An In-depth Look at the Distribution and Redistribution of Welfare
DISSERTATION of the University of St. Gallen, Graduate School of Business Administration, Economics, Law and Social Sciences (HSG) to obtain the title of Doctor of Economics
submitted by Monika Engler from Sevelen (St. Gallen)
Approved on the application of Prof. Dr. Monika Bütler and Prof. Dr. Franz Jaeger
Dissertation no. 3682 Difo-Druck GmbH
The University of St. Gallen, Graduate School of Business Administration, Economics, Law and Social Sciences (HSG) hereby consents to the printing of the present dissertation, without hereby expressing any opinion on the views herein expressed. St. Gallen, October 19, 2009 The President
Prof. Ernst Mohr, PhD
Acknowledgements First of all, I would like to thank my supervisor, Monika Bütler, for giving me the opportunity to write this thesis and develop my skills as a researcher. This work has greatly benefited from her guidance and support. I especially thank Shamika Ravi who generously invited me to work with her at the Indian School of Business, a great place for doing research. Special thanks also to Franz Jaeger for co-reviewing my work and providing valuable comments. I gratefully acknowledge financial support from the Swiss National Science Foundation, and I also thank the Swiss Federal Statistical Office for giving me access to the data used in Chapter 2. Furthermore, I would like to thank all my colleagues from the SEW-HSG for the positive and inspiring work atmosphere. In particular, I thank Stefan Staubli for his extensive input as my coauthor. Finally, I am deeply indebted to my family who spared no effort to make my education possible. And last but not least, I thank Angelo Busa for his invaluable love and encouragement.
October 2009
Monika Engler
Contents 1 Introduction
1
2 Measuring a State’s Overall Impact on Welfare: Public Income Redistribution in Switzerland
5
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6
2.2
Methodological Approach and Data . . . . . . . . . . . . . . . . . . . . . . .
9
2.2.1
The Concept of Budget Incidence Analysis . . . . . . . . . . . . . . .
9
2.2.2
Counterfactual Income . . . . . . . . . . . . . . . . . . . . . . . . . .
11
2.2.3
Incidence Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2.2.4
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
2.2.5
Rationale and Construction of the Pseudo Panel . . . . . . . . . . . .
17
2.2.6
Measurement of Redistribution . . . . . . . . . . . . . . . . . . . . .
18
2.2.7
Statistical Test for the Significance of the Redistribution . . . . . . .
20
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
2.3.1
Redistribution within a Year: General Overview . . . . . . . . . . . .
20
2.3.2
Incidence of Individual Public Interventions . . . . . . . . . . . . . .
22
2.3.3
Evolution of the Incidence Over Time . . . . . . . . . . . . . . . . . .
29
2.3.4
Significance of the Redistributive Effects . . . . . . . . . . . . . . . .
32
2.3.5
Redistribution across Socioeconomic Groups . . . . . . . . . . . . . .
35
2.3.6
Redistribution across Households versus Redistribution across Stages
2.3
of Life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
Sensitivity Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . .
44
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
44
2.4.1
Cross-Comparison with Previous Evidence . . . . . . . . . . . . . . .
44
2.4.2
Limitations of the Analysis . . . . . . . . . . . . . . . . . . . . . . . .
47
2.5
Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
2.6
Appendix of Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
2.3.7 2.4
i
3 Measuring Welfare Distribution Differently: The Distribution of Leisure Time Across Countries and Over Time
57
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
58
3.2
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
3.3
Trends in the Allocation of Time . . . . . . . . . . . . . . . . . . . . . . . .
62
3.3.1
Trends in Nonmarket Work . . . . . . . . . . . . . . . . . . . . . . .
67
3.3.2
Trends in Leisure . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
3.3.3
Demographic versus Behavioral Differences across Countries . . . . .
70
The Distribution of Leisure Time . . . . . . . . . . . . . . . . . . . . . . . .
73
3.4.1
Leisure Time Differences across Educational Categories . . . . . . . .
74
3.4.2
Leisure Time Differences within Educational Categories . . . . . . . .
77
3.5
Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
3.6
Appendix of Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
81
3.4
4 Measuring the Impact of a Single Policy Intervention on Welfare: The Example of India’s National Workfare Scheme
87
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
88
4.2
India’s Poverty and the Rationale for the NREGS . . . . . . . . . . . . . . .
90
4.3
Empirical Strategy and Data . . . . . . . . . . . . . . . . . . . . . . . . . . .
96
4.4
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
99
4.4.1
Impact of NREGS on Consumption and Savings . . . . . . . . . . . .
99
4.4.2
Are there Negative Spillover Effects on the Labor Market? . . . . . . 112
4.4.3
Impact on Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.5
Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5 Concluding Remarks
119
ii
List of Figures 2.1
The distribution of pre- and post-fisc income, 2005
. . . . . . . . . . . . . .
21
2.2
Lorenz curves for pre- and post-fisc income, 2005 . . . . . . . . . . . . . . .
23
2.3
Lorenz curves for public expenditures and revenues, 2005 . . . . . . . . . . .
27
2.4
Incidence of public expenditures, comparison of 1990, 2000, 2005
. . . . . .
30
2.5
Incidence of public revenues, comparison of 1990, 2000, 2005 . . . . . . . . .
31
2.6
Average course of pre- and post-fisc lifetime income . . . . . . . . . . . . . .
40
3.1
Trends in the main time use aggregates, average hours per week . . . . . . .
63
3.2
Nonmarket activities, full sample . . . . . . . . . . . . . . . . . . . . . . . .
68
3.3
Leisure activities, full sample . . . . . . . . . . . . . . . . . . . . . . . . . . .
69
3.4
Composition of the time use differences between the United States and non-US countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
72
3.5
Evolution of the leisure II distribution . . . . . . . . . . . . . . . . . . . . .
73
4.1
Seasonal patterns of NREGS employment
93
4.2
Cumulative distribution of monthly total consumption expenditures, NREGS
. . . . . . . . . . . . . . . . . . .
participants 2007 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
iii
iv
List of Tables 2.1
Consolidated budget of the confederation, cantons, municipalities, and social insurance institutions, 1990, 2000, 2005 . . . . . . . . . . . . . . . . . . . . .
8
2.2
Average pre-fisc household income, 2005 . . . . . . . . . . . . . . . . . . . .
12
2.3
Criteria for the allocation of public revenues and expenditures . . . . . . . .
14
2.4
Key figures of pre- and post-fisc income distributions, 2005 . . . . . . . . . .
22
2.5
Incidence of public expenditures and revenues by quintile of pre-fisc household income, 2005
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
2.6
Transfer-induced change in measured inequality, 2005 . . . . . . . . . . . . .
33
2.7
Transfer-induced change in measured inequality, 1990-2005 . . . . . . . . . .
34
2.8
Incidence of public expenditures and revenues by type of household, households in working age group, 2005 . . . . . . . . . . . . . . . . . . . . . . . .
2.9
36
Incidence of public expenditures and revenues by work load, households in working age group, 2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
38
2.10 Transfer-induced change in measured inequality within and between cohort income, period 1990-2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
42
2.11 Incidence of public expenditures and revenues by quintiles of pre-fisc household income, households in working age group, 2005 . . . . . . . . . . . . . .
51
2.12 Incidence of public expenditures and revenues by quintiles of pre-fisc household income, total sample, 1990 and 2000 . . . . . . . . . . . . . . . . . . . .
52
2.13 Transfer-induced change in measured inequality, details, 2005 . . . . . . . . .
54
2.14 Pseudo panel of pre- and post-fisc income, 1990-2005 . . . . . . . . . . . . .
55
3.1
Description of the time use data . . . . . . . . . . . . . . . . . . . . . . . . .
61
3.2
Average hours per week spent in the main time use aggregates . . . . . . . .
66
3.3
Time allocation by educational attainment . . . . . . . . . . . . . . . . . . .
76
3.4
Decomposition of differences in leisure II due to changes in education . . . .
78
3.5
Leisure II distribution within different educational categories . . . . . . . . .
79
3.6
Time use categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
81
v
3.7
Average hours per week spent in different nonmarket activities . . . . . . . .
82
3.8
Average hours per week spent in different leisure activities . . . . . . . . . .
83
3.9
Blinder-Oaxaca decomposition of changes in time use . . . . . . . . . . . . .
84
3.10 Time allocation by educational attainment for men and women . . . . . . . .
85
3.11 Time allocation by educational attainment for employed individuals . . . . .
86
4.1
Expansion of the NREGS, 2006/07 to 2007/08 . . . . . . . . . . . . . . . . .
92
4.2
Comparison of rural wages, financial year 2007/08 . . . . . . . . . . . . . . .
95
4.3
Descriptive statistics of the 2007 and 2008 surveys . . . . . . . . . . . . . . . 100
4.4
Logit regression to generate the propensity score, 2007 sample . . . . . . . . 102
4.5
Average impact of the NREGS on monthly per capita expenditures, propensity score-matched single differences, 2007 sample . . . . . . . . . . . . . . . . . . 104
4.6
Average change in the monthly per capita expenditures between the 2007 and 2008 surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.7
Impact on monthly per capita food expenses, double differences . . . . . . . 108
4.8
Average impact of the NREGS on monthly per capita food expenses, tripledifference estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.9
Average impact of the NREGS on monthly per capita expenditure subcategories, triple-difference estimates . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.10 Attitudes towards the NREGS, 2008 survey . . . . . . . . . . . . . . . . . . 113 4.11 Descriptive statistics of health indicators . . . . . . . . . . . . . . . . . . . . 115 4.12 Average impact of the NREGS on health outcomes, propensity score-matched single differences, 2007 sample . . . . . . . . . . . . . . . . . . . . . . . . . . 116
vi
Abstract This dissertation examines the distribution of welfare in a population and its redistribution by the government. Motivated by the large and growing role of the public sector – and with it the redistributive potential – this paper makes three distinct contributions to show ways that decision-makers can glean necessary information about distributional situations. As such, this work offers both aggregated welfare analyses (Chapters 2 and 3), which cover public activities as completely as possible, as well as a detailed evaluation of the welfare impact of a single policy intervention (Chapter 4) that precisely captures selected outcomes. In Chapter 2, the example of Switzerland is used to examine the extent to which governments in highly developed welfare states alter the distribution of household income through the state budget and social security system. By means of budget incidence analyses, this contribution first determines who, over the course of one year, benefits from monetary and real public payments and services, and who bears the direct and indirect public costs. Second, from a long-term perspective, the overall redistribution is decomposed into social-balancing redistribution among households with different lifetime income and household internal income shifts across different stages of life, thus smoothing the lifetime income. Chapter 3 complements the first contribution by using leisure time as an alternative indicator to assess individual welfare. For a selection of industrialized countries, this analysis first demonstrates how leisure time, apart from market and nonmarket working hours, has evolved over the last three decades. Second, this analysis addresses the specific relationship between leisure time and income potential to estimate the extent to which the conventional, incomebased determination of welfare distribution should be put into perspective. Chapter 4 switches to the context of the less advanced world and analyzes the impact of the Indian workfare program, the National Rural Employment Guarantee Scheme. Applying non-experimental methods of program evaluation, this contribution measures the effect of participation in the scheme in terms of various expenditure categories as well as selected health indicators. In addition, this analysis addresses the question of whether public employment provisions interfere with regular economic activities by crowding out private employment, for example.
vii
Zusammenfassung Diese Dissertation handelt von der Evaluation der Wohlfahrtsverteilung in einer Bevölkerung und der Wohlfahrtsumverteilung durch den Staat. Motiviert durch die grosse und wachsende Rolle des Staatssektors – und damit des Umverteilungspotentials – zeigt sie anhand dreier Beispiele, wie Entscheidungsträger mit notwendigen Informationen zur Verteilungssituation versorgt werden können. Sie wechselt dabei zwischen stark aggregierten, möglichst das gesamte Staatshandeln umfassenden Wohlfahrtsanalysen (Kapitel 2 und 3) und einer detaillierten, die Wirkung einer einzelnen Staatsintervention präzise erfassenden Untersuchung (Kapitel 4). Kapitel 2 untersucht am Beispiel der Schweiz die von Staat und Sozialversicherungen bewirkte Umverteilung der Einkommen im weit entwickelten Wohlfahrtsstaat. Mittels Budgetinzidenzanalysen wird erstens ermittelt, wer jährlich durch die monetären und realen Staatsleistungen begünstigt und wer durch die direkten und indirekten Abgaben an den Staat belastet wird. In einer Langzeitperspektive wird zweitens die Umverteilung zwischen Haushalten getrennt von den haushaltsinternen Einkommenstransfers zwischen Lebensphasen, die nicht dem sozialen Ausgleich dienen, sondern der Glättung des individuellen Lebenseinkommens. Kapitel 3 ergänzt den ersten Beitrag indem es frei verfügbare Zeit als Indikator für die individuelle Wohlfahrt verwendet. Als erstes wird aufgezeigt, wie sich in den Industrieländern die Freizeit neben der Markt- und Nichtmarktarbeitszeit in den letzten drei Jahrzehnten entwickelt hat. In einem zweiten Schritt geht das Kapitel spezifisch auf das Verhältnis zwischen Freizeit und Einkommenspotential ein, um Hinweise darüber zu erhalten, inwiefern die konventionelle, einkommensbasierte Bestimmung der Wohlfahrtsverteilung durch die (inverse) Freizeitverteilung relativiert werden muss. Kapitel 4 wechselt zum Kontext der weniger entwickelten Welt und analysiert den Impact des indischen Workfare-Programms National Rural Employment Guarantee Scheme. Mithilfe von nicht-experimentellen Methoden der Programmevaluierung wird untersucht, wie sich die Programmteilnahme auf das Niveau verschiedener Ausgabenkategorien und ausgewählter Gesundheitsindikatoren auswirkt. Daneben wird der Frage nachgegangen, ob die staatliche Arbeitsbereitstellung das reguläre Wirtschaftsgeschehen beeinträchtigt indem sie etwa private Arbeit verdrängt.
viii
Chapter 1 Introduction
1
The promotion of general welfare is one of the principal purposes of democratic governments (Krane and Marshall, 2005). Therefore, the distribution and redistribution of welfare determines a major part of the political agenda (Boadway and Keen, 2000). On the one hand, governments, even if dedicated to free market principles, intervene for efficiency reasons. Market failures, incomplete information, and uncertainties constrain the full exploitation of the economic potential because individuals face obstacles such as uninsured income risks or disadvantageous starting positions. Public interventions that balance opportunities and provide security can bring about improvements. On the other hand, governments intervene when the free market results in welfare distributions that a majority of the population perceives as unjust. In so doing, governments complement the economic political goal of efficiency with the sociopolitical objective that each individual is guaranteed a certain standard of living, independent of his or her market success. In industrialized countries, governments intervene in several ways (e.g., see Barr, 2004). The first leverage tool is the social security system. Given the fact that the majority of the population has only minor reserves for stages of life that are without earned income, the authorities, since the end of the 19th century, have steadily expanded social insurance institutions to cover the main risks of loss of earning power, i.e., illness, age, military service, disability, unemployment, and maternity. Because social security generally includes the entire population on equal terms, regardless of individual risk potential, it combines the concept of insurance with that of solidarity, adding a redistributive element. Second, governments steer the welfare distribution through their fiscal budget. The public provision, or subsidization of goods and services, allows the promotion of the consumption of certain goods (e.g., education) as well as the support of weaker parts of the population (e.g., through public transport in peripheral regions). The financing of public expenditures through various tax types opens a third method of redistributive intervention. Finally, welfare distribution is sensitive to a wide range of regulations and policies, such as labor or price regulations, and monetary or competition policies. Less developed countries naturally have less leeway to adjust the welfare distribution (e.g., see Ray, 1998). The imposition of social security contributions is an unrealistic undertaking if common household income does not even satisfy daily needs. Similarly, the possibilities for the public provision of goods are limited if tax revenues are low and elusive. Consequently, sociopolitical interventions often are restricted to the alleviation of absolute poverty. A promising and increasingly followed strategy in this context is the promotion of workfare programs that disburse benefits in return for work (von Braun, 1995; Drèze and Sen, 1999). In doing so, workfare programs channel the scarce financial means to the most needy members of society, because this sector of the population is the only one prepared to engage in 2
low-paid work. Moreover, the workforce generated by these programs can be employed in the production of public goods, namely the development of the basic infrastructure. This strategy improves the groundwork for more productive economic activities and facilitates a self-supporting growth path. Independent of the development level of a given country, the size and scope of the state’s activities – and therewith the redistributive volume – generally grows (rapidly) over time (Tanzi and Schuknecht, 2000; Akitoby et al., 2006). In industrialized countries, public expenditures accounted for between 5 and 10 percent of the national income in the second half of the 19th century. By 2005, this proportion had risen to 35 to 55 percent (Florio and Colautti, 2005). Given this expansion in public expenditure, studying the welfare distribution before and after public interventions is of central importance, all the more so because the destitution of large parts of the world’s population persists, and strong forces continue to push towards more inequality. Amongst these forces are rising economic productivity, which leaves many less skilled people without work and fuels the disparity of market wages (Blau and Kahn, 1996; Katz and Autor, 1999), the decreasing taxability of capital, which increases the burden on the less mobile labor force (Afheldt, 2003), and unfavorable demographics wherein the share of the working population is shrinking. The understanding of welfare distribution and the distributional effects of public interventions is critical for making the normative decision as to whether there should be more or less redistribution, as well as the choice of the most efficient redistributive instruments. The informational need is twofold: On the one hand, precise impact evaluations of single interventions are necessary. Specific policy decisions cannot be made without a profound understanding of the ways an intervention affects the intended outcome. Rigor in the measurement approach, namely the choice of the appropriate counterfactual among the policy-unaffected, is thereby indispensable. On the other hand, however, there is also a demand for less precise but more comprehensive estimates of the combined effects of several, if not all, state activities. Particularly in large welfare states, interventions act in different directions so that the net effect is often unclear, leaving the general direction of a state’s impact in the dark. This dissertation focuses on the evaluation of the distribution and redistribution of welfare. The first two types of analysis address the need for comprehensive assessments in well-developed welfare states. Using the example of Switzerland, Chapter 2 analyzes current ways that the state changes the income distribution, whereby income is, as usual, interpreted as a close approximation to welfare. By means of budget incidence analysis, the question first addressed is who, over the course of one year, benefits from monetary and real public payments and services, and who bears the direct and indirect public costs. In a second step, 3
the evaluation period is extended from one year to a lifetime. As a result, the overall redistribution can be decomposed into redistribution among households with different lifetime income as well as income shifts across different stages in life. The latter redistributions are household-internal redistributions that smooth the individual lifetime income and, in the long run, do not contribute to the social balance. The results suggest that public interventions induce substantial redistribution, which is due primarily, however, to income transfers within households and not to redistribution among households. Chapter 3 complements the first analysis by using leisure time as an alternative indicator for welfare. This approach accounts for the fact that, in many cases, income and the resulting purchase of goods are insufficient to generate utility. Rather, free time resources are also necessary to make use of these goods. Specifically, Chapter 3 studies how leisure time as well as market and nonmarket work hours evolved between 1970 and 2000 in Canada, the Netherlands, Norway, the United Kingdom, and the United States. Besides filtering out the behavioral differences across these countries, the question focuses on how leisure time has developed in various educational categories. As the level of education approximates the level of income, the extent to which higher-income individuals are also better off in terms of leisure can be estimated. The results indicate that, over the last few decades, the income gains of the highly skilled population have occurred at the expense of a relative loss in leisure time. This finding suggests that welfare inequality is less pronounced than income distribution alone would suggest. Chapter 4 switches from the context of the developed world to that of a less advanced country and switches also from a bird’s eye view over all public expenditures and revenues to the precise impact analysis of a single policy. The object of investigation is India’s official workfare program, the National Rural Employment Guarantee Scheme (NREGS). Introduced in 2006, the program is the main antipoverty scheme of India’s current administration. By means of non-experimental methods of program evaluation, this third form of analysis closely examines the extent to which the program improves living conditions in rural areas. This evaluation involves the measurement of the changes seen in various expenditure categories, especially food spending, as well as improvements in health status. Another issue addressed by this analysis is whether public employment provisions interfere with regular economic activities, thereby crowding out private employment, for example. The findings suggest that the NREGS improves the economic situation of the participants without restricting households that have opportunities in the regular labor market. Further, the NREGS lowers mental distress in the form of anxiety or tension, which implies additional nonmonetary benefits as a result of improved income security.
4
Chapter 2 Measuring a State’s Overall Impact on Welfare: Public Income Redistribution in Switzerland
5
2.1
Introduction
Over the last few years, the Swiss national budget has steadily increased. In 1990, the expenditures of the state and social security institutions absorbed 40 percent of Switzerland’s gross domestic product (GDP); by 2005, this share had risen to 51 percent (compare Table 2.1). This upsurge can certainly be traced to the slowdown of economic growth. However, it also mirrors the continuous expansion of state activities. The growth of social security and social assistance benefits has been particularly dynamic, but the costs of other sectors, such as higher education (+78.6 percent), public transport (+61.2 percent), and health services (+46.6 percent), have risen substantially as well. Against this background, questions arise as to who benefits from the expanding public services and who bears the costs. These questions suggest the more basic question concerning welfare distribution in a state and the overall redistribution that is brought about by public interventions. In recent years, several analyses have addressed the issue of income distribution and redistribution in Switzerland (Economiesuisse, 2007; Künzi and Schärrer, 2004; Ecoplan, 2004; Suter and Mathey, 2000). These studies focus on monetary transfers and produced results that are in line with the international literature (for an overview, see Atkinson et al., 1995). Taxes, social security contributions, and social benefits reduce income inequalities as they induce redistribution from high-income to low-income segments of the population. For example, in their study of redistribution within the social insurance system, Künzi and Schärrer (2004) conclude that the greatest net payers in Switzerland are those in working-age households with high income, and the greatest net beneficiaries are those in retired multi-person households. Economiesuisse (2007) calculated that 22 percent of public revenues are financed by the corporate sector and 35 percent by the richest fifth of private individuals. However, despite the consistent picture that emerges from these analyses, they all have a serious shortcoming: They are limited to monetary and annual transfers. This limitation stands to reason with regard to the available data, but it comes with two important disadvantages. First, real, or in-kind public transfers, such as free education, subsidized health services, and the availability of infrastructure facilities, are ignored and with them a major part of the redistributive volume. According to the consolidated budget for the Swiss state and social insurance, real transfers amount to almost half of the total public spending (see Table 2.1). An evaluation of the redistributive impact of the state is incomplete without assessing the ways that nonmonetary transfers benefit the different segments of the population. Second, the exclusive consideration of annual transfers prohibits the consideration of two separate, but conceptually different, redistributive mechanisms. One is the inter -household 6
redistribution across households that have different income levels over the long run. The other mechanism is the intra-household redistribution of income across different phases of life within one household. While the inter-household redistribution aims at a convergence of lifetime income across households, the objective of the intra-household redistribution is to smooth the income over the life cycle. Such smoothing is achieved by shifting the income from one stage of life to the next, primarily from the working stage to the retirement stage. Not distinguishing between these two mechanisms risks overestimating the redistribution across households. The most evident example is the old-age provision in which contributions serve primarily to maintain the income level after retirement. In particular, the occupational pension system is unlikely to reduce inter-household income differences because, in capitalbased systems, those citizens who make high contributions are also those who have high entitlements in the future. In contrast, and not least because of the tax deductibility of above-compulsory contributions, regressive redistribution effects are also possible. This analysis focuses on these analytical shortcomings and examines the redistributive impact of the state, inclusive of social insurance at the level of individual households. Two sets of questions are examined: 1. What are the redistributive effects of the entire state as well as of individual revenues and expenditures? Who, over the course of a year, benefits from monetary and real public payments and services, and who bears the direct and indirect public costs? And who are the net beneficiaries? 2. To what extent does an annual redistribution involve (a) redistribution across households with different lifetime or long-term income, and (b) redistribution across different phases of life within the same households in order to smooth lifetime income? Whereas the first set of questions is connected to the first redistribution analysis conducted for Switzerland, undertaken by Leu et al. (1988), and aims to update and expand upon those results, the second question addresses a field of research that has been widely neglected in Swiss as well as international literature. The methodological approach is thus twofold. The first step consists of conventional analysis of the annual budget incidence for the years 1990, 1998, and 2000 to 2005. Secondly, these annual analyses are used to generate a pseudo panel on the basis of which the course of lifetime income before and after public transfers can be constructed. A decomposable inequality measure allows for the separation of inter-household income redistribution from intra-household income shifts over the life cycle.
7
Table 2.1: Consolidated budget of the confederation, cantons, municipalities, and social insurance institutions, 1990, 2000, 2005 EXPENDITURES (prices 2005)
8
General administration Justice, police, fire brigade National defence International affairs Regional and district planning Environment Education State schools, primary education General education schools, vocational training University-level institutions Culture and recreation Health Transport Road traffic Public transport Subsidies Agriculture Others Payments-out of social insurance (SI) Old-age insurance (AHV) Pension plans (BVG; incl. pre-retirem. benefits) Disability insurance (IV) Mandat. health ins. (incl. premium reductions) Unemployment insurance (ALV) Family and child allowances Accident insurance (UVG) Income compensations (EO) Other expenditures Other social welfare Social assistance Other benefits of social welfare Finance, debt service Total expenditures state and SI In % of GDP
1990 7.6 6.2 8.4 2.0 2.0 2.7 21.1 10.8 6.5 3.8 3.9 13.5 11.6 6.8 4.8 6.5 4.0 2.5 69.7 24.5 14.4 5.4 9.4 0.5 3.7 3.5 1.1 7.2 5.3 2.6 2.7 6.9 167.6 40.1%
in bn CHF 2000 2005 8.2 9.0 7.1 8.1 5.6 4.9 2.4 2.4 2.1 2.1 2.9 2.8 23.8 27.4 12.2 13.7 6.6 7.0 5.0 6.7 4.0 4.2 16.3 19.8 13.6 14.8 6.9 7.1 6.7 7.7 7.1 6.2 4.4 4.2 2.7 2.0 100.8 114.1 30.3 32.9 26.3 28.2 9.6 12.3 13.8 17.4 2.9 5.8 4.4 4.7 4.1 4.7 0.7 0.8 8.8 7.2 8.2 9.3 5.5 6.0 2.8 3.3 11.3 9.5 213.5 234.8 48.5% 50.7%
change 1990-2005 18.2% 30.0% -41.7% 22.5% 7.7% 2.7% 30.2% 26.5% 8.1% 78.6% 7.4% 46.6% 28.0% 4.6% 61.2% -5.1% 4.7% -21.0% 63.6% 34.1% 96.4% 127.0% 84.6% 1040.5% 27.3% 35.0% -25.2% 0.4% 73.4% 130.9% 18.8% 38.3% 40.1%
REVENUES (prices 2005) Taxes on income and property Income and property taxes Earnings and capital taxes (corporate taxes) Withhold. tax, stamp duty (excl. foreign contr.) Real estate taxes Taxes on property: Tax on motor vehicles Excise duties Value-added tax Mineral oil and fuel taxes Tax on tobacco products Vehicle taxes (heavy traffic, nat. routes) Customs duties Royalties and concessions Other taxes Remunerations Financial and investment income Payments-in to SI (insurants & employers) Old-age/invalidity insurance (AHV/IV), EO Occupational pension plans (BVG) Mandatory health insurance ALV (excl. refunding to border crossers) Family and child allowances Accident insurance (UVG) Capital and other revenues of SI Occupational benefit institutions Other SI
Total revenues state and SI In % of GDP Surplus revenue state and SI due to state budget due to SI budget
1990 54.9 37.5 9.4 3.9 4.1 1.5 17.6 12.5 3.9 1.2 0.4 1.5 1.1 2.0 16.2 6.0 69.2 24.4 27.7 8.4 0.8 3.7 4.2 16.5 13.9 2.7
in bn CHF 2000 64.4 43.0 13.0 5.6 2.8 1.7 24.6 17.3 5.5 1.7 0.7 1.1 1.8 1.6 21.9 14.2 83.6 25.7 30.8 11.3 6.5 4.5 4.9 20.4 17.3 3.1
2005 67.5 48.4 12.8 3.5 2.9 1.9 25.5 18.1 5.3 2.1 1.5 1.0 1.3 1.6 25.0 13.1 94.0 28.0 35.8 15.3 4.3 4.8 5.8 17.7 14.9 2.7
187.0 44.7%
236.0 53.6%
250.1 54.0%
19.4 -8.3 27.2
22.5 3.0 20.0
15.3 -1.8 17.5
change 1990-2005 22.9% 28.8% 35.8% -9.9% -30.2% 29.5% 44.6% 45.2% 36.4% 64.2% 271.8% -35.8% 15.6% -16.3% 54.0% 117.5% 35.9% 14.8% 29.3% 82.2% 430.2% 29.4% 38.3% 6.8% 7.6% 2.6%
33.7%
Note: Revenues exceed expenditures in the consolidated budget of state and social insurance because the latter records a surplus revenue. The reason for this lies primarily in the second pillar of the old-age provision system (occupational pension plans, BVG), which is still in the expansion phase and, therefore, pre-finances future benefits to a higher degree than it pays out in current pensions. Considering the budget without social insurance shows familiar deficits for the years 1990 and 2005 (see the lower three rows in the right side of the table). Abbreviations: SI – social insurance; AHV – Alters-und Hinterlassenenversicherung, old-age and survivors’ insurance; ALV – Arbeitslosenversicherung, unemployment insurance; BVG – Berufsvorsorge(gesetz), occupational old-age insurance, i.e. pension plans; EO – Erwerbsersatz(ordnung), income compensations; IV – Invalidenversicherung, disability insurance; UVG –Unfallversicherung, accident insurance. The category ’other taxes’ includes inheritance and gift taxes, incentive taxes, beer taxes, lottery taxes, and other property and excise taxes. Sources: Public finance figures (FFA, 1990-2005), Swiss Social Insurance Statistics (FSIO, 1990-2005)
The results show that the state has a substantial redistributive impact. This impact stems mainly from the public expenditure side, which considerably reduces the income differences in the population. Unlike the implications of short-term effects, however, the convergence of income is not due primarily to the social balancing across households but to the stateprescribed transfers of income across the different stages of life. Income peaks during the working years are broken markedly by an enhanced contribution burden during that period. In return, household income is prevented from declining not only after retirement, but also in times when the labor market or health-related or family-related issues disrupt earning capacity. As a consequence, the variance in lifetime income is clearly reduced. This chapter is structured as follows. The next Section 2.2 discusses the methodological basis, especially the concept of budget incidence analysis, the generation of the pseudo panel using the available data, quantification of the inter- and intra-household redistribution via decomposable inequality measures, and the statistical test procedure chosen. Section 2.3 first presents the results from an annual perspective for vertical and horizontal redistributive effects, then from the long-term perspective. Section 2.4 compares the results to those found in the extant literature and provides guidelines for cautious interpretation. Section 2.5 summarizes and concludes this chapter.
2.2 2.2.1
Methodological Approach and Data The Concept of Budget Incidence Analysis
This chapter studies redistribution on the basis of budget (or fiscal) incidence analysis. Therefore, the pre-fisc income before transfers, i.e., the income that would exist if there were no state and social insurance institutions, is determined for each household. This prefisc income is the reference base for a household’s post-fisc income, which is the income that remains after taxes and social security contributions (negative transfers), and after the receipt of public services and benefits (positive transfers). By considering positive and negative transfers, budget incidence analysis combines expenditure incidence analysis and tax or revenue incidence analysis. The former addresses the question of who finances the state’s activities, and the latter as to who enjoys the benefits. The computation of post-fisc income involves two steps. The first step is to determine the total size of each transfer category. For this, the annual governmental financial statements (including social insurance institutions, see Table 2.1) are referenced: The size of the negative and positive transfer categories follow from the positions on the revenue and expenditure 9
sides, respectively. In the second step, the transfer totals are allocated to individual households. The negative transfers are assigned according to the contribution burden borne by the households, and the positive transfers are allocated according to the benefits received by the households. In the case of nonmonetary public services, households are assigned shares of the provision costs according to their service utilization. These shares approximate the amount that the income of the households would have to rise if they had to pay for the public services accessed.1 Formally, the estimation of post-fisc income involves Yh, post-f isc = Yh, pre-f isc +
n X i=1
m
Uih X Bjh − Cj Si Ui Bj j=1
(2.1)
whereby Yh is household income, Si denotes the total transfer size of public service i, and Cj the total transfer size of public charge j. U represents utilization and B represents burden. The public services and goods allocated to a household thus depend on its individual utilization (e.g., the number of school days that are provided to its children) compared to the aggregated utilization of the total population (e.g., the total number of school days provided). Accordingly, the charges assigned correspond to the individual burden of a household (e.g., its individual income tax obligation) as compared to the statewide burden (total public revenues from income taxes). The proportion of total public expenditures that reach a household is n n n 1 X Uih X Si Uih X Si = = si uih S i=1 Ui S U i i=1 i=1
(2.2)
whereas the proportion of all public revenues borne by a household is given by m m m 1 X Bjh X Cj Bjh X Cj = = cj bjh C j=1 Bj C B j j=1 j=1
(2.3)
whereby si denotes the share of a service i in total services, i.e., in total public expenditures, and uih stands for the share of a household’s utilization within the statewide utilization. Analogously, cj is defined as the share of charge j in total charges or public revenues, and bjh signifies the share of a household’s burden within the statewide burden. The two equations indicate that it depends on two factors as to whether public interventions reach a household. The first is the structure of public expenditures and revenues; the second is the household’s 1
More convincing from a theoretical point of view would be the allocation of public services according
to a household’s willingness to pay and the utility derived from the consumption of the services. However, households do not have to reveal such willingness for publicly provided goods (i.e., no price formation takes place).
10
characteristics that determine the extent to which it demands public services and participates in the funding. The juxtaposition of a household’s economic situation before and after the allocation of (all or individual) public expenses and revenues allows an estimation of the extent of the (total or intervention-specific) welfare redistribution. Thus, welfare is approximated by the sum of private and state-provided monetary and in-kind income. Generally, and also in the most parts of this study, (progressive) redistribution is defined as the transfer of resources and purchasing power from the rich to the poor. Hence, if redistribution occurs, high income decreases and low income increases, both leading to fewer income inequalities in the population. A government can redistribute over the expenditure side of its budget by favoring low-income households compared to high-income households. Or, it can trigger redistribution over the revenue side by taxing low-income households relatively less than high-income households. An expenditure-side intervention is called progressive (regressive) if it enhances (lowers) benefits with decreasing income. Accordingly, a revenue-side intervention is progressive (regressive) if it lowers (enhances) the contribution burden with decreasing income.
2.2.2
Counterfactual Income
A basic problem of budget incidence analysis is that a household’s pre-fisc income is not observable, but rather, must be constructed. In this analysis, pre-fisc income is defined as the sum of earned income, capital income, and private transfers before taxes and social security contributions. For homeowners, an imputed rent of the property is added in order to better assess their economic strength compared to that of tenant households.2 Included as well are implicit taxes that other tax debtors roll off and that lower household income compared to a situation without state intervention. Ranked among such implicit charges are employers’ contributions to the social security system. By their nature, these contributions are wage components that, deducted at the source, are transferred directly to the state. Depending on the incidence assumptions (see Section 2.2.3), corporate taxes are also borne indirectly by private households: If companies distribute profits after taxes, their tax burden eventually falls on wealthy households whose profits per share are diminished. Taken together, these components result in an average pre-fisc income, as shown in Table 2.2. For lack of alternatives, such a construction of the pre-fisc counterfactual (also referred to as the zero-budget approach) has been used regularly in budget incidence analysis (Demery, 2000; Ruggeri, 2003). Nevertheless, the approach is not fully satisfactory from a theoretical 2
In some budget incidence analyses, unrealized capital gains are also added to pre-fisc income. This
addition is not possible with the data at hand.
11
Table 2.2: Average pre-fisc household income, 2005
Earned income, cash and in-kind, including expenses (before taxes) Capital and rent income (before taxes) Private transfers, such as private pensions, gifts, alimonies (before taxes) Imputed rent of house owners Employers’ social security contributions (17.2% of employee income) Shifted corporate taxes (50% according to standard scenario of Section 2.2.3) Total pre-fisc income
Mean in CHF 51’000 2’700 3’000 2’800 8’000 1’300 68’800
Note: These figures are based on the Survey of Income and Expenditure (SFSO, 1990-2005). The imputed rent of a homeowner household equals the average rent expenses of nonhomeowners in the same income category minus mortgage expenses. The employers’ social security contributions for year 2005 consist of 4.2% old-age insurance (AHV), 0.7% disability insurance (IV), 0.2% income compensations (EO), 1% unemployment insurance (ALV), 0.9% occupational accident insurance (UVG), 8.6% occupational pension plans (BVG), and 1.7% family allowances (FSIO, 1990-2005).
point of view. Its shortcoming is that it cannot capture the different allocations and distributions of resources that would occur in the absence of the state. Rather, the zero-budget approach is based on the given intensity of public interventions and the resulting adjustments of the economic agents. When comparing pre-fisc with post-fisc income, this situation must be considered. For example, if households without unemployment compensation obtained new employment more quickly after a job loss, their earned income would be higher without public unemployment insurance. As a consequence, the progressivity of the unemployment benefits is likely to be underestimated. In contrast, the progressivity of certain public services might be overestimated if their provision via wages to civil servants primarily benefits middle and higher income groups.3
2.2.3
Incidence Assumptions
A second difficulty of budget incidence analysis is the allocation of public revenues and expenditures to households according to their individual burdens and benefits. The problem 3
To overcome the drawback of having to construct pre-fisc income, some analyses have used “distribu-
tionally neutral income” as the benchmark (e.g. Dyck, 2005). Instead of estimating household income for a situation without state intervention, they determine the income that would result if public revenues and expenditures affected households proportional to their income. In doing so, such analyses do not compare income distributions before and after transfers, but with and without redistributive interventions. This method has the advantage that the compared distributions are based on the same intensity of public interference. However, the new benchmark also has to draw on the observable state-affected income distribution, so that the basic problem is not solved.
12
is that only personal taxes, social security contributions, and monetary benefits are directly evident from available data. In contrast, burdens of implicit taxes as well as benefits drawn from in-kind public services must be estimated. This effort requires assumptions as to who, independent of formal tax liability, ultimately finances the state budget, and for whom public services effectively prove advantageous. As far as possible, this analysis adopts the incidence assumptions that are generally accepted in the literature (see Boadway and Keen, 2000; Fullerton and Metcalf, 2002; Ruggeri, 2003). Table 2.3 shows the standard allocation rules that follow from these assumptions, as well as alternative progressive and regressive scenarios for cases that elicit different views. Revenue incidence assumptions With regard to tax incidence, the general consensus for personal income and property taxes is that no shifting takes place and the formal tax debtor is also the effective taxpayer. Therefore, public revenues from these taxes are distributed among households according to their income and property taxes paid. No such consensus exists for corporate taxes, however. One literature camp considers the capital owners as the payers of these taxes, because companies distribute profits only after taxes. The other camp attributes the tax burden to the consumers because companies can roll off their taxes by raising product prices. This analysis takes a middle position and allocates half of the earnings and capital taxes on capital owners (approximated by the capital income of the households) and half on consumers. For real estate taxes, where a similar disagreement exists, the same approach applies: 50 percent of the tax burden is allocated to the real estate owners according to their real estate taxes paid, and 50 percent to the tenants according to their rent expenses. In the progressive scenario, the corporate and real estate taxes shift entirely to the capital and real estate owners, respectively. In the regressive case, the consumers fully bear the corporate taxes. The standard assumption with regard to excise duties is that they are shifted to consumers via higher prices. Accordingly, value-added taxes, fuel taxes, taxes on tobacco products, as well as vehicle taxes, customs duties, royalties and concessions are distributed in proportion to the (specific) consumption expenditures of the households. However, the progressive scenario allows for price increases to regularly induce wage adjustments and social transfers so that the excise duties should be assigned to household income instead. For remunerations, two positions are considered as well. One reflects the view that remunerations compensate the state for providing its services and are therefore borne equally by the individuals in each household. The other position implies that the well-off populace more frequently makes use of chargeable services and also is more likely to buy more expensive 13
Table 2.3: Criteria for the allocation of public revenues and expenditures Standard scenario Taxes on income and property Income and property taxes Earnings and capital taxes Withholding tax, stamp duty Real estate taxes Property taxes: Tax on motor vehicles Excise duties Value-added tax Fuel tax Tax on tobacco products Vehicle taxes (heavy traffic, nat. routes) Customs duties Royalties and concessions Other taxes Remunerations Payments-in SI (insurants & employers) Old-age/invalidty insurance (AHV/IV), EO Occupational pension plans (BVG) Health insurance, mandatory part Unemployment insurance (ALV) Family and child allowances Accident insurance (UVG) Capital and other revenues of SI Occupational benefit institutions Other SI Financial and investment income General administration Justice, police, fire brigade National defence International affairs Regional and district planning Environment Education State schools, primary education General education schools, vocational training University-level institutions Culture and recreation Health Transport Road traffic Public transport Subsidies Agriculture Others Payments-out of social insurance (SI) Old-age insurance (AHV) Occupational pension plans (BVG) Disability insurance (IV) Health insurance, mandatory part Unemployment insurance (ALV) Family and child allowances Accident insurance (UVG) Income compensations (EO) Other expenditures Other social welfare Social assistance Other benefits of social welfare Finance, debt service
Progressive
Regressive
Paid income and property taxes 50% capital inc./ 50% consumption exp. Capital income Capital income 50% real estate taxes/ 50% rent exp. Real estate taxes Paid taxes on motor vehicles Total consumption expenses Expenses for fuel Expenses for tobacco products Total consumption expenses Total consumption expenses Total consumption expenses Total income 50% total income/ 50% household size Employees’ Employees’ Health ins. Employees’ Employees’ Employees’
Earned&SI Earned&SI Earned&SI Earned&SI Earned&SI Earned&SI
Consump. exp.
income income income income income income
Total income
Household size
Household Household Household Household Household Household
Total Total Total Total Total Total
AHV/IV/EO contributions BVG contributions rate minus rate reductions ALV contributions ALV contributions UVG contributions
Employees’ BVG contributions Allocated SI contributions Allocated other taxes and charges 50% 50% 50% 50% 50% 50%
total total total total total total
income/ income/ income/ income/ income/ income/
50% 50% 50% 50% 50% 50%
household household household household household household
size size size size size size
size size size size size size
income income income income income income
No. of school children No. of persons in secondary education No. of persons in tertiary education Exp. for entertainm./recreation/culture Age- and gender-specific hhd. profile Expenses for fuel Expenses for public transportation Income from agriculture Total consumption expenses Received AHV benefits Received BVG benefits Received IV benefits Received health insurance benefits Received ALV benefits Received family allowances Age- and gender-specific hhd. profile No. of men in military/civil service age Received social benefits Received social assistance Received social asst./benefits ex BVG Allocated public exp. and social benefits
Capital income
Note: ALV – unemployment insurance; AHV – old-age and survivors’ insurance; BVG – occupational old-age insurance; EO – income compensations; IV – disability insurance; SI – social insurance institutions; UVG – accident insurance. The public expenditure category ‘Health’ contains subsidies to the health sector (e.g., the financing of the medical infrastructure), whereas the category ‘Health insurance’ includes insurance benefits in the case of damage.
14
services (e.g., in the old-age and nursing home sector which fees constitute an important part of the remunerations). Again, the standard scenario adopted in this analysis represents the middle ground and allocates remunerations one half each according to the size and income of the households. The alternative scenarios adopt the extreme positions. Social security fees are attributed entirely to the insurants proportional to their contributions. This arrangement reflects the common understanding that the employers’ contributions are an implicit part of salaries. The capital revenues of the social insurance system, which are accumulated mostly by the occupational benefit institutions, are thought of as indirect contributions and therefore are distributed among households according to paid contributions as well. This arrangement reflects the opinion that, in the absence of pension plans, capital gains would be garnered privately and could be used at liberty. Similarly, the public financial and investment income is charged according to taxes paid, because, in the end, the necessary funds must be procured by the taxpayer. Expenditure incidence assumptions On the side of public expenditures, a distinction must first be made between public goods in the economic sense and goods or services that are publicly provided but privately used. The former includes, for example, the general administration, national security, or involvement in international affairs. The allocation of expenses for these typically generally accessible, nonrivaling goods is not evident. Incidence analysis often employs the concept of allocation per capita, which is based on the conventional assumption of the collective goods theory whereby everybody benefits equally from public goods. Apart from this, public goods are also assigned according to household income, thereby assuming that they benefit those who participate in economic production and rely on a functioning business context. The standard scenario apportions the public goods one half each according to household size and income, whereas the progressive and regressive variants again take the extreme positions. For privately consumed public services, identification of the beneficiaries is easier. In the case of public transport services, road facilities, or subsidized cultural institutions, private outlays are necessary in order to benefit from the state-provided goods. The allocation of these services can then be oriented towards specific household expenses. Other services do not require a cost participation, but the recipients of the benefits are identifiable. This scenario applies directly to the education sector where public expenditures can be assigned according to the number of students per household. In the case of the health sector, the costs are allocable according to the age-gender profiles of the household, because these profiles determine the utilization of health care services.
15
The allocation of monetary benefits rests upon the hypothesis that the recipient of a payment equals the effective beneficiary. Social insurance and social assistance are thus assigned proportionally to the transfers received. For agricultural subsidies, it is assumed that they, in the form of direct payments or production grants, stay with the farmers. Accordingly, their allocation is based on farming income. The remaining subsidies are distributed according to household consumption expenditures, assuming that they lead to price reductions. Finally, the expenses required to serve the national debt, in the standard scenario, are apportioned to households according to the public expenditures assigned to the households up to that point. This arrangement reflects the view that the public debt favors mainly those who also receive the most benefits from the state. Departing from this arrangement, the regressive scenario takes the position that debt service benefits the owners of government securities and thus allocates the debt interests according to capital income.
2.2.4
Data
The first database for this study comprises the financial statements of the confederation, the cantons, the municipalities and social insurance institutions; these statements are consolidated to obtain the overall national budget shown in Table 2.1. The (consolidated) data for the regional corporate bodies are obtained from the Swiss Federal Finance Administration, and those for the social security institutions come from the Federal Social Insurance Office.4 The calculation of the pre- and post-fisc household income calls for detailed information about income and expenses, as well as demographic and socioeconomic characteristics, such as age, gender, education, occupation, and employment status, of each household. This information can be extracted from the Survey of Income and Expenditure (SIE)5 , which the Federal Statistical Office conducts yearly and is based on 3000 to 4000 voluntarily participating households. These data contain necessary information both to calculate the pre-fisc income and to assign public revenues and expenditures according to the aforementioned allocation rules.6 This study uses the annual SIE survey data collected between 2000 to 2005, as well as data from two earlier surveys conducted in 1990 and 1998. 4
To break down the health sector costs into age- and gender-specific outlays, hospital surveys conducted
by the Federal Statistical Office are additionally used. 5 Einkommens- und Verbrauchserhebung (EVE). 6 Specifically, for each category of public revenues and expenditures, first, each household’s share in the total of the respective allocation criteria is calculated. In the case of primary education, for instance, each household’s share in the total (SIE) population of schoolchildren is computed. In this way, distribution keys for all revenue- and expenditure-side transfers are generated, which, in the second step, serve to allocate the state budget (scaled down to the SIE sample size) to individual households.
16
Because pre- and post-fisc income is computed on a household level, each refers to households of different size and composition. This factor raises the question of inter-household comparability of the results and the effective welfare level of the individual household members.7 The common approach taken to address this issue is the utilization of equivalence scales. This device assigns a weight (equivalence value) to each member in a household according to his or her consumer needs, thereby taking into account that large households realize economies of scale. In this study, which uses the OECD equivalence scale, household heads are weighted as 1, household members above the age of 15 as 0.5, and children as 0.3. Dividing the household income by the sum of these values results in comparable single-person income, that is, income per adult equivalent.
2.2.5
Rationale and Construction of the Pseudo Panel
Similar to previous budget incidence analyses, this work first computes pre- and post-fisc household income on an annual basis. A comparison of their distributions allows the assessment of social cohesion in the population at a certain time, in this case, within a year. For an appropriate picture of the public redistribution, however, these snapshots must be enhanced by a long-term perspective. As mentioned in the introduction, part of the transfer volume is caused by streams of funds from working to nonworking and retired households. Because the contributors become entitled to future disbursements, these transfers induce intra-household income shifts across the different phases of life rather than redistribution between high- and low-income households. Therefore, a comparison of annual pre- and post-fisc income is, in a second step, supplemented by a comparison of lifetime, or, more precise, long-term income before and after transfers. This comparison requires information on the yearly income of individual households for a long period of time, ideally for a whole life cycle. In other words, this part of the analysis demands panel data by which households can be tracked over time. Unfortunately, redistribution analysis can seldom draw on such data. An exception is the work of Björklund and Palme (2002) in which market income is contrasted against income after taxes and social transfers for Sweden and a (incomplete) life cycle. The method of pseudo panel construction, developed by Deaton (1985) and Browning et al. (1985), nonetheless provides a possible way to conduct longitudinal analysis. The idea is that in a series of cross-sectional data, although a household cannot be followed up over time, 7
Private households are usually subdivided into single- and multiple-person households. The latter are
further broken down into family households (i.e., couples with children, single parents, and adults with parents) and nonfamily households.
17
a cohort, i.e., a clearly distinguishable population segment, such as persons born in the same year, can be tracked. If annual surveys are carried out with a representative selection of the population (which the SIE does), the average behavior of a cohort in one period can be related to its average behavior in the next period. Thus, instead of tracking individual households over time, cohorts are observed over time. Given that cohorts are big enough (more than 100 individuals per cohort; see Verbeek and Nijman, 1992), the cohorts’ means form a data panel that represents the behavior of the underlying households. The present work adopts this procedure for the available cross-sectional data sets. In each survey, households are allocated to one of nine cohorts according to the year of birth of the household head (-1934, 1935-1939, 1940-44, ..., 1970-74). The average pre-fisc and post-fisc income per cohort is then combined into a pseudo panel (see Table 2.14 in the Appendix), which has two uses: First, the pseudo panel serves to simulate a representative life cycle of income before and after transfers. Secondly, as discussed in the next section, it allows an estimation of the shares of inter- and intra-household income redistribution in total redistribution.
2.2.6
Measurement of Redistribution
In order to capture the comparisons of the pre- and post-fisc income distributions quantitatively, it is useful to express a given income distribution through a single measure. This task is performed by an inequality measure, which increases in value when a monetary unit flows from a poor household to a rich household. For the purpose of this analysis, a decomposable measure is chosen whereby total inequality in the population can be divided into the inequalities within individual, mutually exclusive groups and into the inequalities across these groups. Cowell (1984) shows that a generalized entropy measure meets these requirements:
Iα =
1 n
Pn
yi α ) − 1] µ α2 − α
i=1 [(
(2.4)
where yi is the income of household i, n the number of households, and µ the average income. The parameter α determines if the inequality measure reacts more strongly to inequalities in the upper (α > 0) or lower (α < 0) income segments. α thus indicates the degree of poverty aversion. The decomposition is defined as K X nk µk α nk 1−α k ) ( ) Iα + IαB = IαW + IαB Iα = ( nµ n k=1
18
(2.5)
where k is an index for K mutually exclusive groups, nk denotes the number of households in group k, and µk is the average income in group k. Consequently, nk µk /nµ is the income share of group k in total income, and nk /n its share in total population. Iαk denotes the generalized entropy measure within group k, IαW is the weighted average of these group-internal inequality values (within group inequality), and IαB is the generalized entropy measure of the distribution of the group means µk (between group inequality).8 In this analysis, the undecomposed inequality measure is used as a simple means to quantify the redistribution in the annual analysis. By comparing the inequality measures before transfers, Iα,pre-f isc , with those after (all or individual) transfers, Iα,post-f isc , and statistically testing the resulting differences, the extent to which public interventions affect income distribution can be determined. The decomposition of the inequality measure is used to separate the inter-household from the intra-household redistribution. To this end, total inequality Iα is calculated for the pooled sample (all households in all cross-sectional data sets) before and after transfers (i.e., Iα,pre-f isc and Iα,post-f isc ). Then, the cohorts described in the previous section are taken as the mutually exclusive groups. The inequality between these cohorts, IαB , is the inequality of the average income that was earned within the individual cohorts from 1990 to 2005. The B B comparison of the IαB before and after transfers (Iα,pre-f isc , Iα,post-f isc ) shows which portion of
the total redistribution is due to redistribution across cohorts. This comparison indicates the redistribution across households with different long-term income. The residual Iα − IαB = IαW shows the inequality within the individual cohorts and represents the income variations over W W time. The comparison of IαW before and after transfers (Iα,pre-f isc , Iα,post-f isc ) indicates which
shares of the total redistribution are household-internal income shifts across the different stages of life. 8
In case of α = 0 and α = 1, the weights of Iαk reduce to nk /n (population weights) and nk µk /nµ (income
weights), respectively:
lim Iα = I0 =
α→0
nk n K K X X 1X µ nk X 1 µk nk µ ln( ) = ln( ) + ln( ) n i=1 yi n i=1 nk yi n µk k=1 | {z } |k=1 {z } I0k I0B
(2.6)
and nk n K K X X yi nk µk X yi yi nk µk µk 1 X yi ln( ) = ln( ) + ln( ) . lim Iα = I1 = α→1 n i=1 µ µ nµ i=1 nk µk µk nµ µ k=1 {z } |k=1 | {z } I1k I1B
I1 is also referred to as Theil index.
19
(2.7)
2.2.7
Statistical Test for the Significance of the Redistribution
If the discussed inequality measure is to answer the question of whether state interventions actually level out income distribution, then changes Iα,pre-f isc − Iα,post-f isc = ∆Iα must be tested for their positive deviance from zero. In the present analysis, this test is based on a nonparametric bootstrap. Being a very general approach, it has the advantage of not requiring any assumptions regarding the distribution of the variable to be tested. A bootstrap test attributes the same probability of occurrence to each observation (i.e., household). Then, out of these equiprobable observations, many new samples of the same size as the original sample are drawn (1000 draws, with replacement). Thereafter, both for the original sample and for the resamples, the differences between the pre-fisc and post-fisc inequality measures are calculated. In this manner, ∆Iα is replicated many times, which eventually allows the determination of the distribution and confidence intervals of ∆Iα for different significance levels. If zero is not part of these intervals, an intervention can be said to have a significant redistributive impact.
2.3 2.3.1
Results Redistribution within a Year: General Overview
In Switzerland, the state produces considerable redistribution. If the pre- and post-fisc income is computed, as described in the previous section, then the income distributions shown in Figure 2.1 result.9 The reduced income inequality after allocation of the public transfers is unmistakable: The high share of households, which in the absence of state and social insurance institutions would have no or only a very small income, has virtually disappeared. Likewise, income beyond 140’000 francs (CHF) occurs only rarely. As a consequence, postfisc distribution concentrates more strongly on mid-level income. This picture is confirmed by the location and dispersion parameters shown in Table 2.4. Through the transfers, the income variance decreases by two-thirds. The smaller difference between the first quintile and the mean indicates that the transfers lessen the right skewness of the income distribution, i.e., the concentration on relatively low income.10 From the breakdown of income by pre-fisc income deciles (lower part of Table 2.4), it can be seen 9 10
The results are reported as income per adult equivalent throughout Chapter 2. The 1st decile/quintile/median indicates the income level which 10/20/50 percent of households does
not surpass.
20
1.0e−05 5.0e−06 0
Density
1.5e−05
Figure 2.1: The distribution of pre- and post-fisc income, 2005
0
100’000
200’000
300’000
200’000
300’000
Density
0
5.0e−06 1.0e−05 1.5e−05
Pre−fisc income in CHF
0
100’000
Post−fisc income in CHF
that public interventions reduce the income of the upper six deciles, whereas they enhance the income in the lower four deciles. As expected, the burden of the net contributors rises with income. For example, households in the fifth decile with a mean pre-fisc income of 57’900 francs lose 5’400 francs on average, whereas those in the tenth decile forfeit 80’200 francs of their pre-fisc income of 182’000 francs. On the side of the net beneficiaries, on the other hand, the lowest decile gains the most from the state; on average, it receives 67’500 francs. All in all, 40 percent of households receive a positive net transfer, i. e. they are net recipients. Figure 2.2 illustrates the redistribution by means of Lorenz curves. The solid curves represent the ordinary Lorenz curves before and after transfers and show the income shares that fall to different population shares. The 45 degree line serves as a reference and represents the equal distribution of income. Due to the transfers, the Lorenz curve moves closer to the 45 degree line, which points out the reduction in income inequalities. For instance, in the absence of state interventions, the 20 percent poorest households together would receive barely 3 percent of the total income, whereas the 20 percent poorest households in the present situation actually receive 12 percent of the total income. 21
Table 2.4: Key figures of pre- and post-fisc income distributions, 2005 (in CHF) Variance Mean Median First quintile Average income per pre-fisc decile 1 2 3 4 5 6 7 8 9 10
Pre-fisc income 3.44E+09 68’788 62’656 20’786
Post-fisc income 1.13E+09 65’813 59’228 44’194
Net transfer
70’000 72’500 61’500 54’900 52’500 54’900 57’600 64’600 68’000 101’800
67’500 59’100 31’500 9’100 -5’400 -13’700 -22’900 -29’100 -45’900 -80’200
2’500 13’400 30’000 45’800 57’900 68’600 80’500 93’700 113’900 182’000
Note: The reduction of mean income through transfers is due to the fact that the consolidated state and social insurance budget is allocated to households. As mentioned, the social insurance sector accounts for a surplus revenue because of the occupational pension system that is still in the expansion phase. Consequently, more public revenues than expenditures are assigned to households, thus leading to a negative average net transfer.
Particularly interesting is the ‘modified’ Lorenz curve. It shows the proportion of total postfisc income that falls upon the households, which are ordered by increasing pre-fisc income (as opposed to increasing post-fisc income, as is the case of the ordinary post-fisc Lorenz curve). The modified Lorenz curve indicates that the poorest 40 percent of households, as measured by pre-fisc income, receive only about 13 percent of the total income before transfers, but 40 percent of the total income after transfers. This distribution suggests that for low-income groups, state interventions even lead to an elimination of income inequalities.
2.3.2
Incidence of Individual Public Interventions
The results derived so far indicate that the entirety of public transfers brings the annual resources available to households closer together. Next, this general picture is divided into the redistributive impacts of the individual interventions. Absolute expenditure incidence Table 2.5 shows, per income quintile, the benefits and burdens that accrue to an average 22
100%
Figure 2.2: Lorenz curves for pre- and post-fisc income, 2005
Post−fisc Lorenz curve Modified Lorenz curve
40%
60%
45 degree line
0
20%
Cumulative proportion of income
80%
Pre−fisc Lorenz curve
0
20%
40%
60%
80%
100%
Cumulative proportion of households Note: The pre-fisc and post-fisc Lorenz curves show, respectively, which proportions of the (pre-fisc or post-fisc) income fall upon different households shares for households that are ordered according to pre-fisc and post-fisc income. The modified Lorenz curve maintains the order of the households according to pre-fisc income and shows which proportion of post-fisc income falls upon the households.
household from the public expenditures and revenues, respectively. If the expenditure side and the absolute values are considered first, it is seen (upper left part of Table 2.5) that public spending favors households with the lowest pre-fisc income most and decreases this favoritism with increasing income. The combined benefits for the first quintile sum up to 88’000 francs on average and are thus almost twice as high as the benefits to the average household. The benefits for the second quintile with 51’960 francs are still 12 percent above the average. The households in the third to fifth quintiles each receive about 30’000 francs from public expenditures, which is about one-third below the average gain. The breakdown of expenditures into individual parts shows that the provision of general public goods, the expenditures for individual and for public transport, as well as subsidization of culture resources, reaches high-income households to a greater extent than it reaches lowincome households. In the case of public goods, this finding expresses the assumption of the 23
standard incidence scenario, wherein households with high income and wealth benefit more from a favorable (business) environment created, for example, by an efficient public administration, a functioning legal system, or good foreign relations. In the case of transportation infrastructure and cultural subsidies, well-off households profit more than low-income households because they are better able to afford the private outlays necessary to make use of the public services. In contrast, public spending for the health sector as well as for social security transfers is concentrated in low-income households. As indicated by the benefits of the old-age provisions (AHV and BVG), the state effectively protects income during times when earning ability is no longer feasible. In the first quintile, which consists mainly of retired households, old-age benefits boost income by almost 45’000 francs. In the second quintile, the remittances still amount to 13’000 francs and can be traced back largely to partly retired households. For the upper income quintiles, which contain only a few pensioner households, the AHV and BVG benefits are small and explained mainly by widows’ pensions, children’s allowances, and disability benefits from occupational pension plans. In the case of public health expenditures, the relatively high support of the first quintile reflects the higher utilization of health care services by the elderly. The other social security transfers also tend to benefit the lowest two quintiles, but reach the middle- and high-income groups as well. This finding reflects compensations for temporary disruptions in the ability to work, such as illness, unemployment, or maternity. Finally, expenditures to service the public debt appear in favor of the low-income groups. This finding is in accordance with the incidence assumption that groups who receive most from the state are also those who contribute most to the spending deficit. The structure of the expenditure incidence changes somewhat if only the households in the working age group are considered (see Table 2.11 in the Appendix). In particular, then, education benefits also are oriented towards the low-income quintiles. In contrast, the distribution of health expenditures evens out, which is not unexpected as the age structure becomes fairly similar across the different income quintiles. In the case of subsidies, the beneficiaries are located in the lower two quintiles as well as in the topmost quintile. The former is caused by the support of farming households (note that 70 percent of total subsidies go to the agricultural sector), whereas the latter can be traced to nonagricultural subsidies which, via price reductions, reduce the costs of consumption. Finally, social benefits are more focused on low-income groups, and social assistance goes almost exclusively to the lowest quintile.
24
Table 2.5: Incidence of public expenditures and revenues by quintile of pre-fisc household income, 2005 Q1
Q2
Q3
Q4
Q5
Mean
Q1
Q2
per adult equivalent, in CHF PRE-FISC INCOME
25
PUBLIC SPENDING Public goods Education Culture Health Transport Subsidies Old-age insurance (AHV) benefits Pension plan benefits (BVG) Disability insurance (IV) benefits Health insurance benefits Other social insurance benefits Social assistance Debt service
Q3
Q4
Q5
Mean
in % of pre-fisc income
7’940
37’880
63’220
87’090
147’930
68’790
100.0
100.0
100.0
100.0
100.0
100.0
88’000 4’320 2’100 600 6’390 2’210 560 24’750 19’940 4’670 3’750 9’930 6’990 3’570
51’960 5’030 7’460 620 3’710 2’280 2’380 5’890 7’600 4’410 3’390 6’080 1’780 2’110
32’160 5’490 6’570 740 2’920 2’810 950 900 1’440 1’890 3’190 3’880 310 1’300
29’610 5’970 5’510 980 2’740 3’060 420 710 1’420 1’170 3’260 2’870 460 1’200
30’690 7’860 3’310 1’370 2’930 4’370 970 1’140 1’090 540 3’300 2’530 150 1’250
46’490 5’730 4’990 860 3’740 2’950 1’050 6’680 6’300 2’540 3’380 5’060 1’940 1’890
1108.3 54.4 26.4 7.6 80.4 27.9 7.0 311.7 251.1 58.8 47.3 125.0 88.0 45.0
137.2 13.3 19.7 1.6 9.8 6.0 6.3 15.6 20.1 11.6 9.0 16.0 4.7 5.6
50.9 8.7 10.4 1.2 4.6 4.5 1.5 1.4 2.3 3.0 5.0 6.1 0.5 2.1
34.0 6.9 6.3 1.1 3.1 3.5 0.5 0.8 1.6 1.3 3.7 3.3 0.5 1.4
20.7 5.3 2.2 0.9 2.0 3.0 0.7 0.8 0.7 0.4 2.2 1.7 0.1 0.8
67.6 8.3 7.3 1.3 5.4 4.3 1.5 9.7 9.2 3.7 4.9 7.4 2.8 2.7
in % of pre-fisc income + benefits (real and monetary) PUBLIC REVENUES Income and property taxes Earnings and capital taxes Excise duties Other taxes Remunerations Public investments Contributions to AHV and IV Contributions to pension plans Contributions to health insurance Contributions to other social insurance Capital income of social insurance
24’720 6’460 2’010 4’020 2’290 3’680 1’930 810 70 3’170 130 150
31’660 5’570 2’370 3’980 2’340 4’280 1’940 3’050 2’510 2’820 1’470 1’330
41’670 6’110 1’700 4’660 2’110 4’670 2’010 5’290 6’170 2’870 3’010 3’070
55’600 9’280 2’040 5’440 2’560 5’090 2’550 7’030 9’710 2’980 4’180 4’750
93’740 20’710 4’710 7’250 4’460 6’690 4’570 11’020 17’300 3’050 5’680 8’290
49’460 9’620 2’560 5’070 2’750 4’880 2’600 5’440 7’150 2’980 2’890 3’520
25.8 6.7 2.1 4.2 2.4 3.8 2.0 0.8 0.1 3.3 0.1 0.2
35.2 6.2 2.6 4.4 2.6 4.8 2.2 3.4 2.8 3.1 1.6 1.5
43.7 6.4 1.8 4.9 2.2 4.9 2.1 5.5 6.5 3.0 3.2 3.2
47.6 8.0 1.7 4.7 2.2 4.4 2.2 6.0 8.3 2.6 3.6 4.1
52.5 11.6 2.6 4.1 2.5 3.7 2.6 6.2 9.7 1.7 3.2 4.6
42.9 8.3 2.2 4.4 2.4 4.2 2.3 4.7 6.2 2.6 2.5 3.1
Note: The allocation of public expenditures and revenues to households is based on the incidence assumptions of the standard scenario defined in Section 2.2.3 and Table 2.3. Q1 to Q5 denote the quintiles of the pre-fisc household income distribution. The expenditure category ‘public goods’ combines the subcategories of general administration, justice, police, national defence, international affairs, regional planning, and environment. ‘Other social insurance’ includes unemployment insurance, family and child allowances, accident insurance and income compensation. ‘Excise duties’ includes value-added taxes, mineral oil and fuel taxes, and taxes on tobacco products. The category ‘other taxes’ includes withholding taxes, stamp duties, real estate taxes, taxes on motor vehicles, customs duties, royalties and concessions, inheritance and gift taxes, incentive taxes, and other property and excise taxes.
Absolute revenue incidence On the contribution side, payments to the state increase with income (lower left part of Table 2.5). Whereas the representative household of the first quintile with 24’720 francs contributes only half as much as the average household, the household in the fifth quintile with 93’740 pays 1.9 times as much. This inequality in the burden is still more pronounced for social security contributions from which the (often retired) households in the first quintile are practically exempt. Only health insurance premiums are on a similar level of about 3’000 francs in all income categories. The situation is different for taxes. The first quintile does not bear the least burden; instead, it is the second or third quintile. The reason for this situation lies again in the fact that the first quintile contains mainly households whose income is derived primarily from social transfers. Because such income is taxable, the income taxes are relatively high in comparison to the pre-fisc income. Moreover, for the often wealthy pensioner households, property taxes, as well as the earnings and capital tax of the business sector, carry weight, half of which is shifted to the capital owners in the standard scenario. If the sample is limited to the working-age population, the absolute burden in the first quintile decreases for all tax types, but nevertheless falls only slightly below that of the second quintile (see Table 2.11 in the Appendix). Illustration with Lorenz curves Figure 2.3 illustrates the incidence of the main expenditure and revenue categories, again by means of Lorenz curves. Here, these curves show which proportions of the public expenditures (left panel) and revenues (right panel) fall upon different shares of households that are sorted by pre-fisc income. The 45 degree line represents the locus of the even distribution of the benefits and burdens. On the expenditure side, a Lorenz curve lying below the 45 degree line indicates a pro-rich intervention, while one lying above the line signals larger benefits to low-income households. Consistent with the observations made above, the Lorenz curve for transport expenditures, for example, shows that these expenditures intensify the inequalities. Almost 40 percent of the total outlay for transportation infrastructure accrues to the highest income quintile only. On the other side, the Lorenz curve for social assistance confirms its strong concentration on the lowest income households. The 5 percent poorest households receive a full third of these funds. Conversely, the 50 percent well-off households together receive only about 10 percent of this support. Exceptions aside, either curve suggests the high target accuracy of social assistance.
26
100% 80% 60% 40% 0
20%
Cumulative proportion of public revenues
80% 60% 40% 20% 0
27
Cumulative proportion of public expenditures
100%
Figure 2.3: Lorenz curves for public expenditures and revenues, 2005
0
20%
40%
60%
80%
100%
0
20%
Cumulative proportion of households
40%
60%
80%
100%
Cumulative proportion of households
45°
Pre−fisc income
45°
Pre−fisc income
Education
Health
Income/property taxes
Earnings/capital taxes
Transport
Old age ins. (AHV) benefits
Excise duties
Contrib. to AHV/IV
Occup. provision benefits
Invalidity ins. (IV) benefits
Contrib. to occup. provision
Contrib. to health ins.
Health ins. benefits
Social assistance
Note: This figure shows in the left (right) panel which proportion of the public expenditures (revenues) benefit (burden) different shares of the households. The households are ordered according to their pre-fisc income.
On the revenue side, a Lorenz curve lying below the 45 degree line indicates a pro-poor intervention. Moreover, the greater the distance between the two curves, the more the lowincome households are spared from having to contribute. From the right side of Figure 2.3, it can be seen that health insurance premiums are tantamount to a head tax. This scenario is in contrast to contributions to pension schemes which, to a large extent, are made by highincome households. Sixty percent of payments into occupational pension plans comes from the 20 percent richest households. The Lorenz curve for income and property taxes indicates the relatively high contributions of the lowest income quintile and the comparatively modest contributions of the second and third quintiles. Incidence relative to pre-fisc income Having analyzed the incidences in absolute values, the question remains as to the conclusions that can be drawn with regard to the redistributive effects. Redistribution, as mentioned, takes place when the financially weak improve their position vis-a-vis the financially strong. On the expenditure side, redistribution means that poor households must receive more benefits relative to their pre-fisc income. In the case where they receive larger benefits in absolute terms, the expenditure-side intervention is thus unequivocally progressive and redistributive, as poor households also receive more benefits relative to their pre-fisc income. If low-income households receive smaller benefits in absolute terms, a public intervention can still work progressively as long as the poor receive more benefits relative to their pre-fisc income. On the side of public revenues, a contribution is progressive if poor households pay less than rich ones, not only in absolute terms but also in relation to pre-fisc income. If all households pay the same absolute amount, the contribution has a regressive effect, because poor households are charged more than rich households relative to the pre-fisc situation. The upper right part of Table 2.5 shows the public expenditures per household in terms of percentage of the pre-fisc income for each quintile. The table shows that the lowest income quintile receives public benefits that, on average, multiply the pre-fisc income by a factor of eleven. Meanwhile, the topmost quintile at the other end of the spectrum increases its income by just 20.7 percent. As indicated by the detailed figures, the social benefits work progressively, and, to a lesser extent, the rest of the public spending does, too. Although welloff households in part receive more in francs, they benefit less than low-income households in terms of the proportion of their pre-fisc income. For example, cultural subsidies, where the absolute values are the most widespread, household income increases by 7.6 percent in the first quintile, but only by 0.9 percent in the fifth quintile. In the lower right part of Table 2.5, household contribution burdens are related to the household’s economic strength. Note that, deviating from the public expenditure side, pre28
fisc income plus real and monetary benefits received from the state are used as the reference base. In so doing, it is taken into account that the tax requirements also depend on the economic capacity after the receipt of the positive transfers, and that the progressivity of a levy is usually judged by that measure. Moreover, this approach prevents tax rates quickly becoming very high for low-income households because the tax contributions are divided by near-zero pre-fisc income. With the broader reference base, the contribution system as a whole shows to be progressive. The first income quintile with an average contribution rate of 25.8 percent pays half as much as the highest quintile. If only taxes are considered, the detailed figures reveal that, at best, income and property taxes have a progressive effect, however. There, the lower three quintiles remit taxes in the amount of 6 to 7 percent of the sum of the pre-fisc income and benefits, whereas the fourth and fifth quintiles pay 8.0 and 11.6 percent, respectively. The other taxes do not show a lower charge for the less well-off households; their effect is rather neutral to regressive. In contrast, the social contributions appear to have a progressive effect, even though sparing the low-income households is less pronounced if the sample is restricted to working households (see Table 2.11 in the Appendix). An exception is health insurance contributions, as they, despite widespread publicly financed premium reductions, work regressively. The findings compiled so far indicate that first, out of all the public interventions, the monetary social benefits have the strongest redistributive effect, because the low-income groups, in absolute and relative terms, benefit most from the state. Second, the real public benefits appear to have the next strongest redistributive effect. Although they do not always favor low-income households in terms of absolute value, they invariably do so if related to the pre-fisc situation. Third might be the income and property taxes as well as the social contributions (apart from health insurance rates), which do not tax low-income households less relative to pre-fisc income, but less in terms of pre-fisc income plus monetary and real benefits received. The other tax types, as well as health insurance premiums, suggest few or regressive effects.
2.3.3
Evolution of the Incidence Over Time
If the incidence for 2005 is compared with that of 1990 and 2000, it is seen that the distribution of the benefits and burdens across the different income groups tends to become more pro-poor over time. Figures 2.4 and 2.5 give an overview and compare the incidence of the main expenditure and revenue aggregates of 2005 with those of 1990 and 2000. The detailed incidence records for 1990 and 2000 are presented in Table 2.12 in the Appendix. 29
Figure 2.4: Incidence of public expenditures, comparison of 1990, 2000, 2005
5000
Social benefits in % of pre−fisc income
5000
Real benefits in % of pre−fisc income
5000
Post−fisc income in % of pre−fisc income
2000
2005
2005
2005
1
2
3
4
5
6
7
Deciles of pre−fisc income
8
9
10
200 3
3
5
10
20
40 60 80100
200 40 60 80100 10
20
40 60 80100 20 10 3
30
200
500
1990
2000
500
1990
2000
500
1990
1
2
3
4
5
6
7
Deciles of pre−fisc income
8
9
10
1
2
3
4
5
6
7
8
9
10
Deciles of pre−fisc income
Note: The y-axis is shown in logarithmic scale. The real benefits include all the expenditure-side positions of the state budget (see Table 2.1), except for the benefits of social insurance and the other social welfare institutions. These two positions are subsumed in the social benefits aggregate.
Figure 2.5: Incidence of public revenues, comparison of 1990, 2000, 2005
30 25 20 15 10 1990
1990
2000
2000
5
5
10
15
20
25
30
35
Social contributions in % of pre−fisc inc. + benefits
35
Taxes in % of pre−fisc inc. + benefits
0
2005
0
2005
1
2
3
4
5
6
7
8
9
10
Deciles of pre−fisc income
1
2
3
4
5
6
7
8
9
10
Deciles of pre−fisc income
Note: The taxes include all revenue-side positions of the state budget (see Table 2.1), except for social security contributions and capital and other revenues of social insurance. These positions are pooled in the social contributions aggregate.
For each decile, Figure 2.4 shows the post-fisc income, the aggregated real public benefits, as well as the aggregated monetary social security and social assistance benefits in percentage of pre-fisc income. The y-axis is shown in logarithmic scale in order to make the differences in the upper income range more visible. For the post-fisc income, the figure shows that, apart from the lowest and highest decile, the income level has risen overall since 1990. As a consequence, in 2005, unlike in the past, the post-fisc income also lies above the pre-fisc income in the fourth decile. In addition, the households in the lower deciles succeeded in increasing their positive net transfer more than the upper deciles were able to do. For instance, the post-fisc income of the second decile was 3.9 times the pre-fisc income (390 percent of the pre-fisc income) in 2000, and 5.4 times the pre-fisc income in 2005. In the ninth decile, that same factor rose only a little, from 0.56 in 2000 to 0.60 in 2005. Developments at the lower end of the income distribution only partly fit this picture. Although outliers make an exact quantification difficult, the results suggest that today the poorest households, all in all, benefit less from the state than they did in 1990. This result implies that the general increase of the progressivity of the public transfers system left out the lowest income groups. Meanwhile, however, as the comparison between the years 2000 and 2005 indicates, public interventions tend to stronger progressivity over the whole income spectrum. 31
Decomposing the post-fisc income into its components reveals that the reason for these developments lies in the spending as well as in the revenue policies. As indicated from the middle panel of Figure 2.4, the real benefits enjoyed today reach the upper half of the income groups to a lesser extent than they did in 1990, whereas the second to fifth deciles have largely won in this respect. The social benefits (right panel) have led to income gains in all but the first and tenth deciles since 1990, yet to a decreasing extent with increasing income. The households in the lowest decile, compared to 1990, have lost in terms of both real and monetary benefits and have improved their situation only in recent times. On the contribution side, it can be noted that in the two lowest deciles, the total direct and indirect tax burden has risen since 1990, while it has dropped by 2 to 5 percentage points in the fourth to ninth deciles. This scenario can be seen in the left panel of Figure 2.5 where the contribution aggregates are related to the sum of the pre-fisc and benefits income. The right panel of Figure 5 illustrates that, today, payments into social security demand a smaller share of income after positive transfers than in the past. Exempt from that relief, however, is again the lowest income decile, which, in 2005, made the same contributions as the second decile.
2.3.4
Significance of the Redistributive Effects
Up to this point, redistribution through public interventions has been determined on an intuitive basis. To achieve reliable results, the redistributive effects must be examined for their statistical significance. As discussed in Sections 2.2.6 and 2.2.7, whether or not public interventions significantly change the inequality measure of the pre-fisc income distribution must be tested. Table 2.6 shows the results for the main public expenditure and revenue aggregates; the detailed results are specified in Table 2.13 in the Appendix. The upper part of Table 2.6 lists, for three different values of α, the inequality measure (Iα ) for pre-fisc income, as well as for the income that result when individual expenditure and revenue categories are added to and deducted from, respectively, the pre-fisc income. The ‘∆’ column shows the divergence from the pre-fisc situation, and the ‘Sign.’ column contains the results of the tests that determine whether or not such divergence is significant. In the lower part of Table 2.6, the contribution categories are set also in relation to the sum of the pre-fisc income and benefits so as to account for the broader tax basis. The results outlined in the upper part of Table 2.6 confirm the high redistributive effect of social benefits (social security and social assistance benefits). The social benefits lower 32
Table 2.6: Transfer-induced change in measured inequality, 2005
PRE-FISC INCOME
Iα=0 ∆ Sign. Iα=1 ∆ Sign. Iα=2 ∆ Sign. 0.765 0.306 0.364
Pre-fisc income + Pre-fisc income + Pre-fisc income − Pre-fisc income − Post-fisc income
0.202 -74% ** 0.109 -86% ** 1.439 88% ** 0.865 13% * 0.107 -86% **
0.185 -40% ** 0.118 -61% ** 0.433 42% ** 0.345 13% ** 0.100 -67% **
0.232 -36% ** 0.164 -55% ** 0.423 16% 0.471 30% ** 0.128 -65% **
PRE-FISC INC. + BENEFITS
0.079
0.088
0.121
Pre-fisc Pre-fisc Pre-fisc Pre-fisc
0.080 2% 0.069 -12% ** 0.086 9% ** 0.081 3%
0.082 -7% 0.076 -14% ** 0.097 10% ** 0.092 4% **
0.101 -16% . 0.096 -21% ** 0.141 16% ** 0.133 10% **
inc. inc. inc. inc.
real benefits social benefits taxes social contributions
+benef. +benef. +benef. +benef.
− − − −
taxes income/property taxes social contributions social contrib. excl. health ins.
Note: This table shows different general entropy measures for the distribution of the main income aggregates. α = 0 reflects the highest sensitivity to inequalities in the low-income households. The ‘∆’ column in the upper part of the table shows the change in the inequality measure compared to the income situation before any transfer. The lower part of the table presents the change in the inequality measure in comparison to the income situation after the positive, but before the negative, transfers. **/*/. in the ‘Sign.’ column represent significant changes on a 0.01/0.05/0.10 level. Definitions of the revenue and expenditure aggregates are the same as for Figures 2.4 and 2.5.
the Iα=0 measure, which reacts particularly to inequalities in the lower parts of the income distribution, by 86 percent. Also, the Iα=1 and Iα=2 inequality measures, which are more sensitive to inequalities in the middle and upper income groups, are clearly and significantly reduced. Somewhat less, but still considerable, are the redistributive effects of the real public benefits. If the inequality measures of the pre-fisc situation are compared with those after taxes or social contributions, a regressive effect of the contribution system arises because of the taxation of the transfer income. If, however, the contribution aggregates are related to the broader base of pre-fisc income and received benefits (lower part of Table 2.6), it follows that the tax system as a whole hardly generates redistributive effects. At most, there is a reduction in inequality among the high-income groups, as the weakly significant decline of Iα=2 signals. A glance at the individual tax types confirms the above-mentioned impression that this redistributive neutrality is a result of opposing effects. Specifically, the significant progressivity of the income and property taxes is absorbed by the inverse redistributive effects of the other taxes, above all the regressive excise taxes. In contrast, the notion that income inequality is reduced because low-income households pay fewer social contributions is proven false. In fact, the opposite is true. Income-indifferent 33
health insurance premiums clearly play a part in this finding; however, when they are excluded, regressive effects still remain in the middle and upper income groups. More important are the comparatively high contributions of the (upper) middle class to pension funds. As shown in Table 2.13 in the Appendix, these households have a greater propensity to save within occupational provision plans than do the highest income households. As a consequence, payments to pension funds (direct and indirect ones in the form of capital gains of the pension funds) lose their leveling effect in the upper part of the income distribution. Conducted also for earlier years, these tests confirm that public interventions have become more progressive over time. Table 2.7 shows the pre- and post-fisc inequality measures for 1990 to 2005 and different values of α. The differences Ipre−f isc − Ipost−f isc are all highly significant. In the case of Iα=1 and Iα=2 , which are sensitive to redistributive effects in the middle and upper income groups, inequality has decreased with few exceptions. In 1990, the post-fisc Iα=1 was 38 percent below the pre-fisc value, whereas the reduction amounted to 67 percent in 2005. On the other hand, the less regular pattern for Iα=0 indicates that, in the low-income range, public interventions were temporarily more progressive in earlier years than in 2005. Remarkably, this situation concerns the years of economic downturn (1990, 2002, 2003) in which the inequality of pre-fisc income was relatively high. Table 2.7: Transfer-induced change in measured inequality, 1990-2005 1990
1998
2000
2001
2002
2003
2004
2005
Ipre-fisc, α=0
1.667
0.781
0.665
0.787
0.906
0.812
0.748
0.765
Ipost-fisc, α=0
0.182
0.172
0.154
0.134
0.101
0.092
0.123
0.107
-89%
-78%
-77%
-83%
-89%
-89%
-84%
-86%
Ipre-fisc, α=1
0.312
0.270
0.310
0.274
0.294
0.282
0.277
0.306
Ipost-fisc, α=1
0.193
0.152
0.156
0.115
0.094
0.087
0.102
0.100
-38%
-44%
-50%
-58%
-68%
-69%
-63%
-67%
Ipre-fisc, α=2
0.396
0.335
0.550
0.319
0.328
0.281
0.292
0.364
Ipost-fisc, α=2
0.324
0.223
0.294
0.151
0.118
0.101
0.125
0.128
-18%
-34%
-47%
-53%
-64%
-64%
-57%
-65%
Note: The changes in the inequality measures are all statistically significant on a 0.01 level.
34
2.3.5
Redistribution across Socioeconomic Groups
The results shown so far have focused on the vertical redistribution between high-income and low-income households. The objective of this section is to give an indication of the horizontal redistribution across socioeconomic groups. The incidence results shown in Table 2.5 already have revealed the redistribution from working to retired households. For the working-age population, the following examines, first, to what extent the state brings about redistribution between childless households and families with children. Second, the following addresses ways that resources are redistributed across households with different work loads. Redistribution across different household types Table 2.8 presents the expenditure and revenue incidence for different types of households. It shows that couples without children (school age or in education) realize the highest pre-fisc income. With 96’070 francs per adult equivalent, this group lies about 20 percent above the average pre-fisc income of nearly 80’000 francs. The second highest income earners are couples with one child as well as single-person households. With an increase in the number of children, the pre-fisc income declines, whereby the income losses are greater with the first and second child than with subsequent children. This finding reflects the frequent withdrawal from work by mothers after the birth of the first child or, at the latest, second child. Consequently, households with more than one child generate below-average pre-fisc income. However, income before transfers are by far the lowest for single-parent households. With 51’930 francs they earn about 35 percent less than average. State transfers smooth the income distribution in favor of families with children. Although childless couples still derive the highest income, their lead over the average household falls to 8 percent. Even greater is the loss for single-person households whose post-fisc income is below average. In contrast, couples with more than one child slightly benefit as their income approaches the average from below. By far the most favored group is single-parent households, however. With a post-fisc income of 65’570 francs, they not only receive an above average income, but are also the only household type that is better off in comparison to its pre-fisc situation. As observed from the right section of Table 2.8, the resource endowment of single-parent households increases by 26 percent. Interventions that lead to redistribution towards households with children can be found in both public expenditures and revenues. On the expenditure side, the provision of free primary education is at the fore and, to a lesser extent, secondary and vocational education. Higher education, in contrast, often benefits childless couples as well. Disability insurance, besides supporting households without children, of which there are many in the advanced
35
Table 2.8: Incidence of public expenditures and revenues by type of household, households in working age group, 2005 1 Adult
1 Adult 1+child
Couple
Couple 1child
Couple 2child.
Couple 2+child.
Mean
1 Adult
1 Adult 1+child
per adult equivalent, in CHF
Couple
Couple 1child
Couple 2child.
Couple 2+child.
Mean
in % of pre-fisc income
36
PRE-FISC INCOME POST-FISC INCOME
83’010 61’660
51’930 65’570
96’070 68’690
83’770 64’670
69’300 59’420
60’520 59’140
79’790 63’650
100.0 74.3
100.0 126.3
100.0 71.5
100.0 77.2
100.0 85.7
100.0 97.7
100.0 79.8
PUBLIC SPENDING Public goods Education, primary Education, secondary Education, tertiary Culture Health Transport Subsidies Old-age ins. (AHV)/ pension plans Disability insurance (IV) benefits Health insurance benefits Other social insurance benefits Social assistance Debt service
33’360 5’250 0 1’040 1’830 930 2’110 3’530 760 3’210 4’700 2’780 4’530 1’850 1’350
46’340 5’370 6’390 4’560 2’380 780 2’710 2’460 750 610 5’990 3’010 5’420 4’650 1’880
37’730 6’340 0 920 1’270 1’000 3’000 3’460 1’080 6’820 3’430 3’650 3’970 1’640 1’530
36’010 6’330 3’700 1’570 1’210 940 2’940 3’050 1’290 1’410 1’420 3’570 5’370 1’900 1’460
35’480 6’200 7’460 1’770 1’090 800 3’020 2’410 1’260 200 1’000 3’260 4’560 1’120 1’440
39’190 6’250 10’300 2’150 570 720 3’110 2’320 2’900 40 810 3’020 4’890 620 1’590
36’870 5’970 3’180 1’500 1’410 890 2’800 3’060 1’220 2’990 3’120 3’270 4’590 1’710 1’500
40.2 6.3 0.0 1.2 2.2 1.1 2.5 4.2 0.9 3.9 5.7 3.3 5.5 2.2 1.6
89.2 10.3 12.3 8.8 4.6 1.5 5.2 4.7 1.5 1.2 11.5 5.8 10.4 9.0 3.6
39.3 6.6 0.0 1.0 1.3 1.0 3.1 3.6 1.1 7.1 3.6 3.8 4.1 1.7 1.6
43.0 7.6 4.4 1.9 1.4 1.1 3.5 3.6 1.5 1.7 1.7 4.3 6.4 2.3 1.7
51.2 8.9 10.8 2.6 1.6 1.1 4.4 3.5 1.8 0.3 1.4 4.7 6.6 1.6 2.1
64.8 10.3 17.0 3.5 0.9 1.2 5.1 3.8 4.8 0.1 1.3 5.0 8.1 1.0 2.6
46.2 7.5 4.0 1.9 1.8 1.1 3.5 3.8 1.5 3.8 3.9 4.1 5.8 2.1 1.9
in % of pre-fisc income + benefits (real & monetary) PUBLIC REVENUES Income and property taxes Earnings and capital taxes Excise duties Other taxes Remunerations Public investments Contributions to AHV and IV Contributions to pension plans Contributions to health insurance Contrib. to other social insurance Capital income of social insurance
54’710 10’610 2’310 6’030 2’790 4’470 2’740 6’690 8’690 2’470 3’670 4’250
32’690 4’730 1’420 4’540 1’810 4’570 1’780 3’430 3’990 2’420 1’990 2’010
65’120 12’950 2’460 6’080 2’910 5’400 3’110 8’140 11’030 3’330 4’330 5’380
55’100 9’870 1’820 5’070 2’330 5’390 2’560 6’970 9’490 3’200 3’760 4’640
45’360 6’910 1’520 4’360 2’010 5’280 2’100 5’830 7’640 2’830 3’130 3’750
40’570 5’880 1’270 3’850 1’670 5’330 1’880 5’230 7’040 2’490 2’490 3’440
53’020 9’620 1’990 5’290 2’450 5’090 2’550 6’600 8’730 2’890 3’550 4’270
47.0 9.1 2.0 5.2 2.4 3.8 2.4 5.7 7.5 2.1 3.2 3.7
33.3 4.8 1.4 4.6 1.8 4.7 1.8 3.5 4.1 2.5 2.0 2.0
48.7 9.7 1.8 4.5 2.2 4.0 2.3 6.1 8.2 2.5 3.2 4.0
46.0 8.2 1.5 4.2 1.9 4.5 2.1 5.8 7.9 2.7 3.1 3.9
43.3 6.6 1.4 4.2 1.9 5.0 2.0 5.6 7.3 2.7 3.0 3.6
40.7 5.9 1.3 3.9 1.7 5.3 1.9 5.2 7.1 2.5 2.5 3.4
Note: The allocation of public expenditures and revenues to households corresponds to the incidence assumptions of the standard scenario defined in Section 2.2.3 and Table 2.3.
45.4 8.2 1.7 4.5 2.1 4.4 2.2 5.7 7.5 2.5 3.0 3.7
working age group, favors single parents. Other forms of social insurance and social assistance strongly target single parents, too. This finding comes to light particularly if the benefits are related to pre-fisc income (right section of Table 2.8). On the revenue side, family households, in terms of absolute value as well as in relation to pre-fisc income, bear the lowest tax burden. The same is true for social security contributions, with the exception of health insurance premiums. Again, single-parent households are afforded the most tax relief. Redistribution across households with different work loads Table 2.9 shows the expenditure and revenue incidence for working-age households with a full-time (>90%), middle (50-90%) and small (90%, w. child’ category shows, this trend is driven partly by farming households. With respect to different work loads in the labor market as well, the analysis shows that public transfers smooth income considerably. The beneficiaries are the households with short or missing working hours. While full-time employed households, irrespective of the presence of children, arrive at a post-fisc income that is up to 7 percent higher than that of households with a middle work load, both categories fall short of the income attained by households with the fewest work hours. Partly and early retired households clearly play a role in this somewhat surprising result, as indicated by the large benefits from pension plans to the ‘90%
50-90%
90% w. child
50-90% w. child
90%
50-90%
per adult equivalent, in CHF PRE-FISC INCOME POST-FISC INCOME
38
PUBLIC SPENDING Public goods Education Culture Health Transport Subsidies Old-age insurance (AHV) benefits Pension plans Disability insurance (IV) benefits Health insurance benefits Other social insurance benefits Social assistance Debt service
90% w. child
50-90% w. child
Chi2
0.000
Log likelihood
* *
*
*
956
LR Chi2 (37 degrees of freedom) Pseudo R2
.
0.104 -547.5
Significance codes: * 0.05 . 0.1 Note: The dependent variable is the binomial indicator for whether a household participated in the NREGS (1 = participation).
102
work. On the other hand, owning land increases the likelihood of NREGS participation, indicating that independent agricultural production leaves people (seasonally) underemployed, which leads to ancillary income activities to reach a higher or sufficient income. Finally, being selected for the SKS program also appears to have a positive effect on program participation, which might be due to the accompanying training sessions that raise awareness about the NREGS. Based on propensity scores, Table 4.5 presents single-difference calculations for the average impact of the NREGS on monthly per capita consumption expenditures and savings behavior. The results indicate that participation in the NREGS increased total consumption by 25 Rs or 6 percent of the pre-intervention income. The breakdown into subcategories shows that the increase is most noticeable for nonfood consumables for which expenditures rose by 40 percent. They are followed by an increase in clothing purchases (+11 percent) and food expenditures (+7 percent). On the other hand, participants appear to have been able to reduce their transportation outlays by 18 percent, which suggests that the NREGS effectively provided nearby work. The changes in the other expenditure categories are not significant on at least a 10 percent level. With regard to savings, the results from the descriptive statistics are confirmed that participants and nonparticipants do not differ in terms of the amount saved but in the probability of saving; 68 percent of the participating households saved money versus 59 percent of the nonparticipants. Figure 4.2 illustrates the consumption-increasing effect of the NREGS by comparing the postintervention cumulative distribution function of total consumption (lower curve) with the pre-intervention equivalent calculated from the estimates (upper curve). The x-axis shows the official poverty line of 356 Rs monthly per capita consumption expenditure (MPCE), as well as a low threshold of 0.75 *MPCE (“extreme poor”) and a high threshold of 2 *MPCE (“marginal”). It follows that, among participants, the NREGS caused the percentage of households living in extreme poverty to fall from 23 percent to 16 percent and lowered the poverty incidence as measured by the official poverty line from 44 percent to 37 percent. For higher consumption levels, the impact, as intended by the NREGS, was smaller. At the line of marginal poverty, the poverty rate fell by only one percentage point from 90 percent to 89 percent.
103
Table 4.5: Average impact of the NREGS on monthly per capita expenditures, propensity score-matched single differences, 2007 sample Participants Nonparticipants Y2007 |D2007 =1 Y2007 |D2007 =0 absolute value
absolute value
Total consumption (in Rs)
448.3
422.9
Food
277.6
260.2
Tobacco, alcohol
14.7
13.0
Nonfood consumables
32.3
23.1
Clothing
29.6
26.8
Energy
12.5
13.2
Transport
10.7
13.1
Health
41.2
44.8
Education
12.6
13.5
Other expenses
17.0
15.5
Savings level (in Rs)
8.8
9.5
Savings probability
0.68
0.59
Single Difference [Y2007 |D2007 =1]-[Y2007 |D2007 =0] absolute value
in % of the average
(standard error)
expendit. of the nonpart.
25.3 * (12) 17.5 . (9.2) 1.7 (1.9) 9.2 *** (2.1) 2.8 * (1.5) -0.7 (1) -2.3 * (1.1) -3.5 (4.2) -0.9 (3.8) 1.5 (1.4) -0.8 (1.04) 0.09 * (0.04)
6.0% 6.7% 13.1% 39.8% 10.5% -5.3% -17.6% -7.8% -6.7% 9.7% -7.9% 15.3%
Note: The estimates are based on kernel-based matching. ***/**/*/. represent significant changes on a 0.001/0.01/0.05/0.10 level based on t-statistics bootstrapped with 500 repetitions. ‘Nonfood consumables’ include expendable articles of daily use, such as toiletries, cleaning agents, or detergents. ‘Other expenses’ include rates and taxes, expenses for personal services (e.g., haircuts and shaves), entertainment, phone calls, household and kitchen equipment.
104
Figure 4.2: Cumulative distribution of monthly total consumption expenditures, NREGS
.7 .6 .5
Pre−intervention consumption
.3
.4
Post−intervention consumption
0
.1
.2
Cumulative proportion
.8
.9
1
participants 2007
0
356 (poor) 356*0.75 (extreme poor)
356*2 (marginal)
Per capita consumption (in Rs)
Panel-data analysis As discussed, matching estimates come with the shortcoming that they correct only for observable differences but not for unobservable heterogeneity across households. For example, households seeking NREGS employment might exhibit better health, higher economic ambitions, or a stronger determination to escape poverty than nonparticipating households, with all of these factors potentially having a positive impact on the outcomes of interest. Therefore, the impact analysis is repeated for the 314 households that were interviewed twice. This second analysis comes with the advantage of breaking down participants and nonparticipants into finer subgroups. Table 4.6 shows the changes over time of the monthly per capita consumption (i.e., Yt=2 − Yt=1 ) for the households that never participated in the NREGS, left the scheme after the first survey (leavers), joined only after the first survey (new joiners), or participated during both surveys (stayers). The breakdown reveals the considerable heterogeneity within the participant and nonparticipant groups of single cross-sections. As indicated by the first four rows of Table 4.6, stayers and leavers, both classified as participants in the 2007 survey, performed very differently over time. Whereas the stayers saw the least growth in total 105
Table 4.6: Average change in the monthly per capita expenditures between the 2007 and 2008 surveys (in Rs, prices May 2007) Nonparticipants, both surveys (n=119)
Food
Nonfood Clothing Energy Transport Health Education Other Total consumables expenses Expenses 55.2 25.3 -3.8 16.2 17.4 -16.4 -0.9 12.6 105.6 (18.4) (4.8) (2.6) (3.2) (4.2) (9.6) (5.9) (3.5) (28.9) 53.7 (23.7)
38 (9.6)
-5.6 (4.7)
15.6 (3.8)
44.9 (9.3)
-10.7 (9.7)
-22.4 (11.3)
10.3 (3.7)
123.9 (43.2)
Participants, only 2nd survey, new joiners (n=80)
43 (16.1)
19.1 (6.2)
-2.2 (3.3)
13.8 (3)
22.3 (6.5)
-22.9 (5.6)
-3.6 (3.4)
6.6 (2.3)
76.1 (24.8)
Participants, both surveys, stayers (n=66)
5.6 (18.3)
7.3 (5.5)
-5.1 (3.4)
16.4 (4.6)
31.3 (8.6)
-25.6 (10.2)
6.3 (5.4)
11.1 (2.7)
47.4 (31.6)
Involuntary nonparticipants who failed to get work (n=48)
-6.2 (26.8)
15.8 (6.7)
-3.7 (5)
20.4 (5.3)
20 (8.3)
-0.1 (21.4)
5.4 (9.4)
9.5 (3.2)
61.1 (51.9)
Voluntary nonparticipants with enough other work (n=43)
86.2 (29.2)
30.7 (9.4)
-1.1 (3.4)
13.6 (5.9)
15 (4.9)
-31.1 (9.9)
-9.3 (12.6)
13.7 (8.4)
117.8 (41.9)
Involuntary leavers who failed to get work (n=20)
25.4 (26.7)
13.1 (14.2)
-10.8 (7.9)
15.5 (4.9)
23.6 (5.7)
-18.1 (15.2)
-11.6 (10.1)
10.2 (6)
47.3 (43.4)
Voluntary leavers with enough other work (n=19)
106.2 (49.3)
53.4 (16.1)
-4.8 (8.3)
23.9 (7.7)
66.3 (19.9)
-23.3 (16.4)
-42.9 (26.9)
14.6 (5.6)
193.5 (94.5)
106
Participants, only 1st survey, leavers (n=49)
Note: Numbers in parentheses are standard deviations. ‘Nonfood consumables’ include expendable articles of daily use, such as toiletries, cleaning agents, or detergents. ‘Other expenses’ include rates and taxes, expenses for personal services (e.g., haircuts and shaves), entertainment, phone calls, household and kitchen equipment.
expenditures, which is less than half of the growth in total expenditures exhibited by the never-participating households, the leavers fared best, surpassing the nonparticipants’ performance by 17 percent. The non-consideration of longitudinal information explains the somewhat startling reversal of the higher expenditures by participants in the first survey to those of nonparticipants in the second survey, which is observed in the descriptive statistics outlined in Table 4.3. On the one hand, the 2008 nonparticipants outperformed the participants because they include program leavers with above-average consumption growth. On the other hand, the 2008 participants include permanent NREGS subscribers who form the weakest part of the population, unable to quit the program. The second survey includes information about reasons for nonparticipation. Based on these data, rows four to eight of Table 4.6 further subdivide the nonparticipants and leavers into involuntary (because of non-provision of a NREGS job) and voluntary (because of other work opportunities) groups. As expected, the households with no interest in the NREGS increased their total expenditures more than the average nonparticipant, whereas those households that involuntarily stayed outside the program did better than only the long-term participants and the involuntary leavers. In contrast, the voluntary leavers, i.e., those who left the NREGS as soon as better work opportunities arose, appear to have performed by far the best. Double differences are employed to estimate the program impact corrected by possible unobservable selection biases. Accordingly, Table 4.7 presents the differences in expenditure changes over time for selected pairs of participant and nonparticipant categories. The focus is first on the category of food expenditures, as all households reported regular outlays in this category, so that this category contains the most nonzero observations. The upper part of Table 4.7 contains unconditional double differences, and the lower part controls for observable household characteristics. If the households that have ever participated in the NREGS are taken together, they show to have fared worse than the never-participants. Their increase in the monthly per capita per food expenditures between the 2007 and 2008 surveys lagged 40 Rs on average (lower part of Table 4.7), which amounts to 8 percent of the mean consumption level in the second survey. However, a comparison of the more similar groups of new joiners and stayers suggests that entry into the program had a significant positive first-stage effect. The double difference between the new joiners and involuntary nonparticipants indicates a positive program impact as well, even though the weak statistical significance (upper part of Table 4.7) disappears when household controls are included in the model. (These two groups, however, can be expected to be similar even if not controlled for observable differences.)
107
Table 4.7: Impact on monthly per capita food expenses, double differences (In Rs) Without control variables (1) [∆Y |D08 =1 or D07 =1] −[∆Y |D08 =D07 =0] (2) [∆Y |D08 =1 & D07 =0] −[∆Y |D08 =D07 =1]
DD1 “part. in 2007 or 2008 vs. nonparticipants” “new joiners vs. stayers”
(3)
[∆Y |D08 =1 & D07 =0] −[∆Y |D08 =D07 =0]
“new joiners vs. nonparticipants”
(4)
[∆Y |D08 =1 & D07 =0] −[∆Y |D08 =D07 =0 & requested work in 2008] [∆Y |D08 =0 & D07 =1] −[∆Y |D08 =D07 =1]
“new joiners vs. involuntary nonpart.”
(6)
[∆Y |D08 =0 & D07 =1 & requested work in 2008] −[∆Y |D08 =D07 =1]
“involuntary leavers vs. stayers”
(7)
[∆Y |D08 =0 & D07 =1] −[∆Y |D08 =D07 =0]
“leavers vs. nonparticipants”
(8)
[∆Y |D08 =0 & D07 =1 & no interest in 2008] −[∆Y |D08 =D07 =0 & no interest in 2008]
“voluntary leavers vs. voluntary nonpart.”
(5)
With control variables (1’) [∆Y |D08 =1 or D07 =1] −[∆Y |D08 =D07 =0] (2’) [∆Y |D08 =1 & D07 =0] −[∆Y |D08 =D07 =1]
DD3
DD4
“part. in 2007 or 2008 vs. nonparticipants” “new joiners vs. stayers”
[∆Y |D08 =1 & D07 =0] −[∆Y |D08 =D07 =0]
“new joiners vs. nonparticipants”
(4’)
[∆Y |D08 =1 & D07 =0] −[∆Y |D08 =D07 =0 & requested work in 2008] [∆Y |D08 =0 & D07 =1] −[∆Y |D08 =D07 =1]
“new joiners vs. involuntary nonpart.”
(6’)
[∆Y |D08 =0 & D07 =1 & requested work in 2008] −[∆Y |D08 =D07 =1]
“involuntary leavers vs. stayers”
(7’)
[∆Y |D08 =0 & D07 =1] −[∆Y |D08 =D07 =0]
“leavers vs. nonparticipants”
(8’)
[∆Y |D08 =0 & D07 =1 & no interest in 2008] −[∆Y |D08 =D07 =0 & no interest in 2008] Number of observations F-statistic p-value R2
“voluntary leavers vs. voluntary nonpart.”
DD5
DD6
DD7
DD8
-22.2 (20) 37.4 (24.3) -12.3 (26) 49.1 . (29.3)
“leavers vs. stayers”
(3’)
(5’)
DD2
48.1 . (29.5) 19.9 (36.4)
-1.5 (32.4) 20.0 (54.8)
-40.1 . (21.3) 45.1 . (26) -24.8 (27.4) 14.6 (30.8)
“leavers vs. stayers”
71.9 * (31.5) 31.8 (39.8)
-23.2 (35.6) 81.5 (69.3)
314 3.132 0.000 0.064
146 2.023 0.023 0.091
199 2.721 0.002 0.087
128 3.905 0.000 0.182
115 2.624 0.003 0.144
86 1.955 0.038 0.150
168 2.147 0.014 0.083
62 1.303 0.245 0.150
Note: ∆Y = Yt=2008 − Yt=2007 ; Dt denotes the NREGS participation in t. Numbers in parentheses are standard errors. */. represent significant changes on a 0.05/0.10 level. Household controls are: place of residence (mandal ), household size, occupation, and landownership. More extensive models come at the cost of further losses of observations.
108
With respect to program leavers, the conditional double differences back up the better performance of the program leavers compared to the permanent participants. The significance disappears, though, if only the involuntary leavers are taken into consideration. No significant difference appears between leavers and nonparticipants. Although the voluntary leavers increased their food consumption by 82 Rs more than the voluntary participants, their lead in this regard is not statistically significant. Table 4.8 consolidates the impact estimates and presents the triple differences for the new joiners versus the stayers, as well as for the new joiners versus the involuntary nonparticipants. By first comparing each of these groups to an observationally similar (i.e., propensity score-matched) group of households that were never part of the NREGS, possible changes in the expenditure levels that occurred independent of the NREGS can be controlled for. For example, because new joiners might include those who voluntarily refrained from the scheme during the first survey, it is possible that they had better economic prospects on average than those who participated in both surveys. If this situation is the case, the double difference of new joiners versus stayers overestimates the impact of the program.4 Calculating the triple differences between the new joiners and the stayers leads to a downward correction of the double differences presented in Table 4.7 and results in a 35.4 Rs gain in food expenditures per capita and month, which corresponds to 15 percent of the new joiners’ average food expenditures during the first survey. The same process applied to the comparison between new joiners and involuntary nonparticipants boosts the estimates of the program impact to a highly significant increase in food expenditures of 96.6 Rs. This figure is certainly exaggerated due to the fact that the nonparticipants that remain, in addition to the involuntary nonparticipants, do not represent enough comparable households. The small number of matched observations make the calculations very unstable. In this case, the (unconditional) double differences produce more reliable results than the triple differences. Table 4.9 presents the triple-difference estimators based on the new joiners/stayers comparison for all expenditure categories. It shows that the NREGS not only had a significant impact on food expenditures, but also substantially increased (though on a low level) spending for nonfood consumables (+48.8 percent) and clothing (+43.9 percent), as well as virtually eliminated the outlays for transportation (-99.2 percent). These results fit well with the propensity score matching outcomes of the 2007 data, although they are more pronounced. This accentuation is not unexpected, however, as the triple-difference estimator, by comparing new joiners with stayers, focuses more on the weak part of the population, 4
In fact, as seen from the expenditure changes over time, presented in the third row of Table 4.8, the
matched nonparticipants for the new joiners performed slightly better (66.5 Rs) than the matched nonparticipants for the program stayers (64.5 Rs).
109
Table 4.8: Average impact of the NREGS on monthly per capita food expenses, triple-difference estimates New
Matched
joiners
nonparticipants
Y t |(D08 =1, D07 =0)
Y t |(D08 =D07 =0)
Y t |(D08 =D07 =1)
Y t |(D08 =D07 =0)
239.9
240.5
261.9
229.5
288.8
250.2
(12.4)
(14.4)
(14.8)
(13.9)
(25.5)
(16.1)
282.9
307.0
267.5
294.0
282.7
364.2
(15.2)
(17.3)
(15.8)
(17.3)
(25.0)
(25.6)
T
t = 2007 t = 2008
C
Stayers
T
Matched
Involuntary
Matched
nonparticipants
nonparticipants
nonparticipants
C
NT ∗
Yt
|(D08 =D07 =0)
C
Y t |(D08 =D07 =0)
Single Differences
43 **
66.5 ***
5.6
64.5 ***
-6.2
114.0 ***
SD = Yt=2 − Yt=1
(16.1)
(18.3)
(18.3)
(18.2)
(26.8)
(27.1)
Double Differences
-23.5
-58.9 ***
-120.1 ***
(17.4)
(18.2)
(27.0)
DD =
SDT
−
SDC
110
Triple Difference I, new joiners
35.4 *
vs. stayers
(17.8)
DDDI = DD|(D08 =1,D07 =0)− DD|(D08 =D07 =1)
Triple Difference II, new joiners
96.6 ***
vs. involuntary nonparticipants
(21.4)
DDDII = DD|(D08 =1,D07 =0)− DD|(D08 =D07 =0, involuntary)
Note: To obtain the triple difference between new participants and stayers, the difference between the changes over time (or the single differences) of new participants and nonparticipants as well as stayers and nonparticipants must first be calculated. In a second step, the difference of these resulting (double) differences is determined. The calculation of the triple difference between new participants and involuntary nonparticipants is carried out accordingly. Values in parentheses are standard errors. ***/**/*/. represent significant differences on a 0.001/0.01/0.05/0.10 level. The nonparticipants are propensity score-matched based on their place of residence (mandal ), household size, occupation, and landownership. Involuntary nonparticipants (N T ∗ ) are households that reported in the second survey that they requested NREGS work but were not provided it.
which was not able to find a better earning opportunity outside the NREGS by the time of the second survey. In contrast, the cross-sectional analysis results combine the program’s impact on the more permanent NREGS participants with its impact on the more temporary NREGS participants, who tend to benefit (relatively) less from the scheme. Interestingly, the triple-difference results also suggest a reduction in spending for children’s education as a result of participation in the NREGS, suggesting that school-age household members must take over tasks in and around the house that were previously done by the adults who became engaged in the NREGS. On the other hand, the small but unique increase in educational expenditures between the first and second surveys, as shown by the long-term participants in Table 4.6, could also imply that educational funding efforts might recover in the long run. Unfortunately, the insufficient sample size, which results when keeping only the households with children, prohibits testing for these hypotheses. Table 4.9: Average impact of the NREGS on monthly per capita expenditure subcategories, triple-difference estimates DDDI , new joiners vs. stayers in Rs
in % of the average 2007
(standard error)
expenditures of the new joiners
Total Consumption
16.5
4.4%
(exclusive alcohol & tobacco)
(29)
Food
35.4 *
14.8%
(17.8) Nonfood consumables
11.4 *
48.8%
(5.4) Clothing
11.2 ***
43.9%
(2.9) Energy
-0.4
-3.8%
(3.6) Transport
-11.8 *
-99.2%
(5.5) Health
-11.9
-28.9%
(9.5) Education
-8.7 .
-62.8%
(4.7) Other expenses
-8.7 *
-67.4%
(3.3) Note: The triple differences for the different expenditure categories are calculated as illustrated in Table 4.8 for the category of food. ***/**/*/. represent significant changes on a 0.001/0.01/0.05/0.10 level. ‘Nonfood consumables’ include expendable articles of daily use, such as toiletries, cleaning agents, or detergents. ‘Other expenses’ include taxes, expenses for personal services, entertainment, phone calls, household and kitchen equipment.
111
In addition to consumption expenditures, the savings outcomes were recalculated on the basis of the panel data. However, as the probability of savings was close to 100 percent for all twice-interviewed households due to their SKS program participation, the analysis was restricted to the level of savings. Consistent with the findings of the cross-sectional analysis, no significant impact of the NREGS was found. To summarize, both the cross-sectional and panel data analyses disclose a positive impact of the NREGS on relevant household expenditures. Most important is the significant increase in food expenditures. As the main item in the household budget, this category gains by far the most in absolute terms, which indicates that NREGS payments first of all improve essential living conditions, i.e., in terms of food security, before flowing into alternative uses. This fact is all the more clear because the dominance of the food expenditure category becomes even more pronounced when the potentially less poor, temporary NREGS participants are not taken into account. In addition, the positive impact of the NREGS on the propensity to save, which is evident from the cross-sectional analysis, further underscores the economical use of the earnings from the program. Besides these findings, the longitudinal analysis provides suggestive evidence that the relatively better-off population benefits from the NREGS, too. Although the comparison is statistically insignificant (due to the small sample size), those respondents who voluntary left the program after the first survey recorded higher gains over time in all expenditure categories than those who voluntarily stayed outside the NREGS during the entire observation time (Table 4.6). This result indicates that bridging income shortfalls by moving quickly in and out of the NREGS, and thereby using the program as unemployment insurance, helps households to increase their income.
4.4.2
Are there Negative Spillover Effects on the Labor Market?
The positive impact of the NREGS on the economic situation of its participants does not supersede the question as to the extent that this poverty relief comes at the cost of interfering in the regular labor market, thereby crowding out private employment. The results do not support this concern, however. The low consumption level and slow economic betterment of long-term participants suggests that the NREGS does not so much improve their economic situation as it allows them to catch up with the households that are not dependent on public work. Together with the considerable fluctuations within the NREGS workforce – 17 percent of the first-round participants had voluntarily left the scheme by the time of the second survey – these arguments contradict the notion that the program attracts people who can find regular jobs. 112
When asked about the merits of the NREGS, more than two-thirds of the participants first mentioned the provision of a subsidiary income source when no other work was available (see Table 4.10). This response strengthens the impression of the emergency and temporary character of NREGS participation and indicates that the program acts primarily as basic income insurance. Moreover, as shown by the answers of another 12 percent of the participants, the NREGS appears to fight un- and underemployment by offering previously unengaged household members an earning opportunity, so that households can generate additional income. Table 4.10: Attitudes towards the NREGS, 2008 survey Share of participants who consider the NREGS a useful program What are the merits of the NREGS? Income substitute when no other work is available Increased income as more household members can work Provision of nearby jobs Improvement of the local infrastructure Low search efforts to get a job Changes that would make the NREGS more useful? Provision of more work days (not restricted to 100 days) Faster (daily) payment of wages Higher wages Prevention of irregularities and fraud Higher reliability of providing work (within the entitlement period of 100 days)
96.3%
67.4% 12.1% 8.3% 6.8% 5.3%
37.3% 35.8% 11.9% 9.0% 5.2%
What would have been the alternative time use and annual household income if the NREGS did not exist? Nonagricultural employment Agricultural self-employment, tending own livestock Agricultural employment, tending other’s livestock Other
78.0% 15.2% 5.3% 1.5%
What are the reasons for not participating? Requested work, but not provided it Have enough other work Health reasons; the work offered is too hard Payment of the salary is not reliable Salary is too low
42.6% 36.4% 17.9% 1.9% 1.2%
113
(mean income) 7514 Rs 7857 Rs 4847 Rs 7000 Rs
In addition, the restriction of 100 work days per annum per household prevents households from living solely on NREGS earnings. The 37 percent of participants who wanted more days in the NREGS and the 12 percent who would have appreciated higher wages indicate that the limitation effectively ensures that the participants are forced to look for other income sources. Furthermore, the high share of respondents who requested faster processing of their wages, as well as the proper and reliable implementation of the NREGS, underlines the importance of the scheme as a temporary last resort rather than a long-term solution. The only indication of a labor drain from the private sector to the NREGS that arises from these data comes from the participants’ reports of alternative time use. According to that indicator, almost 80 percent of the NREGS labor force would have worked in nonagricultural informal employment if the scheme had not existed. The earnings would have totalled 7500 Rs for 100 days, which is less than the 8000 Rs the NREGS stipulated in Andhra Pradesh in 2008 (80 Rs daily minimum wage). However, even this indication is weak as it is questionable whether these laborers would have found a regular job during the period of NREGS participation – or, conversely, whether they had actually refused a regular position in exchange for a NREGS job of limited duration. The propensity score estimates presented in Table 4.4, which show that the probability of participation in the NREGS is significantly lower for households with nonagricultural employment, support these doubts.
4.4.3
Impact on Health
In an attempt to capture the impact of the NREGS beyond its effect on the economic situation of its participants and to allow for a broader concept of welfare, this chapter concludes with an analysis of the scheme’s effect on different health outcomes. Table 4.11 shows descriptive statistics for selected indicators of the physical and mental health of NREGS participants and nonparticipants. In the first survey, 27 percent of the participating households self-reported an improvement of their health status in the year preceding the survey, compared to 21 percent of the nonparticipating households. Also, participants less often reported a deterioration of their health than nonparticipants. A summarizing indicator of the ability to carry out daily physical activities, such as dressing, climbing a small hill, walking for 5 kilometers, bending, squatting or kneeling, also suggests that participants were in better health than nonparticipants. There seems to be no difference, however, between the health of children of participants and nonparticipants as measured by their height and weight during the second survey.
114
Table 4.11: Descriptive statistics of health indicators May-July 2007 survey
Nov.-Dec. 2008 survey
Participants Nonpart. Participants Nonpart. (n=359=33.7%) ( n=707) (n=146=46.5%) (n=168 ) mean (st.dev)/ mean (st.dev)/ mean (st.dev)/ mean (st.dev)/ frequency frequency frequency frequency Physical health % of households with improving health in previous year % of households with worsening health in previous year % of households with health-related work inabilities Indicator of physical health problems, scale from 0 (best) - 42 (worst) Average height of children < 14 years (in cm) Average weight of children < 14 years (in kg) Mental health % of households whose respondent felt worried/tense/ anxious for at least one month in previous year Indicator of mental health problems, scale from 0 (best) - 15 (worst)
27.0% 24.8% 37.6%
21.4% 30.1% 40.5%
-
-
2.7 (4.2) -
3.3 (5.1) -
127.6 (20.4) 24.4 (8.8)
127.4 (23.5) 24.8 (9)
22.0%
33.8%
-
-
3.9 (5.2)
4.0 (5.0)
-
-
Note: The indicator of physical health problems summarizes the level of difficulty (not difficult/ difficult but able to do without help/ able to do only with help/ not able to do) to perform 14 daily physical activities, such as dressing, bathing, climbing a small hill, or drawing water from a well. A value of 0 means that a household, on average, has no problem with any of these activities, while 42 indicates that its members are not able to carry out any of these activities. Similarly, the indicator of mental health problems summarizes the frequency (never/ hardly ever/ sometimes/ always) of feelings of sadness, bouts of weeping, loss of appetite, lack of work motivation, and sleeplessness. The scale ranges from 0 to 15, with 0 indicating that a household does not experience any of these problems, and 15 indicating that a household experiences all of these problems.
In terms of mental health, participants appear to have been in better condition than nonparticipants, as they reported feeling worried, tense, or anxious less often than nonparticipants. On the other hand, there is no visible difference in the indicator that summarizes the frequency of depressive feelings, such as sadness, lack of appetite, or sleeplessness. Table 4.12 verifies whether the differences in the health outcomes of the 2007 survey persist if propensity score-matched groups of participants and nonparticipants are compared with each other. It can be seen that the differences in the physical health indicators are not significant. In contrast, the better mental condition of the participants compared to nonparticipants, as measured by their probability to be worried, tense, or anxious for long periods in a year, is highly significant. Whereas 32 percent of the nonparticipants display such problems, the percentage is 23 percent among the NREGS participants, which is a reduction of one-fourth. This finding strongly suggests that the NREGS succeeds in providing a sense of security from which the participating households benefit, irrespective of whether they are temporarily or repeatedly enrolled in the scheme.
115
Table 4.12: Average impact of the NREGS on health outcomes, propensity score-matched single differences, 2007 sample Participants Nonparticipants Y2007 |D2007 =1 Y2007 |D2007 =0
Single Difference [Y2007 |D2007 =1]-[Y2007 |D2007 =0]
absolute value
absolute value
0.263
0.212
0.051 (0.037)
24.1%
2.61
2.49
0.12 (0.31)
4.9%
0.233
0.310
-0.08 * (0.03)
-24.5%
3.98
3.78
0.203 (0.378)
5.4%
Physical health Probability that health status improved in previous year Indicator of physical health problems, scale from 0 (best) - 42 (worst) Mental health Probability that household respondent felt worried/ tense/anxious for at least 1 month in previous year Indicator of mental health problems, scale from 0 (best) - 42 (worst)
absolute value in % of the average (standard error) value of the nonpart.
Note: For the definitions of the summarizing physical and mental health indicators, see the notes for Table 4.11. The estimates are based on kernel-based matching. */. represent significant changes on a 0.05/0.10 level based on t-statistics bootstrapped with 500 repetitions.
4.5
Summary and Conclusions
This Chapter 4 examines the welfare impact of the Indian workfare initiative NREGS as measured by changes in expenditure levels and physical and mental health indicators. Using two consecutive household surveys conducted in the years 2007 and 2008 in the Medak district of Andhra Pradesh, propensity score matching as well as difference-in-difference methods are applied to compare the outcomes of interest of participants and nonparticipants adjusted for selection bias due to observable and unobservable heterogeneity. The results suggest that the NREGS has a significant impact on alleviating rural poverty and confirm the general statement from 96 percent of the participants that it is a useful program (see Table 4.10). First of all, the NREGS appears to have a substantial effect on the food security of the participating households. Food expenditures, which generally account for about 60 percent of total consumption, increased by 15 percent for the poorest members of the population. The effect is about half as much if the households that participate only temporarily and are able to leave the NREGS after some time are not separated from the economically weak households that repeatedly resort to the program. This finding indicates that the effect of the NREGS increases with the vulnerability of a household and confirms the targeting capability of the program. 116
Besides its impact on food security, the NREGS increases the probability of saving money as well as spending money in other categories. Spending for nonfood consumables and clothing increased by an impressive 40 percent to 50 percent among the less well-off participants. At first sight, these expenditures might appear to be a wasteful use of funds. However, it should be borne in mind that the starting level of these expenditure categories is extremely low, so that in absolute numbers the increase is often no more than 10 Rs (0.25 USD) per capita and month. Finally, thanks to the decentralized implementation of the NREGS, which helps provide work opportunities that are close by its participants, the transport outlays of the participants decrease, and among the poorest households, they even disappear. One of the main reservations concerning the NREGS is that the scheme is too generous in terms of wages and therefore attracts too many households, thereby crowding out private employment and boosting program costs or leading to forced wage hikes, which in turn give rise to inflation or reduced private labor demand. Although these hypotheses are not directly tested in this study, the results seem to contradict their validity. Most important, the high proportion of NREGS leavers, who have a higher expenditure level than their continuously participating counterparts, suggests that the NREGS is not able to keep households with basically intact chances on the labor market. For these participants, the NREGS seems rather to act as unemployment insurance, the prompt utilization of which, nota bene, is likely to prove beneficial compared to trying to cope with a jobless period independently. This situation is seen in the household spending of the temporary participants, which tends to be higher than the expenditures of the never-participating households. Finally, besides the improvement in the financial situation of participants, the securityproviding role of the NREGS results in a significant decrease of emotional distress in the form of anxiety, tension, and worries. Although this outcome is certainly a benefit in itself, experience from other workfare schemes shows that it is not unlikely that such higher security promotes the forward planning of the participants and their ability to pursue higherrisk/higher-return strategies, such as investing in productivity-enhancing equipment or children’s education. Hence, in the long run, the NREGS could provide not only instant poverty relief, but also induce multiplier effects that foster the self-sustained development of the actual and potential beneficiaries.
117
118
Chapter 5 Concluding Remarks
119
Given the generally great and growing importance of the state intervention, the public impact on the welfare level of individual households is a constant concern. Thus, investigation into various levels of abstraction, measuring units, and examination methods are necessary, depending on the specific situation and question. This dissertation presents three examples of welfare distribution and redistribution analysis. In contemporary research, the use of impact analysis of single policy interventions, in the manner of the last contribution on India’s workfare program (Chapter 4), is clearly in the fore. Using experimental or nonexperimental evaluation tools, such analysis bases the impact of an intervention on the differences between certain welfare indicators with and without intervention. The appeal of this type of analysis is undeniable. The variations in welfare status, as measured by the chosen indicators, evolve directly from the data without any further assumptions, are precisely quantifiable, and account for behavioral changes as a result of the intervention. For policy decision-making and the design of efficient interventions, such information is indispensable. Nevertheless, comprehensive and comparative examinations, which are the analysis types of the first two contributions presented in this dissertation (Chapters 2 and 3), are central as well. Sure, strong assumptions are necessary to estimate the combined impact of a wide range of interventions, and the results cannot be expected to be more than approximations of the complex reality. Still, they allow an assessment of ways that various, partly counteracting, interventions affect specific population segments. In welfare states where more than half of the national income is absorbed by the public sector, each household is to some extent a contributor to and beneficiary of the state. A priori statements regarding which and how much redistribution results from that entire system are hardly possible, leaving policy makers and citizens in the dark as to whether the size and scope of the state efforts will meet their expectations. The discipline of economics can contribute to closing this information gap, even if, based on the field’s criteria, the implications are generally ambiguous. On the one hand, redistribution has its advantages, also from an economic point of view. It comes with a wide array of options that enables individuals to better exploit their economic potential. Higher income security allows people to take more risks and invest more in, for example, human capital, which pays off with better economic outcomes in the long run. On the other hand, loss of efficiency is inevitable within redistribution schemes. In comprehensive welfare systems, for instance, individual dedication in the labor market is not fully rewarded, withdrawal is only partly punished, and reintegration is not uncompromisingly demanded. Similarly, people insufficiently weigh the costs and benefits of free medical services or fall short of 120
properly evaluating investments into higher education. Due to such behavioral adjustments, each public intervention is accompanied by a loss of resources, i.e., Okun’s famous “leaking bucket”: More equality comes at the cost of overall welfare losses (Okun, 1975). Apart from the guarantee of a subsistence level, there is no explicit, i.e., unchallenged, solution for this dilemma. It is ultimately left to political and public debate to determine the appropriate degree of public intervention and redistribution. In this political context, the level of welfare, brought about by, e.g., income, leisure time, health, or education, must be determined, i.e., the level of welfare that each individual can count on in different situations of life, independent of economic and social position. Yet, economic insights might still serve as guidance. Welfare differences are not permanent as long as people have opportunities to change their situation. Moreover, the more equally these opportunities are distributed, the less permanently poor households remain at the lowest welfare level. Macpherson (1966) once stated: “The notion of democracy has always contained the notion of equality. Not arithmetical equality of income or wealth, but equality of opportunity to realize one’s human capabilities” (p. 47). Similarly, Nobel laureate Sen, defining poverty as a deprivation of capabilities rather than a deprivation of income, advocates for the promotion of capabilities that allow people to lead a life they value (Sen, 1999). Hence, besides, and often prior to, economic support, the essential engagement of a state lies in the facilitation of “enabling conditions”, such as health, education, political participation, or the awareness of a person’s own opportunities. This perception of welfare promotion has been highly influential in development politics. It has not least paved the way for the wide acceptance of the Human Development Index, which combines measures of life expectancy, literacy, educational attainment, and income per capita to determine well-being, as well as acceptance of the multidimensional United Nations’ millennium development goals. The idea of workfare as help to self-help builds on the same basis. The idea of having limited capabilities as a precondition for individual ill-being (or, conversely, having capabilities as a precondition for well-being) applies in more affluent societies as well. Income security provided by a tight social security net does not prevent the marginalization of parts of the population, in particular as a result of the inability to find employment. From this perspective, the education and qualifications of the future workforce remain principal areas of public intervention. As such, higher education is probably not as important as the promotion of vocational training. The creation of manually demanding yet intellectually less challenging opportunities to learn a trade is crucial not least for providing opportunities to academically weak school leavers. In addition, the integration of those citizens who are currently excluded from the labor market, particularly (single-parent) mothers, 121
groups of foreigners, and older employees, is a further key area of public action. This area includes, for instance, the provision of sufficient child care facilities (not necessarily for free), the promotion and requirement of language skills for low-educated migrant households, the facilitation of old-age provision schemes that allow a stepwise withdrawal from work, or the encouragement of alternative, market-unrelated ways to contribute to society (e.g., by strengthening the prestige of voluntary service). In Europe, Sinn (2006) has initiated the discussion of a stronger focus on supplementing the wages of households that have an insufficient market income instead of substituting their income, which has the effect of sponsoring “idleness”. This approach would help to ensure that individual efforts (again) make a difference in the individual level of welfare achieved. If a state succeeds in furnishing its citizens with the necessary skills to improve their own well-being, Sinn’s call gains persuasiveness. Also, if it is within people’s power to change their lot in the foreseeable future, the actual distribution of welfare becomes somewhat less important. Such an outcome would broaden the range of discussible options to define scope and size of the welfare state.
122
Bibliography Aaron, Henry and Martin McGuire (1970): Public Goods and Income Distribution, Econometrica, Vol. 38, No. 6, pp. 907–920. Afheldt, Horst (2003):
Wirtschaft, die arm macht. Vom Sozialstaat zur gespaltenen
Gesellschaft, Munich: Kunstmann. Aguiar, Mark and Erik Hurst (2007): Measuring Trends in Leisure: The Allocation of Time over Five Decades, The Quarterly Journal of Economics, Vol. 122, No. 3, pp. 969–1006. Aguiar, Mark and Erik Hurst (2008): The Increase in Leisure Inequality, NBER Working Paper, No. 13837. Akitoby, Bernardin, Benedict Clements, Sanjeev Gupta, and Gabriela Inchauste (2006): Public Spending, Voracity, and Wagner’s Law in Developing Countries, European Journal of Political Economy, Vol. 22, pp. 908–924. Alesina, Alberto, Edward Glaeser, and Bruce Sacerdote (2005): Work and Leisure in the U.S. and Europe: Why So Different?, NBER Working Paper, No. 11278. Atkinson, Anthony B., Lee Rainwater, and Timothy M. Smeeding (1995): Income Distribution in OECD Countries : Evidence from the Luxembourg Income Study, OECD Social policy studies, No. 18. Barr, Nicholas (2004): The Economics of the Welfare State, 4th Edition, Oxford: Oxford University Press. Barrett, Christopher B., Mesfin Bezuneh, and Abdillahi Aboud (2001): Income Diversification, Poverty Traps and Policy Shocks in Côte d’Ivoire and Kenya, Food Policy, Vol. 26, No. 4, pp. 367–384. Bell, Linda and Richard B. Freeman (2001): The Incentive for Working Hard: Explaining Hours Worked Differences in the US and Germany, Labour Economics, Vol. 8, No. 2, pp. 181–202. 123
Besley, Timothy and Stephen Coate (1992): Workfare versus Welfare: Incentive Arguments for Work Requirements in Poverty-Alleviation Programs, The American Economic Review, Vol. 82, No. 1, pp. 249–261. Bezuneh, Mesfin, Brady J. Deaton, and George W. Norton (1988): Food Aid Impacts in Rural Kenya, American Journal of Agricultural Economics, Vol. 70, No. 1, pp. 181–191. Bianchi, Suzanne M., Melissa A. Milkie, Liana C. Sayer, and John P. Robinson (2000): Is Anyone Doing the Housework? Trends in the Gender Division of Household Labor, Social Forces, Vol. 79, No. 1, pp. 191–228. Björklund, Anders and Marten Palme (2002): Income Redistribution Within the Life Cycle Versus Between Individuals: Emprical Evidence Using Swedish Panel Data, in: Daniel Cohen, Thomas Piketty, and Gilles Saint-Paul (Eds.), The Economics of Rising Inequalities, pp. 205–224, Oxford: Oxford University Press. Blau, Francine D. and Lawrence M. Kahn (1996): International Differences in Male Wage Inequality: Institutions versus Market Forces, Journal of Political Economy, Vol. 104, No. 4, pp. 791–837. Boadway, Robin and Michael Keen (2000): Redistribution, in: Anthony B. Atkinson and François Bourguignon (Eds.), Handbook of Income Distribution, Vol. 1, Amsterdam: Elsevier. von Braun, Joachim (Ed.) (1995): Employment for Poverty Reduction and Food Security, Washington: International Food Policy Research Institue. van den Broek, Andries, Koen Breedveld, Jos de Haan, Joep de Hart, and Frank Huysmans (2004): Trends in Time. The Use and Organization of Time in the Netherlands, 1975-2000, Den Haag: Social and Cultural Planning Office. Browning, Martin, Angus Deaton, and Margaret Irish (1985): A Profitable Approach to Labor Supply and Commodity Demands over the Life-Cycle, Econometrica, Vol. 53, No. 3, pp. 503–544. Burda, Michael C., Daniel S. Hamermesh, and Philippe Weil (2006): The Distribution of Total Work in the EU and US, IZA Discussion Paper, No. 2270. Centre for Time Use Research (2006): Multinational Time Use Study, Version World 5.5, retrieved from http://www.timeuse.org/mtus. 124
Chakraborty, Pinaki (2007): Implementation of the National Rural Employment Guarantee Act in India: Spatial Dimensions and Fiscal Implications, The Levy Economics Institute Working Paper, No. 505. Chenu, Alain (2003): Les usages du temps en France, Futuribles, Vol. 285, pp. 21–31. Chenu, Alain and Laurent Lesnard (2006): Time Use Surveys: a Review of their Aims, Methods, and Results, European Journal of Sociology, Vol. 47, pp. 335–359. Costa, Dora (2000): The Wage and the Length of the Work Day: From the 1890s to 1991, Journal of Labor Economics, Vol. 18, No. 1, pp. 156–181. Cowell, Frank Alan (1984): The Structure of American Income Inequality, Review of Income and Wealth, Vol. 30, pp. 351–75. Davies, James, France St.-Hilaire, and John Whalley (1984): Some Calculations of Lifetime Tax Incidence, American Economic Review, Vol. 74, pp. 633–649. Deaton, Angus (1985): Panel Data from Time Series of Cross-Sections, Journal of Econometrics, Vol. 30, pp. 109–126. Demery, Lionel (2000): Benefit Incidence: A Practitioner’s Guide, Mimeo, The World Bank Poverty and Social Development Group Africa Region. Demery, Lionel (2003): Benefit Incidence Analysis, in: F. Bourguignon, L. da Silva, and N. Stern (Eds.), Tool Kit: Techniques and Tools for Evaluating the Poverty Impact of Economic Policies, Ch. 2, Washington, D.C.: World Bank. Dev, S. Mahendra (1995): India’s (Maharashtra) Employment Guarantee Scheme: Lessons from Long Experience, in: Joachim von Braun (Ed.), Employment for Poverty Reduction and Food Security, pp. 108–143, Washington: International Food Policy Research Institue. Devereux, Stephen and Colette Solomon (2006): Employment Creation Programmes: The International Experience, ILO Issues in Employment and Poverty Discussion Paper, No. 24, Economic and Labour Market Analysis Department. Drèze, Jean, Reetika Khera, and Siddhartha (2008): Corruption in NREGA: Myths and Reality, The Hindu, January 22, 2008. Drèze, Jean and Amartya Sen (1999): Hunger and Public Action, in: The Amartya Sen and Jean Drèze Omnibus, pp. 258–373, Oxford: Oxford University Press. Duflo, Esther, Rachel Glennerster, and Michael Kremer (2008): Using Randomization in Development Economics Research: A Toolkit, in: T. Paul Schultz and John A. Strauss 125
(Eds.), Handbook of Development Economics, Vol. 4, pp. 3895–3962, Amsterdam: Elsevier. Dyck, Dagmar (2005): Fiscal Redistribution in Canada, 1994-2000, Canadian Tax Journal, Vol. 53, No. 4, pp. 974–1006. Economiesuisse (2007): Wer finanziert den Staat in der Schweiz?, Zurich: economiesuisse. Ecoplan (2004): Verteilung des Wohlstands in der Schweiz, Bericht in Erfüllung des Postulates Fehr vom 9. Mai 2001 (01.3246), Berne: Eidgenössische Steuerverwaltung. FFA, Federal Finance Administration (1990-2005): Ausgaben und Einnahmen von Bund, Kantonen und Gemeinden nach Funktionen, Berne: FFA. Fisher, Kimberley, Muriel Egerton, Jonathan I. Gershuny, and John P. Robinson (2007): Gender Convergence in the American Heritage Time Use Study (AHTUS), Social Indicators Research, Vol. 82, pp. 1–33. Florio, Massimo and Sara Colautti (2005): A Logistic Growth Theory of Public Expenditures: A study of Five Countries Over 100 Years, Public Choice, Vol. 122, pp. 355–393. Freeman, Richard B. and Ronald Schettkat (2005): Marketization of Household Production and the EU-US Gap in Work, Economic Policy, Vol. 20, No. 41, pp. 6–50. FSIO, Federal Social Insurance Office (1990-2005): Swiss Social Insurance Statistics, Berne: FSIO. Fullerton, Don and Gilbert E. Metcalf (2002): Tax Incidence, in: A. J. Auerbach and M. Feldstein (Eds.), Handbook of Public Economics, Vol. 4, pp. 1787–1872, Amsterdam: Elsevier. Gaiha, Raghav (1996): Wages, Participation and Targeting – the Case of the Employment Guarantee Scheme in India, Journal of International Development, Vol. 8, No. 6, pp. 785–803. Gaiha, Raghav (2000): On the Targeting of the Employment Guarantee Scheme in the Indian State of Maharashtra, Economics of Planning, Vol. 33, No. 3, pp. 203–219. Galasso, Emanuela and Martin Ravallion (2004): Social Protection in a Crisis: Argentina’s Plan Jefes y Jefas, World Bank Economic Review, Vol. 18, No. 3, pp. 367–399. Gershuny, Jonathan (2000): Changing Times: Work and Leisure in Postindustrial Society, Oxford: Oxford University Press. 126
Gillespie, W. Irwin (1980): The Redistribution of Income in Canada, Ottawa: Gage Publishing. Goodman-Bacon, Andrew and Leslie McGranahan (2008): How do EITC recipients spend their refunds?, Federal Reserve Bank of Chicago, Economic Perspectives, Vol. 2Q, pp. 17–32. Haddad, Lawrence James and Michelle Adato (2001): How effectively do public works programs transfer benefits to the poor? FCND Discussion Paper No. 108, International Food Policy Research Institute. Hammer, Jeffrey S., Ijaz Nabi, and James A. Cercone. (1995): Distributional Effects of Social Sector Expenditures in Malaysia, 1974-89, in: Dominique van de Walle and Kimberly Nead (Eds.), Public Spending and the Poor – Theory and Evidence, pp. 154–186, Baltimore: Johns Hopkins University Press. Hauser, Martin, Annik Bänziger, and Sabine Stebler (1994): Budgetinzidenz: Wie wird der Mittelstand betroffen? Pilot study, Schweizerische Kader-Organisation SKO. Heckman, James J., Hidehiko Ichimura, and Petra E. Todd (1997): Matching as an Econometric Evaluation Estimator: Evidence from Evaluating a Job Training Program, Review of Economic Studies, Vol. 64, No. 4, pp. 605– 654. Heckman, James J., Hidehiko Ichimura Jeffrey A. Smith, and Petra E. Todd (1998): Characterizing selection bias using experimental data, Econometrica, Vol. 66, pp. 1017–1099. Hirway, Indira. (2004): Providing Employment Guarantee in India: Some Critical Issues, Economic and Political Weekly, November 27, pp. 5117-24. Hochschild, Arlie Russel (1997): The Time Bind: When Work Becomes Home and Home Becomes Work, New York: Metropolitan Books. Holden, Stein T., Christopher B. Barrett, and Fitsum Hagos (2006): Food-for-Work for Poverty Reduction and the Promotion of Sustainable Land Use: Can it Work?, Environment and Development Economics, Vol. 11, pp. 15–38. IIPS, International Institute for Population Sciences (2007): National Family Health Survey (NFHS-3), 2005-06, India: Key Findings, Mumbai: IIPS. Indiastat.com (2008): Selected State-wise Average Daily Wage Rates for Non-Agricultural Occupation : Unskilled Labourers (Unspecified) by Sex in Rural India, several months, retrieved November 27, 2008, from http://www.indiastat.com/. 127
Jacobs, Jerry A. and Kathleen Gerson (2001): Overworked Individuals or Overworked Families?, Work and Occupations, Vol. 28, No. 1, pp. 40–63. Jalan, Jyotsna and Martin Ravallion (1999): Income Gains to the Poor from Workfare: Estimates for Argentina’s Trabajar Program, The World Bank Policy Research Working Papers, No. 2149. Juster, F. Thomas and Frank P. Stafford (1991): The Allocation of Time: Empirical Findings, Behavioral Models, and Problems of Measurement, Journal of Economic Literature, Vol. 29, No. 2, pp. 471–522. Katz, Lawrence F. and David H. Autor (1999): Changes in the Wage Structure and Earnings Inequality, in: O. Ashenfelter and D. Card (Eds.), Handbook of Labor Economics, Vol. 3A, pp. 1463–1555, Amsterdam: Elsevier. Künzi, Kilian and Markus Schärrer (2004): Wer zahlt für die Soziale Sicherheit und wer profitiert davon? Eine Analyse der Sozialtransfers in der Schweiz, Zurich: Rüegger. Krane, Dale and Gary S. Marshall (2005): Democracy and Public Policy, in: Jack Rabin (Ed.), Encyclopedia of Public Administration and Public Policy, pp. 78–89, Boca Raton: Taylor & Franics Group. Labour Bureau (2008): Consumer Price Index for Agricultural Labourers, retrieved Decenmber 1, 2008, from www.labourbureau.nic.in/indtab.html. Le Grand, Julien (1982): The Strategy of Equality: Redistribution and the Social Services, London: Allen and Unwin. Leu, Robert E., René L. Frey, and Brigitte Buhmann (1988): Budgetinzidenz: Wer profitiert von den öffentlichen Leistungen und wer zahlt dafür?, in: René L. Frey and Robert E. Leu (Eds.), Der Sozialstaat unter der Lupe. Wohlstandsverteilung und Wohlstandsumverteilung in der Schweiz, pp. 149–174, Basel: Helbing & Lichtenhahn. Linder, Staffan (1970): The Harried Leisure Class, New York: Columbia University Press. Macpherson, Crawford B. (1966): The Real World of Democracy: The Massey Lectures, Oxford: Oxford University Press. McGratten, Ellen and Richard Rogerson (2004): Changes in Hours Worked: 1950-2000, Federal Reserve Bank of Minneapolis Quarterly Review, pp. 14–33. Meerman, Jacob (1979): Public Expenditures in Malaysia: Who Benefits and Why?, New York: Oxford University Press. Ministry of Finance (2008): Economic Survey 2007-2008, Delhi: Ministry of Finance. 128
MRD, Ministry of Rural Development (2006): The National Rural Employment Guarantee Act 2005 (NREGA): Operational Guidelines 2006, 2nd Edition, Delhi: MRD. MRD, Ministry of Rural Development (2008a): Budget, retrieved September 29, 2008, from http://rural.nic.in/budget/Budgetframe.htm. MRD, Ministry of Rural Development (2008b): NREGA Report from MIS, retrieved November 26, 2008, from http://nrega.nic.in/MISreport.htm. MRD, Ministry of Rural Development (2008c): Prevailing Minimum Wages & Schedule of Rates (SOR) for Agricultural Labourers, retrieved November 27, 2008, from http://nrega.nic.in/wages.pdf. Mujeri, Mustafa K. (2002): Bangladesh: Bringing Poverty Focus in Rural Infrastructure Development, ILO Issues in Employment and Poverty Discussion Paper, No. 6. Nayyer, R. (2002) (2002): The Contribution of Public Works and Other Labour-Based Infrastructure to Poverty Alleviation: The Indian Experience, ILO Issues in Employment and Poverty Discussion Paper, No. 3. Nelissen, Jan H.M. (1998): Annual Versus Lifetime Income Redistribution by Social Security, Journal of Public Economics, Vol. 68, pp. 223–249. O’Donnell, Owen, Eddy van Doorslaer, Ravi P. Rannan-Eliya, Aparnaa Somanathan, Shiva Raj Adhikari, Deni Harbianto, Charu C. Garg, Piya Hanvoravongchai, Mohammed N. Huq, Anup Karan, Gabriel M. Leung, Chiu Wan Ng, Badri Raj Pande, Keith Tin, Kanjana Tisayaticom, Laksono Trisnantoro, Yuhui Zhang, and Yuxin Zhao (2007): The Incidence of Public Spending on Healthcare: Comparative Evidence from Asia, The World Bank Economic Review, Vol. 21, No. 1, pp. 93–123. Okun, Arthur M. (1975): Equality and Efficiency, the Big Tradeoff, Washington, D.C.: Brookings Institution. Planning Commission (2008a): Eleventh Five Year Plan 2007-12, Volume I : Inclusive Growth, New Delhi: Oxford University Press. Planning Commission (2008b): Eleventh Five Year Plan 2007-12, Volume III : Agriculture, Rural Development, Industry, Services and Physical Infrastructure, New Delhi: Oxford University Press. Prescott, Edward C. (2004): Why do Americans Work so Much More Than Europeans?, Federal Reserve Bank of Minneapolis Quarterly Review, Vol. 28, No. 1, pp. 2–13. 129
Ravallion, Martin (2008): Evaluating Anti-Poverty Programs, in: T. Paul Schultz and John A. Strauss (Eds.), Handbook of Development Economics, Vol. 4, pp. 3787–3846, Amsterdam: Elsevier. Ravallion, Martin and Gaurav Datt (1995): Is Targeting through a Work Requirement Efficient? Some Evidence for Rural India, in: Dominique van de Walle (Ed.), Public Spending and the Poor – Theory and Evidence, pp. 413–444, Baltimore: The Johns Hopkins University Press. Ravallion, Martin, Gaurav Datt, and Shubham Chaudhuri (1993): Does Maharashtra’s Employment Guarantee Scheme Guarantee Employment? Effects of the 1988 Wage Increase, Economic Development and Cultural Change, Vol. 41, No. 2, pp. 251–275. Ravallion, Martin, Emanuela Galasso, Teodoro Lazo, and Ernesto Philipp (2005): What Can Ex-Participants Reveal about a Program’s Impact?, The Journal of Human Resources, Vol. XL, No. 1, pp. 208–230. Ray, Debraj (1998): Development Economics, Princeton: Princeton University Press. Reid, Margaret G. (1934): Economics of Household Production, New York: John Wiley & Sons. Reynolds, Morgan O. and Eugene Smolensky (1977): Public Expenditures, Taxes, and the Distribution of Income: The United States, 1950, 1961, 1970, New York: Academic Press. Rice, James Mahmud, Robert E. Goodin, and Antti Parpo (2006): The Temporal Welfare State: A Crossnational Comparison, Journal of Public Policy, Vol. 26, No. 3, pp. 195–228. Robinson, John P. and Geoffrey Godbey (1996): The Great American Slowdown, American Demographics, Vol. 18, No. 6, pp. 42–48. Robinson, John P. and Geoffrey Godbey (1999): Time for Life: The Surprising Ways Americans Use Their Time, University Park: The Pennsylvania State University Press. Rosenbaum, Paul R. and Donald B. Rubin (1983): The Central Role of the Propensity Score in Observational Studies for Causal Effects, Biometrika, Vol. 70, pp. 41–55. Ruggeri, Giuseppe C. (2003): Public Expenditure Incidence Analysis, in: Anwar Shah (Ed.), Bringing Civility in Governance, Vol. 3 of Handbook on Public Sector Performance Reviews, Ch. 1, Washington, D.C.: The World Bank. Sayer, Liana C., Suzanne M. Bianchi, and John P. Robinson (2004): Are Parents Investing Less in Children? Trends in Mothers’ and Fathers’ Time with Children, American Journal of Sociology, Vol. 110, No. 1, pp. 1–43. 130
Schor, Juliet B. (1991): The Overworked American: The Uunexpected Decline of Leisure, New York: Basic Books. Selden, Thomas M. and Michael J. Wasylenko (1995): Measuring the Distributional Effects of Public Education in Peru, in: Dominique van de Walle and Kimberly Nead (Eds.), Public Spending and the Poor – Theory and Evidence, pp. 154–186, Baltimore: Johns Hopkins University Press. Sen, Amartya (1999): Development as Freedom, New Delhi: Oxford University Press. SFSO, Swiss Federal Statistical Office (1990-2005): Survey of Income and Expenditure (SIE), Bern: SFSO. Sinn, Hans-Werner (2006): Der neue Sozialstaat, ifo Standpunkt, No. 72. Subbarao, Kalanidhi (1997): Public Works as an Anti-Poverty Program: An Overview of Cross-Country Experience, American Journal of Agricultural Economics, Vol. 79, No. 2, pp. 678–683. Subbarao, Kalanidhi (2003): Systemic Shocks and Social Protection: Role and Effectiveness of Public Works Programs, The World Bank Social Protection Discussion Paper Series, No. 302. Sullivan, Oriel and Jonathan Gershuny (2001): Cross-National Changes in Time-use: Some Sociological (Hi)stories Re-examined, British Journal of Sociology, Vol. 52, No. 2, pp. 331–347. Suter, Christian and Marie-Claire Mathey (2000): Wirksamkeit und Umverteilungseffekte staatlicher Sozialleistungen, info:social 3:00, Swiss Federal Statistical Office. Tanzi, Vito and Ludger Schuknecht (2000): Public Spending in the 20th Century: A Global Perspective, Cambridge: Cambridge University Press. van de Walle, Dominique (1995): The Distribution of Subsidies through Public Health Services in Indonesia, 1978-87, in: Dominique van de Walle and Kimberly Nead (Eds.), Public Spending and the Poor – Theory and Evidence, pp. 154–186, Baltimore: Johns Hopkins University Press. van de Walle, Dominique (1996): Assessing the Welfare Impacts of Public Spending, The World Bank Policy Research Working Paper, No. 1670. van de Walle, Dominique and Kimberly Nead (Eds.) (1995): Public Spending and the Poor – Theory and Evidence, Baltimore: Johns Hopkins University Press. 131
Verbeek, Marno and Theo Nijman (1992): Can Cohort Data be Treated as Genuine Panel Data?, Empirical Economics, Vol. 17, No. 1, pp. 9–23. Vermaeten, Arndt Vermaeten; W. Irwin Gillespie; Frank (1995): Who Paid the Taxes in Canada, 1951-1988?, Canadian Public Policy, Vol. 21, No. 3, pp. 317–343. Webb, Patrick (1995): Employment Programs for Food Security in Rural and Urgan Africa: Experiences in Niger and Zimbabwe, in: Joachim von Braun (Ed.), Employment for Poverty Reduction and Food Security, pp. 174–200, Washington: International Food Policy Research Institue (IFPRI). Whalley, John (1984): Regression or Progression: The Taxing Question of Incidence Analysis, The Canadian Journal of Economics, Vol. 17, No. 4, pp. 654–682. World Bank (2006): Inclusive Growth and Service delivery: Building on India’s Success, Development Policy Review, Report No. 34580-IN, Washington, D.C.: World Bank. Zeckhauser, Richard (1973): Time as the Ultimate Source of Utility, The Quarterly Journal of Economics, Vol. 87, No. 4, pp. 668–675.
132
Curriculum Vitae Born March 3, 1978 in Chur, Switzerland
Education 2006-2009
University of St. Gallen (HSG), Switzerland Doctorate in Economics
2008
Indian School of Business, Hyderabad, India Visiting Scholar, Department of Economics
2005-2007
Swiss Federal Institute of Technology (ETH), Zurich, Switzerland Diploma of Advanced Studies in Applied Statistics
1998-2003
University of St. Gallen (HSG), Switzerland Licentiate in Economics (lic.oec.)
Professional Experience Since 2009
Department of Finance of the Canton of St. Gallen, Switzerland Economist, Budgetary Planning
Since 2006
Swiss Institute for Empirical Economic Research, St. Gallen, Switzerland Research Assistant, Chair of Prof. Dr. Monika Bütler
2003-2006
Credit Suisse Group, Zurich, Switzerland Economist, Economic and Policy Research Department
