70570 SOCIAL INSURANCE AND LABOR SUPPLY: ASSESSING INCENTIVES AND REDISTRIBUTION TECHNICAL REPORT December 30, 2009 Human Development Sector Management Unit Latin America and the Caribbean Region Document of the World Bank Acronyms and Abbreviations AAA Analytical and Advisory Activities BNDES Banco Nacional de Desenvolvimento Econômico e Social BPC Beneficio de Prestação Continuada (Pensions for the Elderly Poor) CEF Caixa Econômica Federal (Federal Savings Union) FGTS Fundo de Garantia por Tempo de Serviço (Link of. Cerzisse Guarantee Fund) GDP Gross Domestic Product GOB Government of Brazil INSS Instituto Nacional de Seguridade Social (National Institute of Social Security) IPEA Instituto for Applied Economic Research IRR Internal Rate of Return (on contributions) IT Information Technology LOC Length of Contribution Pension MTE Ministry of Labor and Employment OECD Organization for Economic Cooperation and Development PLOC Proportional Length of Contribution Pension RGPS Regime Geral de Proteção Social (General Scheme for Social Protection) SI Social Insurance SP Social Protection MPS Ministry of Social Security UI Unemployment Insurance TTL Task Team Leader Vice-President: Pamela Cox Country Director: Makhtar Diop Sector Director: Evangeline Javier Sector Manager: Helena Ribe Task Team Leaders: David Robalino/ Bénédicte de la Brière i 1. MOTIVATIONS, OBJECTIVES AND CONTEXT ....................................................................................... 1 1.1. Why this Technical Report? 1 1.2. Some Stylized Facts About Labor Market Dynamics in Brazil 3 2. SOCIAL INSURANCE, REDISTRIBUTION AND ECONOMIC INCENTIVES ................................................. 9 2.1. Financing Mechanisms, Tax Wedge and Employment 10 2.2. Pensions 14 The mandate of the system ................................................................................................... 15 Retirement decisions ............................................................................................................. 16 Sector choice and entry into the labor force ......................................................................... 21 Redistribution........................................................................................................................ 24 2.3. Income Protection System for Workers 26 The mandate of the system ................................................................................................... 27 Incentives to work and sector choice .................................................................................... 28 Redistribution........................................................................................................................ 32 Other issues that deserve attention ........................................................................................ 33 3. IMPROVING REDISTRIBUTION AND INCENTIVES................................................................................. 34 3.1. Policy Framework for Social Insurance Policy 34 3.2. Implications for Brazil 36 Pensions ................................................................................................................................ 36 Income protection ................................................................................................................. 38 3.3. Potential Effects on Behaviors and Program Costs 39 REFERENCES .............................................................................................................................................. 44 TABLES Table 1: Summary of Pay-Roll Taxes and Social Security Contributions ................................................. 11 Table 2: Expenditures in Social Insurance (excluding health)................................................................... 12 FIGURES Figure 1: Labor Force Age/Gender Composition and Growth .................................................................... 4 Figure 2: Labor Force by Occupation and Distribution of Net Jobs ............................................................ 4 Figure 3: Labor Force by Level of Education .............................................................................................. 5 Figure 4: Distribution of Earnings for Workers with and Without Carteira ................................................ 5 Figure 5: Monthly Probabilities of Separation in the Formal and Informal Sector ..................................... 6 Figure 6: Distribution of a “Typical� Cohort of 25 Year-Old Males by State (Urban Region) ................... 9 Figure 7: Tax Wedge in Brazil and Selected Countries ............................................................................. 13 Figure 8: Gross Replacement Rates by Retirement Age ............................................................................ 16 Figure 9: Retirement Ages and Life Expectancies Around the World ...................................................... 17 Figure 10: Distribution of Type of Pension, Vesting Period and Retirement Probabilities ....................... 18 Figure 11: Change in Pension Wealth for Each Additional Year of Contribution .................................... 19 Figure 12: Internal Rates of Return on Contributions by Retirement Age and Level of Income .............. 20 Figure 13: Informal Sector Productivity, Payroll Taxes, and Incentives for Informality .......................... 22 ii Figure 14: Expected Net Life-Time Wealth for Low Income Individuals ................................................. 24 Figure 15: Relative and Absolute Implicit Subsidies................................................................................. 26 Figure 16: Replacement Rates and Accumulations from UI and FGTS .................................................... 28 Figure 17: Replacement Rates and Benefit Duration in Selected Countries.............................................. 28 Figure 18: Relative Speed of Exit from Unemployment by Type of Worker and Sector Choice.............. 30 Figure 19: Take-up Rates and Share of UI Payments by Income Level .................................................... 33 Figure 20: Projected Expenditures on Unemployment Benefits ................................................................ 33 Figure 21: Designing Subsidies for Individuals with No or Limited Savings Capacity ............................ 38 Figure 22: Probabilities of Contributing to INSS and Retiring.................................................................. 40 Figure 23: Effects of SI Programs on Contribution Densities and Retirement Ages (Average Earner) .... 41 Figure 24: Effects of Matching Contributions ........................................................................................... 42 Figure 25: Costs of Matching Contributions and Minimum Pensions ....................................................... 43 iii This Technical Report is the result of a joint effort between the Government of Brazil, local academic and research institutions, and the World Bank. The report builds on the considerable amount of empirical analysis by Brazilian researchers compiled during the first phase of the PESW on Labor Markets. The new analytical work and the preparation of the Report was conducted by a team including: David A. Robalino (TTL and main author), Benedicte De La Briere (Co-TTL), Wendy Cunningham, Jason Hobbs, Stanislao Maldonado, Lerick Kebeck, Carla Zardo (World Bank), Hélio Zylberstajn and Luis Eduardo Afonso (University of Sao Paulo), André Portela, Vladimir Ponczek and Eduardo Zylberstajn (Getúlio Vargas Foundation/São Paulo), and David Margolis and Raul Samponagro (University of Paris I). The team worked under the guidance of Helena Ribe (Sector Manager Social Protection). The team would like to thank Luis Enrique da Silva de Paiva (MPS), Adriana Maria Giubertti (MTE), Antônio Sérgio A. Vidigal (Secretário de Políticas Públicas de Emprego, MTE), Rodolfo Péres Torelly (Seguro-Desemprego - MTE) and Milena Souto Maior de Medeiros (MTE) for helpful discussions and for providing access to data. Also our thanks go to the teams at IPEA Rio de Janeiro and IPEA Brazilia for insightful comments at various stages in the preparation of the analysis. The peer reviewers for this Technical Report were Bill Maloney, Polly Jones, and Gordon Betcherman. iv 1. MOTIVATIONS, OBJECTIVES AND CONTEXT This Technical Report analyzes the potential effects of the pensions and income protection systems on labor supply decisions and through this channel the coverage and cost of the programs. The focus is on old-age pensions provided as part of the General Scheme for Social Protection (RGPS) by the National Institute of Social Security (INSS), and the income protection programs managed by the Caixa Econômica Federal (CEF), which include the Unemployment Insurance (UI) system and the Length of Service Guarantee Fund (FGTS) -- essentially funded unemployment individual savings accounts. The Report is part of Pillar I of the Brazil PESW on Labor Markets, a three year program of analytical and advisory services aiming at improving the design of social protection programs to promote the creation of high quality jobs and help low income individuals graduate from welfare. Pillar I investigates issues related to the effects of public transfers on labor supply, in particular, whether there are reforms that need to be considered to improve incentives for work and stimulate formal vs. informal employment. The Report addresses five policy questions: (i) how is the pension system influencing decisions regarding entry into the labor market, sector choice (formal/informal), and retirement? (ii) How are current income protection programs affecting turnover; job search efforts and sector choice?; (iii) Are there interactions between the two systems that aggravate or mitigate incentive effects?; (iv) What is the role of redistributive policies in determining observed outcomes?; and (v) Are there policy interventions that could be considered to correct incentives, while securing adequate income protection for different population groups (including workers in the informal sector)? The Report is organized in three sections. This first section discusses the motivations for the study and sets the context by briefly describing some key stylized facts about the Brazilian labor market. The focus is on the composition of the labor force, its distribution by occupational categories, unemployment risks, and transitions between employment states -- including workers flows between the formal and informal sectors. The second section analyzes the Brazilian social insurance system in terms of incentives and redistribution. It starts with an overview of institutional arrangements, programs, and financing mechanisms, which flags the problem of a high tax-wedge. The Report then analyzes the rules of the pensions and income protection systems and their potential effects on behaviors. The final section proposes a policy framework to guide reforms that could “correct� incentives by making redistribution more transparent and progressive. A behavioral life-cycle model estimated for Brazil is used to illustrate how the application of this framework could affect contribution densities (and therefore the time spent working in formal sector jobs), retirement ages, savings, and programs cost. A companion Policy Note on Social Insurance and Labor Supply summarizes the main conclusions and recommendations from the analysis. 1.1. WHY THIS TECHNICAL REPORT? The main motivation for preparing this Report is a general concern in Brazil with the unintended economic consequences of public transfers. Indeed, transfers through social insurance, active labor markets and social assistance programs are estimated at 13 percent of GDP.1 Some of these transfers are explicit, like those related to social assistance programs; others are implicit, like in the case of pensions and unemployment insurance when internal rates of return on contributions for certain population groups are above market rates. In general, there is evidence that the various programs have played an important role in helping individuals manage risks such as unemployment and inflation, and have contributed to reduce poverty.2 At the same time, however, there are concerns among Brazilian policymakers about the potential effects of these programs on incentives to work and save and the associated economic and fiscal 1 This number excludes the “Sistema S,� which represents the core of the public training programs in Brazil. 2 See World Bank, 2002 and World Bank 2007. 1 costs. Indeed, experiences elsewhere show that some of these effects can be important, including in the case of insurance programs.3 In OECD countries, for instance, mandatory earnings related to pension systems are shown to encourage early retirement, which costs the economies around 7 percent of GDP. In Brazil, several studies reviewed throughout the Report also flag problems with incentives in the pensions and income protection systems.4 This Report argues that incentives problems depend, to a large extent, on the type of redistribution embedded in the various programs. In principle, programs that do not redistribute income are less likely to affect behaviors. By design, however, social insurance programs are redistributive. Redistributive in the sense that, systematically, some population groups “get out� of the programs more than what they “put in,� which means that others get less. In essence, the programs involve subsidies and taxes which depend not only on intrinsic characteristics of the individuals but also on their behaviors. Individuals might attempt to exploit subsidies and avoid taxes. And as a general rule, the less explicit the form of redistribution, the more prone the system is to abuse or strategic gaming and the more likely that it will induce undesirable behaviors. For instance, individuals might have incentives to avoid the formal sector, strategically manipulate wages, retire early, reduce job search efforts, and, in general, reduce labor supply and savings. This type of behaviors increase the fiscal costs of the programs, can reduce aggregate productivity and output, and at the end also generate adverse redistribution. In countries like Brazil, where the formal and informal sectors are reasonably integrated, problems with the design of the social insurance system can be amplified. Indeed, in a country where the labor market is segmented into a formal and informal sector and workers flows among the two are thin or non-existent, benefit formulas and eligibility conditions in the Social Insurance System (SI) system will do little to affect the relative sizes of the two sectors5 and/or contribution densities. The Social Insurance (SI) system can still influence when individuals retire, how much they save, and how much effort they put into exiting unemployment, but it is unlikely to affect flows between the two sectors or relative job finding rates. Things are more complex in a country like Brazil where the empirical evidence suggests that the informal sector, at least in part, can be seen as an unregulated, voluntary self-employed sector.6 In this case, workers seem to make explicit choices about formal vs. informal jobs based on an assessment of expected costs and benefits. The SI system can affect these costs and benefits and thus influence relative separation and job finding rates and the size of the formal sector. There is evidence, for instance, that the constitutional reforms implemented in 1988 which increased the direct and indirect costs of labor, contributed to the contraction of the formal sector during the 90s.7 In part, this can be explained by lower hiring rates in the formal sector, but as this Report will argue, reduced efforts to find/keep jobs in the formal sector can also be part of the story. The Report will illustrate how current SI programs can be upgraded to improve both equity and economic efficiency by making redistribution explicit. Redistribution per se is not a problem. Having redistributive public programs is a choice that most societies make to different degrees. Some losses in efficiency are to be expected, but these are compensated by the social benefits that redistribution brings. The problem is when redistribution brings excessive and unnecessary losses. The Report argues that this can be the case when the insurance and redistributive functions of the programs are not well defined. Improving efficiency then implies separating these two functions and ensuring, to the extent possible, that the risk- pooling/savings programs are incentive neutral8 and that public subsidies are then used to “top up� the benefits or contributions of individuals with limited or no savings capacity. These top-ups would still 3 See review of the literature prepared during Phase I of the PESW (Olinto et al. 2007). 4 For a discussion of the unintended consequence of public transfers see also Levy (2007). 5 The tax-wedge can still create a ceiling for low-productivity firms that are then forced into the informal sector (see Section 2). 6 See Bosh and Maloney (2008). 7 See Bosh et al. (2007). 8 This would imply having an actuarial link between the benefits paid by the system and the contributions. This type of neutrality can be achieved, in principle, regardless of the type of financing mechanism. 2 affect incentives for those with incomes close to the eligibility line, but the effects would be more localized and would concern fewer and less productive workers.9 Moreover, negative effects can be minimized by appropriately designing the “claw-back� for the transfer (i.e., the marginal tax on the transfer) and/or introducing conditionalities. The Report is based on a series of analytical pieces looking at how benefit formulas and eligibility conditions in the pensions and income protection systems can affect individual behaviors. In the case of pensions, the analysis is first based on the calculation of four indicators that provide information about incentives (not actual behaviors) to participate in the labor force, save and contribute to the social security. These are: the pension wealth, life-time net earnings, internal rates of return on contributions, and implicit taxes and subsidies received by different plan members. For the income protection system, an econometric model was estimated looking at the potential effect of the unemployment insurance system and the unemployment individual FGTS savings on employment duration and sector choice in Metropolitan areas.10 A review of previous studies was also conducted to better understand the impact that FGTS might be having on hiring and turn-over rates. Finally, the interactions between the pensions and income protection systems and the potential impacts of changes in benefit formulas and eligibility conditions on individuals work choices were analyzed on the basis of a life cycle behavioral model. The model takes into account a large range of possible behavioral responses that could affect contribution densities, the distribution of retirement ages, savings, and ultimately program costs.11 The focus is on how different features of the programs (i.e., benefit formulas and eligibility conditions) affect job search/job preservation efforts in the formal sector, which in turn affect separation and job finding rates. This is in contrast to the more traditional analyses looking at how the presence of a given SI program (e.g., the unemployment system) affects outcomes such as the duration of unemployment spells or separation/job finding rates in the formal sector, without taking into account, explicitly, individual choices. 1.2. SOME STYLIZED FACTS ABOUT LABOR MARKET DYNAMICS IN BRAZIL Today, there are around 88 million people in the Brazilian labor force, a number that has been growing quite rapidly over the past decade. Between years 2001 and 2006 the labor force grew at an average of 3.4 percent per year, among the fastest expansions observed in the world. If current trends continue, by year 2010 the number of people working or looking for jobs could surpass 95 million. The labor force remains very young with more than 90 percent of its members below age 50, compared to 74 percent in OECD or 64 percent in Japan. There has also been a considerable increase in the share of women who now represent 35 percent of its members. Their participation rate, however, is still considerably lower than in OECD countries. Only around 50 percent of Brazilian women between ages 25 and 50 participate in the labor market compared to an average of 80 percent in OECD. Like in other countries in the region, at any point in time, a large share of the labor force is outside the formal sector. In Brazil, formal sector workers are classified in the household survey as those with “carteira.� The “carteira� is a document indicating that the contract was registered by the employer with the Ministry of Labor. Thus, the majority of workers with carteira have social security coverage; most workers without carteira do not. Workers with carteira represented in 2006 only 33.7 percent of the labor force down from 40 percent in 1990. Workers without carteira and the self-employed, on the other hand, represented respectively 28 and 8 percent of the labor force. This group of informal sector workers has been expanding. Indeed, between 1990 and 2006 the economy created, net, 31.2 million jobs. A large part was jobs without carteira (27 percent) and self-employment (14 percent). 9 On this see also Progress Report on Phase I of the PESW (Linder et al. 2007). 10 See Margolis (2008). 11 See Robalino et al. (2008). 3 Figure 1: Labor Force Age/Gender Composition and Growth 90 4.0% 40% Growth rate (left) 80 60+ 3.5% 35% 70 50-59 3.0% 30% Labor Force (million) 60 Share of Women 2.5% 25% Growth Rate 50 ` 2.0% 20% 40 1.5% 15% 30 25-49 Share of women (right) 1.0% 10% 20 10 0.5% 5% 16-24 0 0.0% 0% 1950 1960 1970 1980 1990 1995 2000 2006 50's 60' 70' 80' 90' 00' Source: Authors’ calculations based on the PNADs 1990 and 2006. Figure 2: Labor Force by Occupation and Distribution of Net Jobs 2006 Retired and working Without carteira 1990 Unemployed With carteira Employers Self-employed Self-employed Retired and working Civil servants & mili Civil servants & mili Without carteira With carteira Employers 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 0% 5% 10% 15% 20% 25% 30% 35% Share in the Labor Force Share of Net Jobs Created (1990-2006) Source: Authors’ calculations based on the PNADs 1990 and 2006; and Censo Demográfico 1950, 1960,1970, 1980, and 2000. Although there have been important improvements over the last 15 years the level of education of the labor force remains low. Over 60 percent of the labor force has less than 8 years of schooling; only 12 percent have more than twelve years (see Figure 3). Not surprisingly, the most educated workers are in the civil service or the military, followed by the private formal sector (i.e., workers with carteira). The least educated are the self employed followed by workers in the informal sector. In fact, the probability of being in the informal sector goes down with the number of years of schooling. Still, to a certain degree, both the formal and informal sectors have a mixture of educated and non-educated workers. For instance, 40 percent of workers in the informal sector have secondary education relative to 60 percent in the formal sector. And 7 percent have a diploma from tertiary education relative to 14 percent in the formal sector. The unequal distribution of skills is matched by skewed earnings distributions both in the formal and informal sectors. Close to 70 percent of workers in the formal sector and 85 percent of workers in the informal sector earn less than the average wage. In the formal sector the median worker earns around 60 percent of the economy wide average wage. The minimum wage is around 42 percent of the average and 20 percent of workers seem to have earnings below this minimum. Only 20 percent of workers have earnings above the average (see Figure 4). In the informal sector the distribution of earnings is even more compact; the median worker has earnings which are close to the minimum wage. Thus, 50 percent of workers in the informal sector earn less than the minimum and only around 8 percent have earnings above the average. Among the self employed the distribution is broader. The median worker earns around 50 percent of average earnings and close to 24 percent have earnings above the average. These distributions, however, do not tell us about the “wage premium� that different individuals working in the informal sector would obtain if they move to the informal sector. Some recent studies argue that the average 4 formal sector premium (relative to informal salaried work) is only 7.8 percent and that it vanishes for high income workers.12 Figure 3: Labor Force by Level of Education 1-8 years 9-11 12+ 1-8 years 9-11 12+ Total All labor force Others Others Retired Not-Working Retired Not-Working Retired and working Retired and working Unemployed Unemployed Employers Employers Self-employed Self-employed Civil servants & mili Civil servants & mili Without carteira Without carteira With carteira With carteira 0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100% Source: Authors’ calculations based on the PNADs 1990 and 2006. Figure 4: Distribution of Earnings for Workers with and Without Carteira With Carteira Without Carteira 0.50 0.50 0.45 0.45 Minimum 0.40 0.40 wage 0.35 0.35 0.30 0.30 Minimum 0.25 wage 0.25 0.20 0.20 0.15 0.15 Median 0.10 0.10 Median 0.05 0.05 0.00 0.00 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Income (proportion of average earnings) Income (proportion of average earnings) Self-employed 0.50 0.45 0.40 0.35 0.30 Minimum wage 0.25 0.20 Median 0.15 0.10 0.05 0.00 0 0.5 1 1.5 2 Income (proportion of average earnings) Source: Authors’ calculations based on the PNAD (2006) Around 9 percent of the labor force is unemployed today, with the risk of unemployment being higher for low-income workers and those in the informal sector. Previous studies have estimated the probability of 12 See Botelho and Ponczek (2008). For a methodological discussion about the estimation of wage premiums and applications to other countries in Latina America see also Perry et al. (2007). 5 becoming unemployed conditional on individual characteristics and found that the groups facing the highest risk are young, unskilled, informal sector workers.13 A tabulation of the surveys for workers in metropolitan areas between August 2002 and July 2007 also shows that income levels and sector make a difference.14 On average, on any given month, less than one percent of high income formal sector workers (i.e., earning between 1 and 1.5 times the average wage) leave their job (see Figure 5). The share is higher, close to 1.5 percent, in the case of median and low income workers (earning between 0.5 and 0.75 times the average). Thus, over the course of one year, around 8 percent of high earners and 20 percent of median-low earners would have left their job. In the informal sector the chances of leaving a job are higher. In one year, close to 26 percent of high earners and 34 percent of median-low earners would have left their jobs.15 A recent study shows that fluctuations in unemployment rates are mainly driven by separation rates in the informal sector and low job finding rates in the formal sector.16 Transition probabilities into unemployment are countercyclical (transition rates increase during downturns). However, contrary to what is observed in high income countries (e.g., US and UK) the variation in unemployment rates is mainly explained by higher separation rates and not by lower job finding rates. Job separations can explain around 40 percent of all cyclical fluctuations in unemployment, almost twice as much as the explanatory power of fluctuations in the job finding rate. At the same time, the study confirms that higher separation rates during downturns are mainly driven by higher separation rates in the informal sector; separation rates in the formal sector tend to be a-cyclical. This has important implications for the design of the income protection system in Brazil, which today focuses on formal sector workers. Informal sector workers facing the highest risk of unemployment are not adequately protected. Once unemployed the average duration of the spell is close to 6 months in metropolitan areas. The path out of unemployment is correlated with the level of education and age. Individuals with low education (8 years or less) have a higher propensity to go into informal sector jobs or self employment than educated workers. The opposite is true for formal sector jobs. Also, the likelihood of entry into the informal sector goes down with age, but increases with age in the case of self-employment. For formal sector jobs the likelihood increases up to age 40 and declines afterwards.17 Figure 5: Monthly Probabilities of Separation in the Formal and Informal Sector 10.0 10.0 9.0 9.0 8.0 8.0 7.0 7.0 50%-75% 6.0 6.0 5.0 5.0 4.0 4.0 3.0 3.0 50%-75% 2.0 100%-150% 2.0 1.0 1.0 100%-150% 0.0 0.0 Ju 2 Ju 3 Ju 4 Ju 5 Ju 6 Ju 7 Ju 2 Ju 3 Ju 4 Ju 5 Ju 6 Ju 7 Ja 02 A 03 Ja 03 A 04 Ja 04 O 03 O 04 A 05 Ja 05 O 02 O 05 A 06 Ja 06 O 06 A 07 Ja 02 Ja 03 Ja 04 Ja 05 Ja 06 7 O 02 A 03 O 03 A 04 O 04 A 05 O 05 A 06 O 06 A 07 7 -0 -0 -0 -0 -0 -0 -0 -0 -0 -0 -0 -0 l-0 l-0 l- - n- l- - n- l- - n- l- - n- l- - n- l- - n- l- - n- l- - n- l- - n- l- - n- pr pr pr pr pr pr pr pr pr pr pr pr ct ct ct ct ct ct ct ct ct ct A A Source: Authors’ calculations using PME. 13 See Domeland and Fiess (2002) and Orellano and Picheti (2002). 14 Pesquiza Mensual de Emprego (PME). 15 These numbers, however, do not tell the extent of voluntary and involuntary separations, which seem to decline with age and education. 16 Bosh and Maloney (2008). 17 Bosh and Maloney (2007). 6 Overall, the Brazilian labor market seems to be quite vibrant; the duration of employment spells is relatively short particularly in the informal sector and in self-employment. The last in depth assessment of the labor market had pointed to a very high turnover rate. Thus, during the nineties, while in other countries in the region as well as in the United States the average worker in the manufacturing sector spent 8-9 years in the same job, in Brazil the average duration was only 6 years.18 Other studies looking at labor dynamics find job reallocation rates in the order of 33 percent per year.19 Job destruction rates are estimated at about 15 percent. In metropolitan areas, the average duration of formal sector jobs is around 4.5 years, while, the duration of self-employment and informal sectors jobs is respectively 2.3 years and a little less than one year. 20 Not surprisingly, female are more mobile than men, young workers more mobile than older workers, and those with low education more mobile than those with high education. An analysis of workers transitions across employment states also challenges the view of the informal sector as a warehouse for residual labor.21 The analysis in question exploits changes in transition probabilities across states during the economic cycle. It shows that formal employment as a share of working age population is pro-cyclical; during downturns formal employment is reduced while informal employment increases.22 On average, a 1 percent fall in output (from its trend) increases the share of informal employment by 0.20 percent and decreases the share of formal labor by 0.5 percent. This result would be consistent with the view of the informal sector as an employer of last resort. However, the study also shows that lower formal employment is mainly explained by lower job finding rates in the formal sector and not by higher separation rates. Moreover, movements from informal to formal sector jobs and from formal to informal sector jobs are both pro-cyclical, which contradicts the view that during downturns transition rates from the formal to the informal sector increase (i.e., are countercyclical). In fact, during a downturn, the informal sector does not act as a safety net. Flows from the formal to the informal sector actually decrease. On the contrary, there is evidence that the formal and informal sectors are quite integrated. 23 Workers move in and out of formal and informal sector jobs. For instance, around 46 percent of informal sector workers would transit to formal sector jobs in the course of the year. The opposite movement is less likely, only around 7 percent of formal sector workers go to the informal sector in the course of a year. But this reflects, in part, a lower probability of separation from formal sector jobs. In fact, among individuals who leave formal sector jobs, around 30 percent would become wage earners in the informal sector and 16.7 percent self-employed. The results of self-reports also show that a significant share of workers in the informal sector do not consider their jobs as inferior options. Indeed, around 30 percent of males and 37 percent of females in informal salaried jobs report not wanting to have a formal sector job.24 This share is positively correlated with age. Thus, 23 percent of males and 28 percent of females between ages 19 and 24 prefer the informal sector job vs. 55 percent of males and 70 percent of females between ages 55 and 70. Among the reasons for preferring informal salaried jobs are higher earnings (6 percent of males and 2 percent of females), having time to do other activities (7 percent of males and females), having time to take care for home (18 percent of females), and simply liking the activity in question (67 percent of males and 60 percent of females). Among the self-employed, the shares of those not wanting formal sector jobs are even higher: 68 percent of males and 55 percent of females. Among males the reasons are higher earnings (18 percent of respondents) and liking the activity (65 percent). Among females, also preferring 18 See World Bank, 2002. 19 See Corseuil et al. (2003), Menezes-Filho and Fernandes (2003) and Gonzaga (2003). 20 Bosh and Maloney (2007b). 21 See Bosh and Maloney (2008). 22 This refers to both self employment and informal salaried work. 23 Bosh and Maloney (2007). 24 See Chapter 2 in Perry et al. (2007) 7 higher earnings (10 percent) and liking the activity (44 percent), but in addition, having the flexibility to take care of children (26 percent). The most recent study comparing earnings in the formal and informal sectors also finds relative small premiums to formal sector jobs that vanish with the individual level of earnings.25 Thus, transitions between formal and informal sector jobs (including self-employment) seem to reflect, at least in part, explicit individual choices regarding where to work. This issue has been analyzed by looking at transitions between jobs, after controlling for separations and the extent of job opportunities.26 Flows into and from self-employment and/or informal sector jobs that are not very different would support the view that workers make choices about where they work, which presumably reflect expected costs and benefits.27 The evidence for Brazil is that flows are very similar between formal sector jobs and self-employment, suggesting choice. Flows between formal and informal salaried work are also symmetric for women. Flows, however, are asymmetric for men indicating a preference for formal vs. informal sector jobs. One reason could be the non-pecuniary benefits received through the social insurance system. For instance, given high internal rates of return on contributions individuals would have incentives to look for formal sector jobs up to the point where they qualify for a pension -- if they are able to afford the contributions (see Section 2). But, working continuously in the formal sector might not be the best strategy (see Section 3). In fact, the majority of people who retire from the private sector do not contribute continuously. The average vesting period is 20 years for females and 25 years for males a number that is difficult to explain simply by “bad luck.� Thus the importance of looking at benefit formulas and eligibility conditions in the pensions system and assess whether these can be modified to improve incentives to contribute and extend contribution densities. In summary, a representative member of a given age-cohort entering the labor market is likely to alternate between formal sector work, work in the informal sector, and unemployment, retiring in most cases with a public pension. The typical distribution of the cohort by states is strongly correlated with income (see Figure 6). Low income individuals (first panel), for instance, spend less time contributing to the social security than high income workers (third panel). They also spend more time unemployed and are likely to retire latter. In the middle, the likelihood of contributing to the social security is higher, but participation seems to go down after age 30, most likely as a result of more frequent transitions into self- employment. Middle income workers retire latter than all the others. In all cases, and not surprisingly, unemployment is more likely in younger ages. The next section will investigate to what extent the current pension and income protection systems can affect these transitions. Of particular interest are: (i) transitions between formal jobs and informal jobs/self-employment; (ii) transitions out of unemployment; and (iii) transitions out of the labor force into retirement. The main hypothesis is that, by affecting net earnings in various employment states, the pensions and unemployment insurance systems can influence individual decisions about how much and where to work, and when to retire. The goal is then to identify designs that are more likely to provide incentives for formal versus informal sector work and to reduce incentives for retirement over work. 25 See Bothelo and Ponczek (2008). 26 Ibid. 27 On this issue see also Perry et al. (2007). 8 Figure 6: Distribution of a “Typical� Cohort of 25 Year-Old Males by State (Urban Region) Income <50% of average Income 50% to 75% of average 100% 100% Unemployed Retired Retired 90% Unemployed 90% 80% 80% 70% 70% Outside of the social security 60% 60% 50% 50% 40% Outside of the social security 40% 30% 30% 20% 20% Contributing to the Social Security Contributing to the Social Security 10% 10% 0% 0% 16-20 21-25 26-30 31-35 36-40 41-45 46-50 51-55 56-60 61-65 66-70 70+ 16-20 21-25 26-30 31-35 36-40 41-45 46-50 51-55 56-60 61-65 66-70 70+ Age Group Age Group Income > 75% of average 100% Unemployed Retired 90% 80% Outside of the social security 70% 60% 50% 40% 30% 20% Contributing to the Social Security 10% 0% 16-20 21-25 26-30 31-35 36-40 41-45 46-50 51-55 56-60 61-65 66-70 70+ Age Group Source: Household surveys 1990-2006 PNAD. For methodology to move from cross-sectional to longitudinal cohorts see Robalino et al. (2008). Civil servants and the military are excluded from the chart. 2. SOCIAL INSURANCE, REDISTRIBUTION AND ECONOMIC INCENTIVES The Brazilian social insurance system provides comprehensive coverage of standard risks through the National Social Security Institute (INSS) 28 and the unemployment benefits programs managed by the Ministry of Labor through the Caixa Econômica Federal (CEF). The INSS is an autonomous institution that was created in 1990. It is governed by a tripartite board with representatives from employers and employees. It covers private sector workers and provides old-age, disability and survivorship pensions (RGPS benefits), insurance for work accidents, various transfers related to maternity and sickness leave, as well as non-contributory pensions to the elderly poor and disabled. The Caixa Econômica Federal (CEF) manages the unemployment insurance system and the Length of Service Guarantee Fund (FGTS), which are essentially funded unemployment individual savings accounts. Hence, 60 percent of the taxes collected to finance Unemployment Insurance (UI) and 100 percent of the pay-roll taxes collected to finance the Fundo de Garantia por Tempo de Serviço (FGTS) are transferred to the CEF. The remaining 40 percent of the UI taxes are allocated to the Banco Nacional de Desenvolvimento Econômico e Social (BNDES) that finances entrepreneurs and small companies with subsidized interest rates. Excluding health insurance and FGTS,29 overall expenditures of the SI system are around 7.6 percent of Gross Domestic Product (GDP) (R$174.4 billion) for a covered population of around 44.7 million (48 percent of the labor force). This Report focuses on the income protection system (unemployment 28 For a more detailed overview see Zylberstajn (2008). 29 The section does not deal with social insurance arrangements for civil servants, military and security forces. The health insurance system is an important component of the SI system and should also be expected to influence labor supply decisions. Due to time and resources constraints, however, it is not addressed in this note. 9 insurance and FGTS) and contributory and non-contributory old-age pensions, which constitute the two largest outlets: R$115.2 billion (66 percent of the total). The other programs managed by the Instituto Nacional de Seguridade Social (INSS) are only considered to the extent that they contribute to the fiscal burden of the system. There is evidence, however, that some of these programs do distort incentives, particularly in terms of labor supply decisions.30 This section is organized in three parts. The first part discusses briefly the level of spending of the system, the size of the resulting tax-wedge and potential impacts on labor markets. The second and third parts analyze incentives and the type of redistribution within the pensions and income protection systems. In the case of pensions three types of incentives are analyzed: (i) incentives for retirement over work; (ii) incentives for work over education in young ages; and (iii) incentives for work in the informal vs. formal sector. The focus is on incentives created by current benefit formulas and eligibility conditions – not actual behaviors. In the case of the unemployment insurance system the incentives studied are related to: (i) job search efforts; and, like in the case of pensions, (ii) preferences for formal vs. informal sector jobs. This part is based on an econometric analysis using the Pesquisa Mensual de Emprego (PME) as well as previous studies looking at the impacts of the FGTS. Section 3 then uses a life-cycle behavioral model to look at the interactions between the pensions and income protection system in determining choices between formal and informal sector work, retirement decisions and savings. 2.1. FINANCING MECHANISMS, TAX WEDGE AND EMPLOYMENT Like the majority of countries, Brazil finances social insurance programs trough pay-roll taxes and employees’ contributions. The burden of these pay-roll taxes and contributions is not easy to determine because the rates vary with the level of income of covered individuals and the type of employer and/or economic sector. The typical employer, however, pays between 21 and 23 percent of the wage bill to finance pensions and other assistance programs through the INSS (Banks pay 2.5 percent more) and 8 percent to finance the FGTS. In addition, there are taxes de-linked from the pay-roll that are used to finance UI and some other transfers31 (0.65 of gross revenues in the case of the service sector or 1.65 percent of value added in the case of the industry sector). Finally, there are pay-roll taxes to finance assistance programs outside the INSS and various training and employment generation programs, which add around 3.3 percentage points to the total pay-roll tax.32 Individuals, on the other hand, pay contributions as a function of their level of income; these range between 8 and 11 percent (see Table 1). 30 See for instance Pereira (2008) for an analysis of the distortions created by disability pensions. Bosh et al. (2007) also suggest that universal health insurance may have become an incentive to informality, but research to date is still limited. 31 Forty percent of the revenue from taxes goes to the BNDES. The reminder sixty percent finances UI and one month of salary per year for individuals earning up to 2.7 times of the minimum wage. 32 These additional taxes are allocated as follows: 2.5 percentage points for training and social assistance programs (SENAI- SESI/SENAC/SESC); 0.6 percentage points for the Small Business Program (SEBRAE); and 0.2 percentage points for the Agrarian Reform Programs (INCRA). 10 Table 1: Summary of Pay-Roll Taxes and Social Security Contributions Employer Pay-roll taxes to the INSS 21-23% Pensions and assistance 20%a Workers compensation 1% to 3% (depending on risk) Pay-roll tax to FGTS 8% Pay-roll tax for assistance, training and 5.8% employment generation programs outside INSS and Education Salary Tax on gross revenues or valued-added for UI 0.65 of gross revenues (services) 1.65 of value added (industry) Employeesb Up to R$ 911.70 per month 8% R$ 911.70 to R$ 1,519.50 per month 9% R$ 1,519.51 to R$ 3,38.99 per month 11% Income tax Up to R$12,694 per year 0% Up to R$25,380 per year 15% More than R$25,380 per year 27.5% Average earnings per year R$10,734 Notes: (a) Exceptions are Banks that pay 22%, rural employers who pay 2% on revenues, and labor cooperatives that pay 15%. (b) Self employed and employers contribute with 20% of their labor income. The rate is reduced to 11% when the income is equal to the minimum wage. Small business owners may contribute with 11% of their income, if the yearly business revenue is under R$ 6,000. Source: Various regulations. These taxes mobilize revenues of around R$207 billion (8 percent of GDP); pensions and UI capture more than 70 percent of these. Within the INSS, RGPS benefits account for 85.4 percent of total expenditures. Old-age pensions and disability account for 56 percent of the total, followed by survivorship pensions (20.6 percent), assistance benefits (8.1 percent) and unemployment benefits (6.5 percent). The other transfers, including bonuses and maternity benefits account for less than 10 percent of expenditures (see Table 2). Within the assistance programs, by far, the most important are the “social pensions� for the elderly and disabled, which account for 85 percent of expenditures. The resulting tax-wedge in Brazil (excluding the taxes based on gross revenues or value added), is close to 32 percent, which although average by international standards can be considered high.33 The average tax-wedge in EU15 countries is 44 percent and it reaches 37.8 percent in Eastern Europe and Central Asia. The average for the Middle East and North Africa region is 30 percent. However, several of these countries, more recently Egypt and Turkey, are considering policies to reduce social security contributions and the total tax wedge. At the other extreme, countries such as Mexico, Ireland, Jordan, Korea and Vietnam have tax wedges below 25 percent (see Figure 7). It is also important to notice that if, in Brazil, the 40 percent dismissal fine and the earmarked taxes on revenues and value added were taken into account and expressed as a share of the payroll, the tax-wedge could approximate 37 percent. 33 The tax-wedge is defined as (employers’ costs of labor – employees take home pay) and is expressed as a share of total labor costs. Formally, we have: tw  w  w�e   w  w�l  w�   �e  �l � where w is the gross wage, � is the contribution rate paid by the employer (e) and w  w�e 1  �e the employee (l) and � is the income tax. 11 Table 2: Expenditures in Social Insurance (excluding health) Benefits Amount (R$) % Urban Rural TOTAL 174,421,109,948 100% 130,936,133,364 32,152,196,088 Share of GDP 7.6% RGPS BENEFITS 148,882,664,520 85.4% 117,550,764,696 31,331,899,824 Pensions/Survivors 144,323,912,472 82.7% 113,089,991,292 31,233,921,180 Pensions 97,944,592,716 56.2% 75,422,915,736 22,521,676,980 Age 34,088,512,356 19.5% 13,457,488,884 20,631,023,472 Disability 16,482,424,836 9.4% 14,671,284,396 1,811,140,440 Length of contribution 47,373,655,524 27.2% 47,294,142,456 79,513,068 Survivors benefits 35,980,814,796 20.6% 27,876,569,388 8,104,245,408 Bonus 10,206,341,124 5.9% 9,646,396,392 559,944,732 Health 10,018,752,504 5.7% 9,486,808,464 531,944,040 Accident 69,111,828 0.0% 54,482,004 14,629,824 Reclusion 118,476,792 0.1% 105,105,924 13,370,868 Maternity benefit 186,299,316 0.1% 138,245,256 48,054,060 Others 5,864,520 0.0% 5,864,520 - Workers Compensation 4,558,752,048 2.6% 4,460,773,404 97,978,644 Disability pension 1,193,568,096 0.7% 1,152,364,404 41,203,692 Death pension 920,844,948 0.5% 900,962,760 19,882,188 Health benefit 1,203,292,416 0.7% 1,180,079,808 23,212,608 Accident benefit 1,132,430,772 0.6% 1,118,750,616 13,680,156 Supplemental benefit 108,615,816 0.1% 108,615,816 - ASSISTANCE BENEFITS 14,205,664,932 8.1% 13,385,368,668 820,296,264 LOAS 12,255,113,124 7.0% 12,255,113,124 - Elderly 5,925,709,728 3.4% 5,925,709,728 - Handicapped benefit 6,329,403,396 3.6% 6,329,403,396 - Life-cycle pension 122,648,544 0.1% 122,648,544 - Life-cycle monthly income 1,827,903,264 1.0% 1,007,607,000 820,296,264 Age 530,572,392 0.3% 260,946,684 269,625,708 Disability 1,297,330,860 0.7% 746,660,316 550,670,544 UNEMPLOYEMENT BENEFITS 11,332,780,496 6.5% Unemployment Insurance 11,332,780,496 Unemployment Savings Note: Expenditures on unemployment insurance are only until November 2007. The total for the year is more likely to be around R$12.3 billion (0.7 percent of GDP). Source: INSS and Ministry of Labor. How the current tax-wedge has affected employment levels in Brazil is a question that remains elusive, but it is likely that further increases would have negative impacts on employment. Evidence from other countries suggests that, other things being equal, the higher the tax wedge the lower the level of employment. Effects tend to be more important among low-skilled/low productivity workers, who face a more elastic demand. Estimates for OECD countries, for instance, show that a 10 percentage point increase in the tax wedge reduces labor input by somewhere between 1 and 3 percent of the population of working age.34 Hence, the difference in the tax-wedge could explain around one quarter of the overall difference in the employment rate between the United States and the big three countries of continental Europe (France, Germany and Italy). Another study finds that the observed rise of 14 percentage points in labor taxes in EU between 1965 and 1995 account for a rise in EU unemployment of roughly 4 percentage points.35 In the case of countries in Easter Europe and Central Asia a 10 percentage points increase in the tax wedge could be associated with a reduction of 3 to 6 percentage points in the 34 See Nickell (2003). 35 See Daveri and Tabellini (2000). 12 employment/population ratio.36 The most recent study for Turkey finds that labor costs significantly affect employment levels, but that, on average, higher (lower) social security contributions are more likely to be reflected in lower (higher) wages than in lower (higher) employment levels (i.e., the “pass- through� effect is high). In the case of low-income workers, however, wages are less likely to change and thus the impact on employment levels would be higher.37 Back to the region, the sharp rise in payroll taxes in Colombia—over 10 percent from 41 to 51.5 percent—between 1989 and 1996 may have lead to a decline in formal employment of 4–5 percent.38 Figure 7: Tax Wedge in Brazil and Selected Countries Egypt Hungary Romania EU15 Czech R. Poland Turkey Turkey Denmark Croatia Netherlands Bulgaria Spain Algeria Tunisia Morocco UK Iran Brazil Russian Federation USA Ireland Jordan Lebanon Korea Vietnam Mexico Bahrain 0 10 20 30 40 50 60 70 Source: Authors’ calculations In general, the effect of the “tax wedge� on employment and unemployment levels depends on the structure of the labor market and its institutions. Three factors matter the most. First, how sensitive are the demand and supply of labor to changes in the cost of labor and take home pay respectively.39 The more sensitive, or elastic, the larger is the impact that an increase in labor taxes can have on employment. Second, the institutions that affect the relative bargaining power of employers and employees. In general, the presence of unemployment benefits and minimum wages can add downward rigidity to wages increasing the likelihood that a higher tax-wedge would reduce employment levels. The third factor is the perception that employees have on social security contributions. In principle, these contributions are linked to benefits and therefore should not be considered as a tax by workers. In practice, however, social security programs tend to be redistributive and thus at least part of the contribution rate is a tax. The larger the share of this “implicit tax,� the larger the impact that an increase in the social security contribution rate can have on employment levels (this issue is discussed in the next section). A high tax-wedge can also provide incentives for “informality� by reducing the net earnings gap between the formal and informal sectors or self-employment. This idea is discussed in more detail in the following sections. The point to make here is that for workers who have the choice between working in the formal vs. informal sector or self-employment, other things being equal, reducing net earnings in the formal sector makes it less attractive. Employees are then more likely to collude with employers and avoid 36 See Rutkowski (2007). 37 See Betcherman and Pages (2008). 38 See Kugler and Kugler (2003). 39 Technically this sensitivity is measured by labor demand and labor supply elasticities. These elasticities give the percentage change in the demand (supply) of labor as a percentage change in the cost of labor (take home pay). 13 registering with the social security, choose jobs that do not enroll workers, or become self-employed.40 Low labor productivity in small production units might also “force� evasion if the value of output per worker is below the minimum official cost of labor (minimum wage plus social charges).41 Simulations using matching models suggest, for instance, that pay-roll taxes would reduce incentives to create vacancies in the formal sector thus forcing workers, particularly middle and low productivity workers, into the informal sector. 42 In conclusion, when discussing the financial sustainability of the social insurance system in Brazil, it seems important to consider policies that ensure that the tax-wedge, if not reduced, would at least not increase. From the expenditure side there are interventions that can be considered at each level of the social insurance system and that require coordination. In general, a better use of subsidies which involves better targeting of explicit transfers, the reduction in the size of implicit transfers (particularly for middle and high income workers) and more efficient governance/administration that provides incentives to reduce costs and improve the quality of services. From the revenue side there are also things that the Government could do. Expected revenues will depend essentially on labor force growth, the share of formal employment, and average labor productivity growth. Little can be done at this stage to affect the growth rate of the labor force. But there are policies that could affect labor productivity growth (the topic of Pillar II of the PESW) as well as the share of the labor force employed in the formal sector. Regarding the latter, the main argument is that better incentives can, to a certain degree, increase enrollment in the social security and extend contribution densities. This does not only imply more revenues today, but also lower subsidies related to minimum pension guarantees for future retirees. Indeed, individuals with some savings capacity would have contributed to finance at least part of the targeted minimum level of income for old-age. These issues are discussed in the rest of this section. 2.2. PENSIONS The government of Brazil and the World Bank recently completed a comprehensive assessment of the pension system,43 both the mandatory and voluntary schemes. In terms of the mandatory schemes and in particular the Regime Geral de Proteção Social (RGPS), the report pointed to very high levels of expenditures relative to countries with the same level of income; a deteriorating cash-balance and mounting unfunded pension liabilities; and, in general, benefit formulas and eligibility conditions that are too liberal and generous given contributions. The report presented various recommendations including parametric reforms that would unify retirement ages, lower survivorship and disability pensions, and the review of the minimum pension guarantee aiming, among others, to de-link it from the minimum wage. The purpose of this section is to extend/deepen the analysis of incentives and redistribution within the RGPS, mainly looking at: (i) how current benefit formulas and eligibility conditions reward or penalize behaviors related to retirement, entry into the labor force, and sector choice (formal vs. informal); and (ii) how the system redistributes implicit taxes and subsidies across plan members. The last part of the Report will then look at how changes in benefit formulas and eligibility conditions could affect these behaviors and therefore program costs. 40 Some studies have attempted to estimate the semi-elasticity of self-employment with respect to a change in relative formal sector earnings, which would range between 0.03 (Maloney, 1998) to 0.05 (Loayza and Rigolini, 2006). See also discussion in Chapter 4 of Perry et al. (2007). 41 This is an issue that is being analyzed in the case of Brazil. There is evidence for other countries, however, showing that a sizable share of firms in the manufacturing sector have productivity levels below the official cost of labor (see World Bank, 2007b, 2007c). 42 See Albrecht, Navarro, and Vroman (2006). 43 World Bank (2007a). 14 The mandate of the system The Brazilian pension system has a unique 3 tier arrangement when it comes to eligibility conditions for retirement. There are three regimes: (i) retirement base on a minimum age (53M/48W) and a minimum number of years of contributions (30M/25W); (ii) retirement base on a number of years of contributions (35M/30W) and no minimum age; and (iii) retirement based on age (65M/60W) and a minimum number of years of contributions (15M/15W).44 Benefit formulas vary substantially by regime.45 The first regime pays a so called Proportional Length of Contribution (PLOC) Pension, the second a Length of Contribution (LOC) Pension, and the third an Aging Pension. In all cases, the pension system guarantees a top-up so that the minimum pension (Piso Previdenciário) is equal to the minimum wage.46 Otherwise, pensions above this minimum are indexed by inflation – which creates a flattening of the distribution of pensions. To better understand the interplay between the three regimes it is useful to look at the gross replacement rate that would be received by representative workers at various retirement ages (see Figure 8). The example used here refers to two male individuals who join the pension system at age 25 and contribute continuously up to retirement age. The difference between the two individuals is their level of income: the first earns the economy wide average wage while the second earns 50 percent of the average wage. The earliest age at which these two individuals can retire is 54; by then they would have accumulated 30 years of contributions and would be eligible for the _Proportional Length of Contribution (PLOC) pension. By age 59 they would be eligible for the LOC pension (since they would have accumulated 35 years of contributions) and by age 65 they would also be eligible for the aging pension (the aging pension is represented in the figure by the discontinuous line). What is important to observe in the figure is that: (1) for the average worker the pension system offers considerably lower replacement rates through the PLOC pension; (2) that low income workers retiring early receive higher replacement rates than middle or high income workers due the existence of the minimum pension; (3) that replacement rates under the LOC pension growth “rapidly� with each additional year of contribution and can even surpass 100%; and (4) that the aging pension is considerably lower than the LOC pension (particularly for middle/high income individuals who are not eligible for the minimum pension guarantee) and therefore would only be chosen by those who do not meet the vesting requirement for the former. 44 Rural workers (i.e., individuals performing an activity that is considered as rural work) only retire under the aging pension without any minimum requirement on the vesting period. After the 1988 constitution, however, rural workers are also eligible to the LOC (see Law 8212 and 8213). 45 See the Annex for a description of the benefit formulas. 46 Contrary to most countries, adjustments to the minimum wage in Brazil are based on explicit rules that minimize the discretion of policymakers. The minimum wage is adjusted once a year and the adjustment rate is determined by two components: the inflation rate plus the GDP rate of increase observed two years before. This police will be in place until year 2011. After that, a new administration will define the policy. Since March 1st 2008, the minimum wage has been fixed at R$415 per month. 15 Figure 8: Gross Replacement Rates by Retirement Age 120 Individual earning 100 50% of the average Gross Replacement Rate 80 60 40 Individual with average earnings 20 Not eligible Proportional LoC LoC pension Aging pension 0 50 55 60 65 70 Retirement Age Source: Authors’ calculations based on current legislation. Retirement decisions There are no standards or best practices in terms of how countries should set the retirement age in a mandatory pension system. In principle, the mandatory system should offer flexibility to individuals regarding retirement as long as: (i) the pension they obtain is sufficient to finance an adequate standard of living; and (ii) contributions made while active are enough to cover the resulting liabilities. The first condition is related to the core objective of the pension system, which is to ensure a certain level of forced savings to prevent that some individuals under save while young (assuming some degree of “myopia�). The second condition is necessary to ensure the financial sustainability of the system. Clearly, some individuals might have low or no savings capacity and would need some form of subsidy to prevent poverty during old age. Qualifying conditions in this case would need to be more stringent, for instance, meeting a minimum retirement age and some form means test. In practice, however, retirement ages vary widely across countries and, in the case of most DB-PAYG systems, tend to be disconnected from the level of the pension and the contribution rate (see Figure 9). Moreover, DB-PAYG pension systems tend to implicitly reward early retirement and/or penalize delayed retirement. This happens in various ways but, in general, it is due to high replacement rates at early ages that grow only slowly as the retirement age increases or that are capped. The results are lower contribution flows and higher pension flows that compromise the financial sustainability of the systems. This section briefly assesses the situation in the case of Brazil. 16 Figure 9: Retirement Ages and Life Expectancies Around the World F ranc e Canada 22.5 Belgium Aus tr ia Expected duration of retirem ent New Z ealand Finland Netherlands K orea J apan 20 S witzer lan d S we den C zec h R A us tralia OE CD average T urk ey S lov ak R Iceland H ungary Italy P ortug al Nor way 17.5 UK US Spa in Ireland Gr eece Ger m any P oland Denm ark Luxem bo urg M ex ico 15 55 60 65 67 Pension eligibility age Source: World Bank pensions database. In Brazil, the majority of workers retire with the aging pension, which indicates that workers cannot or prefer not to complete the vesting period for the LOC. This can be seen in the first two panels of Figure 10 which give the distribution of new retirees by age and type of pension in rural and urban areas. Not surprisingly, very few people retire with the PLOC which can be received earlier but with high penalties. In rural areas individuals can only retire with the aging pension and 60 percent do it before age 65 (20 percent before age 60 and 40 percent between ages 60 and 65). Then, in urban areas, around 35 percent of new retirees retire before age 60 under the LOC. Close to half of the new retirees retire with the aging pension after age 60. Because, as shown above, this pension is considerably lower than the LOC pension, this raises the question of why workers do not “go the extra mile� and complete the vesting period for the LOC? The evidence would suggest that many workers with incomes below the median are choosing to retire with the aging pension. Indeed, the third panel of Figure 10 graphs the distribution of the vesting period among new retirees for males and females. There are two peaks. The first occurs at ages 35 (males) and 30 (females), which correspond to the eligibility threshold for the LOC pension. More than half of the new retirees do not reach this threshold. The other peak is around 32 years for males and 26 years for females – relatively close to the LOC threshold. Given the large difference between the LOC and the aging pension it is puzzling that workers are not willing to delay retirement and retire under the LOC. One interpretation would be that, given labor market conditions, it is not easy to do so. The other would be that workers prefer to retire with the aging pension which requires fewer contributions. This would be the rational choice if under the LOC or aging pension individuals would retire with a pension close to the minimum. It would seem, therefore, that those who retire with the aging pension are mainly workers with incomes equal or below the median. And, in fact, 35 percent of new retirees are retiring with the minimum pension and half with a pension below two times the minimum (see fourth panel of Figure 10). The analysis of incentives presented below supports this view. 17 Figure 10: Distribution of Type of Pension, Vesting Period and Retirement Probabilities Urban Rural 40% 45% PLOC PLOC 35% LOC 40% LOC Aging 35% Aging Percentage of New Pensions Percentage of New Pensions 30% 30% 25% 25% 20% 20% 15% 15% 10% 10% 5% 5% 0% 0% <50 50-54 55-59 60-64 65+ <50 50-54 55-59 60-64 65+ Age Group Age Group National 35% 100% Women Men Urban 90% Rural 30% 80% 25% 70% 20% 60% 50% ` 15% 40% 10% 30% 20% 5% 10% 0% 0% 15 20 25 30 35 40 [0-1] ]1,2] ]2,3] ]3,4] ]4,5] ]5,+[ Vesting Period at Retirement (years) Pension Level (proportion minimum wage) Source: Authors’ calculations based on data from the Ministry of the Previdencia Social. The standard indicator to measure incentives for retirement, the change in pension wealth, suggests that middle and high income workers can gain from delaying retirement, but workers with incomes equal or below the median face important losses. The pension wealth is defined as the expected present value of future pensions at the time of retirement. The change in the pension wealth between two consecutive ages (say age 60 and 61) gives an idea of how much individuals would gain if they wait one year to retire.47 The results of these calculations are presented in Figure 11 for four representative individuals.48 The change in pension wealth is expressed as a share of pre-retirement earnings. During the period of eligibility for the proportional LOC individuals are in general penalized if they delay retirement (i.e., the increase in the pension wealth does not compensate for the cost of waiting). There are very strong incentives, however, to wait to retire until eligibility for the LOC pension (including for low income workers). Afterwards, for middle and high income workers the gains from delaying retirement are equivalent to 3-4 months of salary (25 percent-30 percent of yearly earnings) per year. For an individual earning 50 percent of average earnings, however, delaying retirement has an important cost – close to 6 months of salaries per year. This is in essence because before age 65 this individual 47 When computing the change in pension wealth between age a and age a+1, the pension wealth at age a+1 is discounted to take into account that the individual has to wait 12 months to receive it. The contributions paid during those twelve months are also subtracted. 48 The four cases refer to male workers who enroll in the pension system at age 25 and contribute continuously to retirement age. Workers differ on the initial level of income and the growth rate of wages. Three workers face a growth rate of 3 percent real per year and earn respectively 50 percent of average earnings, average earnings, and 2 times average earnings. The fourth worker also earns average earnings, but faces a growth rate of wages of 4 percent per year. 18 would most likely retire with the minimum pension. Thus, there is no increase in pension wealth from one year to the next to compensate for the payment of contributions. Only after age 65, when the contributory pension goes above the minimum because of the increase in the vesting period (40 years for the full career worker enrolling at age 25), would low income workers receive the same rewards for delaying retirement as middle and high income workers. Clearly, a worker that has a contribution density below 1 (the most likely case) would still be on the minimum pension and would therefore face no incentive to delay retirement, unless, as shown above, the individual decides to stop contributing and retires under the aging pension with less than 35 (males) or 30 (females) years of contributions. But in that case, why not to retire with PLOC pension and still claim the minimum. As discussed in Section 3, this can be explained by the interactions of the pension system with other programs, in particular UI, that provide incentives to delay retirement. The data thus suggest that there are two different forces at play: low income workers facing low incentives to contribute up to the LOC and preferring to retire under the aging pension; and middle and high income workers facing incentives to delay retirement but not being willing to do so. Workers with incomes equal or below the median are likely to retire with the aging pension and short vesting periods, while iterating between formal and informal sector jobs (see next section). Among middle and high income workers, on the other hand, very few would retire with the aging pension. Most will thus accumulate the minimum vesting period for the LOC. Their retirement decision then depends on their preferences over future and current consumption, leisure, and risks; the level of the pension; and the availability of other sources of income during old-age. It would seem that even in the presence of rewards many of these individuals might not be willing to delay retirement beyond a certain age, which is close to 60 years. Figure 11: Change in Pension Wealth for Each Additional Year of Contribution Marginal change in pension wealth (proportion earnings). 200% average earnings 150% 2 * average earnings average earnings (+1pp growth in salaries) 100% 50% 0% 54 56 58 60 62 64 66 68 -50% 0.5 * average earnings -100% Source: Authors’ calculations. An important problem is that, regardless of the income of the individual and the retirement age, internal rates of return on contributions appear unsustainable. Internal rates of return can be interpreted as the implicit interest rate that the pension system pays on contributions. Indeed, the pension system can be associated to a bank account: while active individuals “deposit� contributions, when they retire they “withdraw� pensions. It is thus possible to compute what is the interest rate that the system is paying on the contributions. It turns out that in a pay-as-you-go system the sustainable rate of return on contributions can be approximated by the long-term growth rate of the covered wage bill (or the long- 19 term growth rate of the economy).49 Real rates of return above 3 to 4 percent are in general not sustainable, in the sense that they create pension liabilities that cannot be financed out of future contributions (see Figure 12). Future generations, in this case, must face the burden through higher explicit or implicit taxes and/or lower expenditures on public goods. In the RGPS workers earning the average wage and retiring at age 60 under the LOC pension would receive a real rate of return on contributions of 4 percent per year. Rates of return would be in general higher for low-income workers, due to the minimum pension, and lower for high income workers given that they pay higher contributions. In fact, for individuals earning twice the minimum wage and retiring with the proportional LOC pension, rates of return could be below sustainable levels. This would make the Brazilian pension system very redistributive. At the same time, as discussed above, there are few individuals who retire with the PLOC pension. Most individuals retire around age 60 when rates of return are above sustainable levels. For middle and high income workers these rates of return increase as a function of the retirement age (as replacement rates increase) while for low income workers they decrease (since the replacement rate remains constant). Figure 12: Internal Rates of Return on Contributions by Retirement Age and Level of Income 7 0.5 * average earnings 6 average earnings 5 4 3 2 * average earnings 2 Sutainable level average earnings (+1pp growth in salaries) 1 0 54 56 58 60 62 64 66 68 70 Source: Authors’ calculations. In conclusion the RGPS faces two challenges when it comes to the regulation of retirement: (i) to realign replacement rates with retirement ages, given the contribution rate and life expectancy; and (ii) to improve incentives to contribute for individuals with earnings equal or below the median. Specific policy recommendations are provided in the last section of the Report. Some of the principles, however, can already be outlined. For middle and high income workers the current implicit subsidy to encourage delayed retirement does not seem to be working. Moreover, in a country like Brazil with a young labor force that is growing fast (see Section 1) there is little rationale for this subsidy. Two things are important then: (i) to ensure that when individuals retire a certain level of income replacement can be secured (one that would most likely fall as income rises); and (ii) to ensure that regardless of the retirement age the Internal Rate of Return (IRR) is sustainable. For individuals with limited savings capacity it might be necessary to have in place some form of subsidy to ensure an adequate level of income during old age. 49 The calculation of the correct sustainable internal rate of return on contributions is in reality more complex (see Robalino and Bodor, 2008). 20 But, the form of this subsidy matters. In its current form the minimum pension guarantee reduces incentives to contribute (see next section) and provides incentives for early retirement. Sector choice and entry into the labor force One of the important challenges facing pension systems in middle income countries are low coverage rates and sparse contribution densities. These imply not only lost revenues from taxes but also higher future expenditures as some workers might not be able to build sufficient savings for old-age and then require some form of minimum pension guarantee. As shown in the previous section, in Brazil, half of new retirees do not meet the minimum vesting period to retire under the LOC (35 years for male and 30 for women). But the country is not unique in this respect. In middle income countries observed vesting periods are below 30 years of contributions. In the region, recent studies in Chile show that up to 60 percent of new retirees end-up with the minimum pension, in part due to short career histories.50 In Uruguay, 35 percent of workers are not able to meet the minimum vesting period for retirement.51 The problem is that individuals who have access to the social security52 tend to transit in and out during their active lives.53 Part of these transitions depends on factors outside the control of workers (e.g., firms that close down, introduction of new technologies that change the composition of the labor force, economic restructuring). As discussed in the last section, however, individuals also make choices that influence these transitions. Indeed, because it is very difficult to fully enforce the mandate to contribute to the pension system (social security in general) some individuals might choose to have incomplete contribution densities, based on an assessment of expected costs and benefits. In general, decisions about how long to contribute can be influenced by relative net earnings between the formal and informal sectors, expectations about future pension benefits (which depend on the form of the minimum pension guarantee), and perceptions about the capacity of the system to deliver on its promises. Because enrollment in the pension system is usually bundled with enrollment in other social insurance/transfer programs, the expected benefits and utility delivered from these other programs also plays a role.54 In the case of the pension system, whether individuals prefer jobs covered by the social security to jobs that are not covered55 will depend on the productivity of labor in the informal sector relative to the formal sector; and pay-roll taxes & social security contributions. If the productivity of the formal sector is very high, then even in the presence of social security contributions individuals are likely to prefer formal sector jobs; net earnings there would always be higher. But if the difference in productivity is not able to compensate for the tax wedge, then workers might prefer jobs that are not covered by the social security. Thus, for a given level of productivity in the two sectors, the higher the tax-wedge, the higher the likelihood of informality is. This idea is captured in Figure 13, where the downward slopping line provides the combinations between the productivity of labor in the informal sector and the contribution rate paid by the employee that generate the same level of net earnings. It is assumed that productivity in the formal sector can be measured by the marginal cost of labor to the employer: wages plus payroll taxes. Productivity in the informal sector (vertical dimension) is then expressed as a fraction of the productivity of the formal sector. Thus, the lower the level of productivity in the informal sector, the higher the social security contribution can be without providing incentive for informal sector work. In general, when the combination of the productivity in the informal sector and the social security contribution rate falls below 50 Valdez (forthcoming). 51 See Forteza et al. (2008). 52 Large groups of individuals can also be permanently outside. Current statistics make this difficult to disentangle, but the available evidence suggests that these are low income workers with very limited or no savings capacity. 53 Ibid 29. 54 See Perry et al. (2007). 55 As a reminder, in this note, jobs covered by the social security are considered formal sector jobs and imply that the worker has a carteira. 21 the line, workers might prefer formal sector jobs. But if the combination is above the line, given a relatively high level of productivity in the informal sector and/or a high contribution, then workers might choose informal sector jobs depending on what they receive in exchange for the forgone revenue. In the case of the pension system, the decision would depend on the expected internal rate of return on contributions. 56 Figure 13: Informal Sector Productivity, Payroll Taxes, and Incentives for Informality 1 Productivity in the Informal Sector Relative to. 0.9 0.8 Workers might take informal sector jobs 0.7 the Formal Sector 0.6 0.5 0.4 0.3 Workers prefer formal 0.2 8% sector jobs 0.1 0 0 0.2 0.4 0.6 0.8 1 Employee Social Security Contribution Note: The slope of the line depends on the employer’s pay-roll-tax; around 28% in the case of Brazil. Along the line, productivity in the informal sector is equal to the tax wedge. The equation of the line is Productivity=1/(1+pay-roll tax ) – Social Security Contribution/(1+pay-roll tax ). Source: Authors’ calculations. In the RGPS the Expected Net Life-time Wealth varies little as a function of the contribution density indicating that, in general, individuals have weak incentives to contribute. The ENLW is defined as the present value of earnings from work and pensions net of taxes and social security contributions. 57 Its value depends, in part, on the time that workers spend contributing to the social security (which influences the value of future pensions), as well as decisions about whether to enter early in the labor force or delay entry and invest in education. Figure 14 presents simulations of the ENLW as a function of contribution densities, under various assumptions regarding the productivity of the informal sector and the percentage increase in initial wages that could be achieved for each year that entry into the labor force is delayed. It shows that, in general, ENLW do not increase with an increase in the contribution density. For the median worker there are no incentives to contribute even when the productivity of the informal sector is low. For a level of productivity equal to 80 percent of that in the formal sector,58 somebody entering the labor market at age 20 would face a flat ENLW if retiring at age 65. Additional contributions would bring nothing (see first panel Figure 14). If retiring at age 60, either the individual contributes at least 70 percent of the time to be eligible for the minimum pension or it evades all the time. In any case, the ENLW would not reach the ENLW that he/she would receive if retiring at age 65. If the productivity 56 See Valdez (forthcoming) for a more elaborated presentation of this framework and its rationale and for an application to the analysis of the incentive effects of the minimum pension guarantee in Chile. 57 Because in practice individuals are more likely making decisions in order to maximize the present value of consumption – not wealth – the analysis here is only meant to provide some preliminary insights about the incentives provided by current benefit formulas and eligibility conditions. The last section will look at actual behaviors in a more comprehensive framework where individuals also make simultaneous choices about savings and retirement. 58 Assuming that average productivity can be approximated by average earnings, in Brazil, the average productivity of the informal sector would be around 70 percent of the productivity of the formal sector (see Section 1). 22 in the informal sector is as low as 60 percent that of the formal sector, the ENLW would increase with the contribution density but mildly (see last panel in Figure 14). Grosso-modo, the ENLW would increase by the equivalent of 6 average wages for each 10 percentage points increase in the contribution density. The main issue is the minimum pension guarantee that in its current form implies a 100 percent marginal tax on contributions after eligibility. In all charts in Figure 13, the ENLW without the minimum pension is represented by the dotted lines. Unequivocally, when the minimum pension is removed, incentives improve, in the sense that the ENLW increases when the contribution density increases. This is because the minimum pension is implemented though a top-up for individuals who end-up with a contributory pension below the minimum. The implication is that the top-up is reduced by one R$ for each one R$ increase in the value of the contributory pension. In essence, the top-up faces a 100 percent marginal tax. Each additional year of contributions increases the contributory pension but reduces the top-up by the same amount. The last chapter of this Report will propose alternative mechanisms to guarantee a minimum level of income during old age while preserving incentives to contribute. In terms of the decision to enter the labor force the pension system does not seem to play any significant role. Individuals delaying entry into the labor force to go to school or college can achieve a higher or lower ENLW depending on how additional years of education affect earnings. With high returns to education individuals can generate a higher ENWL, but would still face very week incentives to contribute to the pension system at the margin. If retiring at age 65 the incentive would be to contribute 40 percent of the time to become eligible for the minimum pension (see second panel Figure 13). To increase the ENLW after that the individual would need to contribute 80 percent of the time to become eligible for the LOC; this would add around 2 to 3 years of earnings. Nonetheless, an 80 percent contribution density is high and individuals might prefer to take a more conservative stance. With low returns to education delaying entry into the labor market does not pay off, but this is for the most part unrelated to current benefit formulas and eligibility conditions in the pension system. Somebody entering the labor force at age 30 would face a large loss in the ENLW mainly given forgone revenues while studying that are not compensated by higher earnings while working. If retiring at age 65 incentives would be to contribute 40 percent of the time to become eligible for the minimum pension. To retire at age 60 the contribution density would have to be of at least 90 percent to make a difference (see third panel Figure 13). 23 Figure 14: Expected Net Life-Time Wealth for Low Income Individuals Entry Age 20 Entry Age 25 (high returns to education) 30 30 Retirement age 65 (without minimum) Retirement age 65 Expected Life Time Net Wealth Expected Life Time Net Wealth 25 (without minimum) 25 20 20 Retirement age 60 (without mininum) 15 15 (withouth pension) Retirement age 60 10 10 5 5 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Contribution Density (fraction of a year) Contribution Density (fraction of a year) Entry age 30 (low returns to education) Entry age 20 – Productivity Informal = 0.6% 30 30 Retirement age 65 Expected Life Time Net Wealth Expected Life Time Net Wealth 25 25 (without minimum) 20 20 Retirement age 65 (without minimum) 15 15 (without mininum) 10 10 Retirement age 60 Retirement age 60 (withouth pension) 5 5 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Contribution Density (fraction of a year) Contribution Density (fraction of a year) Note: The Expected Net Life-time Wealth (ENLW) is expressed as a share of average earnings. Simulations are for the median individual who earns 50 percent of the average wage. Except in the last panel, productivity in the informal sector is assumed to be 80 percent of the productivity in the formal sector. The scenario “high returns to education� assumes that initial real earnings increase by 10 percent with each year of education. The scenario “low returns to education� assumes that earnings increase only by one percent. Source: Authors’ calculations. Redistribution This sub-section deals with the systematic redistribution of income within the pension system. Systematic means that, consistently, some population groups transfer income to others. It differs from the non- systematic redistribution of income that takes place in any insurance arrangement (including in the private sector) through risk pooling. Income redistribution is an important function of the pension system (and the social security system as a whole) in most countries. By in large, the type of redistribution depends on benefit formulas and eligibility conditions. Defined benefit arrangements are usually redistributive (albeit sometimes regressive), defined contribution arrangements are not. In both cases, however, the type of financing mechanism pay-as-you-go or funded plays no role. Or in other words, redistribution can exist within pay- as-you-go or funded systems. The analysis developed here is based on the calculation of implicit taxes and subsidies for individuals with different levels of earnings. A tax (or subsidy) is computed as the difference between the current contribution rate and the contribution rate required to finance the annuity implicit in the pension benefit. A positive value indicates that the individual is paying more than what it should for a given level of benefit. A negative value indicates the individual is paying less and therefore is receiving a transfer from other plan members. The results are presented in Figure 15. The left panels graph relative subsidies/taxes 24 (i.e., expressed as a share of individual earnings), while the right panels graph absolute subsidies/taxes (i.e., expressed as a share of average earnings).59 The analysis shows that full-career workers with earnings below the median receive subsidies ranging between 5 and 35 percent of individual wages depending on when they retire. This subsidy is linked to the minimum pension guarantee. As seen in the previous section, the earlier the individual retires (the shorter the contribution period) the higher the share of the minimum pension that the individual receives, the higher its internal rate of return on contributions, and the higher the subsidy. As the retirement age is delayed, the value of the top-up diminishes and so does the subsidy (see first panel of Figure 15). Because few individuals retire before age 60 and, as discussed above, many tend to retire with the aging pension, the effective subsidy is more likely to be in the 5-10 percent range. Middle and high income full-career workers receive subsidies as well as long as they do not retire with the proportional Length of Contribution Pension (LOC) pension. Indeed, as discussed in the previous section, the proportional LOC imposes heavy penalties on early retirement and implicit taxes in this case can range between 5 and 10 percent. This induces the majority of workers to retire after age 60. In this case, the size of the subsidy is more or less constant across retirement ages and ranges between 2.5 and 5 percent (see first panel of Figure 15). The only exception in the example is an individual with earnings growing at 4 percent per year. This individual does not receive subsidies when delaying retirement and would more or less break even. For partial career workers subsidies can increase for median and low-income workers and decrease for high income workers. For instance, for the median worker with a contribution density of 70 percent, these subsidies can range between 20 and 40 percent of individual earnings – even when retiring at age 65 or above. This would seem to be the most common case (see third panel of Figure 14). For high income workers, on the other hand, short contribution histories mean no access to the LOC pension. This implies retiring either with the PLOC, the aging pension, or the minimum pension. In all cases, it would involve a large reduction in benefits relative to contributions and thus lower subsidies/higher taxes. When expressed in absolute terms, however, subsidies can be higher for middle and high income workers than for median and low income workers, depending on the retirement age and career histories. Having information about subsidies as a share of individual earnings is useful, mainly to determine the part of the social security contribution that can be considered a tax or a subsidy. Ultimately, however, policymakers care about the absolute value of the subsidy (tax) that each individual would receive (pay). These absolute subsidies/taxes expressed as a share of average earnings are presented in the two right panels of Figure 15. The patterns are more or less preserved but in the case of full-career workers, absolute subsidies for low income workers who retire at or after age 65 are lower than the subsidies of high income workers. The main conclusion from the analysis is that while redistribution is designed to be progressive, in practice it might not always be. The problem is that redistribution is opaque as much depends on earnings and career histories. If low income workers are more likely to retire with the minimum pension at age 65 or above, and middle and high income workers retire with the LOC, subsidies could be higher for the latter depending on vesting periods. Low income workers with relatively long vesting periods would be the ones receiving the smallest implicit subsidies – and as discussed above this provides incentives to have sparse contribution densities. What seems clear, however, is that in all cases the large majority of plan members might be receiving implicit subsidies regardless of income levels. This means that redistribution is not being financed within the system but through intergenerational transfers. 59 Clearly, the results depend on the relation between the discount rate and the growth rate of wages. The assumption here is that real wages grow at 3 percent per year while the real discount rate is 4 percent per year – closer to a market rate. The main messages from the analysis, however, are unlikely to change if these rates change within reasonable boundaries. 25 Figure 15: Relative and Absolute Implicit Subsidies Full career 40% 40% 35% age=30; w=0.5; g=3% 35% 30% 30% Subsidiy (percent of earnings) Subsidiy (percent of earnings) 25% age=30; w=1; g=3% 25% age=30; w=0.5; g=3%; 20% age=25; w=1; g=3% 20% d=1 age=25; w=2; g=3%; d=1 15% age=25; w=2; g=3% 15% age=30; w=1; g=3%; d=1 age=25; w=1; g=3%; d=1 10% 10% 5% 5% 0% 0% 55 57 59 61 63 65 67 69 55 57 59 61 63 65 67 69 -5% -5% age=25; w=1; g=4% age=25; w=1; g=4% -10% -10% Retirement Age Retirement Age Partial career (contributes 70 percent of the time) 40% age=30; w=0.5; g=3% 40% 30% 30% age=30; w=0.5; g=3% Subsidiy (percent of earnings) Subsidiy (percent of earnings) 20% 20% age=30; w=1; g=3% age=30; w=1; g=3% age=25; w=1; g=3% 10% age=25; w=1; g=3% 10% ` ` 0% 0% 65 66 67 68 69 70 65 66 67 68 69 70 ` -10% -10% age=25; w=1; g=4% age=25; w=1; g=4% -20% age=25; w=2; g=3% -20% age=25; w=2; g=3% -30% -30% Retirement Age Retirement Age Note: In the left panels taxes/subsidies are expressed as a share of individual earnings. In the right panels, taxes/subsidies are expressed as a share of average earnings. Source: Authors’ calculations based on current legislation. 2.3. INCOME PROTECTION SYSTEM FOR WORKERS Brazil has been one of the international leaders in the reform of income protection system for workers. While the majority of middle income countries in the region and the world still base their income protection system on severance pay arrangements, starting almost 50 years ago Brazil implemented a system that now combines unemployment savings accounts (created in 1967) with unemployment insurance (created in 1986). While savings and risk pooling can still have certain unintended consequences that are discussed below, they tend to be a superior option to traditional severance pay. Indeed, international experiences show that the latter, by focusing on job as opposed to income protection, negatively affects the mobility of the labor force, limits firms capacity to restructure, can reduce incentives to hire, and ultimately impact labor productivity growth and employment levels.60 Thus, by moving away from severance pay, Brazil has added flexibility to the labor market while implementing two programs that help formal sector workers to smooth consumption when transiting between jobs. 60 The results of the empirical literature are reviewed in Holzmann (2005). Among the early studies, Lazear (1990) finds that severance pay increases unemployment and reduces both employment and labor force participation (in their “update� of his findings, Addison and Teixeira (2003) confirm the first finding, but cast some doubt on the others). Fallon and Lucas (1991) show that strengthening of job security regulations led to a strong decline of employment in India and Zimbabwe. More recent studies confirming the link between job security and lower employment levels include Haffner et al. (2001), for OECD countries, Heckman and Pages (2000), for OECD and Latin American countries, Besley and Burgess (2002) for India, and Haltiwanger, Scarpetta and Vodopivec (2003) for OECD and transition countries. See also Davis and Haltiwanger, 1999, for a survey of the effects of job reallocation on aggregate productivity growth. 26 This section analyses the current system and discusses some areas where some improvements in design could be considered to generate better incentives and redistribution. The section starts by reviewing the mandate of the income protection system in terms of income replacement and the duration of benefits. It then looks at the effects of the system on incentives to work, sector choice and turnover. Although not at the core of the Terms of Reference (TOR) for this Report, the section also discusses some issues related to financial sustainability and the management of FGTS reserves. The mandate of the system Formal sector workers who loose their jobs involuntarily can have access to the balance in their unemployment savings accounts and unemployment benefits. To be eligible for unemployment benefits workers need to have held a formal sector job (trabalho with carteira) for at least 6 months in the previous 36 months period to the start of the unemployment spell. The duration of the benefit ranges between 3 and 5 months depending on the number of contributions. With 6 to 11 months workers receive 3 months of benefits, with 12 to 23 they receive 4 months, and with 24 to 36 they receive 5 months. The benefit itself depends on earnings and ranges between one minimum wage (R$415 since March 2008) and a maximum of R$710 per month (around 80 percent of average earnings). At the same time, workers receive a lump sum equal to the balance accumulated in their individual accounts while working in their last job plus the value of a “dismissal fine� equivalent to 40 percent of the FGTS capital. As previously mentioned the FGTS accumulations are financed with an 8 percent contribution rate61 that over a period of 12 months yields a capital more or less equal to one month of salary. Overall, the replacement rates offered by the UI system range between 40 and 100 percent depending on the level of income. The benefit formula ensures that replacement rates are higher for low than for high income workers. Hence, an individual earning the minimum wage receives a 100 percent replacement. At the other extreme, individuals earning 2 times the average wage receive a 40 percent replacement rate. A worker with average earnings has a replacement rate that is closer to 70 percent and the median worker a replacement rate of around 85 percent (see left panel of Figure 16). Thus, relative to other countries which have implemented UI systems, Brazil replacement rates would be more on the high end (see left panel Figure 17).62 Taking both UI and FGTS together, the median worker can finance between 3.5 and 8 months of salaries depending on the number of months of contributions. The number of months of salaries that a worker receives from the UI system during a given spell is given by the replacement rate times the maximum duration of the benefit – which depends on the number of contributions. A worker earning the minimum wage can receive between 3 and 5 months of salaries, the median worker between 2.5 and 4, and a worker earning 2 times the average wages only one to two months. On the other hand, the number of months of salaries that a worker can accumulate through the FGTS system is independent from their level of income. Assuming no real growth in wages during the short term and a 4 percent real interest rate per year, after 12 months of contributions workers can have a capital equivalent to one month of salaries. This relationship is linear in the short-term, so after 36 months of continuous contributions workers can accumulate a capital equivalent to 3 months of salary. The dismissal fine that the employer pays to the employee (40 percent of the capital) would then add around 1.7 weeks of benefits per month. Thus, taking both UI and FGTS together accumulations can range between 1.8 and 6.2 months for workers earning 2 time the average wage, up to 4 to 9 months for those earnings the minimum wage (see right 61 Among countries that have mandatory unemployment insurance savings account the contribution rate varies between 3 percent in Chile (where system is linked to a solidarity fund), to 13 percent in Uruguay. Colombia mandates a 9.5 percent contribution, and Argentina, Ecuador, and Peru charge like Brazil 8 percent. 62 It should be added that Brazilian workers are entitled to a 30-day advance notice. As a general practice, employers prefer to pay the additional one-month salary rather than have somebody already dismissed working in the same place with the other employees. Thus, one may add one additional salary to the figures presented in this paragraph. 27 panel Figure 16). In general, the duration of unemployment insurance benefits in Brazil is above the average for countries with similar levels of income (see right panel Figure 17). Figure 16: Replacement Rates and Accumulations from UI and FGTS 120% 9 0.5 times average wage 8 100% Replacement rate (% earnings) Average wage Number of months of salaries 7 80% 6 5 60% 2 times average wage 4 40% 3 2 20% 1 Accumulations related to FGTS 0% 0 0.4 1 1.6 6 11 16 21 26 31 36 Level of Earnings (proprotion average wage) Number of months of contributions Source: Authors’ calculations based on current legislation. Figure 17: Replacement Rates and Benefit Duration in Selected Countries 90 Lower-middle Upper-middle High income 70 Lower-middle Upper-middle High income 80 income income income income 60 Replacement rate (% wage) 70 Duration (months) 50 60 50 40 40 30 30 20 20 10 10 0 0 So Hun lic St ) Uk rica rtu s es Ca ela h Br a Ve urk a ge ne ce Af ia ne ey Gr ce h ry er n Fr da en S al So R nia Un (V ain pu il m t Ne J ly ce Af ia ge ne h Br a ne ey Uk rica Un (Vo ain Ca ela Gr ce er an rtu s So Hun ic es h ry T ea Fr da pu il m t Ne J ly en S al So Ru nia St ) d nt. Ru gyp Ru gyp Po and d nt. T re in Re az in Po nd th apa Re az g ut uss Ita ut ss g Ita bl ut ga b at ut ga ee an ee Ar rai Ve urk na an at ap r Ar rai na zu nt zu Ko ite olu a p p nt a Ko ite lu la E E l h h th ec ec ed ed Cz Cz Sw Sw Source: Vodopivec (2004) and authors’ calculations for Brazil and MENA. Incentives to work and sector choice The literature on the effects of unemployment insurance on labor markets is quite rich at the international level and, in general, shows that increasing “generosity� has negative effects on the unemployment rate and the length of unemployment spells.63 In principle, a UI system can improve search efficiency leading to shorter unemployment spells. This tends to be more a function of placement and training services/programs than UI benefits per-se. At the same time, UI benefits increase reservation wages. On one hand this reduces incentives to provide search effort (and can increase the length of the unemployment spell), on the other, it can lead to “better quality matches.� The empirical literature to date, however, finds little evidence for better matching, including in the case of Brazil.64 On the contrary, there is evidence that UI increases the length of the unemployment spell. Estimates of the benefit elasticity (the percentage change in duration of the spell resulting from a one percent increase in the 63 See Olinto et al. (2007) for a comprehensive review of the literature both at the international level and in Brazil. Here we refer only to the micro effects on labor markets. The social benefits resulting from better consumption smoothing or UI as a macroeconomic stabilizer are not discussed here. 64 See Cunningham (2000). 28 benefit) range between 0.2 and 0.9, while the duration elasticity ranges between 0.4 and 0.5.65 But, overall, these effects seem to be relatively modest. For instance, if the baseline unemployment duration is 3 months, increasing benefits by 10 percent would prolong the unemployment spell by around 2 to 8 days. Increasing the duration of benefits by one month would add 2 to 3 weeks.66 Individual unemployment savings accounts can improve incentives and eliminate/reduce contingent liabilities to the government but the cost can be high for the median worker. Like in the case of pensions, UISAs improve incentives by strengthening the link between contributions and benefits. Indeed, workers face the tradeoff between using their savings to finance one additional month of unemployment or future consumption. The lack of redistribution also implies, in principle, that contributions to UISAs are not perceived as taxes. But at the same time, in the absence of risk pooling, savings rates to finance adequate benefits during the unemployment spell can be high -- 8 percent of wages in the case of Brazil to finance one month of salaries after 12 months of contributions. For median and low income workers facing credit constraints, this level of mandatory savings for precautionary motifs could be excessive. Workers can then face strong incentives to terminate contracts to cash-out unemployment savings. In the case of Brazil, the effects of UI on work incentives are not easy to disentangle and the evidence is mixed. The main study that addressed this question by exploiting changes in regulations in year 1994, found that there were not significant effects of UI on the duration of unemployment spells except in the case of transitions to self-employment.67 In this case, higher UI benefits were associated with shorter spells – as if UI benefits were helping finance a new business. The study also did not find evidence of a positive impact of UI on wages or the probability of finding a formal sector job. So the increase in reservation wages associated with UI did not materialize in better jobs. The most recent study conducted in the context of this project finds some evidence that the presence of UI and FGTS reduces somewhat the probability of exiting unemployment through informal sector jobs, but no evidence of major impacts on the duration of the spell.68 After controlling for observed and unobserved characteristics that could affect individuals’ likelihood to work in the formal sector and therefore be eligible for unemployment insurance, the study shows that individuals who are eligible for UI and/or FGTS are less likely to take informal sector jobs69 (see first panel Figure 18). At the same time, the study shows that benefit duration does not have significant effects on the duration of unemployment spells both in the case of transitions to informal or formal sector jobs. Basically, other things being equal, workers who only have FGTS behave in similar ways to workers who are eligible for 3, 4 or 5 months of unemployment benefits (see second and third panels of Figure 18). In essence, the joint unemployment benefit package might delay the unemployment spell because workers take more time to find a “good match� (and probably reduce job search efforts in matches with low potential), but the additional benefits provided by UI on top of those already available through FGTS seem to have little influence at the margin. 65 See World Bank (2004). 66 A third concern, which has been less documented, is that as a redistributive program the UI system increases the tax-wedge, which can have effects on employment levels and provide incentives to evade (see Section 1). 67 See Cunnighman (2000). The author exploits the changes in regulations introduced in 1994. She uses a different-in-difference methodology to analyze the effects of UI on the duration of the unemployment spell, post unemployment wages, and the likelihood of finding jobs in the formal sector. 68 See Margolis (2008). The study is based on the PME survey that covers workers in Brazil’s 6 largest metropolitan areas (Recife, Salvador, Belo Horizonte, Rio de Janeiro, São Paulo and Porto Alegre). The study distinguishes transitions to jobs in the formal sector from transitions to jobs in the informal sector and thus uses a competing risks duration model (one risk being formal sector employment, the other being informal sector employment). The estimation methodology takes into account that access to UI is endogenous, as it is correlated with unobservable characteristics that make workers more or less likely to work in the formal sector. Thus, correlation between the risks is introduced through unobserved individual-specific heterogeneity components in the model. 69 Informal sector here refers to wage earners without carteira and the self-employed. 29 One interpretation would be that the minimum package facilitates the job search process and allows individuals to avoid bad jobs but that when good jobs are offered these are generally taken. Indeed, if good jobs are scarce, employees might prefer to forgo a few months of unemployment insurance than to risk missing a good job opportunity. In fact, the study shows that a high unemployment rate in the metropolitan area and a high participation rate in the labor force (both of which can be interpreted as indicators of how “crowded� the labor market is), both slowdown exits from unemployment. Interestingly, a higher average real wage also tends to delay the unemployment spell. Another result is that transitions from informal to formal sector jobs have a lower probability than transitions from formal to formal sector jobs. Other things being equal, workers in the informal sector who become unemployed are less likely to exit into the formal sector (see first and third panels of Figure 18). This is again after controlling for observed and unobserved characteristics that could influence individual preferences. One interpretation is that employment opportunities in the formal sector for them are, to some extent, scarcer. Other things being equal, employers may be less likely to offer a job to somebody coming from the informal sector than to somebody coming out of the formal sector. The difference in the transition probabilities is not dramatic, but it could indicate some form of discrimination or labor market segmentation – that in a way contradicts some of the evidence discussed in Section 1. It is an issue that deserves more attention. One explanation could simply be unobserved heterogeneity that is not controlled for. Figure 18: Relative Speed of Exit from Unemployment by Type of Worker and Sector Choice 7 Ex-informal to informal 6 Log(Baseline Hazard Rate) Ex-formal to informal 5 4 Ex-formal to formal 3 2 1 Ex-informal to formal 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Months of Unemployment Exit to an Informal Job Exit to a Formal Job 7 7 UI 5 months UI 4 months 6 6 UI UI 3 months Log(Baseline Hazard Rate) Log(Baseline Hazard Rate) 5 Ex-informal 5 UI 3, 5 and 5 months 4 4 Ex-informal 3 3 FGTS only 2 2 1 FGTS only 1 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Months of Unemployment Months of unemployment Differences in the figures between FGTS and different durations for UI benefits are not statistically significant. Source: Margolis (2008). When it comes to FGTS and its economic impacts several studies have been conducted showing that there are design problems affecting both employees and employers. From the point of view of employees, one issue is whether current arrangements are reducing the value of FGTS savings to workers and thus providing incentives to force dismissals. There are three main possible reasons for this. First, individuals might be willing to put aside some precautionary savings but if the mandate of the government is too high 30 (even after taking into account some level of myopia) they might start to find ways around it. Workers could dis-save (i.e., borrow), but imperfections in capital markets might create credit constraints. Even when possible, borrowing rates tend to be higher than the rate of return on FGTS savings (more on this below). The second reason is that the presence of another consumption smoothing mechanisms (UI) reduces the value of pre-cautionary savings. In essence, there can be a redundancy that provides incentives to cash-out savings.70 The third reason is that the rates of return on FGTS savings systematically underperform the market; workers could be better-off by moving their savings elsewhere. Thus, in practice, current employer contributions to the FGTS incorporate an implicit tax that increases the tax-wedge (see Section 2). Finally, the large dismissal tax paid directly to employees who loose their jobs exacerbates the problem. From the point of view of employers incentives effects – mainly through the dismissal tax -- are more complex, but could result in higher, not lower turnover rates overall. The tax first provides incentives for employers to be more careful when hiring. This implies more selectivity (which would reduce matching rates) but also perhaps a reduction in the number of vacancies. A second effect is that during the first few months employers have more incentives to monitor the quality of the match. On one hand, this might imply increasing investments in training in the case of matches that have good potential, on the other, more rapid actions to terminate a contract in the case of poor matches. If as a result of better initial screening the share of good matches increases, one would observe lower separation rates over the short term (when employers are investing in the human capital of potentially good matches) and probably higher separation rates after a period when the quality of the match has been revealed. This effect could be exacerbated if employees themselves have incentives to collude to cash-out growing FGTS balances. But the dismissal tax could also force certain employers to keep workers that become redundant or did not perform as expected. At the end, the effect of the tax would be indeterminate. To date, there is some evidence that following the 1998 constitutional reform that increased the severance fine to 40 percent dismissals became more prevalent. Studies find that dismissal rates were multiplied by four71 and that employers and employees have incentives to collude to turn voluntary quits into dismissals -- the former to avoid taxes and the latter to cash benefits.72 In fact, there is evidence that 2/3 of the workers who voluntary quit jobs in the formal sector access their FGTS, which means that these quits are officially registered as dismissals.73 Nonetheless, an in depth analysis of the constitutional changes to labor market regulations found that the effect on turnover rates actually depends on the length of the employment spell.74 Separation rates fell for short spells (between 6 and 12 months) and increased for longer spells (6 to 12 months). The same study, did not find structural changes in the demand for labor. There is also evidence that the constitutional changes reduced job finding rates in the formal sector. A study looking at secular movements in gross flows of workers and the volatility of these flows found that the expansion of the informal sector in Metropolitan areas between 1982 and 2002 responded mainly to a drop in job finding rates in the formal sector.75 The authors attribute a small part of this drop to trade liberalization and the reminder to rising labor costs – in part resulting from the higher dismissal tax. One interpretation would be that vacancies in the formal sector fell. Some have even argued that because employers have incentives to spend more time screening workers and this increases the cost of hiring, they might attempt to substitute labor to control costs.76 Another, not mutually exclusive, interpretation might be that job search efforts for formal sector jobs fell – which would be consistent with an unchanged labor demand function. 70 Anecdotal evidence suggests that FGTS benefits are often used up-front to make extra purchases of relatively expensive consumption (or capital) goods rather than to smooth consumption. It would seem that UI benefits are used for that purpose. 71 See Barros et al (1999). 72 See Gonzaga (2003). 73 Barros, Corseuil and Foguel (1999). 74 Barros and Corseuil (2001). 75 See Bosh et al. (2007). 76 See Carneiro and Faria (2001). 31 Beyond the problem of the dismissal fine, FGTS has two conflicting functions that can also distort incentives: mandating precautionary savings in the case of job-loss and promoting saving to finance investments. The first objective calls for monthly payments that plan members can use to finance consumption. The second calls for lump-sum payments to purchases a house or start a new business. Whether FGTS should serve the second function is an important policy question. In principle, if there are failures in credit markets that need to be addressed, separate policy instruments would have to be considered, ideally not linked to a consumption smoothing scheme. Redistribution As discussed above, and like in the case of pensions, the UI system is designed to be progressive: low income workers receive higher replacement rates than high income workers. If the UI system was financed by contribution rates that were the same for low and high income workers, then indeed, redistribution could be progressive. But in the case of Brazil the UI system is financed essentially through an ear marked tax – general revenues. Then the important question to assess progressivity within the system is whether, during a given period of time, low income workers receive a higher share of total UI payments than high income workers. And this depends on two factors: (i) the utilization of UI benefits by income level; and (ii) the absolute amount of the benefits received. This sub-section addresses this question using the PME survey. The first finding is that low income workers use UI benefits less than high income workers. Take-up rates, that are the percentage of individuals receiving UI payments, are higher for high income workers than for low income workers. Thus, while only 6 percent of individuals with earnings between 25 and 50 percent of the average wage receive benefits in a given month, the take-up rate among individuals with earnings between 75 percent and 100 percent of the average is 12 percent. The take-up rate among individuals earning between 100 and 150 percent of the average wage is 22 percent. Take up rates decrease for higher levels of income (see left panel of Figure 19). These numbers do not imply that low income individuals go through fewer episodes of unemployment. In fact, Section 1 showed that unemployment risks are higher for low income individuals. The explanation seems to be, instead, that low income workers when they become unemployed are less likely to have accumulated the necessary number of contributions to be eligible for UI. Low income workers do not necessarily spend more time in unemployment – the opposite tends to be true – but they are likely to spend more time in informal sector jobs where they do not accrue UI benefits. The results of the analysis also show that the average payment by income level is higher for high income workers than for low income workers. This is to be expected. The benefit formulas presented at the beginning of this section offer higher replacement rates to low income workers but the absolute level of the benefit increases with income – up to a maximum. Hence, the average transfer among individuals with earnings above 2 times the average wage is 30 percent higher than the transfer for individuals with earnings below 0.5 times the average wage (see right panel of Figure 19). The corollary is that there is room to better target the implicit subsidies associated with UI. As it will be discussed in the next section, this is possible in the case of Brazil given the presence of FGST. For high income workers it is reasonable to think that FGTS could be the main source of income to smooth consumption during unemployment periods. This being the case, UI subsidies could be reallocated to low income individuals who in addition to be more vulnerable to unemployment have more limited means to cope with it. 32 Figure 19: Take-up Rates and Share of UI Payments by Income Level 25% 2.5 20% 2 (proportion average earnings) Average UI transfer Take-up Rate 15% 1.5 10% 1 5% 0.5 0% < 25% 25%-50% 50%-75% 75%-100% 100%- 150%- >200% 0 150% 200% 25%-50% 50%-75% 75%-100% 100%-150% 150%-200% >200% Income (percent of average earnings) Income (percentage average earnings) Source: Authors calculations based on PME data using model BRALAMMO designed by Zylberstajn et al. (2008) Other issues that deserve attention Another issue that is important to consider is the financial sustainability of the UI system. By design, the system does not link benefits to contributions. At this stage it is unclear whether the regulations prescribe a certain course of action when the system displays, or is close to displaying, a deficit. But most likely, there is no explicit balancing mechanism in place. This should be a cause of concern because projections conducted in the context of this project show that expenditures are growing very fast and could soon surpass revenues. The projection model was first tested to see its accuracy in predicting past expenditures and the results were remarkably accurate (see Figure 20). Projections for the future then indicate that expenditure would go up from R$12.3 billion in 2007 to close to R$14.4 in 2008 and around R$18 billion in 2010. This is an average increase of 14 percent per year, relative to revenues which have been growing at 11 percent per year. Figure 20: Projected Expenditures on Unemployment Benefits 20,000 18,000 16,000 Predicted Expenditures (Million R$) 14,000 12,000 10,000 8,000 6,000 Actual 4,000 2,000 - 2002 2003 2004 2005 2006 2007 2008 2009 2010 Source: Authors calculations based on PME data using model BRALAMMO designed by Zylberstajn et al. (2008) Finally, there is the question of whether the portfolio of investments of FGTS is the most adequate for its function. FGTS investments represent around 0.5 percent of GDP. Most of these investments today go to housing and sanitation projects. In the case of housing, FGTS provides subsidized loans to plan members, which represent 65 percent of the portfolio. A priori, there is little rationale for having the FGTS engage in this type of activity instead of focusing on helping individuals to replace income during 33 periods of unemployment. The current policy not only reduces the rates of return on investments but can also increase administration costs. In fact, the rate of return on FGTS savings has systematically under- performed market rates. Today, the equivalent to the Fed Funds interest rate, the so called SELIC rate, is close to 6.25 percent (real), while the annual real interest rate paid by FGTS is around 3 percent per year. 3. IMPROVING REDISTRIBUTION AND INCENTIVES The previous section showed that there is room to improve redistribution and incentives within the pensions and income protection systems. In the case of pensions the following problems were identified: (i) the system provides weak incentives to contribute for the median worker; (ii) the system attempts to reward delayed retirement for middle and high income workers but this comes at a cost that does not seem justified; (iii) in most cases internal rates of return on contributions, regardless of retirement ages and income levels, are not sustainable; (iv) redistribution is opaque and low income workers with full-careers could be receiving the lowest transfers from the system; and finally (v) current arrangements do not seem adequate to reach out to individuals with limited savings capacity outside the “formal sector.� Regarding the income protection system an important finding from the analysis -- which confirms the results from previous studies -- is that the effect of UI on the duration of the unemployment spell tends to be small. If anything, UI seems to be allowing workers to find better jobs. The main problems are related instead to: (i) incentives for both employers and employees to reduce the duration of employment spells; (ii) lower job finding rates in the formal sector as a result of a high dismissal tax; (iii) conflicting functions of FGST as a system that mandates precautionary savings and savings to finance investments; (iv) redistribution arrangements that can be regressive; and (v) a financing mechanism that is not sustainable. While the problems of the two systems are different, the causes are similar and are related to the lack of a clear distinction between consumption smoothing/insurance and redistributive functions. This section proposes a general policy framework that could guide adjustments in the social insurance system to address the problems outlined above. The section starts with a general discussion of the objectives of the social insurance system and certain principles that would need to be considered in the case of any reform. It then proposes a set of general policy interventions that could be further analyzed and eventually adopted in the case of Brazil. The last part illustrates the potential impact that these policies could have on behaviors and program costs. 3.1. POLICY FRAMEWORK FOR SOCIAL INSURANCE POLICY Social insurance systems have two core objectives. First, to help individuals smooth consumption in the presence of various shocks (e.g., unemployment, disease, incapacity to work, longevity, death). Second, to prevent poverty by ensuring that, in any state of nature, individuals are able to finance a minimum level of consumption. The first objective is often referred as the insurance function of the system, the second as the adequacy/redistributive function. The first function can be achieved through two mechanisms: risks pooling (classic insurance) and savings. Both have pros and cons; insurance seems to be a superior option for risks which are rare but with expensive consequences, and savings a better alternative for risks which are frequent but impose relatively low costs.77 In general, there is an agreement in that robust social insurance systems might need to combine both -- and in the case of the income protection system Brazil is already doing it. The second function involves inevitably some form of subsidy and a counterpart implicit or explicit tax. The challenge there is to find an arrangement that transfers the 77 See Ehrlich and Becker (1972). 34 subsidies to those that need them the most while trying to contain the negative effects that both taxes and subsidies can have on behaviors. Different social insurance systems can achieve these two objectives/functions in different ways. There are choices to be made not only in terms of savings Vs. risk pooling, but also regarding financing mechanisms (e.g., funding, pay-as-you-go, pay-roll taxes, general revenues), and institutional arrangements (e.g., public/private management). No universal model exists; different countries opt for different designs. It is, however, important to have in place a set of guiding principles that can help compare the performance of two systems and to assess whether certain policies should be preferred over others. In this Report, five guiding principles are proposed: (i) adequacy of benefits and universality; (ii) financial sustainability and affordability; (iii) predictability and robustness; (iv) progressive and transparent redistribution; and (v) administrative and economic efficiency.78 Adequacy of benefits and universality. The point here is that the benefits provided by the social insurance system need to be sufficient to maintain a certain standard of living and be designed taking into account the needs of different population groups. There is no rule to determine what the adequate level of a given benefit is. Countries make different choices that reflect, in part, social and cultural preferences. In all cases, however, it is important to define these choices explicitly and to understand their financial and fiscal consequences. In general, for each benefit, there are three parameters that need to be defined: minimum level; average replacement rate; and the ceiling on the covered wage. Ideally, minimums and ceilings should be linked to real variables that reflect standards of living (e.g., economy wide average earnings). Financial sustainability and affordability. This is an obvious principle. First, the level of the benefits offered needs to be in line with the revenues that the system will mobilize. Second, the tax-burden necessary to mobilize these revenues needs to “affordable.� A social insurance system that offers 100 net replacement rates on pensions or UI benefits can be sustainable if the tax-wedge is high enough, but it is unlikely to be affordable. It is important to Report that the SI system does not have to be financed exclusively out of pay-roll taxes and social security contributions. General revenues have an important role financing the redistributive part of the system – if coverage is large. Projected public spending, however, needs to be consistent with the fiscal framework and the broader objectives of the government regarding the production of public goods. Predictability and robustness. Ideally, the SI insurance system should not be subject to sudden and/or discretionary changes in contributions, benefit formulas and eligibility conditions. Plan members (and employers) should be able to foresee what their obligations with the system are and the types of benefits that they will receive. Retirees should not face uncertainty in terms of how their pensions will be adjusted each year. Workers should not face uncertainty in terms of how unemployment benefits will evolve each year. Employers should not face uncertainty about how pay-roll taxes will evolve. At the same time, because the financial equilibrium of the system depends on unknown factors such as the future evolution of the population and labor market dynamics, it is also important to have explicit rules that define how certain parameters of the system would adapt in the presence of macroeconomic or demographic shocks. As an example, a robust pension system would automatically index the statutory retirement age with changes in life-expectancy. Progressive and explicit redistribution. This idea has already been discussed. Redistribution is an important function of the SI system and should be preserved. The form of redistribution, however, matters both in terms of equity and incentives. To the extent possible, redistribution should be progressive and focus transfers on the most vulnerable individuals. It should also be explicit in the sense that the beneficiaries, the costs, and the financing mechanisms are identified up front. 78 For similar guiding principles applied to the case of pensions see Holzmann and Hinz (2005). 35 Economic and administrative efficiency. This principle states that the objectives of the social insurance system should be implemented at the lowest possible cost. At the level of the economy this implies paying attention to how the SI system affects incentives to work, save, invest, and create jobs. Because subsidies and taxes are involved in the design of the various programs some negative effects are to be expected, but these can be contained if the redistributive function of the system is well targeted. Finally, how the system is governed and administered makes a difference. Governance and administrative structures affect incentives and behaviors among managers, employees in the social security system, and providers of different types of services, thus ultimately influencing administrative costs and the quality of services. Because the same five principles apply to all programs, it follows naturally that their design should be carefully coordinated and ideally integrated. Integration does not necessarily mean that the programs share common administrative and IT systems, but more importantly, that they share the same financing arrangements and redistributive mechanisms. Hence, there would be similar rules to balance risk-pooling and savings across risks and to allocate subsidies to individuals with no or limited savings capacity.79 An integrated framework would also help address two problems that tend to be pervasive in social protection systems: the bundling of non-essential services in the “social security package�; and the lack of integration between certain non-contributory and contributory programs (e.g., free services provided by the Ministry of Health and health insurance or social pensions and contributory pensions).80 3.2. IMPLICATIONS FOR BRAZIL To a large extent the current pensions and income protection systems in Brazil are consistent with the principles outlined above. As previously discussed, however, there are adjustments that could be considered, particularly to improve redistribution, economic efficiency, financial sustainability, affordability, and predictability. These are summarized below. Issues related to the adequacy of benefits or institutional arrangements and management are not discussed here. Pensions Three general policy interventions are suggested, which are in line with the recommendations developed in the recent GOB-World Bank assessment of the pension system.81 Unifying benefit formulas into one that is incentives neutral. This would simplify the current system and make it more transparent. The new formula would ensure that: (i) individuals receive the same internal rate of return on contributions regardless of career histories and retirement ages; (ii) that the resulting IRR is sustainable; and (iii) that, like today, benefits are able to adapt to changes in life expectancy.82 Reviewing eligibility conditions to give individuals the flexibility to choose when to retire while ensuring adequate income protection. There would be a minimum retirement age (or one for man and one for women) for eligibility to the minimum pension guarantee, but otherwise individuals would be able to 79 For some of the theoretical arguments in favor of integration also see Stiglitz and Yun (2005). 80 For a discussion of the issue in the case of Latin American countries see Chapter 7 in Perry et al. (2007). 81 See World Bank 2007a. The main differences between the policies outlined in the report and the policies outlined here are: (i) a more elaborated proposal for the treatment of old-age subsidies and the expansion of coverage to the informal sector; and (ii) some additional suggestions in terms of benefit formulas and eligibility conditions to improve incentives. 82 The benefit formula could take the form: �  �e  �g  R   w . 1  irr  w R a p    R   G R irr i a a  where pR is the pension paid by the system at retirement age R; βw, βe, and βg are the contribution rates paid to the system respectively by the employee, the employer and the government (when there are explicit subsidies); a is the age when the individual joins the system, irr is the rate of return that the system pays on contributions; and GR(irr) is an annuity factor that also depends on irr. 36 choose when to retire as long as they are able to finance a pension that replaces a given share of their earnings (and is higher than the minimum pension). It is expected that more flexibility to choose the retirement age will improve incentives to enroll and contribute. Also, given the benefit formula presented above, the finances of the system would not be affected by early retirement. Those individuals choosing to retire early would receive lower replacement rates. At the same time, those individuals choosing to delay retirement would not receive a higher IRR. Reviewing current redistributive arrangements to make them more progressive, improve incentives to contribute, and better accommodate workers in the informal sector. The end goal is to integrate the pension for the elderly poor, the rural pension, and the minimum pension guarantee of RGPS in the context of a general strategy to expand coverage. Like today, a minimum benefit would be defined for all individuals above a certain age, ideally means tested as opposed to income tested. In addition, the system would reward low income individuals who contribute to the system by providing a subsidy above the minimum. This subsidy can take the form of a pension top-up or matching contribution;83 it would also be open to individuals in the informal sector who join voluntarily – and who are more likely to contribute lump sums (instead of a given percentage of earnings) and on an irregular basis.84 Figure 21 illustrates this idea. It graphs the value of the pension (expressed as a share of average earnings) of an individual who earns 50 percent of the average wage and who retires at age 65 after a certain number of years of contributions (between 20 and 42). The subsidy, in the example, is calculated as a 150 percent matching on contributions while active or a top-up equal to 1.5 times the contributory pension. Both are equivalent and have the same cost in present value. There is also a maximum for the subsidy equal to 20 percent of average earnings. In addition, there is a 30 percent “claw-back� (i.e., the subsidy is reduced by 30 cents for each R$ increase in the contributory pension). The proposed arrangement is equivalent to offering a minimum pension guarantee of 20 percent of average earnings after 30 years of contributions – except that the guarantee is not reduced with each additional year of contribution. In the example, it is also assumed that there is a flat means-tested pension equal to 15 percent of average earnings. Hence, the individual in question has to decide between staying out of the system, saving aside (or consuming more) and then applying for the means tested flat pension; or contributing to the system (which might imply lower consumption while active) but then having access to a somewhat higher subsidy and a substantially higher pension. In the next section we will see that individuals are more likely to prefer enrolling in the system and contributing. 83 These subsidies play a similar role than tax credits related to voluntary long term savings accounts (or health insurance). Tax credits, however, would be ineffective to motivate contributions from individuals with no or limited savings capacity. Direct subsidies are used instead. 84 This additional flexibility also requires changing the benefit formula to the one described in footnote 82. For those plan members contributing on the basis of lump sums, the contribution rates would drop from the formula and the variable wage would be replaced by the lump sum. 37 Figure 21: Designing Subsidies for Individuals with No or Limited Savings Capacity 0.6 Pension (share of average earnings) 0.5 0.4 Threshold when the matching contribution stops 0.3 Subsidy 0.2 Flat means tested pension 0.1 Contributory pension 0 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 Vesting Period Note: Example refers to an individual who earns 50 percent of the average wages, retires at age 65 and contributes 5 percent of average earnings. The subsidy is equal to 1.5 times the pension at retirement (or 1.5 times the contributions while active). Source: Authors calculations. Income protection The general idea here is to better integrate UI and FGTS and focus the mandate of the systems on consumption smoothing and redistribution. The following general policy interventions could then be considered. Centralize the payment of unemployment benefits through FGTS accounts and use the revenues of the UI system to top-up the accounts/benefits of individuals with limited savings capacity and/or those of individuals experiencing longer unemployment spells. When individuals lose their job involuntarily, they would receive benefits only from FGTS. Moreover, benefits would take the form of monthly payments up to a maximum replacement rate (to be defined by policy). FGTS would be focusing its role on consumption smoothing and not on the promotion of savings to finance certain investments. The revenues from the UI system could then be used to top-up the contributions to the FGTS accounts of workers with incomes below a given threshold, using different top-up levels for different income brackets. Alternatively, UI revenues could subsidize negative balances in the accounts of individuals who experience longer unemployment spells. To address the financial problem of the UI system, all subsidies would be expressed as a function of a “numéraire� that is adjusted periodically to ensure the solvency of the system. The system could then be gradually expanded to informal sector workers who would contribute to both their FGTS savings account and the “solidarity� account – the previous UI system. Better integrate the income protection system and pension systems. At the time of retirement, balances in the FGTS accounts would be transferred to the pension system. Workers would then have the choice between using the balances to finance a higher pension or receive a lump-sum. It would also be possible to consider using the “excess wealth� in the pension system (i.e., savings above those needed to finance a given replacement rate) to cover “deficits� in the unemployment savings accounts for individuals who experience particularly long periods of unemployment. Review the size and form of payment of the dismissal fine. The dismissal fine can have a role in internalizing the social costs associated with the loss of jobs. It is important, however, to review its level to ensure that the fine does not impose an excessive burden on firms that can then discourage investments and employment creation or excessive turnover. Since the tax is internalizing social costs, then it would also be more appropriate to use it to finance redistribution as opposed to direct benefits for workers. 38 Revenues from the tax, for instance, could also top-up the FGTS account of individuals with limited savings capacity. These are only general policy guidelines that would need to be developed and analyzed in detail prior to adoption and implementation. The next section presents a preliminary assessment of how some of these policies could affect incentives, behaviors and program costs. 3.3. POTENTIAL EFFECTS ON BEHAVIORS AND PROGRAM COSTS Analysis using a life cycle behavioral model estimated for Brazil provides several insights.85 Movements in and out of the social security and retirement decisions are highly dependent on individual preferences. This can be seen in Figure 22 which gives the probability of contributing to the social security or being retired at a given age. In the figure, each line refers to one point of the joint distribution of model parameters.86 The figure shows that, under the status-quo, on average, around 30-35 percent of the high-income workers and 45-50 percent of the low income would be outside of the formal sector between ages 35 and 45. Afterwards, the probability of formal sector work declines for both high and low earners. This is consistent with the current distribution of age cohorts as discussed in Section 1. Thus, several sets of preferences reproduce the pattern of a declining probability of formal work with age which is found in the empirical analysis of labor market transitions over the life-cycle.87 In essence, the “marginal utility� of formal jobs relative to informal jobs goes down with time – while the marginal disutility linked to the effort invested in finding and keeping jobs remains constant (given the shape of the utility function). In terms of retirement, the model predicts that around half of the high earners would retire between ages 55 and 60. Low income individuals, on the other hand, tend to retire later – between ages 60 and 65. This is also consistent with the analysis of cohorts presented in Section 1. But again, the variation in retirement patterns can be considerable. Some individuals can retire as early as 53, others can delay retirement until 70 (see bottom two panels of Figure 22). 85 The model incorporates the current rules for RGTS, unemployment insurance, and FGTS. The core assumption is that individuals react to changes in these rules in order to maximize and smooth consumption over time. More precisely, during the relevant time period, individuals choose: (i) how much to save; (ii) whether they retire; and (iii) the level of effort they put into finding or preserving formal sector jobs. Thus, transition probabilities in and out of the social security (i.e., formal sector) are assumed to have two determinants. First, factors which are independent from workers decisions (e.g., hiring and/or job destruction rates in the formal sector) although they can be correlated with workers characteristics such as their education level. Second, the time and energy that individuals allocate to keep or find formal sector jobs and which depend, in part, on their preferences/predisposition for formal vs. informal sector jobs. Thus, the modeling exercise assumes individuals choose whether to engage in formal or informal sector work. The freedom of choice though is correlated with individuals’ characteristics. For some, transition probabilities might be largely explained by exogenous factors as opposed to the level of effort invested in finding/keeping formal sector jobs (see Robalino et al., 2009 for a description of the model and detail results of the policy analysis). 86 Te key parameters estimated are: (i) the coefficient of risk aversion; (ii) the preference for consumption over leisure; (iii) an exogenous probability of formal sector work; (iv) the rate of time preference; (v) the disutility of efforts to find/keep formal sector jobs; and (vi) and the probability of working when retired. 87 See Cunnignham, 2006 and Perry et al. 2007 39 Figure 22: Probabilities of Contributing to INSS and Retiring Earnings = 100% Average Earnings=50% Average 1.0 1.0 0.9 0.9 0.8 0.8 Probability of Formal Work Probability of Formal Work 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0 35 40 45 50 55 60 65 35 40 45 50 55 60 65 Age Age 1.0 1.0 0.9 0.9 0.8 0.8 Probability of Retirement Probability of Retirement 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0 45 50 55 60 65 45 50 55 60 65 Age Age Note: The dark lines with dots give the “average� path for the cohort. Source: Robalino et al., (2009). Another message from the simulations is that there are important interactions between the pensions and income protection systems, which emphasize the need for an integrated approach to the design of eventual reforms. For instance, without the pension system, individuals with earnings equal to the average would likely retire earlier but also spend more time in the formal sector (see first panel of Figure 23) – which would therefore affect the performance of the unemployment insurance system. 88 Similarly, in the absence of unemployment insurance participation in the formal sector would increase and retirement ages would decrease (see second panel of Figure 23). This would also affect the performance of the pension system. In the absence of FGTS, on the other hand, individuals would be less likely to participate in the formal sector (see third panel of Figure 23), which would affect both the performance of the unemployment insurance and pension systems the presence of FGTS does not seem to meaningfully affect retirement ages). 88 Robalino et al., (2009) show that the effects would be different for low income individuals who, for instance, would spend less time in the formal sector. 40 Figure 23: Effects of SI Programs on Contribution Densities and Retirement Ages (Average Earner) Pensions Unemployment Insurance 0 1 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 -1 0 Change in Retirement Age (# of Years) Change in Retirement Age (# of Years) -0.30 -0.25 -0.20 -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 -2 -1 Discount rate (+) -3 -2 -4 -3 -5 -4 -6 -5 -7 High cost of working in the formal sector; high -6 -8 probability of work after retirement -9 -7 Change in Contribution Density (% of Time) Change in Contribution Density (% of Time) FGTS 1 Change in Retirement Age (# of Years) 0 -0.08 -0.07 -0.06 -0.05 -0.04 -0.03 -0.02 -0.01 0.00 -1 -2 -3 High discount rate and high -4 cost of working in the formal sector -5 Change in Contribution Density (% of Time) Source: Robalino et al., (2009). Simulations also show that there are gains to be made by moving towards explicit forms of redistribution that could combine matching contributions with targeted anti-poverty transfers. In the context of a unified pension formula, for instance, matching contributions could be a promising instrument to expand pension coverage to the informal sector.89 In essence, the program involves matching part of the contributions made by employees (and in this case the employer) as an incentive to promote enrollment and contributions, and thus help finance an adequate pension at retirement for individuals with low savings capacity. Simulations in the case of Brazil show that matching contributions could increase contributions among low income workers relative to the case of a pure minimum pension guarantee (see Figure 24). The most significant effects would take place when there are no restrictions on the retirement age. Contribution densities in that case can increase, on average, by up to 30 percentage points depending on preferences. The tradeoff, however, is a reduction in the retirement age that can decline by up to 10 years. In essence, individuals who before delayed retirement to benefit from the minimum pension offered at age 65 now are able to retire early and benefit from the matching. One policy implication would be that, within a strategy to expand coverage and promote formality, matching contributions can play a role as long as individuals are allowed to decide when to retire. As long as the pension system is actuarially fair and there is a maximum level of matching, this would not threaten the financial sustainability of the system. Imposing restrictions on the retirement age and, in particular, setting a high retirement age would reduce incentives to contribute and participate in the social security. Clearly, not imposing a higher minimum retirement age and not having a minimum pension can result in pension values that are too low relative to earnings. The alternative then would be to set a retirement restriction that is based on the value of the pension. Individuals who after a certain age have not met the restriction, would be eligible for a minimum pension guarantee tested on consumption (or total earnings) not simply pension income. 89 see Palacios and Robalino, 2009 41 Figure 24: Effects of Matching Contributions Matching = 75% Matching=225% 4 4 2 2 Minimum retirement age = 65 Minimum retirement age = 65 Condition based on value of pension (> min pension) 0 0 Change in retirement age Change in retirement age -0,10 0,00 0,10 0,20 0,30 0,40 0,50 -0,10 0,00 0,10 0,20 0,30 0,40 0,50 -2 -2 -4 -4 No mininimum No mininimum retirement age retirement age -6 -6 -8 -8 -10 -10 -12 Change in contribuiton density -12 Change in contribuiton density Note: The simulations look at two matching levels: 75% and 225%. In both cases there are two retirement ages, minimum 55 and minimum 65. The analysis is only applied to low income workers. The two panels give the changes in contribution densities and retirement ages for the two matching levels relative to the case of the minimum pension guarantee at age 65. In the figures the circles correspond to the case with a minimum retirement age at 55 and the squares to the case with a minimum retirement of 65 years. In the second panel the triangles refer to a case where individuals can retire after age 55 but before age 65, as long as they have a pension that is above or equal to the minimum pension. Source: Robalino et al., 2009. A final issue to consider is that the cost of matching contributions can be lower than the cost of a minimum pension guarantee – depending on the level of the targeted replacement rate. This can be seen in Figure 25. In the case of the two minimum pensions graphed (42 and 25 percent of average earnings) the majority of workers retire with a pension at least equal to the minimum (there are a few exceptions of workers retiring before age 65 with no minimum pension), but this means that the pension system needs to subsidize a large part of the total pension received. The subsidy is by definition larger than in the case of matching contributions since workers contribute less (i.e., have lower contribution densities). It is also important to note that while matching contributions may fail to bring most individuals to retire with, effectively, a replacement rate equal to 84 percent of pre-retirement earnings (or 42 percent of the average wage) it can bring many workers to retire with a replacement rate of 50 percent (or 25 percent of the average). And replacement rates with matching are considerably higher than replacement rates without matching (the maximum replacement rate without matching is represented in the figure by the dotted horizontal line). Hence, the effectiveness of matching contributions also depends on the policy objective. By international standards a minimum pension equal to 42 percent of average earnings is high.90 A 25 percent target would be more affordable and could be more easily achievable through matching contributions. It is also possible to think about a system where the minimum pension is 25 percent of average earnings but where individuals who contribute more can still finance higher pensions without losing the subsidies. 90 See Whitehouse, 2007 42 Figure 25: Costs of Matching Contributions and Minimum Pensions 0,90 Minimum pension (42% AW) 0,80 0,70 Average replacement rate Matching 225% no Matching 225% age 65 minimum age 0,60 Matching 225% with Minimum pension (25% AW) restriction in the 0,50 value of the pension before age 65 0,40 Matching 75% age 65 0,30 Maximum replacement rate in the 0,20 absence of subsidies 0,10 0,00 0 1 2 3 4 5 Present value of subsidies Note: The cost of the subsidy has been normalized as a percentage of total earnings in the base year. Source: Robalino et al., 2009. Looking forward, the analytical tools developed in the context of this study can be refined and used to deepen the analysis of the behavioral, fiscal, and welfare impact of alternative reforms to the pensions and income protection systems. The main message at this stage is that there is room to improve the design of the programs to generate, both, a more progressive redistribution of income and better incentives for formal sector work. A key ingredient would be to move from implicit to explicit forms of redistribution. In addition, simplifying and better integrating the various programs. 43 REFERENCES Addison, John T. and Paulino Teixeira. 2005. “What Have We Learned about the Employment Effects of Severance Pay? Further Iterations of Lazear et al.� Empirica Vol. 32, pp. 345–368. Albrecht, J., L. Navarro, and S. Vroman. 2006. “The Effects of Labor Market Policies in an Economy with an Informal Sector.� IZA Discussion Paper 2141, Institute for the Study of Labor, Bonn, Germany. de Barros, Ricardo Paes, Carlos Henrique Corseuil, and Mônica Bahia. 1999. “Regulamentacão do mercado de trabalho e duração do emprego no Brasil,� Pesquisa e Planejamiento Econômico, 29 (3). de Barros, Ricardo Paes, Carlos .Henrique Corseuil and M. Foguel. 1999. “Os incentivos adversos dos programas de seguro-desemprego e do FGTS no Brasil.� Rio de Janeiro, Brazil: IPEA. Mimeographed document. de Barros Ricardo Paes and Calso Henrique Corseui. 2001. “The Impact of Regulations on Brazilian Labor Market Performance.� (Instituto de Pesquisa Economica Aplicada) Research Network Working paper series ; R-427. Inter American Development Bank. Besley, Timothy and Robin Burgess. 2004. “Can Labor Regulation Hinder Economic Performance? Evidence From India.� Quarterly Journal of Economics Vol. 119, No. 1, pp 91-134. Bodor, Andras, David A. Robalino and Michal Rutkowski. 2007. “Assessing the Distortions of Mandatory Pensions on Labor Supply Decisions and Human Capital Accumulation: How to Bridge the Gap between Economic Theory and Policy Analysis.� (forthcoming). Bosh Mariano, Edwin Goñi, and William Maloney. 2007. “The Determinants of Rising Informality in Brazil: Evidence from Gross Worker Flows.� Mimeo. World Bank. Washington. DC. Bosh Mariano and William Maloney. 2007. “Comparative Analysis of Labor Market Dynamics Using Markov Processes: An Application to Informality.� IZA Discussion Paper Series. IZA DP No. 3038. Bosh Mariano and William Maloney. 2008. “Cyclical Movements in Unemployment and Informality in Developing Countries.� Mimeo. World Bank. Washington. DC. Bothelo Fernando and Vladimir Ponczek. 2008. “Segmentation in the Brazilian Labor Market.� Mimeo. Getulio Vargas Foundation. Sao Paolo. Brazil. Carvalho, Irineu E. 2002: "Old-Age Benefits and Retirement Decisions of Rural Elderly in Brazil" mimeo, Department of Economics, Massachusetts Institute of Technology, Cambridge, MA, November 15, 2002. Corseuil, Carlos Henrique, Eduardo Pontual Ribeiro, and Daniel Santos. 2003. “Job and Worker Flows in Brazil,� Mimeo. Cunningham,Wendy V. 2000. “Unemployment Insurance in Brazil: Unemployment Duration, Wages and Sectoral Choice,� Technical Report, World Bank. Daveri and Tabellini. 2000. “Unemployment Growth and Taxation in Industrial Countries. Discussion Paper� 9706. Brescia University. Brescia. Italy. 44 Davis, S., and J. Haltiwanger. 1999. “Gross Job Flows.� In Orley Ashenfelter and David Card, eds, Handbook of Labor Economics. Amsterdam. Domeland, Dorte and Norbert Fiess. 2002. World Bank Brazil Job Reports, Vol. 2, 2002. Ehrlich Isaac and Gary S. Becker. 1972. “Market Insurance, Self-Insurance, and Self-Protection.� The Journal of Political Economy, Vol. 80, No. 4. (Jul. - Aug., 1972), pp. 623-648. Fallon, Peter R. and Robert E. B. Lucas. 1992. “The Impact of Changes in Job Security Regulations in India and Zimbabwe.� The World Bank Economic Review Vol. 5, No. 3, pp. 395-413. Gonzaga, Gustavo. 2003. “Labor Turnover and Labor Legislation,� Economía, Journal of the LACEA, 4 (1). Gruber, Jonathan and David A. Wise. 2005. “Social Security Programs and Retirement Around the World: Fiscal Implications.� NBER Working Paper Series No. 11290. Haffner, R., S. Nickell, G. Nicoletti, S. Scarpetta, and G. Zoega. 2001. “European Integration, Liberalization and Labour Market Performance.� In G. Bertola, T. Boeri, and G. Nicoletti, eds, Welfare and Employment in a United Europe. Cambridge, Mass.: MIT Press. Haltiwanger John, Stefano Scarpetta and Milan Vodopivec. 2003. “How Institutions Affect Labor Market Outcomes: Evidence from Transition Countries.� Paper presented at the World Bank Economist Forum. Washington DC. April 10. 2003. Heckman, J.J. and C. Pagés. 2000. “The Cost of Job Security Regulation: Evidence from Latin American Countries.� National Bureau of Economic Research Working Paper No. 7773. Holzmann, Robert. 2005. “Reforming Severance Pay: Toward an Understanding of Program Rationale, Economic Impact and Reform Options.� Empirica Vol. 32, pp. 251–253. Kugler, A., and M. Kugler. 2003. “The Labor Market Effects of Payroll Taxes in a Middle-Income Country: Evidence from Colombia.� IZA Discussion Paper 852, Institute for the Study of Labor, Bonn, Germany. Lazear, E.P. 1990. “Job Security Provisions and Employment.� Quarterly Journal of Economics Vol. 105, pp. 699–726. Levy Santiago. 2008. “Good Intentions, Bad Outcomes.� Brooking Institution. Washington. DC. Loayza, N., and J. Rigolini. 2006. “InformalityTrends and Cycles.� Policy Research Working Paper 4078, World Bank, Washington, DC. Maloney, W. 1998. “Are Labor Markets in Developing Countries Dualistic?� Policy Research Working Paper 1941, World Bank, Washington, DC. Nickell Stephen 2003. “Unemployment and Labor Market Rigidities: Europe versus America� Journal of Economic Perspectives. 11 (3): 55-74. 45 Palacios and Sluchynsky. 2006. “The Role of Social Pensions.,� Pensions Primer. World Bank. Washington DC. Margolis David N. 2008: “The Effect of Unemployment Insurance on Formal and Informal Sector Employment in Brazil.� Draft Manuscript. World Bank. Washington DC. Menezes-Filho, Naercio Aquino and Paolo Picchetti (2000), “Os determinantes da duracão do desemprego em São Paulo,� Technical Report, Report no.24408-BR 2000. Pereira Priscilla. 2008. “Disincentive Effects of Social Security System with Disability Insurance Coverage on Labor Market.� Unpublished Manuscript. University of Illinois at Urbana-Champaign. Olinto Pedro, Kathy Lindert, Rita Almeida, Jason Hobbs and Rodrigo Garcia Verdu. 2007. “The Impacts of Public Transfers on Labor Supply in Brazil and Abroad: A Review of the Evidence.� Background Paper – Phase I. PESW On Labor Markets. Washington DC. Orellano, Verônica and Paulo Picchetti. 2002. “A Bi-variate Probit Analysis of JobTurnover in Brazil,� Technical Report, Universidade de São Paulo. Queiroz, Bernardo Lanza. 2007: “Retirement Incentives: Pension Wealth, Accrueal and Implicit Tax,� Unpublished paper, Department of Demography, Centro de Desenvolvimento e Planejamento Regional, (Cedeplar), Universidade Federal de Minas Gerais, Departamento de Demografia, Brazil. Queiroz, Bernardo Lanza. 2005: “Labor Force Participation and Retirement Behavior in Brazil,� Unpublished Ph.D. dissertation, Department of Demography, University of California at Berkeley, Berkeley, CA, December 2005. Robalino David and Andras Bodor. 2008: “On the Financial Sustainability of Earnings Related Pay-as- you-go Systems and the Role of Government Indexed Bonds�. Forthcoming. Journal of Pension Economics and Finance. Robalino David, Eduardo Zylberstajn, Helio Zylberstajn, and Luis Eduardo Afonso. 2009. “An Ex-Ante Evaluation of the Impact of Social Insurance Policies on Labor Supply in Brazil: The Case for Explicit over Implicit Redistribution.� Policy Research Working Papers Series. World Bank. Washington DC. Rutkowski, Jan. 2007. “Labor Taxes and Employment in ECA.� Unpublished Manuscript. World Bank. Washington DC. Vodopivec, M. 2004. Income Support for the Unemployed: Issues and Options. World Bank, Washington, D.C. World Bank. 2002: “Brazil Jobs Report.� Vol. 1 of Report no.24408-BR, World Bank, Washington DC. World Bank. 2004. “Unemployment Insurance: Efficiency Effects and Lessons for Developing Countries.� Employment Policy Primer. April 2004 n No. 5. World Bank 2007a. “Brazil: Towards a Sustainable and Fair Pension System.� Report. Washington DC. World Bank 2007b. “Morocco: Skills Development, Social Protection and Employment� Report. Washington DC. 46 World Bank 2007c. “Algeria: Consolidating the Move from Jobs to Income Protection.� Policy Note. Washington DC. 47