Brazil From Stability to Growth through Public Employment Reform

Transcription

1 Report No BR Brazil From Stability to Growth through Public Employment Reform (In Two Volumes) Volume Il: Annex February 17, 1998 Brazil Country Management Unit Poverty Reduction and Economic Management Unit Latin America and the Caribbean Regional Office Document of the World Bank

2 CURRENCY AND EQUIVALENTS Currency Unit: The Real R$1.005 = US$l 1996 R$1.078 = US$l 1997 WEIGHTS AND MEASURES Metric System FISCAL YEAR January I - December 31 ABBREVIATIONS AND ACRONYMS BANDEPE BANRISUL BNDES CLT FGTS GFS IBGE IFS INSS IPEA PDV PROER SECEX VAT Banco do Estado de Pernambuco Banco do Estado do Rio Grande do Sul Banco Nacional do Desenvolvimento Economico e Social Consolidagao das Leis Trabalhistas Fundo de Garantia por Tempo de Servigo Govemment Finance Statistics Instituto Brasileiro de Geografia e Estatistica Intemational Financial Statistics Instituto Nacional de Seguridade Social Instituto de Pesquisa Econ6mica Aplicada Programas de Demissao Voluntaria Programa de Estimulo a Restructura9ao e ao Fortalecimento do Sistema Financeiro Secretaria da Receita Federal - Banco do Brasil Value Added Tax Vice President: Shahid Javed Burki Director: Gobind T. Nankani Lead Economist: Suman K. Bery This Report was prepared by a team consisting of Gautam Datta (Task Manager, LCSPR), Indermit Gill (LCC5C) and Craig Burnside (DECRG). Indermit Gill is the author of Chapters 3 and 4 of the main report and the Annex. Homi Kharas (PRMEP) helped formulate the framework and provided guidance throughout. A background paper for the World Bank authored by Nissan Liviatan (Bank of Israel and Hebrew University) was extensively drawn on for the macroeconomic sections. The public employment sections of the report are the result of a collaborative effort with IPEA, Rio de Janeiro. Contributions were also made by Francisco Carneiro (University of Brasilia) and Edward Amadeo (PUC, Rio). Michael Walton (PRMPO) was the peer reviewer.

3 Annex Public Employment Reform Table of Contents 1. THE MAIN PROBLEM, AND A SIMPLE ANALYTICAL FRAMEWORK I EMPHASIZING PRIVATE-PUBLIC DIFFERENTIALS IN EMPLOYMENT LEVELS EMPHASIZING PRIVATE-PUBLIC DIFFERENTIALS IN COMPENSATION FORMALIZING PUBLIC-PRIVATE DIFFERENCES IN COMPENSATION INSTRUMENTS FOR CHANGING S PUBLIC AND PRIVATE EMPLOYMENT AND EARNINGS EMPLOYMENT IN PUBLIC AND PRIVATE SECTORS EARNINGS AND THE WAGE BILL IN PUBLIC AND PRIVATE SECTORS PUBLIC-PRIVATE DIFFERENCES IN EARNINGS THE PUBLIC-PRIVATE GAP IN EARNINGS PUBLIC-PRIVATE DIFFERENCES IN JOB SECURITY PUBLIC-PRIVATE DIFFERENCES IN PENSIONS PUBLIC EMPLOYMENT REFORM: WHAT IS BEING DONE? OVERVIEW: VOLUNTARY SEVERANCE PROGRAMS CRITICAL FEATURES OF PUBLIC ADMINISTRATION PDVS CRITICAL FEATURES OF PDVS OF PUBLIC ENTERPRISES IMPLICATIONS FOR FUTURE PDVS PUBLIC EMPLOYMENT REFORM: WHAT IS NEEDED? A SUMMARY STATEMENT OF THE PROBLEM PRINCIPLES FOR RENEWED EFFORTS REDUCING PUBLIC EMPLOYMENT UNDER CURRENT CONDITIONS SELECTED PENSION REFORMS CAN ELIMINATE PUBLIC SECTOR PENSION PREMIUM SELECTED LABOR REFORMS CAN REDUCE ARTIFICIALLY HIGH INFORMALITY AND TURNOVER OF PRIVATE EMPLOYMENT ADMINISTRATIVE REFORMS CAN REDUCE PUBLIC-PRIVATE JOB STABILITY DIFFERENTIALS LESSONS OF INTERNATIONAL EXPERIENCE SOCIAL SECURITY REFORMS PUBLIC SECTOR RETRENCHMENT SCHEMES ACTIVE LABOR PROGRAMS... 46

4

5 THE MAIN PROBLEM, AND A SIMPLE ANALYTICAL FRAMEWORK 1. The problem of reducing the fiscal burden of public employment is viewed as central to a sustainable reduction of the high public deficit, which in turn is required to sustain the successful stabilization. Government efforts to reduce the wage bill attributable to civil servants - administration workers at the federal, state, and municipal levels, education and health workers at all levels of government, and judicial and legislative workers - through reduction in real salary levels is increasingly difficult with the decline in inflation to single digit levels since Reductions in employment are believed to be constrained by the presence of tenure or neartenure rights of most public sector workers, guaranteed under the Constitution. The payment of generous pensions to civil servants results in their becoming a financial burden for government for life or longer (because of survivor benefit schemes) but reduction of pensions or cost-sharing through employee contributions is difficult because these rights are unassailable under the Constitution. 1.1 EMPHASIZING PRIVATE-PUBLIC DIFFERENTIALS IN EMPLOYMENT LEVELS 2. In general, consolidated government employment (excluding public enterprises) in Brazil does not appear to be either extraordinarily high or low by international standards. Rama (1997), in the most up-to-date study of public sector employment in 90 countries, finds that the predicted level of government employment for Brazil is about 9.6% for the 1990s. This is close to the actual figure from nationwide PNAD surveys, which ranges between 9.0% and 9.4% during Including employment in Brazil's public enterprises, the share of government goes up to about 12%, which is also not usually high by international standards. 3. At the most aggregate level, therefore, the problem of high payroll expenses in Brazil is not due primarily to excessive government employment, but more so because of extraordinarily high salaries and benefits. However, this does not rule out the possibility that public sector employment growth has outstripped that in the private sector, or of public sector redundancies in particular regions (e.g., in some northeastern states), at particular levels of government (e.g., at the municipal level) or for particular classes of workers (e.g., for judicial and legislative workers). In this report, we focus on public-private differentials in employment at a more disaggregated level: * Using PNAD surveys, we examine public-private employment differences in selected states and the federal district; the states are Pernambuco, Minas Gerais, Sao Paulo, Rio de Janeiro, and Rio Grande do Sul. * For the country as a whole, and for each of these states (and the federal district), we examine the share of public employment at different levels of government - federal, state, and municipal, and public enterprises.

6 -2 - * For the country as a whole, and for these selected states, we examine public and private employment by type of contract (e.g., com carteira, or sem carteira). 4. This analysis helps in identifying where public employment is relatively high and where, alternatively, the problem of high public sector payroll costs is the result of private-public differentials in compensation. 1.2 EMPHASIZING PRIVATE-PUBLIC DIFFERENTIALS IN COMPENSATION 5. Under the constraints imposed by the 1988 Constitution, state and federal governments have relied largely on incentives to tenured public employees to voluntarily leave government employment. In most cases, the size of severance packages required for employees to leave depends not simply on the level of public earnings, but upon public-private diferences in earnings, pensions, job stability, and other benefits. Much of the discussion surrounding fiscal adjustment, however, has focussed on the former. In this report, we explicitly focus on the latter: * Based on nationwide household surveys, we compute public-private differences in monthly earnings, adjusting for worker characteristics such as education, age, tenure, sex, and race. * Using experience-earnings profiles, and rules for determining pensions, we compute publicprivate differentials in pensions. * Using data on tenure in the current job, we estimate public private differentials injob stability. 6. This approach lends itself easily to policy discussions. Labor market reforms that reduce the cost of labor and raise the demand for labor (e.g., by lowering payroll tax rates), or social security reforms that make INSS pension benefits conform more closely with contributions, or reforms that lower artificially high turnover rates and increase private sector job stability (e.g., social security reforms that lower the informality of employment or a redesign of thefundo garantia por tempo de servico - FGTS) would reduce private-public differentials in earnings, pensions, and job stability respectively, and make it easier to reduce civil service employment. Similarly, an administrative reform bill for the public sector that enforces longer working hours, requires civil servants to make contributions for pensions at the same rate as for INSS benefits, or eliminates tenure, would also reduce public-private differentials in earnings, pensions, and job stability and make it easier to reduce public employment. 7. Focussing on public-private differentials also helps to illustrate that there is more than one way to achieve the same objective. Thus, the incentives for public sector employees to give up government jobs can be made stronger by reducingpublic sector earnings, pensions, and job stability; or by increasingprivate earnings, formality of employment, or job stability. 8. Finally, the analysis can also be used to distinguish the effects of policy measures that require constitutional reforms, and those that are feasible under current conditions. Thus, the severance packages that are required to compensate workers for possible losses in earnings and

7 -3 - pensions can be estimated. The fiscal effects of specific constitutional reforms can be identified as the reduction in severance packages due to these changes. The efficiency effects of constitutional reforms are harder to estimate, and are not discussed in detail in this report. 1.3 FORMALIZING PUBLIC-PRIVATE DIFFERENCES IN COMPENSATION 9. A simple equation can be used to formalize these issues. If Cpub is the compensation in the public sector, Cpvt is the compensation in the private sector, and S is the level of severance benefits required to make public servants indifferent between staying in or leaving government employment, then: S = Cpub - Cpvt Or, if C = E + P + T, where E represents the level of earnings, P is pension benefits, and T is the value attached by the worker to job stability, S = (Epub-Epri) + (Ppub-Ppvt) + (Tpub-Tpvt) - For earnings differences, the most important factors to consider are worker characteristics such as gender, education, and work experience. v For pensions, worker characteristics that determine earnings are important, because they determine pension levels as well. But also important is the regime distinction between estatutario and CLTista, because it determines whether or not the worker contributes towards these benefits, and whether or not pension benefits have a ceiling. - For job stability, the most important factor is the regime distinction betweenestatutario and CLTista. For CLTistas, the years of tenure may also be important because those who have been on the job for more than five years have near tenure rights; but perhaps even more important is the existence of labor regulations (e.g., payroll taxes, and the FGTS scheme) that may artificially increase turnover for CLTistas. 10. Thus, S should be viewed as a function of how worker attributes translate into earnings and pensions in the public and private sectors, the employment regime in the public sector, and labor regulations in the private sector. Note also that since Ei and Pi are expected earning and pensions over the remaining working life, calculating them requires using a rate of discount. If workers discount the future at a higher rate than governments, then the level of S that induces voluntary departure may result in significant savings for the government. 1.4 INSTRUMENTS FOR CHANGING S 11. The problem of public employment reform, in order to reduce the public sector fiscal burden, can therefore be seen deciding on which measures will reduce S. The instruments for reducing S are:

8 -4- * Fiscal restraint, including reducing the earnings of some public sector employees by deliberately keeping nominal wage adjustments below the rate of inflation. * Administrative reforms that reduce tenure rights of civil servants. * Public pension reforms that change pension rules (e.g., increasing required years of service, introducing contributions for civil servants, or reducing replacement rates). * Economic reforms that increase the demand for labor in the private sector and increase private earnings relative to those of similar workers in the public sector. * Social security reforms that make workers view INSS-related payroll deductions as contributions rather than taxes and eliminate some of the perverse incentives for both employers and employees to become informal. * Labor market reforms to eliminate distortions that artificially raise turnover rates in the private formal sector. 12. In this report, we examine which of these broadly defined strategies are likely to be effective, and within each strategy, the exact measures needed.

9 -5-2. PUBLIC AND PRIVATE EMPLOYMENT AND EARNINGS 13. This chapter examines the patterns of employment and average earnings in the private and public sectors, the growth of the public sector wage bill in recent years EMPLOYMENT IN PUBLIC AND PRIVATE SECTORS 2.1.a All-Brazil, Pesquisa Nacional por Amostra de Domicilios (PNAD) surveys indicate that total employment in Brazil was million in September Of this, public sector employment - other than public enterprises - was approximately 5.6 million (2.2 million in public administration, more than half at the municipal level, 0.6 million in the judiciary, legislature and military, and 2.9 million in the education and health sectors). The enterprise sector had 56.9 million workers (16.1 million in agriculture, 8.4 million in industry, and 28.4 million in services). Table 2.1: Sectoral Shares in Employment, All Brazil EMPLOYMENT (in percent of total) Public Employment Direct Administration Federal State Municipal Education & Health Judiciary & Legislative Military Private Employment* Agriculture Industry Services Construction Other private Total (percent) Total (millions) n.a.** n.a.** n.a.** * Enterprise employment includes employment in public enterprises. ** Total employment figures before 1991 are not accurately estimated. Source: Pesquisa Nacional por Amostra de Domicilios (selected years). 15. The share of the "public sector" - civil servants and the military - as defined above rose from about 7% to 9% of the total between 1981 and 1990, and has stayed above 9% since then.

10 -6- Between 1992 and 1995, the number in public administration increased from about 2.07 million to 2.16 million - almost all of this increase was accounted for by an increase in municipal employment. State employment fell by 1%, and federal employment rose about 7% over these two years. The size of the military stayed roughly constant at about 0.27 million. Both judicial/legislative and education/health sectors registered increases in employment between 1992 and Judicial/legislative staff increased from 0.26 to 0.32 million, and employment in education and health increased from 2.66 to 2.85 million. 16. In the enterprise sector, agricultural employment fell, employment in industry and construction stayed constant, and employment in services (especially personal services) rose between 1992 and Using the class of worker distinction, the share of the public sector was about 12% during the period This definition thus includes workers in public enterprises and autonomous agencies (fundacoes). Almost all government employees are either public servants (estatutarios)or have a signed work card (CLTistas, com carteira). Workers in the public sector who do not fall into this category - i.e., are not estatutarios and do not have a signed work card (sem carteira) form about 1.5% of total employment in the country. Table 2.2: Share of workers, by contract type (percent), 1995 All Brazil Public Servants Non Public Servants Total Share With Card Without Card Public Sector Workers 11.5 Federal State Municipal Military Private Sector Workers 88.5 Salaried SelJemployed Employers Unpaid Total Source: Pesquisa Nacional por Amostra de Domicilios, b Employment in SelectedStates, We analyze employment patterns in the federal district and six states (Bahia, Minas Gerais, Pernambuco, Rio de Janeiro, Rio Grande do Sul, and Sao Paulo) to examine if there are considerable state-level differences in these patterns and trends. Other than in Rio de Janeiro and the federal district, the public sector (as defined above) is about 8% of total employment. Rio de Janeiro's ratio is about 11%, and the federal district's twice that. Public administration as a ratio of total employment is highest in Pernambuco, while Rio has the highest ratios for education and health. In the private sector, agriculture has the lowest share in Rio and the federal district, and

11 -7- the highest in the Northeastern states. Sao Paulo and Rio Grande do Sul have the highest ratios for industry, while Rio, the federal district and Sao Paulo have the highest ratio for services. These ratios did not change very much over the period For the six states and the federal district, while the largest share of government employment is of public servants (estatutarios), considerable government employment in all states is for workers without tenure. This ratio.yanges from about 3% of total employment in Pernambuco, to more than 6 percent in Bahia and'the federal district. Government employees without a signed working card are about 1% overall, but are close to 2% of total employment in Bahia and the federal district. Much of this "informality" of government employment is at the municipal level. In the case of municipal employees, the two northeastern states appear to have a considerable number who do not have a signed working card. Table 2.3: Sectoral Shares in Employment, 1995 Selected States Distrito Bahia Pernam- Minas Rio Rio de Sao Federal buco Gerais Grande do Janeiro Paulo Sul Public Employment Direct Administration Federal State Municipal Education & Health Judiciary & Legislative Military Private Employment* Agriculture Industry Services Construction Other private Total (percent) Total (millions) * Private employment includes employment in public enterprises. ** Total employment figures before 1991 are not accurately estimated. Source: Pesquisa Nacionalpor Amostra de Domicilios, There is also considerable variation across states in the degree of formality of private employment across states. Northeastern states have largest proportions of private workers who fall into this category. While the share of workers with a signed card in private employment is 35-50% for southern states and the federal district, it is 15-20% for the northeastern states. Minas Gerais, with a ratio of about 30%, falls in the middle.

12 -8 - Table 2.4: Share of workers, by contract type (percent), 1995 Selected States Distrito Bahia Pernam- Minas Rio Rio de Sao Federal buco Gerais Grande do Janeiro Paulo Government* Public service Others With signed card Without signed card Private Employment** With signed card Without signed card Selfemployed Employers Unpaid Total (percent) Total (millions) * Includes federal, state, municipal, and military personnel. * Private employment includes employment in public enterprises. Source: Pesquisa Nacional por Amostra de Domicilios, Sul 2.2 EARNINGS AND THE WAGE BILL IN PUBLIC AND PRIVATE SECTORS -.2.a Earnings, Average real monthly earnings by sector were computed using PNAD surveys and the INPC deflator. Between 1981 and 1995, real monthly earnings for increased about 29% for federal employees and 8% for state employees, and fell by 12% for municipal employees. Earnings for judicial and legislative employees rose by more than 40% during these years, by about 7% for the military, and stayed roughly constant for education and health workers in the public sector. In the private sector, earnings in agriculture fell by about 10%, in industry by 6%, but rose in distributive and productive services by 6%, and in personal services and construction about 16%. The largest increase in real earnings was therefore for judicial and legislative employees, and the largest fall was for municipal employees. 22. In absolute terms, and unadjusted for worker characteristics, earnings were highest for judicial and legislative workers in 1995, who earned almost R$1500 per month. Earnings were 25% lower than this for federal workers, 50% lower for state and military personnel, 62% lower for education and health workers, and 75% lower for municipal workers. In the private sector, earnings were highest in productive services at about R$950 per month. Earnings in distributive services and manufacturing were about 45% lower than this, those in personal services and construction about 60% lower, and those of agricultural workers about 75% lower.

13 - 9- Table 2.5: Average Monthly Earnings, All Brazil, in constant September 1995 reais* Public Employment Direct Administration Federal State Municipal Education & Health Judiciary & Legislative Military Private Employment* Agriculture Manufacturing Services Distributive Productive Personal & other Construction The deflator used is INPC (Brazil) -* Privatemployment includes employment in public enterprises. Source: Pesquisa Nacional por Amostra de Domicilios (selected years). 23. Considerable differences exist in average eamings in different parts of the country (Table 2.6). In 1995, average earnings in almost all sectors were highest in the federal district and Sao Paulo than in other states, and earnings in Pernambuco and Bahia were generally the lowest. Earnings in Minas Gerais were somewhat higher than for the two northeastern states, and earnings in Rio Grande do Sul and Rio de Janeiro higher still. Relative to salaries in the manufacturing sector, state employees received roughly 67-80% more in Bahia and Pemrnambuco, and 40-45% more in the other states. Municipal workers' earnings relative to private manufacturing earnings, in contrast, were highest in Rio Grande do Sul and Rio de Janeiro, somewhat lower in Minas Gerais and Bahia, and lowest in Pemambuco and Sao Paulo. In absolute terms, however, earnings of Sao Paulo's municipal workers were more than double those in Pernambuco. 24. In the private sector, relative earnings by sector are uniform across regions. The only exception is that in Rio Grande do Sul, eamings of agricultural workers is comparatively high. But this is hardly surprising; what is more striking is that sectoral earnings relative to those in manufacturing are similar for all states.

14 - 10- Table 2.6: Average Monthly Earnings, September 1995 Selected States, in current reais* Distrito Bahia Pernam- Minas Rio Rio de Sao Federal buco Gerais Grande do Janeiro Paulo Sul Public Employment Direct Administration Federal State Municipal Education & Health Judiciary & Legislative Military Private Employment* Agriculture Manufacturing Services Distributive Productive Personal & other Construction * The deflator used is INPC (Brazil) * Private employment includes employment in public enterprises. Source: Pesquisa Nacional por Amostra de Domicilios, b The Wage Bill, Using employment and average earnings for each sector, we computed the sectoral wage bill in Education and health workers, who are about 50% of all government workers, absorb about 45% of the wage bill. Municipal employees are about 20% of total public employment, but absorb only 12.5% of total wages; state employees, who are about 15% of employment, absorb about 17% of total wage. Federal and military workers are relatively small fractions of both total employment and wage bill. The most dramatic difference between these two shares is for judicial and legislative workers, who are about 5% of total employment, but almost 14% of the total wage bill. Thus, while employment of judicial/legislative workers may be relatively small, they are potentially critical for fiscal reasons. 26. To examine the trends in wage-related expenditures in the public sector, we computed the growth of the wage bill in the public and private sectors. Because of inaccurate weighting before 1991, wage bill figures are reliable only for the three PNAD surveys for 1992, 1993, and Average earnings grew in every part of the private sector, especially in personal and distributive services and construction. In contrast, industry and agriculture registered modest earnings growth, and employment actually fell in agriculture. In general, only employment growth in services exceeded the national average of about 2% during this period. In the public

15 sector, on the other hand, employment of federal, municipal, judicial and legislative, and education and health workers rose faster than this average rate. Only state and military employment grew slower than the national average. 27. In the public sector, the 20% annual growth of the wage bill for judicial and legislative workers dwarfs that of other workers. This growth in personnel expenditures was due to both an increase in their number (about 7.5% annually) and increased average earnings (about 12% annually). The growth of state wage bill exceeded 9%, despite a small decrease in state employment. The wage bill for education and health workers increased by 7.5% annually, due to increases in both average real earnings and employment. The federal wage bill also increased by 4.5% because of both new hires and higher wages. Military employment stayed roughly constant, but earnings increased moderately. 28. Without knowing what the wage gap between the public and the private sector is, however, it is difficult to determine whether earnings growth of government employees has been excessive. Thus, for example, the relatively high average earnings of judicial and legislative employees may be because of their higher skill levels relative to other public employees and workers in the private sector. And the relatively rapid growth in their earnings may be because they have been historically underpaid relative to their private sector counterparts. To examine whether or not public sector employees are under- or overpaid relative to those in the private sector, we compute the earnings gap in the next chapter. This would help in determining whether the solution to the problem of a high public sector wage bill lies in reducing public sector employment, or average earnings and pensions, or both.

16 PUBLIC-PRIVATE DIFFERENCES IN EARNINGS, PENSIONS, AND JOB STABILITY 29. In this chapter, we present the results of estimates of public-private differences in E, P, and T - as defined in Chapter I - for the country, and for selected states. We estimate the wage premium eamed by public sector workers over their private sector counterparts, and decompose this gap into the component due to differences in worker- and location-specific attributes and that which can be considered a purer premium. We also examine public-private differences in other job characteristics, primarily job security. Finally, we provide imputations of differences in pension levels and duration for public servants and private sector workers. 3.1 THE PUBLIC-PRIVATE GAP IN EARNINGS 30. In this section, we investigate the earnings gap between the private and public sectors in Brazil using PNAD data. 3.1.a. Approach The analysis consists of four steps: * We begin by estimating the overall wage gap between the private and public sectors for the entire Brazilian labor market. * In the second step, we take into account the heterogeneity within the public sector. We estimate the wage gap between the segments of the public sector and the overall private sector. Although these estimates for the overall wage gap represent an important starting point, they have serious limitations as a measure of the actual degree of labor market segmentation, since they do not necessarily represent the actual difference in payment between equally productive workers in the public and private sectors. * The third step is dedicated to analyze the impact of differences in the spatial location of the public and private labor force on the wage gap between the sectors. More specifically, we investigate to which extent the over-representation of public employment in the northeast tends to make overall measures of the wage gap between the public and private sectors an under-estimate of the actual gap faced by local labor markets. * In the fourth step we estimate the wage gap between workers with identical observed characteristics in the public and private sectors. To make this analysis empirically feasible we limit the scope to the six major Brazilian Metropolitan Areas and the Federal District. Since the wage gap between the public and private sectors is likely to differ according to the type of worker, in this fourth step we also investigate how the wage gap varies with the workers' main observed characteristics. 31. All the empirical analysis conducted in this section is based on the PNAD survey for September We used two measures of labor income. The first measure is the total monthly income normally received by a worker in his/her current main job. The second measure seeks to

17 - 13- standardized for the number of hours usually worked. It is defined as the total monthly income normally received in the main job, divided by the number of hours usually worked per week in main job and multiplied by a standard work week of 40 hours. Since the average number of hours worked by public employees tends to be significantly smaller than the corresponding average for workers in the private sector, it does make a difference for level of the wage gap between these sectors whether a standardization for hours of work is conducted or not. To ensure that the standardized wage is a good approximation of what worker would get if they work 40 hours a week, workers working less than 20 hours per week were excluded from the analysis. We also exclude workers with zero labor income. 3.1.b Tle overall wage gap 32. Table 3.1 presents estimates for the wage gap between the public and private sectors for Brazil. Average earnings tend to be much higher in the public than in the private sector. This table also shows that the wage gap between the sectors is much greater when labor income is not standardized for the number of hours worked, indicating that workers in the public sector work fewer hours per week than workers in the private sector. The gap in non-standardized wages tends to be between 15% and 20% smaller than the corresponding gap in standardized wages. The results using the standardized wages indicate that the relative wage gap between the sectors is equivalent to 33% of the average wage in the public sector or 48% of the average wage in the private sector. Table 3.1: Measures of the Public Sector Wage Premium Unadjusted for Sector, Region and Worker Attributes, Percent Measure Public-Total Private Public-Formal Private Adjusted* Not adjusted Adjusted Not adjusted Relative Wage Gap Baseline: Public Sector Wage Baseline: Private Sector Wage Logarithmic Measure of Gap** Log average-wages Average log-wage * Adjusted refers to earnings adjusted for differences in hours worked. ** These measures refer to the wage gap relative to a weighted average of the levels of wages in the two sectors, and have the advantage of not being dependent on whether the comparator is the public or private sector wage level. 33. Table 3.1 also reports estimates of the gap between the wage of employees in the public sector and employees in the privateformal sector, which may be the part of the private sector that those leaving government employment are more likely to seek employment in. This table reveals that this wage gap is smaller than the overall standardized gap between the public and private sectors. At this stage of the analysis, it is very difficult to interpret these large wage gaps. Three factors complicate such a comparison: first, there are large differences in average earnings within the public (e.g., between municipal and legislative workers) and private sectors (e.g., between agricultural and service sector workers); second, there are regional differences in average earnings (e.g., between state employees in Pernambuco and Sao Paulo); and third, there are large differences across workers by individual attributes (e.g., workers in the public sector tend to be

18 - 14- better educated and older). Henceforth, only wage gaps adjusted for hours worked will be reported. 3.1.c. The heterogeneity within the public sector 34. Thus, the natural next step is to estimate the wage gap between workers in the public and private sectors with identical observed characteristics. Nevertheless, before we adjust for worker-specific characteristics, we examine the degree of wage heterogeneity within the public sector. To do this, we use two alternative decompositions. The first uses a desegregation based on the sector of economic activity where we distinguish between the three levels of public administration (federal, state, and municipal), public health and education, and the legislative and judiciary activities. This desegregation of the public sector is not exhaustive since it cannot distinguish employment in public enterprises from private employment. The second decomposition, which is based on information on both the type of labor contract and on the level of government, is exhaustive and permits decompose the employment at each level of government into public servants and employees with and without a signed work card. Table 3.2: Measures of the Public Sector Wage Premium Unadjusted for Region and Worker Attributes Measure Relative Wage Gap(%): Relative Wage Gap(%): Baseline:Public Wage Baseline:Private Wage By Sector of Activity: Federal Administration State Administrations Municipal Administrations Judicial & Legislative Military Education and Health By Class of Worker: Federal Public Servants Federal other, with signed card Federal other, no signed card State Public Servants State other, with signed card State other, no signed card Municipal Public Servants Municipal other, with signed card Municipal other, no signed card Military * All estimates are for earnings adjusted for differences in hours worked. 35. Table 3.2 presents estimates for the wage gap between segments of the public sector and the overall private sector. This table uses both desegregation procedures for the employment in the public sector. The results indicate a great level of heterogeneity within the public sector, with the average wage being much higher in certain segments of the public sector than in others. Wages are particularly higher in federal administration, in particular, among those in the legislative and judiciary sectors. In these sectors, average wages are between 150 and 250% higher than the average for the private sector. At the other extreme, workers in the health and education activities, in particular those at the municipal level have earnings that are close to the

19 - 15- average for the private sector or even below it. The average wage in municipal administration is 16% smaller than the average in the private sector. Occupying an intermediate position are the state administration and military personnel, for whom wages are close to 70% above the average in the private sector. 36. Similar results are obtained using the alternative desegregation of the public sector based on the nature of labor contracts and level of government. Wages are particularly high among federal public servants, and employees in federal enterprisesl. In both cases the wage gap is greater than 150% of the average wage in the private sector. For military personnel, state public servants and employees of state enterprises, the average wage gap is about 75% to 100% of the average wage in the private sector. The average wage of municipal public servants and employees of municipal enterprises is below the average for the private sector, with the wages of employees in municipal enterprises being particularly low. 37. Overall, the average wages in all segments of the public sector tend to be above those paid in the private sector, the only exceptions are the wages paid by municipal administrations and enterprises. But these comparisons also reveal a large degree of heterogeneity in wages among subsectors within the public sector, with the average wage in federal administration being more than 200% higher than the corresponding average paid in municipal administrations. 3.1.d Regional differences in the public-private wage-gap 38. In previous sections we have considered the magnitude of the overall wage gap between the public and private sectors. This overall gap, however, captures both intrinsic differences in wages between these sectors and differences in the characteristics of workers in the two sectors. In this section we consider the role of spatial differences in the distribution of public and private employment. More specifically we investigate three topics: first, we investigate whether public employment is over-represented in the poorest states; second, we investigate how the uneven distribution of public employment across states affects the overall level of the wage gap between the public and private sectors; and finally we investigate how the within-state wage gap between the public and private sectors varies across states. In this subsection we only consider the hoursstandardized measure for wages. 39. As far as differences in the spatial distribution of the labor force are concerned, this table reveals that public employment tends to be only marginally over-represented in areas with lower average wage (the northeastern states). In fact, public employment is over-represented in the Federal District and in Rio de Janeiro, which are among the states with the highest average wage. Overall, however, employment in the public sector is slightly over-represented in the poorest areas: 49% of public employment and 46% of private employment are in states with average wage below average. The results indicate that the gap in average log-wage would increase from 44% to 46% if public employment had the same spatial distribution as private employment, so that public employment has a "locational disadvantage" of only 2%. 40. Table 3.3 presents alternative measures for the wage gap for the entire country and for the states and metropolitan areas chosen for this report. This table reveals that the wage gap tends to be considerably higher for the metropolitan areas than for the corresponding states. This is For simplicity we identify workers in the public sector with signed working card with employees in pubiic enterprises.

20 - 16- especially true in the northeastern states of Pemambuco and Bahia, and in the State of Minas Gerais. Sao Paulo is the only state where the wage gap for the metropolitan area is smaller than for the corresponding gap for the entire State. The average gap between the public and private sectors is much higher in Brasilia than in any state. The average wage gap tends to be high in the northeast. and lower in the more developed states in the South and Southeast of Brazil, particularly in Sao Paulo, Parana and Santa Catarina. But it is also very low in a few poor states like Ceara and Alagoas (see tables in Statistical Annex). Table 3.3: Measures of the Public-Private Wage Gap in September 1995, By Location Unadjusted for Sector and Worker Attributes, Percent* Baseline: Baseline: Private Wage Public Wage All Brazil Distrito Federal Metropolitan Area Pernambuco State Recife Bahia State Salvador Minas Gerais State Belo Horizonte Rio de Janeiro State Metropolitan Area Sao Paulo State Metropolitan Area Rio Grande do Sul State Metropolitan Area All estimates are for eamings adjusted for differences in hours worked. 3.1.e The role of differences in worker characteristics 41. In this section we investigate the role of differences in the characteristics of workers between the public and private sectors in explaining the wage gap between the two sectors. The fundamental question being answered here is: what fraction of the wage gap between the two sectors is simply due to sectoral differences in the characteristics of the labor force employed in the two sectors? 42. To investigate this question we consider the labor market of the major six Brazilian metropolitan areas and Brasilia. The restriction of the analysis to these seven well-defined local labor markets is useful to isolate the wage gap between the public and private sectors from possible spatial differences among local markets. This restriction, however, also has some

21 - 17- important disadvantages. The major disadvantage is the fact that the nature of public employment in large metropolitan areas is likely to be different from smaller urban areas. The contribution of federal and state employment is likely to be greater. Moreover, municipal jobs in metropolitan areas are more likely to be better paid. As a consequence, the wage gap estimated for these areas is likely to over-estimate the gap for the entire urban labor market in the respective state. 43. The basic set of observed characteristics includes gender, race, schooling, and age. We also work with a version of this set that includes tenure at the current job. Workers in the public sector tend, on average, to have been in their current job for a longer period of time. Since to a considerable extent this is not a result of any differential merit between employees in the public and private sectors, but one of the major advantages of jobs in the public sector, it is unclear whether we should control for this characteristic. Hence, we present all results including and excluding tenure from the analysis. 44. The methodology used to compare the wage of observably identical workers in the public and private sectors consists of three steps. In the first, we regress the log-wages of workers on their characteristics and a dummy variable for whether they are in the public or in the private sector. In the second step, based on the results of these regressions, we compute what would be the average wage of public employees if they were in the private sector, given their personal characteristics. Finally, we obtain the wage gap between workers with identical observed skills as the difference between the actual average log-wage in the public sector and the average log-wage they would receive if they were in the private sector, as estimated in the second step. 45. Table 2.5 in chapter 2 showed that differences in the composition of the labor force are a major explanatory factor of the wage gap between the public and private sector. These differences explain almost all the wage gap between the public and private sectors, revealing that a large fraction of the overall wage gap between the public and private sector just reflects differences in the skill composition of their labor force. Therefore, overall measures for the wage gap that do not control for differences in the characteristics of the labor force are misleading indicators of the actual wage advantage of workers in the public sector. 46. About 70% of the gross wage gap is explained by the higher educational level of the labor force in the public sector. The remaining 30% is explained by the fact that public employees tend to be older and have longer tenure, with each of these two factors being responsible for 10% to 20% of the differential. The impact of the differences in the composition by gender and race is small. Differences in race composition between the private and public sectors are very small and, as a consequence, they have a negligible impact on the wage gap. The impact of gender differences is also small because though women have lower wages than men in either sector, they tend to be over-represented in the public sector which pays higher wages. 47. In sum, differences in the sectoral composition of the labor force are a major explanation for the wage gap between the public and private sector. Once the differences in worker attributes have been eliminated, the high wage gap between the two sectors becomes smaller but remains considerable for most states: * In Brasilia, the wage gap declines to a still very large public-private wage gap of 63% among workers withi identical observed characteristics. * In Recife and Salvador, workers in the public sector receive salaries that are 32% and 21% higher respectively.

22 - 18- * The public-private wage gap is 17% in Belo Horizonte, 14% in Porto Alegre, and 12% in Rio de Janeiro. * Only in Sao Paulo is the pattern reversed: among equally qualified workers, workers in the public sector receive salaries that are 13% lower than private sector salaries. 3.1.f Controlled wage gaps by sector and class of worker 48. In this section we estimate the wage gap between each segment of the public sector and overall private sector controlling for differences in worker-specific attributes. This is perhaps the best estimate of the "pure" premium enjoyed by public sector employees over what they would have earned had they held private sector jobs. Table 3.4 presents estimates for the controlled wage between each segment of the public sector and the private sector for the country as a whole. Table 3.4: Measures of the Public Sector Earnings Premium, September 1995 Adjusted for Worker Characteristics*, by Sector/Class, Percent Measure Monthly** Hourly** By Sector of Activity: Federal Administration State Administrations Municipal Administrations Judicial & Legislative Military Education and Health By Class of Worker: Federal Public Servants Federal other, with signed card State Public Servants State other, with signed card Municipal Public Servants Municipal other, with signed card All workers without signed card Military * Worker characteristics adjusted for are age, sex, race, tenure, and education. ** Monthly earnings premia are not adjusted for differences in hours worked. Hourly earnings comparisons are adjusted for differences in hours worked. 49. The table reveals large differences among segments of the public sector with respect to their wage advantage. The wage advantage, measured by the controlled log-wage gap, is larger at the federal level for both among public servants and among employees in public enterprises. For these groups the average salaries are 35% to 65% higher than for comparable workers in the private sector. At the state level, the wage advantage is close to zero for public servants; for employees in public enterprises the wage advantage is positive and between 10% and 22%. At the municipal level, public servants and employees in public enterprises get lower wages than workers with similar observed characteristics. Overall, the evidence corroborates the existence of significant wage advantage of some segments of the public sector relative to the private sector.

23 PUBLIC-PRIVATE DIFFERENCES IN JOB SECURITY 50. Table 3.5 presents the average tenure on the current job in months, for the seven metropolitan areas. The table presents the average tenure in the public and private sectors, the "raw" gap in average tenure, and then the gap controlling for differences in worker-specific attributes. Not surprisingly, it appears that public sector workers have more job security, as measured by the tenure at the current job, adjusting for characteristics such as age, education, and gender. This is only a crude measure of job security, because it does not measure expected tenure. Nevertheless, it does suggest that public sector workers, besides earning higher wages in all states except Sao Paulo, also have jobs that are more secure. Estimates of the public-private gap in job security range between 50% for the federal district, and between 20-35% in the six states. Table 3.5: Measures of the Public-Private Tenure Gap, By State In September 1995 Metropolitan Region Public Private Gap Controlled Gap* Average Tenure (Months) Distrito Federal Pernambuco Bahia Minas Gerais Rio de Janeiro Sao Paulo Rio Grande do Sul * Estimates are for differences for workers with identical characteristics This advantage of public sector workers appears to be strengthened by the lower unemployment rate of workers, classified by their sector of previous employment. But this pattern is reversed when worker-specific attributes are controlled for. Combined with the finding for average tenure differences, this may indicate that while public sector workers are less likely to lose their jobs, they are more likely to stay unemployed if in fact they do. 3.3 PUBLIC-PRIVATE DIFFERENCES IN PENSIONS 52. The calculation of public private differentials is somewhat complicated, and we present the technique only in outline here, focusing the discussion on results instead. To estimate differences in pensions for public servants and those receiving social security benefits from the INSS Time of Service and the Old Age schemes, we use the following technique: * First, using 1995 PNAD survey data for the seven metropolitan areas used in this report, we estimate the profile of monthly earnings for male and females, divided further into four worker categories (public servants, private com carteira, private sem carteira, and private selfemployed) and three education groups (0-8 years, 9-11 years, and 12+ years of schooling).

24 * Second, using these estimated earnings profiles, and assuming that all males retire after 35 years of service, and all women after 30 years, we estimate the average pension levels for each group. For public servants, we use the 36th year's salary in the 36th year as the estimated monthly pension (because pensions for public servants are based on the last month's salary). For com carteira salaried workers and the selfemployed, we use the 34th year's salary as the estimated pension level (because INSS time of service benefits are based on salary levels in the last 36 months of service). Note that we make the assumption that selfemployed workers do actually contribute to and receive social security. Forsem carteira workers, we assume that these workers will receive minimum pensions from other INSS programs: their pension levels are assumed to be equal to one minimum wage in 1995, which we assume to be R$ 100. * Third, we make adjustments for the fact that INSS pensions are capped at 10 minimum wages. This results in pensions for com carteira men with 12+ years of education being restricted to equal R$1000. * Fourth, using weights obtained from nationwide PNAD data for 1995 on the shares of com carteira and sem carteira employees, and selfemployed workers, we compute the average expected pensions for private sector men and women. We do not take into account pensions from the supplementary social security system (i.e., closed or company funds). * Finally, these numbers are compared with expected monthly pensions for public servants with the same education level. 53. Table 3.6 reports the results of these estimations. The results show that due to the nature of the estimated profiles and differences in rules basically the pension ceiling of 8 minimum salaries, public-private pension differences are greater for men than for women, and greatest for the most educated male workers. Thus, the average CLTista male worker with 0-8, 9-1 1, and 12+ years of schooling will receive, respectively, 75%, 72% and 50% of the monthly pension level of his estatutario counterpart. The corresponding numbers for women are 73%, 73%, and 71%. The findings on private pay and pension differences, combined with the fact that pension contribution rates are 0-2% for public servants but range between 8-11% for CLTistas, have important implications for the structure of optimal severance packages, and the likely success and full costs of current voluntary severance programs.

25 Table 3.6: Estimated Monthly Pensions, September 1995 Reais Gender/Education Public Private With Card Without Selfemployed Total* Card Males 0-8 years schooling years schooling years schooling ** n.a. n.a Females 0-8 years schooling years schooling years schooling Source: PNAD Surveys, IPEA-World Bank calculations. Men are assumed to retire after 35 years of service, and women after 30 years. * Using weights derived from nationwide PNAD employment distribution data. ** Capped at 10 times the minimum monthly salary level, assumed to be R$100 in September n.a. signifies that the sample was too small to permit reliable estimation of experience-earnings profiles.

26 PUBLIC EMPLOYMENT REFORM: WHAT IS BEING DONE? 54. This chapter discusses what is being done to address the problem of high personnelrelated expenses at the federal, state, and municipal levels. Principally, this has involved voluntary severance programs in public administrations and enterprises, and the use of attrition to reduce employment levels. We analyze ten voluntary severance programs in public administrations and enterprises launched since 1995, and the concurrent measures taken by these agencies/enterprises. 55. We examine the costs and effectiveness of voluntary severance programs (called programas de demissao voluntaria or PDVs) in tackling the problems of a growing public salary and pension bill. The programs selected are from the states of Sao Paulo, Rio Grande do Sul, and Pernambuco, thus representing a broad economic spectrum; also included is the Federal Government and Banco do Brasil. Not by coincidence, public private earnings and employment differentials were also examined for these three states and the federal district in Chapters 3 and 4. The programs were chosen so that the sample was representative: four of the programs are for public administration, and six are for state-owned enterprises (three banks and three public utilities). The three public enterprises are FEPASA - Sao Paulo railways company, CEEE- Rio Grande do Sul electricity company, and CESP - Sao Paulo electricity company. The public sector banks are included in the study: Banco do Brasil; BANDEPE - Bank of the State of Pernambuco; and BANRISUL - Bank of the State of Rio Grande do Sul. 4.1 OVERVIEW: VOLUNTARY SEVERANCE PROGRAMS 56. Tables 4.1 and 4.2, which summarize the results of the detailed study of these programs prepared for this report, can be used to identify the performance of the PDVs which have been implemented. A detailed description of these programs can be found in Carneiro and Gill (1997). Overall, state enterprises and banks have had more volunteers for their severance programs as they were able to offer more attractive packages than those offered by federal and state governments. 4.1.a Take-up rates. 57. The federal government's program had the lowest take-up rate (1.7%), followed by Sao Paulo administration (2.2%). Pernambuco and Rio Grande do Sul administrations obtained more than double these rates: the ratio of total job separations to baseline employment was about 5% and 6.5%, respectively. Amongst public enterprises, FEPASA managed to reduce employment by almost 27% while CEEE's ratio was only about 6%. Official banks, with substantially greater financial incentives, averaged higher take-up rates. Bandepe had the highest take up rate of about 34% of pre-program employment among the PDVs analyzed. Banco do Brasil averaged about 12% and Banrisul 10.5%. Overall, governments aimed to lay off low-skilled and negligent workers while the main target of official banks were workers with above average tenure (i.e., older workers). Public enterprises simply aimed at reducing baseline employment without specifying any particular target group.

27 b Main design features. 58. Financial incentives varied from case to case but were usually based on the number of years of service. BANDEPE was the exception to this rule, paying only the equivalent to four gross monthly salaries to those who joined the voluntary separation program, regardless of years of service. Enrollment for the programs in the federal and state administrations was treated as a voluntary quit or a justified dismissal. As a consequence CLTista workers leaving public employment through the PDVs were not paid the 40% penalty over the balance of their FGTS accounts. On the other hand, public enterprises and official banks have treated voluntary separations as unjust dismissals and paid substantially higher financial incentives relatively to the ones offered by public administrations. Due to a collective bargaining agreement, FEPASA paid 180% of FGTS account balances for PDV takers. Table 4.1: Voluntary Severance Programs: Size and Target Group Total Takers Total as % of Target Group Avg. Tenure of (Thousands) Baseline Employment PDV Takers (Years) Administrations Federal Workers with less than n.a. secondary education Pernambuco Mild pressure on 10 negligent workers Sao Paulo No target group 15 Rio Grande do Sul Pressure on negligent n.a. workers Public Enterprises FEPASA CEEE No target group 6.33 No target group - aim n.a. 22 CESP 0.49 was to have 10% cut n.a. No target group n.a. Public Sector Banks Banco do Brasil Workers with more than 16 Bandepe years of service, and in branches to be closed Workers with more than 10 years of service and 10 poor performers Banrisul No target group - aim to have 10% cut c Average benefits. 59. Among public enterprises and official banks, Banrisul paid the highest package per person (R$58,700) followed by Banco do Brasil (R$50,000) and CEEE (R$41,885). Among public sector enterprises, average benefits were lowest for Bandepe (R$21,930). The federal government offered the highest package amongst public administrations paying roughly R$23,500 per person; Rio Grande do Sul paid an average of R$17,055 while Pemambuco and

28 Sao Paulo paid R$10,665 and R$5,500, respectively (see Table 4.2). Nonfinancial assistance for PDV takers was offered in all cases except FEPASA. The most common form of assistance was extended medical benefits (usually for one year after leaving employment), training programs, and help with business start-ups. Banco do Brasil contracted a firm specializing in job search assistance to help PDV takers to find other jobs while Bandepe only offered special loans for PDV takers interested in starting small businesses. Table 4.2: Financial Costs and Recovery Periods Total Costs Unit Costs Monthly Recovery (R$ million) (R$ thousand) Wage Bill Period (R$ million) (Months) Administration Federal Pemambuco Sao Paulo Rio Grande do Sul Public Enterprises FEPASA CEEE CESP n.a. n.a. n.a. n.a. Public Banks Banco do Brasil Bandepe Banrisul d Totalfinancial costs, savings, and reported recovery periods. 60. Table 4.2 provides a summary of financial costs of PDV programs. The biggest outlay was for Banco do Brasil program which had a total cost of R$350 million, and the smallest was that of Bandepe at R$30 million. Following Banco do Brasil, other large programs were those of Rio Grande do Sul (R$216 million), the federal government (R$183 million) and FEPASA (R$135 million). The highest saving in monthly wage bill was obtained by Banco do Brasil which managed to reduce its monthly payroll by some R$35 million. The lowest saving was achieved by Pernambuco State government (R$3.3 million). The federal government's PDV program has had limited success in reducing payroll expenditures with the monthly wage bill for PDV takers amounting to R$9 million (out of a total monthly payroll of about R$800 million). In the case of Pernambuco, considerable delays in releasing the financial benefits due to PDV takers undermined the credibility of the program and limited the number of volunteers. 61. With regard to thle period necessary to recover PDV expenses - as reported by the enterprise/agency - public sector banks presented the best performance relative to both public enterprises and administrations. BANDEPE, for example, estimates that it would recover PDV spending in 5 months; for Banco do Brasil, the recovery period was estimated at 10 months and for BANRISUL at 19 months. In the case of public enterprises, the average recovery period was estimated by its administrators at 24 months for FEPASA and 14 months for CEEE. Amongst state governments, Sao Paulo had the shortest recovery period of about 5 months. Pernambuco and Rio Grande do Sul report recovery periods of 12 and 16 months, respectively, and the federal government's program had an estimated recovery period of about 20 months.

29 CRITICAL FEATURES OF PUBLIC ADMINISTRATION PDVS 4.2.a Thzere is little or no pressure on employees to take the program. 62. In general, state governments do not apply pressure on redundant employees to accept severance packages. It cannot be fully determined whether this was due to legal reasons (i.e., it was unlawful for these programs to have even a semblance of coercion for estatutarios) or because of a lack of political will on the part of program administrators ("cultura"). But the relative success of the Rio Grande do Sul program indicates that political will is the more important factor. This program is exceptional among the public administration PDVs in that negligent or redundant workers were explicitly targeted by the program and pressured into leaving (e.g., by institutinig a new system for monitoring attendance for all workers, informing workers whose performance was less than satisfactory that they were being monitored and by reassigning them to other jobs - see Box 1). The program had the highest take-up rate (about 7% of baseline employment), both because of this pressure and because the payments were somewhat higher than those of other administrations. 4.2.b Nontenured employees obtain better severance packages than those with tenure. 63. In most cases, CLTista employees of public administrations - who technically do not have tenure or "estabilidade"- obtained higher severance benefits than tenured estatutarios. This is because while special incentives (indemniity payments) under the schemes were the same for CLTistas and estatutarios, CLTistas also obtained at least 100% of accumulated FGTS balance (i.e., the separation was treated as a justified dismissal). For example, in the Sao Paulo program, any worker with a monthly salary of $500 and ten years of service got R$6,250 if s/he volunteered during the first two weeks (one month's pay for each year of service, plus the 25% early volunteer bonus). But if the worker was a CLTista, s/he would also receive more thani R$5,000 in FGTS benefits, with the exact figure depending upon the profile of earings over the last ten years. The Rio Grande do Sul state program was exceptional in that it did not allow participants to withdraw funds from their FGTS accounts, i.e., treated the separation as a voluntary quit. 4.2.c Wlhen potentiarl pension payments are consilerel, benefits of state PD V programs are greater thlan if only saving in salanries is considere(l. 64. Government saves more when an estatutario takes a PDV package than when a CLTista leaves. This is because public pension benefits are more generous than INSS pensions. Figure 2 illustrates a methodology for calculating the reductions in salary and pension expenditures when an estatutcario leaves public employment to work in the private sector. Using this methodology, we estimated - under reasonable assumptions - the saving in salaries and pensions for groups of workers according to their sex, education, and tenure. Table 4 lists results of this exercise for six groups: men and women with, respectively, 0-8, 9-11, and 12+ years of education. In the case of men, it is assumed that the PDV taker has 17.5 years of service. In the case of women, this is assumed to be 15 years. These numbers are similar to actual tenure levels of participants in these programs. When pensions are considered, the savings due to an estatutario leaving government employment are about 12%, 10% and 23% higher for men with 0-8, 9-11 and 12+ years of education. For women, the corresponding numbers are 25%, 21%, and 21%. Pension-related savings are greater for highest wage males because private sector pensions are capped at 8 times the minlimum salary, and the cap is binding only for this group. Other than this group, savings

30 are greater for women because the expected length of retirement is greater for women. These numbers would be even greater if we take into account the fact that while pension contributions are zero in the public sector, they can be as much as 20 percent of earnings for private employees. 4.2.d Tle severance amounts paid are small relative to saving in payroll expenses. 65. The most striking finding is that the potential benefits of PDV programs - in terms of reduced salary and pension bills - dwarf the costs of existing programs. While the potential saving for these groups ranges between R$34,000 (for least educated women) and R$265,000 (for educated men), the average benefits being paid range from $5,000 and $25,000. One of the main reasons is that governments are cash-strapped, and cannot afford to pay much more. As a consequence, though, take-up rates are low, and a promising source for reducing the long-term fiscal burden on state and federal governments remains largely untapped. Table 4.3: Implications of Reducing Public Employment (Average earnings, transfer of obligations, and DPV* of savings, in thousand reais) Education Consolidated Government Transfer of StateGovernment Level Saving Obligations to Saving INSS** Salaries Pensions Total Pensions Total Males, with 17.5 years of service*** 0-8 years years years Females, with 15 years of service*** 0-8 years years years These estimates are based on estimated earnings profiles of private and public sector workers in a nationwide sample drawn from the 1995 PNAD survey. * A 6% discount rate was used to calculate present values. ** These transfers do not take into accounthe expected contributions of these workers (and their employers in the case of salaried employees) to the INSS system. *** Men are assumed to retire from public employment after 35 years of service, and receive pensions for 20 years; women are assumed to retire after 30 years of service and receive pensions for 25 years. 4.2.e Voluntary separation programs for estatutarios involve large state-to-feleral government transfers of pension obligations. 66. For estatutarios who leave under these programs but do not qualify for early retirement, there is an uncompensated transfer of pension obligations from the state administration to the national INSS scheme. This is because these employees carry over their years of service into the private system and, after a relatively short period of contributions, become eligible for private pensions. Thus a male PDV participant with 17.5 years of service would be eligible for reduced INSS pensions after only 12.5 more years of work in the private sector, and for full pensions after only 17.5 years. The present value of these obligations for men and women leaving public

31 employment mid-career are about R15,000 to R$17,500 for low wage workers, R$25,000 to R$30,000 for medium wage workers, and R$42,500 to R$55,000 for high wage workers. Thus, for example, a state PDV for 10,000 estatutarios drawn from all education/salary categories could imply a transfer of obligations of more than R$300 million in current terms from the state to the federal pension system. RIO GRANDE DO SUL STATE REFORM In Rio Grande do Sul, the voluntary separation program was launched with the explicit objective of reducing personnel expenditures, which had risen to more than 80% of net revenues in Concurrent measures included a cut of 25% in the number of commissioned positions, phasing out some activities and the payment of the 13th salary in two installments. The program was open from April 16 to May 17, Before the program was actually launched, the state government put some pressure on workers it would like to see joining the PDV. The strategy of pressure included the transfer of workers to functions which they were not used to, sending out short-notice letters ("aviso-previo") to targeted workers, implementation of an electronic time monitoring system, revoking the concession of unpaid leave and special privileges for a period of two years, and defining targets to be met by employees. The PDV helped to reduce the monthly wage bill by about R$5 million. The initial target of the PDV program was to make 13,700 workers redundant (10,300 estatutarios and 3,400 CLTistas); actual figures added up to 9,953 estaturarios and 2,718 CLTistas totalling 12,671 takers. On average, PDV takers (of which 56% were female) had 15 years of service and were 38 years of age. The costs for Rio Grande do Sul separation program amounted to R$216 million and the monthly wage bill for PDV takers was estimated at R$14.3 million. The average package per person totalled R$ 17,055 and the recovery period was estimated at 16 months. Volunteers were offered medical benefits for one year after separation, and subsidies for training courses provided by SENAI, SENAC and SENAR (for which the government paid up to R$300 per course). Targeted workers could take one of three options: (i) voluntary separation (Demissao Voluntdria); (ii) special leave to qualify workers for early retirement (Reconversdo Funcionao; and (iii) voluntary retirement (Aposentadoria Voluntdria): * In the case of voluntary separation, the taker would receive I month's salary per year of service, capped at 20 monthly salaries; there was an extra cash bonus of 25% of the benefit if adhesion to the PDV occurred between April 16-30, or 15% ifjoined between May * Under the special leave option, eligible workers were given paid leave of up to 5 years, plus an extra benefit of 20% of monthly salary. To qualify for this scheme, male workers had to have years of service, and female workers years of service. As soon as the worker qualified for a proportional retirement scheme, s/he would have to resume his/her duties and within two months apply for proportional retirement or return to the government all money received during the leave period. * Men with more than 30 years of service and women with more than 25 years could opt for the voluntary retirement scheme and receive an indemnity of 5% of remaining wages; that is, 5% of last monthly salary multiplied by the number of months remaining for retiring on full pension. Box 1

32 f Expected social security contributions of PDV participants do not make up for the transfer ofpension obligations. 67. To some extent, this transfer of pension obligations would be offset by payroll contributions. But under the current rules, the INSS system never recovers the "lost contribution" of these estatutarios, viz., what they and their employers would have contributed had they belonged to the INSS system during their years of service with the government. If one assumes that the INSS system is actuarially fair for workers who make full contributions for 30/35 years, this means that the INSS is "owed" employee and employer contributions (which add up to about 30% of earnings) for their years of tenure. The estimated present values of "lost" contributions are R$16,000, R$42,000, and R$78,000 for men with 0-8, 9-11 and 12+ years of schooling who take PDV packages after 17.5 years of service; and R$13,000, R$25,000 and R$54,000 for the corresponding groups of women with 15 years of service. Naturally, these numbers would be smaller for PDV takers with lower tenure levels, and greater for those who have been civil servants for longer. 4.2.g Administrators are aware of the problem of adverse selection, and have taken some measures to address it. 68. Despite being legally or constitutionally constrained, program administrators have taken some measures to counter the problem of "good" workers leaving and "bad" workers staying on. The main instruments were encouraging some workers to volunteer for the program, and by discouraging "good" workers who had volunteered from leaving, and even refusing them the package. Among these two, the latter appears to have been the more effective way to address this problem. 4.3 CRITICAL FEATURES OF PDVS OF PUBLIC ENTERPRISES 69. Matters are both simpler and more complicated for calculating the saving from PDVs in public enterprises. They are simpler because these workers are all CLTistas, so the issue of transfer of pension obligations from states to the federal government does not arise. They are more complicated because many public enterprises have "closed" or company pension funds (called the "complementary" social security system), which give generous benefits that imply increasing company contributions. Severance schemes that entail shedding company responsibility to pay pensions thus imply a saving for the government. But because these closed pension schemes vary from company to company, we do not address this problem here. 4.3.a Generally, there is little or no pressure on employees to leave. 70. As with PDVs of public administrations, state enterprises generally do not apply pressure on redundant employees to accept severance packages. In this case, however, it is clear that this is not due to legal reasons: the law permits firing of public employees because they arecltistas. So the voluntary nature of downsizing programs is obviously because of political reasons. Bandepe's program is exceptional among the public administration PDVs in that negligent or redundant workers were explicitly targeted by the program and pressured into leaving (mainly by instituting an evaluation system to identify unmotivated or mismatched workers, who were advised to take severance packages - see Box 2). Despite having the lowest average indemnity payment, this pressure resulted in Bandepe's program having a high take-up rate of all PDVs (about 33% of baseline employment).

33 BANDEPE'S PROGRAMA DEMISSAO INCENTIVADA The redundancy program at Bandepe (the official bank of the State of Pernambuco) was implemented between February 21-28, The main objective of the program was to reduce personnel expenditures. Bandepe carried out an extensive performance evaluation of all its employees and those with either unsatisfactory performance or a poor match with the bank's planned operations were urged to take redundancy packages. The bank used pressure to get workers to join the redundancy program, mainly by signaling that further personnel cuts were planned as part of ongoing restructuring of operations. Total employment at Bandepe before the program was launched was 3,268 employees. The program targeted 2,000 employees, most of them with more than 10 years of service and/or poor performance. 1,117 employees enrolled in the program, and the reduction in the monthly wage bill obtained was about R$5 million, which represented a cut of 46% in payroll. 40% of those who joined the redundancy program were women. Program participants were generally older than 30 years with more than 10 years of service. The average package paid was R$21,930 and the total cost of the program amounted to R$30 million. The monthly wage bill for PDV workers was R$5 million, so the estimated period for the recovery of program expenditures was about 6 months. Separations in the context of the redundancy program were treated as unjust dismissal and thus workers received an additional 40% of their FGTS account. The financial benefit offered to leavers was 4 gross monthly salaries, which were paid in 18 months: the severance package made no distinction between workers with different years of service. The only tenure-related component was due to the balance in FGTS accounts. Other benefits included medical benefits and foodstuff allowance for a period of 18 months after leaving employment. Special loans for small businesses (totaling R$1.3 million) were provided by the bank as part of an assistance program. Special courses on setting up own businesses were also offered with the support of SEBRAE, SENAI and SENAC. These were not very popular. For those workers who would have retired within 18 months following the redundancy program, the bank paid the employer's share to BANDEPREV (the company pension fund) up to retirement; for those retiring in the following 3 years, Bandepe paid both shares from the 18th month up to the 36th month prior to retirement. Workers taking up INSS early retirement received 2 gross monthly salaries as bonus; if s/he was under the age of 55, s/he received 80% of her/his gross monthly salary, and the also bank paid the employer's share to BANDEPREV until legal requirements to receive full pension were met. Box b Severance programs always provide more than what is legally required. 71. For employees of public enterprises, the FGTS scheme provides the mandated minimum benefits to be paid to workers upon dismissal. Collective bargaining agreements - e.g., in the case of FEPASA - specify additional payments. But in all the schemes studied, workers were paid indemnity payments in addition to their FGTS-related benefits. These payments were usually a multiple of monthly wages: thus higher wage workers got considerably more than workers with low salaries. The payments also generally increased with years of service, implying added compensation to what they already received under the FGTS scheme. The only case where workers received indemnity payments that were not related to their tenure was Bandepe's program: not coincidently, this program also was the only one where participation was not entirely voluntary.

34 c Voluntary separations are treated as unjust dismissals. 72. Despite there being little or no pressure on employees to participate in PDVs, those volunteering to leave were treated as though they had been dismissed sem justa causa As a result, they were allowed to withdraw their FGTS balances, and get 40% more as a penalty from the firmn. In FEPASA's case, where participation in the program was somewhat involuntary, a collective bargaining agreement resulted in the penalty being doubled, so that participants got 180% of their FGTS balances, plus special indemnity payments. 4.3.d Severance payments are considerably greater than in PDVs of public administrations. 73. As Figure 1 shows, the average benefits as a multiple of monthly wage was considerably higher for public enterprises than for administrations. The exceptions are the federal administration - where benefits were 24 times the average monthly salary and hence greater than PDVs of public enterprises - and Bandepe, where benefits were only five times average monthly salaries and hence smaller than any other PDV (these benefits do not include the FGTS account, and the 40% penalty). Participants in PDVs of public enterprises also had higher average salaries than those of administratiot,s. This is consistent with the greater disincentives facingestatutarios who expect to earn very high pensions to leave voluntarily, due to the ceiling (eight times the minimum salary) on private sector pensions being binding. 4.3.e Severance amounts paid are small relative to saving in payroll expenses. 74. As with the PDVs of state administrations, the potential benefits of PDV programs - in terms of reduced salary bills - appear to be significantly greater than the costs of existing programs. While the potential saving ranges between R$27,000 (for least educated women) and R$215,000 (for better educated men), the average benefits being paid range from $22,000 and $52,000, a large part of which are from FGTS balances and not government or company funds. 4.4 IMPLICATIONS FOR FUTURE PDVS 75. The range of experience with voluntary severance programs within Brazil is broad. No two programs were alike, so it is difficult to generalize and draw implications for future programs. Nonetheless, some findings appear to be quite general and, if interpreted carefully, will help in guiding future efforts at downsizing. The main implications of these findings for future PDVs are: 4.4.a Downsizing is a financially profitable strategy. 76. Under reasonable assumptions, financial benefit-to-cost ratios exceed ten for many worker categories. Current benefit levels imply that the costs of severance packages are recovered within three years. When both the saving in salaries and pensions are considered, few investments are as rewarding as downsizing programs, even under fairly constrained legal circumstances. 4.4.b Reducing salaries has a much smaller financialpayoffthan reducing employment. 77. Comparing the reduction in payroll expenses of lowering employment by one person, with the alternative of lowering the same worker's salary by 10%, we find that downsizing has

35 about 7.5 times the financial return, when severance payments are not considered. Even when severance payments twice as large (e.g., with payments averaging 48 months' salaries) as those found in the programs studied are used, downsizing is between 5.5 and 6.5 times more rewarding than a 10% cut in the worker's salary. 4.4.c costs. An element of involuntariness is necessary for increasing take-up and containing 78. The programs surveyed show that some pressure has to be used to ensure that the take-up rates are reasonably high, and that those leaving are largely those who should leave. For PDVs of both public administrations and enterprises, those that used pressure (Rio Grande do Sul administration and Bandepe) were most successful in reducing employment and payroll expenses. 4.4.d Even under the existing laws, it is possible to use pressure on negligent public employees to leave with modest severance benefits. 79. The cases of Rio Grande do Sul and Bandepe again show that constitutional reform is not strictly necessary to carry out a successful public sector downsizing operation. With careful planning and execution, both programs have reduced the number of workers and avoided litigation. The experience of state PDVs, combined with points 4.4.a and 4.4.b, implies that administrative reforms that are being debated in congress - while lowering the cost of downsizing somewhat - may not be strictly necessary for successfully reducing the public sector wage bill. 4.4.e Severance payments for CLTista employees are more than what is legally necessary, and those for estatutarios less than what is sufficient under current tenure and pension laws. 80. On equity and efficiency grounds, estatutarios should get at least as much as similarly qualified CLTistas. But because under current laws they cannot easily be fired, estatutarios have to be bought out. Other things being equal, therefore, estatutarios should be getting higher severance benefits than CLTistas. In fact, the opposite holds. States could reallocate some of what is now being spent on CLTistas to get estatutarios to volunteer for the program. While on equity considerations these payments should at least be equal, on efficiency grounds it makes sense to spend more to induce an estatutario to leave than a CLTista with similar attributes (salary, age, sex, education, tenure, etc.),.

36 PUBLIC EMPLOYMENT REFORM: WHAT IS NEEDED? 81. In this chapter, using the framework set out in Chapter 1, and the findings discussed chapters 2, 3 and 4, we determine the steps that are needed to address the problem of high and rising personnel costs, both those that do not require constitutional amendments, and those that require the passage of reform bills by Congress. The critical components of these reform bills (e.g., tenure revocation clauses in the Administrative Reform Bill) are identified. Among the measures that need not await the passage of these bills, we suggest the course of action that is most appropriate given the current state of private and public employment and pay, and previous experience in tackling public employment reform. Relevant international experience in the areas for which reform is recommended is provided in Chapter A SUMMARY STATEMENT OF THE PROBLEM 5.1.a Wltile Brazil does not have unusually high public employment levels, the public wage and pension bill has risen to unsustainable levels 82. At the national level, employment in federal, state, and municipal governments was about 9% of total employment in September 1995, approximately the level predicted by a study of 90 countries. Including employment in public enterprises, this ratio increases to about 12%, close to the average level for countries similar to Brazil. Thus, though public sector employment is out of line in some states and municipalities, and some branches of the federal government have had high employment growth in recent years, Brazil's problem is not one of general overemployment in the public sector. Despite the fact that Brazil's public sector does not suffer from general over-employment, public payroll-related expenses are about 12% of GDP in Brazil, which is considerably above the international average for comparable countries. These findings point to misalignment of compensation levels relative to those in the private sector. In fact, we find that salary and pension levels in the federal and state governments and enterprises are high relative to those in the private sector. Reducing wages and pension benefits of public employees has proved to be politically difficult since the Real Plan, so reduction of public employment levels is viewed as the primary potential recourse for a sizable fiscal adjustment. S.1.b In the civil service, the fiscal adjustment efforts have been modest 83. Wage growth for federal government officials appears to have slowed down since 1992, and some irregularities in pension payments are being eliminated at all levels of government (e.g., the 20% increment in salary given to imminent retirees from government). But real salary levels continued to grow between 1992 and 1995 by more than 10% annually for judicial and legislative workers and for state administration employees. Salary levels for public teachers and health workers and the military grew by about 5% annually. Even at the municipal level, where considerable overemployment exists, average real earnings grew by 3%. Some states and the federal government have attempted to reduce employment through voluntary separation schemes, but the take-up has been low. Participants in these programs are generally no more than 5% of baseline employment. In fact, employment increased between 1992 and 1995 in all branches of government except for state administrations, where it fell by a small amount.

37 c In public enterprises, the efforts have been somewhat more successful 84. For public enterprises, there are more options. Privatization of public enterprises will both reduce the wage bill of government and yield revenues. Also, voluntary separation schemes in enterprises that are not being privatized appear to have had more success than public administration PDVs, in terms of inducing higher take-up rates that have been as high as 33% of baseline employment. The main reasons for this are that these enterprises, especially public sector banks, have been able to offer more generous severance benefits, because their workers belong to the FGTS scheme and usually obtain 140% of their accumulated severance fund account balance in addition to the special package, and because workers in public enterprises often have better job prospects than their civil service counterparts. 5.1.d But these efforts have to be deepened now 85. Past efforts have not yielded enough saving of public expenditures: the share of payroll in government expenditures has remained high, and the share of inativos in public employment keeps growing. The growing ratio of pensioners to public employees means that with each passing year, public employment reform will become less and less meaningful as a way to reduce govemment spending. Also, it means that reforms requiring public employees to contribute towards their pensions will be less and less effective in balancing the public pension system the longer one waits. 5.2 PRINCIPLES FOR RENEWED EFFORTS 86. The principle underlying an effective approach to a sustainable reduction of the fiscal burden of public employment is one of reduction of differences in public and private earnings, pensions, and job stability. Since 1994, these have diverged instead of converging, as would have happened if labor market efficiency were growing. The aim of administrative, social security and labor market reforms should be to reduce differences in compensation between the public and private sectors, and not simply a reduction of public employment. But in some parts of government, especially where salary levels of workers are roughly the same as equally qualified workers in the private sector (e.g., in municipal administration and enterprises, for education and health workers in many states) substantial reduction of employment will be necessary in order to control growth of payroll-related spending even if salary and pension adjustments are feasible. 5.2.a Since 1994, public employment and pay has been unresponsive to economic changes 87. Public employment grew between 1992 and 1995 at all levels except state administrations, and real earnings continue to grow at all levels, and pension payments - that are linked to these earnings also have grown. Public employees have retained rights to lifelong employment. Thus while the macroeconomic setting has changed considerably, public employment levels and salaries have not shown any systematic response to these changes. 5.2.b In contrast, Brazil's private sector has displayed considerableflexibility 88. In sharp contrast, adjustment to a more stable and open economy in the private sector has been shared by wages and employment. The main mechanism that has facilitated this is the degree of informality of employment, which has allowed considerable changes in payroll

38 expenses of firms without a sharp rise in general unemployment levels. Many private formal sector firms have downsized successfully, but unemployment rates have only climbed slowly to about 6%. However, underemployment - i.e., people working fewer hours than they would like to at current wage levels, a phenomenon often associated with work in the informal sector - appears to have risen in recent years to about 15%. The strongest correlate of informality in Brazil appears to be non-payment of taxes to the government and not disregard for laws governing wages and employment conditions. Payroll taxes that fund the pension, severance, disability and other systems in Brazil are the highest in Latin America. The costs of shouldering this burden are lower earnings, pensions, and job stability in the private sector than would result if these taxes on employment are lower. 5.2.c Efforts shouldfocus on reducing public-private differences in earnings, pensions, and job stability 89. The result of these patterns in the public and private sectors is a growing gap between employment conditions and pay in the public and private sectors. While there is "too much" stability in the public sector, job security in the private sector is "artificially" low due to the nature of labor legislation. Artificially inflated informal employment is more than 50% of urban private employment in some states. While earnings of public sector employees are generally higher than those of similarly qualified private workers - estimates of the public sector premium for federal government workers are almost 30%, while for judicial and legislative workers this is higher than 50% - but the earnings gap between public and private formal sector is smaller than for total private employment. Since the aim of policy reforms is to improve the functioning of the Brazilian labor market while reducing the fiscal burden on the government, the focus of administrative reform efforts should be as much on the public-private earnings and job stability gaps as on earnings and tenure levels of public employees. Similarly, the efforts should aim to reduce both the public-private pension gap and the level of public pensions, which are excessively generous both relative to private employment in Brazil and by international standards. 5.3 REDUCING PUBLIC EMPLOYMENT UNDER CURRENT CONDITIONS 90. With constitutional and other restrictions on firing public employees or reducing their salaries or pensions, the use of voluntary separation programs initiated by state administrations and enterprises has been the major - perhaps overemphasized - avenue for reducing the fiscal burden of public employment. The analysis of ten voluntary separation programs suggests the following: 5.3.a Separation programsfor public administration should be redesigned and expanded 91. Based on comparisons of the costs and benefits of voluntary separation programs in three states and the federal government, downsizing appears to be a very profitable strategy for government. Our calculations suggest the benefits (in terms of reduced salary and pension bill) can be about seven to ten times the full financial costs of severance packages. Many state governments have resorted to relatively expensive early retirement schemes (that allow a spreading out of costs over time) because voluntary separation schemes involve large one-time expenses. This provides a rationale for expanding federal support for state downsizing reforms, in return for guarantees from states that they incorporate the lessons learned from past experience in future schemes. Our report suggests that, first, separation programs can be made cheaper and

39 more effective by introducing an element of involuntariness; while it is unconstitutional to fire public employees except for disciplinary reasons, the experience in Rio Grande do Sul shows that some pressure can be legally brought to bear on negligent or redundant employees to quit; and second, another design change that can result in lower costs is to make packages for nontenured government employees smaller than for those with tenure; currently estatutarios get lower severance benefits than CLTistas who - in addition to the special indemnity payments - can also withdraw their FGTS account balances. 5.3.b States should bear the full costs of state voluntary separation programs 92. One of the main findings is that these state downsizing efforts for estatutarios contain substantial hidden expenses for the federal government because of an implicit transfer of pension obligations from state treasuries to the INSS time of service program. Under current conditions, this leads to wedge between the amounts that any one state would be willing to pay to remove an employee from their payroll, and the amount that an agency representing the consolidated government would pay. Mechanisms to make these transfers (e.g., recognition bonds) explicit should be put in place and these transfers should be counted as state debt. Table 5.1: Fiscal Payoff to Public Employment Reform, in billion Reals Eliminating Public Sector Earnings Premium and Reducing Public Employment Baseline (1995 Levels) Wage Bill Saving Employ Earnings Wage Remove Reduce Both ment premium Premium Emplt. * Direct Administration Federal State Municipal Education & Health Judiciary & Legislative Military * Enterprisemployment includes employment in public enterprises. Employment is reduced by 10%; ** Total employment figures before 1991 are not accurately estimated. Source: Pesquisa Nacional por Amostra de Domiclios (selected years). 5.3.c Separation programsfor public enterprises can be made lessgenerous 93. Downsizing efforts in public enterprises - despite generally paying more than what is legally required to redundant or negligent employees - are a profitable strategy. The benefit-tocost ratios can be as high as seven in many cases. Since these employees - who fall under the CLT - can technically be dismissed, special indemnity payments to induce workers to voluntarily quit can be viewed as unnecessary. If special indemnity payments are paid, then these workers should not have access to their FGTS balances. In fact, not only are workers are allowed to withdraw these funds (i.e., the separation is treated as involuntary), state enterprises also pay them an additional 40% of the accumulated balance as a (i.e., the separation is treated as an "unjustified dismissal"). Cash-strapped enterprises have also relied on relatively expensive early retirement schemes in order to spread out the costs of downsizing over time. Our report suggests three main lessons. First, the costs can be reduced by introducing an element of involuntariness -

40 the experience of Bandepe, the state-owned bank in Pernambuco, suggests that this can be done with little or no labor unrest and litigation. Second, given the growing actuarial deficits in closed company funds, public enterprises should be encouraged to rely more on outright severance than early retirement, e.g., through financial support for lump-sum payments in return for adherence to general rules in designing these programs. Third, if special indemnity payments are made, then departures should be treated as voluntary quits, and workers should not be given access to their FGTS accounts; in the presence of these payments, it is difficult to justify treating these separations as unjustified dismissals. 5.3.d Keeping salary increases small will reduce the incentive to stay in government 94. Efforts to reduce employment under the current legal and political constraints should be assisted by keeping public sector salary increases to a minimum. Declining real earnings of public employees would encourage their voluntary departures as private sector employment and earnings increase, keep payroll expenses for current employees from rising, and - since pension payments to inativos are linked to salary levels of ativos - keep pension-related expenses from rising. 5.4 SELECTED PENSION REFORMS CAN ELIMINATE PUBLIC SECTOR PENSION PREMIUM 95. While these measures can provide some short-term relief, longer-term sustainability requires constitutional reform in the areas of social security, private sector labor regulations, and administration. Of these, perhaps the most important are measures to reduce the high and growing bill for both public and private sector retirees. Immediate measures to reduce the inequality between public and private pensions are most beneficial from fiscal, efficiency, and equity viewpoints. 96. Reform of private pensions will have a fiscal impact, in that contributions will rise and eliminate the growing INSS deficit. But these reforms are likely to be effective only if the system is individualized and capitalized. Reform ofpublic pensions is perhaps the most critical measure for immediate and long-term fiscal reasons. Pensions of public servants are between 33% and 100% higher than pensions of similarly qualified private sector employees. The generosity of public pensions relative to those in the INSS time of service program has three harmful effects: (a) it prevents successful retrenchment of all types of public employees at all levels of government, but especially those with salaries that will exceed 10 minimum wages during their lifetime; (b) it adds to the growing fiscal deficit (c) it reduces the fiscal benefits of employment reduction in the future, even if the tenure of civil servants is revoked. 5.4.a Lower replacement ratios to internationalevels 97. The most effective way to prevent the pension burden from growing is to lower the replacement ratio from 100% to more modest levels. For equity reasons, replacement ratios can be higher for low wage workers, e.g., about 50% for workers earning up to 5 minimum monthly salaries, 45% for workers earning between 5 and 10 minimum salaries, and 40% for those earning more.

41 b Public pension levels should be no more generous than private pensions 98. We find that salaries of public employees in state and municipal governments are approximately the same as private sector workers with similar attributes (education, age, gender, race, tenure and location), and the former enjoy considerably greater job stability. Thus, there is no justification for public servants not having the same basic pension benefits as private employees. Thus, the ceiling for benefits (eight times the monthly minimum salary) that applies for INSS pensioners should apply for public employees as well, and same rate of replacement - preferably below the current 100% rate - should apply. Equalizing pension benefits will also reduce some labor market distortions, e.g., the practice of switching to public service late in the career to avail of higher pension benefits. 5.4.c Public employees should contribute at the same rate as private employees 99. An effective way to begin equalizing public-private pay and benefits is to require public sector employees to make pension contributions at the same rates as private sector employees, viz., between 8% and 10% depending on their salary levels. Given small public-private differences in salary levels, there is little justification for not requiring public employees to contribute at the same fraction of their salaries as private workers. Requiring estatutarios to contribute at the same rate as CLTistas will reduce the fiscal burden for government, make public servants view pensions as deferred compensation and not an entitlement, reduce the attractiveness of public employment and aid downsizing programs. Equalizing the contribution rates would also mitigate the problem of transfer of pension obligations from state governments to the INSS system when workers leave public employment but are allowed to carry over their years of service into the private pension system. 5.4.d Make formula for calculating public pensions less end-loaded 100. There are both efficiency and equity reasons to make the formula for calculating pensions less end-loaded. The period used to calculate the salary base for pensions should be the same for public and private employees; the best solution would be to use entire salary histories to calculate pensions, rather than the last month's salary (as for civil servants) or the last 36 months' salary (as for INSS pensions). Even without a change in the replacement ratios, given the slope of experience-earnings profiles of public and private sector workers, this would reduce the pension bill in both the systems by more than 25% for new retirees. Since earnings profiles of less educated workers tend to be flatter, the reduction in pensions would be smaller for those with lower wages. 5.4.d Eligibility rules should be the same for private and public employees, and a minimum age at which retirement benefits commence should be instituted 101. The rules regarding eligibility for pensions should be made uniform for private and public employees. Raising retirement ages for arbitrarily selected occupations (such as teachers, legislators, and judicial workers) to the same levels as for other workers, basing eligibility in both the public and private systems on the years of contribution rather than years of service, and introducing a minimum age at which retirement benefits commence could result in large savings annually at current replacement rates and eligibility periods.

42 SELECTED LABOR REFORMS CAN REDUCE ARTIFICMILLY HIGH INFORMALITY AND TURNOVER OF PRIVATE EMPLOYMENT 102. Measures to increase private employment and earnings may be as important as monetary incentives to induce public sector employees to voluntarily leave government jobs. Realistically, most government employees are likely to view formal sector jobs (that offer pension benefits and more job stability than work in the informal sector) as the alternative to their current jobs. Labor market measures to improve the likelihood and attractiveness of privateformal employment are an important but underrated instrument for reducing the burden of public payroll expenses. 5.5.a Reduction of payroll taxes will reduce informality of employment 103. There is mounting evidence that the high level of payroll levies of about 45% of gross wages, and the structure of systems that these levies finance, increase informality of employment in Brazil to artificially high levels. Since more than half of these taxes fund the INSS pension system, reform that lead to the perception of these levies as contributions may lower labor costs in the private formal sector, and lead to employment growth. Reforms that encourage workers to view these taxes as contributing to their own pensions also increase the incentives to workers to formalize their labor contracts, leading to a growth in the share of the formal sector in private employment. Given that differentials in earnings, pensions, and job stability are smaller for public and formal private employmenthan those observed in comparisons of public and informal private employment, these reforms will narrow the public-private gap in compensation. 5.5.b Reform of FGTS scheme will reduce turnover in the formal sector 104. While there are distinct similarities between the US and Brazilian labor markets, one of the striking differences is that while the annual US turnover rate is about 25%, the average rate in Brazil is more than 33%. While no rigorous empirical evidence is available, the structure of severance scheme for private sector workers - a fund financed by employer contributions of 8% of payroll - is widely believed to increase turnover levels to artificially high levels in the formal sector, and even to increase informality of employment. Reforming the FGTS system (e.g., by reducing the penalty for unfair dismissal to the pre-1988 level of 10%) to reduce the perverse incentives to workers to "get fired" will reduce turnover rates and, by increasing investments in worker skills by firms, lower turnover rates to normal levels. By narrowing the gap in job stability between public and private sectors, these reforms will also help to lower the costs of public sector downsizing. 5.5.c Reform of selected labor laws will increase employment and earnings 105. Reform of relatively strict hiring and firing laws will lower the duration of unemployment in the formal sector, and thus reduce the payments required to induce public employees to voluntarily leave the public sector. De-linking minimum wages from pension benefits will reduce the sharp increases in fiscal burden that accompany even small increases in the minimum monthly salary. Additionally, since the ceiling on private sector pensions is fixed at eight times the minimum salary, it can be argued that relatively wealthy workers gain more from this parity relationship than poorer workers. A more efficient labor market will accelerate the growth of earnings of private workers, and narrow the public-private gap in earnings.

43 ADMINISTRATIVE REFORMS CAN REDUCE PUBLIC-PRIVATE JOB STABILITY DIFFERENTIALS 106. Constitutional changes proposed in the administrative reform bill being debated in Congress have received a lot of attention, and are being monitored as a signal of the government's resolve to accelerate the fiscal adjustment process initiated via the privatization program. The most contentious clause in this bill is the revocation of job stability for civil servants, which is regarded as necessary to facilitate cost-efficient downsizing a Immeliate fiscal benefits of tenure revocation are uncertain 107. A review of the experience with downsizing casts doubt on the likelihood of immediate fiscal benefits of this clause. Even employees who do not have tenure (CLTistas in public administrations and enterprises) are being "bought out" with indemnity payments, even though many of tlhem can legally be fired. The main constraint to successful and sizable downsizing of public employment appears to be political will, rather than constitutional. Thus the passage of this bill is unlikely to encourage state governments that lack the political will to adjust, though state governments thiat have displayed the political will to reform will obtain somewhat higher fiscal benefits because they will be able to reduce indemnity payments. For state governments that are reluctant to reform, passage of this bill removes an excuse to delay adjustment. But state governments facing high and rising payroll expenses may be best advised not to wait for the passage of these reforms b Though there may be some efficiency gains for determined reformers 108. Revocation of tenure for civil servants would allow governments to put pressure on negligenit workers, and hence result in greater efficiency in provision of government services. Here again, our review of state reform efforts indicates that governments with adequate political will have done so even under current laws, using measures such as time monitoring systems and performance evaluations to encourage grossly negligent workers to take relatively modest packages and quit. For such governments, the efficiency gains resulting from passage of the administrative reform will be modest. 5.6.c But cost of waiting for constitutional reform is high 109. While the fiscal and efficiency gains from the administrative reform bill are uncertain or modest - even if it is passed in its current form and made applicable for all workers immediately (i.e., not grandfathered) - the cost of waiting is certain and high. The two main factors are the high rate of interest paid on debt, and because of the rising share of pensions in payroll expenses, a later public sector downsizing/reform will be less fiscally rewarding than one today.

44 LESSONS OF INTERNATIONAL EXPERIENCE 110. This chapter summarizes the experience of other countries with social security reforms, severance schemes for public employees, and the effectiveness of active labor programs such as retraining programs, job search assistance, selfemployment startups, and temporary work creation. 6.1 SOCIAL SECURITY REFORMS 111. Countries all over the world are reforming their social security systems. The old systems were generally publicly managed, payroll tax financed, defined benefit, pay-as-you-go schemes. The main factors prompting these reforms are unsustainable fiscal imbalances, the need to reduce distortions and evasion, the desire to increase savings rates, and the desire to mitigate generational or socioeconomic conflicts. The common goals of social security reforms are: First, a closer link between benefits and contributions to minimize distortions; second, including a funded element which would generate more savings and help growth, and third, emphasizing long-term fiscal sustainability and minimizing intergenerational conflicts by eliminating scope for manipulation of the system for short-term redistributive gains. The overall objective of the reforms is to protect the old and unfortunate and promote growth Instead of relying solely on a intergenerational redistribution scheme for providing for old age, countries are diversifying their old age and disability pension provision by providing some public benefits (e.g., as a fixed universal benefit, or means-tested transfers to ensure a minimum pension level) but supplementing them with mandatory privately managed savings scheme ("sistema de capitalizacao"). Most countries already have voluntary pension plans managed either by companies or individuals themselves. In pension terminology, these different sources of pension finance and provision are termed "pillars of old age income security". A multi-pillar system has three pillars: first, a tax-financed mandatory publicly-managed pillar, second, a mandatory, fully-funded, privately-managed pillar, and third, a voluntary privatelymanaged pillar. 6.1.a The Models of Social Security Reform 113. From the experience of reformers, three models of social security reform can be identified: The Latin American Model 114. Countries such as Chile (1981), Peru (1993), Argentina (1994), and Uruguay (1996) have adopted multi-pillar systems, and several others such as Hungary are currently implementing such reforms. There are important differences in the nature of these reforms, though. In Chile, the old system is being phased out by making it compulsory for new entrants to the labor force to enter the new system, while older workers have a choice of system; in Argentina and Peru, the old system (consisting of the first pillar) is optional for all; in Uruguay the old system was "repartitioned", being mandatory for workers under 40 years, and optional for those above 40. In all cases, when the transition is complete, the new system will have three pillars: a defined

45 contribution, capitalized, privately managed second pillar, some form of minimum pension guarantee which can be thought of as the first pillar, and a voluntary third private savings pillar There can be considerable costs of transition in the shift from an unfunded pay-as-you-go system to a fully capitalized scheme. The problem arises because of the need to set aside enough funds to fully finance the pension obligations that have been already accumulated, i.e., for current retirees, who receive benefits but do not contribute to the system anymore, and current workers who have contributed and are still contributing. The cost of transition depends on the magnitude of these liabilities of the old system, which in turn depends on how generous the defined benefits were, and how old the population is. Chile, Argentina, and Peru had relatively young populations at the time of reforms, and not overly generous defined benefits, so the costs of transition were manageable. To further ease the transition, Chile and Peru issued "recognition bonds" to currently employed workers who switched from the old to the new system which formally acknowledged their previous contributions. In Peru's case, the total cost of recognition bonds was about 3% of GDP. Brazil's relatively young population would imply low costs of transition, but the generosity of defined benefits potentially raise these costs. The OECD Model 116. A few industrialized countries have already made major structural changes, and their experience may be considered the "OECD model"of social security reform. The reforming countries are Australia, Switzerland, Denmark and the UK. The main difference between the OECD and Latin American models of reform is that in case of the former - because it was small and affordable - the first pillar remains essentially the same, i.e., universal, pay-as-you-go and defined benefit. In Australia the first pillar is a means-tested pension paid from general revenues, in the UK it is a small flat benefit, and in Switzerland it is a compressed earnings-related benefit. The original contribution rate was also small, so a second pillar could be added on with additional contributions. For historical reasons, the investment decisions of funds of the second pillar are entrusted to employers and/or labor unions, instead of individual workers OECD countries such as France and Germany that already have large pay-as-you-go pension systems and relatively old populations, would face large costs of transition in adopting the Latin American (capitalized) model of reform. But these countries have also managed to avoid large scale evasion of payroll taxes that fund these systems. While real pension benefits in these countries have been eroded over time due to unanticipated increases in retirement duration, being wealthy has enabled these countries to prop up their social security deficits. For all these reasons - none of which resemble conditions in Brazil today - these countries have been exceptional in that they have not moved away from defined benefit system that characterized older social security schemes towards a system relying on defined contribution. While these countries have postponed reform, they retain the potential distortions of the old system. The Swedish "Notional Accounts " Model 118. The third approach seeks to combine the more attractive features (from a political economy standpoint) of the OECD model, viz., low transition costs, with those of the Latin American model, viz., low levels of labor and capital market distortions. This approach was pioneered by Sweden in the 1990s. Faced by a growing deficit and an increasingly distorted labor market, but unwilling to bear the high costs of transition of moving to a capitalized system,

46 Sweden devised a system of "notional accounting". Countries such as Italy, Poland, and Latvia are considering such a system as well In this system, contributors have notional or virtual accounts that earn a notional interest rate, which is based on economic and demographic projections. But it is still a pay-as-you-go system: while contributors are aware of their account balances, their contributions are in fact used to pay current retirees: no pension fund exists except to bridge the gap between current contributions and benefits. Thus the government is not required to, as in the Latin American model, contribute large sums of money to start a fund with which accrued pension rights are met. The notional link between contributions and benefits eliminates some disincentive effects associated with traditional pay-as-you-go systems: the somewhat tighter link between contributions and benefits may reduce evasion and annuity calculations can be used to help workers understand why retiring early will result in low replacement rates. But the system remains pay-as-you-go, so it does not offer the benefits of the Latin American model: there is no increase in saving and financial market development and contribution rates will have to increase as the population ages so intergenerational transfers stay large The important points that emerge from this experience are: e Long-term thinking is required: The very nature of the problem of social security implies that long-term thinking, which goes beyond the concerns of one or even two generations, is required for sustainable solutions. * Reforms will be difficult to implement: The reasons are, first, politicians are shy to propose deep reforms because costs have to be borne now while benefits come later and second, the potential losers from reform (e.g., old workers and retirees) are often well-organized and vocal, while the gainers (young workers) are unaware of the benefits or disinterested because the gains are so far in the future. * But reforms are being implemented successfully: Nevertheless, many countries, even those that were welfare states (e.g., Sweden) or socialist economies (e.g., Poland and Hungary) are implementing reforms. * Solutions are unique, but problems are the same: The solutions are often unique, but the problems that prompt these reforms are the same:financial unsustainability, inefficiency, and inequality. These are the same problems that Brazil is confronted with. 6.1.b LessonsfromArgentina, Hungary, and the USA 121. Argentina's reforms. In 1994, Argentina undertook comprehensive reforms to move towards a multi-pillar system. This involved setting up a capitalized second pillar; workers had the option of a publicly-managed system and a privately-managed regulated system. For Brazil, the experience of Argentina has the following lessons: First, because of political reasons, Argentina's reformers could not make it mandatory for anyone, including new entrants to the labor force, to join the private capitalized system. But almost all new entrants are going into the private part of the second pillar. Because they cannot switch back to the public part, this implies that in the long term everyone will have both a private account and the basic minimum from the government. So while this transition will take longer than in Chile, it has the benefit of being completely voluntary. Second, Argentina has maintained relative high payroll tax rates (of about 40-45%), and because only about one-fourth of these are linked directly to future benefits, evasion rates may not come down much. In Brazil, where high evasion rates are a problem, bolder measures may be required.

47 Hungary's planned reforms. Hungary's reforms may provide some guidance in reducing evasion of payroll taxes. Hungary has a high system dependency ratio of about two pensioners for every three contributors. Payroll taxes to finance the pension and health system are therefore very high, totaling almost 55 percent. These high rates have led to widespread and increasing evasion, and in 1996, social security took up more than one-third of central government expenditures, and contributed to a deficit of about 3.3% of GDP. The important point is that Hungary's problems with social security stem not from an aging population but from labor market distortions. The reforms aim to modernize the first pillar, introduce a second pillar, give a boost to the third pillar, and tighten labor and disability regulations. While the reforms of the first and second pillar are only now being implemented, these reforms and other changes have already encouraged the growth of the third voluntary pillar. In designing these reforms, Hungary has sought to learn from the experiences of countries all over the world. Hungary's attempt to return to an official economy from a shadow economy by lowering the payroll tax rate and establishing a tight link between today's contributions and future benefits, while retaining a redistributive component, should be studied by Brazil which has similar problems and aspirations USA's social security debate. The design of the US system, and the discussion surrounding it, is notable for long-term thinking: First, legislators have the mandate to keep the system in good health for the next 75 years. In contrast, Brazil's constitution requires lawmakers to ensure that benefits are paid, but not that the system is financially viable. Second, the social security debate in the USA has heated up because the system is believed to have registered an actuarial deficit (when accumulation of assets falls short of accumulation of liabilities; the US system will not go into a current deficit (i.e., when social security payments exceed contributions) for approximately another 25 years, but the situation is being treated as a crisis. In contrast, Brazil's INSS system registered a deficit last year. Finally, social security benefit levels - for both private and government workers - are calculated using individual earnings for the last 35 years of work, in essence the entire working life of contributors. In contrast, INSS pensions use only the last 36 months' earnings, and public pensions use only the last monthly salary. 6.1.c Summary of experience and its relevance for Brazil 124. In light of this experience, it is clear that Brazil's problems are neither unique nor beyond repair. As compared to countries such as the US, Sweden, and Uruguay, Brazil has a young population. This means better prospects for moving to a multi-pillar system, since it implies lower transition costs to a capitalized system. But the window of opportunity is small. The population, while young, is aging quickly; the rate of increase of pension debt is high because of high benefit levels; the absolute number of retirees, and the ratio of inativos to ativos is rising rapidly; and Brazil is facing strong pressures to increase labor market competitiveness The only countries not reforming are some OECD countries, because they have avoided large scale evasion of taxes that fund their systems, because the age distribution of their populations would imply very high costs of transition, and because they are wealthy enough to fund deficits of the system from general revenues. These circumstances clearly do notfit those of Brazil, where there are serious concerns about evasion of payroll taxes and high levels of informality of employment, the population is relatively young, and because propping up the existing system would both drain resources from investments in education, health and infrastructure and make it impossible to raise saving rates. Even the experience of richer countries shows an increasing emphasis on reducing distortions that raise the cost of labor and discourage saving. Economic concerns about social security are no longerpolitically incorrect.

48 PUBLIC SECTOR RETRENCHMENT SCHEMES 126. The relevant lessons from a World Bank study (Rama, 1997) on about 40 episodes of public sector retrenchment are: 6.2.a. Appropriate targeting must precede the offering of severance pay The most important message from the study is that appropriate targeting mustprecede the offering of any severance packages. This appears to be confirmed by the relative success of Bandepe's and Rio Grande do Sul's downsizing programs. But too often it is the other way round: the structure of public sector employment is the implicit outcome of down-sizing. The sequence should be the opposite: first identify the services one wants to leave to the private sector (e.g., printshops) as well as the occupational categories which are in excess (e.g., drivers or janitors); in the meantime, implement mechanisms to chase ghost workers at low cost (e.g., enforce working hours and other rules). 6.2.b Severance pay helps to overcome legal or political obstacles, but does not qualify on poverty reduction or equity criteria Severance pay may be needed because of legal or political constraints. In some countries public sector workers cannot legally fired, so that one has to "buy" their departure. This is far from trivial in the Brazilian case, where a court ruling once forced the government to re-hire 100,000 dismissed civil servants. Similarly, politically powerful labor unions may successfully oppose any restructuring if their members do not get appropriate compensation. But apart from these two considerations, the rationale to pay severance benefits may be weak. Public sector workers are usually not among the most poor. 6.2.c. Tailoring severance pay to characteristics of workers may substantially reduce the costs of downsizing Using the observable characteristics of public employees, such as sex, age, education, and comparing their salaries to private sector workers with similar attributes help in substantially reducing the cost of downsizing. Such comparisons can help in deciding both the absolute minimum amounts that people should be offered (to leave voluntarily, or to ensure adequate support until displaced workers find private sector employment). Usually, public sector agencies and firms are bound by precedence or legislation to make the package a multiple of the current salary, with the multiple depending on seniority. Such packages can overcompensate the people who accept them: estimates of over-compensation range from 15 to 40 percent. Household surveys, if not available, can be implemented with much less than these amounts. In the case of Brazil, as this report illustrates, PNAD surveys offer a rich source of information for the larger states and cities. Smaller states will require special surveys in determining the level and structure of severance benefits. Such surveys can be administered and analyzed during the targeting process, while implementing interim measures such as removal of ghost workers. 6.2.d Cost of severance schemes should be borne by restructuring agency When severance pay is offered, its cost should be borne by the agency and not by other levels of government, e.g., the federal government in case of state administration downsizing, or the state treasury in the case of downsizing of public enterprises. If the restructuring agency does

49 -45 - not into account the costs being bome by other levels of government, it may retrench more workers than is socially optimal. In fact, the correct calculations are quite complicated than suggested here. Salaries, pensions, and other benefits in the private sector differ from those in the public sector. The study recommends the use of a shadow pricing approach to make decisions on retrenchment. The shadow price of labor for each group of workers has to take into account, among other things, the cost of transition to a new job, the earnings gap between public and private sector jobs, the surplus or deficit of the restructuring agency, the marginal tax burden, and any externalities that may arise from restructuring (e.g., the impact of downsizing on the social security system). 6.2.e. The restructuring of public sectorpay shouldgo hand-in-hand with the downsizing Decompressing the wage structure may be a way to retain the most motivated and qualified civil servants. But this does not work if the restructuring of public sector pay takes place after the downsizing. The decision to stay or to leave when offered a severance pay package depends on the package itself, but also on salaries in government. The research actually points to taking advantage of the restructuring of public sector pay to offer something more elaborate than a standard severance pay package. For example, civil servants could be given the following choice: (a) stay where and as you are, (b) leave and get severance pay, or (c) stay in the public sector, with a higher salary but without tenure, e.g., a fixed-term contract. This menu should help retain hard-working civil servants, who may not be afraid of losing tenure. 6.2.f. "Adverse selection" is a serious but commonproblenm 132. Severance pay packages tend to be flawed with this kind of adverse selection, whereby the most productive people leave. The total output of the private sector may not change much because of their departure, but there may be a collapse of public sector output. Quite often, the best civil servants have to be re-hired (this is called the "revolving door" syndrome), which represents an expensive way of getting back to square one. This problem was found iii about a quarter of the cases studied by the World Bank study. Preventing individuals who accepted severance pay packages to work for the public sector in subsequent years is a poor solution to the problem. It may reduce the risk of people abusing the system, by getting first a golden handshake and then returning to their jobs, but does nothing to restore the productivity of the public sector once the "wrong" people have left. 6.2.g. The timing of downsizing in public enterprises depends on the state of labor legislation and the strength of labor unions A somewhat resolved issue is when downsizing should take place in public firms that are being privatized. The new (private) managers will generally have different employment targets than the bureaucrats in charge of the downsizing, so downsizing may be better left to the new owners. Also, one of the main reasons to privatize is the belief that private management is generally more efficient than public management; if this is true in general, it should be true for the downsizing process as well. The only rationale to downsize prior to privatization is to overcome the resistance of labor unions and make the firm more attractive to potential buyers: the rule should be that the full costs of severance packages are not greater than the expected increase in privatization proceeds.

50 h. Active labor market programs such as retraining are common but seldom effective Often, considerable sums of money are spent in active labor market programs, such as retraining, job search assistance, and help with selfemployment and small business startups. These programs are seldom effective (see next section for details). Their promoters are often public sector providers of training, information, or microcredit, who are given a second life by such retrenchment operations. It is appropriate to inform the displaced or potentially displaced workers about problems they may face, and about assistance that is available. But these workers should be allowed to get either such program benefits or their cash equivalent. Moreover, for those who prefer a program, displaced workers should be allowed to pick a service provider of their choice, among those who are accredited for these tasks (e.g., through a voucher scheme). 6.3 ACTIVE LABOR PROGRAMS 6.3.a. OECD experience with public retraining programs and job search assistance 135. In OECD countries, public retraining schemes and job search assistance are usually an important component of the total package of assistance for those laid-off due to plant closure or restructuring, as well as programs to assist the long-term unemployed. However, these schemes are seldom evaluated rigorously. Where they have been evaluated, studies vary in the sophistication of their treatment of impact and costs. These evaluations fall into three categories: experimental evaluations that OE(J: P C 1 compare the effects Diplaced Lng-term of programs for Wkers 0tUnemployed participants and aprvoretr e,poen MnfatrgHtrgeou $141TRAIN SCI XHMSf control group chosen prior to the Lormrecoitiosd Generally Generally scheme's initiation; deterioratjng improving quasi-experimental Training venue Mainly Casroom evaluations that cro and on-hejob choose the control Ot provided Job search JSA group after the assistance program is completed; and non- Noi; ofevatuatfin s scientific evaluations Qai eno 5 4 that use no control p3 (all US) group and are at best = R dtp unreliable. Sadly, ehius P-Nncetfc 1O'Sit Generally positive out of over twenty -- techiques Negative GScintific Generally negative evaluations -- Relative to S No moe e more efctive reviewed by Dar and Gill (1997), half -- Tratment Scaty Scanty were non-scientific - ic Alaticeas and none contained ; ascostlyas costasjsa a rigorous analysis Sorc Dar : and Gl ( 997).

51 of costs Distinguishing displaced workers from the long-term unemployed Retraining schemes for workers who have been displaced due to plant closures or restructuring should be distinguished from those for the long-term unemployed, because these groups differ in important aspects. The long-term unemployed are a relatively heterogenous group of individuals, are generally more dispersed geographically, and - almost by definition - have been unemployed longer than those displaced en masse Retraining for displaced workers. For those displaced en masse, retraining programs address layoffs and plant closures in the manufacturing sector, generally during deteriorating conditions in industry or aggregate employment. Second, the programs were mostly classroombased, and accompanied by job search assistance (JSA). Third, scientific evaluations show poor results for retraining programs: some programs resulted in modest gains in re-employment probabilities, but wage changes are generally insignificant. Fourth, scientific evaluations indicate that retraining programs are generally no more effective than JSA in increasing either reemployment probabilities or earnings. Finally, retraining programs for displaced workers appear to be two to four times more expensive than JSA. Combined with the previous finding, this implies that JSA may be more cost-effective than retraining in assisting displaced workers get jobs Implications for Brazil. The paucity of rigorous evidence on the costs and effectiveness of retraining programs do not allow us to conclude definitively on the economic justification of these interventions. The evidence that we do have, however, does not appear to justify indiscriminate expansion of retraining programs to cover more of the unemployed. The main lessons that other countries can draw from OECD experience are: = First, it is necessary to evaluate retraining (and other) public interventions using sound techniques. It is instructive that while non-scientific evaluations of retraining programs present a rosy picture, scientific evaluations are quite discouraging. - Second, rigorous evaluations, while not necessarily allowing a complete social cost-benefit analysis, can be useful for policymakers in allocating public expenditure. Reviews of evaluations find, for example, that job search assistance measures - which cost less than retraining but appear equally effective - may be a more cost-effective device in assisting displaced workers. * Third, OECD experience of retraining programs for workers displaced en masse, e.g., from public enterprises, may be useful in designing assistance programs in countries that expect labor-shedding in the manufacturing and mining sectors. Principally, this involves recognizing that retraining should not be the main form of assistance. * Fourth, public retraining programs are more beneficial for some groups than for others. However, it is difficult to predict a priori who will benefit most from retraining: principally, these results call for using modest pilot programs, evaluating them rigorously, and then tightly targeting retraining to those for whom it is found most cost-effective.

52 b. Self-employment programs 139. In recent years, self-employment has been proposed as a way to move some of those displaced from the public sector into gainful private employment. In these schemes, retrenched workers are encouraged to formulate a business idea for which they receive financial assistance and counselling. Lessons for the design and financing of self-employment programs have been distilled from the study of self-employment in nine OECD and two transition economies, Hungary and Poland. In the case of the OECD countries, 140. Who participates? Support for self-employment should not be seen as a panacea for unemployment. Experience shows that well-designed self-employment programs will attract at most about five percent of the unemployed. Those who participate are likely to be primarily male, better educated, and in their 30's. The programs tend not to attract women. Different strategies are needed to introduce women to self-employment. Self-employment does not provide a ready solution for unemployed or displaced youth. Not only are the young less inclined to participate in these programs when self-screening is used, the success and survival of enterprises appears to be greater for those 35 years of age and older What services are provided? In OECD countries, once admitted to a self-employment program, participants are usually paid a periodic benefit that is equivalent to a regular unemployment benefit. This avoids distorting the unemployed person's choice of employment assistance. Training and business advisory services are frequently offered, but the evidence of success of these services is scanty. Access to credit is generally not featured in countries with efficient banking sectors. Lump-sum payments, however, may be offered to provide an immediate source of capital. Programs offering lump-sum payments feature new businesses with larger amounts of capital per worker. In Poland, for example, a loan was available five times the size of unemployment benefits, and half the loan was forgiven if payments were made on time during the first two years How are costs controlled? Countries control the cost of self-employment assistance by introducing screening instruments that lead potential participants to weigh the private costs and benefits of the assistance. These instruments include: information sessions, applications, interviews, pre-entry business advisory services and training, and development of business plans. Instruments such as information sessions, formal applications, and interviews are effective selfscreening instruments that minimize the need for third party screening and "picking winners." 143. What are program outcomes and benefits? In the programs studied, roughly one of every four small businesses would have started without program assistance: this is the "deadweight effect" of these programs. An estimated one of every two businesses started fail during the first year. The businesses started, on average, create 1.5 jobs each. There is a tradeoff between controlling costs more through self-screening, and increasing deadweight costs. Even when a business fails, however, the experience may be beneficial in moving the owner back into wage employment: participants of self-employment programs have a greater chance of being employed later than those who remain in regular unemployment benefit programs. A rigorous randomized experiments design evaluating self-employment programs in the US indicates that such assistance can be cost beneficial when unemployment benefits have to be paid. Program costs, excluding administration, are equivalent in most cases to paying unemployment benefits. This conclusion may not hold in countries such as Brazil where unemployment benefits are relatively low.

53 c. Direct job creation or public works programs 144. Public work programs are an important part of labor market policy in many industrialized countries and transition economies. They involve employment in a government agency, with some programs providing government services and others aiming to develop physical infrastructure. Public work programs are of considerable interest to countries where workers are increasingly being displaced as a consequence of the necessary restructuring and privatization of the state-owned enterprises. It is argued that an advantage of these programs is that because government agencies control both the selection of workers and the type of job, they can minimize "displacement effects", which occur when a government program-funded employment replaces public or private sector employment. It is also argued that the infrastructure created by public works leads to productivity gains in the economy as a whole. There is a risk, however, of overestimating the job-creating and productivity-enhancing potential of such initiatives, and of underestimating their costs. The following are the main results of a joint OECD-World Bank study on the subject, which examined direct job creation programs in five OECD and two formerly communist countries (Poland and Hungary) Who participates, and who delivers? Direct job creation programs are primarily for the registered unemployed. If the program is countercyclical, the participants tend to be short-term unemployed. If the program is structural, the participants are long-term unemployed. A review of participants' education shows that they are among the lesser skilled. Most programs have a high majority of men, in part due to the design of the projects and the work undertaken. However, if the program is targeting economic hardship cases, the share of women tends to rise. In fact, in some of the countries studied (e.g., Germany, Poland), a significant number of participants are clients of the social welfare office. Direct job creation programs are generally under the responsibility of Ministries of Labor and are delivered, at the local level, by the local labor office. Often, these offices contract with the local authorities to hire the participants to provide government services. There is evidence that this leads to considerable displacement of government workers. In a number of countries, the Ministry allows the program to be delivered by non-profit associations: this can lower displacement rates since these organizations tend to provide services not otherwise available What benefits do participants receive? In OECD countries, participants generally receive the minimum wage or the wage rate stipulated by unions. This contributes to high program costs because wages tend to be much higher than regular unemployment benefits. In addition, the social security contribution or unemployment contribution is covered by the program. This allows the participant to requalify for unemployment insurance once the program has ended. Evidence in some countries (e.g., Germany and Poland) suggests that these programs are being used to "requalify" those who have exhausted their unemployment insurance benefits. That is, the participants enter the program so that they may be eligible for unemployment insurance benefits once they exit the program. There is little evidence that the programs give training to the participants or improve their skills, even though this is often a stated program objective What are program outcomes? Experience in OECD countries shows that public work programs are the least effective in moving the unemployed back into employment. Although there may be an increase in the flow out of short-term unemployment, there is little impact on