مزایای سن بالا و تصمیم گیری های دوران بازنشستگی سالمندان روستایی در برزیل

I estimate the impact of social security benefits on retirement decisions of rural workers by studying changes in the rules governing old-age benefits for rural workers in Brazil. I focus on a reform implemented in 1991, which reduced the minimum eligibility age, increased benefits, and extended the program to non-heads of households. Because those benefits come with no strings attached — they are not means or retirement tested — any behavioral response is a pure income effect. The main finding of the paper is that access to old-age benefits is a strong determinant of retirement of rural workers in Brazil: receiving old-age benefits increases the probability of not working by about thirty-eight percentage points and reduces total hours per week by 22½ h.

The Brazilian Constitution of 1988 established the guidelines for a reform of the Social Security system, to be regulated by ordinary law. Those guidelines require, among other things, that rural workers' old-age benefits be extended to women who were not household heads; that length-of-service eligibility be extended to rural workers; that no benefit be smaller than one minimum wage; and that the minimum eligibility age for old-age benefits for rural workers be reduced. Male rural workers were affected by: (1) a reduction in the minimum eligibility age from 65 to 60; (2) an increase in benefits from 50% to 100% of a minimum wage; and, (3) extension of access to length-of-service benefits (see Table 1). Female rural workers were affected by: (1) a reduction in the minimum eligibility age from 65 to 55; and, (2) the end of a restriction of one pensioner per household, which generally allowed married females to take-up benefits too.The timing of the events can be summarized as follows: In 1988, the Constitutional change was passed and more informed workers became aware of the future expansion of their entitlements. Immediately after July 1991, when the ordinary law (Lei #8212/8213) was passed, benefit payments to rural beneficiaries of old-age pensions increased automatically in general from 50 to 100% of the minimum wage, and newly eligible rural workers (e.g. 60 to 64 year old males) began to apply for benefits. In September 1992, the month of reference of the 1992 household survey that I will use in estimation, take-up of new benefits was still incomplete, either for bureaucratic reasons or because of delays in the spread of information. Finally, by September 1993, the month of reference of the 1993 survey, almost all of the take-up process was completed and newly eligible workers were already receiving their benefits. Administrative data from Ministério da Previdência Social (1998) confirms this pattern of sluggish take-up: by the end of 1992, 129,953 newly eligible males aged 60–64 were receiving rural old-age benefits; by the end of 1993, 326,158 were; by the end of 1994 that figure was 358,761. Fig. 1 shows the flow of new granted old-age benefits overtime and Appendix Table 1 shows the evolution of the stock of rural benefits outstanding, by type (old-age or length-of-service), gender and age of the receiver. Although the change in the law happened in July 1991, there is no apparent increase in the yearly flow of new granted benefits before 1992, until a spike is apparent in 1993 and 1994 (Ministério da Previdência Social, 1998). The figure also shows that the yearly flow of new disability benefits shrinks with the extension of old-age benefits, suggesting that disability and old-age benefits are substitutes for the age group affected by the reform. The number of granted rural length-of-service benefits is only about ½% of all rural benefits and is hardly visible in the picture. Therefore, the impact of the extension of access to length-of-service pensions is likely to be small or negligible.To a first approximation, social security for rural workers has a flat benefit schedule. Before the reform, rural old-age benefits were flat and equal to 50% of the minimum wage, with no incentive for rural workers to postpone their application to benefits. With the reform, benefits are either a function of documented past earnings, or the minimum wage, whichever the greater. Because the vast majority of rural workers do not have a long history of documented earnings, the near totality of rural old-age benefits is equal to the minimum wage (according to Ministério da Previdência Social, 1998, the average rural old-age benefit was R$121.37 in 1997, when the minimum wage was R$120). It is worthwhile noting that aposentadorias rurais (rural benefits) contain no incentives for total withdrawal from the labor force. There is no earnings test and, unlike their urban counterparts, rural workers do not have to quit their jobs to become eligible for benefits. 5 Therefore, examining only discrete measures of labor supply (as is common for studies of retirement in the United States) might miss the adjustment in the intensity margin, captured by total hours. I use the Pesquisa Nacional por Amostra de Domicílios (PNAD) to estimate the impact of the extension of social security benefits on labor supply. The PNAD is a yearly household survey, with sample size equal to 1/500 of the Brazilian population (about 100,000 households), designed to produce a picture of the living conditions and economic life of the Brazilian population, rural and urban. For every individual I observe characteristics such as age, race, education, school enrollment, income from different sources, housing and living arrangements, family structure, work, fertility, migration and other topics. There are several measures of labor participation, including hours of work, labor force non-participation and earnings. Unfortunately, the PNAD data do not allow me to determine the type of social security benefits paid to each beneficiary. The PNAD classifies social security benefits only into two broad categories: aposentadorias, which comprise disability, old-age and length-of-service benefits, and pensões, which comprise military and survivors' income maintenance benefits. In my baseline specification, I identified rural and urban workers based on their occupations. I observe every worker's current occupation and I observe past occupation up to a four years recall.6 Individuals who do not report their occupation or who have not worked in the last 4 years are classified as having “undefined” occupation. My results are robust to the use of location of residence as a proxy for occupation for those workers whose occupation is undefined, or to assigning urban occupation to all workers with undefined occupation.7Table 2 reports the observation counts by year, age groups 55–59, 60–64 and 65–69 and rural, urban and undefined occupation. The sample is composed of males 55–69 years old; with less than 12 years of schooling; single or whose wife is younger than 50 years old, and comprise 15,573 observations, of which 3,180 reported not to have worked in the last four years and to have an undefined occupation.The labor supply measures analyzed in this paper describe both the market and non-market dimensions of work. The basic binary indicator of retirement decision is the variable “did not work in the week of reference”. This variable does not consider agriculture work for own consumption or own construction. In order to explore the intensity of work, I use total hours per week, which includes own consumption and own construction activities.8 This paper will study male workers age 55–69, because compulsory retirement age is 70, and workers younger than 55 might have important unobservable differences from workers receiving old-age benefits. Because there were simultaneous changes in eligibility and benefits for females, the sample is restricted to male workers unmarried or whose wives are not older than 50.9 Hence analysis of the joint retirement decision of married couples is postponed to subsequent work. Because less than ½% of elderly rural workers had 12 years of schooling or more (equivalent to high school completion or more) whereas 9% of elderly urban workers achieved that schooling level, I restrict both the rural and urban samples to workers who had less than 12 years of schooling. Since there are no data for 1991 and 1994, in the econometric analysis, 1989 and 1990 are the pre-reform years — those are the latest years before the reform implementation for which data is available, — and 1992 and 1993 are the post-reform years — those are the earliest years after the reform implementation started.

The reform in the Brazilian Social Security System in 1991 provides a unique opportunity to learn about labor supply responses to reductions in the minimum eligibility age for old-age benefits, increases in benefit levels, and about how social programs reach groups with different education levels in developing countries. This paper estimates a pure income effect — the vast majority of the existing literature has estimated total effects that are the sum of an income and a substitution effect, the latter due to means, income, earnings or retirement tests. In the standard case of leisure as a normal good, the income and substitution effects of means tested unearned income point to the same direction. All else being equal, one should expect smaller total effects in the program analyzed in this paper than if means testing were in place. However, this paper shows that a reduction in the minimum eligibility age for old-age benefits had remarkably strong consequences on the work behavior of those affected. Moreover, because my estimates are based on a comparison between the just-made-eligible by the reform (elderly age 60–64) with the soon-to-be-eligible (elderly age 55–59), they are likely an underestimate of the actual responses. The main result of this paper contradicts the hypothesis that working as long as one's body permits is a common strategy among rural elderly in Brazil — when given the opportunity of earning a benefit that guarantees their subsistence, withdrawal of the labor force is found to be the preferred decision for about 40% of benefit receivers. The finding of a sizeable pure income effect suggests that means testing is more desirable than otherwise, since one important cost of means testing is its negative labor supply consequences.12 These results indicate that estimates of the poverty reduction effect of old-age benefits in developing countries should take into account the negative labor supply impact of those benefits. For instance, a simple back of the envelope calculation based on the IV estimates can help us gauge the magnitude of the aggregate effect of reducing the old-age eligibility age for male rural workers from 65 to 60. Based on the PNAD of 1989, this group comprised 460,000 people or 0.42% of the population over 10 years old, supplied 0.8% of total hours in the economy, and 3.2% of total hours in the rural sector. With an estimated reduction in hours of 22.6 hours per week for this group, this would imply a drop in total hours in the rural sector of 1.6%. With the knowledge that even pure income transfers might generate sizeable labor supply responses among elderly rural workers in developing countries, it remains to be investigated whether implementation of universal old-age benefits might bring about substantial output losses. If labor markets are characterized by high unemployment or low participation rates of the youth, output losses due to retirement may be minimized if the non-employed youth replace the retired elderly.13 Finally, this paper finds inconclusive results about differences in behavior across education groups, perhaps because the standard errors of estimates are too large. However, the point estimates for the differences across education groups are large enough to suggest that more research on education differences might be worth pursuing.