich and Melly (2008) and Frandsen (2008) for comparison reasons. This estimator differs in several aspects to the one proposed here. First, it corresponds to a standard regression discontinuity design and is not designed to be used in a difference-indifferences fashion. In our setup we implement this estimator by comparing only treated (born after SIMPLES) and non-treated (born before SIMPLES) considering a discontinuity in age of the firm. Second, as a nonparametric estimator, it posses difficulties with a large set of covariates. Thus, we implement the estimator without covariates and then, following Frolich and Melly (2008), we use an alternative parametric specification using the propensity score (Prob½ti  tjd; x; jti  tjo) as a unique conditioning variable. Third, standard errors are available only for the case without covariates, and therefore only point estimates are provided for the case with covariates. Finally, the choice of bandwidth is always an important concern in nonparametric and semiparametric estimation, and estimates may have large variation depending on the bandwidth. We therefore use three different choices of bandwidth. We estimate the model using the subsample of all micro-firms.24 The results for both estimators, with and without covariates, are presented in Table 3. Regarding the case with no covariates, there are only a few quantiles where the point estimates are statistically different from zero. The point estimates for the bandwidths two and three are somehow similar to the IVQR estimates, while those for a bandwidth of four are negative and are not statistically different from zero, evidencing the sensitivity to the bandwidth choice. When covariates are used through the propensity score, the point estimates are reduced to 1.1 on average. These point estimates provide additional evidence on formality having a positive effect on revenues. As mentioned above, the lack of a measure of dispersion precludes us to provide any inference on these estimates. Thus, we are not able to statistically analyze the question posed in the chapter regarding which firms benefit from the reduction in formality costs. However, given the large standard errors for the IV estimates presented in Tables 1 and 2, in most cases, these nonparametric estimates are included in the 95% confidence intervals of the estimates discussed above.

127

Quantile Regression Discontinuity Analysis

Table 3.

Nonparametric Analysis Without and With Covariates – All Micro-Firms.

Quantiles

t ¼ 0.1 t ¼ 0.2 t ¼ 0.3 t ¼ 0.4 t ¼ 0.5 t ¼ 0.6 t ¼ 0.7 t ¼ 0.8 t ¼ 0.9

Without Covariates

With Covariates

Band ¼ 2

Band ¼ 3

Band ¼ 4

Band ¼ 2

Band ¼ 3

Band ¼ 4

5.586 (5.92) 4.500 (7.16) 4.605 (6.17) 4.700 (11.94) 4.423 (2.30) 4.423 (2.40) 4.423 (2.49) 4.423 (2.59) 4.605 (9.23)

3.832 (3.74) 3.832 (3.70) 3.817 (5.16) 4.209 (3.83) 4.081 (3.89) 4.159 (4.27) 4.338 (4.65) 4.232 (3.48) 4.232 (3.47)

3.011 (2.30) 2.606 (2.99) 2.548 (1.50) 2.534 (1.36) 2.485 (1.24) 2.659 (2.05) 3.079 (2.69) 3.344 (1.77) 2.784 (1.78)

1.194 – 1.099 – 1.066 – 1.130 – 1.110 – 1.099 – 1.163 – 1.139 – 1.124 –

1.281 – 1.099 – 0.971 – 1.003 – 1.110 – 1.012 – 1.163 – 1.139 – 1.046 –

1.281 – 1.099 – 1.012 – 1.099 – 1.099 – 1.107 – 1.163 – 1.281 – 1.225 –

Note: 6,741 observations. Band ¼ Bandwidth. Standard errors in parenthesis.

Significant at the asymptotic 10% level; Significant at the asymptotic 5% level; Significant at the asymptotic 1% level.

CONCLUSION AND POLICY IMPLICATIONS The econometric results are summarized as follows. First, the results show positive point estimates evidencing that formality has a positive effect on revenues. Overall this confirms the effect of formality on firm performance is positive and suggests that formality gains are potentially large. From a policy perspective this implies that improving institutions to increase participation benefits the micro-firm sector. Reducing the cost of formality allows firms to approach the steady state size dictated by their intrinsic entrepreneurial ability. Second, the answer to the question ‘‘which firms benefit from the tax reduction and simplification?’’ is given by the estimates from the empirical exercise showing that the target population corresponds to t quantiles in 0.10rtr0.50 or 0.10rtr0.60 depending on the sample. This means that

128

TOMMASO GABRIELI ET AL.

SIMPLES had a potential effect on 40–50% of the micro-entrepreneur population, mostly concentrated on low-ability firms. Note that this corresponds to benefits in terms of changing formality status (i.e., becoming formal) not on the overall effect of SIMPLES, because SIMPLES also had benefits for those already formal that would face lower taxes. The theoretical model also shows that the larger is the tax reduction, the larger will be the segment of firms that will change their formality status. Third, for the target group where the effect of formality can be identified, we find evidence of heterogeneity across quantiles on the impact of license on the conditional distribution of revenues. These estimates suggest that reducing the cost of formality might significantly benefit lowability firms more. However, these effects can only be studied for the quantiles where the effect of formality can be identified, and therefore, we cannot offer a complete analysis of the heterogeneity in the effect of formality on revenues.

NOTES 1. Re´gimen Simplificado para Pequen˜os Contribuyentes, see Gonza´lez (2006). 2. SARE stands for ‘‘Sistema de Apertura Ra´pida de Empresas.’’ It was implemented in selected municipalities and consolidated in single local offices all the federal, state, and municipal procedures needed to register a firm, reducing the total duration of the process to at most 48 hours. 3. SIMPLES stands for ‘‘Sistema Integrado de Pagamento de Impostos e Contribuc- o˜es as Microempresas e Empresas de Pequeno Porte.’’ See section ‘‘The SIMPLES Program and Identification Strategy’’ for a detailed description of the program. 4. This is the definition adopted in Fajnzylber et al. (2009, 2011). 5. Ability is thus only relevant when managing a firm. Modeling the salaried sector exceeds the scope of this chapter. 6. The results of the model would still apply with any concave production function. 7. The functional form of the probability of detection could be more general: Paula and Scheinkman (2007) show that as long as p is an increasing function of k there is still a threshold level of ability such that entrepreneurs go from informal to formal and therefore the same conclusions hold. 8. As in Rauch (1991) and Paula and Scheinkman (2007, 2010) the weakly monotonic relationship between exogenous ability and optimal level of formality is implied by the standard assumption of convex technology. Non-convex profit functions could imply more than one crossing point; hence, a non-monotonic relationship over a certain range of ability, but the relationship would still be monotonic for high levels of ability if formality constraints the production set.

Quantile Regression Discontinuity Analysis

129

Moreover, it could be an interesting avenue for future research to analyze the possibility that ability is not exogenous but is affected by the formality/informality decision, for instance by learning dynamics. 9. As the tax rate f changes, the equilibrium wage may in principle change. Ceteris paribus, a decrease in the tax fosters a larger formal sector, but this effect increases in turn the demand for labor. We abstract from the possibility of a change in the salaried wage. 10. We use a ¼ 0.2, b ¼ 0.7, r ¼ 3, w ¼ 5. Then, it can be computed that k ¼ 3.123 and y ¼ 10. Fig. 1 shows the informal entrepreneurs’ profit (thick line) and those of formal entrepreneurs given f ¼ 0.2 (thin line) and given fu ¼ 0.1 (dashed line). It can be computed that the threshold value of ability is y ¼ 16:1 for f ¼ 0.2 and decreases to y ¼ 13:2 for fu ¼ 0.1. 11. These would be the case with the non-convexities described in McKenzie and Woodruff (2006), where the return to capital is higher for low-capital firms. 12. Given values a ¼ 0.2, b ¼ 0.7, r ¼ 3, w ¼ 5, it can be computed that k ¼ 3.123 and y ¼ 10. Fig. 2 shows a plot of the informal entrepreneurs’ profit (thick line) and those of formal entrepreneurs given f ¼ 500 (thin line) and given fu ¼ 250 (dashed line). It can be computed that the threshold value of ability is y ¼ 16 for f ¼ 500 and decreases to y ¼ 14:5 for fu ¼ 250. 13. This corresponds to the indicator variable ELIG below. 14. See the analysis for micro-firms in Mexico and other evidence for Latin American countries in Fajnzylber et al. (2009). 15. We thank Tiziano Razzolini and an anonymous referee for pointing this out. 16. A formal analysis of the choice non-eligible vs. eligible sector and of the general equilibrium effects of a reduction in the cost of formality (see Note 9) goes beyond the scope of the present chapter. 17. We thank an anonymous referee for pointing this out. 18. As discussed in Chernozhukov and Hansen (2006), the exact form of A is irrelevant when the model is exactly identified, but it is desirable to set A equal to the asymptotic variance–covariance matrix of g^ ðaðtÞ; tÞ otherwise. 19. Chernozhukov and Hansen (2005) show that it is possible to achieve identification with IVQR using a weaker assumption called rank similarity. Rank similarity relaxes exact rank invariance by allowing unsystematic deviations, ‘‘slippages’’ in one’s rank away from some common level. 20. The implied first-stage regression is Licensei ¼ a1AFTERi þ a2ELIGi þ a3(AFTERi ELIGi) þ a4xi þ ei. 21. Kolmogorov–Smirnov test in QR are discussed in Chernozhukov and Hansen (2006) and Koenker (2005). 22. In general the index used for Kolmogorov–Smirnov tests in QR is symmetric of the form [e,1e]. However, in some situations it is desirable to restrict the interval of estimation to a subinterval, as [t0,t1]A(0,1). As Koenker (2005) discusses, this can be easily accommodated by using a renormalized statistic. 23. However, as argued by an anonymous referee, it is also the case that women engage in less risky activities, and it is not necessarily the case that more risk is optimal. 24. Similar results are obtained for the owners-only subsample. Results are available from the authors upon request.

130

TOMMASO GABRIELI ET AL.

ACKNOWLEDGMENTS The authors would like to express their appreciation to two anonymous referees, Tiziano Razzolini, Kostantinos Tatsiramos, Hartmut Lehmann, Scott Adams, John Heywood, William Maloney, Blaise Melly, and participants at the IZA/World Bank Workshop on Institutions and Informal Employment in Emerging and Transition Economies, Bonn, the 2011 North American Summer Meeting of the Econometric Society, St Louis, the 2011 Conference of the Royal Economic Society, Royal Holloway, and seminars at City University London, Queen Mary University London and Universidad Auto´noma de Barcelona for helpful comments and discussions. All the remaining errors are ours.

REFERENCES Chernozhukov, V., & Hansen, C. (2005). An IV model of quantile treatment effects. Econometrica, 73, 245–261. Chernozhukov, V., & Hansen, C. (2006). Instrumental quantile regression inference for structural and treatment effects models. Journal of Econometrics, 132, 491–525. Chernozhukov, V., & Hansen, C. (2008). Instrumental variable quantile regression: A robust inference approach. Journal of Econometrics, 142, 379–398. Chesher, A. (2003). Identification in nonseparable models. Econometrica, 71, 1405–1441. Chesher, A. (2005). Nonparametric identification under discrete variation. Econometrica, 73, 1525–1550. de Soto, H. (1989). The other path: The invisible revolution in the third world. New York, NY: Harper and Row. Fajnzylber, P., Maloney, W. F., & Montes-Rojas, G. (2006). Microenterprise dynamics in developing countries: How similar are they to those in the industrialized world? Evidence from Mexico. World Bank Economic Review, 20, 389–419. Fajnzylber, P., Maloney, W. F., & Montes-Rojas, G. (2009). Releasing constraints to growth or pushing on a string? Policies and performance of Mexican micro-firms. Journal of Development Studies, 45, 1027–1047. Fajnzylber, P., Maloney, W. F., & Montes-Rojas, G. (2011). Does formality improve microfirm performance? Evidence from the Brazilian SIMPLES program. Journal of Development Economics, 94, 262–276. Frandsen, B. R. (2008). A nonparametric estimator for local quantile treatment effects in the regression discontinuity design. Mimeo, MIT, Cambridge. Frolich, M., & Melly, B. (2008). Quantile treatment effects in the regression discontinuity design. IZA Discussion Paper No. 3638, IZA, Bonn. Gerxhani, K. (2004). The informal sector in developed and less developed countries: A literature survey. Public Choice, 120, 267–300. Gonza´lez, D. (2006). Regı´menes Especiales de Tributacio´n para Pequen˜os Contribuyentes en Ame´rica Latina. Banco Interamericano de Desarrollo.

Quantile Regression Discontinuity Analysis

131

Guiteras, R. (2008). Estimating quantile treatment effects in a regression discontinuity design. Mimeo, University of Maryland. Hahn, J., Todd, P., & van der Klaauw, W. (2001). Identification and estimation of treatment effects with a regression-discontinuity design. Econometrica, 69, 201–209. Harris, J., & Todaro, M. (1970). Migration, unemployment and development: A two-sector analysis. American Economic Review, 40, 126–142. Kaplan, D., Piedra, E., & Seira, E. (2006). Are burdensome registration procedures an important barrier on firm creation? Evidence from Mexico. SIEPR Discussion Paper No. 06-13. Stanford University, Stanford. Koenker, R. (2005). Quantile regression. Cambridge: Cambridge University Press. Koenker, R., & Hallock, K. (2001). Quantile regression. Journal of Economic Perspectives, 15(1), 143–156. Maloney, W. (1999). Does informality imply segmentation in urban labor markets? Evidence from sectoral transitions in Mexico. World Bank Economic Review, 13, 279–302. Maloney, W. (2004). Informality revisited. World Development, 32, 1159–1178. Mandelman, F., & Montes-Rojas, G. (2009). Is self-employment and micro-entrepreneurship a desired outcome? World Development, 37, 1914–1925. McKenzie, D., & Woodruff, C. (2006). Do entry costs provide an empirical basis for poverty traps? Evidence from Mexican microenterprises. Economic Development and Cultural Change, 55, 3–42. Monteiro, J. C. M., & Assunc- a˜o, J. J. (2006). Outgoing the shadows: Estimating the impact of bureaucracy simplification and tax cut on formality and investment. European Meeting of the Econometric Society, Vienna. Paula, A., & Scheinkman, J. (2007). The informal sector. NBER Working Paper No. 13486. National Bureau of Economic Research, Cambridge. Paula, A., & Scheinkman, J. (2010). Value added taxes, chain effects and informality. American Economic Journal: Macroeconomics, 2, 195–221. Pereda-Fernandez, S. (2010). Quantile regression discontinuity: Estimating the effect of class size on scholastic achievement. Master Thesis CEMFI No. 1002. CEMFI, Madrid. Rauch, J. (1991). Modelling the informal sector formally. Journal of Development Economics, 35, 33–47. van der Klaauw, W. (2002). Estimating the effect of financial aid offers on college enrollment: A regression-discontinuity approach. International Economic Review, 43, 1249–1287.

132

TOMMASO GABRIELI ET AL.

APPENDIX Proof of Proposition 1 An entrepreneur with ability yi y always finds optimal to be informal. An entrepreneur with ability yiWy finds optimal to become formal if and only if pFi  pIi . Plugging the first-order conditions into Eq. (1) we obtain that  a=ð1abÞ  b b=ð1abÞ ð1=ð1abÞÞ a I  p ðy Þ ¼ ð1  a  bÞy r w and F

p ðyi Þ ¼ ð1  a  bÞðð1  fÞyi Þ

1=ð1abÞ

aa=ð1abÞ  b b=ð1abÞ r w

An entrepreneur with ability yiWy who decides to be informal will choose capital k and labor l I ðk ; yi Þ ¼ ðbyi ka =wÞð1=ð1bÞÞ . Defining gi  yi =y  1 we can re-express yi ¼ ð1 þ gi Þy and l I ðk ; yi Þ ¼ ð1 þ gi Þl  . Plugging k and l I ðk ; yi Þ into the expression for the profit of a formal entrepreneur we obtain that pI ðyi Þ ¼ ð1 þ gi Þ1=ð1bÞ ð1  a=ð1 þ gi Þ1=ð1bÞ  bÞ yð1=ð1abÞÞ ða=rÞa=ð1abÞ ðb=wÞb=ð1abÞ . Therefore, we obtain that pI ðyi Þ4pF ðyi Þ if and only if ðð1 þ gi Þa=ðð1þbÞð1abÞÞ Þ=ð1  ða=ðð1 þ gi Þ1=ð1bÞ Þ  bÞo1=ðð1  a  bÞ ð1  fÞ1=ð1abÞ Þ. The left-hand side ð1 þ gi Þa=ð1þbÞð1abÞ 1  a=ð1 þ gi Þ1=ð1bÞ  b

(5)

of the inequality above increases in gi as the derivative of Eq. (5) dðÞ=dgi ¼ ðað1  xÞ=ð1  a  bÞÞxða=ð1bÞð1abÞÞ1 Þ=D2 , where D  denominator of Eq. (5), x  ð1 þ gÞð1=ð1bÞÞ and 0oxo1. Define g such that the condition above is satisfied with equality. This condition identifies a threshold level of ability y ¼ ð1 þ g Þy such that an  entrepreneur i decides to become formal if and only if yi 4y. Notice that the right-hand side of the inequality increases in f therefore g and y increase in f. QED

133

Quantile Regression Discontinuity Analysis

Proof of Proposition 2 The second cross-derivative d 2 pF ðÞ=dydf is negative. Therefore, the difference ðpF ðf0 Þ  pF ðfÞÞ, where f0 of, increases in y. This proves the proposition for formal entrepreneurs. pF ðf0 Þ increases in y at a faster rate than pF(f) as d 2 pF ðÞ=ðdyÞ2 is decreasing in f. The result of Proposition 1 (single crossing between pF and pI) implies that pF(f) increases at a faster  Therefore, it must be the case that pF ðf0 Þ increases at a rate than pI for yoy. 0 I  where y 0 is the new cut-off level of ability faster rate than p for y 2 ½y ; y, given fu. Therefore, this proves the proposition also for those entrepreneurs that change their status from informal to formal as a result of the policy change. Plugging the first-order conditions into the expression for output/ revenue yi ¼ yi kai l bi we obtain that I



y ðy Þ ¼ y

ð1=ð1abÞÞ

aa=ð1abÞ  b b=ð1abÞ r

w

and F

1=ð1abÞ

y ðyi Þ ¼ ðð1  fÞyi Þ

aa=ð1abÞ  b b=ð1abÞ r

w

represent respectively revenues for informal and formal entrepreneurs. It is immediate to notice that the revenue functions behave exactly as the profit functions. QED.

CHAPTER 4 DETECTING WAGE UNDER-REPORTING USING A DOUBLE-HURDLE MODEL + Bala´zs Reizer Pe´ter Elek, Ja´nos Ko¨llo, and Pe´ter A. Szabo´ ABSTRACT We estimate a double-hurdle (DH) model of the Hungarian wage distribution assuming censoring at the minimum wage and wage underreporting (i.e. compensation consisting of the minimum wage, subject to taxation and an unreported cash supplement). We estimate the probability of under-reporting for minimum wage earners, simulate their genuine earnings and classify them and their employers as ‘cheaters’ and ‘non-cheaters’. In the possession of the classification, we check how cheaters and non-cheaters reacted to the introduction of a minimum social security contribution base, equal to 200 per cent of the minimum wage, in 2007. The findings suggest that cheaters were more likely to raise the wages of their minimum wage earners to 200 per cent of the minimum wage, thereby reducing the risk of tax audit. Cheating firms also experienced faster average wage growth and slower output growth. The results suggest that the DH model is able to identify the loci of wage under-reporting with some precision.

Informal Employment in Emerging and Transition Economies Research in Labor Economics, Volume 34, 135–166 Copyright r 2012 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1108/S0147-9121(2012)0000034007

135

PE´TER ELEK ET AL.

136

Keywords: Double-hurdle model; wage under-reporting; minimum contribution base JEL classification: C24; H26; J30; J38

INTRODUCTION The evasion of payroll taxes has two main forms. One is unreported (black) employment, when the employee is not registered and neither she nor her employer pays any taxes. The other main form is the under-reporting of wages, or grey employment, when the compensation consists of an officially paid amount, subject to taxation, and an unreported supplement also known as an ‘envelope wage’ or ‘under-the-counter payment’. In order to maximize the total evaded tax, the officially paid wage is often (but not always) chosen as the minimum wage (MW). In this chapter, we estimate the prevalence of disguised MW earners with the double-hurdle (DH) model, first proposed by Cragg (1971), using linked employer–employee data. The DH is a potentially suitable method for disentangling genuine from ‘fake’ MW earners, relying on the assumption that MW payment is governed by two different processes: market imperfections implying censoring at the MW, on the one hand, and nonrandom selection to wage under-reporting, on the other. Our application of the DH for Hungary assumes that a spike at the MW was observed for two reasons: (i) because of constraints and costs preventing firms from firing all low-productivity workers after a wave of exceptionally large hikes in the MW and (ii) because of tax fraud. That said, a worker’s genuine wage is observed only if her productivity exceeds the MW and her wage is fully reported. The DH model simultaneously deals with the censoring problem and selection to tax fraud, and estimates the probability of cheating for each MW earner. In the possession of the parameters, one can also simulate the ‘genuine’ wages of MW earners. The DH model’s reliance on distributional properties (as well as the difficulty in finding exclusion restrictions for the selection equation) warns us not to take the estimates at face value. Therefore, we test the validity of the DH results by exploiting a unique episode of Hungary’s unconventional MW policies. The test examines the introduction of a minimum contribution base amounting to 200 per cent of the MW (2 MW), in 2007. After the introduction of the reform, firms paying wages lower than 2 MW faced an

Detecting Wage Under-Reporting Using a Double-Hurdle Model

137

increased probability of tax authority audit and a higher risk of being detected as cheaters. Firms were required to report that they paid wages below 2 MW and provide evidence, upon request, that their low-wage workers were paid at the going market rate. The reform created incentives for cheating firms to raise the reported wages of MW earners to 2 MW, while non-cheaters (those paying genuine minimum wages) had no interest to do so. We distinguish cheaters from non-cheaters on the basis of DH estimates for 2006 and check how the cheating proxies affected the probability that a worker earning the MW in 2006 earned 2 MW in 2007. We also study how the wages of MW earners changed in 2006–2007. We find that suspected cheaters were more likely to shift their workers from MW to 2 MW compared to non-cheating firms. Furthermore, we find that the sales revenues of cheating firms were adversely affected by the reform. At least in the East and South-East of Europe, MW policies are strongly influenced by the conviction that nearly all MW workers earn untaxed side payments. Our results suggest that while the suspicions are not groundless they are overstated: we estimate the share of ‘disguised’ MW earners to be around 50 per cent and the share of cheating enterprises to fall short of 40 per cent. The high share of non-cheating firms and genuine MW earners warns against radical, fiscally motivated experiments with the MW, which may put unskilled jobs at risk. While the statistical profiles derived from the DH model may help the better targeting of tax authority inspection, they can also facilitate more circumspect MW policies. The chapter is organized as follows. The second section gives a brief overview of the literature, while the third section gives the MW regulations and the wage distribution in Hungary. The fourth section introduces the DH model, explains the estimation of its parameters, shows how the probability of cheating and ‘genuine’ wages are simulated and how we classify workers and firms on the basis of the DH estimates. The fifth section introduces the data. The sixth section presents the estimates of the DH model. The seventh section presents the methods, data and results of the test and the eighth section concludes.

WAGE UNDER-REPORTING AND THE MINIMUM WAGE – AN UNDER-RESEARCHED AREA Compared to the vast literature on income under-reporting and MW regulations, respectively, the body of research on how these two areas relate to each other seems rather thin. Most of what we know empirically about

138

PE´TER ELEK ET AL.

this relationship comes from anecdotal evidence, inspection of aggregate data, scarce survey results and a few attempts to identify the incidence of envelope wages indirectly. Theoretical work is largely missing. Although several mechanisms may cause a spike of the wage distribution at the MW, including the tacit collusion of employers (Shelkova, 2008) or the extrusion of wages due to the effective MW (DiNardo, Fortin, & Lemieux, 1996), grey employment is certainly among the suspects. Cross-country data suggest a positive correlation between the size of the spike and estimated size of the informal economy (Tonin, 2007). Several accession countries including Hungary, Latvia, Lithuania and Romania have (or had) high shares of MW earners, while their Kaitz indices are (were) in the middle range, suggesting that disguised MWs may be particularly widespread in these countries. Similar observations are interpreted in a similar way in World Bank (2005). Erdogdu (2009) reports on the basis of several surveys that under-thecounter payments are prevalent in the wage policy of Turkish firms. There is a relatively extensive literature focusing on grey employment in the Baltic states. Relying on survey results, Masso and Krillo (2009) point out that 16–23 per cent of the MW earners received envelope wages in Estonia and Latvia but only 8 per cent in Lithuania in 1998. Meriku¨ll and Staehr (2010) show that young employees and people working in construction and trade are most likely to get unreported cash supplement on top of their official salary in the three Baltic countries. Kriz, Meriku¨ll, Paulus, and Staehr (2007) present similar results on the distribution of envelope wages using three different Estonian data sets. According to the Eurobarometer survey conducted by the European Commission in 2007 (European Commission, 2007), 5 per cent of employees in the EU receive part or all of their regular income untaxed and this ratio is over 10 per cent in some central and eastern European countries (8 per cent in Hungary), but there is no information on how many of them are officially paid the MW. Some studies obtain evidence on disguised MWs indirectly, by comparing the reported consumption–income profiles of households. Using household budget survey data from Hungary, Benedek, Rigo´, Scharle, and Szabo´ (2006) looked at the winners and losers from the 2001 to 2002 MW hikes (see also Szabo´, 2007). They observed income loss without the loss of a wage earner in the high-income brackets where substantial under-reporting is most likely to occur. For these households, the increasing MW may have implied higher taxes and lower net income. Based on the same data set, Tonin (2011) analysed changes in the food consumption of households affected by the MW hike compared to unaffected households of similar

Detecting Wage Under-Reporting Using a Double-Hurdle Model

139

income. He found that food consumption fell in the treatment group relative to the controls – a fact potentially explained by a fall in their unreported income in response to the MW hike and growth of the associated tax burden. The theories of wage under-reporting (Allingham & Sandmo, 1972; Yaniv, 1988) shed light on the incentives to engage in tax fraud under alternative penalty and withdrawal schemes, but they do not explicitly discuss the case of reporting the MW to tax authorities. This is the costminimizing choice for the firm (unless MW payment provokes audits thereby decreasing the expected gain from cheating), but it also requires the cooperation of workers. As Madzharova (2010) notes: if the actual or perceived linkages between contribution payments and pensions or access to health services are weak and/or workers see that their payments feed corruption rather than are used to finance public services, they will be willing to accept the lowest possible reported wage. Theoretical models explicitly addressing the issue of wage underreporting cum MW regulations include Tonin (2011) and Shelkova (2008). Tonin argues that the MW induces some workers whose productivity is above the MW, but who would have declared less if there was no MW, to increase their declared earnings to the MW level. Workers with productivity below the MW either work in the black market or withdraw from the labour force, while high-productivity workers are unaffected. This is a possible explanation of why a spike at the MW appears in the distribution of declared earnings. Shelkova assumes that low-productivity labour is homogenous and easy to replace thanks to the low fixed costs of hiring. If a non-binding MW exists and employers act symmetrically, then tacit collusion and offering the MW to low-productivity workers is profit maximizing and dominant strategy for the companies. An increase in the MW increases the probability of collusion since the incentive for deviation is weaker. This implies that a higher MW increases the spike there. Our empirical work attributes the sudden nascence and decease in a huge spike at the MW to state intervention, on the one hand, and tax evasion, on the other. We look at a unique period in Hungary’s MW history, which quadrupled the spike at the MW in only two years (when the MW was nearly doubled in 2001–2002) and decreased it by a factor of 2.5 in only one year (when a double contribution base was introduced in 2007). We do not believe that these sudden and enormous changes could be explained by the established strategic behaviour of enterprises underlying Shelkova’s model. It is also hard to trust that Tonin’s assumption,

140

PE´TER ELEK ET AL.

stating that the marginal products of those at the spike exceed the MW, was valid in the period we are looking at. When the plan of increasing the MW from Ft 25,500 to Ft 50,000 was announced, 32.7 per cent of the private sector employees earned less than that. When the idea of the minimum contribution base came up, 58 per cent had wages below 2 MW. It is quite obvious that the vast majority of the affected workers remained in employment for a protracted period (or until recently) and many of them had productivity below the aforementioned thresholds after the hikes. It took time until mobility between jobs, changes of the product mix and technology, adult training and other forms of adjustment could restore (if at all) the optimum condition for mutually gainful employment without causing massive unemployment in between.1 Therefore, we stick to the assumption that in the period under examination the spike at the MW was explained by under-reporting and the continuing employment of many low-productivity workers – two different processes that we try to model following the DH approach.

THE MINIMUM WAGE AND THE WAGE DISTRIBUTION IN HUNGARY MW regulations had minor impact on the Hungarian wage distribution until the millennium.2 As shown in Fig. 1, the MW–average wage (AW) ratio slightly decreased in 1992–2000 and fell short of Spain’s, the laggard within the EU in that period. The fraction of workers paid 95–105 per cent of the minimum amounted to 5 per cent, a ratio similar to those reported by Dolado et al. (1996) for Austria, Belgium, the Netherlands, Denmark and the United States. In 2001–2002 the MW was nearly doubled in nominal terms, resulting in a 14 percentage point rise in the Kaitz index.3 The fraction of private sector employees earning near the MW jumped to 11 per cent in 2001 and 18 per cent in 2002. The wage distribution preserved its distorted shape until 2007, when a second spike appeared at 200 per cent of the MW, as shown in Fig. 2.4 That year, the Hungarian government introduced a minimum social security contribution base amounting to 2 MW. Firms were allowed to pay wages lower than 2 MW, but in case they did so they faced an increased probability of tax authority audit and a higher risk of being detected as cheaters (for paying disguised MW or for other reasons).5

141

Detecting Wage Under-Reporting Using a Double-Hurdle Model (A)

0

.1

5

.3

10

.5

15

.7

20

(B)

1990

1995

2000 MW/AW

2005

1990

2010

MW/MEDW

1995 >5 employees

2000 >10 employees

2005

2010

>20 employees

Fig. 1. The Minimum Wage and Minimum Wage Earners in Hungary 1992–2009. (A) The MW Compared to the Average Wage and the Median Wage 1992–2009. (B) Fraction Paid 95–105 Per Cent of the MW, 1992–2009. The data relate to gross monthly earnings in the private sector. Source: Wage Surveys.

.12

2000 .12

.1 .1 .08 Fraction

Fraction

.08 .06

.06

.04

.04

.02

.02

0

0 –2

0 Relative earnings (log)

–2

2

–1

2006

0 1 Relative earnings (log)

2

3

2

3

2007 .12

.12 .1 .08

.08

Fraction

Fraction

.1

.06

.06

.04

.04

.02

.02 0

0 –2

Fig. 2.

–1

0 1 Relative earnings (log)

2

3

–2

–1

0 1 Relative earnings (log)

The Wage Distribution in Selected Years. Source: Wage Surveys. Samples: Full-timers in the private sector.

PE´TER ELEK ET AL.

142

Table 1.

Fraction Paid the Exact Amount of the Minimum Wage in 2006. Per cent Paid the MW

Composition All MW Earners ¼ 100

Top managers Managers (heads of department, foremen, etc.) Managers of small firms (5–20 employees) Engineers Architects and construction technicians Professionals in health, education and social services (private) Other professionals Lawyers, business and tax advisors, accountants Freelance cultural occupations (musicians, actors, writers, etc.) Technicians Administrators Agents, brokers Office workers Blue collars in retail trade and catering Blue collars in transport Services A (other than B and C) Services B (health and social services, private) Services C (personal services) Farmers and farm workers Blue collars in heavy industry and engineering Blue collars in light industry Blue collars in construction (house building) Blue collars in civil engineering (roads, railways, bridges) Assemblers and machine operators Truck drivers Porters, guards, cleaners Unskilled labourers, casual workers

9.7 3.6 18.0 2.4 9.5 3.5

1.6 2.2 1.6 0.6 0.3 0.0

3.0 8.8

0.5 0.7

16.5

0.5

7.3 8.3 12.6 11.5 22.5 7.7 12.7 0.0 17.7 20.9 8.9 14.6 21.0 20.0

2.7 6.8 0.8 5.6 15.3 0.1 1.5 0.0 0.7 5.1 6.6 9.2 10.0 0.6

4.7 20.5 18.2 37.6

4.3 3.8 7.6 11.2

Total

10.8

100.0

Source: Wage Survey, 2006, estimation sample of the DH model. Number of observations ¼ 91,240. Note: For this table, some occupations were divided into parts on the basis of industrial affiliation and firm size in order to capture differences in the scope for cash transactions with customers (personal versus other types of services, small firm versus large firm managers).

Detecting Wage Under-Reporting Using a Double-Hurdle Model

143

The suspicion that the crowding of workers at the MW in 2001–2006 was partly explained by wage under-reporting is difficult to avert. In 2006, the fraction of MW earners amounted to 18 per cent among small firm managers, and close to 10 per cent among top managers in larger firms also earned the MW. High shares could be observed in a number of freelance occupations such as architects, lawyers, accountants, business and tax advisors, agents, brokers, artists, writers, film-makers, actors and musicians (13–17 per cent). The fraction was particularly high in those sectors, where cash transactions with customers frequently occur such as shops, hotels and restaurants (23 per cent), house building (21 per cent), personal services (18 per cent) and farming (21 per cent). In some low-wage occupations such as cleaners, porters and guards, the fraction earning the MW fell short of the above-mentioned levels (Table 1). Further doubts arise if we look at the wage distribution within occupations (Fig. 3). In 2006, the distribution for unskilled workers was strongly skewed at the MW with a small number of workers earning substantially more than that. By contrast, the wage distribution of managers, for instance, had a spike at the MW and a mode at 440 percent of the MW, clearly pointing to a minority of managers under-reporting their earnings. With the help of the DH model, we can utilize the information content of the different shapes of the wage distributions. In the next section, we summarize how the estimation proceeds, how the probability of underreporting and the MW earners’ ‘genuine’ wages are derived and how we classify workers and firms as cheaters or non-cheaters.

Managers

0

0

.02

.05

Fraction .04 .06

Fraction .1 .15

.2

.08

Unskilled workers

10

11

12 13 Log gross monthly earnings

14

10

12 14 Log gross monthly earnings

16

Fig. 3. The Wage Distribution in Two Occupations, 2006. Source: Wage Survey 2006, Private Sector. Occupational codes: Managers 1311–1429, Unskilled workers 9190.

PE´TER ELEK ET AL.

144

THE DOUBLE-HURDLE MODEL The Set-Up of the Model Let us use the notation y for the (normalized) logarithm of the ‘true’ wage, i.e. of the wage which would prevail in the absence of MW and underreporting. (We normalize y to be zero at the true MW.) The value of y is determined by some characteristics X of the employee and the firm, and we assume that its distribution is conditionally normal with expectation Xb and variance s2. (This is a standard assumption in the literature; see e.g. Meyer & Wise, 1983a, 1983b.) In the presence of MW and underreporting, a spike appears at the MW in the wage distribution. The observed wage (the logarithm of which – normalized again to be zero at the MW – will be denoted by y) may be equal to the MW for two reasons: because of constraints and costs preventing firms from firing low-productivity workers (in the simplest case those whose genuine wage would fall below the MW) or because of tax fraud (when the MW is reported to the authorities but an unobserved cash supplement is also given). The probability of cheating is determined by some characteristics Z of the employee and the firm, and X may be different from Z. Formally, omitting subscript i for the individual, the following model governs y and y: y ¼ Xb þ u

(1)

and we observe the reported log wage y according to the rule: 

y ¼



y 0

if Xb þ u40 and Zg þ v40 otherwise

(2)

Under-reporting occurs when both Xb þ uW0 and Zg þ vp0 hold, and in this case the observed wage is equal to the MW. The residuals u and v are zero-mean normally distributed, possibly correlated (r) random variables. s2 stands for the variance of u, while the variance of v is set equal to unity without loss of generality, hence the covariance matrix of (u, v) is given by: ! s2 rs2 (3) S¼ rs2 1 This is the DH model first proposed by Cragg (1971), with the restriction r ¼ 0, to model the purchase of consumer goods in a setting where the

Detecting Wage Under-Reporting Using a Double-Hurdle Model

145

decision to buy and the decision of how much to buy are governed by different processes. The name of the model comes from the fact that the spike of the distribution (in our case at the MW) is determined by two ‘hurdles’: a standard Tobit-type constraint (in our case following from the wage equation: Xb þ ur0) and a different second hurdle (following from the selection equation: Zg þ vr0). Note that the standard Tobit model is obtained as a special case when the second hurdle is not effective, e.g. when Z contains a sufficiently large constant and all other terms in g are zero, or when X ¼ Zs, b ¼ g (apart from a constant) and r ¼ 1. In our case, a second hurdle is needed because under-reporting and wage determination are governed by partly different processes. Since the paper of Cragg the model and its extensions have been widely used to analyse consumer and producer behaviour as well as problems in environmental and agricultural economics and banking (e.g. Labeaga, 1999; Martinez-Espineira, 2006; Moffatt, 2005; Saz-Salazar & Rausell-Ko¨ster, 2006; Teklewold, Dadi, Yami, & Dana, 2006). However, to our knowledge, only Shelkova (2008) used the model to analyse wage distributions, in a setting discussed earlier. In our application, the baseline DH model (Eqs. (1)–(3)) has to be slightly modified in order to better capture the features of the wage formation process. The first problem to be addressed is that the log wage distribution is not censored normal because of the crowding of wage earners just above the MW6 (see Panel A in Fig. 4). While at and above the median the distribution is close to the normal, we have more workers on the left tail than expected under normality. This poses a problem because – as usual for non-linear models – maximum likelihood estimation of the DH model yields consistent results only if the underlying distributions are well specified. Therefore, we apply a preliminary transformation that is roughly linear at higher wages and accounts for ‘crowding’ at lower wages. We assume that instead of y we observe g(y), where r is a coefficient to be determined:  x gðxÞ ¼ x þ r  exp  r r

if x  0

(4)

By the preliminary transformation g1, we can ensure that y is close to a (censored) normal distribution and hence the DH model can be applied. Our approach is in line with the DH literature, where a preliminary transformation is often needed to achieve normality: Martinez-Espineira (2006) and Moffatt (2005) use the Box-Cox, while Yen and Jones (1997) apply the inverse hyperbolic sine transformation.

PE´TER ELEK ET AL.

146

1.5 1 0

.5

Density

2

2.5

Panel A

0

2 4 log wages (normalized)

6

1.5 1 0

.5

Density

2

2.5

Panel B

0

Fig. 4.

2 4 transformed log wages

6

Wage Distribution Before and After the Transformation. Source: Wage Survey 2006, Private Sector. Full-timers.

The second possible problem concerns our assumption that cheating employers report the MW (and not a larger wage) to the authorities. This is a reasonable assumption for 2001–2006 because firms could maximize the evaded tax this way and the chance of tax audit was not increased for MWreporting firms before 2007. The model can be extended to allow for cheating above the MW (see Elek, Scharle, Szabo´, & Szabo´, 2009) but external (e.g. survey-based) information is needed to identify its parameters. In this chapter, we use the simpler formulation.

Parameter Estimation Firstly, the parameter r of the preliminary transformation (Eq. (4)) should be determined. Instead of a likelihood-based statistical procedure, we make use of the fact that the wage distribution was close to log-normal in 2000

147

Detecting Wage Under-Reporting Using a Double-Hurdle Model

2 1.5 1 0

.5

normalized log wage (2002)

2.5

(see Fig. 2), changed substantially because of the MW increase and spillover effects in 2001–2002, and – in the absence of further drastic MW hikes – was practically unaltered in 2003–2006. Thus, we create a quasi-panel subsample of the LEED data for 2000–2002, and assign the median of the 2002 logarithmic wages (normalized to be zero at the MW) to the median of the 2000 logarithmic wages (again normalized) for each percentile of the wage distribution in 2000 (see the fifth section for details of the LEED data set). This graph hence shows the change in (normalized) logarithmic wages between 2000 and 2002 by percentiles. Finally, the function g (with unknown parameter r) is fitted to the percentile graph with non-linear least squares. This function gives a transformation for the log wages in 2002 and – for the reasons mentioned above – for 2006 as well. Our method yields r ¼ 0.49. Fig. 5 displays the function and its appropriate fit to the 2000–2002 wage percentiles, while Panel B in Fig. 4 shows that the transformed log wages (g1(y)) are approximately censored normal. For ease of notation, in what follows, we refer to g1(y) as y. Using the properties of the conditional distributions of the bivariate normal distribution, the likelihood function of the DH model (Eqs. (1)–(3)) can be shown to have the following form (for the sake of clarity, here we use

0

Fig. 5.

1 2 normalized log wage (2000)

3

The Function g(x) for r ¼ 0.49 and Its Fit to the Percentile Graph 2000–2002.

PE´TER ELEK ET AL.

148

subscripts i for the individuals): L¼

Y yi ¼0

½1  Fr;s;1 ðxi b; zi gÞ 

Y yi 40

"

!   # zi g þ rs ðyi  xi bÞ 1 yi  x i b pffiffiffiffiffiffiffiffiffiffiffiffiffi f F s s 1  r2 (5)

where Fr;s;1 denotes the bivariate normal distribution with covariance matrix given in Eq. (3), while F and f stand for the univariate standard normal distribution and density, respectively. Parameter estimation can be carried out with maximum likelihood, where we use cluster–robust standard errors to tackle the potential within-firm correlation in the error terms. If the DH model is correctly specified (including the distributional assumptions), then identification can be carried out even if X ¼ Z, i.e. based merely on non-linearities. However, to make the results more robust to deviations from the distributional assumptions, it is worth including variables that only influence the selection equation but not the wage equation (i.e. making valid exclusion restrictions). Therefore, in the wage equation we include the usual variables thought of as influencing the productivity of a worker such as her individual characteristics (experience, education, sex) and the characteristics of her firm (industry, productivity, fixed assets, location, size and ownership).7 Since the majority of these variables affect cheating behaviour as well, they are also present in the selection equation (e.g. for larger firms it is more difficult to hide envelope wages from the tax authority thus they tend to be less involved in grey employment.) In the selection equation we also include individual- and firm-level proxies directly affecting the decision to evade taxes. In particular, we distinguish some occupational categories that are more prone to cheating than others, mainly due to the lower risk of being caught such as managerial and freelance occupations, occupations with frequent cash transactions or jobs in trade, hotels and restaurants (see Table 2 for definitions). We also choose proxies for tax evasion from the corporate tax returns. It is expected that wage-underreporting firms tend to evade corporate taxation, thus tax liability correlates negatively with cheating. Another proxy is ‘other personnel related expenses’ which contain fringe benefits: these are rather complementary to wage payments hence a high share of personnel-related costs indicates compliance to the tax rules. The chosen indicators are indicative of compliance with the tax rules in fields other than wage payment. It is reasonable to assume that, after controlling for the usual factors in the wage equation, the firm-level instruments only influence the probability of cheating but not the genuine wages thus we have valid exclusion restrictions in the model.

Table 2.

DH Estimates of Wages for 2006. Standard Errora

Coefficient Wage equation for normalized log wages (also includes industry controls) Experience/10 0.327 Experience squared/100 0.049 Male 0.205 Vocational education 0.183 Secondary education 0.485 Higher education 1.191 Budapest 0.135 Value added per worker log 0.147 Fixed assets per worker log 0.007 Firm of foreign ownership 0.255 Firm with 5–10 employees 0.404 Firm with 11–20 employees 0.371 Firm with 21–50 employees 0.233 Firm with 51–300 employees 0.112 Constant 0.255

0.013 0.002 0.011 0.013 0.015 0.019 0.023 0.010 0.005 0.020 0.037 0.026 0.024 0.020 0.020

Selection equation Experience/10 Experience squared/100 Male Vocational education Secondary education Higher education Managerial and freelanceb Cash transactionsc Retail traded Budapest Works in a city Works in a village Corporate tax payment/sales revenues Other personnel-related expenses/payroll Firm of foreign ownership Firm with 5–10 employees Firm with 11–20 employees Firm with 21–50 employees Firm with 51–300 employees Constant

0.408 0.108 0.214 0.050 0.128 0.012 0.392 0.226 0.333 0.244 0.112 0.133 10.03 2.261 0.778 2.007 1.709 1.404 0.930 3.074

0.070 0.015 0.050 0.138 0.127 0.136 0.161 0.113 0.100 0.123 0.091 0.105 4.00 0.844 0.187 0.272 0.264 0.270 0.270 0.315

0.302 0.547

0.047 0.008

Rho Sigma No. of observations

91,240

Source: Wage survey 2006, private sector. a Cluster–robust standard errors, adjusted for firm-level clustering. b Managerial and freelance occupations (the latter includes professionals in culture and arts, agents and brokers). c Occupations where cash transactions occur frequently. Includes car mechanics, electricians, plumbers, household employees, couriers, truck drivers and workers in personal services and house building. d Occupations in retail trade. po0.1, po0.05, po0.01.

PE´TER ELEK ET AL.

150

Under-Reporting Probabilities, ‘Genuine’ Wages and Classification of Workers and Firms In the possession of the DH parameters, the probability of cheating for each MW earner can be estimated as: P ðunderreportingÞ ¼ PðXb þ u40; Zg þ v 0 y ¼ 0Þ Pðu4  XbÞ  Pðu4  Xb; v4  ZgÞ ¼ 1  Pðu4  Xb; v4  ZgÞ FðXb=sÞ  Fr;s;1 ðXb; ZgÞ ¼ 1  Fr;s;1 ðXb; ZgÞ

ð6Þ

Also, we can simulate the genuine wage of each MW earner as follows. We generate independent copies of bivariate normal random variables (u, v) with covariance matrix given in Eq. (3), and accept max (Xb þ u, 0) as the normalized genuine log wage of an MW earner if Xb þ ur0 or Zg þ vr0. If none of these conditions hold, the person cannot earn MW according to the model. Technically, for each MW earner, the (u, v) variables are simulated until at least one condition holds. Let us denote the estimated probability of under-reporting by a MW earner with P and the simulated wage with w (i.e. w ¼ MWexp(g(y)). As a benchmark definition cheating behaviour is assumed in case of PW0.5, but wWMW and wW1.5 MW will also be used for robustness checks.8 If we find at least one MW earner classified as ‘cheater’ in a firm, we treat the firm as a cheater. Since the majority of cheating firms are small, the use of other, more advanced criteria such as a certain threshold for the ratio of cheaters would be of limited practical importance.

DATA Throughout the chapter, we rely on the Wage Survey (WS) of the National Employment Service. The WS is a linked employer–employee data set recently comprising observations on over 150,000 individuals in about 20,000 firms and budget institutions. The survey was carried out tri-annually until 1992 and annually since then. In the enterprise sector, the WS covers businesses employing at least 5 workers. All Hungarian firms employing more than 20 workers are obliged to report data for the WS, while smaller firms are randomly selected from the census of

Detecting Wage Under-Reporting Using a Double-Hurdle Model

151

enterprises. In the years considered in our chapter, firms employing 5–20 workers had to report individual data on each employee, while larger ones reported data on a (roughly 10 per cent) random sample of their workers, selected on the basis of their day of birth. The observations are weighted by the Employment Service to correct for the selection of firms and individuals. The survey contains information on the wages and demographic and human capital variables of the workers and their job characteristics. The firm-level variables comprise industry, region, firm size, location, ownership, union coverage and financial variables including sales revenues, the net value of fixed assets, profits and several cost items. Our estimation sample covers the private sector and comprises 92,140 observations. In the seventh section, we use panels of individual- and firm-level observations. Firms in the WS can be directly linked and followed over time. Individuals cannot be linked directly, but they can be identified across waves with acceptable precision using data on their firm identifier, location of their workplace, year of birth, gender, education and four-digit occupational code. The worker and firm panels are non-randomly selected from the base period (2006) populations because of the survey design, on the one hand, and group-specific differences in firm survival, job destruction and quits, on the other. We control for selection on observables by estimating probit equations and using the inverse of the predicted probabilities of being in the panel as weights in those models, where weighting is allowed (for the method used, see e.g. Moffit, Fitzgerald, & Gottschalk, 1999). The probits are presented in Tables A2 and A3 of Appendix A. While the probability of making it to the panel was clearly non-random, weighting still had negligible effect on the estimated parameters.

RESULTS OF THE DOUBLE-HURDLE MODEL Table 2 presents the parameter estimates of the DH model. The parameters of the selection equation largely conform to intuition. ‘Grey’ occupations, male workers and employees in Budapest tend to underreport wages significantly, while foreign ownership, firm size, higher corporate tax liability and larger ‘other personnel-related expenses’ of the firm are positively correlated with labour tax compliance. After controlling for other factors, education does not seem to have a direct effect on cheating. The correlation between the error terms (r) is significantly negative, implying that unobserved factors leading to higher genuine wages

PE´TER ELEK ET AL.

152

tend to increase the probability of cheating. Similar results are obtained in the alternative specification, when occupation dummies (defined in Table A1 of Appendix A) are used instead of industry dummies in the wage equation. Using the estimated parameters, the probability of under-reporting among MW earners and their genuine wages were calculated. The results suggest that around half of all workers paid the MW hid part of their earnings from the tax authority. We estimate that the average ‘genuine’ wage of the MW earners amounted to approximately 170 per cent of the MW and the AW of cheating MW employees (using wWMW as the criterion for cheating) was around 250 per cent of MW. We should note that the exact share of cheaters and their simulated genuine wages are quite sensitive to the parameter r of the preliminary transformation but – more importantly from a modelling point of view – the partial effects of the different factors (occupations, etc.) are robust across different specifications. Table 3 displays the estimated probability of under-reporting among MW earners, their average genuine wage and a ‘cheating indicator’ by occupation, industry and firm size for the two different specifications.9 (The cheating indicator is defined as the share of cheating MW earners among all employees.) Looking at occupations, the estimated fraction of cheaters among MW earners is small for cleaners (10–20 per cent), unskilled labourers and agricultural workers (20–30 per cent), while it is much larger than average, e.g. for drivers, and approaches 100 per cent for managers and professionals. It is also clear that the share of MW earners is not a good indicator of underreporting because fraud is relatively frequent for some occupations with a high share of MW earners (e.g. in construction), while infrequent for others (e.g. among cleaners, unskilled labourers). The cheating indicator, which is the product of these two terms, is substantially higher than average in construction and trade professions and among drivers. As far as firm characteristics are concerned, Table 3 also displays the relation of economic branch and firm size to under-reporting. The cheating indicator is higher than average in construction, trade and hotels and restaurants, while it is the lowest in financial services (where the share of MW earners is the smallest as well). Both the ratio of MW earners and cheating behaviour are strongly negatively correlated with firm size: the cheating indicator is 10 times higher for firms with 5–10 employees than for larger firms with more than 50 employees. Foreign-owned enterprises tend to employ much less workers at the MW than domestic and mixed ones, but the ratio of under-reporting among them does not differ substantially. .

Table 3.

Predictions of the DH Model for 2006. Probability of Share of Under-Reporting MW Among MW Earners (%) Earners (%)

Cheating Indicator (%)b

Simulated Wage of Cheaters (MW ¼ 1.0)

(1)a

(2)a

(1)a

(2)a

(1)a

(2)a

46

48

11.9

5.5

5.7

2.6

2.4

31 45 40 52 41

29 56 43 39 49

27.5 23.4 6.7 20.5 12.8

8.6 7.9 10.6 13.0 2.7 2.9 10.6 7.9 5.3 6.3

2.3 1.9 2.4 2.2 2.1

1.8 1.8 2.2 1.8 2.0

18 30 35 38 59

13 22 45 24 72

23.8 33.3 5.7 15.6 15.8

4.2 3.2 10.1 7.5 2.0 2.5 5.9 3.7 9.3 11.4

2.5 2.0 2.1 2.5 2.2

1.6 1.6 2.1 1.6 2.2

52 72 64 74 94

59 84 78 96 97

11.0 5.3 6.4 5.2 2.5

6.5 4.4 5.0 5.0 2.4

2.7 3.2 3.1 3.7 5.1

2.4 2.8 2.9 4.4 5.1

Industries Agriculture, fishing 34 Manufacturing 40 Construction 42 Trade 52 Hotels, restaurants 40 Transport 68 Financial services 72 Real estate, business activities 51 Other 49

37 43 49 57 36 61 35 47 42

15.9 7.5 27.9 17.4 22.2 6.3 2.4 12.0 8.1

5.4 5.9 3.0 3.2 11.6 13.8 9.1 9.9 8.8 7.9 4.3 3.8 1.7 0.8 6.2 5.6 4.0 3.4

2.4 2.6 2.1 2.4 2.2 4.0 3.6 3.1 2.8

2.5 2.4 2.0 2.4 2.3 2.5 3.6 2.8 2.3

Firm size 5–10 employees 11–20 employees 21–50 employees 51–300 employees 300 þ employees

60 52 46 36 7

32.3 23.3 14.1 6.9 1.0

18.9 19.5 11.6 12.2 6.2 6.6 2.1 2.5 0.1 0.1

2.2 2.3 2.7 3.1 5.8

2.1 2.2 2.5 3.1 7.7

Total Occupationsc Agriculture Construction Services Trade Industry Other blue collar Cleaners Unskilled labourers Machine operators Porters and guards Drivers White collar Office clerks Technicians, assistants Administrators Managers Professionals

58 50 44 30 7

5.7 3.8 4.1 3.8 2.4

Source: Wage Survey 2006, private sector. Number of observations: 91,240. a Models: Wage equation with (1) industry dummies, (2) with occupation dummies. b Cheating indicator: Share of cheating MW earners among all employees. c Occupations: See Table A1 in Appendix A.

154

PE´TER ELEK ET AL.

The proportion of cheating firms (i.e. firms with at least one cheating employee) amounted to 17.3 per cent of all firms and 37.0 per cent among enterprises having at least one MW earner. While the estimates confirm that, in 2006, envelope wages existed at a large scale, they suggest that more than half of the MW earners did not receive cash supplement, and the majority of firms paying MW did not cheat on taxes. However, for reasons discussed earlier, the estimates should be treated with caution and the model’s predictive power needs to be checked.

TESTING THE PREDICTIONS OF THE DH: RESPONSES TO THE INTRODUCTION OF A MINIMUM CONTRIBUTION BASE As was briefly discussed earlier, the 2007 reform created incentives to raise the reported wages of disguised MW earners. Cheating firms could fully avert the risk of audit by officially paying 2 MW or more to their grey employees instead of MW. Furthermore, the public debate preceding the reform gave a clear warning that the tax authority would treat MW payment as a signal of tax evasion. Therefore, cheating firms had stronger motivation to shift their grey employees away from the MW, while noncheating enterprises, in the position to demonstrate that they pay ‘genuine’ MWs, had less incentive to raise the wages of their MW earners. The sudden shift of the spike of the wage distribution from MW to 2 MW in 2007 (shown earlier by Fig. 2) clearly indicated that firms – especially smaller ones – considered tax audit a credible threat. Before 2007, tax inspections were rather lax in Hungary. While firms employing more than 50 workers were checked by independent auditors and/or the tax authority annually and the monitoring activities of the tax authority concentrated on ‘accentuated tax payers’ (companies having the largest tax liabilities), entities without legal personality were monitored only in every 7th year and individual entrepreneurs only in every 23rd year on average. Penalties were insignificant. Consistent with the reform’s intentions, the new regulation changed the wage distribution of small firms dramatically while larger firms were weakly affected.10 We check how the wages of grey employees changed in response to the reform by estimating a probit equation (7) for a quasi-panel of individuals earning the MW in May 2006 and also observed in May 2007: Pðw1 ¼ 2MW1 w0 ¼ MW0 Þ ¼ FðCb þ ZgÞ

(7)

155

Detecting Wage Under-Reporting Using a Double-Hurdle Model

In the above equation, C denotes the dummy for cheating, Z comprises worker and firm characteristics and MW0 and MW1 stand for the minimum wage in 2006 and 2007. Base period MW earners are defined as those earning the exact amount of the minimum and those earning 95–105 per cent of the minimum, alternatively. The expectation is that bW0. We use fraud indicators defined on the individual level since the reform affected only the fake MW earners within firms: by shifting these particular employees away from the MW, the enterprise could reduce the risk of audit. The cheating proxies in Eq. (7) come from the DH model hence they are predicted regressors, and the estimation of their effect by simple maximum likelihood would not yield valid results. Therefore, in calculating the standard errors in the equation, we follow a two-step procedure. Firstly, we simulate the parameter vector of the DH model from its asymptotic normal distribution with its variance matrix, and create 100 simulated draws of firm-level cheating variables from the models. Secondly, using the different cheater classifications, we estimate Eq. (7) by bootstrap and finally take the sample mean and standard deviation of all simulated parameters. This way, the cumulated parameter uncertainty of the two stages is quantified – by using the asymptotic variance matrix in the first stage and direct bootstrap in the second. For simplicity, in the following text and tables, we refer to this procedure as ‘two-step bootstrap’. Note also that the resulting standard errors are only about 5 per cent larger than the ML standard errors of Eq. (7) because the error of the DH model, based on nearly 100 thousand observations, is negligible compared to the error of the test equation. The descriptive statistics in Table 4 yield preliminary support to our hypothesis: cheating enterprises were more likely to move their (apparent) low-wage workers away from the MW and shift them to 2 MW than non-

Table 4. The Wages of Year 2006 MW Earners in 2007. Wage in 2007

MW Between MW and 2 MW 2 MW Above 2 MW Total

Earned the MW in 2006 and Estimated to be Non-cheater (%)

Cheater (%)

23.2 70.0 5.9 0.9 100.0

14.3 71.0 13.4 1.3 100.0

Source: Wage Survey, MW earners in the worker panel of 2006–2007; number of observations ¼ 3,940.

PE´TER ELEK ET AL.

156

cheating firms. The MW earners (as of 2006) employed by fraudulent firms were 40 per cent less likely to earn the MW in 2007, 2.3 times more likely to earn 2 MW and 2.2 times more likely to earn 2 MW or more. The results from Eq. (7) are presented in Table 5. As shown in the first row, base period MW earners classified as cheaters (victims of cheating) were significantly more likely to earn 2 MW in 2007 than non-cheating MW earners. The estimated marginal effect of being a cheater amounts to 2.4 per cent when all controls are included – a remarkable impact if we take into account that the probability of earning 2 MW1 in 2007 conditional on earning MW0 in 2006 amounted to approximately 13 per cent. In the second row of the table an alternative to Eq. (7) estimates the probability that a MW0 earner was shifted to or beyond 2 MW1, i.e. the worker was moved out of the ‘danger zone’. The partial effects are positive and significant but lower. We may try to assess the magnitude of change induced by the 2007 reform and evaluate its economic significance in two ways. Firstly, one can make back-on-the-envelope calculations relying on the results in Table 4, and taking into consideration that the share of MW earners amounted to 47.9

Table 5.

The Effect of Estimated Cheating Behavioura on Wage Adjustment Between May 2006 and May 2007. Probit Marginal Effects at the Sample Means

c

Controlsb

Model

No

Probit 1 Probit 2

Education

Number of observations All

Partial effect

Z-valued Partial effect

Z-valued Partial effect

Z-valued

0.072 0.049

9.511 0.049 8.219 0.026

6.263 0.024 4.262 0.009

3.580 2.124

3,940 3,940

Source: Wage survey, MW earners in the worker panel of 2006–2007. a Individuals suspected of cheating in 2006 on the basis of the DH model. b Controls (all variables relate to 2006): dummies for education (college graduate, secondary school and vocational school), work experience in years, dummies for gender, municipality and the logarithm of firm size. c At probit1, the dependent variable is Pðw1 ¼ 2MW1 jw0 ¼ MW0 Þ, at probit2 Pðw1  2MW1 jw0 ¼ MW0 Þ. d Based on two-step bootstrap standard errors, adjusted for clustering by firms. po0.1, po0.05, po0.01.

Detecting Wage Under-Reporting Using a Double-Hurdle Model

157

per cent in cheating firms and 5.7 per cent in non-cheating ones. This implies that the reported wages of 6.4 per cent and 0.3 per cent of the employees were doubled in the two groups of firms, respectively.11 Holding other wages constant these pay rises implied 4.9 and 0.2 per cent increase in the average reported wages, respectively. Secondly, one may try to estimate the effect of cheating behaviour on firm-level outcomes by estimating regressions of the form: D ln x ¼ bC þ Zg þ 

(8)

where Dln x stands for the log changes in AWs, sales revenues and employment, alternatively, while C and Z denote the firm-level cheater dummy and the controls, respectively. The equations are estimated for 5,230 firms observed in 2006 and 2007, and the standard errors are estimated with the two-step bootstrap procedure described earlier. In the wage equation, we expect bW0 since raising the reported wages of grey employees must have increased the average reported wages of the cheating firms to some extent. The question of how actual costs and, therefore, output and employment were affected is more difficult to answer a priori. Firstly, firms may have cut the cash payments of the affected workers, offsetting the impact of increased payroll taxes. Secondly, some of them may have increased the share of cash transactions in order to economize on VAT instead of payroll taxes. The results presented in Table 6 suggest that the firm-level cheating proxy had positive effect on the change of observed AWs. Reported wages grew faster by 12 percentage points after controlling for industry, region, firm size, ownership and skill composition. The estimated gap between honest and dishonest firms is larger than the 5 percentage points difference calculated beforehand. This may result from the effect of the reform on other reported wages, or from unobserved shocks, for which we cannot effectively control with the firm-level variables at hand. In either case, the budgetary effect of the reform seems modest. According to the DH estimates, approximately 170,000 workers, or 11 per cent of the labour force represented by the WS, were employed by cheating firms. The AWs of these firms equalled 1.3 times the MW. Starting from these data and considering that the combined (employer and employee) social security contribution rate was 49 per cent and the lowest personal income tax rate was 18 per cent, we can estimate that the excess increase in reported wages in fraudulent firms resulted in an extra revenue of 12 billion Ft, or about 0.05 per cent of GDP. If we accept the back-on-the-

PE´TER ELEK ET AL.

158

Table 6. The Effects of Estimated Cheating Behavioura on the Changes of Selected Firm-Level Indicators in 2006–2007. OLS Regressions

Change of average wage (log) Change of employment (log) Change of sales revenues (log)

Controlsb

Partial effect

Z-valuec

Number of observations

No Yes No Yes No Yes

0.1294 0.1146 0.0073 0.0048 0.0454 0.0352

14.37 11.41 1.02 0.79 3.21 2.33

5,230 5,230 4,824

Source: Panel of firms observed in the Wage Survey in 2006 and 2007. a Firms suspected of cheating in 2006 on the basis of the DH model. b Controls include skill shares, average wage, average age and dummies for sectors, regions, type of municipality and state ownership. c Based on two-step bootstrap standard errors, adjusted for clustering by firms. po0.1, po0.05, po0.01.

envelope calculations, the budgetary effect is proportionally lower (about 6 billion Ft). The results indicate a significant negative effect on sales revenues and no effect on employment. A possible interpretation of this result is that the 2007 reform directed cheating enterprises to alternative forms of tax evasion and/ or urged them to cut envelop wages. The results presented in this section proved robust to changes in the definition of cheating and specification of the individual- and firm-level regressions. Weighting had practically no impact on the parameters. Using the exact amounts of the MWs rather than brackets around them, left the qualitative results unchanged in the individual regressions. We also examined the sensitivity of results to alternative cheating indicators based on the simulated wage (wWMW, wW1.1 MW, wW1.5 MW and wW2 MW). Since there was no significant deviation from the presented results, the regressions using alternative indicators are not presented.

CONCLUSIONS While grey employment and disguised MWs are widely debated issues in many emerging market economies, few attempts have been made to

Detecting Wage Under-Reporting Using a Double-Hurdle Model

159

measure their magnitude and distribution. We applied a DH model to this issue for Hungary in a period in which the presumptions of the model seemed to fit, i.e. censoring at the MW and wage under-reporting (at the MW) occurred simultaneously. If these preconditions are met, a properly specified DH model can estimate the ‘genuine’ wage distribution, permits the calculation of cheating probabilities and allows the simulation of ‘true’ earnings. The DH results for 2006 suggest that employers paid cash supplement to around half of all MW employees, and hinted at a wide (150 per cent) gap between reported and actual wages in these cases. The estimated distribution of under-reporting across occupations, industries and firm size seem to be consistent with the anecdotal evidence and survey-based results. The DH model makes strong assumptions about the wage distribution, and finding variables, which affect selection to cheating without affecting wages, is also rather difficult. Driven by the resulting uncertainty of the estimates, we conducted an experiment aimed at testing if the DH estimates have predictive power. It seems that the estimates worked well in the quasi-experimental setting analysed in the chapter: firms and workers suspected of tax evasion responded differently to the strong shock under investigation. We obviously make both type 1 and type 2 errors in disentangling cheaters from non-cheaters, but the results are encouraging for the analysis of ‘grey employment’ and, we believe, they also have practical importance. On the one hand, audits may be targeted by statistical profiles derived from the DH model, thereby improving compliance. However, by showing the loci of under-reporting, the DH estimates also draw attention to the limits of tax enforcement. Disguised minimum wages have high shares in services provided to households and small businesses, freelance occupations and small firm management – an attribute that limits the potential budgetary intakes from more stringent inspection. Cash transactions between households and the providers of personal services are difficult, if not impossible, to detect. Grey transactions of this kind can rather be whitened indirectly, by creating incentives to require receipts and making clear the link between reported income and access to publicly financed services and transfers such as pensions. On the other hand, the DH results call for more cautious MW policies. The micro-data do not support the popular belief that in Hungary ‘millions’ are fraudulently paid the MW – an assumption that served as a justification for regulations like the minimum contribution to be paid after 2 MW. Reducing the under-reporting of wages by means of substantially increasing the MW and/or the tax burden on it is an undoubtedly cheap alternative to

PE´TER ELEK ET AL.

160

independent checks and carefully designed presumptive taxation. However, raising the costs of low-wage employment across the board is a poorly targeted policy, which can further reduce unskilled job opportunities: an undesirable outcome in a country, where 6 out of 10 low-educated prime-age adults are out of work.

NOTES + ¨ si, and Vincze (2004) and Kertesi 1. Independent studies by Halpern, Koren, Koro and Ko¨llo+ (2003) estimated the short-run aggregate disemployment effect of the first MW hike to fall to the range of 1–1.5 per cent in 2001. 2. See Appendix B for further details of Hungary’s MW regulations. 3. The MW increased from Ft 25,500 in 2000 to Ft 40,000 on 1 January 2001 and Ft 50,000 on 1 January 2002. See Kertesi and Ko¨llo+ (2003) on the motives and aftermaths of the large hikes. 4. At the same time further minima were introduced for young and older skilled workers (1.2 MW, 1.25 MW) that flattened the spike near the MW. 5. Similar minimum contribution levels were introduced in Bulgaria and Croatia in 2003. The Hungarian regulations remained in effect until January 2010. 6. This is explained by spillover effects as argued in Dickens, Machin, and Manning (1994) and elsewhere. 7. For robustness check, in an alternative specification, we use occupation dummies instead of industry dummies in the wage equation. 8. The definition PW0.5 is preferable to e.g. wWMW because the latter includes some extra simulation uncertainties. 9. One is the baseline specification containing industry dummies in the wage equation, while the other contains occupation dummies instead. 10. The reform was initiated by a high (close to 10 per cent) budget deficit in 2006, and might be regarded as a simple form of presumptive taxation. For a discussion of the idea of presumptive taxation, practices in Italy and an application to Bulgaria, see Tanzi and Casanegra de Jantscher (1987), Arachi and Santoro (2007) and Pashev (2006), respectively. 11. Recall that cheaters shifted 13.4 per cent of their MW earners to 2 MW, while the respective share was only 5.9 per cent with non-cheaters. 12. All data quoted in Appendix B come from the Wage Survey.

ACKNOWLEDGMENTS The authors thank Tiziano Razzolini, two anonymous referees and seminar participants at IZA, Bonn and in Budapest for helpful comments on earlier versions. We are also grateful to the editors of this volume for their support and helpful suggestions.

Detecting Wage Under-Reporting Using a Double-Hurdle Model

161

REFERENCES Allingham, M., & Sandmo, A. (1972). Income tax evasion: A theoretical analysis. Journal of Public Economics, 1, 323–338. Arachi, G., & Santoro, A. (2007). Tax enforcement for SMEs. Lessons from the Italian experience? eJournal of Tax Research, 5(2), 225–243. Benedek, D., Rigo´, M., Scharle, A´., & Szabo´, P. (2006). Increases in the minimum wage in Hungary, 2001–2006. Working Papers in Public Finance No. 16. Ministry of Finance, Budapest. Cragg, J. (1971). Some statistical models for limited dependent variables with application to the demand for durable goods. Econometrica, 39, 829–844. Dickens, R., Machin, S., & Manning, A. (1994). The effects of minimum wages on employment: Theory and evidence from the UK. NBER Working Paper No. 4742, Cambridge, MA. DiNardo, J., Fortin, N. M., & Lemieux, T. (1996). Labor market institutions and the distribution of wages, 1973–1992. Econometrica, 64(5), 1001–1044. Dolado, J., Karamarz, F., Machin, S., Manning, A., Margolis, D., & Teulings, C. (1996). The economic impact of minimum wages in Europe. Economic Policy, 23, 319–372. Elek, P., Scharle, A´., Szabo´, B., & Szabo´, P. A. (2009). Wage-related tax evasion in Hungary (in Hungarian), Ko¨zpe´nzu¨gyi Fu¨zetek No. 23. Eo¨tvo¨s Lora´nd University, School of Social Sciences, Budapest. Erdogdu, S. (2009). Turkey – Minimum wage in tension between economic and social concerns. In D. Vaughan-Whitehead (Ed.), The minimum wage in the enlarged EU. Edward Elgar. European Commission. (2007). Undeclared Work in the European Union, Special Eurobarometer 284, European Commission. Downloadable at http://ec.europa.eu/public_ opinion/archives/ebs/ebs_284_en.pdf + ¨ si, G., & Vincze, J. (2004). A minima´lbe´r ko¨ltse´gvete´si hata´sai Halpern, L., Koren, M., Koro (The budgetary effects of the minimum wage, in Hungarian). Ko¨zgazdasa´gi Szemle, 51(4), 325–345. + J. (2003). Fighting low equilibria by doubling the minimum wage – Kertesi, G., & Ko¨llo, Hungary’s experiment. IZA Discussion Paper No. 970, IZA, Bonn. Kriz, K. A., Meriku¨ll, J., Paulus, A., & Staehr, K. (2007). Why do individuals evade payroll and income taxation in Estonia? University of Tartu – Faculty of Economics and Business Administration Working Paper Series 49, Faculty of Economics and Business Administration, University of Tartu (Estonia). Labeaga, J. M. (1999). A double-hurdle rational addiction model with heterogeneity: Estimating the demand for tobacco. Journal of Econometrics, 93, 49–72. Madzharova, B. (2010, Nov. 12). The effect of low corporate income tax on social security evasion. Cerge-EI Prague, EPRN Seminars, University of Copenhagen. Retrieved from http://web.econ.ku.dk/eprn_epru/Seminar/EPRU%20Nov12.Madzharova.pdf Martinez-Espineira, R. (2006). A Box-Cox double-hurdle model of wildlife valuation: The citizen’s perspective. Ecological Economics, 58, 192–208. Masso, J., & Krillo, K. (2009). Estonia, Latvia and Lithuania – Minimum wages in the context of migration and labour shortages. In D. Vaughan-Whitehead (Ed.), The minimum wage in the enlarged EU. Edward Elgar. Meriku¨ll, J., & Staehr, K. (2010). Unreported employment and envelope wages in midtransition: Comparing developments and causes in the Baltic countries. Comparative Economic Studies, 52(September), 637–670.

162

PE´TER ELEK ET AL.

Meyer, R. H., & Wise, D. A. (1983a). Discontinuous distributions and missing persons: The minimum wage and unemployed youth. Econometrica, 51, 1677–1698. Meyer, R. H., & Wise, D. A. (1983b). The effects of minimum wage on employment and earnings of youth. Journal of Labor Economics, 1, 66–100. Moffatt, P. G. (2005). Hurdle models of loan default. Journal of the Operational Research Society, 56, 1063–1071. Moffit, R., Fitzgerald, J., & Gottschalk, P. (1999). Sample attrition in panel data: The role of selection on observables. Annales d’E´conomie et de Statistique, 55–56, 129–152. Pashev, K. V. (2006). Presumptive taxation: Lessons from Bulgaria. Post-Communist Economies, 18(4), 399–418. Saz-Salazar, S. D., & Rausell-Ko¨ster, P. (2006). A double-hurdle model of urban green areas valuation: Dealing with zero responses. Landscape and Urban Planning, 84, 241–251. Shelkova, N. Y. (2008). Low wage labor markets and the power of suggestion. Working Paper No. 2008/33, Department of Economics, University of Connecticut. Szabo´, P. A. (2007). A 2000–2001. e´vi minima´lbe´r-emele´s hata´sa a jo¨vedelemeloszla´sra (The effect of the 2000–2001 minimum wage increase on income distribution, in Hungarian). Ko¨zgazdasa´gi Szemle, 54(May), 397–414. Tanzi, V., & Casanegra de Jantscher, M. (1987). Presumptive income taxation: Administrative, efficiency, and equity aspects. IMF Working Paper No. WP/87/54, International Monetary Fund, Washington, DC. Teklewold, H., Dadi, L., Yami, A., & Dana, N. (2006). Determinants of adoption of poultry technology: A double-hurdle approach. Livestock Research for Rural Development 18(40). Available from http://www.lrrd.org/lrrd18/3/tekl18040.htm. Retrieved on February 22, 2012. Tonin, M. (2007). Minimum wage and tax evasion: Theory and evidence. Institute of Economics, Budapest, DP 2007/1. Tonin, M. (2011). Minimum wage and tax evasion: Theory and evidence. Journal of Public Economics, 95, 1635–1651. World Bank. (2005). Enhancing job opportunities: Eastern Europe and the Former Soviet Union. Washington, DC: World Bank. Yaniv, G. (1988). Withholding and non-withheld tax evasion. Journal of Public Economics, 35, 183–204. Yen, S. T., & Jones, A. M. (1997). Household consumption of cheese, an inverse hyperbolic sine double hurdle approach with dependent errors. American Journal of Agricultural Economics, 79, 246–251.

Detecting Wage Under-Reporting Using a Double-Hurdle Model

163

APPENDIX A Table A1.

Occupational Classification Used in the Double-Hurdle Model.

Occupations

Typea

Definition (Based on Standard Classification of Occupations)

Agricultural

E

Construction Service

S S

Trade

S

Industrial

S

Codes 61–64 and 92 comprising the drivers of agricultural vehicles Code 76 Codes 52–53 except 532, 533 and 536. Includes transport, mail and telecommunication Codes 51 and 421, 422 and 429 comprising cashiers Codes 71–75

Other blue collar Cleaners Unskilled labourers Machine operators

E E E

Porters and guards

E

Drivers

S

White collar Office clerks Technicians, assistants Administrators Managers Professionals a

W W W W W

Code 911 Codes 913–919 Codes 81–83. Includes the operators of mobile machines such as cranes Codes 912 and 536 comprising porters and security guards, respectively Code 833, 835, 836 car, truck and bus. Excludes the drivers of agricultural vehicles Codes 41–42 and 532–533 comprising officebased jobs in health and social services Codes 31–34 Codes 35–39 Codes 11–14 Codes 21–29

E: elementary; S: secondary; W: white collar.

PE´TER ELEK ET AL.

164

Table A2.

Selection to the Worker Panel Used in the Test (Probit).

Dependent Variable: 1 If Made It to the Panel, 0 Otherwise Male Years in school Experience Experience squared Earned more than the MW log wage Earned the MW Firm size: 5–20 employees Firm size: 21–50 employees Firm size: 51–300 employees Firm size: 301–1,000 employees Ownership: majority domestic private Ownership: majority foreign Ownership: mixed Sales revenues per worker (log) Negative value added Micro-region unemployment rate (log) Western Hungary Northern Transdanubia Southern Transdanubia Southern Plain Northern Plain Northern Hungary Agriculture, forestry, fishing Mining Construction Trade, tourism Transport Financial services Services Education and health (private establishments) Observations LR chi2 (30), significance

Marginal Effect

Z-Value

0.013 0.000 0.009 0.000 0.013 0.099 0.110 0.102 0.138 0.003 0.048 0.013 0.021 0.008 0.009 0.414 0.027 0.064 0.059 0.024 0.091 0.114 0.131 0.154 0.013 0.030 0.051 0.032 0.164 0.035 132,115 8,473.98

6.05 1.27 25.68 19.94 5.52 3.13 28.95 32.89 65.07 1.25 19.92 4.54 4.21 8.53 0.64 5.69 7.23 16.20 12.15 5.26 19.69 25.78 24.92 6.69 3.41 11.49 8.17 8.50 4.46 6.22 0.0000

Source: Wage Survey 2006, enterprise sector. All variables relate to May 2006. Reference categories: female, more than 1,000 employees, majority state-owned, Central Hungary, manufacturing. Significant at 10%; Significant at 5%; Significant at 1%.

165

Detecting Wage Under-Reporting Using a Double-Hurdle Model

Table A3.

Selection to the Firm Panel Used in the Test (Probit).

Dependent Variable: 1 If Made It to the Panel, 0 Otherwise

Marginal Effect

Z-Value

Share of men Average years in school Average experience Average wage Share of workers affected by the 2001 MW hike Firm size: 5–20 employees Firm size: 21–50 employees Firm size: 51–300 employees Firm size: 301–1,000 employees Ownership: majority domestic private Ownership: majority foreign Ownership: mixed Sales revenues per worker (log) Negative value added Micro-region unemployment rate (log) Western Hungary Northern Transdanubia Southern Transdanubia Southern Plain Northern Plain Northern Hungary Agriculture, forestry, fishing Mining Construction Trade, tourism Transport Financial services Services Education and health (private establishments) Firms in WS 2006 Firms also observed in WS 2007

0.000 0.000 0.000 0.000 0.000 0.368 0.355 0.375 0.053 0.036 0.058 0.059 0.000 0.101 0.008 0.022 0.018 0.031 0.019 0.011 0.050 0.013 0.091 0.037 0.021 0.053 0.020 0.248 0.030 9,574 6,348

0.01 0.01 0.02 0.00 0.000 12.47 12.27 13.49 2.17 2.54 3.08 1.58 0.00 1.20 0.02 0.96 0.75 1.29 0.83 0.51 2.38 0.61 1.05 1.93 1.53 1.75 1.11 1.37 1.19

Source: Wage Survey 2006, enterprise sector. All variables relate to May 2006. Reference categories: more than 1000 employees, majority state-owned, Central Hungary, manufacturing Significant at 10%; Significant at 5%; Significant at 1%.

166

PE´TER ELEK ET AL.

APPENDIX B: MINIMUM WAGE REGULATIONS IN HUNGARY Target and coverage. A single national monthly gross minimum wage (MW) was introduced by Hungary’s last communist-led government in 1989. The MW relates to monthly pre-tax base wages, that is, total monthly earnings net of overtime pay, shift pay and bonuses. Starting from 2007 weekly, daily and hourly levels are determined, too. The MW is legally binding and covers all wages, including those paid to the self-employed by their own businesses. For part-timers, who account for about 5 per cent of total employment, the wage floor is proportionately lower. In 2006–2008, further minima applied to skilled workers (1.25 MW) and young skilled workers (1.2 MW). In 2009, the minimum for young skilled workers was abolished. MW setting. The MW is negotiated in a consultative body of employers and unions (Council of the Reconciliation of Interests). The government usually steps into the process at the end, by accepting the recommendations of the Council, but it is authorized to make a unilateral decision in case the negotiations fail, as it happened in 2001. Level of the MW. At its introduction the MW amounted to 35 per cent of the average wage (AW), while in 2000 it stood at 29 per cent. Viktor Orba´n’s first government (1998–2002) nearly doubled the MW, by raising it from Ft 25,500 in December 2000 to Ft 40,000 in January 2001 and Ft 50,000 in January 2002. The two hikes raised the MW – AW ratio to 39 per cent and 43 per cent, respectively. Since 2003, the MW/AW ratio slightly fell but remained above its pre-hike level.12 Compliance. The Wage Survey’s data suggest that sub-minimum wages accounted for less than 1 per cent of all wages in each year since 1989. Estimates based on personal income tax reports and pension contributions hint at higher rates, but these data do not allow proper adjustment for time out of work during the year. Fraction of employees affected. The fraction of workers paid 95–105 per cent of the MW amounted to 5 per cent in 2000. It jumped to 19 per cent in May 2002 in firms employing five or more workers and increased substantially in larger firms, too. The ratio fell to 10–12 per cent by 2004 and fell further substantially after 2006, when the tax authority started to interpret MW payment as a signal of wage under-reporting. Taxing the MW. In 1989–2001, the MW was subject to linear social security contribution and progressive personal income tax. In 2002, it became free of personal income tax. In 2007, a minimum social security contribution base amounting to 2 MW was introduced, as discussed in third section of the text. This measure was abandoned in 2010.

CHAPTER 5 DOES FORMAL WORK PAY? THE ROLE OF LABOR TAXATION AND SOCIAL BENEFIT DESIGN IN THE NEW EU MEMBER STATES$ Johannes Koettl and Michael Weber ABSTRACT The analysis presented in this chapter defines three different synthetic measurements of disincentives for formal work: two standard measurements, namely, the tax wedge and the marginal effective tax rate (METR); and a new, innovative measurement called formalization tax rate (FTR). The novelty of the latter is that it measures disincentives stemming not only from labor taxation but also from benefit withdrawal due to formalization. A descriptive analysis across a large number of OECD and Eastern European countries reveals that the disincentives for formal work – when measured through the FTR – are especially high for low-wage earners. This suggests that formal work might not pay in this

$

This chapter is part of background technical analysis for a forthcoming World Bank regional study ‘‘In from the Shadow: Integrating Europe’s Informal Labor’’. The views and opinions expressed in this chapter are solely those of the authors and do not represent the views and opinions of the World Bank, its board of Executive Directors, or the countries they represent.

Informal Employment in Emerging and Transition Economies Research in Labor Economics, Volume 34, 167–204 Copyright r 2012 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1108/S0147-9121(2012)0000034008

167

168

JOHANNES KOETTL AND MICHAEL WEBER

segment of the labor market, in particular for the so-called mini-jobs and midi-jobs (low-paying part-time work). Another novelty of the chapter is its empirical approach. Using EU-SILC 2008 data and OECD Tax and Benefit data for six Eastern European countries (Bulgaria, the Czech Republic, Estonia, Latvia, Poland, and Slovakia), we match disincentives for formal work to individual observations in a large data set. Applying a probit regression, the analysis finds a significant positive correlation between FTR or METR and the incidence of being informal. In other words, controlling for individual and job characteristics, the higher the FTR or the METR that individuals are facing is, the more likely they are to work informally. The tax wedge, on the other hand, yields a negative correlation. This indicates that the tax wedge is not sufficiently capturing disincentives for formal work. We also conclude that in cross-country analysis, it might be more useful to use the tax wedge that applies to low-wage earners as opposed to average wage earners. Keywords: Informal employment; measurement disincentives for formal work; tax wedge; marginal effective tax rate; formalization tax rate JEL classification: H26; J32; O17

INTRODUCTION This chapter investigates the question if – given the high levels of informality and inactivity in some European countries – it is actually ‘‘worthwhile’’ for the working-age population to engage in income-generating activities. If so, what are the incentives that employers, the self-employed, and workers have to actually register these activities and pay taxes and contributions on the income generated? There are a number of reasons why employers, the self-employed, and workers might decide not to register their activities. First, regulations in the product and labor market – like product licensing, employment protection legislation (EPL), and minimum wages – might be too stringent, so in order to circumvent these regulations, people might decide to operate outside the formal economy (Bertola, Jimeno, Marimon, & Pissarides, 2001; OECD,

Does Formal Work Pay?

169

2004; Perry et al., 2007; Schneider & Klingelmair, 2004; Terrell & Grindling, 2002). Second, certain administrative procedures related to paying taxes, accounting, and completing statistical questionnaires might deter people from operating in the formal sector (Bassanini & Duval 2006; Djankov, Lieberman, Mukherjee, & Nenova, 2002). Third, people and firms might want to avoid paying taxes on revenues, income, profit, or property and social security contributions (Betcherman & Page´s, 2007; Davis & Henrekson, 2004). Fourth, formal income might lead to a withdrawal of social benefits – like social assistance or unemployment benefits – so that people might prefer informal or no work over formal work (EC, 2003, 2004; OECD, 2004). Fifth, enforcement of existing legislation on regulations and taxation might be lax; therefore, the risks of circumventing regulations and avoiding taxes might be low (Hanousek & Palda, 2003; Schneider & Enste, 2000). To answer the main question of the chapter whether formal work pays, we focus on the role of labor taxation and social benefit design in explaining the incidence of informal employment. More specifically, it investigates what disincentives for formal work might be provided to people in working age so they choose to ‘‘exit’’ into informality (Perry et al., 2007). Since labor taxation and social benefit design only partly explain informality, the analysis presented below highlights how for lower wage earners the value of formal social security benefits that come with formal employment would at times have to be enormously high to offset the opportunity costs of formalization. The conclusion is that formal (part-time) jobs at low-wage levels – so-called mini-jobs and midi-jobs – are not economically viable for low-wage earners in some countries. This lack of economic viability might exclude a substantial part of the working-age population from formal employment and social security coverage. In this latter sense, informality and inactivity might not only be a deliberate choice of exit but are also a matter of ‘‘exclusion’’ (Perry et al., 2007). The analysis starts by examining which incentives or disincentives the informally employed and their employers face when considering formal work. For the informally employed, switching to formality will have a number of implications for both workers and firms. First, it implies that workers and their employers will enter as contributors to social security. This means that both the employer and the worker have to contribute to pension funds, health insurance funds, and unemployment insurance funds. The decision on contributions will strongly be influenced by the value that informal workers attach to being covered by social security. Second, workers will have to pay personal income tax on their formal gross wages. This decision will be influenced by the value informal workers put on public

170

JOHANNES KOETTL AND MICHAEL WEBER

services and social norms about paying taxes (Torgler, 2003; Torgler & Schneider, 2009). Paying social security contributions and income tax combined will decrease workers’ take-home pay when compared to their informal wage. Third, informal workers after formalizing might no longer be eligible to a number of income-tested benefits such as social assistance, family benefits, or housing benefits. Firms, finally, which are formalizing informal workers, will have to generate additional formal revenues by switching informal revenues to formal revenues. This implies paying additional taxes in the product market, like sales or value-added taxes. In the New Eastern European Member States of the European Union (NMS) informality is high,1 and the demographic transition will considerably increase the need for participation in the formal sector in the future. For the social contract of these countries to survive, more people will need to contribute through taxes and social security contributions. Those who currently do not work and those who work informally will need to be activated and convinced to participate in the formal sector of the labor market. Arguably, one precondition for participation in the formal sector is that formal work has to pay. In other words, the incentives for formal work that originate in the tax and benefit system of a country have to be aligned to encourage formal work. Incentives alone, though, might perhaps not be the binding constraint. Labor taxation and benefit design are but two pieces of the puzzle that explain high levels of informality among the working-age population of the NMS. We are not trying to identify which of these potential reasons are the main causes for high levels of informality, but we narrowly focus on the incentives and disincentives for formal employment provided by the labor taxation and benefit system. For a very recent discussion on drivers of informal employment and potential policy options to address them, see Lehmann (2010).2 One indication that at least taxation in general plays a prominent role in income-generating activities comes from enterprise surveys. For example, the World Bank Enterprise survey for 2009 (World Bank, 2011b) reveals that on average, 45 percent of firms in the NMS cite tax rates as a one of the major obstacles for doing business. The question in the survey refers to all types of taxes, and, hence, not specifically to labor taxes. Nevertheless, the results indicate that in the NMS, tax rates are perceived as a greater obstacle to doing business than any other impediment in the ranking, like tax administration (24 percent), competition from informal enterprises (23 percent), licensing (14 percent), labor regulations (14 percent), and trade regulations (8 percent). Therefore, although the results of this enterprise survey are not a direct

Does Formal Work Pay?

171

measurement of obstacles to formal employment, they give an indication that tax rates could be perceived as a constraint for creating new formal jobs. The contribution of this chapter is twofold: First, we present a detailed descriptive analysis of the tax and benefit systems of the NMS, describing disincentives for formal employment stemming from labor taxation and social benefit design. In doing so, we not only analyze standard measurements of disincentives like the tax wedge and the marginal effective tax rate (METR) but also develop a new and innovative measurement of disincentives for formal work, the so-called formalization tax rate (FTR). The FTR goes beyond the usual measurements of the tax wedge and METR by combining both. It expresses the opportunity costs of formal employment by measuring what share of informal income is being taxed away – in terms of income tax, social security contributions, and withdrawn benefits – when formalizing, and therefore how much workers have to gain in return for formalization in terms of social security benefits and employment protection. For this exercise, the analysis relies on the Organization of Economic Co-operation and Development (OECD) Tax and Benefit model, which is available for most OECD countries as well as the Baltics, Bosnia and Herzegovina, Bulgaria, Macedonia, Romania, and Serbia (Immervoll, 2007; OECD, 2011).3 This descriptive analysis is presented in the second section. Second, we present an empirical analysis that combines the synthetic measurements of the FTR, the METR, and the tax wedge with actual informality patterns by using data from the European Union Statistics on Income and Living Conditions (EU-SILC; Eurostat, 2008), investigating the question if and how much these disincentives matter for informal employment. To this end, the analysis matches these synthetic measurements to individual observations for six countries in the EU-SILC 2008 data: Bulgaria, the Czech Republic, Estonia, Latvia, Poland, and Slovakia. In other words, for each individual in the survey, we define synthetic measurements of disincentives for formal work stemming from the tax and benefit system of his or her country of residence. The empirical analysis then explores the correlation between disincentives for formal work and the incidence of informal employment in these six countries. This analysis is presented in the third section. The main finding of the descriptive analysis is that the disincentives for formal work – when measured through the FTR – are especially high for low-wage earners. This suggests that formal work might not pay in this segment of the labor market, in particular for the so-called mini-jobs and

172

JOHANNES KOETTL AND MICHAEL WEBER

midi-jobs (low-paying part-time work). The empirical analysis finds a significant positive correlation between FTR or METR and the incidence of being informal. Controlling for individual and job characteristics, the higher the FTR or the METR that individuals are facing is – that is, the higher disincentives for formal work – the higher the likelihood to work informally. The tax wedge, on the other hand, yields a negative correlation, suggesting that the tax wedge is not sufficiently capturing disincentives for formal work.

MEASUREMENTS OF INFORMALITY AND DISINCENTIVES FOR FORMAL WORK This section offers a descriptive analysis of the tax and benefit systems in the NMS and benchmarks them against other OECD countries. The section starts by discussing various definitions of informal employment. Next, the theoretical foundations of the decision between formal and informal employment are outlined. Finally, three different measurements of work disincentives are introduced and discussed. Definition of Informal Employment The definition of informal employment is not straightforward. A comprehensive and widely accepted definition was provided by the International Labor Organization (ILO) in 2003. It includes (i) own-account workers and employers employed in their own informal sector enterprises; (ii) unpaid family workers, irrespective of whether they work in formal or informal sector enterprises; (iii) members of informal producers’ cooperatives; (iv) own-account workers engaged in the production of goods exclusively for own final use by their household; and, finally, (v) employees holding informal jobs in formal sector enterprises, informal sector enterprises, or as paid domestic workers employed by households. In this last category, informal jobs are those not subject to legislation, income taxation, social protection, nor entitlements to codified benefits such as advance notice of dismissal, severance pay, paid annual, or sick leave (Hussmanns, 2004; ILO, 2003). This comprehensive definition of informality is constrained by data availability. Particular measurement challenges are (i) measuring the status of self-employed, own-account workers, and employers, as this requires

Does Formal Work Pay?

173

identifying if their enterprise is registered or not and (ii) measuring the status of employees, as this requires identifying the nature of their employment contract. With regard to the self-employed, if it is not possible to determine the legal status of the enterprise, a variety of alternative measures have been developed. The most widely used is the size of the enterprise, with employers or self-employed of enterprises with less than 5 or 10 employees defined as informal, all others as formal. Another measurement developed by Hazans (2011) qualifies the firm size measurement by combining it with the professional status of self-employed or employers – only nonprofessional employers or self-employed with a small enterprise are defined as informal. With regard to employees, the most common measurement defines those with no written employment contract as informal. Alternatively, informal employment is sometimes also measured as those employees who do not contribute to social security. If neither measurement is available, the firm size is again the most common measurement, defining all employees in firms with less than 5 or 10 employees as informal. Unpaid family workers, finally, are relatively easy to measure directly from surveys. In this chapter, two related definitions of informal employment are used. For the descriptive analysis of this section, we use the legalistic definition: informally employed are all those who do not report any of their labor income to tax and social security authorities. These are all self-employed, employers, own-account workers, unpaid family workers, and employees who completely conceal their labor income and pay neither social security contributions nor personal income taxes. The second section uses only data from the OECD Tax and Benefit model (OECD, 2011) and analyzes synthetic measurements. These synthetic measurements are exclusively derived from legislation – that is, they measure what individuals should pay in taxes and what benefits individuals should to be entitled to according to the law; it does not measure the actual incidence of tax payments and entitlement consumption. For the econometric analysis of third section, survey data from EU-SILC (Eurostat, 2008) is used. Data limitations require deviating from the preferred legalistic definition of informal employment with regard to the self-employed. Informal employees, as previously defined, are as those for whom no social security contributions are being paid by their employers. Also, unpaid family workers are identified as informal. However, for the self-employed, the definition deviates from the legalistic definition of the descriptive analysis. Because of data restrictions, a less precise ‘‘productivity’’ definition as suggested by Hazans (2011) is applied. Accordingly, all

174

JOHANNES KOETTL AND MICHAEL WEBER

nonprofessional employers who employ five or fewer workers (including those with no employees) are identified as informal. Applying this productivity definition to the informal self-employed yields rather high informality rates among the self-employed (see below). This productivity definition could therefore potentially overestimate informality rates for this group. As a robustness check, the analysis is performed restricting the sample to either only the self-employed or only employees. The results do not differ substantially from those for the whole sample. Hence, the behavior of the self-employed, using the productivity definition, corresponds to the behavior of employees with regard to work disincentives and informal employment. This suggests that the productivity definition used to identify informal self-employed does not conflict with the results using the legalistic definition for employees.

The Comparison Between Formal and Informal Employment This chapter attempts a comparison between two theoretically identical workers: one working formally and one informally. The formal worker faces certain costs for working formally which are measured by the overall and marginal taxes that need to be paid. The informal worker does not have to pay taxes. So, how much the informal worker’s theoretical formal twin has to pay in overall taxes (the tax wedge) or in marginal taxes (the METR) are important indicators for formal work disincentives. Similarly, the difference in net income of the informal worker and the informal worker’s formal twin is a comprehensive indicator for how much income is taxed away through formalization (the FTR). Each of these three measurements centers on the individual worker. Therefore, the focus of the analysis is primarily on the worker’s choice between formal and informal employment. The key question is how to construct two identical workers. That is, what income would an informal worker have in the formal sector, and what income would a formal worker have in the informal sector? We assume that two workers are identical if they generate the same total labor cost for the employer. This approach has some important implications. In particular, it implies that the entire tax wedge is born by the formal worker and therefore enters in its entirety as an opportunity cost of formal work. Such an interpretation assumes that the employer has a very strong bargaining position and is able to roll over all taxes to the worker. In other words, the labor supply curve is infinitely inelastic. Empirical evidence suggests that labor supply could indeed be relatively inelastic in

Does Formal Work Pay?

175

the low-wage segment, but not so much in the higher wage segment (see, e.g., Betcherman & Page´s, 2007). This could suggest that in the higher wage segment, the empirical analysis might be less adequate. The suggested methodology might make it difficult to interpret the decision process between formal and informal employment. We model this primarily as the worker’s decision, which in turn depends on the financial disincentives provided by the tax and benefit system. In reality, the decision will clearly also build upon preferences – in particular time preferences – about the value of social security benefits, employment protection, and risk. Yet, for the analysis these preferences cannot be measured with the available data, which means that we assume that the marginal product of labor is equal among the two identical workers. This further implies that the decision between formal and informal employment can only take place within the same firm, if one further assumes that the marginal product of labor of the worker is firm specific. In that sense, the informality decision as modeled in this chapter would only apply to the limited case where workers are formalized within the same firm. If, on the other hand, one assumes that the marginal product of labor is not firm specific, but tied to the worker’s abilities, the informal worker could also move to other firms when formalizing. This position implies a more competitive labor market in which workers are able to move relatively freely between jobs and, therefore, are confronted with an integrated labor market (Maloney, 1999, 2004). This assumption also implies that there is no wage gap between the formal and informal sector. Empirical evidence suggests that the latter might not be the case, at least not for employees. Perry et al. (2007) find that there is a considerable wage gap between formal and informal salaried workers in Latin American countries and a considerable segmentation in the labor market. Despite this imprecision and difficulties in interpretation, we consider the measurement concepts put forward as sufficiently adequate to allow for a meaningful analysis. The central question then is how to measure these opportunity costs or disincentives of formal work. As mentioned above, we investigate three measurements: first, the tax wedge; second, a new indicator, called the FTR; and third, a marginal measurement, the METR.

Labor Taxation: The Tax Wedge Labor taxes in the NMS are high at lower wage levels. A comparison with other EU, OECD, and neighboring countries shows that the tax wedge on labor at lower wage levels (33 percent of a country’s average wage) tends to

176

JOHANNES KOETTL AND MICHAEL WEBER

be relatively high (see Fig. 1).4 The tax wedge is a so-called synthetic measurement, meaning it is purely based on legislation and therefore measures what individuals are supposed to pay, not what they actually pay, in taxes and social security contributions. It measures the difference between total labor costs and take-home pay as a percentage of total labor costs. This difference consists of personal income tax and social security contributions paid by the worker and the employer. Hence, the tax wedge t is defined as t¼

TLCE  NIW ITW þ SSCW þ SSCE ¼ TLCE TLCE

where TLCE is the total labor costs paid by the employer, NIW is the net income of the worker, ITW is the income tax paid by the worker, SSCW is the social security contribution paid by the worker, and SSCE is the social security contribution paid by the employer. It therefore expresses the costs of social security contributions by employers and employees and the personal income tax of employees as a share of total labor costs, or the share of total labor costs that is ‘‘taxed away.’’

Ireland

Switzerland

United Kingom

Japan

United States

Spain

Norway

Portugal

Macedonia

Bosnia - RepublikaSrpska

Italy

Slovak Republic

France

Netherlands

Austria

Slovenia

Estonia

Latvia

Greece

Poland

Bulgaria

Finland

Czech Republic

Belgium

Lithuania

Serbia

Germany

Romania

Bosnia - Federation

Sweden

Hungary

50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0%

Fig. 1. Labor Taxation Tends to be Relatively High for Low-Wage Earners (at 33 Percent of Average Wage) in the NMS. Note: Columns represent the tax wedge for low-income earners (singles with no children at 33 percent of average wage) in 2008 (for Bosnia, Macedonia, and Serbia, 2009). NMS depicted in black. Source: Authors’ calculation based on OECD Tax and Benefit model (OECD, 2011).

Does Formal Work Pay?

177

In the case of informal employment, neither the worker nor the employer has to pay any taxes or social security contributions so that the worker’s net income equals total labor costs and the tax wedge is zero. The tax wedge is therefore a good first approximation to measure incentives faced by employers and workers for informal employment. It measures the gains of informal employment. Although it cannot give any information on how these gains can potentially be shared between the worker and the employer, it nevertheless gives valuable information on the incentives that workers and employers face when choosing between formal or informal employment. The tax wedge varies depending on family type and wage level. The OECD Tax and Benefit model calculates (i) income tax and social security contributions paid by the employer and the worker, (ii) for 10 different family types, (iii) for individuals earning between 0 and up to 367 percent of average wage, and (iv) for most OECD countries. This data reveals that for single persons with no children receiving a gross wage of 33 percent of the average wage, only a few EU-15 countries – like Sweden, Germany, Belgium, and Finland – charge higher taxes than most of the NMS. Also, labor taxation in the NMS is not very progressive. While in most other countries, labor taxes increase significantly with the wage level in the NMS, labor taxes increase by less than 10 percentage points. For most EU-15 countries, taxes increase by over 10 percentage points between 33 and 100 percent of average wage level. Although countries with a high tax wedge at lower wage levels can be expected to display less progressivity, the NMS display especially low levels of progressivity. All NMS except for Hungary and Slovenia are below the trend line in a cross-county comparison (see Fig. 2).5 In particular, for single workers without children, Bulgaria stands out with zero progressivity of labor taxes. This is important because low progressivity means that there is some room for lowering the tax wedge for low-wage earners in a fiscally neutral way by increasing progressivity. Nevertheless, with the exception of Bulgaria in all countries, labor taxation displays some degree of progressivity.6 A typical graph of the tax wedge over the wage level for the NMS is depicted in Fig. 3 – in this case, for Estonia and Latvia.7 As can be seen, the tax wedge is lower for low-wage earners (around 26 percent for Estonia and Latvia) and starts to significantly increase from a certain wage level onward (around 20 percent of average wage) to levels of about 40 to 45 percent of total labor costs. What is interesting, though, is that some countries display much lower tax wedges for low-wage earners, as in the case of Australia and the United Kingdom (see Fig. 3).8 Both have a tax wedge of 0 percent for low-wage earners, and only for wage levels above 20 percent of average wage the tax wedge is increasing significantly.

Progrssivity of tax wedge (percetnage points)

178

JOHANNES KOETTL AND MICHAEL WEBER 25

20

Belgium

Ireland France Switzerland

15

Italy

10

Austria Germany

Hungary

Netherlands

United Kingom

Norway United States

5 Japan

Slovenia Spain Finland Portugal Slovak RepublicCzech Republic Greece Latvia Sweden Estonia Lithuania Macedonia Poland Romania Bosnia - Federation Bosnia - Republika Srpska Serbia

0

Bulgaria

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

Tax wedge for single at 33 percent of averge wage

50%

45%

45%

40%

40%

35%

35%

30%

30%

25%

25%

20%

20%

15%

15%

10%

10%

5%

5%

0%

0% 1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161 171 181 191

50%

Percent of average wage

1 10 19 28 37 46 55 64 73 82 91 100 109 118 127 136 145 154 163 172 181 190 199

Fig. 2. In the NMS, Labor Taxation Tends to be Not Very Progressive. Note: Data points represent the tax wedge for low-income earners (single person with no children at 33 percent of average wage; x-axis) in relation to a country’s progressivity of the tax wedge (the percentage point increase of the tax wedge between 33 and 100 percent of average wage; y-axis) in 2008 (for Bosnia, Macedonia, and Serbia, 2009). Source: Authors’ calculation based on OECD Tax and Benefit model (OECD, 2011).

Percent of average wage

Tax wedge (Estonia)

Minmum wage (Estonia)

Tax wedge (Australia)

Minimum wage (Australia)

Tax wedge (Latvia)

Minmum wage (Latvia)

Tax wedge (UK)

Minimum wage (UK)

Fig. 3. In Estonia and Latvia, the Tax Wedge for Low-Wage Earners Is Higher Than in Australia or the United Kingdom. Note: Graphs show the tax wedge for single person with no children. Countries have been chosen to illustrate contrasting examples. Source: Authors’ calculation based on OECD Tax and Benefit model (OECD, 2011).

179

40

Tax wedge at 1 percent of average wage

Tax wedge at 1 percent of average wage

Does Formal Work Pay?

35 30 25 20 15 10 5 0 0

10

20

30

40

50

60

70

Wage level (percent of average wage) where tax wedge starts to increase significantly

40 35 30 25 20 15 10 5 0 0

20

40

60

80

100

120

140

Wage level (percent of average wage) where tax wedge starts to increase significantly

Fig. 4. In the NMS, the Tax Wedge for the Lowest Wage Earners Tends to be High, and the Wage Level Where the Tax Wedge Increases Significantly Is Relatively High. Note: The scatter plot depicts the wage level where the tax wedge starts to increase (x-axis) versus the tax wedge at 1 percent of average wages (y-axis). Hungary, the Netherlands, and Serbia feature falling tax wedges at low-wage levels and are not depicted, just like Bulgaria that has a flat tax wedge. Austria, Belgium, and Canada have partly negative tax wedges at low-wage levels, especially for families, and are also not included in the right scatter plot (Canada also in the left). Source: Authors’ calculation based on OECD Tax and Benefit model (OECD, 2011).

A closer look at Fig. 4 reveals that in the NMS, the tax wedge tends to be high for a relatively large spectrum of low-wage earners. The wage level from where onward the tax wedge starts to increase significantly is also relatively high. In many high-income OECD countries, to the contrary, the tax wedge is low for the lowest wage earners, but the tax wedge also tends to increase across the whole wage spectrum.

Social Benefits Aside from the tax wedge, the withdrawal of social benefits is the main contributor to the opportunity costs of taking up formal work for individuals with low skills/earnings potential. Consider an informal worker who earns a certain level of informal wage.9 If this worker were to work in the formal sector, various implicit opportunity costs occur to the worker and the worker’s employer. First, assuming that the marginal labor product does not change because of formalization, total labor costs of the informal worker have to be the same as for the formalized worker. For the informal

180

JOHANNES KOETTL AND MICHAEL WEBER

worker, total labor costs are the informal wage. For the formalized worker, total labor costs are the net wage plus the income tax and both the worker’s and the employer’s social security contributions. That is, the net wage plus the entire tax wedge. Comparing the informal wage with the worker’s potential formal net wage, the entire tax wedge enters as an opportunity cost of formal work for the informal worker and the worker’s employer. Second, informal workers also face implicit opportunity costs because they might lose parts of certain income-tested benefits – most importantly social assistance, housing benefits, and family benefits – once they have a formal income on record. For example, if an informal worker receives a certain amount of social assistance, this benefit will be decreased or completely withdrawn if the worker formalizes. This amount of the withdrawn benefit also enters as an opportunity cost of formal work. Therefore, both of these losses – the tax wedge and withdrawn benefits – have to be taken into account when considering the implicit opportunity costs of formalization. At the same time, though, informal workers gain from formalization: they gain a future right to an old-age pension and they gain immediate rights with regard to disability insurance, workers compensation, health insurance, and unemployment insurance.10 Arguably, the most important of these potential gains are old-age pension and health insurance. With regard to old-age pensions, though, one has to keep in mind that especially low-wage earners tend to discount future benefits more because their concerns are focused on short-term income, and in cases of poverty, day-to-day consumption.11 Also, any means-tested social pensions for the elderly might further discount the value of a vested old-age pension. In addition, especially in developing and transition countries, workers might discount the value of pension benefits because of a lack of trust in social security contributions.12 Finally, the value attached to the most important benefit, health insurance, could be low because in many countries it can be accessed for free. Nevertheless, there is no reliable concept of measurement available at this point that can properly quantify the value that workers attach to these formal benefits – may it be short-term or long-term social security benefits or other benefits like employment protection. In the analysis presented below, this shortcoming has to be kept in mind. Regarding the potential role of benefit withdrawal, Tables 1 and 2 provide information on the coverage and generosity of the social benefits considered for some of the countries analyzed in the third section. Overall the family benefits seem to be the most important type of benefit, reaching between 21 and 47 percent of the total population. Social assistance and housing benefits seem less important for the population at large but do reach

181

Does Formal Work Pay?

Table 1.

Coveragea of Social Benefit Programs (Percent, Various Years). Social Assistance

Bulgaria 2007 Estonia 2004b Latvia 2009b Poland 2008

Family Benefits

Housing Benefits

Total population

Poorest quintile

Total population

Poorest quintile

Total population

Poorest quintile

3.8 2.0 2.7 4.4

14.5 6.7 5.9 14.2

20.8 47.2 43.9 21.3

35.3 57.5 60.7 51.2

4.1 — — 3.8

13.8 — — 9.4

Source: World Bank (2011a). a Coverage indicates the percentage of individuals who receive social assistance transfers. b Social assistance includes housing benefit.

Table 2.

Generositya of Social Benefit Programs (Percent, Various Years). Social Assistance

Bulgaria 2007 Estonia 2004b Latvia 2009b Poland 2008

Family Benefits

Housing Benefits

Total population

Poorest quintile

Total population

Poorest quintile

Total population

Poorest quintile

21.5 26.3 6.7 7.7

33.8 39.9 12.1 9.6

3.6 7.0 2.5 3.1

7.6 18.0 7.7 5.2

5.8 — — 3.4

8.0 — — 4.4

Source: World Bank (2011a). a Generosity indicates the transfer as a share (percent) of total consumption of beneficiary households. b Social assistance includes housing benefit.

between 6 and 15 percent of the poorest income quintile. For these poorest households, social assistance benefits finance between 10 (Poland) and 40 (Estonia) percent of total household consumption.

The Formalization Tax Rate As discussed above, the implicit costs of formalization for informal workers are a measurement of the necessary minimum value of social security

182

JOHANNES KOETTL AND MICHAEL WEBER

benefits and employment protection they receive in return for formalization. The value of rights to pension and unemployment insurance – but also from formal EPL – they gain from formalization must exceed their implicit opportunity costs from formalization. Fig. 5 depicts this implicit cost to the informal worker as a share of informal income, the so-called FTR. The FTR measures the difference between informal income (informal wage, social assistance, and family and housing benefits at the level of no formal wage) and formal net income (formal net wage, in-work benefits, social assistance, and family and housing benefits at formal wage level) as a share of informal income.13 Just like the tax wedge, it is a synthetic measurement that is based on legislation on taxes and social benefits, measuring the difference in net income between an informal worker (informal wage and social benefits) and the theoretical income of the same informal worker if he or she were to formalize. More precisely, the FTR tF(w) is defined as tF ðwÞ ¼

NII ðwÞ  NIF ðwÞ NII ðwÞ

80%

70%

70%

60%

60%

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0% 1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161 171 181 191

80%

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161 171 181 191

where NII(w) is the net income of an informal worker and NIF(w) is the income of a formal worker at a certain (gross) wage level w. The net income of an informal worker is the wage the worker receives plus any benefits the worker is entitled to. Since the benefit level depends on the

Percent of average wage

Percent of average wage

FTR (Bulgaria)

Minmum wage (Bulgaria)

FTR (Australia)

Minimum wage (Australia)

FTR (Romania)

Minimum wage (Romania)

FTR (US)

Minimum wage (US)

Fig. 5. For Low-Wage Earners, the Opportunity Costs of Formal Work (Formalization Tax Rate, FTR) Are Higher in Bulgaria and Latvia Than in Australia and the United States. Note: Graphs show the FTR for single person with no children. Countries have been chosen to illustrate contrasting examples. Source: Authors’ calculation based on OECD Tax and Benefit model (OECD, 2011).

183

Does Formal Work Pay?

level of formal wage, an informal worker can claim benefits at formal wage level w ¼ 0, or NII ðwÞ ¼ w þ Bð0Þ where B(w) is the benefit function. The question that arises is what the net income of a comparable formal worker at the same wage level would look like. Note that an employer will not consider two workers to be equal based on their gross wage, but based on total labor costs; in other words, two workers are equal if their marginal product of labor is equal. Hence, it will be useful to compare an informal worker with a formal worker whose total labor costs (and not gross wage) equal the informal worker’s wage. This means that the net income of a comparable worker in the formal sector not only has to be net of income tax and social security contributions paid by the worker but also by the employer. Therefore, the net income of a comparable formal worker is NIF ðwÞ ¼ w  ITðwÞ  SSCW ðwÞ  SSCE ðwÞ þ BðwÞ where IT(w) is the income tax, SSCW(w) is the social security contribution paid by the worker, SSCE(w) is the employer’s social security contribution, and B(w) is the benefits claimed by the worker, all at formal gross wage level w. Comparing these two workers described above leads to the following interpretation. A worker who earns a certain wage in the informal sector is discussing with his or her employer to formalize. Since the worker does not have any formal income, the worker also claims benefits as if he or she had no income. The employer and the worker now consider how much the worker would earn, net of income tax, and social security contributions in the formal sector. They also take into account any changes in benefits the worker could claim given his or her formal wage. The FTR measures how much of the informal income would be taxed away, through taxes, social security contributions, and changes in benefits if the worker were formalized. How does the FTR look in different countries? In order to calculate and illustrate the FTR, the OECD Tax and Benefit model also provides all the necessary information on social assistance, housing, family, and in-work benefits. Consider the contrasting examples of Bulgaria and Romania on the one hand and Australia and the United States on the other (see Fig. 5). For lower wage levels, the FTR in Bulgaria and Romania is higher than in Australia and the United States. In Bulgaria, the FTR for a single person

184

JOHANNES KOETTL AND MICHAEL WEBER

with no children peaks at around 70 percent (around 60 percent for Romania) at a wage level of about 10 percent of average wage. This means that in Bulgaria, a single person with no children who earns less than the minimum wage in the informal sector has to give up between 50 and 70 percent of income to formalize. By contrast, in Australia and the United States, the FTR peaks at a lower value – around 40 percent in Australia and 30 percent in the United States – and at a higher wage level of around 30–40 percent (although in the case of the United States, the FTR continuous to increase at higher wage levels, yet at a slow rate). A more comprehensive comparison shows that in the NMS the opportunity costs of formal work tend to peak at lower wage levels than in highincome OECD countries. Fig. 6 reveals that both for single persons and one-earner couples with two children, the costs of formalization in the NMS generally tend to be highest for low-wage earners (less than 30 percent of average wage for singles). In some countries, like Bulgaria, Hungary, and Romania, the FTR for singles is particularly high and peaks at around 70 percent. For families, the FTR tends to be lower and peaks at somewhat higher wage levels.

65

90

60

80

Peak value of FTR

Peak value of FTR

55 70

60

50

50 45 40

40

35

30

30 0

10

20

30

40

50

60

Wage level (percent of average wage) where FTR peaks

0

50

100

150

200

Wage level (percent of average wage) where FTR peaks

Fig. 6. In the NMS, the Opportunity Costs of Formal Work Tend to be Highest at Lower Wage Levels. Note: The scatter plot depicts the wage level where the formalization tax rate (FTR) peaks (x-axis) versus the peak value of the FTR (yaxis). Countries with a continuiously and significantly increasing FTR were omitted. In countries where the FTR froms a plateau and increases only slightly with the wage level, the lowest wage level at which the FTR stops to increase significantly was chosen as the peak. Source: Authors’ calculation based on OECD Tax and Benefit model (OECD, 2011).

Does Formal Work Pay?

185

The main reasons for the high opportunity costs of formal work are labor taxation and the sudden withdrawal of social assistance and family benefits at higher wage levels. Labor taxation has already been discussed above as one of the potential obstacles to formal employment at the lower wage levels. The design of income-tested benefits also plays an important role; social assistance is often paid out as a top-up to earned gross income to guarantee a minimum gross income. Any earned household gross income is subtracted from social assistance that is paid out. This means that any formal mini-job at low-wage levels does not pay. Likewise, for higher-paid midi-jobs, the net gain in income might not be very high because of the sudden loss of social assistance. A more phased-in withdrawal of social assistance through (formal) income disregards for all household members could decrease this disincentive. Income-tested family and housing benefits also contribute to the FTR if the formal income would exceed the threshold for eligibility.

The Marginal Effective Tax Rate The METR also suggests that formal work does not pay at lower wage levels. The METR measures at a given wage levels how much of an additional dollar earned in formal gross wage is taxed away, either as labor tax or in the form of withdrawn benefits. It is therefore an indication of how much it pays for workers to earn more gross income, either by increasing work hours or receiving higher wages. Just like the tax wedge and FTR, also the METR is a synthetic measurement of disincentives for formal work, based on legislation – what people ought to pay in taxes and are entitled to – and not on actual income, tax, and benefit consumption data. In many countries, at low-wage levels (below 10 percent of average wage), every dollar earned is subtracted from entitlements to social assistance; hence a 100 percent of any additional dollar earned is taxed away. For example, in the Czech Republic and Slovenia, every additional dollar earned in formal income is a 100 percent taxed away through withdrawal of social assistance at wage levels below 20 percent of average wage (see Fig. 7). In other countries, like Portugal and the United States, incentives for formal work are better for low-wage earners in these countries. In Portugal, only 50 percent of every additional earned dollar is taxed away and in the United States it is significantly less. In the United States, this is mainly achieved through so-called in-work benefits and tax credits that subsidize work at low-wage levels.

186

JOHANNES KOETTL AND MICHAEL WEBER 1.20

1.00

1.00

0.80

0.80

0.60

0.60

0.40

0.40

0.20

0.20

0.00

0.00 1 10 19 28 37 46 55 64 73 82 91 100 109 118 127 136 145 154 163 172 181 190 199

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161 171 181 191

1.20

Percent of average wage

Percent of average wage

METR (Czech Republic)

Minmum wage (Czech Republic)

METR (Portugal)

Minmum wage (Portugal)

METR (Slovenia)

Minmum wage (Slovenia)

METR (United States)

Minimum wage (US)

Fig. 7. For Low-Wage Earners, the Marginal Effective Tax Rate (METR) Is at 100 Percent in the Czech Republic and Slovenia, While It Is Much Lower in Portugal and the United States. Note: Graphs show the METR for single with no children. Countries have been chosen to illustrate contrasting examples. Source: Authors’ calculation based on OECD Tax and Benefit model (OECD, 2011). 140 METR at 5 percent of average wage

METR at 5 percent of average wage

140

120

100

80

60

40

20

120

100

80

60

40

20 0

10

20

30

40

50

Wage level (percent of average wage) where METR drops significantly

0

20

40

60

80

100

120

Wage level (percent of average wage) where METR drops significantly

Fig. 8. In the NMS, the Marginal Effective Tax Rate (METR) Tends to be High at Low-Wage Levels, but Also Tends to Drop Significantly at Lower Wage Levels than in High-Income OECD Countries. Note: The scatter plot depicts the wage level where the METR drops significantly (x-axis) versus the value of the METR at 5 percent of average wage (y-axis). Countries with a METR that increases with the wage level even at lowest wage levels were omitted (Greece, Hungary, Italy, and the United States). Source: Authors’ calculation based on OECD Tax and Benefit model (OECD, 2011).

Overall, the NMS tend to have high METRs – usually at 100 percent – at low-wage levels, although the METR tends to drop at lower wage levels than in high-income OECD countries (Fig. 8). A notable exception is Poland, which according to the OECD Tax and Benefit model has the

187

Does Formal Work Pay?

lowest METR at low-wage levels of all countries. This is due to an apparent lack of a comprehensive, federally administered social assistance program. Nevertheless, Poland might have a locally administered social assistance program. If this is indeed the case, it is unfortunately not captured by the OECD Tax and Benefit model. So, the actual METR could be higher than predicted.14

DO DISINCENTIVES FOR FORMAL WORK MATTER? This section performs an empirical analysis at the individual level that investigates how incentives correlate with informality, controlling for individual and job characteristics. Data The analysis is based on the EU-SILC for the year 2008. The survey covers a wide range of European countries and includes detailed questions on employment, income, taxes, and social security contributions. This allows to apply a comprehensive definition of informal workers and selfemployed for a number of countries. In particular, the survey includes a question on the amount of social security contributions paid by the employer on behalf of the worker. This question allows the identification of informal workers as those dependent employees for whom no social security contributions are paid. In addition, unpaid family workers are identified as informal. Finally, nonprofessional employers who employ five or fewer workers (including those with no employees) are also identified as informal.15 The analysis is performed for six NMS – Bulgaria, the Czech Republic, Estonia, Latvia, Poland, and Slovakia – which yields a total sample size of 48,865 employed individuals. The dependent variable is a binary indicator that takes the value of one for informal workers and zero for formal ones. The independent variables for the regression are age, gender, education, geography (degree of urbanization), employment status of the spouse, citizenship, income, and sector. The age variable is grouped into five categories: age 15–24, 25–39, 40–54, 55–64, and 65 or older. Similarly, the education variable is grouped into three categories: high (postsecondary or tertiary education), medium (secondary education), or low (primary or preprimary education).16 For geography, the three categories are densely,

188

JOHANNES KOETTL AND MICHAEL WEBER

intermediate, or sparsely populated area. The employment status of the spouse has four categories, namely formally employed, informally employed, inactive, or no spouse. Citizenship can either be the same as the country of residence (local), or of another EU country, or a non-EU country. The sector variable follows the NACE standard.17 Income groups are categorized based on income as a percentage of average wage of the country of residence. That is, income (employee and self-employment cash or near cash income) is calculated as a share of the official average wage. The average wage data is taken from OECD (2011). Note that for unpaid family workers, income is 0, while for some selfemployed and employers, it can also be negative (in the case of a loss from self-employment or business activities). Income groups are then defined as those earning (i) 0 or less; (ii) more than 0 but less than 25 percent of average wage; (iii) 25 percent or more, but less than 50 percent of average wage; (iv) 50 percent or more, but less than 100 percent of average wage; (v) 100 percent or more, but less than 200 percent of average wage; and (vi) 200 percent of average wage or more. The main innovation of the analysis stems from the attempt to measure incentives and disincentives for formal work that are being provided by the tax and benefit system at the individual level; that is, the FTR, METR, and the tax wedge are defined for each individual in the sample. This yields a synthetic measurement – purely based on de jure tax obligations and entitlements – of incentives and disincentives for formal work at the individual level. To this end, we use the OECD Tax and Benefit model (OECD, 2011) for the year 2008 for the six countries in the sample. The OECD Tax and Benefit model already provides the METR, and the FTR is calculated using the same model and according to the methodology developed by Koettl (2009) and described above. Both FTR and METR depend on three variables: (i) individual income, expressed as percent of average wage; (ii) household type (single or family); and (iii) the income of the spouse, if applicable. First, individual income as percent of average wage is calculated as outlined above, expressing the individual’s cash or near cash income from dependent work and self-employment as a percent of average wage. The model is limited to the extent that the OECD Tax and Benefit model only provides calculations up to a certain level of income – for individuals, up to 200 percent of average wage, for certain types of families up to 367 percent of average wage. Since FTR and METR vary mainly at lower wage levels and are fairly constant from a certain income level onward, we assume that individuals with income above the limitations set by the OECD Tax and Benefit model face the same incentives as those individuals at the boundary.18

189

Does Formal Work Pay?

Second, the OECD Tax and Benefit model is provided for 10 household types, from singles with or without children to one- and two-earner couples with or without children. For the latter type, the model is provided for three different income levels for the spouse.19 These 10 OECD household types are matched to the household types provided in the EU-SILC data set. Certain assumptions have to be made in doing so. For example, the number of children is not taken into account, all individuals with children are assumed to face the same incentives as those with two children. In households with children, a couple, and additional adults, the children are matched to the couple while the additional adults are assumed to be singles. Finally, in households with children, but not couples, children are matched to those singles in a certain age group (35–45). Third, for individuals with a spouse working in the formal sector, the spouse’s income also has to be taken into account. The OECD Tax and Benefit model does so for three income levels of the spouse: 67, 100, and 167 percent of average wage. The spouse’s income is then matched to 67 percent of average wage for all those earning more than 0 but less than 83.5 percent of average wage; to 100 percent of average wage for all those earning more than 83.5 but less than 133.5 percent of average wage; and to 167 percent of average wage for all those earning more than 133.5 percent of average wage. That is, we assume that all individuals face the same FTR and METR who have a certain income level and whose spouse works in the formal sector within the income brackets of (i) 1 and 83.5 percent of average wage, (ii) 83.5 and 133.5 percent of average wage, and (iii) above 133.5 percent of average wage. Descriptive Statistics According to EU-SILC 2008, the sample of around 49,000 is representative for about 27 million workers in Bulgaria, the Czech Republic, Estonia, Latvia, Poland, and Slovakia. Of these, 28 percent are engaged in informal work (see Table 3). These 28 percent are largely made up of noncontributing workers (13 percent) and nonprofessional own-account workers (11 percent); the rest are family workers (2 percent) and nonprofessional employers in small firms (2 percent). The complement of 72 percent formal workers comprises largely of contributing workers (70 percent). Employers (in large firms or professionals in small firms) and professional own-account workers account for only 1 percent of all represented workers each. Looking at gender, the share of males among informal workers is 63 percent. This is much higher than the corresponding share for formal

190

JOHANNES KOETTL AND MICHAEL WEBER

Table 3. Informality Rates Across Different Groups and Countries. Bulgaria (%)

The Czech Republic (%)

Estonia Latvia Poland Slovakia Total (%) (%) (%) (%) (%)

By employment status Self-employed Employees

71.4 7.3

82.8 0.1

71.5 3.3

70.9 6.3

87.1 27.1

81.5 2.2

83.9 16.2

By sex Male Female

19.5 12.9

17.1 8.9

10.8 4.6

15.9 8.6

42.6 35.5

14.2 6.0

31.5 23.7

By age 15–24 25–39 40–54 55–64 65 or more

19.3 15.4 15.7 17.9 34.4

7.8 12.2 15.8 13.2 30.8

6.9 7.4 7.8 7.2 13.2

11.0 12.6 13.4 9.7 11.1

44.6 35.6 40.5 45.2 73.7

7.7 10.2 11.0 10.5 25.2

30.2 26.2 28.8 28.0 48.4

By income group 0% of AW or less 1–24% of AW 25–49% of AW 50–99% of AW 100–200% of AW 200% of AW or more

79.8 37.6 17.1 11.8 11.2 29.8

100.0 23.3 14.9 10.3 13.7 27.1

67.8 29.1 5.9 3.0 6.9 24.2

82.8 29.0 15.2 9.7 7.1 9.8

92.7 66.6 40.4 29.2 29.3 25.7

80.5 18.5 11.8 8.0 10.9 21.6

91.7 55.4 30.0 19.5 20.9 25.4

3.3 7.6

2.1 7.7

1.9 3.9

2.6 7.8

17.8 27.5

4.2 8.3

11.2 18.3

22.6 24.4 13.6

31.2 19.0 9.2

14.4 8.4 10.3

21.1 10.9 11.0

49.7 42.9 40.1

27.4 13.7 8.1

38.2 32.0 26.7

By sector Health services Mining, manufacturing Construction Trade and repair Transport and storage Accommodation and food ICT Financial services Professional services Public sector Education Agriculture

19.0

19.1

6.8

10.9

49.8

7.6

28.1

10.0 1.3 14.1 1.6 1.1 54.3

13.0 22.6 17.8 0.7 2.0 26.9

9.4 7.1 5.0 1.0 1.1 33.2

4.5 5.1 10.8 2.5 1.8 46.7

25.6 24.1 31.7 21.1 15.8 93.2

6.5 15.6 12.2 2.1 2.7 15.4

18.0 20.1 24.4 11.5 10.2 80.6

Overall

16.5

13.6

7.7

12.3

39.5

10.4

28.0

Source: Authors’ calculations based on EU SILC (2008) and OECD (2011).

Does Formal Work Pay?

191

workers (53 percent). This structural composition can be found in all countries represented in the sample. With respect to age, the group of 40- to 54-year olds holds the highest share (42 percent) among informal workers followed by the 25- to 39-year olds (37 percent). The age groups 15–24 and 50–64 account for 10 and 9 percent, respectively, while the age group of 65 and older comprises only 2 percent of all informal workers. This distribution of workers and age is rather similar across all countries. Only the Czech Republic and Slovakia have slightly lower shares of informal youth (15–24 years) that are compensated by higher shares of workers in the age bracket of 40–54 years. Comparing the age distribution of informal workers with those of formal workers, no distinctive difference can be found in the sample. The overall informality rate of 28 percent is within a range of 72 percent for all age groups except for those workers that are 65 years and more. The informality rate for this age group that only represents 1 percent of all workers is 48 percent. The informality rate of the young is 30 percent; the rate for the age group of 25- to 39-year olds is 26 percent, and around 28 percent for the remaining groups. In addition to these demographic considerations, the composition of the informal workers by sectors and income groups provides further interesting insights. The agricultural sector holds with 26.1 percent the highest share of informal workers, while it represents only 2.4 percent of all formal workers. The informality rate for agriculture is consequently high, it amounts to 80.6 percent.20 The second highest informality rate can be found in the construction sector. The rate is 38.2 percent and is followed by the rates of the categories ‘‘Others’’ (33 percent), ‘‘Trade and Repair’’ (32 percent), and the ‘‘Accommodation and Food Services’’ (28.1 percent). All other sectors are below the overall informality rate of 28 percent. By income group, informality rates are highest for the three lowest income groups. These are ‘‘0 income of the average wage or less’’ with 92 percent, ‘‘1–24 percent of the average wage’’ with 55 percent, and ‘‘25–49 percent of the average wage’’ with an informality rate of 30 percent. The other income groups are below the average informality rate. The highest share of informal workers can be attributed to the groups ‘‘50–100’’ and ‘‘25–49’’ percent of the average wage. Their shares are 32 and 23 percent, respectively, with the latter group covering a lower percentage range. In contrast, the groups ‘‘50–100’’ and ‘‘25–49’’ percent of the average wage represent 48 and 20 percent in the corresponding categories for formal workers. Consequently, the distribution of informal workers across the income groups is distinctively different from the distribution of formal workers with particularly high informality rates for the three lowest income groups.

192

JOHANNES KOETTL AND MICHAEL WEBER

Econometric Analysis To quantify the effects of the explanatory variables on the dichotomous outcome variable, a generalized linear model with a probit link function was applied. For a binary outcome, the probit equation is PðY i ¼ 1Þ ¼ FðX 0i  b þ i Þ with outcome variable Yi and explanatory variables Xi for respondent i. F(  ) stands for the cumulative distribution function of the standard normal (probit model) distribution.21 It is important to highlight that characteristics such as preferences for working independently, for flexible working hours, and the possibility to receive on-the-job training are not observed, which potentially could cause an omitted variable bias. However, we have made an effort to control for a large number of characteristics to reduce the potential bias as much as possible. In addition, the analysis does not establish a causal relationship, but mere correlations and should be interpreted with some caution. Causal relationships could point in either direction. High work disincentives might cause a high incidence of informal employment or high levels of informal employment might lead to high tax rates on formal labor, thereby increasing disincentives for formal work.

Expected Results The interpretation and use of the tax wedge as a measurement of disincentives for formal work suggests that the higher the tax wedge, the higher the disincentives for formal work and, hence, the higher the incidence for informality. Consequently, one would expect a positive correlation between the tax wedge and the incidence of informality. However, the descriptive statistics regarding informality rates (Table 3) and the tax wedge (Figs. 1–3) point at a negative correlation. The tax wedge is lowest with low-wage earners and strictly increases with income level (Fig. 3). At the same time, the incidence of informality is highest among low-wage earners (Table 3), which should lead to a negative correlation. Does this imply that work disincentives do not matter? Not necessarily, since there could be an income effect or the tax wedge could be an insufficient measurement of disincentives for formal work. With regard to the income effect, low-wage earners might earn so little that they cannot afford to

193

Does Formal Work Pay?

forego any amount of income. So even giving up a relative small amount of income could effectively be a higher disincentive for a low-wage earner than giving up a relatively larger amount of income for a high-wage earner. With regard to insufficient measurement, the tax wedge does not take any loss of benefits into account, as already discussed. Figs. 5 and 7 show a less straightforward relationship between income on the one hand and FTR or METR on the other when compared to the tax wedge.22 Assuming that FTR and METR address the shortcomings of the tax wedge, a positive correlation between FTR or METR and the incidence of informality is expected. Furthermore, the marginal effects of FTR and METR are expected to be higher for low-wage earners than middle- or high-wage earners for two reasons. One is the income effect as described above with forgone income having a higher importance for low-wage earners, resulting in a higher correlation between work disincentives and informality. The second reason is that the marginal effect not only depends on the coefficient but also on the incidence of informality. An estimation with a restricted sample of only lowwage earners should yield higher marginal effects when compared to an estimation with the full sample.

Results The results of the regression are presented in Table 4 for the full sample and Table 5 for low-wage earners only. We report marginal effects, computed at the mean characteristics. Controlling for individual and job characteristics (income and sector), there is a significant positive correlation between FTR or METR and the probability of being informal. In particular, a 1 percentage point increase in the FTR (METR) increases the probability of being informal by 1.1 percent (0.8 percent) for the full sample. For the lowwage earners, the correlation is stronger for both, FTR and METR. The marginal effects for the restricted sample are more than double compared to the ones for the full sample: 2.5 percent for FTR and 1.6 percent for METR. Estimations with a restricted sample of only high-wage earners (100 percent of average wage or above) confirm that the correlation is less strong, with coefficients either smaller or negative or insignificant. This confirms the expectations as outlined above. The marginal effect for the tax wedge is negative for both samples. As discussed above, the tax wedge increases with income, and because informality rates are lower at higher income levels, the overall correlation turns out to be negative. Therefore, it seems that the tax wedge might not

194

JOHANNES KOETTL AND MICHAEL WEBER

Table 4. Probit Regression Results with Informality Dummy as the Dependent Variable, Reporting Average Marginal Effects (Full Sample). (i) Formalization tax rate Marginal effective tax rate Tax wedge Country Bulgariaa The Czech Republic Estonia Latvia Poland Slovakia

0.011

(ii) 0.008

(iii)

0.084

0.095 0.110 0.086 0.163 0.028

0.077 0.101 0.113 0.123 0.071

0.846 0.176 0.554 0.404 0.372

0.017 0.030 0.015 0.120

0.010 0.027 0.020 0.137

0.029 0.012 0.026 0.181

0.076

0.080

0.079

0.053 0.111

0.058 0.107

0.042 0.089

Employment status of spouse Formala Informal Inactive No spouse

0.239 0.002 0.018

0.272 0.032 0.017

0.381 0.144 0.365

Degree of urbanization Densely populateda Intermediate Thinly populated

0.023 0.040

Age group 15–24a 25–39 40–54 55–64 65þ Sex Malea Female Education level Higha Medium Low

Citizenship Locala Other EU country Others

0.051 0.037

0.021 0.038

0.051 0.039

0.060 0.066

0.043 0.030

195

Does Formal Work Pay?

Table 4. (Continued ) (i)

(ii)

(iii)

Income group 0% of AW or lessa 1–24% of AW 25–49% of AW 50–99% of AW 100–200% of AW 200% of AW or more

0.132 0.259 0.389 0.272 0.202

0.183 0.289 0.393 0.278 0.209

0.198 0.274 0.417 0.266 0.179

Sector Health servicesa Mining, manufacturing, utilities Construction Trade and repair Transport and storage Accommodation and food services ICT Financial services Professional services Public sector Education Agriculture Others

0.055 0.281 0.224 0.150 0.231 0.113 0.142 0.155 0.004 0.022 0.502 0.298

0.053 0.282 0.225 0.151 0.233 0.116 0.143 0.146 0.008 0.020 0.509 0.296

0.046 0.248 0.205 0.144 0.208 0.096 0.116 0.140 0.002 0.024 0.444 0.284

Number of observations ¼ Wald chi2(37) ¼ Prob W chi2 ¼ Pseudo R2 ¼ Log pseudo-likelihood ¼ Mean dependent variable

47,065 5,865.19 0 0.2642 20,023.73 0.2650287

47,065 6,393.39 0 0.2878 19,382.73 0.2650287

47,065 4,876.38 0 0.5138 13,231.89 0.2650287

a Baseline category. Regressions based on individual data from EU-SILC (Eurostat, 2008) with matched data for individual FTR, METR, and tax wedge from OECD Tax and Benefit model (OECD, 2011). Coefficients are interpreted as follows: In the specification with FTR, an increase of 1 percentage point of the FTR increases the probability of being informal by 1.1 percent; living in the Czech Republic decreases probability of being informal by 9.5 percent when compared to living in Bulgaria; being female decreases probability by 7.6 percent; having low education increases probability by 11.1 percent when compared to someone with high education; and so on.  Statistically significant at: 10 percent.  Statistically significant at: 5 percent.  Statistically significant at: 1 percent.

196

JOHANNES KOETTL AND MICHAEL WEBER

Table 5. Probit Regression Results with Informality Dummy as the Dependent Variable, Reporting Average Marginal Effects (Low-Wage Earner Sample). (i) Formalization tax rate Marginal effective tax rate Tax wedge Country Bulgariaa The Czech Republic Estonia Latvia Poland Slovakia Age group 15–24a 25–39 40–54 55–64 65 þ Sex Malea Female Education level Higha Medium Low Employment status of spouse Formala Informal Inactive No spouse Degree of urbanization Densely populateda Intermediate Thinly populated Citizenship Locala Other EU country Others

0.025

(ii) 0.016

0.197 0.158 0.109 0.268 0.003

0.193 0.193 0.234 0.088 0.180

0.029 0.083 0.095 0.162

0.010 0.089 0.127 0.205

0.088

0.091

0.022 0.036

0.017 0.060

(iii)

0.092

0.687 0.042 0.289 0.422 0.197

0.038 0.167 0.056 0.127

0.120

0.003 0.054

0.154 0.085 0.102

0.083 0.132 0.047

0.426 0.212 0.356

0.072 0.046

0.073 0.051

0.080 0.046

0.096 0.009

0.090 0.029

0.012 0.177

197

Does Formal Work Pay?

Table 5. (Continued ) (i)

(ii)

(iii)

Income group 0% of AW or lessa 1–24% of AW 25–49% of AW

0.158 0.450

0.320 0.529

0.380 0.557

Sector Health servicesa Mining, manufacturing, utilities Construction Trade and repair Transport and storage Accommodation and food services ICT Financial services Professional services Public sector Education Agriculture Others

0.131 0.332 0.224 0.231 0.193 0.246 0.333 0.197 0.039 0.011 0.545 0.298

0.127 0.318 0.232 0.209 0.184 0.225 0.332 0.169 0.091 0.008 0.558 0.308

0.112 0.273 0.220 0.227 0.180 0.197 0.291 0.180 0.032 0.012 0.500 0.350

Number of observations ¼ Wald chi2(34) ¼ Prob W chi2 ¼ Pseudo R2 ¼ Log pseudo-likelihood ¼ Mean dependent variable

14,328 2,267.19 0 0.3409 6,373.44 0.4047872

14,328 2,810.70 0 0.3976 5,824.83 0.4047872

14,328 4,261.33 0 0.5669 4,187.99 0.4047872

a

Baseline category. Regressions based on individual data from EU-SILC (Eurostat, 2008) with matched data for individual FTR, METR, and tax wedge from OECD Tax and Benefit model (OECD, 2011). Coefficients are interpreted as follows: For low-wage earners, in the specification with FTR, an increase of 1 percentage point of the FTR increases the probability of being informal by 2.5 percent; living in the Czech Republic decreases probability of being informal by 19.7 percent when compared to living in Bulgaria; being female decreases probability by 8.8 percent; having low education increases probability by 3.6 percent when compared to someone with high education; and so on.  Statistically significant at: 10 percent.  Statistically significant at: 5 percent.  Statistically significant at: 1 percent.

sufficiently capture individual differences in disincentives for formal work. In that sense, the tax wedge does not seem appropriate as a measure for individual disincentives for formal work. The rest of the discussion will therefore focus on the specifications using FTR and METR.

198

JOHANNES KOETTL AND MICHAEL WEBER

As for the individual characteristics, the correlation with sex (male being the baseline category) clearly stands out as significant and negative. That is, women are clearly less likely to work informally. For the full sample, being female decreases the probability of being informal by 7.6 (specification with FTR) and 8 percent (METR). For low-wage earners, the correlation is stronger with 8.8 and 9.1 percent, respectively. This is in contrast to findings in other developing countries. Perry et al. (2007), for example, find that in Latin America women seem to value informal self-employment because the flexibility of the informal sector allows them to better balance their home and income-earning roles. Accordingly, informality rates in Latin America tend to be higher among women when compared to men. In transition countries, though, this seems not to be the case, as also confirmed by Hazans (2011). In terms of age, there seems to be no clear correlation between informality and age. The only consistent effect across specifications is observed for the age group 65 and older, which is more likely to be in informal employment compared to the 15 to 24 age group. With regard to education, the low and medium educated are significantly more likely to be informal when compared to the highly educated in the full sample. The results regarding the employment status of the spouse are somewhat surprising: there is a clear positive correlation between working informally and having an informally working spouse. This suggests that households do not make strategic decisions along the line of one partner working formally (and receiving employment and social protection, including for dependents), while the other one works informally. This result is robust with regard to restricting the sample to various subgroups, the agricultural or nonagricultural sector, low-wage earners, employees, or self-employed. Regarding inactive spouses or being single, the results are more ambiguous. Other individual characteristics like geography (rural or urban) and citizenship did not yield any clear significant correlations. Although for the sample there seems to be no correlation between informality and geography, for low-wage earners the probability of being informal is higher when living in less densely populated areas. With regard to citizenship, the full sample suggests a highly significant positive relationship between foreign citizenship (both EU and non-EU) and being informal, yet this is not the case in the restricted sample. The income variable confirms the expected relationship between income and informality. Those with no or negative income are clearly most likely to be informal, which is certainly explained by all unpaid family workers falling into this category. For the other income groups, the likelihood of

Does Formal Work Pay?

199

being informal decreases with income and is lowest for the income group earning 50–99 percent of average wage. For those earning more than average wage, though, the likelihood of being informal seems to increase again. One possible explanation is that enforcement of tax and contribution collection is highest for those earning around average wage, because they are the largest group of contributors. In other words, tax revenue collection is geared toward the middle class. Regarding job characteristics, agriculture consistently yields a significant and highly positive relationship with being informal. Other sectors with similar results are construction, trade and repair, transport and storage, and accommodation and food services. Public sector and education do not seem to have a significantly higher probability of being informal, but one has to bear in mind that the baseline category is health services, which by itself is mainly public. This also applies to the education sector.

CONCLUSIONS AND POLICY IMPLICATIONS The analysis presented defines three different synthetic measurements of disincentives for formal work: two standard measurements, namely the tax wedge and the METR; and a new, innovative measurement called FTR. The novelty of the latter is that it measures disincentives stemming not only from labor taxation but also from benefit withdrawal due to formalization. A descriptive analysis across a large number of OECD and Eastern European countries reveals that the disincentives for formal work – when measured through the FTR – are especially high for low-wage earners. This suggests that formal work might not pay in this segment of the labor market, in particular for the so-called mini-jobs and midi-jobs (low-paying part-time work). Another novelty is the empirical analysis that complements the descriptive analysis. Using EU-SILC 2008 data and OECD Tax and Benefit data for six Eastern European countries, we match disincentives for formal work that depend on income and family type to individual observations in a large data set. More precisely, EU-SILC data is matched with individual synthetic measurements of disincentives for formal work – namely, the tax wedge, METR, and the FTR – coming from the OECD Tax and Benefit model. Applying a probit regression, the analysis finds a significant positive correlation between FTR or METR and the probability of being informal. In other words, controlling for individual and job characteristics, the higher the FTR or the METR is that individuals are facing, the more likely they are to work informally. The tax wedge, on the other hand, yields a negative

200

JOHANNES KOETTL AND MICHAEL WEBER

correlation, indicating that the tax wedge is not sufficiently capturing disincentives for formal work. This also suggests that in cross-country analysis in which the tax wedge is used as an independent variable to explain informality levels, it might be more appropriate to use a tax wedge that is more representative for low-wage earners, not for average wage earners. For example, calculating the tax wedge at 33 percent of the average wage would represent better the actual tax wedge faced by most informal workers. These results lead to the question on how formal work can be made more viable for low-wage earners. The two main policy tools to make formal work pay are to decrease labor taxation at the lower wage levels and to reform benefit design for social assistance, housing, and family benefits. On the former, policies linked to wage subsidies, social security contribution credits, in-work or employment-conditional benefits (cash benefits or refundable income tax credits conditional on formal) for low-wage earners could play an important role (Immervoll & Pearson, 2009). With regard to reforming the design of social assistance, housing, and family benefits, the key is to keep the METR in mind when designing benefit withdrawal. Finally, it should be pointed out that most of the reforms discussed above have fiscal costs. Given fiscal constraints, there might be little fiscal space available to push through these reforms. In particular, reforms that aim at making work pay at the low wage end – like wage subsidies or tax credits – can considerably reduce tax revenues, including social security contributions, or increase public expenditures. In this regard, though, the NMS are in a favorable position: as shown above, their tax systems are relatively nonprogressive. Making the relatively nonprogressive tax system more progressive could make any future reforms along these lines fiscally neutral to a large extent.

NOTES 1. See, for example, Hazans (2011), as well as descriptive statistics presented further below in this text. 2. Lehmann (2010) establishes that EPL, the tax wedge, and union density have positive impacts on the level of informal employment in transition countries and quantifies these impacts. 3. The models for the non-OECD countries were developed under a recent research partnership between the OECD and the World Bank. 4. In many countries, full-time work at 33 percent of the average wage is below the legal minimum wage. Nevertheless, the same tax wedge applies to someone receiving average wage, but working 33 percent part-time, although there can be slight variations in the tax wedge for part-time workers when compared to full-time workers.

Does Formal Work Pay?

201

5. The assumed relationship is that tax systems need to raise a certain fixed amount of resources, and those that put higher taxes on lower wages have less need to increase taxes at higher wages and hence display less progressivity. 6. Exceptions are Hungary and the Netherlands, which have a social security contribution floor. Such a floor has to be paid independent of actual wages earned and therefore increases the tax wedge significantly at lower wage levels. 7. For the sake of illustration, the countries chosen in this and the following graphs are rather contrasting. That is, for the NMS, countries with fairly high disincentives in the low-wage sector have been chosen; for the OECD, countries with rather low disincentives in the low-wage sector have been chosen. 8. The four countries are chosen as illustrating examples, with typical yet contrasting tax wedges. 9. Only workers who are not registered at all are considered; partially formal workers who underreport their wages are not considered. 10. To the extent that these benefits are also offered to informal workers for free or at low cost, they do not enter as opportunity costs of formal work (Levy, 2008). 11. See, for example, Lawrance (1991) on the relationship between time preference and poverty. 12. For an analysis on Bulgaria, see Perotti and Sanchez Puerta (2009). 13. For a more detailed definition and discussion, see Koettl (2009). 14. Koettl and Weber (forthcoming) – an earlier version of this chapter – provide a detailed overview of the tax wedge, the FTR, and the METR for a large number of OECD and Eastern European countries, including the NMS and some Western Balkan countries. It presents a graphical depiction of these synthetic measurements for disincentives for formal work for two family types across the 0–200 percent of average wage spectrum, country by country. Where available, an estimate of the informality rate – based on EU-SILC 2008 – is included in the graphs. 15. The definition of informality for the self-employed and employers follows Hazans (2011). 16. For the Czech Republic and Slovakia, the employed are categorized only in the first two categories. 17. NACE stands for ‘‘Nomenclature generale des activite´s e´conomiques dans la Communaute´ europe´enne’’ and codes economic activity into various sectors and subsectors. 18. This is obviously a simplifying assumption. Yet, the only variation that could occur at income levels beyond the boundary are higher income tax brackets or ceilings on social security contributions, which could shift both the FTR and METR to some limited extent. 19. More precisely, the OECD Tax and Benefit model is provided for (i) single; (ii) single with two children; (iii) one-earner couple with no children; (iv) one-earner couple with two children; (v) two-earner couple with no children, spouse earning 67 percent of average wage; (vi) two-earner couple with no children, spouse earning 100 percent of average wage; (vii) two-earner couple with no children, spouse earning 167 percent of average wage; (viii) two-earner couple with two children, spouse earning 67 percent of average wage; (ix) two-earner couple with two children, spouse earning 100 percent of average wage; and (x) two-earner couple with two children, spouse earning 167 percent of average wage.

202

JOHANNES KOETTL AND MICHAEL WEBER

20. In the literature, analysis on informal employment is often restricted to the nonagricultural sector. The analysis presented below focuses on disincentives in general and will therefore include the agricultural sector. Nevertheless, the analysis controls for sectors and therefore takes into particularly high incidence of informal employment in agriculture. 21. A correlation analysis helped to explore pairwise collinearity. Moreover, multicollinearity of the explanatory variables was explored by variance inflation factors (VIFs; Fox & Monette, 1992). None of the variables showed pairwise collinearity or multicollinearity. 22. For a detailed description of the tax wedge, FTR, and METR across many OECD and transition countries, see Koettl and Weber (forthcoming).

ACKNOWLEDGMENTS The chapter greatly benefited from research support from Isil Oral and Victoria Strokova; comments by Hartmut Lehmann, Truman Packard, Konstantinos Tatsiramos, two anonymous referees, and participants at an IZA workshop on informal employment; and a close cooperation with Claudio Montenegro as well as with Tatiana Goridne, Herwig Immervoll, and Dominique Partout. All errors are those of the authors.

REFERENCES Bassanini, A., & Duval, R. (2006). Employment patterns in OECD countries: Reassessing the role of policies and institutions. Economics Department Working Paper No. 486. OECD, Paris. Bertola, G., Jimeno, J. F., Marimon, R., & Pissarides, C. (2001). EU welfare systems and labor markets: Diverse in the past, integrated in the future? In G. Bertola, T. Boeri & G. Nicoletti (Eds.), Welfare and employment in Europe. Cambridge, MA: MIT Press. Betcherman, G., & Page´s, C. (2007). Estimating the impact of labor taxes on employment and the balances of the social insurance funds in Turkey. Washington, DC: World Bank. Davis, S. J., & Henrekson, M. (2004). Tax effects on work activity, industry mix and shadow economy size: Evidence from rich-country comparisons. NBER Working Paper No. 10509. National Bureau of Economic Research, Cambridge, MA. Djankov, S., Lieberman, I., Mukherjee, J., & Nenova, T. (2002). Going informal: Benefits and costs. Washington, DC: World Bank. European Commission (EC). (2003). European employment strategy: 2003 national action plans. Brussels: EC. Retrieved from http://europa.eu.int/comm/employment_social/ employment_strategy/index_en.htm European Commission (EC). (2004). Undeclared work in an enlarged union: An analysis of undeclared work: An in-depth study of specific items. Brussels: EC.

Does Formal Work Pay?

203

Eurostat. (2008). EU-SILC – European Union statistics on income and living conditions. Luxembourg: Eurostat. Retrieved from http://www.eui.eu/Research/Library/Research Guides/Economics/Statistics/DataPortal/EU-SILC.aspx Fox, J., & Monette, G. (1992). Generalized collinearity diagnostics. Journal of the American Statistical Association, 87(417), 178–183. Hanousek, J., & Palda, F. (2003). Why people evade taxes in the Czech and Slovak Republics: A tale of twins. In B. Belev (Ed.), The informal economy in the EU accession countries: Size, scope, trends and challenges in the process of EU enlargement. Sofia: Center for the Study of Democracy. Hazans, M. (2011). Informal workers across Europe: Evidence from 30 European countries. IZA Discussion Paper No. 5871. IZA, Bonn, Germany. Hussmanns, R. (2004). Statistical definition of informal employment: Guidelines endorsed by the seventeenth international conference of labour statisticians (2003). Geneva: International Labour Office. Retrieved from http://ilo.org/public/english/bureau/stat/download/ papers/def.pdf Immervoll, H. (2007). The OECD tax-benefit model and policy database. In A. Gupta, & A. Harding (Eds.), Modelling our future: Population ageing, health and aged care. International Symposia in Economic Theory and Econometrics (Vol. 16, pp. 503–506). Amsterdam, The Netherlands: Elsevier. Immervoll, H., & Pearson, M. (2009). A good time for making work pay? Taking stock of in-work benefits and related measures across the OECD. IZA Policy Paper No. 3. IZA, Bonn, Germany. International Labour Organization (ILO). (2003). General report of the seventeenth international conference of labour statisticians. Geneva, November 14–December 3. Koettl, J. (2009). The role of income taxes and social protection in providing incentives for informality. A first glance at the Czech Republic, Hungary, Poland, and Slovakia. Background paper prepared for World Bank. Mimeo. Koettl, J. & Weber, M. (Forthcoming). Disincentives for formal work in OECD and Eastern European countries. A descriptive and empirical analysis of the tax wedge, the marginal effective tax rate, and the formalization tax rate. Policy Research Working Paper Series. World Bank. Washington, DC. Lawrance, E. C. (1991). Poverty and the rate of time preference: Evidence from panel data. Journal of Political Economy, 99(1), 54–77. Lehmann, H. (2010). Policies to combat informality and to broaden the tax base: Lessons for transition countries. Background paper prepared for World Bank. Mimeo. Levy, S. (2008). Good intentions, bad outcomes: Social policy, informality, and economic growth in Mexico. Washington, DC: Brookings Institution Press. Maloney, W. F. (1999). Does informality imply segmentation in urban labor markets? Evidence from sectoral transitions in Mexico. World Bank Economic Review, 13, 275–302. Maloney, W. F. (2004). Informality revisited. World Development, 32, 1159–1178. Organisation for Economic Development and Co-operation (OECD). (2004). Employment outlook. Paris: OECD Publishing. Organisation for Economic Development and Co-operation (OECD). (2011). Tax/benefit policies: Detailed descriptions and reforms since 2001. Online database. Retrieved from http://www.oecd.org/document/29/0,3746,en_2649_34637_39618653_1_1_1_1,00.html Perotti, V., & Sanchez Puerta, M. L. (2009). Personal opinions about the social security system and informal employment: Evidence from Bulgaria. Social Protection Discussion Paper Series No. 0915. World Bank, Washington, DC.

204

JOHANNES KOETTL AND MICHAEL WEBER

Perry, G. E., Maloney, W. F., Arias, O. S., Fajnzylber, P., Mason, A. D., & Saavedra-Chanduvi, J. (2007). Informality: Exit and exclusion. Washington, DC: World Bank. Schneider, F., & Enste, D. H. (2000). Shadow economies: Size, causes, and consequences. The Journal of Economic Literature, XXXVIII. Schneider, F., & Klingelmair, R. (2004). Shadow economies around the world: What do we know? IZA Discussion Paper No. 1043. IZA, Bonn, Germany. Terrell, K., & Grindling, T. H. (2002). The effect of minimum wages on the formal and informal sector: Evidence from Costa Rica. Washington, DC: World Bank. Torgler, B. (2003). Tax morale in transition countries. Post-Communist Economies, 15, 357–381. Torgler, B., & Schneider, F. (2009). The impact of tax morale and institutional quality on the shadow economy. Journal of Economic Psychology, 30, 228–245. World Bank. (2011a). ECA social protection database. Washington, DC: World Bank. World Bank. (2011b). Enterprise surveys 2009. Washington, DC: World Bank. Retrieved from http://www.enterprisesurveys.org

CHAPTER 6 MIGRATION AS A SUBSTITUTE FOR INFORMAL ACTIVITIES: EVIDENCE FROM TAJIKISTAN Ilhom Abdulloev, Ira N. Gang and John Landon-Lane ABSTRACT How is migration related to informal activities? They may be complementary since new migrants may have difficulty finding employment in formal work, so many of them end up informally employed. Alternatively, migration and informality may be substitutes since migrants’ incomes in their new locations and income earned in the home informal economy (without migration) are an imperfect trade-off. Tajikistan possesses both a very large informal sector and extensive international emigration. Using the gap between household expenditure and income as an indicator of informal activity, we find negative significant correlations between informal activities and migration: the gap between expenditure and income falls in the presence of migration. Furthermore, Tajikistan’s professional workers’ ability to engage in informal activities enables them to forgo migration, while low-skilled nonprofessionals without postsecondary education choose to migrate

Informal Employment in Emerging and Transition Economies Research in Labor Economics, Volume 34, 205–227 Copyright r 2012 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1108/S0147-9121(2012)0000034009

205

206

ILHOM ABDULLOEV ET AL.

instead of working in the informal sector. Our empirical evidence suggests migration and informality substitute for one another. Keywords: Informal; migration; remittances; Tajikistan JEL classification: O17; J61; P23

INTRODUCTION All economies contain some form of informal/unreported activity. This paper considers the influence migration has on this type of activity. We argue that migration and informal sector activity are viable options for the household. The migration literature going back to at least Harris and Todaro (1970) and the papers by, for example, Fields (1975, 1976, 1979) and Gang and Gangopadhyay (1987a, 1987b, 1987c) generally introduce the informal sector as a complement to migration – that is, the informal sector becomes a staging ground for those trying to get formal sector jobs, part of the process that drives modern economic growth, and, frequently, urbanization. In these models, informal sector work and migration are complementary: migrants have difficulty finding employment in formal work in ‘‘new’’ places, so many of them end up informally employed. The informal sector is in the migrant’s destination location, along with the good jobs the migrant is hoping to get. It is also possible for the informal sector – if it pays enough – to be the migrant’s desired employment. Our approach is somewhat different as we consider informality and migration as possible alternatives to one another. While the informal sector may be part of the process of economic growth and growing urbanization described in the previous paragraph, the informal sector may also be a home for entrepreneurs, a place to supplement ‘‘regular’’ earnings, or, alternatively, a home of last resort where the vulnerable end up during periods of economic hardship. This is local informal activity and this is the focus of our investigation into a trade-off between migration and informality. As substitutes, migration may effectively ‘‘crowd out’’ informality: migrant’s earnings help improve families’ finances encouraging their members to be less involved in informal employment. This structure has not been generally addressed in migration models. Thorbecke (1999) describes the coexistence

Migration as a Substitute for Informal Activities

207

of modern and informal/traditional sectors in both urban and rural regions, modeling their linkages via social accounting matrices. Building on the Harris and Todaro (1970) model, Gang and Gangopadhyay (1987c) allow for regular and informal employment in both urban and rural regions, with the possibility of open urban unemployment. With this extra complexity, whether migration out of the rural region and rural informality are substitutes or complements depends on relative wages and the various labor supply and demand elasticities. Using the Living Standards Measurement Survey (LSMS), our stage is Tajikistan, a poor Central Asian economy and former Soviet Republic possessing both a very large informal sector and extensive external migration. Our aim is to define the direction of correlation between informality and migration, and to examine nuances of the relationship. In the next section, we provide a short introductory background to the major economic events in Tajikistan’s recent history, emphasizing elements that are important to our story. We then discuss our approach to measuring informal sector activity; we discuss the data used in this study, report on the results, and draw our conclusions.

BACKGROUND Tajikistan underwent severe economic, social, and political changes following its separation from the USSR. Independence in 1991, with its rupture of economic ties, was followed by civil war among rival regional clans from 1992 to 1997 and then an initially tenuous peace. Tajikistan’s GDP fell by 65% from US$2.6 billion in 1990 to US$921.8 million in 1997, while inflation peaked at 1207.2% using the GDP deflator in 1993, two years after independence, and was still at 65.2% in 1997 (World Bank, 2011a, 2011b). After reaching reconciliation in 1997, the joint government initiated strict fiscal and monetary policies, along with the privatization of small and medium state-owned enterprises, and price and trade liberalization.1 For the last decade, annual real GDP growth has averaged 8.4%, and the inflation rate was also moderated at average annual rate of 20.5% over the decade 2001–2010 (World Bank, 2011a, 2011b). Despite these positive developments, Tajikistan remains the poorest country among former Soviet countries with 47.2% of its population living below the poverty line in 2009 (United Nations, 2011). Average monthly wages were US$83 in 2010; 8.5 times lower than that in Russia (Statistical Committee of CIS, 2011).

208

ILHOM ABDULLOEV ET AL.

For agriculture, forestry, and fisheries, which provide jobs to 50% of the employed population, monthly wages average US$42 (Statistical Agency of Tajikistan, 2011). The institutional transformation in Tajikistan was slowed by its civil war. The absence or weaknesses of newly established institutions spurred the increase of the informal sector in Tajikistan. Severe economic conditions during the war and post-war recovery period reduced the number employed in state enterprises. Extremely low wages and economic recession drove many employees of state-owned enterprises and kolkhozes (collective farms) to self-employment and migration. Tajikistan’s Statistical Agency reports the official unemployment rate as increasing from 0.4% in 1992 to 2.9% in 1998, though this is generally recognized as an understatement. An informal consensus suggests in 1999 the unemployment rate was above 40%, including hidden unemployment (Noda, 1999). Financial constraints for families increased after the loss of savings collected during the Soviet period due to high inflation. Families in Tajikistan were not able to solely rely on wages as a source of income, as they did during the Soviet time. The average monthly wage in 1998 was 8,287 Tajik rubles (US$9.9 at the official National Bank rate), far less than the internationally recognized subsistence level of ‘‘one dollar per day.’’ In 1996, the real monetary income was 38.9% of the 1991 level (Robertson, 1999). Such conditions led to the increase of the shadow economy and informal sector in Tajikistan. In 2006, the size of the shadow economy in Tajikistan reached 60.9% of GDP, tax avoidance amounted to about one-third of GDP, and home production of food was 14.7% of GDP, while income from in-kind wages and barter exchange was 13.1% of GDP. Informal employment is common in Tajikistan, with only 46% of household members who are in the labor force employed in formal sector work in 2006. Moreover, 45.4% of respondents received income from informal employment that was 2.7 times higher than the income from formal employment (United Nations Development Programme, 2007). Tajikistan is a country with significant external migration, such that approximately 37% of the labor force is working outside of the country. Most emigrants go to Russia (95.3% of migrants, 2007 World Bank Living Standard Measurement Survey (2007 WB LSMS)). Increasing migration led to the increasing inflow of remittances into Tajikistan, which in its turn helped to support positive economic growth. Tajikistan became the most remittance-dependent country in the world. In 2009, the total received remittances were counted as 35% of its GDP (World Bank, 2011a, 2011b).

Migration as a Substitute for Informal Activities

209

According to 2007 WB LSMS, the international labor migration from Tajikistan is dominated by men (93.5%), from rural areas (76.4% of all migrants), and ethnically Tajik (81.4% of all migrants). Only 10.7% of migrants had obtained postsecondary schooling; 76.2% graduated from secondary schools. The majority of current migrants were unemployed, 66.5%, and only 26.6% of migrants were working before migration; and, the remaining were students, pupils, or militants. The percentage of migrants who remitted both in kind and in cash in last 12 months was 6.6%; 74.2% remitted in cash only, and 1.0% remitted in kind only. We can draw out of this that Tajikistan was an economy in crisis during most of the first decade of separation from the Soviet Union. Over the second decade, the economy has become stable and growing, yet marked by two potentially problematic features: a very large informal sector and extensive emigration. The remainder of this paper analyzes the relationship between these two phenomena and examines their implications for households and the economy.

MEASURING INFORMAL SECTOR/UNREPORTED ACTIVITY The purpose of this paper is to document the impact migration has on informal and unreported activity. To do so we follow the approach used in Dimova, Gang, and Landon-Lane (2006) that looks at income and expenditure information at the household level to determine the amount of informal/unreported activity for each household. There are many definitions of informal activity including, but not limited to, activity in organizations that have less than five employees, activity in organizations that do not use modern production techniques (sometimes referred to as traditional sector employment), employment in activities that do not have employment protections, and employment in organizations that do not have access to formal capital markets. In this paper, we do not make a distinction among these definitions but rather look for evidence that a household is spending considerably more than its total income. This, we believe, is a good indicator of the unreported activity in an economy. A large component of this unreported activity is informal sector activity. To measure the size of unreported activity, we turn to income and expenditure data at the household level. Total income is computed as including total receipts from employment, net transfers from government

210

ILHOM ABDULLOEV ET AL.

agencies, remittances from household members living away from home, the market value of assets consumed (e.g., livestock, vegetables), and the market value of labor services rendered for which payment was in kind. Total expenditure for a household includes total payments for good and services consumed, the market value of goods and services consumed where payment was made in kind, the market value of assets consumed, and the value of savings (or asset accumulation). We measure total reported income and total reported expenditures, with the excess of total expenditures over income regarded as unreported income. There are many reasons why there would be a discrepancy between households’ reported expenditures and income, such as nonreporting of informal sector income, memory recollection problems, or problems assigning market prices to in-kind consumption or income. Our analysis looks at the variation in this discrepancy across different households, and in particular we look at the differences between households that contain migrants and those that do not. Our assumption is that the only major difference between these households is that households with migrants receive observed remittance income. We use the household as the unit of analysis since expenditures are difficult to assign to any one individual. While the source of formal sector income can often be assigned to an individual, in keeping with our idea of the informal sector, informal income invariably cannot. Formal sector employees may have a second informal sector job; an apparently nonworking member of the household may in fact be employed in the informal sector; or children may be participating in the informal sector. People may participate in both formal and informal activities. Our approach is different from much recent work on the informal economy, which has followed a paradigm set out by International Labour Organization (ILO) and World Bank staff. The approach is nicely summarized in Perry et al. (2007), and synthesized with some earlier approaches especially in Box 1.1 (p. 27). The idea is that there are two main definitional strands: the earlier ‘‘productive’’ and the more recently fashioned ‘‘legalistic’’. The productive categorization defines informality, as its label implies, by the production attributes of a firm: for example, a firm might be defined as informal if it employs less than 5 people and uses mechanical power or less than 20 people if it does not use such power. The legalistic categorization essentially distinguishes people who have social protection from those who do not. This approach has been used in labor economics and occasionally in international trade for decades, especially in distinguishing between covered and uncovered sectors. These two ways use

Migration as a Substitute for Informal Activities

211

information about the firm in which the individual is working to identify whether an individual is working in the informal sector. The approaches overlap with one another in their identification of who is in the informal sector, for example, workers employed by firms having limited capital and offering no formal labor market protections are counted by both approaches. Our approach overlaps with these categorizations, but the overlap is not defined along the same rows and columns useful for comparing the productive and legalistic categorizations. For example, our measure will capture those working in formal jobs as their first job, and who work in informal jobs as second or third jobs. The other approaches have difficulty with second and third jobs, even when reported, as individuals may report industry characteristics of their first job that would make it look to the researcher that they were engaging in formal activity when in fact the majority of their income was sourced from informal activity. On the other hand, our approach does not capture informal activity by a household reporting income equal to expenditure.2 The main advantage of our approach is that it allows the use of the rich trove of survey data to examine informality and the link between informality and other aspects of the economy. By looking at the disparity between reported income and reported expenditures as evidence of informal sector activity, our approach does not need detailed information about the working environment (whether it be firm characteristics or worker protections) in order to assign individuals (or households) to informal sector activity.

EMPIRICAL ANALYSIS Data Used This study uses the 2007 WBLSMS on Tajikistan.3 The survey data is based on a representative probability sampling on (i) Tajikistan as a whole; (ii) total urban and total rural areas, and (iii) five main administrative regions (oblasts) of the country: Dushanbe (the capital), Regions of Republican Subordination (RRS), Sogd Oblast, Khatlon Oblast, and Gorno-Badakhshan Autonomous Oblast (GBAO). This data provides a good basis for our analysis as it incorporates all relevant information on the flow of resources in and out of the household. The data is collected by interviewing 4,860 households in two rounds from September to November

212

ILHOM ABDULLOEV ET AL.

2007. The first round of interviews was conducted in September–October 2007, during the Ramadan period. The second round was conducted in October–November 2007 to gather additional information, and, to readminister food consumption to take into account its changes because of Ramadan. This survey asks questions on migration, education, health, labor market, housing, transfer and social assistance, subjective poverty and food security, as well as data for household’s expenditure and income. Income variables include both cash and in-kind forms of remittances, scholarships, wages and bonuses, individual transfers, social assistance, pensions, income from selling harvest, farm animals and poultry (or their product), and other income. Expenditures include payments for food, education, transportation, payments for health and medication, mortgage payments, house utilities and rent, assistance provided to other individuals, payments for the land use, purchases related with land cultivation and harvesting, purchases of farm animal and poultry breeding, and their food. All income and expenditure variables are converted to monthly equivalent for each household in our estimations. Table 1 reports the definitions of the variables used in the regression analysis. The dependent variable used is the natural logarithm of the ratio of reported expenditures to reported income, where income includes remittances from members of the household living away from the household. As described above, the income and expenditure variables are computed from self-reported income and expenditure data that includes good or services given or received in kind. Households that report a larger expenditure than income clearly have unreported income. Based on earlier work, this discrepancy includes informal sector income that is not reported in the income reports but does show up in the expenditure reports (see Dimova et al., 2006). We investigate the relationship between remittance income (a household’s income derived from a household member working away from home and sending money to the household) and informal sector income. We aim to see if these two sources of income are substitutes or complements. To do so we look to see if the presence of a migrant in a household or a recently returned migrant – that is, a member of the household has left the home and is potentially remitting income – has an impact on the amount of excess expenditure over income. Our assumption here is that this excess expenditure over income, while due to many factors including measurement error and recall error, is mainly due to the presence of unreported income. If a household substitutes informal sector income for remittance income, then this would show up as a decrease in the excess expenditure over income.

213

Migration as a Substitute for Informal Activities

Table 1. Variable Name log(expenditure/income)

Migrant (abroad) Migrant (returned)

Borrow Vocational

University

Single No. of children (o15) No. of elderly (W65) Ethnic Urban Land Self-employed Professional

Variables Used in Regressions. Description

Difference of log of totally reported expenditure and log of totally reported income; the income and the expenditure are defined at monthly rates from all reported sources A dummy variable taking a value of 1 if a household has any current migrant who is currently abroad and 0 otherwise A dummy variable taking a value of 1 if a household has an external migrant who was abroad for less than 12 months and recently returned and 0 otherwise A dummy variable taking a value of 1 if the household borrowed money and 0 otherwise A dummy variable taking a value of 1 if the highest level of education for the head of household is a vocational qualification and 0 otherwise A dummy variable taking a value of 1 if the highest level of education for the head of household is a university degree and 0 otherwise A dichotomous variable taking a value of 1 if the head of household is single and 0 otherwise Number of children in the household with ages less than 15 Number of elders in the household with ages greater than 65 A dummy variable taking a value of 1 if a head of household is a member of an ethnic minority group and 0 otherwise A dummy variable taking a value of 1 if a household lives in urban area and 0 otherwise A dummy variable taking a value of 1 if the household has access to land and 0 otherwise A dummy variable taking a value of 1 if any member of the household owns his/her business or farm and 0 otherwise A dummy variable taking a value of 1 if a head of the household is employed in a professional occupation and 0 otherwise

We find that households with migrants have a lower ratio of expenditure to income. We interpret this as a substitution of unreported informal sector income for reported remittance income. We argue that as the difference is so large it is hard to believe that this drop in excess expenditure is due to systematic differences in potential reasons for an excess expenditure over income such as recall error and mispricing of in-kind consumption. Another possible reason for this is that there are systematic differences between migrant and nonmigrant households with respect to their savings. It could be argued that migration causes income to increase to an extent that migrant households save some of their reported income thus lowering the observed

214

ILHOM ABDULLOEV ET AL.

excess expenditure over income. Of course, this is a possibility but for the reasons outlined below is a extremely unlikely event. Tajikistan is poor with a large proportion of the population living on or below the poverty line. The average increase in expenditure shown in Table 2 is approximately 25% between returned migrant households and nonmigrant households. If all the reduction in excess expenditure is due to increased savings, then that would mean households save on average 10% of their income. This is an implausibly high number for a developing country whose population is living close to or below the poverty line. A second consideration is that while we do not have information regarding household savings in our sample, there is evidence from other studies that suggest that very few Tajik households have bank accounts. The survey asks whether the household has a bank account and in the survey 99% of respondents did not have a bank account. Also, the ILO (2010, p. 33) reports ‘‘It is interesting to note that whether or not one receives remittances appears to have little impact on the likelihood of having a bank account. Of all households who receive remittances, 98% do not have a bank account, while 99% of households who do not receive remittances do not have an active bank account.’’ Moreover, there are no differences between migrant and nonmigrant households in terms of house ownership. It is very hard to argue that the observed reduction in the excess expenditure over income is caused by migrant households saving some of their new income. We start with a model that has the difference in log expenditures to log incomes as the dependent variable and variables indicating whether there is a current migrant or a recently returned migrant in the household as an independent variable. We also include other household characteristics to check the robustness of our regression results. The additional variables used in the regression include whether the household had taken a loan to capture whether it was credit constrained (Borrow), and household demographic variables such as whether the household is a single household, that is, the head of the household is single and has never been married (Single), the number of children under the age of 15 present in the household, and the number of adults over the age of 65 in the household. We also include information as to whether the head of the household is a member of an ethnic minority as we want to allow forth potential that ethnic minorities are discriminated against in the informal economy as well as the formal economy. We include education indicator variables to control for the possibility that unreported activity may be a function of one’s education. The education variables we use is an indicator variable that takes the value of 1 if the

Log (expenditure/income) Total income Total expenditure Migrant (abroad) Migrant (returned) Borrow Vocational University Single No. of children (o15) No. of elderly (W65) Ethnic Urban Land Self-employment No. of observations

Variable

0.97 558.06 1,371.11 0.17 0.10 0.05 0.11 0.19 0.01 2.20 0.30 0.21 0.35 0.65 0.51 4,391

Mean 1.36 811.05 1,851.37 0.37 0.30 0.23 0.31 0.39 0.11 1.70 0.58 0.41 0.48 0.48 0.50

SD 0.75 575.14 1,305.72 1.00 0.09 0.06 0.08 0.13 0.01 2.04 0.30 0.19 0.24 0.74 0.47 733

Mean 1.24 620.72 1,651.79 – 0.29 0.27 0.27 0.33 0.12 1.74 0.58 0.39 0.43 0.44 0.50

SD

Migrant (Abroad)

0.64 911.26 1,686.79 0.15 1.00 0.07 0.13 0.13 0.01 2.48 0.22 0.22 0.25 0.78 0.47 447

Mean 1.48 839.24 2,213.84 0.36 – 0.29 0.34 0.34 0.08 1.75 0.53 0.42 0.43 0.41 0.50

SD

Migrant (Returned)

Summary Statistics: Full Sample and Migration.

Full Sample

Table 2.

1.05 513.42 1,347.85 – – 0.05 0.11 0.21 0.01 2.20 0.31 0.22 0.38 0.62 0.52 3,280

Mean

1.37 833.69 1,838.14 – – 0.22 0.31 0.41 0.11 1.69 0.58 0.41 0.49 0.49 0.50

SD

No Migrant

Migration as a Substitute for Informal Activities 215

216

ILHOM ABDULLOEV ET AL.

highest level of education for the head of the household is a vocational training school (Vocational) or a university education (University). The next set of variables describes the type of work undertaken by the head of household. We use the variable (Self-employment) to reflect whether anyone in the household is self-employed and the variable (Professional) to denote that the head of the household works in a professional job. A feature of Tajikistan is the fact that professionals such as doctors and lawyers are paid low wages by the state, and it is standard for them to augment their income by taking clients ‘‘off the books’’ – low wages in Tajikistan have driven professionals look for additional, informal, earnings. Since professionals have wider social networks, access to information, and flexible time, it is easier to them to open own businesses or provide services to other institutes or people. Such cases are common in developing poor countries. For example, Kinyanjui (2010) discusses a case of disempowered professionals in Kenya who do additional informal work after their normal working time. It is also common in Tajikistan to see a doctor practicing at home after hours, teachers providing after-school tutorship to students, lawyers practicing their clientele beyond their office times, all for ‘‘underthe-table’’ payments. Out-of-pocket payments are also common in hospitals and clinics (Falkingham, 2004). Professionals holding managerial positions in state agencies and enterprises in Tajikistan might also receive an ‘‘unofficial’’ income in forms of gifts or bribes. The corruption and bribery in Tajikistan is common and it has impacted every sector and level of state agencies (UNDP and the Center for Strategic Studies under the President of the Republic of Tajikistan, 2010). Nonprofessionals, on the other hand, have no access to bribes, since they work at lower occupations and have fewer opportunities to be involved into informal sector employment due to their time and physically intensive work. Finally, we include a set of variables that aim to capture the opportunity set of households to find informal sector work. Such variables are whether the household lives in an urban area (Urban), and whether the household has access to land for cultivation (Land). Sample summary statistics (means and standard deviations) for each variable are reported in Tables 2 and 3. There are two samples used: the first is the sample of all households who report their positive total income and the second sample is the first sample restricted to those households who report an occupation. Included in this second sample are all those households who work. We see that the mean log ratio of expenditure to income is 0.97 that equates to a mean ratio of excess expenditure to income greater than 2.5. Thus, there appears to be a large amount of unreported

Log (expenditure/income) Total income Total expenditure Migrant (abroad) Migrant (returned) Borrow Vocational University Single No. of children (o15) No. of elderly (W65) Ethnic Urban Land Self-employment Professional No. of observations

Variable

Table 3.

0.85 611.27 1,422.88 0.14 0.10 0.06 0.13 0.25 0.01 2.15 0.12 0.21 0.36 0.64 0.56 0.27 2,799

Mean

All

1.25 848.36 1,876.75 0.34 0.30 0.25 0.33 0.43 0.10 1.62 0.37 0.41 0.48 0.48 0.50 0.44

SD 0.66 660.53 1,406.10 1.00 0.09 0.07 0.10 0.18 0.02 1.97 0.14 0.17 0.22 0.78 0.61 0.22 381

Mean 1.14 615.52 1,617.43 – 0.29 0.29 0.30 0.38 0.12 1.67 0.38 0.38 0.41 0.42 0.49 0.42

SD

Migrant (Abroad)

0.35 1,052.48 1,693.86 0.13 1.00 0.09 0.14 0.18 0.01 2.38 0.13 0.22 0.24 0.81 0.54 0.11 278

Mean 1.25 859.21 2,092.51 0.33 – 0.33 0.35 0.38 0.10 1.63 0.42 0.41 0.43 0.39 0.50 0.32

SD

Migrant (Returned)

0.93 553.77 1,400.00 – – 0.06 0.13 0.27 0.01 2.16 0.12 0.22 0.40 0.59 0.56 0.30 2,175

Mean

1.25 865.20 1,896.46 – – 0.24 0.34 0.45 0.10 1.61 0.36 0.41 0.49 0.49 0.50 0.46

SD

No Migrant

Summary Statistics: Sample of Reported Occupation by Heads of Households.

Migration as a Substitute for Informal Activities 217

218

ILHOM ABDULLOEV ET AL.

income in Tajikistan. We also break the sample into those households with a migrant who is currently away from home, a household that has a recently returned migrant, and those households with no migrant. The sample means show that households without any migrants have the largest excess expenditure over income with the households with a recently returned migrant having the smallest ratio of excess expenditure over income. This suggests that unreported income of whatever source is being replaced with reported remittances from migrants. We observe the same picture when we look at the incomes and expenditures separately. Households with migrants have a higher mean income and also a higher mean expenditure with households with returned migrants having the largest income and expenditure.4 Approximately a quarter of all households have a member who has migrated with 17% of households having a migrant who is currently abroad and 10% having a recently returned migrant. Note that some households have both a recently returned migrant and a currently abroad migrant. Very few households borrowed money in the survey period (approximately 5%) and less than 1% were households with a nonmarried head. The education levels of migrant and nonmigrant households are somewhat different. For the full sample, only 19% of the households have a university-educated head, whereas for households with migrants only 13% of the households are university educated. Thus, it appears that migrant households have lower education than nonmigrant households. When we add in households with vocational training (nonuniversity postsecondary education), we observe that 32% of nonmigrant households have some form of postsecondary education whereas between 21% and 26% of migrant households have some form of postsecondary education. Table 3 reports the same statistics for the subsample of households that report an occupation. This subsample is dominated by those who work, so it is not surprising that the incomes and expenditures for these households are slightly larger than for the full sample. However, the comparisons between migrant and nonmigrant households are qualitatively similar for the reduced sample as for the full sample. In order to specify the partial, or marginal, impact that migration status of a household has on the differences we observe between log expenditure and log income, we require a multivariate analysis that includes an array of variables that may influence this difference. We now turn to regression analysis, reporting these results in the next section.

Migration as a Substitute for Informal Activities

219

Regression Results Table 4 reports the result of simple regressions with the log of the ratio of expenditures to income as the dependent variable and household characteristics as the independent variables. All models are estimated using ordinary least squares with the reported standard errors computed using 1,000 bootstrapped replications.5 Regression (1) is just the simple linear regression replicating the difference in means test between households with no current or recent migrants and households with current migrants and households with migrants who have recently returned (within the last 12 months). We see that households with current or recent migrants have significantly lower excess expenditure than households without any migrants. Households with a migrant who is currently abroad have excess expenditures that are 26.2% lower than the reference nonmigrant households, while households that have a recently returned migrant have excess expenditures that are 36.9% lower than the reference nonmigration household. This supports our assertion that migrant income is a substitute for informal or nonreported activity. The full effect of the additional migrant income occurs after the migrant has returned, but there is a significant impact even when the migrant is still abroad, most likely through remittances sent back to the household from abroad. This result is obtained using the full sample of households. Regressions (2)–(4) report results for regression models that include the various household characteristics for the same sample. A number of important features are evident from these results. First, the coefficients on the two migrant indicator variables are consistent across specifications and are always significant. The result that income from migrant labor is a substitute for informal or nonreported activity is robust to our different specifications. Regression (2) adds variables that indicate a household’s education level and whether or not they borrowed money in the past month. The coefficient on the variable borrowed is significant and positive with households who borrow having about 27% more excess expenditure than households without any borrowing. This number is consistent across the other specifications as well. This result is not surprising as in this data set reported income does not include loans, while it would be expected that expenditure would reflect the additional income due to loans. Including the borrowing dummy variable, however, does not affect our result that

0.081 (0.066) 0.292 (0.055)

0.083 (0.039) 0.276 (0.052)

0.121 (0.041)

Ethnic

No. of elderly (65 þ )

0.003 (0.015)

0.004 (0.012)

0.015 (0.012)

0.290 (0.057)

0.080 (0.066)

0.006 (0.015)

0.174 (0.231)

0.047 (0.061)

0.173 (0.054) 0.145 (0.252)

0.064 (0.069)

0.265 (0.098)

0.585 (0.080)

0.274 (0.063)

(6)

0.083 (0.070)

No. of children (o15)

0.066 (0.192)

0.060 (0.051)

0.021 (0.065)

0.264 (0.097)

0.617 (0.084)

0.278 (0.063)

(5)

0.009 (0.203)

0.023 (0.047)

0.022 (0.063)

0.260 (0.090)

0.428 (0.071)

0.325 (0.050)

(4)

Working sample

Single

0.001 (0.047)

University

0.279 (0.087)

0.276 (0.088) 0.004 (0.065)

0.369 (0.072)

0.375 (0.070)

0.369 (0.070)

0.260 (0.052)

0.265 (0.053)

0.262 (0.053)

(3)

(2)

Full sample

Regression Results.

(1)

Vocational

Borrowing

Migrant (returned)

Migrant (abroad)

Variables

Dependent Variable: Log (Expenditure/Income)

Table 4.

220 ILHOM ABDULLOEV ET AL.

4,391 0.012

1.047 (0.024)

Significance at 0.05; Significance at 0.01

Observations R2

Constant

Self-employed Professional

Professional

Self-employed

Land

Urban

4,391 0.014

1.033 (0.028) 4,391 0.017

0.955 (0.042)

0.036 (0.048)

0.198 (0.043)

4,391 0.040

2,799 0.051

0.812 (0.088)

0.268 (0.076)

0.294 (0.064)

1.013 (0.075)

0.111 (0.076)

0.124 (0.063)

2,799 0.058

0.700 (0.093)

0.323 (0.100)

0.349 (0.073)

0.086 (0.059)

0.270 (0.076)

0.119 (0.077)

Migration as a Substitute for Informal Activities 221

222

ILHOM ABDULLOEV ET AL.

migration significantly decreases the amount of informal sector or nonreported income for households. The education variables are not significant and therefore do not appear to impact the excess expenditures of a household. Regression (3) augments regression (2) with some household characteristic variables. The new variables that are included are an indicator variable on whether the household head is single and variables that indicate the number of young people in the house and the number of elderly people in the house. The marital status of the head of household and the number of young people under the age of 15 in the household do not significantly affect the excess expenditure for the household but the number of elderly does. This is not surprising as the pension paid to retirees in Tajikistan is very low and below the subsistence wage. Thus, the elderly would need to augment their income. If this additional income was not reported in the formal income, then we would expect to see an increased excess expenditure for the household as the augmented income would be likely to show up in the household expenditure. Regression (4) adds in other household characteristics including the ethnicity of the head, whether the household is situated in an urban area, whether the household has access to cultivatable land, and whether any member of the household is self-employed. All of these variables significantly affect the excess expenditure of a household. Households with an ethnic minority head have excess expenditures that are 27.6% lower than Tajik households. Households that are located in urban areas have excess expenditures that are 12.4% less than households in nonurban areas, while households that have access to cultivatable land have increased excess expenditures of the order of 29.4%. Finally, households with members who are self-employed have lower excess expenditures to the order of 19.8%. The result that ethnic minority households have significantly lower excess expenditures suggests that these households are not able to generate as much informal sector income as other households. The results for the urban households and households who have access to cultivatable land are consistent with each other. Households in rural areas have more opportunity to grow their own food that is included in the expenditures but not included in incomes. The consistent result, however, is that households with migrants have significantly lower excess expenditure than households without migrants. While not all of the discrepancy in reported expenditure over income can be attributed to informal sector activity, it is hard to believe that informal sector activity does not make up a large proportion of the discrepancy.

Migration as a Substitute for Informal Activities

223

Other sources of the reported discrepancy are likely due to unreported consumption of assets (animal stock, food, etc.) or to incorrect pricing of such activities. As reported in Tables 2 and 3 the magnitudes in the differences in excess expenditure between households with migrants and households without migrants are large. It is very hard to believe that this difference is due to differences between households in pricing of in-kind consumption or consumption of assets. In particular, it is hard to believe that households with migrants are systematically better (or worse) at pricing nonmarket activities than households without migrants to such an extent as to explain the large change in excess expenditure seen in the data. Therefore, the decline in the discrepancy between expenditures and income is likely to be due to the fact that remittances are explicitly measured in this survey, while informal sector income is not explicitly measured. Thus, our results are consistent with the hypothesis that remittances from migrants are substitutes for informal sector income rather than complements. Our regressions include controls for the other sources of income that are not included in the income data including loans and access to growing or rearing you own food. One source of additional income that we do not control for is additional income obtained by professionals ‘‘under-the-table’’. This income could not be considered informal sector income as the income is derived from activity that is identical to their formal sector income; the only difference is that it is not reported. In order to control for this, we include a dummy variable for occupation, but since households who are not working or unemployed do not report their occupation, we include only working households in our sample. Regressions (5) and (6) are results using data from this restricted sample. Regression (5) is the same as Regression (4) with the only difference being the estimation sample, while Regression (6) includes variables pertaining to a household’s occupation; in particular, a dummy variable if the household works in a professional occupation and the interaction between the professional dummy variable and the self-employment dummy variable. The results are consistent with the first set of results that show that households with current or recently returned migrants having significantly lower excess expenditures than households without migrants. We also see that when restricting attention to only workers the significance of the elderly variable and the urban dummy variable disappears. This is not surprising in that we lose those households who are retired and not working and those households in the rural areas that are not working. The new result is that households whose head works in a professional occupation have significantly higher excess expenditure, thus suggesting that there is additional

224

ILHOM ABDULLOEV ET AL.

income being collected ‘‘under-the-counter’’. This is reinforced by the result that professional but self-employed households do not show a significant increase in excess expenditure. This is consistent with professionals who work in formal (and nonself-employed) jobs need to augment their income as their pay is low in their formal jobs. The results reported above show that the discrepancy between reported expenditures and income for migrant households are significantly lower than those for nonmigrant households. A possible criticism of the methods used would be that there are possible endogeneity biases present in our estimates that we have not modeled. However, in our specification the dependent variable is the amount of discrepancy between reported expenditure and reported income, and it is not clear that excess expenditure over income would drive the decision to migrate. Certainly the total amount of expenditure (or income) might influence the migration decision, but it is not clear that the component of expenditure that is unreported income would be a driver of the migration decision. Households would have to care about whether and from where their income was sourced for the excess expenditure over income to cause the migration decision.6

CONCLUSION Over the last 20 years, Tajikistan experienced both increasing migration and informal sector employment. They both are relatively new phenomena in Tajikistan, a former Soviet country, where informal sector employment and international migration were strictly controlled and even ‘‘prohibited’’ by the Soviet Government. After the Soviet Union’s collapse, these restrictions quickly untwisted, now involving an appreciably large proportion of the population. Such preconditions make Tajikistan a good case to study, where one does not need to be very concerned about historically well-established patterns and traditions of migration and informality (or look further into the historical and cultural elements of these processes), allowing us to focus on economic issues and factors that help to explain how these two processes interact. Moreover, the very large size of the migration has made it a relative low-cost path for obtaining additional income, and there are many households in the sample who have a current or recently returned migrant. We consider informality and migration as alternative income sources for the household – the two are part of the portfolio of the family. When informality is considered in the context of migration, it is almost always in the context of the migrant working in an informal job, or not. In the

Migration as a Substitute for Informal Activities

225

household model we are implicitly considering in this paper, it is quite reasonable that one member of the household might migrate while another works in the informal sector. This is in line with the portfolio theory of migration, in which family members work in different labor markets as an income diversification strategy. The LSMS allows us to investigate the relationship between external migration and local informal sector activity. To do so we used the discrepancy between reported household expenditure and reported household income as an indication of informal/unreported activity. We understand that all of this discrepancy is not due to only informal sector activity. However, variation across household in this discrepancy is most likely due to differences in informal sector activity – broadly defined – other sources for this discrepancy between expenditure and income, such as measurement errors and memory retention error, are not likely to differ systematically across households. Using this measure of discrepancy between expenditure and income, we investigate the linkages between migration and the size of informal sector activity. We do this by estimating an equation that explains the discrepancy between expenditure and income using household characteristics and migration status. The overall result that we find is that there is consistent evidence that migration (accompanied by remittances) and home region informal sector activities are largely substitutes for one another. This result is robust across all of our regression specifications. We find that households with members who have externally migrated are less likely to participate in the domestic informal sector in that they have a significantly lower discrepancy between their reported expenditure and reported income. We do not believe that there is anything special about migrant households that enable them to better measure or remember their expenditure or income than nonmigrant households, so that we conclude that the significant drop in the discrepancy between expenditure and income is due to the substitution of remittances, which are observed, for informal sector income that is not observed. Our work indicates that the ability of professionals to engage in informal activities enables them to compensate for the discrepancy between their expenditure and income without migrating. Migrants typically are lowskilled nonprofessionals without postsecondary education who lack informal sector opportunities or might have lower earnings in the local informal sector. Migrants find it less costly to migrate (more earning opportunities are abroad) than to be involved in local informal sector. Migration becomes a substitute for informal sector employment. We have documented the existence of this phenomenon and suggested some ways to understand its source. More work is needed – other case

226

ILHOM ABDULLOEV ET AL.

studies, modeling how and why this form of the link arises. The result adds a considerable amount of complexity to our understanding of the decisions faced by households in less-developed economies. Work on informality and work on migration should not continue to ignore the connection.

NOTES 1. The presidential election and the first multiparty elections were held in Tajikistan in 1999 and 2000 respectively, after reaching the reconciliation between confronting parties in 1997. 2. Recall and measurement error may also play a role here. To minimize this, one could consider the difference between expenditure and income as indicating informal sector activity only if expenditure is significantly more than income (Dimova et al., 2006). 3. http://go.worldbank.org/IPLXWMCNJ0 4. This is consistent with the story that migrants who are currently abroad may not have received their full compensation and so their remittances are less than the migrants who have returned and earned their full salary. 5. Using bootstrapped standard errors allow for us to control for unobserved heteroskedasticity without the need to commit to the exact form or commit to the clustering variable needed to compute clustered–robust standard errors. 6. As a check, we have tried some possible instruments, including education variables, and find that the negative sign on migration is robust but that the magnitude is implausible large. The instruments that were tried had little theoretical motivation and were weak in the sense of having low first stage F-statistics. We therefore do not report the IV results for both of these reasons – that is, the available instruments are weak and most likely not valid.

REFERENCES Dimova, R., Gang, I., & Landon-Lane, J. (2006). The informal sector during crisis and transition. In B. Guha-Khasnobis & R. Kanbur (Eds.), Informal labour markets and development (pp. 88–108). Basingstoke: Palgrave-McMillan Press. Falkingham, J. (2004). Poverty, out-of-pocket payments and access to health care: Evidence from Tajikistan. Social Science and Medicine, 58(2), 247–258. Fields, G. (1975). Rural-urban migration, urban unemployment and underemployment, and job-search activity in LDCs. Journal of Development Economics, 2(2), 165–187. [Elsevier] Fields, G. (1976). Labor force migration, unemployment and job turnover. The Review of Economics and Statistics, 58(4), 407–415. [MIT Press] Fields, G. (1979). Place-to-place migration: Some new evidence. The Review of Economics and Statistics, 61(1), 21–32. [MIT Press] Gang, I., & Gangopadhyay, S. (1987a). Employment, output and the choice of techniques: The trade-off revisited. Journal of Development Economics, 25(2), 321–327. [Elsevier]

Migration as a Substitute for Informal Activities

227

Gang, I., & Gangopadhyay, S. (1987b). Optimal policies in a dual economy with open unemployment and surplus labour. Oxford Economic Papers, 39(2), 378–387. [Oxford University Press] Gang, I. N., & Gangopadhyay, S. (1987c). Welfare aspects of a Harris-Todaro economy with underemployment and variable prices. The Developing Economies, 25, 203–219. Harris, J., & Todaro, M. (1970). Migration, unemployment and development: A two-sector analysis. American Economic Review, 60(1), 126–142. International Labor Organization (ILO). (2010). Migrant remittances in Tajikistan. Moscow: ILO Sub Regional Office for Eastern Europe and Central Asia, International Labour. Kinyanjui, M. N. (2010). Social relations and associations in the informal sector in Kenya. Social Policy and Development Programme Paper No. 43, United Nations Research Institute for Social Development, New York. Noda, S. (1999, March). Aftermath of war in Tajikistan: Country profile. ILO News Eastern Europe and Central Asia, p. 3. Perry, G., Maloney, W., Arias, O., Fajnzylber, P., Mason, A., & Saavedra-Chanduvi, J. (2007). Informality: exit and exclusion. Washington, DC: World Bank. Robertson, L. R. (Ed.). (1999). Russian & Eurasia facts & figures annual (Vol. 25(2)). Gulf Breeze, FL: Academic International Press. Statistical Agency of Tajikistan. (2011). Database. Retrieved from Statistical Agency of Tajikistan: http://stat.tj/ru/database/real-sector/. Accessed on September 17, 2011. Statistical Committee of CIS. (2011). Average monthly nominal wage in the CIS countries, in national currency. Retrieved from Statistical Committee of CIS: http://www.cisstat.com/ index.html. Accessed on September 17, 2011. Thorbecke, E. (1999). A dual-dual framework to analyze intersectoral relationships throughout the development process. In G. Ranis, Y.-P. Chu & S.-C. Hu (Eds.), The political economy of comparative development into the 21st century. Cheltenham: Edward Elgar Publishing, Ltd. [Chapter 5] United Nations. (2011). Population below national poverty line, total, percentage. Retrieved from http://data.un.org/Data.aspx?d¼MDG&f¼seriesRowID%3A581. Accessed on September 17, 2011. United Nations Development Programme. (2007). National human development report 2007: Informal economy in Tajikistan. Dushanbe: United Nations Development Programme. United Nations Development Programme and the Center for Strategic Studies under the President of the Republic of Tajikistan. (2010). Corruption in Tajikistan – Public opinion. Retrieved from UNDP: http://www.undp.tj/files/undpeng.pdf. Accessed on October 17, 2011. World Bank. (2011a). Migration and remittances. Factbook 2011, 2nd ed. Retrieved from World Bank: http://siteresources.worldbank.org/INTLAC/Resources/Factbook2011-Ebook.pdf. Accessed on August 12, 2011. World Bank. (2011b). World data bank. Retrieved from World Bank: http://databank.world bank.org/ddp/home.do?Step¼12&id¼4&CNO¼2. Accessed on September 17, 2011.

CHAPTER 7 THE PERSISTENCE OF INFORMALITY: EVIDENCE FROM PANEL DATA Alpaslan Akay and Melanie Khamis ABSTRACT Informality is a growing phenomenon in the developing and transition country labor market context. In particular, it is noticeable that working in an informal employment relationship is often not temporary. The degree of persistence of informality in the labor market might be due to different sources: structural state dependence due to past informality experiences and spurious state dependence due to time-invariant unobserved individual effects, which can alter the propensity of being in the informal sector independently from actual informality experiences. The purpose of our paper is to study the dynamics of informality using a genuine panel data set in the Ukrainian labor market. By estimating a dynamic panel data probit model with endogenous initial conditions, we find a highly significant degree of persistence due to previous informality experiences. This result implies that policies attempting to reduce current levels of informality may have a long-lasting effect on the labor market.

Informal Employment in Emerging and Transition Economies Research in Labor Economics, Volume 34, 229–255 Copyright r 2012 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1108/S0147-9121(2012)0000034010

229

230

ALPASLAN AKAY AND MELANIE KHAMIS

Keywords: Informality; unobserved individual effects; state dependence; transition country JEL classification: D60; I31

INTRODUCTION Informality in the labor market is a prevalent phenomenon in the developing, middle-income, and transition country context. A large proportion of the labor market participants is considered to be working informally at one point in time or permanently in the sector. Subjective data obtained from informality surveys conducted by the World Bank (2007) highlight that the reasons and motivations for working and remaining in an informal employment relationship can be manifold and summarized as follows: institutional barriers, type of human capital, preferences, tastes, ability, wage differences, and previous informality experiences among others.1 Some of these market or individual specific factors may lead the informality status of workers to be persistent over time. For instance, institutional barriers to enter into the formal sector such as registration costs or employers not wanting to register workers could be important for the persistence of informality status (De Soto, 1989). Lack or limited enforcement of registration or cumbersome rules and regulation may also provide incentives for employers to offer informal jobs leading workers to participate and remain in the informal sector for extended spells (De Soto, 1989; Kanbur, 2009). Individual human capital considerations may also be an important factor to enter and stay in the informal sector. For some individuals, the informal sector serves as a training ground to gain skills in order to get a job in the formal sector in later years (Maloney, 2004). Contrary to this, it could also be that working in the informal sector might be perceived as a low-skill task and a signal for low productivity, which renders the individual to be unsuccessful in future job searches in the formal sector. One of the other reasons of persistence in the informality status may be due to different valuation of the benefits of social security system by workers in the informal sector. Workers may be myopic toward the future and not value the benefits of social security. Individuals might not find it beneficial to contribute to the social security system due to a history of noncontribution. This can occur when the cost to pay social security at present is higher,

The Persistence of Informality: Evidence from Panel Data

231

or at least perceived to be higher, than the benefits received in the future (Levy, 2008). These factors may characterize the labor market and lead to persistence in the informality status once the individuals enter into the informal sector. The aim of this paper is to examine whether there is persistence in the informality status of workers in the Ukrainian labor market. Particularly, we aim to disentangle structural state dependence that is caused by past informality experiences of workers from spurious state dependence that is caused by other sources of persistence such as time-invariant unobserved individual characteristics. Individuals might have different unobserved motivations and preferences for flexible work hours and different valuations of the social security benefits (Dohmen, Khamis, & Lehmann, 2011; Maloney, 2004). Individuals may also prefer specific tasks in a job that might only be possible in an informal arrangement (De Mel, McKenzie, & Woodruff, 2010; Maloney, 2004). Hence, these unobserved characteristics might also alter the propensity of being in the informal sector independently from actual informality experiences (Heckman, 1978, 1981a, 1981b). In order to assess the degree of persistence, we estimate a dynamic randomeffects probit model controlling for past informality experience, observed, and unobserved individual characteristics. It is also very important to consider the initial values problem in the estimation of such a model (Heckman, 1981a, 1981b; Wooldridge, 2005). This problem occurs when the process generating the informality states is not observed from the beginning and the initial values could be endogenously determined by observed and unobserved time-invariant individual characteristics. It is crucial to deal with the endogenous initial values especially with a short panel data set (as we have in this paper) to identify the structural state dependence. We deal with this problem using two existing methods: Heckman’s (1981a, 1981b) reducedform approximation and Wooldridge’s (2005) method. Our results can be summarized as follows: the parameter of past informality experience, that is, structural state dependence, is highly significant on the current informality status in the context of the Ukrainian labor market. Our analysis suggests that a failure to control for the endogenous initial values leads to a wrong inference: the models seriously overestimate the structural state dependence, and also the variance of the unobserved individual effect is not identified. Controlling for the endogenous initial values generates lower estimates of structural state dependence and a sizeable variance of the unobserved individual effect. The models estimated here lead to nonlinear conditional expectations, and thus we calculate the average partial effect of past informality experience for various important groups of interest. Our

232

ALPASLAN AKAY AND MELANIE KHAMIS

analysis suggests that a past informality experience increases the current probability of being in the informal sector by 7–9 percentage points, ceteris paribus, compared to the workers who do not have previous informal sector experience. A detailed analysis suggests that the past informal sector experience is more pronounced on the current informality status for young single males with low education. The remaining part of the paper is organized as follows: in the second section the data source and descriptive statistics are presented. The third section introduces the dynamic random-effects probit model and the two solution methods for the initial values problem: the Heckman’s (1981a, 1981b) reduced-form approximation and the method of Wooldridge (2005). The fourth section presents the results. The fifth section discusses the robustness checks. The sixth section concludes.

DATA The Survey To understand the nature of informality over time, we employ the Ukrainian Longitudinal Monitoring Survey (ULMS) for the years 2003, 2004, and 2007. The survey is nationally representative of the Ukrainian work force and contains household and individual questionnaires. The household questionnaire elicits responses regarding income and expenditure patterns and living arrangements and conditions. The individual questionnaire contains detailed information on various individual characteristics and a large section on current and past labor market experiences. From this individual section, we create our dependent variable, informality in the labor market: we focus on informal employment relationships and do not consider other measures of informality in our paper.2 Informality in our data is defined according to nonregistration of work contract, while formal workers are registered. We generate our dependent variable using answers to the following question: Tell me, please, are you officially registered at this job, that is, on a work roster, work agreement, or contract? The answer to the question is either registered (formal) or not registered (informal). Therefore, our dependent variable is an indicator variable that takes the value 1 if the individual is an informal worker and 0 otherwise. The data set includes both employees and selfemployed individuals working either formally or informally. However, we have restricted our analysis to an employee-only sample as this provides the

The Persistence of Informality: Evidence from Panel Data

233

most clear-cut way to distinguish formal and informal employment structure in the labor market. In addition to data on registration, the ULMS allows us to exploit a rich set of explanatory variables in our analysis. Age (from age 15 onward), gender (female/male dummy), marital status (married, single, divorced/ separated/widowed), education levels (10 levels), total household income, sector of occupation (8 different sectors: agriculture, industry, sales, transportation, public administration, education, services, and other sectors that include occupations not included in the previous categories), and regions (Kiev, Center, East, West, South) are the variables that we use in our empirical analysis. These variables are standard in the literature on informality and have also been employed in previous studies with the ULMS (Dohmen et al., 2011; Lehmann & Pignatti, 2007). One of the important limitations of the data set is that the duration between waves is different. We include several history variables, which account for the time gap in the panel data set between waves 2004 and 2007. These variables indicate whether individuals changed jobs, marital status, or residence during that time period not covered by the main survey. We later use these variables to test the sensitivity of our results.

Descriptive Statistics In Table 1, the descriptive statistics of the selected variables from the ULMS data set are presented. The second column presents the statistics for the whole sample. The third and fourth columns relate to the characteristics of individuals split by informal and formal workers. Here, informal workers are counted as the ones that experience at least one spell of informality during the three waves. Formal workers are the number of observations relating to formal work experience, with no informality experience during the three waves of the ULMS. The last three columns of Table 1 show the characteristics of individuals at a point in time of each wave, 2003, 2004, and 2007. About 4 percent of our sample is considered working in the informal sector as an employee at some point during the three waves, while the remaining observations are entirely formal over this time period. Our sample is restricted only to formal and informal employees, and also the data set is organized to be a balanced panel for our empirical analysis.3 Thus, the mean informality rate is lower than the mean informality levels reported in the other studies using the same data set (Lehmann & Pignatti, 2007). One important concern is that the attrition in the sample at use might

South

East

West

Center

Regions Kiev

Age

Education level

Male

Female

Divorced/separated/widowed

Single

Married

Log(household income)

Informal dummy

0.052 (0.222) 0.242 (0.428) 0.152 (0.359) 0.308 (0.462) 0.246 (0.431)

0.041 (0.198) 6.797 (0.989) 0.704 (0.456) 0.130 (0.336) 0.166 (0.372) 0.527 (0.499) 0.473 (0.499) 6.554 (1.829) 42.429 (10.743)

Whole Sample

0.046 (0.210) 0.167 (0.373) 0.093 (0.290) 0.343 (0.475) 0.352 (0.478)

– – 6.795 (0.863) 0.559 (0.497) 0.265 (0.442) 0.176 (0.381) 0.426 (0.495) 0.574 (0.495) 5.481 (1.860) 35.972 (10.028) 0.052 (0.223) 0.248 (0.432) 0.157 (0.364) 0.305 (0.460) 0.237 (0.425)

– – 6.797 (1.000) 0.717 (0.450) 0.118 (0.323) 0.165 (0.371) 0.536 (0.499) 0.464 (0.499) 6.649 (1.796) 43.001 (10.617)

Formal Workers

Descriptive Statistics.

Informal Workers

Table 1.

0.052 (0.222) 0.242 (0.428) 0.152 (0.359) 0.308 (0.462) 0.246 (0.431)

0.038 (0.192) 6.237 (0.793) 0.701 (0.458) 0.148 (0.356) 0.151 (0.358) 0.527 (0.499) 0.473 (0.499) 6.320 (1.969) 40.755 (10.606)

2003

0.052 (0.222) 0.242 (0.428) 0.152 (0.359) 0.308 (0.462) 0.246 (0.431)

0.044 (0.206) 6.627 (0.619) 0.687 (0.464) 0.136 (0.342) 0.178 (0.382) 0.526 (0.499) 0.474 (0.499) 6.609 (1.771) 41.774 (10.619)

2004

0.052 (0.222) 0.242 (0.428) 0.152 (0.359) 0.308 (0.462) 0.246 (0.431)

0.040 (0.196) 7.528 (1.023) 0.725 (0.447) 0.106 (0.308) 0.169 (0.375) 0.526 (0.499) 0.474 (0.499) 6.733 (1.714) 44.758 (10.605)

2007

234 ALPASLAN AKAY AND MELANIE KHAMIS

0.187 (0.390) 0.025 (0.156) 0.030 (0.171) 3,984

0.080 (0.271) 0.257 (0.437) 0.078 (0.268) 0.102 (0.303) 0.047 (0.212) 0.265 (0.441) 0.064 (0.244) 0.107 (0.310) 0.565 (0.497) 0.028 (0.165) 0.046 (0.210) 324

0.102 (0.303) 0.204 (0.403) 0.321 (0.468) 0.068 (0.252) 0.015 (0.123) 0.037 (0.189) 0.114 (0.319) 0.139 (0.346)

Source: Ukrainian Longitudinal Monitoring Survey (2003, 2004, & 2007). Note: Standard deviations are in parenthesis.

No. of observations

Residential change

Marital status change

History 2004–2007 Job leavers

Other

Services

Education

Public administration

Transportation

Sales

Industry

Sectors Agriculture

0.153 (0.360) 0.025 (0.155) 0.029 (0.167) 3,660

0.078 (0.268) 0.262 (0.440) 0.057 (0.231) 0.105 (0.306) 0.050 (0.218) 0.285 (0.451) 0.059 (0.236) 0.105 (0.306) – – – – – – 1,328

0.084 (0.278) 0.259 (0.438) 0.081 (0.272) 0.098 (0.297) 0.042 (0.201) 0.269 (0.444) 0.067 (0.250) 0.100 (0.300) – – – – – – 1,328

0.089 (0.285) 0.262 (0.440) 0.073 (0.260) 0.106 (0.308) 0.048 (0.214) 0.262 (0.440) 0.061 (0.239) 0.099 (0.298) – – – – – – 1,328

0.066 (0.248) 0.251 (0.434) 0.081 (0.272) 0.102 (0.302) 0.051 (0.220) 0.264 (0.441) 0.063 (0.244) 0.123 (0.329)

The Persistence of Informality: Evidence from Panel Data 235

236

ALPASLAN AKAY AND MELANIE KHAMIS

be nonrandom relating to informality. We find that the informality dummy shows a very stable pattern over time that is in line with informality patterns in the nonrestricted sample with self-employed individuals. The mean informality rates in our sample selection are 0.038, 0.044, and 0.040 for 2003, 2004, and 2007, respectively. The descriptive statistics for informal and formal workers are similar to the findings from the literature in the Ukrainian context using the same data set (Dohmen et al., 2011). The characteristics of informal workers differ on several dimensions from workers who were in the formal sector over the three waves. Average age of the overall estimation sample is about 42 years, while the informal workers are younger at about 35 years of age. Informal labor market participants are less likely to be married. More male workers are working in the informal sector. Education levels of informal sector workers are lower than education levels of formal sector participants. Looking at additional characteristics (history variables) for the period between 2004 and 2007, we find that about 2–3 percent of individuals report to have changed residence or to have gotten married. About 20 percent report a job change during that period and this is larger for the ones who worked at some point in the informal sector during this time period. In order to give an impression of the dynamics of informality states in the Ukrainian labor market, we present in Table 2 the run patterns of informality status using only the estimation sample. Here, (0, 0, 0) means that an individual is formally employed across the three waves. On the other

Table 2.

Run Patterns of Informality in Ukraine.

Run Patterns 2003 0 0 0 0 1 1 1 1

2004

2007

0 0 1 1 0 0 1 1

0 1 0 1 0 1 0 1

No. of Observations

%

3,660 78 66 27 51 18 48 36 3,984

91.87 1.96 1.66 0.68 1.28 0.45 1.20 0.90

Note: 1 is informally employed and 0 indicates formal employment based on the above definitions.

The Persistence of Informality: Evidence from Panel Data

237

hand, the triple (1, 1, 1) implies the individual is informally employed during the three waves and this indicates a high degree of persistence. The other possibilities indicate lower levels of persistence to and from informality to formality or vice versa.

ECONOMETRIC FRAMEWORK The Model In order to distinguish structural state dependence from other sources of persistence, we specify a dynamic random-effects probit model by controlling for previous period informality status, observed and unobserved individual characteristics, and endogenous initial values. The dynamic panel random-effects model is specified as follows: d it ¼ 1ðxit b þ ld i;t1 þ uit 40Þ

(1)

uit ¼ Zi þ it  d i1 ¼ 1 xi1 b1 þ ui1 40

(2) (3)

where dit is a binary-dependent variable indicating whether an individual i is informally employed during the current period t (where i ¼ 1; . . . ; I and t ¼ 1; 2; 3), xit is a vector of current sociodemographic and economic characteristics (such as education and marital status), b is the corresponding vector of parameters to be estimated, d i;t1 is an observed binary variable indicating whether an individual i was in the informal sector during the previous period (t1), and the parameter l represents the structural state dependence following Heckman (1981a, 1981b). The error term in the model (1) and (3) has two components as displayed in Eq. (2). The first term (Zi) captures the time-invariant unobserved individual effects (such as motivation and ability). To control for these characteristics is crucial in order to be able to identify structural state dependence (l). The second term (eit) is the usual error term, which is assumed to have a normal distribution with zero mean and unit variance due to the identification of a discrete choice model. The actual disturbance process is assumed to be serially uncorrelated. However, in this model controlling for unobserved individual effects automatically induces a serial correlation. The correlation between two sequential error terms is, Corrðuit ; uis Þ ¼ s2Z =s2Z þ 1, ðt; s ¼ 1; . . . ; T i ; tasÞ, where s2Z is the variance of

238

ALPASLAN AKAY AND MELANIE KHAMIS

unobserved individual effects to be estimated. We follow a fully parameterized random-effects approach with a maximum likelihood estimator. The log-likelihood function that is used in the estimation process is as follows: ) "Z ( # I T 1 Y X T ln f 1 ðd i1 jfxit gt¼1 ; Zi Þ f it ðd it jd i:t1 ; xit ; Zi ; bÞ f ðZi ÞdZi log L ¼ i¼1

1

t¼2

(4)

 f it ðd it jd i;t1 ; xit ; Zi ; bÞ ¼ F ð2d it  1Þðx0it b þ ld i;t1 þ sZ Zi Þ

(5)

where F is the distribution function of standard normal random variable and f 1 ðd i1 jfxit gTt¼1 ; Zi Þ is the conditional distribution of initial values. The specification of this distribution is necessary in order to be able to identify structural state dependence.

Initial Values Problem The likelihood function in Eq. (4) can be easily maximized using a Gaussian–Hermite quadrature when the conditional distribution of initial values f 1 ðd i1 jfxit gTt¼1 ; Zi Þ is known. However, the distribution is unknown since the system given in Eqs. (1)–(3) has started many periods before our sample panel data set was observed. In this case, the initial values would be endogenously determined with the evolution of parameters, observed and unobserved characteristics of individuals. In order to identify the structural state dependence and to disentangle it from other sources of persistence, the initial values should be considered as endogenous with a probability distribution conditioned on observed and unobserved individual characteristics. There are two main methods for doing this: Heckman’s (1981a, 1981b) reduced-form approximation and the Wooldridge’s (2005) method that is a simple alternative. Heckman’s method is based on available presample information with which the conditional distribution of initial values is approximated via a reduced-form. This approximation allows a flexible specification of the relationships between initial values, observed and unobserved individual characteristics. Wooldridge (2005) introduces a simple alternative to Heckman’s reduced-form approximation. The method suggests that the unobserved individual effects should be considered

The Persistence of Informality: Evidence from Panel Data

239

conditional on initial values and the time-varying exogenous variables in a similar way to the correlated random-effects model of Chamberlain (1984). Recent studies suggest that there may be differences in the magnitude of state dependence and the estimated variance of unobserved individual effects between these two methods, especially for very short panels (Akay, 2011; Arulampalam & Stewart, 2009). Our sample size is small with T ¼ 3, and we use both methods to check for the sensitivity of the results. We define the reduced form that is employed in the Heckman’s method as follows: d i1 ¼ 1ðxi1 y þ ui1 40Þ

(6)

ui1 ¼ cZi þ i1

(7)

and the conditional distribution of the initial values is approximated as

 f 1 ðd i1 jfxit gTt¼1 ; Zi Þ ¼ F ð2d i1  1Þðxi1 y þ csZ Zi Þ (8) We estimate the parameters b, y, c, and sZ simultaneously by inserting Eq. (8) into the likelihood function (4) without imposing any restrictions (Heckman, 1981a, 1981b; Hsiao, 2003). The main assumption in this particular formulation of Heckman’s reduced-form approximation method is that the first period error term ui1 is correlated with the unobserved individual effects Zi, but it is uncorrelated with uit for tW1. For the Wooldridge method, we first define an auxiliary distribution of unobserved individual effects as follows: Zi ¼ p0 þ p1 x i þ p2 d i1 þ ai

(9)

wherePp0 (a constant), p1, and p2 are parameters to be estimated, x i ¼ ð1=TÞ Ti¼1 xit is the within-means of time-variant exogenous variables, and ai is the new unobserved individual effect that is assumed to be normally distributed with ai jx i ; d i1  N½p0 þ p1 x i þ p1 d i1 ; s2a . Inserting Eq. (9) into Eq. (4) generates a conditional likelihood that can be estimated as a standard random-effects probit model. This method is also very similar to the Chamberlain’s (1984) correlated-effects (quasi-fixed-effects) model since the auxiliary distribution of unobserved individual effects includes some of time-variant exogenous variables. We only have limited information on the time-varying characteristics of the individuals. Moreover, we only have three waves and some time-varying variables such as education do not show enough variation in the sample across time. We mainly use age and income in the auxiliary distribution of unobserved individual effects given in Eq. (9).

240

ALPASLAN AKAY AND MELANIE KHAMIS

RESULTS Our main aim is to test whether there is significant structural state dependence on the probability of being employed in the informal sector. We first start with a standard static model in order to compare our results to the previous literature. We then extend our analysis to dynamic models in which we check the importance of endogenous initial values under the two different estimation methods. Finally, we test the robustness of our results.

Main Results We present the full set of parameter estimates from various model specifications in Table 3. In each of these specifications, the dependent variable is the current informality status and the models include previous period informality status together with various observed and unobserved individual characteristics. We first present a benchmark specification based on a static random-effects model, which is similar to models estimated in the previous literature (Arias & Khamis, 2008; Dohmen et al., 2011; Lehmann & Pignatti, 2007; World Bank, 2007). This allows us to validate the results from our data set and to see whether our selection generates similar results compared to the literature. We also examine whether timeinvariant unobserved individual characteristics explain an important proportion of the variation in the probability of being employed in the informal sector. This specification is presented in the second column (Model I).4 Most of the results are in line with the literature, which estimates the propensity of being in the informal sector in the Ukraine for the cross-section sample in 2007 or in 2003 and 2004 (Dohmen et al., 2011; Lehmann & Pignatti, 2007). In particular, the results relating to household income, marital status, education, and age exhibit similar patterns in our estimations compared to the results found in the literature (Dohmen et al., 2011; Lehmann & Pignatti, 2007). The household income is negatively correlated with the probability of being in informal sector, but it is not significant here. We find that the relationship between age and informality follows a U-shape, with a high propensity for informality during younger and older ages and a lower propensity for informality for the age cohorts in between. Marital status is significant and negatively correlated with the informal sector participation. Education tends to be a highly significant determinant of informality in the Ukraine. A higher number of years of education (higher education levels in our case) is negatively correlated with

East ¼ 1

West ¼ 1

Center ¼ 1

Age-squared

Age

Education

Female ¼ 1

Single ¼ 1

Married ¼ 1

Log (household income)

Lagged informality status ¼ 1

Variables

(0.092) 0.051 (0.160) 0.030 (0.199) 0.136 (0.126) 0.145 (0.033) 0.058 (0.152) 0.001 (0.002) 0.294 (0.332) 0.263 (0.347) 0.469 (0.318)

(0.103) 0.276 (0.201) 0.066 (0.251) 0.054 (0.172) 0.207 (0.042) 0.103 (0.133) 0.001 (0.002) 0.156 (0.409) 0.002 (0.434) 0.429 (0.387)

(0.092) 0.004 (0.165) 0.078 (0.203) 0.174 (0.129) 0.145 (0.034) 0.057 (0.153) 0.001 (0.002) 0.390 (0.350) 0.399 (0.365) 0.567 (0.337)

(0.210) 0.007

(0.146) 0.010

0.030

–

(0.072) 0.052 (0.078) 0.058 (0.090) 0.168 (0.104) 0.175 (0.034) 0.065 (0.066) 0.001 (0.001) 0.354 (0.240) 0.309 (0.259) 0.558 (0.238)

(0.194) 0.002

0.802

The dynamic probit model (Heckman’s method with correlated random-effects)

The dynamic probit model (Wooldridge method)

0.842

The dynamic probit model (correlated random-effects and exogenous initial values)

The static probit model (correlated random-effects model)

IV

III

1.398

II

Main Results.

I

Table 3.

The Persistence of Informality: Evidence from Panel Data 241

0.625 (0.318)

0.648 (0.390)

(0.139) 0.012 (0.153) 0.000 (0.002)

(0.182) 0.040 (0.144) 0.000 (0.002)

–

–

Log (household income)

Education

–

–

–

0.048

0.085

–

The dynamic probit model (correlated random-effects and exogenous initial values)

The static probit model (correlated random-effects model)

–

II

I

Heckman’s method reduced-form equation Age –

Mean (age-squared)

Mean (age)

Nuisance parameters Wooldridge’s method Initial period informality Mean (household income)

South ¼ 1

Variables

III

–

–

–

(0.140) 0.019 (0.155) 0.000 (0.002)

0.072

(0.217)

0.819

0.750 (0.338)

The dynamic probit model (Wooldridge method)

Table 3. (Continued )

(0.049) 0.201

0.057 (0.013) 0.122

(0.104) 0.005 (0.077) 0.000 (0.001)

0.049

–

0.767 (0.232)

The dynamic probit model (Heckman’s method with correlated random-effects)

IV

242 ALPASLAN AKAY AND MELANIE KHAMIS

–

–

–

–

–

–

Center

West

East

South

Constant

c

471.87 Yes

(0.131) 3,984 0.000 307.77 Yes

(0.039) 2,656 0.000

o0.001

–

–

–

–

–

–

–

–

300.78 Yes

(0.622) 2,656 0.000

0.002

–

–

–

–

–

–

–

–

494.31 Yes

(0.102) 2,656 0.000

(0.062) 0.760 (0.235) 0.583 (0.277) 0.547 (0.242) 1.067 (0.259) 0.352 (0.196) 0.132 (0.154) 0.057 (0.069) 3.447 (0.488) 0.488

Note: We use 30 quadrature points in Gauss–Hermite quadrature for the integrals in likelihood functions. [], [], [] indicate significance levels at 0.01, 0.05, and 0.10, respectively. Standard errors are in parenthesis.

No. of Observations Wald chi-squared test (p-value) Log-likelihood Sector dummies (7 dummies)

1.136

–

Single

Variance of the unobserved effects

–

Married

The Persistence of Informality: Evidence from Panel Data 243

244

ALPASLAN AKAY AND MELANIE KHAMIS

informality. Gender is not a significant determinant of informal sector participation in our initial model (Model I). The results also suggest that unobserved individual characteristics explain an important part of the variation in the probability of being employed in the informal sector (r ¼ 0.52). Our main aim is to identify the structural state dependence and to disentangle it from other sources of persistence. Hence, we are mainly interested in the magnitude and statistical significance of two key parameters: the parameter of the lagged informality status (l) and the variance of unobserved individual effects (s2a ). In the third column of Table 3, we present the estimation results obtained by using the exogenous initial values assumption (Model II). This specification implies that the first period state of being in the informal sector is exogenous (not a function of individual observed and unobserved characteristics) that is not a very plausible assumption itself. The parameter of lagged informality status, that is, structural state dependence, is estimated by the exogenous initial values assumption to be around 1.40 (Table 3, Model II), and it is highly significant. However, it is important to note that the variance of unobserved individual effect is not identified and it is estimated as zero, which is not expected given the large variance found in the case of the static randomeffects model (Model I). In the fourth and fifth columns of Table 3, we present results that control for the endogenous initial values using the Wooldridge’s and Heckman’s methods (Models III and IV). The Wooldridge’s method is a simple way to deal with the initial values problem in the dynamic probit models with random effects. However, it may generate bias for panels of very short durations (Akay, 2011; Arulampalam & Stewart, 2009). We also do not have many time-variant characteristics to specify a flexible conditional distribution for the unobserved individual effects. The estimated parameter of lagged-dependent variable is reduced to 0.84 that is almost half of the size of the parameter generated with the assumption of exogenous initial values. However, as before the variance of unobserved individual effect is not identified and estimated to be very close to zero. The preferred method in this paper to deal with the initial values problem is the Heckman’s reduced-form approximation given the available sample characteristics and the duration of the panel data set. Since we do not have additional presample information of individuals, we follow the suggestions of Heckman (1981a, 1981b) to use the first wave in the reduced-form Eq. (6). The characteristics we include are age, education, marital status, regions, and income. The estimated parameter of lagged informality status is highly

The Persistence of Informality: Evidence from Panel Data

245

significant and estimated as 0.80, which is even smaller compared to the Wooldridge method. An important result is that the variance of the unobserved individual effect is identified and estimated as 0.49. This reflects a large variance for the heterogeneity distribution among informal sector workers in Ukraine.

Average Partial Effects The results reported in Table 3 suggest that there is a highly significant structural state dependence of informality. However, the models estimated in this paper have nonlinear conditional expectations, and the ceteris paribus interpretation is only possible using the partial (marginal) effect of a variable. There are various ways of calculating partial effects. The nonlinearity inherent in the dynamic random-effects probit specification allows us to calculate the partial effects for each individual. The individual partial effects of lagged-dependent variable d i;t1 can then be calculated using the conditional expectation of the probit model:5 ^ i;t1 ¼ 1ÞÞ  Fðxit b^ þ lðd ^ i;t1 ¼ 0ÞÞ m^ d;1 ðxit ; d i;t1 Þ ¼ Fðxit b^ þ lðd it

(10)

where m^ it is the partial effect function, F is the cumulative distribution function of standard normal random variable, and b^ and l^ are the estimated parameters. A consistent estimator of the population average partial effect (ape) can be calculated by simply averaging each individual-time partial effects in the observed sample:6 apeðxit ; d i;t1 Þ ¼

N X T 1 X m^ d;1 ðxit ; d i;t1 Þ NT i¼1 t¼2 it

(11)

We present the estimated ape and their population averaged standard errors in Table 4. The ape generated by the model that assumes exogenous initial values is 0.20 which implies that previous period informality status increases the probability of being in the informal sector in the current period by 20 percentage points. It is large and highly significant as expected. The estimated ape of structural state dependence is substantially reduced once we control for the endogenous initial values. The average partial effect of lagged informality status on the current informality status is around 7–9 percentage point, ceteris paribus. In the second part of Table 4, we sort individual partial effects by using sociodemographic characteristics such as gender and marital status and

AgeW50

AgeW35 and ageo50

Ageo35

Divorced/separated/ widowed

Single

Married

High educated

Low educated

Males

Sorting by sociodemographics Females

Whole sample

Average Partial Effect of Lagged Informality Status for y

0.069 (0.036) 0.109 (0.052) 0.125 (0.057) 0.054 (0.031) 0.082 (0.042) 0.136 (0.061) 0.077 (0.038) 0.132 (0.059) 0.090 (0.045) 0.043 (0.027)

(0.060) 0.285 (0.074) 0.211 (0.066) 0.115 (0.050)

0.088 (0.044)

0.204 (0.064)

0.169 (0.059) 0.242 (0.070) 0.273 (0.074) 0.140 (0.055) 0.193 (0.063) 0.288 (0.076) 0.182

Wooldridge’s Method

(0.032) 0.112 (0.050) 0.066 (0.038) 0.027 (0.021)

0.050 (0.029) 0.085 (0.044) 0.100 (0.050) 0.036 (0.024) 0.060 (0.035) 0.115 (0.050) 0.058

0.067 (0.036)

Heckman’s Method

Average Partial Effects of Lagged Informality Status.

Exogenous Initial Values

Table 4.

1,077

1,905

1,002

660

518

2,806

2,074

1,910

1,886

2,098

3,984

No. of Observations

246 ALPASLAN AKAY AND MELANIE KHAMIS

0.186 (0.062) 0.420 (0.071) 0.220 (0.076) 0.132 (0.084) 0.067 (0.040) 0.355 (0.082)

0.073 (0.039) 0.224 (0.079) 0.090 (0.050) 0.049 (0.042) 0.019 (0.017) 0.169 (0.073)

0.052 (0.032) 0.182 (0.074) 0.067 (0.041) 0.033 (0.030) 0.011 (0.010) 0.139 (0.065)

Note: [], [], [] indicate significance levels at 0.01, 0.05, and 0.10, respectively. Standard errors are in parenthesis.

Services

Education

Public administration

Transportation

Sales

Sorting by sectors Industry

254

1,055

188

406

311

1,025

The Persistence of Informality: Evidence from Panel Data 247

248

ALPASLAN AKAY AND MELANIE KHAMIS

estimate the ape of lagged informality in each case. The estimated ape is significant in most cases and varies in size. It is clearly larger for males. We generate two dummy variables to indicate low educated (less than education level 7) and high educated individuals (education level 7 or higher). The estimated ape of lagged informality status is much larger among low educated individuals. We also sort individuals by marital status and calculate the ape of lagged informality status. It is larger for single people compared to married or divorced/separated/widowed individuals. One of the important findings is that the ape of lagged informality status is larger for young people, and it gradually is reduced by age. We also find that the ape of lagged informality status differs by sectors. As we would expect, it is not significant for the public administration and education sectors. However, it is substantially larger for the sale and service sectors.

ROBUSTNESS CHECKS Alternative Specifications The dynamic probit model with random effects relies on strict parametric assumptions for the auxiliary distribution of unobserved individual effects and initial values. To check the sensitivity of previous results with respect to these assumptions, we estimate a dynamic linear probability model using a Generalized Method of Moments (GMM) estimator with first differences (Arellano & Bond, 1991). This specification provides a semiparametric identification of the model parameters by eliminating the unobserved individual effects and initial values problem (Alessie, Hochguertel, & van Soest, 2004; Stewart, 2007). The full estimation results of the first-difference GMM estimator are presented in the second column of Table 5 (Model I).7 In this specification, we use the same set of independent variables as before. However, due to first differences, the time-invariant variables are swept away and the number of observations in the final estimation sample is also reduced. We experiment with various alternative sets of instruments including the initial informality status di1 as one of the instruments (Stewart, 2007). Table 5 reports the p-value of the overidentifying restrictions based on the Sargan test. The result suggests that the overidentifying restrictions are not rejected at any conventional significance level (p-value ¼ 0.979). This specification generates linear conditional expectations, and the estimated parameters directly correspond to the average partial effects. The ape of state dependence is around 0.36 that is

(0.554) 0.234 (0.284) 0.007 (0.098) 0.023 (0.173) 0.117 (0.211) 0.118 (0.137) 0.163 (0.035) 0.054 (0.159)

(0.520) 0.142 (0.284) 0.003 (0.098) 0.034 (0.167) 0.059 (0.206) 0.077 (0.134) 0.161 (0.035) 0.053 (0.157)

0.012 (0.011) 0.025 (0.011)

Education

Age

Female ¼ 1

Single ¼ 1

Married ¼ 1

0.002 (0.007) 0.002 (0.026) 0.013 (0.032) –

Log(household income)

–

Residential change between 2004 and 2007

(0.120) 1.504

(0.118) 1.279

–

0.711

–

Marital status change between 2004 and 2007

Wooldridge’s method Initial period informality Job leavers between 2004 and 2007

0.789 (0.208) 0.897 (0.218) 0.726

Wooldridge’s method

1.375 (0.148) –

Exogenous initial values

First-difference GMM estimation Arellano–Bond

III

0.358 (0.077) –

II

Model with History Variables

Probability Model I

Dynamic Probit Random-Effects

Dynamic Linear

(0.228) 0.013 (0.084) 0.028 (0.136) 0.092 (0.148) 0.098 (0.158) 0.187 (0.039) 0.058 (0.100)

(0.509) 0.216

(0.132) 1.428

0.765

0.863 (0.210) –

Heckman’s method

IV

Alternative Estimators and the Effect of Unequal Durations Between Periods in the Panel Data Set.

Lagged informality status ¼ 1

Variables

Table 5.

The Persistence of Informality: Evidence from Panel Data 249

0.979 –

Sargan test (p-value) Variance of the unobserved effects

401.84 Yes

(0.033) 2,656 0.000

703.66 Yes

(0.088) 2,656 0.000

0.001 (0.002) 0.539 (0.379) 0.606 (0.396) 0.836 (0.367) 0.941 (0.367) – 0.003

Wooldridge’s method

III

700.71 Yes

(0.125) 2,656 0.000

0.001 (0.001) 0.499 (0.354) 0.511 (0.368) 0.817 (0.338) 0.937 (0.335) – 0.429

Heckman’s method

IV

Note: We use 30 quadrature points in Gauss–Hermite quadrature for the integrals in likelihood functions. [], [], [] indicate significance levels at 0.01, 0.05, and 0.10, respectively. The last three models (II, III, and IV) are specified using Chamberlain approach for the unobserved individual effects. Nuisance parameters are not presented here but they can be reported upon request. Standard errors are in parenthesis.

– Yes

1,328 0.000

–

South

No. of Observations Wald chi-squared test (p-value) Log-likelihood Sector dummies (7 dummies)

–

0.001 (0.002) 0.434 (0.358) 0.450 (0.375) 0.715 (0.345) 0.792 (0.344) – 0.001

o0.0001 (0.0001) –

East

Exogenous initial values

First-difference GMM estimation Arellano–Bond

–

II

Model with History Variables

Probability Model I

Dynamic Probit Random-Effects

Dynamic Linear

West

Center

Age-squared

Variables

Table 5. (Continued )

250 ALPASLAN AKAY AND MELANIE KHAMIS

0.056 (0.030) 0.088 (0.044) 0.103 (0.050) 0.041 (0.025) 0.066 (0.035) 0.111 (0.051) 0.063 (0.032) 0.111 (0.051) 0.071 (0.037) 0.035 (0.023)

(0.055) 0.250 (0.072) 0.180 (0.061) 0.100 (0.046)

0.071 (0.037)

0.176 (0.060) 0.145 (0.054) 0.210 (0.067) 0.240 (0.071) 0.117 (0.049) 0.167 (0.058) 0.246 (0.072) 0.159

Wooldridge’s Method

Exogenous Initial Values

(0.031) 0.117 (0.054) 0.068 (0.037) 0.031 (0.018)

0.054 (0.029) 0.088 (0.044) 0.104 (0.049) 0.039 (0.024) 0.064 (0.034) 0.116 (0.053) 0.062

0.070 (0.036)

Heckman’s Method

Sensitivity Analysis: Average Partial Effects.

Note: [], [], [] indicate significance levels at 0.01, 0.05, and 0.10, respectively. Standard errors are in parenthesis.

AgeW50

AgeW35 and Ageo50

Ageo35

Divorced/separated/ widowed

Single

Married

High educated

Low educated

Males

Sorting by sociodemographics Females

Whole sample

Average Partial Effect of Lagged Informality Status for y

Table 6.

1,077

1,905

1,002

660

518

2,806

2,074

1,910

1,886

2,098

3,984

No. of Observations

The Persistence of Informality: Evidence from Panel Data 251

252

ALPASLAN AKAY AND MELANIE KHAMIS

larger than the results reported above. However, it is positive and highly significant, which is in line with the previous results. This result also confirms that there is a substantial structural state dependence on the informality status of workers in Ukraine.

Unequal Durations Between Periods We only have three years of panel data and these were collected in the years 2003, 2004, and 2007. The time duration for 2003–2004 and 2004–2007 differs. This may bias the estimated parameter of structural state dependence. In order to deal with this issue, we use a history module covered by the 2007 questionnaire about events that take place between 2004 and 2007. To test the sensitivity of previous results, we estimate models using three indicator variables for any job, marital status, and residential changes that occurred between 2004 and 2007. We then compare the magnitude of the structural state dependence and the variance of unobserved individual effects. The results are presented in the third, fourth, and fifth columns of Table 5 (Models II–IV). We find that the estimated parameters are almost the same as before. However, the estimated structural state dependence is slightly reduced (except Heckman’s reduced-form approximation specification). In order to check the sensitivity of the results further, we calculate the ape as previously reported. These results are presented in Table 6 for the whole sample and for some groups of interest. The estimated ape is slightly lower compared to Table 4, but the main results remain stable.

CONCLUSIONS The objective of this paper is to examine the dynamics of informality in the Ukrainian labor market. We attempt to identify persistence or structural state dependence due to past informality experience and to disentangle it from other sources such as the persistence due to unobserved time-invariant individual characteristics. To address this issue, we estimate a dynamic panel data random-effects model with endogenous initial values and quasi-fixedeffects specification based on three waves of genuine panel data on informality. The model that assumes initial informality status of individuals in the labor market to be exogenous substantially overestimates the persistence

The Persistence of Informality: Evidence from Panel Data

253

due to past participation in the informal sector. This assumption leads to zero variance for the unobserved influences on informality status. Once we account for endogenous initial values using Heckman’s reduced-form approximation, we find a significant degree of heterogeneity among the individuals in our sample and a lower degree of persistence due to past informality experience. We calculated individual partial effects for the whole population and for various subgroups and find that the effect of past informality experience on the current informality status is highly significant leading to 7–9 percentage points larger probability to be employed in the informal sector, ceteris paribus. The effect of the past informality experience is more pronounced for young single and less educated males. One of the limitations of this study is the short time dimension of the panel data set. This data set, according to our knowledge, is one of the few genuine panel data, which make it possible to study the dynamics of informality in the developing and transition country contexts. However, we cannot assess the precise nature of structural state dependence because the size of the sample reduces substantially when we consider the information on the reasons for being in the informal sector (e.g., whether nonregistration is voluntary or involuntary, following Dohmen et al., 2011). We also have not included transitions to and from unemployment or inactivity. With these limitations in mind, our results still provide significant insights into the nature of informality over time and the persistence of informality in the labor market in a transition country. Hence, current informality experience of an individual seems to have future effects on the individual’s labor market participation in the informal sector. Our results suggest that policies attempting to reduce current levels of informality may have a long-lasting effect on the labor market.

NOTES 1. The degree to which these reasons and motivations are voluntary or involuntary for the informal sector participants is part of a large debate in the informality literature. For a summary of this debate, see World Bank (2007). 2. This follows Kanbur (2009) who recommends the classification of informality according to a specific regulation. 3. The mean informality rate in the nonrestricted data set is 0.13 (0.09 for 2003, 0.13 for 2004, and 0.15 for 2007). We restrict the data set only to employees by deleting the self-employed individuals. The mean informality rate then becomes 0.086 (0.06 for 2003, 0.10 for 2004, and 0.10 for 2007). We also delete individuals who are observed less than three waves and also the missing values in various

254

ALPASLAN AKAY AND MELANIE KHAMIS

variables. This would explain why our reported informality rates are lower compared to the literature (e.g., Lehmann & Pignatti, 2007) 4. Note that all models estimated here are based on the assumption that the observed and unobserved characteristics are not orthogonal to each other. The correlation is controlled for using Chamberlain’s (1984) correlated-effects model (quasi-fixed-effects model). 5. We calculate the standard errors of the partial effects using the delta method. 6. One of the other alternatives is to integrate the partial effect function over the unobserved individual effects using Monte Carlo integration methods. In this paper, we assume that the effect of unobserved individual effect can be ignored after averaging the partial effects for the whole population (see Wooldridge, 2005). 7. We have estimated also a system-GMM specification and found similar results (Blundell and Bond, 1998). The results can be provided from authors upon request.

ACKNOWLEDGMENTS The authors would like to thank the participants of the IZA/World Bank Workshop on Institutions and Informal Employment in Emerging and Transition Economies in 2011 for valuable comments and suggestions. We also would like to thank the editors of the RLE and two anonymous referees.

REFERENCES Akay, A. (2011). Finite sample comparison of alternative methods for estimating dynamic panel data models. Journal of Applied Econometrics. doi:10.1002/jae.1254 Alessie, R., Hochguertel, S., & van Soest, A. (2004). Ownership of stocks and mutual funds: A panel data analysis. The Review of Economics and Statistics, 86, 783–796. Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The Review of Economic Studies, 58, 277–297. Arias, O., & Khamis, M. (2008). Comparative advantage, segmentation and informal earnings: A marginal treatment effects approach. IZA Discussion Paper no. 3916. Arulampalam, W., & Stewart, M. (2009). Simplified implementation of then Heckman estimator of the dynamic probit model and a comparison with alternative estimators. Oxford Bulletin of Economics and Statistics, 71, 659–681. Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87, 115–143. Chamberlain, G. (1984). Panel data. In Z. Griliches & M. D. Intriligator (Eds.), Handbook of econometrics (Vol. 2, pp. 1247–1320). Amsterdam: North Holland. De Mel, S., McKenzie, D., & Woodruff, C. (2010). Who are the microenterprise owners? Evidence from Sri Lanka and Tokman v. de Soto. In J. Lerner & A. Schoar (Eds.), International differences in entrepreneurship. NBER Conference Report, University of Chicago Press, Chicago.

The Persistence of Informality: Evidence from Panel Data

255

De Soto, H. (1989). The other path—The invisible revolution in the third world. London: I.B. Tauris & Co Ltd Publishers. Dohmen, T., Khamis, M., & Lehmann, H. (2011). Risk attitudes and the incidence of informality among workers. Manuscript. Unpublished Working Paper. Heckman, J. J. (1978). Simple statistical models for discrete panel data developed and applied to test the hypothesis of true state dependence against the hypothesis of spurious state dependence. Annales de l.INSEE, 30–31, 227–269. Heckman, J. J. (1981a). Heterogeneity and state dependence. In S. Rosen (Ed.), Studies in labor markets. Chicago, IL: University of Chicago Press. Heckman, J. J. (1981b). The incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic process. In C. Manski & D. McFadden (Eds.), Structural analysis of discrete panel data with econometric applications (pp. 179–196). Cambridge: MIT Press. Hsiao, C. (2003). Analysis of panel data (2nd ed.). New York, NY: Cambridge University Press. Kanbur, R. (2009). Conceptualizing informality: Regulation and enforcement. Indian Journal of Labour Economics, 52, 33–42. Lehmann, H., & Pignatti, N. (2007). Informal employment and labor market segmentation in transition economies: Evidence from Ukraine. IZA Discussion Paper no. 3269. Levy, S. (2008). Good intentions, bad outcomes: Social policy, informality and economic growth in Mexico. Washington, D.C.: Brookings Institution Press. Maloney, W. F. (2004). Informality revisited. World Development, 32, 1159–1178. Stewart, M. (2007). The interrelated dynamics of unemployment and low-wage unemployment. Journal of Applied Econometrics, 22, 511–531. Wooldridge, J. M. (2005). Simple solutions to the initial conditions problem in dynamic, nonlinear panel-data models with unobserved heterogeneity. Journal of Applied Econometrics, 20, 39–54. World Bank. (2007). Informality: Exit and exclusion. Washington D.C.: The World Bank.

CHAPTER 8 JOB SEPARATIONS AND INFORMALITY IN THE RUSSIAN LABOR MARKET Hartmut Lehmann, Tiziano Razzolini and Anzelika Zaiceva ABSTRACT In the years 2003–2008, the Russian economy experienced a period of strong and sustained growth, which was accompanied by large worker turnover and rising informality. We investigate whether the burden of informality falls disproportionately on job separators (displaced workers and quitters) in the Russian labor market in the form of informal employment and undeclared wages in formal jobs. We also pursue the issues whether displaced workers experience more involuntary informal employment than workers who quit and whether informal employment persists. We find a strong positive link between separations and informal employment as well as shares of undeclared wages in formal jobs. Our results also show that displacement entraps some of the workers in involuntary informal employment. Those who quit, in turn, experience voluntary informality for the most part, but there seems a minority of quitting workers who end up in involuntary informal jobs. This scenario

Informal Employment in Emerging and Transition Economies Research in Labor Economics, Volume 34, 257–290 Copyright r 2012 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1108/S0147-9121(2012)0000034011

257

258

HARTMUT LEHMANN ET AL.

does not fall on all separators but predominantly on those with low human capital. Finally, informal employment is indeed persistent since separating from an informal job considerably raises the probability to be informal in the subsequent job. Keywords: Job separations; informality; Russia JEL classification: J64; J65; P50

INTRODUCTION Russia experienced a period of strong economic growth between 1999 and 2008. This growth, manifesting itself in an average GDP growth rate of roughly 7 percent, was accompanied by substantial worker turnover in the Russian labor market, with annual job separations amounting to up to 20 percent (see Fig. 1). Parallel to these large separations rates, we see a continuous rise in informal employment and informal activities: the number of informally employed workers rose from roughly 8 million in 1999 to about 12 million in 2008, that is, from 13 to 18 percent of total employment (Gimpelson & Zudina, 2011). Schneider, Buehn, and Montenegro (2010) provide evidence that the shadow economy of Russia is large compared to other transition and emerging economies, amounting to roughly 41 percent of official GDP in 2007. Even if the shadow economy and informal employment are substantial, it could well be that they afflict predominantly marginal groups of the workforce. The descriptive statistics of dependent employees in 2009 in Table 1 show that the informally employed indeed have a worse labor market history and, in the case of educational attainment, worse characteristics than their formal counterparts. Preceding the job in 2009, informally employed have substantially longer nonemployment spells and a far lower share of university graduates. Still, nearly 12 percent of the informally employed have finished university education. What is in addition particularly striking in Table 1 is the lack of divergence regarding the other demographics. Thus, rising informal employment is an important phenomenon in the Russian labor market, which is clearly not restricted to marginal groups of the workforce. The main aim of this paper is to investigate the link between job separations and the incidence of informal employment. The first six rows of Table 1 seem to imply such a link since informal employees have roughly

Separations (sample: working age 15-59 years) 0.16 0.14 0.12 0.1 Layoffs

0.08

Quits

0.06 0.04 0.02 0 2003

2004

2005

2006

2007

2008

Layoffs (sample: working age 15-59 years) 0.035 0.03 0.025 0.02

Plant/firm closed Redundant

0.015 0.01 0.005 0 2003

2004

2005

2006

2007

2008

Quits (sample: working age 15-59 years) 0.16 0.14 0.12 Other

0.1

Voluntary

0.08

Parental leave Retirement

0.06

Temporary contract

0.04 0.02 0 2003

2004

2005

2006

2007

2008

Fig. 1. Separations and Layoffs. Source: Authors’ calculations based on RLMS supplement on displacement. Note: Our definition of working age deviates from the official definition, which is 16–59 for men and 16–54 for women.

260

HARTMUT LEHMANN ET AL.

Table 1.

Descriptive Statistics of Dependent Employees.

Variables

All Samples

Displ., 2008 Displ., 2007–2008 Displ., 2003–2008 Quits, 2008 Quits, 2007–2008 Quits, 2003–2008 Months nonempl., 2008 Months nonempl., 2007–2008 Months nonempl., 2003–2008 Age Male City Village Regional center Higher education Secondary education Primary education Children Marital status Moscow/St. Petersburg North-West Central-Volga South East Risk indicator Household income No. of observations

0.025 0.042 0.134 0.095 0.198 0.585 0.438

(0.155) (0.211) (0.394) (0.306) (0.473) (0.917) (1.844)

Employed Officially 0.022 0.039 0.122 0.086 0.184 0.551 0.352

(0.146) (0.205) (0.376) (0.291) (0.457) (0.881) (1.637)

Informal Employees 0.041 0.066 0.231 0.248 0.413 1.116 1.471

(0.199) (0.249) (0.511) (0.469) (0.626) (1.180) (3.310)

1.020 (3.771)

0.841 (3.367)

3.008 (6.736)

2.626 (8.253)

2.225 (7.459)

7.058 (13.625)

42.714 0.431 0.344 0.190 0.466 0.291 0.622 0.087 0.735 0.806 0.182

(9.130) (0.495) (0.475) (0.393) (0.499) (0.454) (0.485) (0.282) (0.787) (0.395) (0.385)

0.069 (0.253) 0.432 (0.495) 0.106 (0.308) 0.212 (0.409) 3.744 (2.816) 33,402.91 (22,074.41) 16,854

42.897 0.423 0.346 0.185 0.469 0.309 0.609 0.081 0.731 0.810 0.186

(9.091) (0.494) (0.476) (0.389) (0.499) (0.462) (0.488) (0.273) (0.788) (0.392) (0.389)

0.072 (0.259) 0.431 (0.495) 0.102 (0.303) 0.209 (0.406) 3.657 (2.789) 33,656.14 (22,044.56) 15,342

41.554 0.537 0.256 0.165 0.579 0.116 0.736 0.149 0.719 0.760 0.264

(9.324) (0.499) (0.437) (0.372) (0.494) (0.320) (0.441) (0.356) (0.742) (0.427) (0.441)

0.017 (0.128) 0.339 (0.474) 0.099 (0.299) 0.281 (0.450) 4.372 (2.733) 33,449.59 (23,522.41) 726

Notes: Sample used in the analysis with the 2009 data. ‘‘Official Employment’’ variable is from the main survey. ‘‘Displ.’’ and ‘‘Quits’’ stand for sum of separation events. Household income includes total income of the family in the last 30 days and is trimmed (the first and the last percentage is dropped); the sample for the household income is 15,702.

twice the displacement and quit rates of formal employees. In a transition economy like the Russian one where informal employment has been growing and where the vast majority of incumbents has a formal employment relationship, it might well be that the burden of rising informal employment falls disproportionately on job separators.1

Job Separations and Informality in the Russian Labor Market

261

We are particularly interested in establishing whether the type of job separation produces a differential impact on informality. In other words, are workers who voluntarily separate from their jobs (quitters) differently affected than their displaced counterparts who lost their jobs involuntarily? We can moot that quitters are less likely to end up in informal employment against their will than displaced workers. Using unique data from a displacement supplement to the Russian Longitudinal Monitoring Survey (RLMS) in 2008 and from an informality supplement to the RLMS in 2009, we are able to test this proposition. We thus can establish important findings about the factors driving the formal–informal divide in the labor market, which have not yet sufficiently been discussed in the literature, by linking mode of job separations and subsequent informal or formal employment.2 Our data are detailed enough to investigate the impact of job separations on type of employment across heterogeneous groups of the workforce. We can also analyze whether informality breeds informality, that is, whether having separated from an informal job raises the likelihood to find oneself subsequently in another informal job. The scarce empirical literature on informality in transition countries finds that most informal employment relationships are not wanted by the affected workers, especially if they are dependent wage earners.3 Given this predominantly involuntary nature of informal employment, its incidence might be perceived as a labor market outcome that imposes a cost on displaced workers. This paper thus contributes to the large literature on the costs of job loss.4 The conventional costs that this literature focuses on are foregone earnings due to less employment and less hours worked but also wage penalties upon reemployment. In a companion paper, we find that the monetary costs of job loss in Russia consist in large foregone earnings due to less employment and less hours worked and not in wage penalties upon reemployment (Lehmann, Muravyev, Razzolini, & Zaiceva, 2011). In addition to these traditional labor market outcomes caused by job loss, researchers have started to look at other outcomes that are related to workers’ welfare as well as the welfare of their families. For example, Sullivan and von Wachter (2009) analyze life expectancy as an outcome and establish that displacement at age 40 will shorten the life expectancy of an average worker in the United States by 1–1.5 years. Leombruni, Razzolini, and Serti (2010) measure the causal effect of displacement on workplace injury rates in Italy, confirming a substantially higher injury rate at subsequent jobs of displaced workers relative to their nondisplaced counterparts. Lindo (2011) investigates parental job loss and infant health in the United States. His analysis reveals that husbands’ job losses have

262

HARTMUT LEHMANN ET AL.

significant negative effects on infant health. Liu and Zhao (2011) study a similar issue in China, looking at the effects of mass layoffs of parents in the mid-1990s on their children’s health. They find that paternal job loss affects children’s health negatively, while maternal job loss does not show any significant effect.5 Adding to this literature, we focus on two nonconventional labor market outcomes for the individual displaced worker: apart from informal employment relationships in subsequent jobs, we also look at unofficial wage payments in formal sector jobs, which are widespread in the Russian economy (Gimpelson & Zudina, 2011). Lehmann et al. (2011) provide some preliminary evidence that displaced workers have a higher probability of having their subsequent jobs in the informal sector than their nondisplaced counterparts. The study here exclusively focuses on the link between job separations and informality using various measures of informal employment from different data sources as well as a measure of unofficial wage payments (so-called ‘‘envelope payments’’). Being able to distinguish between involuntary and voluntary informal employment, our study contributes to the debate in the informality literature on the issue of segmented versus integrated labor markets. We thus contribute not only to the literature on displacement but also to the literature on informality. The remainder of the paper has the following structure. The next section addresses the research questions that we investigate when linking job separations and informal employment relationships, embedding this discussion in the literature on informality, while the third section discusses the data and definitional issues and provides some descriptive analysis of type of job separations and informality. This is followed by a section, which presents the empirical models and our research approach of testing the link between displacement, quits, and informality. These tests are done first for dependent employees only using probit, pooled logit, and fixed effect logit models as well as Ordinary Least Squares (OLS) estimation. In a second part they are extended to formal and informal self-employment and nonemployment within a multinomial logit (MNL) framework. The fifth section presents our empirical findings. We find a significant impact of previous displacement and quits on informality, which is robust to different measures of informality. The central results of our analysis show that displacement entraps some of the workers in involuntary informal employment. Those who quit, in turn, experience voluntary informality for the most part, but there seems a minority of quitting workers who read the labor market incorrectly and thus end up in involuntary informal jobs. This

Job Separations and Informality in the Russian Labor Market

263

scenario of entrapment for the displaced and wrong expectations of some of those who quit does not fall on all the workers who separate but predominantly on workers with low human capital and on those who separate from informal jobs. In a final section, we offer some conclusions and policy implications.

LINKING DISPLACEMENT, QUITS, AND SUBSEQUENT INFORMAL EMPLOYMENT The general literature on informality does not discuss a possible link of the mode of separation from jobs on the one hand and the formality or informality of subsequent jobs on the other. The theoretical search and matching macro models, which explicitly include an informal sector, treat separations from jobs as exogenous.6 Micro studies on informal employment, on the other hand, make no distinction between involuntary displacement and voluntary quits (see, e.g., Boeri & Garibaldi, 2006; Bosch & Maloney, 2010). The scarce literature on informality in transition countries analyzes the generally contentious issue of whether labor markets are segmented and workers are prevented from entering the formal sector, as put forth in an early seminal paper by Harris and Todaro (1970), or whether labor markets are integrated and most workers choose voluntarily the informal sector (see, e.g., De Soto, 1990; Maloney, 2004). For Bosnia and Herzegovina, Krstic and Sanfey (2007) find segmentation as do Bernabe` and Stampini (2008) for Georgia. Lehmann and Pignatti (2007), on the other hand, get mixed results for the Ukrainian labor market: while they establish segmentation for dependent employees, they find a two-tier informal self-employment sector, where the lower tier reflects an integrated labor market, that is, anyone can enter informal activities, while the more remunerative upper tier is segmented, with workers blocked from freely entering this part of informal self-employment.7 None of these studies explicitly take into account previous employment, past informality experience or the type of separation from the previous job, which might have an important impact on whether a worker is formally or informally employed in the current job. It is certainly feasible to moot that displaced workers have a higher probability to end up in informal employment against their will. In turn, those who quit may choose an informal employment relationship voluntarily. However, a fraction of those who quit might read the labor market wrong and consequently also they

264

HARTMUT LEHMANN ET AL.

might end up in informal employment involuntarily. With the data at our disposal, we are, therefore, interested to pose the following four research questions: 1. Do job history and past separations matter for subsequent informal employment and the amount of ‘‘envelope payments’’ and are there any differences between voluntary and involuntary separations? 2. Are displaced workers more likely to be ‘‘trapped’’ in informality while those who quit choose it voluntarily? 3. Is the experience of displaced workers and quitters with little human capital different from those with abundant human capital? 4. Is informality persistent, that is, are workers who separate from informal jobs more likely to be informally employed in their subsequent jobs and are there different likelihoods for those displaced and those who quit from informal jobs? Answers to these questions allow us to better understand the nature of informal employment and what drives it in the Russian labor market. Thus, the value added of this paper does not only consist in establishing whether informality is an additional important cost of displacement but also sheds light on unresolved questions in the literature regarding the factors driving the formal–informal divide in the labor market. In this regard, our analysis especially contributes to the debate on the nature of labor markets in emerging and transition countries, that is, whether these labor markets are segmented or integrated.

DATA SOURCES, MEASUREMENT ISSUES, AND DESCRIPTIVE ANALYSIS Data Sources The analysis uses a database that consists of the panel data of the RLMS for the years 2003–2009 and two special supplements. The first supplement is on displacement that was developed by our team in collaboration with Russian scholars and administered to the 17th round of the RLMS between September and December 2008, while the second one on informality, developed by the same group of researchers, was fielded between September and December 2009. The main RLMS data form a well-known rich panel data set, which has provided the empirical basis of many important papers

Job Separations and Informality in the Russian Labor Market

265

on the Russian labor market. We use the main panel data of the years 2003– 2009 and combine them with the new data from the two supplements. This study and the two supplements focus on the main job of workers, which in the case of multiple job holding is either the job providing the largest income or the job where the worker deposits his or her labor book.8 We also distinguish in our analysis between dependent employees and the self-employed and entrepreneurs. Following Slonimczyk (this volume), we consider respondents as self-employed/entrepreneurs if they report to undertake entrepreneurial activities and to be either owners of firms or self-employed individuals who work on their own account with or without employees. The supplement on displacement provides retrospective information on respondents’ job and nonemployment spells over the years 2003–2008. We have information on the beginning and the end of each job spell and of each nonemployment spell, and we are thus able to construct a complete labor market history for all respondents in the indicated period.9 The panel element of the supplement also allows us to trace informal employment over time.

Separation Events: Their Definition and Profile In order to identify a separation as a quit or a displacement, the supplement on displacement provides information on the reason for separating from a job. The possible answers given in the supplement are reproduced in Table A1 and are very much standard in labor force surveys administered in OECD countries. As respondents are told to only give one answer, it is relatively straightforward to classify job separations into quits and displacements.10 We use the first seven answers in Table A1 to determine a separation as a displacement. These answers all reflect involuntary separations insofar as they occur for reasons, which are extraneous to the worker. Focusing only on the first seven answers gives us a conservative estimate of the displacement rate since it might not be unreasonable to consider the expiring of employment contract or of probation time also a displacement. However, we stick to the narrower definition of displacement when producing the estimates of Fig. 1.11 The upper panel of Fig. 1 shows the estimates of annual quit and displacement rates for the years 2003–2008. Quit rates are generally thought to be procyclical and displacement rates countercyclical (Pissarides, 1994). This supposition is borne out by the presented quit and displacement rates.

266

HARTMUT LEHMANN ET AL.

Since the years 2003–2008 are a boom period, we see very large quit rates that are between four to five times larger than the displacement rates. The latter rates hover between 2 and 3 percent and are thus not negligible, but on the lower end of the spectrum that we observe in OECD countries (Kuhn, 2002). Only a small portion of displacements are caused by plant or firm closure; the vast majority are due to redundancies as the bottom panel of Fig. 1 attests.

Defining and Measuring Informality Defining informal employment is a complex issue (see, e.g., Perry et al., 2007). We use the ‘‘legalistic’’ perspective to determine informal employment in this paper, that is, we consider an employment relationship informal if the employer does not register the job to avoid the payment of taxes and social security contributions.12 The Russian labor code stipulates that all employees must sign a written contract and provide their ‘‘labor book’’ to the employer. Oral agreements are explicitly prohibited. We consider selfemployed workers as informal if their activity is not registered. Also interesting, and thus far little pursued in the literature is informality that arises from ‘‘envelope payments’’, where workers who are formally employed get part of their income as undeclared wages. The main RLMS data survey instrument contains questions that allow the identification of workers who have informal employment relationships. Dependent employees are asked whether they are officially registered at their job, that is, whether they are on a ‘‘work roster, work agreement or contract?’’ A positive response to this question is interpreted as a formal employment relationship. Those workers who say no to this question are considered to be in an informal employment relationship. For those who are determined to be in such a relationship, we can also establish whether they entered it involuntarily or voluntarily.13 From the main data set, we can also recover the percentage of a worker’s salary that is paid officially, that is on which taxes and contributions are paid, thus indirectly establishing the incidence and extent of unofficial wage payments or so-called ‘‘envelope payments’’. The supplement on informality allows us to establish dependent workers who have an oral contract in 2009, which we take as a second measure of an informal employment relationship. The informality supplement also allows us to get at the issue of informal employment from an additional angle, by asking dependent employees whether to their knowledge the employer pays

Job Separations and Informality in the Russian Labor Market

267

social security contributions on the entire wage or only on part of it. In the latter case, the percentage of noncompliance is asked for. We use the answers to these questions to establish the incidence of informal employment. In addition, the displacement supplement contains retrospective questions about the type of contract, which a person has in the period 2003– 2008. Again, we take the existence of an oral contract as an indication of an informal employment relationship All information that we use to construct the informality measures is summarized in Table 2, where we also give the source and the way we use the data in the estimation. The first two measures, informal employment and informality in contributions, are taken from the informality supplement. The percentage of official wage payments, the complement of ‘‘envelope payments,’’ is taken from the 2009 reference week section of the main RLMS data. The information that allows us to construct formal dependent employment as well as involuntary informal dependent employment and voluntary informal dependent employment (item 4) is also taken from the 2009 reference week section of the main RLMS data. To establish informal and formal self-employment, we employ data from both the 2009 informality supplement and from the 2009 reference week section of the main RLMS data. This information and responses that imply nonemployment in the 2009 reference week are the basis for the construction of six mutually exclusive labor market states, in which workers can find themselves in 2009.14 Finally, information from the displacement supplement is used to construct panel data on informal employment for the years 2003–2008, equating an oral contract with an informal employment relationship. We use the retrospective panel data from the displacement supplement since these data allow us to accurately map separation events to informality status.

Job Separations and Destination Labor Market States: A Descriptive Analysis Table 3 shows the link between type of job separation, occurring anytime between 2003 and 2008, and the six labor market states, in which a worker can be found in 2009. Looking at displacement events, the bold numbers give the absolute number and the percentages of events associated with each destination state. For example, 35 displacement events in the years 2003– 2008 (8.4 percent of all displacement events in this period) are associated with nonemployment in 2009. The vast majority of displacement events is

268

HARTMUT LEHMANN ET AL.

Table 2. Measure of Informality

(1) Informal employment: Equals 1 if employee has an oral agreement without documents. (2) Informality in contributions: Equals 1 if the employer does not or is suspected not to pay, at least in part, the social security contributions commensurate with an employee’s wage. (3) Percentage of official wage: Denotes the percentage of the wage the respondent thinks was paid officially, that is, from which the employer paid taxes (set equal to missing if answer is ‘‘don’t know’’).

Informality Measures. Source

Informality supplement 2009

Cross-section

Informality supplement 2009

Cross-section

Main survey 2009

Cross-section

Reference week section

(4) Formal dependent employment plus voluntary nature thereof: Main survey 2009 Equals 1 if an employee is registered at the job officially, that is with labor book/agreement or contract. If informal dependent employment: Voluntary vs. involuntary: Involuntary informal equals 1 if the employer did not want to register, while voluntary informal – if either employee or both employer and employee did not want to register.

Cross-section

Reference week section

(5) Informal and formal self-employment: If the respondent works in an Informality supplement enterprise or organization, is the 2009 and Main survey owner of the firm and considers 2009 himself as an entrepreneur and is Reference week section not officially registered at the job (it is formal if the respondent is registered at the job). (6) Informal employment: Equals 1 if employee has an oral agreement without documents.

Way Data Are Used in Estimations

Displacement supplement 2008

Cross-section

Retrospective panel 2003–2008

269

Job Separations and Informality in the Russian Labor Market

Table 3.

Descriptive Statistics on Types of Separations and Potential Destination States.

Type of Job Separation 2003–2008

Destination State in 2009 0

1

NonFormal employed employee

Displacement events total

Displacement events from formal job

Displacement events from informal job

Quit events total

Quit events from formal job

Quit events from informal job

35 8.4% 0.337 (0.601) 32 8.7% 0.308 (0.576) 3 12.5% 0.029 (0.168) 92 5% 0.885 (0.884) 79 5% 0.760 (0.876) 9 5.5% 0.087 (0.315)

104

342 82.2% 0.119 (0.370) 304 82.8% 0.106 (0.349) 17 70.8% 0.006 (0.085) 1546 83.2% 0.537 (0.869) 1353 85.6% 0.470 (0.791) 105 63.6% 0.037 (0.229)

2

3

Informal involuntary employee

Informal voluntary employee

15 3.6% 0.234 (0.496) 12 3.3% 0.187 (0.467) 2 8.3% 0.0312 (0.175) 71 3.8% 1.109 (1.311) 43 2.7% 0.672 (0.855) 22 13.3% 0.344 (0.930)

13 3.1% 0.220 (0.527) 12 3.3% 0.203 (0.518) 1 4.2% 0.017 (0.130) 65 3.5% 1.102 (1.029) 51 3.2% 0.864 (0.798) 11 6.6% 0.184 (0.473)

4

5

SelfSelfemployed employed informal formal 9 2.2% 0.076 (0.297) 6 1.6% 0.051 (0.221) 0

62 3.2% 0.525 (1.115) 39 2.5% 0.331 (0.693) 13 7.9% 0.110 (0.429)

Number of individuals in respective state (2009) 2,879 64 59 118

2 0.5% 0.039 (0.196) 1 0.2% 0.019 (0.140) 1 4.2% 0.020 (0.140) 22 1.2% 0.431 (0.806) 16 1% 0.314 (0.678) 5 3% 0.098 (0.361)

51

416 100%

367 100%

24 100%

1,858 100%

1,581 100%

165 100%

3,275

Notes: In the top of each cell, we find in bold the total number of type of separation event by potential destination state and its percentage relative to the total number of that event in the entire sample. The second row gives the distribution of individuals in the labor market states in 2009 for the main job. The third row displays the ratio of type of separation event by potential destination state relative to the number of individuals in this destination state in 2009 (NB: These are events not individuals, that is, an individual might be displaced more than once, and all these events enter the ratio.) Standard deviations are in parentheses. NB: The sum of displacement events from formal jobs and from informal jobs does not equal the total number of displacements because of missing information on formality/informality in some cases. The same problem exists with quits.

270

HARTMUT LEHMANN ET AL.

unsurprisingly linked to dependent formal employment, while at a low level slightly more are associated with involuntary than voluntary informal dependent employment. Self-employment is the least likely outcome for workers experiencing displacement, with formal self-employment particularly rare, since of the total 416 displacement events only 2 are associated with formal self-employment in 2009. We see a similar distribution of quit events by destination state, with the vast majority of quits ending up in formal dependent employment and self-employment, in particular formal self-employment, being the least likely destination. When we slice separation events along the formal–informal dimension, the distribution of labor market states changes markedly. For example, comparing the distributions for quit events from formal and informal jobs, we can see that the number of individuals ending up in dependent formal employment drops by more than 20 percentage points when we go from quitting formal jobs to quitting informal jobs. In addition, quits from formal jobs produce a slightly higher percentage of workers ending up as a voluntary informal employee, while quitting from informal jobs is associated with a large majority of involuntary informal jobs within dependent informal employment. Similar changes in the distributions of destination states occur when going from formal to informal job displacement, with the caveat that the absolute numbers are small for the latter type of displacement. Our descriptive analysis clearly points to the persistence of informality and to the fact that some workers previously employed in an informal job seem to subsequently get entrapped in informal jobs against their will. The third entry in each cell of Table 3 gives the ratio of separation events relative to the number of individuals in a destination state in 2009 together with their standard deviations. For example, the total displacement events associated with nonemployment are 35 and the number of individuals in this state in 2009 are 104, leading to a ratio of 0.337. The ratio of total quits to individuals in nonemployment is 0.885. Going down the columns one can see the contribution of separation events of each type to the number of individuals in each state in 2009. Inspection of these ratios with respect to type of separation shows the obvious fact, that the contribution of quit events is much larger than the contribution of displacement events. Also note that the ratio of the total displacement and quit events is larger than the sum of their respective disaggregated events because of missing information regarding the distinction between formal and informal jobs.15 Finally, the sum of the total displacement and quit ratios tells us how much the stocks in the respective states are driven by job turnover brought

Job Separations and Informality in the Russian Labor Market

271

on by displacement and quits. When this sum is less than 1, like, for example, in the case of the destination state of dependent formal employment (0.656), separations do not contribute to a rising stock of the state. In the case of the states of dependent involuntary and voluntary informal employment, the sum of the ratios is far above 1. This implies that displacement and quit events contribute to rising stocks of the two states in question. For informal and formal self-employment, the sums of the ratios are below 1. The upshot of these calculations is that displacement and quit events contribute disproportionately to the stocks of dependent informal employment, but not to informal self-employment.

THE EMPIRICAL MODELS AND OUR RESEARCH APPROACH The decision to be an informal worker can be modeled in the framework of random utility models, where choices are determined by individual characteristics xi and an error term e that includes unobserved attributes. An individual i opts for informality if the utility from this choice, Uinf, is higher than the utility from a formal job, Uform. Thus, the probability of observing individual i in an informal job is form 40Þ PrðInf ¼ 1Þ ¼ PrðU inf 4U form Þ ¼ Prðx0i binf  x0i bform þ inf i  i

¼ Prðx0i b þ 40Þ ¼ Fðx0i bÞ

ð1Þ

Assuming that the unobserved factors e are normally distributed, the binary choice between informality and formality can be estimated using a standard probit model. We start by estimating the set of binary choice equations for different dependent variables in 2009 that define the informal employment relationship employing the probit model (1) as well as standard OLS regressions to estimate the complement of ‘‘envelope payments,’’ that is, the percentage of official wage payments. We begin with the most parsimonious model that includes exogenous covariates only (age and gender), and then extend it by including sequentially other covariates, which are summarized in Table 1. To at least reduce the omitted variables bias, we also control for risk attitudes that are usually unobserved and found to be an important predictor of informality status (Dohmen, Khamis, & Lehmann, 2011). To this purpose, we use a general risk indicator, which runs from 0 (complete unwillingness to take risks in general matters) to 10 (complete willingness to take risks in

272

HARTMUT LEHMANN ET AL.

general matters). Inspection of Table 1 shows that on this measure a majority of employees are risk averse and that informal employees have a substantially higher propensity to take risks than formal employees. The main regressors of interest are, of course, the measures related to job separations. We employ the number of displacement and quit events and link them to informality in 2009.16 These measures of job separations are defined for three different time intervals: job separations occurring in 2008 (t1), in 2007 and 2008 (t2), and in the period 2003–2008. We thus model shorter-term and longer-term effects of job separations on informality, but also ensure that the coefficients on the separation variables in our crosssection regressions do not just pick up the rising trend of informal employment and informality that we have mentioned in the introduction. The sketched regressions that use probit and OLS models can establish correlations between separations and informality; they cannot establish a causal effect of the former on the latter. Assuming that the unobservable factors are fixed over time, the causal effect can be estimated when these unobservables are differenced away. We, therefore, take advantage of the panel dimension of our data, and, in a second step, estimate pooled logit and fixed effects logit models with the separation events occurring at time t1 and t2. The panel data are retrospective data covering the years 2003– 2008, which might raise concerns of recall bias. Preliminary analysis of these retrospective data by Lehmann et al. (2011) shows that recall bias does not drive the results regarding wage developments. Considering that recall bias should be minimal when recalling such a dramatic event as a job separation, we are confident that displacement and quits are measured essentially without error, or if there should exist some measurement error, it will not be systematically correlated with informality. The derivation of the fixed effects logit specification is more complex than the derivation of the probit or the pooled logit model. We estimate a conditional maximum likelihood on the sample of individuals who change status at least once over the six periods that we have at our disposal. For these individuals, the conditional distribution of the sequence of outcomes does not depend on the individual specific and time-invariant unobserved effect (Wooldridge, 2002). As long the time invariance assumption of unobservable factors holds, we can identify a causal effect of separations on informality status. In addition, we also perform robustness checks of the fixed effects logit model by interacting year dummies with region and year dummies with gender and educational attainment. In this way we, at least partially, can account for possible time-varying unobservable factors that have an impact on informality.

Job Separations and Informality in the Russian Labor Market

273

Taken together, the results of the probit, OLS, pooled logit and fixed effects logit regressions, estimated with the sample of dependent employees of working age (16–59), provide a tentative answer to the question whether the type of job separation matters for an informal relationship or ‘‘envelope payments’’ in subsequent jobs. To obtain a better understanding of the voluntary versus involuntary nature of informality, in a last step, we differentiate between six different labor market states – formal employment, involuntary informal employment, voluntary informal employment, formal self-employment, informal self-employment, and nonemployment. Again, random utility models can be used to estimate such multiple choice models. In this framework, the probability of observing outcome j is PrðU j 4U k Þ

for any

kaj

(2)

If the k error terms have an extreme value distribution, this choice can be estimated using a MNL model. This model is estimated with the cross-section data of 2009, where the set of regressors includes displacement and quit measures of separation events, which, however, can occur anytime in the period 2003–2008.17 Estimation of MNL models using the whole sample of the working age population allows us to give an answer to the question whether displaced workers are more likely to get entrapped in informal employment. Slicing the data by level of education and by source of separation (separation from formal or informal employment), we provide an empirical analysis of heterogeneous outcomes along observable characteristics.18

RESULTS Relationship Between Separation Events and Informality Status Using various measures of informal employment from different sources as well as the percentage of official wage payments as the dependent variables representing informality status in 2009, we perform probit and OLS regressions, having a set of control variables and separation (displacement and quit) events as the explanatory variables of interest. The set of separation events that we employ jointly in all our regressions is characterized according to the three different time intervals mentioned in the previous section. The results of this set of regressions will be summarized in a concise fashion by reporting the marginal effects on the separation measures.

274

HARTMUT LEHMANN ET AL.

However, to better understand how we proceed we reproduce the results of probit regressions that link informal employment in 2009 to displacement and quit events occurring in 2008. Table 4 shows the four specifications of this probit model. The first specification only includes truly exogenous covariates. It has a quadratic in age and gender, with older workers having a lower probability, while males

Table 4. The Impact of Displacement and Quit Events Occurring in 2008 on Informal Employment in 2009 – Probit Regressions-Marginal Effects.

Displ. Quits Age Age squared Male

(1)

(2)

(3)

(4)

0.048 (0.013) 0.054 (0.004) 0.008 (0.001) 0.000 (0.000) 0.025 (0.003)

0.074 (0.016) 0.040 (0.004) 0.006 (0.001) 0.000 (0.000) 0.022 (0.003) 0.061 (0.019) 0.052 (0.011) 0.051 (0.003) 0.022 (0.005) 0.001 (0.002) 0.020 (0.005)

0.057 (0.014) 0.041 (0.004) 0.005 (0.001) 0.000 (0.000) 0.019 (0.003) 0.059 (0.019) 0.049 (0.011) 0.050 (0.003) 0.019 (0.005) 0.001 (0.002) 0.022 (0.005) 0.004 (0.000)

Yes 22116

Yes 17442

Yes 16854

0.052 (0.015) 0.046 (0.004) 0.006 (0.001) 0.000 (0.000) 0.016 (0.003) 0.060 (0.020) 0.047 (0.012) 0.048 (0.004) 0.015 (0.005) 0.001 (0.002) 0.015 (0.005) 0.005 (0.001) 0.000 (0.000) Yes 15432

City Village Higher education Secondary education Children Married Risk indicator High income Small regions Observations

Source of dependent variable: Informality supplement in 2009. Note: Robust standard errors in parentheses. Significant at 1%.

Job Separations and Informality in the Russian Labor Market

275

having a higher probability to be in informal employment. Both results are confirmed in the scarce literature on informal employment in transition countries (Bernabe` & Stampini, 2008; Gimpelson & Zudina, 2011; Krstic & Sanfey, 2007; Lehmann & Pignatti, 2007; Page´s & Stampini, 2007). Specification 2 adds variables of educational attainment, of marital status, for the number of children, for living in a village or in a big city. It also controls for local labor market conditions by including small region dummies. Workers living in a village or in a big city have a lower probability of being informally employed by roughly five percentage points than workers living in regional centers. Confirming our priors, workers with higher education have a propensity to be informally employed that is substantially lower than workers with primary education or less. For workers with secondary education, this negative difference in the propensity to be informally employed also exists but is attenuated. In the case of married workers, this propensity is two percentage points lower. Specification 3 adds the general risk measure running from 0 (‘‘unwilling to take any risk’’) to 10 (‘‘always willing to take risk’’). An increase by one unit of this measure will increase the likelihood of being in an informal employment relationship by roughly half a percentage point. This positive relationship between willingness to take risks and informal employment confirms the finding of Dohmen et al. (2011) who study the link between risk attitudes and informality in Ukraine. The final specification adds household income that is negatively related to informal employment, but is not statistically significant.19 We thus report the marginal effects of displacement and quit events of specification 3 when summarizing our regression results. These marginal effects in Table 5 are large but attenuated over time when having an oral contract defines informal employment. A displacement event taking place in 2008 raises the probability of being informally employed by nearly six percentage points. This effect falls to roughly two percentage points if displacement occurs anytime in the period 2003–2008. The effects are smaller for quits but show the same attenuation pattern. If in the opinion of the employee the employer does not pay social security contributions or pays them only partially, the worker is defined to be informally employed. Defining informal employment in this way produces very large marginal effects since displacement occurring in 2008 is associated with a rise of the probability of being informally employed of roughly 15 percentage points falling to about 7 percentage points when the displacement event falls into the 2003–2008 interval. For quits, these effects are substantially smaller. The third block of results deals with the complement of informal employment using the respondent’s assertion that in the

0.030

(0.006) 0.023 (0.003)

(0.015) 0.041 (0.004)

Displ.

Quits

0.003 0.016 (0.001)

0.018

2003–2008

Formal Employment: Reference Week 2009 of Main Survey Marginal effects: probit

(0.024) 0.087 (0.007)

0.151

t1

2008

(0.010) 0.049 (0.005)

0.085

t2

2007–2008

(0.006) 0.038 (0.003)

0.066

2003–2008

t1

2008

t2

2007–2008

2003–2008

9.114 4.708 (0.013) (.006) (0.003) (1.814) (1.414) (0.704) 0.039 0.021 0.015 5.780 4.656 2.965 (0.004) (0.002) (0.001) (0.937) (0.598) (0.320)

t2

2007–2008 2003–2008

Percentage of Official Wage: Reference Week 2009 of Main Survey OLS coefficients

0.034 0.015 0.015 8.341

t1

2008

Regressors separation variables used jointly in the same regression

Informality in Contributions: Informality Supplement 2009 Marginal effects: probit

Notes: Robust standard errors in parentheses. Sample includes employees only. ‘‘Displ.’’ and ‘‘Quits’’ stand for sum of separation events. All regressions include age, age squared, gender, city/village dummies, education, children, marital status, risk indicator, and small regions (primary sample units). The tests of the equality of marginal effects or coefficients suggest that they are statistically different for informality in contributions and percentage of official wage, but not in the case of having an oral contract or being registered at the job oficially. The means of the dependent variables are as follows: informal employment 0.060, informality in contributions 0.143, formal employment 0.948, percentage of official wage –89.469 Significant at 1%.

t2

0.057

2007–2008

2008

Informal Employment: Informality Supplement 2009 Marginal effects: probit

Relationship Between Different Informality Measures and Separation Events.

t1

Dependent Variables

Table 5.

Job Separations and Informality in the Russian Labor Market

277

reference week of 2009 s/he is officially registered at the job. While having the same attenuation patterns with respect to the time intervals as the other two measures, the effects are much smaller (in absolute value) and quits seem to produce a slightly larger reduction in formal employment than displacement events. The final block in panel 1 reports the coefficients on the separation events when the dependent variable is the percentage of officially paid wages. We have the striking result that the large negative effect on the percentage of official wage payments is not attenuated when we use the larger 2007–2008 interval. Attenuation only sets in when separation occurs anytime between 2003 and 2008. Equally striking are the much larger declines associated with displacement events. Since we estimate the effects of displacement and quit events jointly, we are able to test for the equality of the marginal effects. In the case of informal employment captured by a lack of paid contributions and in the case of official wage payments, the null hypotheses of equal marginal effects or equal coefficients are rejected pointing to larger effects associated with displacement events. This assertion is particularly true for displacement events that have occurred in 2007 and 2008 and in the period 2003–2008. Even though the marginal effect of the displacement variable is substantially larger than the marginal effect of the quit variable when informal employment is defined as having an oral contract, the test fails to reject the null hypothesis of equal marginal effects. In the case of formal employment, the marginal effects are quite close or even equal; consequently unsurprisingly the null hypothesis is not rejected in this case.20 The evidence presented in Table 5 provides some tentative answers to our first research question. Job separations are strongly associated with a higher incidence of informal employment no matter which of its measures we use. Job separations also lead to a substantial reduction in official wage payments in subsequent jobs. In two of the four cases, informal employment captured by lack of paid contributions and the percentage of official wage payments, formal tests establish a larger effect for displacement events than for quits. So, displaced workers are more strongly affected by informal employment relationships and ‘‘envelope’’ payments in subsequent jobs than their quitting counterparts.

Establishing a Causal Effect of Separation Events on Informal Employment The cross-section regressions that we performed thus far establish strong correlations between separation events and informality status, no matter

278

HARTMUT LEHMANN ET AL.

which definition is used. We now take the analysis a step further using the retrospective panel data of the 2008 displacement supplement. This is a monthly data set with a complete labor market history of all respondents employed at time of interview, which allows us to identify displacement and quit events up to 12 months (t1) and 24 months (t2) prior to holding an employment relationship. This employment relationship, the dependent variable, is traced back through time, taking the value one if the respondent has an oral contract. We start off with the estimation of pooled logit models and then turn to fixed effects logit models to establish a causal effect of separation events on informal employment. The first two columns of Table 6 present coefficients on the separation variables and other covariates including year dummies of the pooled logit model. A comparison with the marginal effects of the probit regressions in Table 4 shows the same demographic characteristics driving informal employment since the signs and the significance levels are similar. Displacement and quit events have large impacts on informal employment independent of whether we use t1 or t2 as the time interval. The larger coefficients on the quit variables are confirmed by formal tests that reject the null hypothesis of equal coefficients. While the pooled logit model takes advantage of variation between and within individuals, the fixed effects logit model only uses variation within individuals, that is, only uses respondents who move from formality to informality and vice versa. The number of regressors is thus reduced with fixed effects logit estimation,21 but we eliminate unobservable factors that partially determine informal employment as long as these unobservable factors are time invariant. The coefficients on the displacement variables in columns 3 and 4 are slightly larger than the corresponding coefficients in the pooled logit models. In contrast, the coefficients on quits fall dramatically in the fixed effects model. It is of particular interest that the impact of displacement events shows no attenuation when we go from 2008 to the period 2007 and 2008 and is nearly three times as large as the impact of quit events in this time period. Formal tests strongly confirm this larger impact of displacement on informal employment when we control for time-invariant unobserved effects.22 The estimates on the separation events in columns 3 and 4 of Table 6 are preserved when we perform robustness checks of this fixed effects model. As columns 1 and 2 of Table A2 in the Appendix attest, the magnitudes of the coefficients on the displacement and quit measures are very similar to those in Table 6 when we add the interaction of year with region as an additional

Table 6.

Pooled and Fixed Effects Logit Regressions. (1)

(2)

(3)

Pooled logit T1 Displ. Quits

0.437 (0.040) 0.718 (0.028)

Displ. Quits Age Age squared Male Higher education Secondary education Children Married North/west Central/Volga South East City Village Constant Year dummies Observations No. of individuals

0.128 (0.007) 0.001 (0.000) 0.283 (0.020) 1.827 (0.039) 0.538 (0.025) 0.043 (0.016) 0.328 (0.023) 0.000 (0.047) 0.300 (0.031) 0.372 (0.041) 0.614 (0.032) 0.514 (0.023) 0.805 (0.026) 0.806 (0.146) Yes 29,5070

(4) FE logit

t2

0.510 (0.034) 0.772 (0.027) 0.115 (0.007) 0.001 (0.000) 0.293 (0.020) 1.816 (0.039) 0.530 (0.025) 0.043 (0.016) 0.339 (0.023) 0.017 (0.047) 0.307 (0.031) 0.366 (0.041) 0.610 (0.032) 0.513 (0.023) 0.807 (0.026) 1.214 (0.148) Yes 29,5070

t1 0.532 (0.074) 0.310 (0.048)

t2

0.634 (0.070) 0.233 (0.053)

0.006 (0.001)

0.007 (0.001)

2.394 (0.300) 0.094 (0.142) 0.349 (0.079) 0.058 (0.102)

2.424 (0.299) 0.054 (0.142) 0.380 (0.080) 0.077 (0.102)

Yes 18,335 349

Yes 18,335 349

Notes: Robust standard errors in parentheses. The test of equality of coefficients of ‘‘Displ. and ‘‘quits’’ is rejected with all specifications. In the pooled logit, the coefficients on ‘‘Displ.’’ are always significantly smaller than the coefficients on ‘‘Quits’’ (at t1 and t2). In the FE logit, the coefficients on ‘‘Displ.’’ are always significantly larger than the coefficients on ‘‘Quits’’.The dependent variable is informal employment (oral contract) and is taken from the displacement supplement 2008. ‘‘Displ.’’ and ‘‘Quits’’ stand for sum of separation events. This is a monthly data set based on the retrospective panel from the displacement supplement; t1 indicates displacement or quit events in the previous 12 months; t2 indicates displacement or quit events in the previous 24 months. Fixed effects (conditional) logit estimation uses only job changers (i.e., movers from formality to informality and vice versa). Omitted categories: female, primary education, not married, regional center, Moscow/St. Petersburg.  Significant at 1%.

280

HARTMUT LEHMANN ET AL.

regressor. The magnitudes are also maintained if we add the interactions of year with education and with gender to the model (columns 3 and 4 of Table A2). Thus, when we partially take account of the variation in the macroeconomic environment over time and space as well as of time-varying heterogeneity, the baseline effect is clearly not altered. The larger estimated effects of displacement events in the fixed effects logit model and the fact that these effects are not attenuated over time in conjunction with the smaller and attenuated effects of quit events might be interpreted as evidence of a segmented labor market. Essentially those separated from their jobs involuntarily seem to be rationed out of formal employment more than their quitting counterparts. Since we have information on the voluntary nature of informal dependent employment in our data, we analyze this issue of labor market segmentation in what follows together with the question whether displacement imposes a cost on workers in the form of involuntary informal employment.

Job Separations and the Involuntary and Voluntary Nature of Informal Employment Taking formal dependent employment as our base category, we perform MNL regressions varying measures of displacement and quits and allowing for five labor market states in addition to the state of formal employment of dependent workers: involuntary informal employment of dependent workers, voluntary informal employment of dependent workers, informal self-employment, formal self-employment, and nonemployment. We treat both informal self-employment and formal dependent employment as voluntary. The six states shown in Table 7 are given for the year 2009.23 The MNL regressions are cross-section regressions where we estimate the probability of being in a certain state in 2009 using covariates from the same year, including the general risk indicator. The main regressors of interest are measures of job separations, which are defined as separations occurring anytime between 2003 and 2008. We use this time interval to maximize the number of occurring job separations. The evidence in Table 5 implies that it is not really problematic to map separation events in the period 2003–2008 to labor market status in 2009 since the effects of displacement and quits are never reduced to 0 when we choose this longest time interval at our disposal. In addition, the evidence of the fixed effects estimates in Table 6 points to a nondecreasing causal effect of displacement on informal employment as the

281

Job Separations and Informality in the Russian Labor Market

Table 7.

Multinomial Logit Regressions – Marginal Effects of Regressors Measuring Displacement and Quits. NonFormal Formal Informal Involuntary Voluntary Employment SelfSelfInformal Employment Informal Employment Employment Employment Employment

Panel 1: Displacements and quits used jointly Displ. 0.0049 0.0035 0.0085 Quits 0.005 0.0043 0.0139

0.0125 0.0016

0.0094 0.0018

0.0219 0.0081

Panel 2: Displacements and quits by education used jointly Displ_low 0.0049 0.0029 0.0056 Displ_high 0.1157 0.0077 0.0897 Quits_low 0.0043 0.0039 0.0106 Quits_high 0.0009 0.0073 0.0021

0.0146 0.0089 0.0010 0.0257

0.0109 0.0060 0.0057 0.0016

0.0233 0.0154 0.0071 0.0138

Panel 3: Displacements Displ_formal Quits_formal Displ_informal Quits_informal

and quits by informality used jointly 0.0040 0.0044 0.0005 0.0018 0.0039 0.0027 0.0139 0.0048 – 0.0142 0.0090 0.0599

0.0176 0.0092 – 0.0215

0.0143 0.0033 0.0126 0.0083

0.0230 0.0094 0.0373 0.0070

Panel 4: Last separation by informality status used jointly last_displ_formal 0.0024 0.0192 0.07 last_quit_formal 0.0054 0.0152 0.0359 last_displ_informal 0.0679 0.101 0.1768 last_quit_informal 0.0449 0.0571 0.166

0.017 0.0172 0.0269 0.0033

0.0093 0.0025 0.0038 0.007

0.0748 0.035 0.1497 0.0535

Notes: Robust standard errors in parentheses. Marginal effects are reported.‘‘–’’ Refers to the cells where the effects were estimated very imprecisely due to negligible numbers of observations.Other covariates include age, age squared, gender, city, village, education, children, marital status, macroregion and risk indicator.Measures representing various types of separations:‘‘Displ.’’ and ‘‘Quits’’ stand for sum of separation events over 2003–2008.‘‘ Displ_low’’ (‘‘Quits_low’’) and ‘‘Displ_high’’ (‘‘Quits_high’’) stand for the sum of displacement (quits) events for individuals with low (high) education, respectively.‘‘Displ_formal’’ (‘‘Quits_formal’’) and ‘‘Displ_informal’’ (‘‘Quits_informal’’) stand for the sum of displacement (quit) events from a formal and informal job, respectively.‘‘last_displ_formal’’ (‘‘last_quit_ formal’’) and ‘‘last_displ_informal’’ (‘‘last_quit_informal’’) are equal to one if last separation is displacement (quit) from a formal or informal job, respectively; these four dummies represent mutually exclusive events.Significant at 10%; significant at 5%; significant at 1%.

time interval is widened to 24 months, while for quits the effect is only slightly reduced. On the basis of MNL regressions, not shown here but available on request, we calculate the marginal effects of displacement and quits for the six potential states. In panels 1–3 of Table 7, we use variants of the sum of displacement and quit events as the regressors of interest, while panel 4 is based on one MNL regression with four mutually exclusive dummies

282

HARTMUT LEHMANN ET AL.

included: the dummies take the value one if the last separation is a displacement from an informal job, a displacement from a formal job, a quit from an informal job, or a quit from a formal job. In panel 1, both the sum of displacement events and the sum of quit events raise the probability of being involuntarily in informal employment by roughly half a percentage point. In contrast, only quits raise the probability of being in a voluntary informal job. We take these two results as evidence that displaced workers get trapped in informal jobs, while among quitters there are some workers who select themselves into an informal job while others read the labor market wrong and end up involuntarily in such a job. Panel 1 also shows that those who separate voluntarily from their job lower their chances of finding formal dependent employment, while the displaced have a lower probability of being selfemployed formally. It is also striking that displaced workers have a far higher probability to end up in nonemployment than those who quit. Panel 2 shows displacement and quit events interacted with low and high education.24 Displaced workers with low human capital find themselves with a higher probability in involuntary informal employment than their nondisplaced counterparts, while displaced workers with high educational attainment are much less likely to find themselves in this state. While for both groups displacement does not affect the probability to be in voluntary informal employment, it has a positive impact on formal employment for the highly educated displaced workers. In turn displaced workers with low human capital have a lower propensity to end up in informal selfemployment, while they have a larger probability to enter nonemployment. The sum of quit events of workers with low and high education has a somewhat different pattern. Those with low education have an increased likelihood to be in both the involuntary and voluntary sector of dependent informal employment; at the same time these workers are less likely to find themselves in formal dependent and self-employment. Workers with high education who quit their previous jobs have a higher propensity of finding a voluntary informal job, and a substantially lower probability to be involved in informal self-employment, while the states involuntary informal and formal dependent employment are not affected by their quitting actions. The evidence collected in panel 2 can be interpreted in the following way. Some of the workers with low human capital who are displaced get trapped in informal jobs, as they end up in a state they do not want to select. On the other hand, workers with a large amount of human capital upon displacement do not find themselves more frequently in any type of informal

Job Separations and Informality in the Russian Labor Market

283

employment relationships; in actual fact, interacting displacement with high education depresses the probability to be in an involuntary informal job substantially. Workers with low education who quit end up in both involuntary and voluntary informal jobs, so some of them get trapped against their will in informal employment. In turn, workers well endowed with human capital who quit subsequently can avoid informal jobs if they do not want them. Consequently, the results presented in panel 2 imply that informal employment is an important cost of displacement and that it falls predominantly on workers with low education. At the same time, for quitters with low human capital the presented results imply some labor market segmentation. In panel 3, displacement and quit events are sliced differently as we investigate whether there are differences in the probability of occupying states by formal or informal sector of origin. Concentrating on dependent employment as an outcome, we see that being displaced from a formal job does not affect any dependent employment state. Quits from formal employment, on the other hand, raise the probability to be in voluntary informal jobs. We find no effects on dependent informal employment for those workers who are displaced from an informal job. For those who quit from such a job, the likelihood is raised for both involuntary and voluntary informal employment. It is also striking that those who quit from an informal job are not entering nonemployment at an increased rate but informal and formal self-employment, while the three remaining separations in panel 3 cause a higher probability to end up in nonemployment. Panel 4, where the last separation is decomposed in four mutually exclusive events (displacement from a formal job, displacement from an informal job, quits from a formal job, and quits from an informal job), conveys similar information as the previous panel. For example, quits from informal employment translate into higher probabilities of both types of informal jobs. In addition, displacement from an informal job makes it a lot less likely that the new job is of the voluntary informal nature.

CONCLUSIONS The general research question that we investigate focuses on the link between job separations (displacement and quits) and informality. Our empirical analysis explores whether displaced workers and quitters

284

HARTMUT LEHMANN ET AL.

experience more informal employment and ‘‘envelope payments’’ in subsequent jobs than new labor market entrants or incumbents. In a transition economy like the Russian one where informal employment has been growing and where the vast majority of incumbents has a formal employment relationship, it might well be that the burden of rising informal employment falls disproportionately on job separators. We refine this general research question by probing into the question whether workers who are involuntarily separated from their jobs are more likely to become trapped in involuntary informal employment than workers who quit their jobs. We also analyze whether this experience of potentially being trapped in involuntary informal employment differs by the level of human capital. In addition, we look at the persistence of informality, that is, whether past spells in informal employment raise the likelihood to be currently in an informal job. Our central results confirm our contention that displacement entraps some of the workers in involuntary informal employment. Those who quit, in turn, experience voluntary informality for the most part, but there seems to be a minority of quitting workers who end up in involuntary informal jobs because they read the labor market wrong when separating from their previous job. However, this scenario of entrapment for the displaced and wrong expectations of some of those who quit does not fall on all the workers who separate but predominantly on workers with low human capital and on those who separate from informal jobs. This latter result also implies that informal employment is persistent as some workers churn from one informal job to the next. We also find strong evidence that displaced workers are confronted with a larger share of ‘‘envelope payments’’ in formal jobs than quitters. Overall, our results point to informal employment as an important cost of job loss in the Russian labor market. In a companion paper on the monetary and nonmonetary costs of displacement in the Russian labor market, we put forth the policy recommendation to promote policies that help displaced workers to increase their search effectiveness. This recommendation was based on the fact that the main monetary costs of job loss were found to be foregone earnings due to long spells of nonemployment and not wage penalties upon reemployment. Given the results in this study, the policies that we wish to advocate need to be amended. If it is true that above all displaced workers with low human capital end up in informal jobs involuntarily, training and further training policies should also be on the agenda of policy makers who wish to help displaced workers.

Job Separations and Informality in the Russian Labor Market

285

NOTES 1. In principle, rising informal employment could also obtain by changing formal jobs of incumbents into informal ones and by having a high incidence of informal employment for new labor market entrants. 2. Neither the general literature that debates labor market segmentation versus integrated labor markets (e.g., Harris and Todaro 1970 versus de Soto 1990 and Maloney 2004) nor the literature on informality in transition countries (see papers mentioned in Footnote 3) do discuss the link between job separations and informality. 3. See, for example, Krstic and Sanfey (2007) on Bosnia and Herzegovina, Lehmann and Pignatti (2007) on Ukraine, Bernabe` and Stampini (2008) on Georgia, and Page´s and Stampini (2007) on several transition countries. 4. For a survey of older studies on the costs of job loss, see Kuhn (2002); the most recent studies are summarized, for example, in Hijzen, Upward, and Wright (2010). 5. There are many more studies on the health costs of displacement; this growing literature is discussed in Lindo (2011). 6. See, for example, Kolm and Larsen (2003), Albrecht, Navarro, and Vroman (2009), Zenou (2008). 7. This characterization of informal self-employment goes back to Fields (1990). 8. Respondents in the main RLMS and in the displacement supplement are asked to discuss the job that they themselves consider their main job. This can be understood by the respondents in the two ways mentioned in the text. 9. We also have information on the actual weekly hours worked, on occupation, and the sector of employment as well as on the wage at the beginning and the end of each job. 10. For a discussion of the pros and cons of using survey data to define displacement, see the introductory chapter in Kuhn (2002). 11. In our opinion, there certainly exist good arguments to consider job separations voluntary when they occur because of the expiring of a contract. When a worker signs a contract for a temporary job, s/he does so out of her/his own volition. The separation resulting from such a contract can, therefore, be considered voluntary. The same can be said about a contract signed that has a probation period as one of its stipulations as long as the firm evaluates the worker’s performance fairly. 12. The ‘‘productive’’ concept of informal employment, which, for example, links small firm size or self-employment to informal status, can lead in transition economies to large measurement error (Lehmann & Pignatti, 2007). This is not to say that the ‘‘legalistic’’ definition cannot be also plagued by some measurement error. In a middle income transition country like Russia, this type of measurement error strikes us, however, as smaller of an order of magnitude than the measurement error associated with the ‘‘productive’’ definition. 13. Respondents are asked whether (1) the employer did not want a registration of the job, (2) the respondent did not want to register, or (3) both employer and respondent did not want to register. Respondents giving answers (2) or (3) are deemed to be voluntarily in informal jobs.

286

HARTMUT LEHMANN ET AL.

14. These states are as follows: involuntary informal dependent employment, voluntary informal dependent employment, formal dependent employment, formal self-employment, informal self-employment, and nonemployment. 15. For the same reason in the last column of Table 3, total separation events are larger than the sum of these events originating from formal and informal jobs. 16. The small number of displacements caused by firm or plant closure (see panel 2 of Fig. 1) determines our research strategy insofar as we cannot use this measure as our conditioning variable, even though it is thought to be ‘‘more exogenous’’ than displacement due to redundancies. Instead, we have to employ displacement in general as our conditioning regressor, independent of whether it is due to firm/plant closure or redundancies. 17. We have also experimented with estimating pooled multinomial logit models for the 2003–2008 period in order to incorporate more labor market transitions and to check the robustness of our results. We were, however, not able to distinguish between voluntary and involuntary self-employment and had to use self-reported selfemployment status in these regressions. The main results were qualitatively similar to the ones reported in the text and are available from the authors upon request. 18. A major drawback of the multinomial logit model is the assumption that the error terms are mutually independent leading to the independence of irrelevant alternatives (IIA) property. We have conducted several tests excluding each of the outcomes (or a combination of more outcomes) and tested the IIA property between this restricted model and the full model with all the alternatives. The IIA test was implemented with a generalized Hausman test. The null hypothesis of equality of coefficients between the restricted and full model was always rejected. For this reason, we have opted for the full efficient model that includes all outcomes. An alternative route would have been to estimate multinomial probit models, which alas is not possible with the data at hand since we do not have exclusion restrictions, that is, attributes that vary across choices (see Keane, 1992, for identification requirements of multinomial probit models). The second theoretical alternative to the multinomial logit model could be the nested logit model. This model, while solving the IIA problem, in practice converges only in the context of a conditional logit model, that is, a model where there exist characteristics that vary across choices. 19. We performed some sensitivity analysis with this probit model, expanding the classification of displacement by including expiring of contract and of probation time in its definition. The results we very similar to those in Table 4 and are available upon request from the authors. 20. The results of the chi-square tests (in the case of the probit regressions) and F-tests (when using OLS regression) are not shown here but available upon request from the authors. 21. Adding year dummies in the fixed effects, model wipes out the linear term of the quadratic in age. On the other hand, since some workers change education, the number of children and marital status over the span of the panel these variables are not eliminated from the set of regressors. 22. We also estimated the pooled and the fixed effects logit models using the more encompassing definition of displacement. The estimated coefficients on the separation variables are very close to those in Table 6. They are not reported but available on request.

Job Separations and Informality in the Russian Labor Market

287

23. We are confronted here with rather small sample sizes, especially for the formal and informal self-employed. 24. High education means university education; low education is secondary education or less.

ACKNOWLEDGMENTS We are grateful to Alexander Muravyev, Fabian Slonimczyk, two anonymous referees and to participants of the IZA workshop ‘‘Institutions and Informal Employment in Emerging and Transition Economies’’ in June 2011 in Bonn for valuable comments and suggestions. The comments and suggestions of one of the editors, Konstantinos Tatsiramos, were particularly helpful in improving the paper. We thank the MacArthur Foundation for financing the informality supplement. Lehmann is also grateful to the Volkswagen Foundation for financial support within the project ‘‘The political economy of labor market reform in transition economies: a comparative perspective’’. We would also like to express our appreciation to Vladimir Gimpelson, Rostislav Kapeliushnikov, Mikhail Kosolapov, and Polina Kozyreva for their invaluable help in developing the RLMS displacement and informality supplements.

REFERENCES Albrecht, J., Navarro, L., & Vroman, S. (2009). The effects of labor market policies in an economy with an informal sector. Economic Journal, 119, l105–1129. Bernabe`, S., & Stampini, M. (2008). Labour mobility during transition: Evidence from Georgia. LICOS Discussion Paper Series, Discussion Paper No. 206. University of Leuven, Leuven, Belgium. Boeri, T., & Garibaldi, P. (2006). Shadow sorting. CEPR Discussion Paper No. 5487. Centre for Economic Policy Research, London, U.K. Bosch, M., & Maloney, W. F. (2010). Comparative analysis of labor market dynamics using Markov processes: An application to informality. Labour Economics, 17(4), 621–631. De Soto, H. (1990). The other path: The invisible revolution in the Third World. Perennial Library. Dohmen, T., Khamis, M., & Lehmann, H. (2011). Risk attitudes, time preferences and the incidence of informality among workers: Evidence from a transition country, Bonn and Maastricht, unprocessed. Fields, G. S. (1990). Labour market modelling and the urban informal sector: Theory and evidence. In D. Thurnham, B. Salome´ & A. Schwarz (Eds.), The informal sector revisited. Paris: OECD.

288

HARTMUT LEHMANN ET AL.

Gimpelson, V., & Zudina, A. (2011). The ‘‘informal’’ in the Russian economy: How many are there and who are they? Higher School of Economics Working Paper No. WP3/2011/06, Moscow (in Russian). Harris, J. R., & Todaro, M. P. (1970). Migration, unemployment and development: A two sector analysis. American Economic Review, 60, 126–142. Hijzen, A., Upward, R., & Wright, P. W. (2010). The income losses of displaced workers. Journal of Human Resources, 45(1), 243–269. Keane, M. P. (1992). A note on identification in the multinomial probit model. Journal of Business & Economic Statistics, 10(2), 193–200. Kolm, A.-S., & Larsen, B. (2003). Social norm, the informal sector and unemployment. Finanzarchiv, 59, 407–424. Krstic, G., & Sanfey, P. (2007). Mobility, poverty and well-being among the informally employed in Bosnia and Herzegovina. Economic Systems, 31(3), 311–335. Kuhn, P. (Ed.). (2002). Losing work, moving on: Worker displacement, an international perspective. W. E. Upjohn Institute for Employment Research. Lehmann, H., Muravyev, A., Razzolini, T., & Zaiceva, A. (2011). The wage and non-wage costs of displacement: Evidence from Russia. Department of Economics Working Paper No. 734, University of Bologna, March. Lehmann, H., & Pignatti, N. (2007). Informal employment and labor market segmentation in transition economies: Evidence from Ukraine. IZA Discussion Paper No. 3269. Institute for the study of Labor, Bonn, Germany. Leombruni, R., Razzolini, T., & Serti, F. (2010). The pecuniary and non-pecuniary costs of job displacement. The risky job of getting back to work. Ivie Working Paper WP-AD 2010–12. Instituto Valenciano de Investigaciones Econo´micas, Valencia, Spain. Lindo, J. M. (2011). Parental job loss and infant health. Journal of Health Economics, 30(5), 869–879. Liu, H., & Zhao, Z. (2011). Parental job loss and children’s health: Ten years after the massive layoff of the SOE’s workers in China. IZA Discussion Paper No. 5846, July. Institute for the Study of Labor, Bonn, Germany. Maloney, W. F. (2004). Informality revisited. World Development, 32, 1159–1178. Page´s, C., & Stampini, M. (2007). No education, no good jobs? Evidence on the relationship between education and labor market segmentation. IZA Discussion Paper No. 3187. Institute for the Study of Labor, Bonn, Germany. Perry, G., Maloney, W., Arias, O., Fajnzylber, P., Mason, A., & Saavedra-Chanduvi, J. (2007). Informality: Exit and exclusion. Washington, DC: World Bank. Pissarides, C. (1994). Search unemployment with on-the-job search. Review of Economic Studies, 61(3), 457–475. Schneider, F., Buehn, A., & Montenegro, C. E. (2010). Shadow economies all over the world: New estimates for 162 countries from 1999 to 2007. World Bank, mimeo. Sullivan, D., & von Wachter, T. (2009). Job displacement and mortality: An analysis using administrative data. Quarterly Journal of Economics, 124(3), 1265–1306. Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. London: MIT Press. Zenou, Y. (2008). Job search and mobility in developing countries: theory and policy implications. Journal of Development Economics, 86, 336–355.

289

Job Separations and Informality in the Russian Labor Market

APPENDIX Table A1.

Reasons for Leaving Job and Classification as Quit or Displacement.

Reason

Classification

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

Displacement Displacement Displacement Displacement Displacement Displacement Displacement Quit Quit Quit Quit Quit Quit Quit Quit Quit Quit Quit Quit Quit Quit Quit Quit Quit Quit Variable

Closing down of enterprise/organization Moving of enterprise/organization Reorganization of enterprise/organization Bankruptcy of enterprise/organization Privatization of enterprise/organization Dismissal initiated by employer Personnel reduction Expiring of employment contract Expiring of probation time Military service Imprisonment Own illness or injury Studies Retirement Early retirement Marriage Parental leave Need to take care of other members of family Change of residence Wanted/was proposed higher salary Wanted/was proposed better working conditions Wanted/was proposed more interesting work Wanted to start own business Main job became second job End of farming/sole proprietorship Other

290

HARTMUT LEHMANN ET AL.

Table A2.

Fixed Effects Logit with Interactions. (1)

Displ. t1 Quits t1

0.549 (0.075) 0.296 (0.048)

Displ. t2 Quits t2 Year dummies Year region Year male Year education Observations Number of individuals

Yes Yes

18,335 349

(2)

0.685 (0.071) 0.219 (0.054) Yes Yes

18,335 349

(3) 0.523 (0.075) 0.290 (0.048)

(4)

Yes

0.612 (0.071) 0.205 (0.054) Yes

Yes Yes 18,335 349

Yes Yes 18,335 349

Notes: Robust standard errors in parentheses.The dependent variable is informality (oral contract) from displacement supplement 2008.‘‘Displ.’’ and ‘‘Quits’’ stand for sum of separation events.This is a monthly data set based on the retrospective panel from displacement supplement. t1 indicates displacement or quit events in the previous 12 months, t2 indicates displacement or quit events in the previous 24 months.Fixed effects (conditional) logit estimation uses only job changers (i.e., movers from formality to informality and vice versa).The rest of covariates are in Table 6. Significant at 1%.