WCs a2tt POLICY RESEARCH WORKING PAPER 2210 The Slippery Slope During the turbulent years 1976-96, aggregate data for Brazil appear to show only Explaining the Increase in Extreme small changes in mean Poverty in Urban Brazil, 1976-96 income, inequality, and incidence of poverty - Francisco H.G. Ferreira suggesting little change in the distribution of income. But a Ricardo Paes de Barros small group of urban households - excluded from formal labor markets and safety nets - was trapped in indigence. Based on welfare measured in terms of income alone, the poorest part of urban Brazil has experienced two lost decades. The World Bank Poverty Reduction and Economic Management Network Poverty Division H October 1999 POLicY RESEARCH WORKING PAPER 2210 Summary findings Despite tremendous macroeconomic instability in Brazil, negative "growth" effect, and unfortunate changes in the the country's distributions of urban income in 1976 and structure of occupations and participation in the labor 1996 appear, at first glance, deceptively similar. Mean force. household income per capita was stagnant, with rninute * Two factors that tended to reduce poverty - accumulated growth (4.3 percent) over the tvo decades. improved educational endowments across the board, and The Gini coefficient hovered just above 0.59 in both a progressive reduction in dependency ratios. years, and the incidence of poverty (relative to a poverty The net effect was small and negative for measured line of RS60 a month in 1996 prices) remained inequality overall, and negligible for the incidence of effectively unclhanged over the period, at 22 percent. poverty (relative to "high" poverty lines). Behind this apparent stability, however, a powerful But the net effect was to substantially increase extreme combination of labor market, demographic, and poverty - suggesting the creation of a group of urban edticational dynamics was at work, one effect of which households excluded from any labor market and trapped was to generate a substantial increase in extreme urban in indigence. poverty. Above the 15th percentile, urban Brazilians have Using a decomposition methodology based on "stayed put" only by climbing hard up a slippery slope. microsimulation, which endogenizes labor incomes, Counteracting falling returns in both self-employment individual occupational choices, and decisions about and the labor market required substantially reduced education, Ferreira and de Barros show that the fertility rates and an average of two extra years of distribution of income was being affected by: schooling (which still left them undereducated for that - Three factors that tended to increase poverty - a income level). decline in average returns to education and experience, a This paper - a product of the Poverty Division, Poverty Reduction and Economic Management Network -is part of a larger effort in the network to understand the determinants of urban poverty. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Grace Ilogon, room MC4-552, telephone 202-473-3732, fax 202-522-3283, Internet address gilogon@(worldbank.org. Policy Research Working Papers are also posted on the Web at http://wblnOO18.worldbank. org/research/workpapers.nsf/policyresearch ?openform. Francisco Ferreira may be contacted at fferreira Recon.puc-rio.br. Cctober 1999. (71 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the i countries they represent. Produced by the Policy Research Dissemination Center THE SLIPPERY SLOPE: Explaining the Increase in Extreme Poverty in Urban Brazil, 1976 - 1996 Francisco H.G. Ferreira and Ricardo Paes de Barros l Keywords: Brazil; Poverty; Inequality; Labor Markets; Education; Demography. JEL Classification: C15, D31, J22, 121, I32 ' Respectively PUC - Rio and IPEA. We are very grateful to Francois Bourguignon for his guidance and support; and to Nora Lustig, Naercio Menezes Filho and participants at the LACEA 1998 conference in Buenos Aires; the SBE 1998 conference in Vitoria, ES, and a seminar at Cornell University for helpful comments. We owe our largest debt of gratitude to Phillippe George Leite, Roberta Barreto, Carlos Henrique Corseuil, Sergio Firpo, Luis Eduardo Guedes, Cristiana Lopes, Vanessa Moreira, Daniele Reis e Alinne Veiga for superb research assistance and dedication beyond the call of duty. 2 1. Introduction. Both by the standards of its own previous growth record, during the 'Brazilian miracle' years of 1968-1973, and by those of other leading developing countries thereafter, notably in Asia, the two decades between 1974 and 1994 - between the first oil shock and the return of stability with the Real plan - were dismal for Brazil. First and foremost, they were characterized by persistent macroeconomic disequilibrium, the main symptoms of which were stubbornly high and accelerating inflation and a GDP time-series marked by unusual volatility and a very low positive trend. Figure 1 below plots annual inflation and GDP per capita growth rates for the 1976-1996 period. [See Figures la and lb in Appendix 4] The macroeconomic upheaval involved three price and wage freezes (parts of the Cruzado Plan of 1986, the Bresser Plan of 1987 and the Verao Plan of 1989) - all of which were followed by higher inflation rates; one temporary financial asset freeze (with the Collor Plan of 1990); and finally a successful currency reform followed by the adoption of a nominal anchor in 1994 (the Real Plan). The national currency changed name four times.2 Throughout the period, macroeconomic policy was almost without exception characterized by relative fiscal laxity and growing monetary stringency. In addition, substantial structural changes were taking place. Brazil's population grew by 46.6% between 1976 and 19963, and became more urban (the urbanization rate rose from 68% to 77%). Average education of the over-ten population rose from 3.2 to 5.3 effective years of schooling.4 Open unemployment grew steadily more prevalent. The sectoral composition of the labour force changed away from agriculture and manufacturing, and towards services. The degree of formalization of the labour force declined substantially: the proportion of formal workers (wage 2The changes were from Cruzeiro to Cruzado in 1986; from Cruzado to Novo Cruzado in 1989; from Novo Cruzado back to Cruzeiro in 1990, and from Cruzeiro to Real in 1994. 3 See Table Al, Appendix 1, for a complete population series. 4 'Effective' years of schooling are based on the last grade completed, and are thus net of repetition. 3 workers with formal documentation) was almost halved, from just under 60% to just over 30% of all workers. And not least, despite the morass, real GDP per capita and mean household per capita income (for the spatially undeflated national distribution) were both some 22.5% higher in 1996 than 1976.5 Importantly, however, this increase was closely associated with rural-urban migration; accumulated growth in mean urban household per capita incomes was a mere 4.3% (spatially deflated). See Table 1 below. And yet, despite the macroeconomic turmoil and continuing structural changes, however, a casual glance at the headline inequality indicators and poverty incidence measures reported at the bottom of Table 1 might suggest that little had changed in the Brazilian income distribution between 1976 and 1996. __________________________________ l1976 1981 1985 1996 GNP (in constant 1996 Reais - thousands)* 434,059 538,475 599,130 778,820 GNP per capita (in constant 1996 Reais)* 4,040 4,442 4,540 4,945 Annual Inflation Rate1 42% 84% 190% 9% Open Unemployment2 1.82% 4.26% 3.38% 6.95% Average Years of Schooling 3, 4 3.23 4.01 4.36 5.32 Rate of Urbanization 4 67.8% 77.3% 77.3% 77.0% Self-employed as share of Labor Force4 27.03% 26.20% 26.19% 27.21% Share of Formal Employment4* 6 57.76% 37.97% 36.41% 31.51% Mean Household Per Capita Income 4, 5 265.10 239.08 243.15 276.46 Inequality (Gini) 4 0.595 0.561 0.576 0.591 Inequality (Theil - T) 4 0.760 0.610 0.657 0.694 Poverty Incidence (R$ 30/month) 4 0.0681 0.0727 0.0758 0.0922 Poverty Incidence (R$ 60/month) 4 0.2209 0.2149 0.2274 0.2176 Notes: *) Annual figures. 1) Percent, from January to December. Based on the IGP-DI for 1976, and on the INPC-R for all other years. 2) Based on the IBGE Metropolitan Unemployment Index. 3) For all individuals 10 years of age or older, in urban areas. 4) Calculated from the PNAD samples by the authors. See Appendix 1. 5) Urban only, monthly and spatially deflated. Expressed in constant 1996 Reais. 6) Defined as the number of employees 'com carteira' as a fraction of the sum of all wage employees and self- employment workers. 5 See Table 1, and Appendix I for details. 4 But, as is often the case, casual glances may turn out to be misleading. This apparent distributional stability belies a number of powerful, and often countervailing, changes in four realms: the returns to education in the labour markets; the distribution of educational endowments over the population; the pattern of occupational choices; and the demographic structure resulting from household fertility choices. In this paper, we note two puzzles about the evolution of Brazil's urban income distribution in the 1976-1996 period, and suggest explanations for them. The first puzzle is posed by the combination of growth in mean incomes and stable or slightly declining inequality, on the one hand, and rising extreme poverty on the other. We argue that this can only be explained by the growth in the size of a group of very poor households, who appear to be effectively excluded both from the labour markets and from the system of formal safety nets. This group is trapped in indigence at the very bottom of the urban Brazilian income distribution, and contributes to rises in poverty measures particularly sensitive to the depth (FGT(l)) and severity (FGT(2)) of poverty, particularly when poverty is defined with respect to a low poverty line. E(O) fails to respond to this group because of a rise in the share of families reporting (valid) zero incomes.6 Other inequality measures, which also fall slightly between 1976 and 1996, compensate for these increases in poverty by declining dispersion further up along the distribution. But the reality of the loss in income to the poorest group of urban households is starkly captured by Figure 3 (in Section 2), which plots the observed (truncated) Pen parades for the four years being studied. The main endogenous channel through which the marginalization of this group is captured in our model is a shift in their occupational 'decisions' away from either wage or self-employment, towards unemployment or out of the labour force.7 As Table 1 indicates, there was certainly a decline in fornal employment as a share of the labour force. Second, the evidence which we will examine in Section 3 below reveals downward shifts in the earnings-education profile, controlling for age and gender, in both the wage and self-employment 6 See Appendix 1, Table A2. 7 The use of terns such as 'occupational choice' or 'decision' should not be taken to imply an allocation of responsibility. It will become clear, when the model is presented that, as usual, these are choices under constraints. 5 sectors.8 Despite a slight convexification of the profile, the magnitude of the shift implies a decline in the (average) rate of return to education for all relevant education levels. Similarly, average returns to experience also fell unambiguously for 0-50 years of experience (see figure 5). The combined effect of changes in these returns - the 'price effects' - was an increase in simulated poverty, for all measures, and for both lines. Simulated inequality also rose, albeit much more mildly. Both effects were exacerbated by simulating the changes (to 1996) of the determinants of labor force participation decisions were also taken into account. The second puzzle, then, is what forces counterbalance these price and occupational choice effects, so as to explain the observed stability in inequality and 'headline' poverty.9 We find that they were fundamentally the combination of increased education endowments, moving workers up along the flattening earnings-education slopes, with an increase in the correlation between family income and family size, caused by a more than proportional reduction in dependency ratios and family sizes for the poor. This demographic factor had direct effects on per capita income - through a reduction in the denominator - and indirect effects - through participation decisions and higher incomes. Naturally, the co-existence of these two phenomena - or 'puzzles' - implies that this last demographic effect did not extend to all of Brazil's poor. At the very bottom, some are being cut off from the benefits of greater education and economic growth (such as these were), and remain trapped in the sinking valley. We address these issues by means of a micro-simulation-based decomposition of distributional changes, developed by Bourguignon et. al. (1998), which itself builds upon the work of Almeida dos Reis and Paes de Barros (1991), and Juhn, Murphy and Pierce (1993). The approach is described in Section 3 below, and has two distinguishing features. First, unlike other dynamic inequality decompositions such as that proposed by Mookherjee and Shorrocks (1982), it decomposes the effects of changes on an entire distribution, rather than on a scalar sumnmary statistic (such as the mean log deviation). This allows for much greater versatility: within the B This shift is from 1976 to 1996, and takes place after upward shifts in the 1980s. See Figure 4. 9 By 'headline' poverty, we mean poverty incidence computed with respect to the R$60/month poverty line. 6 same framework, a wide range of simulations can be performed to investigate the effects of changes in specific parameters on any number of inequality or poverty measures (and then for any number of poverty lines or assumptions about equivalence scales). More fundamentally, it allows us to observe the effects of particular simulations on the entire distribution, as we do in Figures 6 - 15, rather than merely on a couple of scalar indices. Second, the evolving distribution which it decomposes is a distribution of household incomes per capita (with the recipient unit generally being the individual). Therefore, moving beyond pure labor market studies, the effect of household composition on living standards and participation decisions is explicitly taken into account. As it turns out, these are of great importance for a fuller understanding of the dynamics at hand. The remainder of the paper is organized as follows. Section 2 briefly reviews the main findings of the literature on income distribution in Brazil over the period of study, and presents summary statistics and dominance comparisons for the four observed distributions we analyze: 1976, 1981, 1985 and 1996. Section 3 outlines the basic model and describes the empirical methodology. Section 4 presents the results of the estimation stage and discusses some of its implications. Section 5 presents the main results of the simulation stage, and decomposes the observed changes in poverty and inequality. Section 6 concludes and draws some policy implications. 2. Income Distribution in Brazil from 1976 to 1996: a brief review of the literature and of our data set. There is little disagreement in the existing literature about the broad trends in Brazilian inequality since reasonable data first became available in the 1960s. The Gini coefficient rose substantially during the 1960s, from around 0.500 in 1960 to 0.565 in 1970 (see Bonelli and Sedlacek, 1989). There was a debate over the causes of this increase, spearheaded by Albert Fishlow on the one hand, and Carlos Langoni on the other, but there was general agreement that the sixties had seen substantially increased dispersion in the Brazilian income distribution. 7 The 1970s displayed a more complex evolution. Income inequality rose between 1970 ad 1976, reached a peak on that year, and then fell - both for the distribution of total individual incomes in the economically active population (PEA) and for the complete distribution of household per capita incomes - from 1977 to 1981. This decline was almost monotonic, except for an upward blip in 1980. See Bonelli and Sedlacek (1989), Hoffrnan (1989) and Ramos (1993). The recession year of 1981 was a local minimum in the inequality series, whether measured by the Gini or the Theil-T index. From 1981, it rose during the recession years of 1982 and 1983. Some authors report small declines in some indices in 1984, but the increase was resumed in 1985. By that year, the Gini coefficient for the distribution of household incomes per capita had risen from 0.574 in 1981 to 0.589 (see Ferreira and Litchfield, 1996). Hoffman (1989) and Bonelli and Sedlacek (1989) report similar increases. 1986, the year of the Cruzado Plan, saw a break in the series, caused both by a sudden (if short- lived) decline in inflation, and a large increase in reported household incomes. Stability and economic growth led to a decline in measured inequality, according to all authors. Thereafter, with the failure of the Cruzado stabilization attempt and the return to stagflation, inequality resumed its upward trend, with the Gini finishing the decade at 0.606. Table A2.1 in Appendix 2 summarizes the findings of this literature, both for per capita household incomes and for the distribution of total individual incomes in the economically active population. The general trends identified in the existing literature are mirrored in the statistics for the years with which we concern ourselves in this paper, namely 1976, 1981, 1985 and 1996. The distributions for each of these years come from the Pesquisa Nacional por Amostra de Domicilios (PNAD), run by the Brazilian Geographical and Statistical Institute (IBGE). Except where otherwise explicitly specified, we deal with distributions where the welfare concept is total household income per capita (in constant 1996 Reais, spatially deflated to adjust for regional differences in average cost-of-living), and the unit of analysis is the individual. Details of the PNAD sampling coverage and methodology, sample sizes, the definition of key income variables, spatial and temporal deflation issues, and adjustments with respect to the National Accounts baseline are discussed in Appendix 1. 8 Table 2 below presents a number of summary statistics for these distributions - in addition to the mean, which was provided in Table 1 above. The four inequality indices, which will be used throughout this paper, are the Gini coefficient and three members of the Generalized Entropy Class of inequality indices, E(e). This class of measures satisfies a number of desirable properties, such as the strong Pigou-Dalton transfer principle, scale invariance, population replication invariance and decomposability. See Cowell (1995) for a discussion. Specifically, we have chosen E(O), also known as the mean log deviation or the Theil - L index; E(l), more famously known as the Theil - T index, and E(2), which is half of the square of the coefficient of variation. These provide a useful range of sensitivities to different parts of the distribution. E(O) is more sensitive to the bottom of the distribution, while E(2) is more sensitive to higher incomes. E(1) is roughly neutral, whereas the Gini places greater weight around the mean. We also present three poverty indices from the Foster-Greer-Thorbecke (FGT) additively decomposable class P(a). P(O), also known as the headcount index, measures poverty incidence. P(l) is the normalized poverty deficit; and P(2) is an average of squared normalized deficits, thus placing greater weight on incomes furthest from the poverty line. We calculate each of these with respect to two poverty lines, representing R$1 and R$2 per day, at 1996 prices. Each of these poverty and inequality indices is presented both for the (individual) distribution of total household incomes per capita, and for an equivalized distribution using the Buhmann et. al. (1988) parametric class of equivalence scales, (with 0 = 0.5). This provides a rough test that the trends we describe are robust to different assumptions about the degree of economies of scale in consumption within households. Whereas a per capita distribution does not allow for any such economies of scale, taking the square root of family size allows for them to a rather generous degree. As usual, per capita incomes generate an upper bound for inequality measures, whereas allowing for some extent of local public goods within households raises the income of (predominantly poor) very large households, and lowers inequality. In the case of the poverty measures, in order to concentrate on the household re-ranking effect, and to abstract from the 9 pure mean scaling effect, the poverty lines were adjusted as follows: z* = z[,u(n)]3-, where ,u(n) is the mean household size in the distribution. 1976 1981 1985 1996 Median (1996 R$)* 127.98 124.04 120.83 132.94 Inequality Gini - 0 = 1.0 0.595 0.561 0.576 0.591 Gini - 0 = 0.5 0.566 0.529 0.548 0.567 E(0) - 0 = 1.0 0.648 0.542 0.588 0.586 E(0) - 0 =0.5 0.569 0.472 0.524 0.534 E(1) - 0 = 1.0 0.760 0.610 0.657 0.694 E(1) - 0 =0.5 0.687 0.527 0.580 0.622 E(2)- 0 1.0 2.657 1.191 1.435 1.523 E(2)- 0=0.5 2.254 0.918 1.134 1.242 Poverty - R$30/ month ___ P(0) - 0 1.0 0.0681 0.0727 0.0758 0.0922 P(0) - 0 =0.5 0.0713 0.0707 0.0721 0.0847 P(1) - 0 = 1.0 0.0211 0.0337 0.0326 0.0520 P(1) - 0 0.5 0.0235 0.0315 0.0303 0.0442 P(2) - 0 =1.0 0.0105 0.0246 0.0224 0.0434 P(2)- 0=0.5 0.0132 0.0226 0.0204 0.0357 Poverty - R$601 month P(0) - 0 = 1.0 0.2209 0.2149 0.2274 0.2176 P()_- fl=0.5 0.2407 0.2229 0.2382 0.2179 P(1) - 0 =_1.0 0.0830 0.0879 0.0920 0.1029 P(l) - 0 0.5 0.0901 0.0875 0.0927 0.0960 P(2)_ - 0 1.0 0.0428 0.0525 0.0534 0.0703 P(2)_ - 0_ 0.5 0.0471 0.0508 0.0521 0.0625 Note: *: For urban areas only, and spatially deflated. See Appendix 1. The median incomes in Table 2 behave roughly in tandem with the means reported in Table 1, and in accordance with the macroeconomic cycle: 1981 was a recession year, followed by stagnation in 1982 and a severe recession in 1983, from which the median - unlike the mean - had not yet recovered by 1985. Both subsequently rose to 1996. The table also confirms that the evolution of inequality over the period is marked by a decline from 1976 to 1981, and a subsequent deterioration over the remaining two sub-periods. Furthermore, this trend is robust to 10 the choice of equivalence scale, proxied here by two different values for 0, although the inequality levels are always lower when we allow for economies of scale within households. It is also robust to the choice of inequality measure, at least as regards the inequality increases from 1981 to 1996 and from 1985 to 1996, as the Lorenz dominance results identified in Table 3 indicate. Figure 2a plots the four Lorenz curves. Their proximity suggests, as we have stated, that even where Lorenz dominance is detected, the changes in inequality over this period are not quantitatively very large. Figure 2b truncates the Lorenz curves for the first 40% of the distribution, so as to show the separation between the curves more clearly. The dominance of 1981 and 1985 over 1996 shows clearly. 1976 also lies everywhere above 1996 for this range, but the lines cross at a higher percentile. Be that as it may, the cumulative income share for the poorest four deciles was certainly lower in 1996 than in 1976. [See Figures 2a and 2b in Appendix 41 The results for poverty are more ambiguous. With respect to the higher poverty line, incidence is effectively unchanged throughout the period (and even displays a slight decline for the equivalized distribution). FGT(1 and 2), however, show increases over the period, and these become both more pronounced and more robust with respect to 0, as the concavity of the poverty measure increases. This suggests that depth and severity of poverty, affected mostly by falling incomes at the very bottom of the distribution, were on the rise. This is confirmed by the trend of all three P(a) indicators, with respect to the indigence line. Once again, the trend is more pronounced the higher cc. For P(1) and P(2), the monotonicity of the increase is independent of 0. As a result there is only one welfare dominance result among the years studied. These results are reflected in Table 3 below, where a letter L (F) in cell (i, j) indicates that the distribution for year i Lorenz dominates (first order stochastically dominates) that for year j. 1981 and 1985 both display Lorenz dominance over 1996, as suggested above. There is only one case of first-order welfare dominance throughout the period and, symptomatically, it is not of a 11 later year over an earlier one. Instead, money-metric social welfare was unambiguously higher in 1976 than in 1985, as indicated above. Indeed, all poverty measures reported for both of our lines (and for 0 = 1.0) are higher in 1985 than in 1976.10 This is conspicuously not the case for a comparison between 1976 and 1996. Whereas poverty measures very sensitive to the poorest are higher for 1996, poverty incidence for 'higher' lines fall from 1976 to 1996, suggesting a crossing of the distribution functions. Figure 3 shows this crossing, by plotting the Pen parades (F-'(y)) - truncated at the 60' percentile - for all four years analyzed. Note that whereas 1976 lies everywhere above 1985, all other pairs cross. In particular, 1976 and 1996 cross somewhere near the 17t percentile. *1976 1981 1985 1996 1976 F 1981 L 1985 L 1996 [See Figure 3 in Appendix 4] Before we turn to the model used to decompose changes in the distribution of household incomes, which will shed some light on all of these changes, it will prove helpful to gather some evidence of the evolution of educational attainment, as measured by average effective years of schooling, and for labour force participation, for different groups in the Brazilian population, partitioned by gender and ethnicity. Table 4 presents these statistics. As can be seen, there was some progress in average educational attainment in urban Brazil over this period. Average effective years of schooling, as reported in Table 1, rose from 4.2 to 5.3. In fact, this piece of good news will prove of vital importance in having prevented a more ' Note that this first-order welfare dominance is not robust to a change in 0 to 0.5. 12 pronounced increase in poverty. Table 4 now reveals that the male-female educational gap has been eliminated, with females older than ten being on average slightly more educated than males. Clearly, this must imply a large disparity in favour of girls in recent cohorts. While a cohort analysis of educational trends is beyond the scope of this paper", such a rapid reversal may in fact warrant a shift in public policy towards programmes aimed at keeping boys in school, without in any way discouraging the growth in female schooling. Finally, note the remarkable disparity in educational attainment across ethnic groups, with Asians substantially above average, and blacks and those of mixed race below it. Tabe 4 Fuc~tina1a~ LaorFore artciatin1tat#i,b gedr and race~ 1976 1981 1985 1996 Average Years of Schooling (Males) 4.04 4.36 5.20 Average Years of Schooling (Females) 3.14 3.99 4.37 5.43 Average Years of Schooling (Blacks and MR) - - - 4.20 Average Years of Schooling (Whites) 6.16 Average Years of Schooling (Asians) - - 8.13 Labor Force Participation (Males) 73.36% 74.63% 76.04% 71.31% Labor Force Participation (Females) 28.62% 32.87% 36.87% 42.00% Labor Force Participation (Blacks and MR) - - 55.92% Labor Force Participation (Whites) 56.41% Labor Force Participation (Asians) _ 54.88% Notes: Average 'effective' years of schooling for persons ten years or older, in urban areas. Labor Force Participation in urban areas only. As for labour force participation, the persistent and substantial increase in female participation from 29% to 42% over the two decades, was partly mitigated by a decline in male participation rates. Those trends notwithstanding, the male-female participation gap remains high, at around 30 percentage points. There is little evidence of differential labour force participation across ethnic groups. " See Duryea and Szekely (1998) for such an educational cohort analysis of Brazil and other Latin American 13 3. The Model and the Decomposition Methodology Let us now turn to the Brazilian version of the general semi-reduced-form model for household income and labor supply in Bourguignon et. al. (1998). It is used here to investigate the evolution of the distribution of household incomes per capita over the two decades from the mid- 1970s to the mid-1990s. Specifically, we analyze the distributions of 1976, 1981, 1985 and 1996, and simulate changes between them. The paper covers only Brazil's urban areas (which account for some three quarters of its population). The general model therefore collapses to two occupational sectors: wage earners and self-employed in urban areas."2 Total household income is given by: ( 1 ) n n (1) Y+>wL Vi,,Le +YOh Where w; are the total wage earnings of individual i, LW is a dummy variable that takes the value one if individual i is a wage earner (and zero otherwise); Ti is the self-employment profit of individual i: L" is a dummy that takes the value one if individual i is self-employed (and zero otherwise); and Y0 is income from any other sources, such as transfer or capital incomes. Equation I is not estimated econometrically. It aggregates infornation on right-hand-side term I (from equations 2 and 4), 2 (from equations 3 and 4) and 3 directly from the household data set. The wage-earnings equation is given by: (2) Logw, = Xi 7W± + eW countries. 12 We will eventually extend the model to cover rural areas too, by incorporating two additional sectors: wage earners and self-employed in the rural areas. In Brazil, wage earners include employees with or without formal documentation ('com ou sem carteira'). The self-employed are own-account workers ('conta propria'). 14 where Xi = (ed, ed2, exp, exp2 , Dg ). Ed denotes completed effective years of schooling. Experience (exp) is defined simply as: age - education - 6, since a more desirable definition would require the age when a person first entered employment, a variable which is not available for 1976.13 Dg is a gender dummy, which takes the value 1 for females (and zero for males). wi are the monthly earnings of individual i. This extremely simple specification was chosen so as to make the simulation stage of the decomposition feasible, as described below. It embodies the assumptions that the Brazilian labor market was not segmented by region, firm size, race, or any attribute other than gender. Analogously, the self-employed earnings equation is given by: (3) Logg7, = XPf8se + gse Equations 2 and 3 are estimated by simple OLS. Equation (2) is estimated for all employees, whether or not heads of household, and whether with or without formal sector documentation (com or sem carteira). Equation 3 is estimated for all self-employed individuals (whether or not heads of households). Because the errors s are unlikely to be independent from the exogenous variables, a sample selection bias correction procedure might be used. However, the standard Heckmnan procedure for sample selection bias correction requires equally strong assumptions about the orthogonality between the error terms E and , (from the occupational choice multinomial logit below). The assumptions required to validate OLS estimation of (2) and (3) are not more demanding than those required to validate the results of the Heckman procedure. We assume, therefore, that all errors are independently distributed, and do not correct for sample selection bias in the earnings regressions. We now turn to the labor force participation model. Because we have a two-sector labor market (segmented into the wage employment and self-employment sectors), labor force participation and the choice of sector (occupational choice), could be treated in two different ways. One could 13 Given the fact that education is given by the last grade completed, and is thus net of repetition, this definition will overestimate the experience of those who repeated grades at school, and hence bias the experience coefficient downwards. The numbers involved are not substantial to alter any conclusions on trends. 15 assume that the choices are sequential, with a participation decision independent from the occupational choice, and the latter conditional on the former. This approach, which would be compatible with a sequential probit estimation, was deemed less satisfactory than one in which individuals face a single three-way choice, between staying out of the labor force, working as employees, or in self-employment. Such a choice can be estimated by a multinomial logit model. According to that specification, the probability of being in state s ( 0, w, se) is given by: ezir, (4) Pj where s, j = (0, w, se) ezi,r +le Zr yj j*s where the explanatory variables differ for household heads and other household members, by assumption, as follows. For household heads: x I -;nO-13 nl4-65, n,65 E D14-65ed, 1 D14-65ed ]DI4-65age, h b14-65 -I n14-65 -I 14-65 -I G,_,4-65 age,-E D,4-65Gd, D n,4-65 -1 n,4-65 -1 For other members of the household: XjP;n n, n_6,>65, YD,4 6,ed, 'ED 46ed] D46ae Zh0 13fl14-65 nf>5 n, D14-65ed n D14-65agej zh= 1l4-65 -i nl4-65 -i -i4-65 -i Dl4-65ge: s D, 6Gd, D,s, Ll w, D 1l4-65 1in4-65 -i Where nk., is the number of persons in the households whose age falls between k and m; D,465 iS a dummy that takes the value one for individuals whose age is between 14 and 65; DS' is a dummy for a self-employed head, the penultimate term is the earnings of a wage-earning head; and D is a dummy variable that takes the value one if there are no individuals aged 14-65 in the household. The sums defined over {j } are sums over {Vi E h / j }. 16 The multinomial logit model in (4) corresponds to the following discrete choice process: (5) s=Argmax{Uj=Zhy+±,,j = (O,w,se)} where Z is given above, separately for household heads and other members; the tj are random variables with a double exponential density function and Uj may be interpreted as the utility of alternative j. Once the vector yj is estimated by (4), and a random term , is drawn, each individual chooses an occupation j so as to maximize the above utility function. A Decomposition of Changes in the Distribution of Household Incomel 4 Once equations 2, 3 and 4 have been estimated, we have two vectors of parameters for each of the four years in our sample (t E (1976, 1981, 1985, 1996}): Pt from the earnings equations for both wage earners and the self-employed (including constant terms cc), and y, from the participation equation. In addition, from equation 1, we have Yoht and Yht. Let Xht :={XP, ZIh | Vi e h} and Qb : {gwi, Se.i &i Ii E h }. We can then write the total income of household h at time t as follows: (6) Yht =H(Xht ~Yohti Qht; ft SYt) h=l, ..,m Based on this representation, the distribution of household incomes: (7) Dt = {Yi,, Y2t, ...--Ymt } can be rewritten as: (8) D, = D[{Xh,YOhtYQh, }3hA,Yr,] Where {.} refers to the joint distribution of the corresponding variables over the whole population. We are interested in understanding the evolution of Dt over time, or possibly that of a set of alternative summary poverty or inequality measures defined over it. 17 The decompositions proposed in this project consist of estimating the effects of changing one or more of the arguments of D[. on Dt. The simplest decomposition applies to those arguments which are exogenous to the household: that is, the fis, ~S, and the variance of the various residual terms. Changing the fis amounts to assuming a change in the rate of return on human capital variables in equation (2) and (3). We refer to this as a "price effect". In algebraic terms, it can be expressed as: (9) B= D[{Xh,, Yohm ' 0 ht } lA, ] D[{Xh,D Yoht,Qh, } rfit, ] This expression measures the contribution to the overall change in the distribution Dt' - Dt of a change in ,l between t and t', holding all else constant. Likewise, the "labor supply effect" may be defined by: (10) L,,. = D[{XhtYihD I Y ht },fi, 7rt ]- D[{Xh, X YohtQht }' fit r] The price effect Btt' is obtained by comparing the distribution at date t with the hypothetical distribution obtained by simulating on the population observed at date t the remuneration structure of period t'. A price effect can be computed individually - that is, for one element of the vector f3, or collectively - that is, for all elements of the vector P. Both types of simulations are reported below. Likewise, the labor supply effect, Ltt', is obtained by comparing the initial distribution with the hypothetical distribution obtained by simulating on the population observed at date t the occupational preferences observed at date t'. Again, a labor-supply effect can be computed individually - that is, for one element of the vector y, or collectively - that is, for all elements of the vector y. We only report collective labor-supply decompositions in this paper. '4 This section draws heavily on Bourguignon et. al. (1998), adapting it to our specifications. 18 Considering only the collective price and labor supply effects, one can then write the change in the distribution of household income as the sum of a price effect, a labor supply effect and a residual: (I11) Di - D,, = B,t. + L,t. +R,,, The residual Rt. measures the contribution to the change in the distribution of income of changes in the distributions of observable and unobservable characteristics, respectively all the Xhs and YOhs, and all the es and q7. (11) is a definitionally exact decomposition, but changes in the residual term R,, which encompass all changes in household physical and human capital endowments, changes in the receipts of non-labour incomes, such as capital income or transfers, and demographic changes, are likely to be important. In order to shed light on some of those effects, one must mind the fact that of the variables in {Xht, YOi Qht}, only the residual terms in Qht are (by assumption) orthogonal to all other variables. For any other variable, i.e. elements of the X and Y vectors, a change in distribution must be understood conditionally on all other observable characteristics. Specifically, if we are interested in the effect of a change in the distribution of a single specific variable Xk on the distribution of household incomes between times t and t', it is first necessary to identify the distribution of Xk conditional on other relevant characteristics X.k (and possibly other incomes Y.). This can be done by regressing Xk on X kat dates t and t', as follows: (12) Xki = Xk ,PI + Ukit where k is the variable, i is the individual, and t is the date. The vector of residuals ukit represents the effects of unobservable characteristics (assumed to be orthogonal to X , on Xk. The vector zt is a vector of coefficients capturing the dependency of Xk on the true exogenous variables Xk, at time t. For the sake of simplicity, let us assume that the error terms u are normally distributed with mean zero and a common standard deviation at. 19 The same equation can, of course, be estimated at date t', generating a corresponding vector of coefficients ptw, and a standard error of the residuals given by a,,. We are then ready to simulate the effect of a change in the conditional distribution of Xk from t to t', by replacing the observed values of Xki, in the sample observed at time t, with: (13) X *ki, = X_k,t ' + Ukil at The contribution of the change in the distribution of the variable Xk to the change in the distribution of incomes between t an t' may now be written as: (14) Rx* =D[{X,, , X Y ,YOh,1,ahA} rJ-DY ] D[{XkLk,X ki}iJkit.YQhA,5r1 In this paper, we perform four regression estimations such as (13), and hence four simulations such as (14). The four variables estimated are Xk = {n013, n1465, n>65, ed). In the case of the education regression, the vector of explanatory variables X-d, was (1, age, age2, Gd, regional dummies). In the case of the regressions with the numbers of household members in certain age intervals as dependent variables, the vector X-kit was (1, age, age2, ed, ed2, regional dummies), where age and education are those of the household head. The simulations permitted by these estimations allow us to investigate the effects of the evolution of the distribution of educational attainment and of demographic structure on the distribution of income. We now turn to the results of the estimation stage of the model. 4. Estimating the Model The results of the OLS estimation of equation (2) for wage earners (formal and informal) are shown in Table 5 below. The static results are not surprising. All variables are significant and have the expected signs. The coefficients on education and its square are positive and significant. The effect of experience (defined as [age - education - 6]), is positive but concave. The gender dummy (female =1) is negative, significant and large. 20 The dynamics are more interesting. Between 1976 and 1996, the earnings-education profile changed shape. After rising in the late 1970s, the linear component fell substantially from 1981 to 1996. Meanwhile, the coefficient of squared years of schooling fell to 1981 but then more than doubled to 1996, ending the period substantially above its initial 1976 value. Overall, the relationship became more convex, suggesting a steepening of marginal returns to education at high levels. However, plotting the parabola which models the partial earnings-education relationship from equation (2), the lowering of the linear term dominates. The profile shifts up from 1976 to 1981, and again to 1985, before falling precipitously (while convexifying) to 1996. See figure 4. The net effect across the entire period was a fall in the cumulative returns to education (from zero to t years) for the entire range. This co-existed with increasing marginal returns at high levels of education. The implications for poverty and inequality are clear, with the education price effect leading to an increase in the former and a decline in the latter. Year 1976 1981 1985 1996 Intercept 4.350 4.104 3.877 4.256 (0.0001) (0.0001) (0.0001) (0.0001) Education 0.123 0.136 0.129 0.080 (0.0001) (0.0001) (0.0001) (0.0001) Education2 0.225 0.181 0.283 0.438 (x 100) (0.0001) (0.0001) (0.0001) (0.0001) Experience 0.075 0.085 0.087 0.062 (0.0001) (0.0001) (0.0001) (0.0001) Experience2 -0.105 -0.119 -0.121 -0.080 (x 100) (0.0001) (0.0001) (0.0001) (0.0001) Gender -0.638 -0.590 -0.635 -0.493 (0.0001) (0.0001) (0.0001) (0.0001) _______________ 0.525 0.538 0.547 0.474 Source: Authors' calculations based on the "Pesquisa Nacional por Amostra de Domicilios" (PNAD). P-values in parentheses. [See Figure 4 in Appendix 4] 21 Returns to experience also increased from 1976 to 1981, and from 1981 to 1985, with a concave pattern and a maximum at around 35 years of experience. See Figure 5 below. But from 1985 to 1996, there was a substantial decline in cumulative returns to experience, even with respect to 1976, until 50 years of experience. The relationship became less concave, and the maximum returns moved up to around 40 years. Over the entire period, the experience price effect was mildly unequalizing, and seriously poverty increasing. [See Figure 5 in Appendix 41 The one piece of good news comes from a reduction in the male-female earnings disparity. While female earnings, controlling for both education and experience, remained substantially lower in all four years, suggesting that some labour market discrimination may be at work, there was nevertheless a decline in the this effect between 1976 and 1996. This effect, as we will see from the simulations reported in Section 5, was both mildly equalizing and poverty reducing. Let us now turn to Equation (3), which seeks to explain the earnings of the self-employed with the same set of independent variables as Equation (2). The results are reported in Table 6 below. This table reveals that education is also an important determinant of incomes in the self- employment sector. The coefficient on the linear term has a higher value in all years than for wage-earners, but the quadratic term is lower. This implies that, ceteris paribus, the return to low levels of education might be higher in self-employment than in wage work, but would eventually become lower as years of schooling increase. This will clearly have an impact on occupational choice, estimated through equation (4). Dynamically, the same trend was observed as for wage- earners: the coefficient on the linear term fell over time, but the relationship became more convex.'5 The coefficients on experience and experience squared follow a similar pattern to that observed for wage earners, as shown in Figure 5. Once again, the cumulative return to experience fell over the bulk of the range from 1976 to 1996, contributing to the observed increase in poverty. The effect of being female, ceteris paribus, is even more markedly negative in this sector than in the wage sector. It also fell from 1976 to 1996, despite a temporary increase in disparity in the 1980s. 15 In this case, it actually switched from concave to convex. 22 .... ,.. ......Table 6:Euain 3:-Total Earing Regeso for the self employed-- Year 1976 1981 1985 1996 Intercept 4.319 4.192 3.853 4.250 (0.0001) (0.0001) (0.0001) (0.0001) Education 0.196 0.148 0.165 0.114 ______________ (0.0001) (0.0001) (0.0001) (0.0001) Education2 -0.206 0.021 0.012 0.219 (x 100) (0.0001) (0.4892) (0.6545) (0.0001) Experience 0.074 0.079 0.084 0.063 (0.0001) (0.0001) (0.0001) (0.0001) Experience2 -0.101 -0.108 -0.111 -0.082 (x 100) (0.0001) (0.0001) (0.0001) (0.0001) Gender -1.092 -1.148 -1.131 -0.714 (0.0001) (0.0001) (0.0001) (0.0001) ______________ 0.431 0.434 0.438 0.336 Source: Authors' calculations based on the "Pesquisa Nacional por Amostra de Domicilios" (PNAD). P-values in parentheses. Let us now turn to the estimation of the multinomial logit in equation (4). This was estimated separately for household heads and for others, since the set of explanatory variables was slightly different in each case (see the description of vectors Z, and Z1 in Section 2 above). Table A3.1 in Appendix 3 presents the results for household heads, and Table A3.2 for other household members. In both tables, the results are presented as the effects of other choices versus that of remaining outside the labor force ('unoccupied'). For household heads, education was not significantly related to the likelihood of choosing to work in the wage sector vis-a-vis staying out of the labor force, at any time. In addition, the positive effect of education decreased from 1976 to 1996, to the point where it was no longer statistically significant. The dominant effect on the occupational choices of urban household heads over this period, however, was a substantial decline in the constant term affecting the probability of participating in either productive sector, as opposed to remaining outside the 23 labour force, or in unemployment. Since it is captured by the constant, this effect is not related to the educational or experience characteristics of the head, or to the endowments of his or her household. We interpret it, instead, as the effect of labour market demand side conditions, leading to reduced participation in paid work. This effect will be shown, in the occupational choice simulations reported in the next section, to be both unequalizing and immiserizing . For other members of the household, education did seem to raise the probability of choosing wage work vis-a-vis staying out of the labour force, with the relationship changing from concave to convex (and weak) over the period. It also enhanced the probability of being in self employment vis-A-vis outside the labor force in both periods, although this relationship remained concave. The number of children in the household significantly discouraged participation in both sectors, although more so in the wage-earning one. The change in the constant term was much smaller than for household heads, suggesting that negative labour market conditions hurt primary earners to a greater extent. Consequently, we will observe the effect of the occupational choices of other household members on poverty and inequality to be much milder than those of the heads. This is in contrast to other countries where similar methodologies have been applied, such as Taiwan, where changes in spouse (and particularly women) labour force participation rates had important consequences for the distribution of incomes (See Bourguignon et. al., 1997). Table A3.3 in Appendix 2 reports the results of the estimation of equation (12), with education of individuals ten years old or older as the dependent variable, regressed against the vector (1, age, age2, Gd, regional dummies). Over time, there is a considerable increase in the value of the intercept, which will yield higher predicted values for educational attainment, controlling for age, gender and regional location. Additionally, the gender dummy went from large and negative to positive and significant, suggesting that women have more than caught up with males in educational attainment in Brazil over the last twenty years. The effect of individual age is stable, and regional disparities, with the South and Southeast ahead of the three central and northern regions, persist. 24 Tables A3.4 - A3.6 report the results of regressing the number of household members in the age interval 0-13; 14-65; and above 65 (respectively), on the vector (1, ed, ed2, age, age2, regional dummies). The main finding here is that the schooling of the head has a large, negative and significant effect on the demand for children, so that as education levels rise, family sizes would tend to fall, ceteris paribus. Additionally, some degree of convergence across regions in family size can be inferred, with the positive 1976 regional dummy coefficients for all regions (with respect to the Southeast) declining over time, and more than halving in value to 1996. This picture suggests a possibly important transformation in Brazil's demographic structure, with potential implications for welfare. As we will see in the next section, the role of observed reductions of family size was indeed crucial. 5. The Simulation Results. Having estimated earnings equations for both sectors of the model (wage-earners (2) and the self- employed (3)); participation equations for both household heads and non-heads (4); and 'endowment' equations (13) for the exogenous determination of education and family composition we are now in a position to carry out the decompositions described in equations (9), (10) and (14). These simulations, as discussed above, are carried out for the entire distribution (as in equations 7 and 8). However, the results are summarized below in Table 7, 8 and 9, which report mean household per capita income pt(y), four inequality indices (the Gini coefficient, the Theil-L index [E(0)], the Theil-T index [E(l)], and E(2)), and the standard three members of the Foster-Greer-Thorbecke class of poverty measures, P(a), a = 0, 1, 2, computed with respect to two monthly poverty lines: an indigence line of R$30 and a poverty line of R$60 (both expressed in 1996 RM Sao Paulo prices). 25 Table 7: Simulated Poverty and Inequality for 1976, Using 1996 coefficients. Mean Inequality Poverty __ _____________________ p/c ________ Z = R$30 / month Z = R$ 60 / month Income Gini E(0) E(1) E(2) P(0) P(1) P(2) - P(0) P(1) P(2) 1976 observed 265.101 0.595 0.648 0.760 2.657 0.0681 0.0211 0.0105 0.2209 0.0830 0.0428 1996 observed 276.460 0.591 0.586 0.694 1.523 0.0922 0.0530 0.0434 0.2176 0.1029 0.0703 Price Effects a, 1 for wage earners 218.786 0.598 0.656 0.752 - 2.161 _ 0.0984 0.0304 00.0141 _ 0.2876 0.1129 0.0596 a, 1 for self-employed 250.446 0.597 0.658 0.770 2.787 0.0788 0.0250 0.0121 0.2399 0.0932 0.0490 =,a for both 204.071 0.598 -6 0.655 .i754 =2.90 0.1114 0.0357 0.0169 0.3084 0.1249 0.0673 a only, for both 233.837 0.601 0.664 0.774 2.691 0.0897 0.0275 0.0129 0.2688 0.1040 0.0545 All 1B (but no a) for both 216.876 0.593 0.644 0.736 2.055 0.0972 0.0303 0.0143T- 0.2837 0.1114 0.0590 Education ,B for both 232.830 0.593 0.639 0.759 2.691 0.0779 0.0234 0.0110 0.2531 0.0953 0.0488 Experience 1 for both 240.618 0.600 0.664 0.771 2.694 0.0851 0.0265 0.0125 _ 0.2592 0.1000 0.0525 Gender P for both 270.259 0.595 0.649 0.751 2.590 0.0650 0.0191 0.0090 _ 0.2160 0.0797 0.0404 Occupational Choice Effects y for both sectors (and both 260.323 0.609 0.650 0.788 2.633 _ 0.0944 0.0451 0.0331 _0.2471 0.1082 0.0671 heads + others) y for both sectors (only for 265.643 0.598 0.6 57 0.57 2.482 0.0721 0.0231 0.0119 -0 0.2274 0.0867 0.0454 other members) I _y, a, 1 for both sectors 202.325 0.610 0.649 0.788 2.401 0.1352 0.0597 0.0402 0.3248 0.1466 0.0902 Demographic Patterns _ ,ud only, for all 277.028 0.574 0.585 0.704 2.432 0.0365 0.0113 0.0063 - 0.1711 0.0554 0.0264 __,ud, y, ac, ,B, for all 210.995 0.587 0.577 0.727 2.177 0.0931 0.0433 0.0321 - 0.2724 0.1129 0.0677 Education Endowment Effects 393 54 0 _04 0 6 0 e only, for 339.753 0.594 0.650 0.740 2.485 0.0424 0.0136 0.0073 0.1593 0.0567- 0.0287 [ id, [Le for all 353.248 0.571 0.584 0.688 2.320 0.0225 0.0078 0.0049 0.1131 0.0359 0.0173 1.te, pd, y, oc, 3,B for all 263.676 0.594 0.600 0.727 1.896 0.0735 0.0374 0.0296 =0.2204 0.0913 0.0561 Source: Based on "Pesquisa Nacional por Amostra de Domicilios" (PNAD) of 1976 and 1996. .- 26 Table 8: Simulated Poverty and Inequality for 1981, Using 1996 coefficients. Mean Inequality Poverty P/c Z = R$30 / month Z=R$ 60/ month Income Gini E(0) E(1) E(2) P(0) P(1) P(2) P(0) P(1) P(2) _1981 observed 239.075 0.561 0.542 0.610 1.191 0.0710 0.0321 0.0230 _ 0.2133 0.0862 0.0509 1996 observed 276.460 0.591 0.586 0.694 1.523 0.0922 0.0530 0.0434 0.2176 0.1029 0.0703 Income Generation a, 1 for wage earners 203.978 0.563 0.546 0.624 1.288 0.0925 0.0383 0.0258 0.2648 0.1081 0.0628 _a, , for self-employed 236.511 0.564 0.554 0.618 1.216 0.0772 0.0342 0.0241 _ 0.2229 0.0915 0.0542 a, p for both 201.262 0.568 0.557 0.636 1.325 0.0987 0.0405 0.0269 _ 0.2750 0.1135 0.0662 a only, for both 226.751 0.560 0.541 0.608 1.203 _ 0.0774 0.0340 0.0240 _ 0.2300 0.0927 0.0545 _ All , (but no a) for both 184.150 0.574 0.571 0.656 1.411 _ 0.1179 0.0474 0.0302 _ 0.3126 0.1320 0.0772 Education P for both 206.439 0.554 0.523 0.603 1.232 - 0.0812 0.0351 0.0245 - 0.2463 0.0984 0.0571 Experience 1 for both 201.805 0.570 0.566 0.637 1.301 0.1029 0.0427 0.0282 _ 0.2784 0.1169 0.0687_ Gender P for both 244.918 0.558 0.538 0.602 1.168 0.0676 0.0310 0.0225 0.2052 0.0829 0.0490 Occupational Choice r for both sectors (and both 235.636 0.570 0.548 0.629 1.234 0.0907 0.0479 0.0374 -0.2344 0.1044 0.0675 heads + others) y for both sectors (only for 240.013 0.564 0.552 0.614 1.195 - 0.0756 0.0342 0.0244 _ 0.2207 0.0903 0.0537 other members) Y= a, (1 for both sectors 200.559 0.579 0.566 0.663 1.393 0.1172 0.0562 0.0412 _ 0.2925 0.1307 0.0823 Demographic Patterns Ld only, for all 247.443 0.544 0.496 0.573 1.093 _ 0.0529 0.0275 0.0219 0.1745 0.0688 0.0416 _p4,r, a, 13, for all 207.243 0.560 0.513 0.617 1.256 0.0874 0.0455 0.0359 0.2486 0.1056 0.0663 Education _____ p_ 0only, for all 298.677 0.582 0.592 0.663 1.325 _ 0.0610 0.0300 0.0231 0.1779 0.0735 0.0450 pd, p. for all 310.762 0.569 0.552 0.634 1.248 0.0448 0.0251 0.0208 0.1426 0.0574 0.0361 Jie, id , rY, a, 13, for all 255.032 0.586 0.572 0.681 1.390 _ 0.0775 0.043 1 0.0352 j 0.2155 0.0938 0.0607 Source: Based on "Pesquisa Nacional por Amostra de Domicilios" (PNAD) of 1981 and 1996. l l l l l l l 27 Table 9: Simulated Poverty and Inequality for 1985, Using 1996 coefficients. Mean Inequality Poverty p/c Z = R$30 / month _ Z = R$ 60 / month Income Gini E(O) E(l) E(2) P(0) P(1) P(2) P(0) P(l) P(2) 1985 observed 243.152 0.575 0.588 0.654 1.432 0.0738 0.0307 0.0205 0.2258 0.0901 0.0514 1996 observed 276.460 0.591 0.586 0.694 1.523 _ 0.0922 0.0530 0.0434 0.2176 0.1029 0.0703 Income Generation cX, , for wage earners 221.944 0.563 0.557 0.631 1.403 0.0758 0.0306 0.0203 0.2353 0.0929 0.0524 cc, f3 for self-employed 241.405 0.572 0.58 1 0.647 1.392 0.0725 0.0299 0.0200 0.2236 0.0887 0.0504 a, 1 for both 220.421 0.560 0.549 0.625 1.380 0.0744 0.0299 0.0198 0.2332 0.0915 0.0514 a only, for both 265.972 0.569 0.575 0.636 1.343 0.0599 0.0262 0.0184 0.1936 0.0758 0.0434 All , (but no a) for both 170.654 0.582 0.597 0.698 1.754 0.1308 0.0494 0.0291 0.3484 0.1467 0.0838 Education 13 for both 199.652 0.562 0.552 0.637 1.473 0.0864 0.0343 0.0221 0.2659 0.1054 0.0592 Experience f for both 217.070 0.579 0.594 0.666 1.472 0.1049 0.0521 0.0388 0.2651 0.1189 0.0754 Gender P for both 249.474 0.573 0.583 0.647 1.381 _ 0.0698 0.0290 0.0196 0.2160 0.0855 0.0487 Occupational Choice y for both sectors (and both 237.069 0.591 0.630 0.690 1.502 0.1048 0.0532 0.0398 0.2577 0.1176 0.0756 heads + others) y for both sectors (only for 241.081 0.580 0.603 0.663 1.422 0.0833 0.0344 0.0228 0.2391 0.0982 0.0568 other members) y, a, f for both sectors 217.070 0.579 0.594 0.666 1.472 0.1049 0.0521 0.0388 0.2651 0.1189 0.0754 Demographic Patterns ,u only, for all 275.264 0.573 0.583 0.702 2.420 0.0368 0.0114 0.0063 0.1724 0.0558 0.0266 >, r, a,d, for all 210.838 0.599 0.605 0.761 2.335 0.0997 0.0462 0.0339 0.2910 0.1215 0.0726 Education _p. only, for all 281.427 0.587 0.614 0.680 1.451 0.0648 0.0293 0.0209 _ 0.1985 0.0800 0.0469 _ , p[. for all 292.292 0.579 0.588 0.662 1.385 0.0498 0.0246 0.0188 0.1718 _0.0659 0.0386 p.t, .d, r, cx, 1P, for all 254.675 0.580 0.590 0.666 1.410 0.0774 0.0434 0.0348 0.2151 0.0937 0.0606 Source: Based on "Pesquisa Nacional por Amostra de Domicilios" (PNAD) of 1985 and 1996. = 28 Tables 7 - 9 contain a great wealth of information about a large number of simulated economic changes, always by bringing combinations of 1996 coefficients to the populations of 1976, 1981 and 1985. In order to address the two puzzles posed in the Introduction - namely the increase in extreme urban poverty between 1976 and 1996 despite (sluggish) growth and (mildly) reducing inequality; and the coexistence of a deteriorating labour market with stable 'headline' poverty - we now focus on a comparison of 1976 and 1996. To do so, we plot differences in the (logarithms) of incomes between the simulated distribution and that observed for 1976, for a number of the simulations in Table 7 16 Figure 6 below plots the combined price effects (aX and 0), separately for wage-earners and the self-employed. As can be seen, these effects were negative (i.e. would have implied lower income in 1976) for all percentiles. The losses were greater for wage earners than for the self- employed and, for the latter, were regressive. These losses are exactly what one would have expected from the downward shifts of the partial earnings-education and earnings-experience profiles, shown in Figures 4 and 5. [See Figure 6 in Appendix 41 In Figure 7, we adopt a different tack to the price effects, by plotting the income differences for each individual price effect simulation (for both sectors combined), and then aggregating them. As we would expect from figure 6, the returns to education and experience are both immiserizing. The change in partial returns to education alone is mildly equalizing (as can be seen from table 7). The change in the partial returns to experience are unequalizing, as well as being immiserizing. The change in the intercept, calculated at the mean values of the independent variables, was also negative throughout. This proxies for a 'pure growth' effect, capturing effects on earnings from processes not captured by education, experience, gender, or the unobserved characteristics of individual workers. It is intended to capture the effects of capital accumulation, managerial and technical innovation, macroeconomic policy conditions, and other factors likely to determine economic growth, not included explicitly in the Mincerian equation. Its negative 16 In computing these differences, we compare the percentiles of the two different distributions described above. A different, but equally interesting exercise, is to compare the percentiles of the simulated distribution ranked as in the observed 1976 distribution, with that 1976 distribution. These exercises were also performed, but are not reported 29 effect in this simulation suggests that these factors were immiserizing in urban Brazil, over the period. [See Figure 7 in Appendix 4] The one piece of good news, once again, comes from the gender simulation, which reports a poverty-reducing effect, as a result of the decline in male-female earnings differentials captured in Tables 5 and 6. This effect was far, however, from being sufficient to offset the combined negative effects of the other price effects. As the thick line at the bottom of Figure 7 indicates, the combined effect of imposing the 1996 parameters of the two Mincerian equations on the 1976 population was substantially immiserizing. Figure 8 below reiterates this point, separating the 'growth' effect (associated with a simple oc simulation), from the combined relative price effects (associated with a joint simulation of the vector 1). Note that, when combined in this form, the real price effect is on average incomes (and hence on poverty, but not so much on inequality). An inspection of the rows on table 7 confirms this observation. [See Figure 8 in Appendix 4] Figure 9 plots the (logarithm) of the income differences between the distribution which arises from imposing the 1996 occupational choice parameters (the y vector from the multinomial logit in Equation 4) on the 1976 population, and the observed 1976 distribution. It does so both for all individuals (the lower line), and for non-heads (the upper line). The effect of this simulated change in occupational choice and labour force participation behaviour is both highly immiserizing and unequalizing, as an inspection of the relevant indices in table 7 confirms. It suggest the existence of a group of people who, by voluntarily or involuntarily leaving the labour force, or entering unemployment, or being consigned to very ill-remunerated occupations (likely) in the informal sector, are becoming increasingly impoverished. In the unfavourable conditions of the Brazilian urban labour market of these two decades, which we have just documented above, these are people who appear to be failing to climb the slippery slope, and are becoming trapped in extreme poverty. due to space constraints. In any case, the plots which are presented are those which correspond to the summary statistics presented in tables 7-9. 30 [See Figure 9 in Appendix 4] Combining the negative price and occupational choice effects, one gains a sense of the overall effect of Brazil's urban labour market conditions over this period. This is done graphically in Figure 10, where the lowest curve plots the differences between the incomes from a distribution in which all as, Ps and ys change, and the observed 1976 distribution. It shows the substantially poverty-augmenting (and unequalizing) combined effect of changes in labour market prices and occupational choice parameters on the 1976 distribution. [See Figure 10 Appendix 4] At this point, the second puzzle can be stated clearly: given these labour market circumstances what factors can account for the facts that mean incomes rose, headline poverty did not rise, and inequality appears to have fallen slightly? The first part of the answer is shown graphically in Figure 11, where the upper line plots the differences between the (log) incomes from a distribution arising from imposing on the 1976 population the transformation (13) for the demographic structure of the population. The changes in the parameters ptd (and in the variances of the residuals in the corresponding regression) have a positive effect on incomes for all percentiles, and in an equalizing manner. However, when combined in a simulation in which the values of all as, P3s and ys also change, it can be seen that the positive demographic effect is still overwhelmed. Nevertheless, it is clear that the reduction in dependency ratios, and subsequently in family sizes, in urban Brazil over this period had an important mitigating effect on the distribution of incomes. This can be seen clearly in Figure 12, where the bottom line from Figure 11, which incorporates the demographic effect, is superimposed on the simulations that exclude it, from the changes in labour market parameters only. The line incorporating the demographic effects lies essentially everywhere above the line for all as, Ps and ys only, indicating a smaller loss in incomes everywhere. Note, however, that while this relative gain is particularly pronounced for the 'quasi-poor' (say, from the 5t" to the 25 percentile), it is less pronounced below that level, where the new extreme poor are located. [See Figures 11 and 12 in Appendix 4] 31 There remains one final piece of the puzzle, necessary to explain why the deterioration in labour market conditions did not have a worse impact on poverty. That, as should be evident from the increase in mean years of effective schooling, registered in Table 1, is the rightward shift in the distribution function of education. This is shown in Figure 13, which reveals that gains in educational attainment were particularly pronounced at lower levels of education, and thus presumably among the poor. [See Figure 13 in Appendix 4] A gain in educational endowments across the income distribution, but particularly among the poor, has both direct and indirect effects on incomes. The direct effects are through equations (2) and (3), where earnings are positive functions of schooling. The indirect effects are both through the occupational choices that individuals make, and through the further impact that education has on reducing the demand for children, and hence family size. A simulation of the effect of education is thus quite complex. After it is completed, one observes, in Figure 14, a rather flat improvement in (log) incomes across the distribution (i.e. a scaling effect). However, when this is combined with changes in the parameters of the demographic equations again, the effect gains strength, and becomes not only more poverty-reducing, abut also mildly equalizing. The bottom line in Figure 14, in keeping with the pattern, combines both of these effects with the changing cas, P3s and ys. The result is striking: this complex combined simulation suggests that all of these effects, during twenty turbulent years, cancel out almost exactly from the 159 percentile up. Hence the small changes in headline poverty. However, from around the 12' percentile down, the simulation suggests a prevalence of the negative occupational choice (and to a lesser extent, price) effects, with substantial income losses. These account for the rise in indigence captured by the R$30/month poverty line. [See Figure 14 in Appendix 41 The bottom line in Figure 14 is, in a sense, the final attempt by this methodology to simulate the various changes that led from the 1976 to 1996 distribution. Figure 15 is a graphical test of the approach. Here, the line denoted "1996-1976" plots the differences in actual (log) incomes 32 between the observed 1996 and the observed 1976 distributions. Along with it, we have also plotted every (cumulative) stage of our simulations. First the immiserizing (but roughly equal) price effects; then these combined with the highly immiserizing occupational choice effects; then the slightly less bleak picture arising from a combination of the latter with the parameters of the family size equations. And finally, the curve plotting the differences between the incomes from the simulation with all parameters changing, and observed 1976. As can be seen, it would not appear that the last line replicates the actual differences badly. Of course, the point of the exercise is not to replicate the actual changes perfectly, but rather to learn the different effects of different parameters, and possibly to infer any policy implications from them. But the success of the last simulation in approximately matching the actual changes does provide some extra confidence in the methodology, and in any lessons we may derive from it. [See Figure 15 in Appendix 4] 7. Conclusions In the end, does this exercise help us improve our understanding of the evolution of Brazil's urban income distribution over this turbulent twenty-year period? Whereas many traditional analysts of income distribution dynamics might have inferred, from the small changes in mean income, in various inequality indices, and in poverty incidence1", that there was little - if anything - to investigate, digging a little deeper has unearthed a wealth of economic factors interacting to determine substantial changes in the environment faced by individuals and families, and in their responses. In particular, we have found that, despite a small fall in measured inequality (although the Lorenz curves cross as expected, see Figure 2b) and a small increase in mean income, extreme poverty has increased, for sufficiently low poverty lines or sufficiently poverty aversion parameters. This seems to have been caused by outcomes related to participation decisions and occupational choices, in combination with declines in the labour market returns to education and experience. These changes are associated with greater unemployment and informality, as one would expect, but more research into them seems necessary. While we seem to have identified the existence a group excluded from both the productive labour markets and any substantive 17 With respect to the already low R$ 60/month poverty line, by historical standards for Brazil. 33 form of safety net, we have not been able to fully interpret the determinants of their occupational choices. Issues of mobility - exacerbated by the current monthly income nature of the welfare indicator - will also require further understanding in this context. Policy implications would seem to lie in the area of self-targeted labour programmes, or other safety nets, but it would be foolhardy to go into greater detail before the profile of the group which seems to have fallen into extreme poverty in 1996 is better understood. Secondly, we have found that, even above the 15 percentile, where urban Brazilians have essentially 'stayed put', this was the result of some hard climbing along a slippery slope. They had to gain an average of two extra years of schooling (which still leaves them undereducated for the country's per capita income level), and substantially reduce fertility, in order to counteract falling returns in both the formal labour market and in self-employment. It may well be, as many now claim, that an investigation of non-monetary indicators - such as access to services, or life-expectancy at birth - should lead us to consider the epithet of 'a lost decade' as too harsh for the 1980s. Unfortunately, we find that if one is sufficiently narrow- minded to consider only money-metric welfare, urban Brazil has in fact experienced two, rather than one, lost decades. 34 References. Almeida dos Reis, J. G. and R. Paes de Barros (1991): "Wage Inequality and the Distribution of Education: A Study of the Evolution of Regional Differences in Inequality in Metropolitan Brazil", Journal of Development Economics, 36, pp. 117-143. Bonelli, R and G. L. Sedlacek (1989): "Distribuicao de Renda: evolucao no ultimo quarto de seculo", in Sedlacek, G. L. and R. Paes de Barros, Mercado de Trabalho e Distribuicao de Renda: Uma Coletanea, (Rio de Janeiro: IPEA; Serie Monografica #35). Bourguignon, F., F. Ferreira and N. Lustig (1998): "The Microeconomics of Income Distribution Dynamics in East Asia and Latin America", World Bank Research Proposal, (Washington, DC: World Bank, April). Buhmann, B., L. Rainwater, G. Schmaus and T. Smeeding (1988): "Equivalence Scales, Well- being, Inequality and Poverty: Sensitivity Estimates Across Ten Countries using the Luxembourg Income Study Database", Review of Income and Wealth, 34, pp. 1 15-142. Cowell, F. A. (1995): Measuring Inequality (2nd Edition), (Hemel Hempstead: Harvester Wv7heatsheaf). Duryea, S. and M. Szekely (1998): "Labor Markets in Latin America: A Supply-Side Story", Paper prepared for the IDB/IIC Annual Meeting, Cartagena de Indias, Colombia. Ferreira, F. H. G., P. Lanjouw and M. Neri (1999): "The Urban Poor in Brazil in 1996: A New Poverty Profile Using PPV, PNAD and Census Data", background paper for the World Bank's Brazil Urban Poverty Assessment, 1999; (Washington, DC: The World Bank). Ferreira, F. H. G. and J. A. Litchfield (1996a): "Growing Apart: Inequality and Poverty Trends in Brazil in the 1980s", LSE - STICERD - DARP Discussion Paper No.23, London (August). Hoffiman, R. (1989): "Evolucao da Distribuicao da Renda no Brasil, Entre Pessoas e Entre Familias, 1979/86", in Sedlacek, G. L. and R. Paes de Barros, Mercado de Trabalho e Distribuicao de Renda: Uma Coletanea, (Rio de Janeiro: IPEA; Serie Monografica #35). Juhn, C., K. Murphy and B. Pierce (1993): "Wage Inequality and the Rise in Returns to Skill", Journal of Political Economy, 101 (3), pp. 410-442. Macrometrica (1994): "Inflacao: Primeiros Meses do Real", Boletim Mensal Macrometrica, #111, (July-August). Mookherjee, D. and A. F. Shorrocks (1982): "A Decomposition Analysis of the Trend in UK Income Inequality", Economic Journal, 92, pp. 886-902. Ramos, L. (1993): A Distribuicao de Rendimentos no Brasil: 1976/85 (IPEA: Rio de Janeiro). 35 Appendix 1: Data and Methodology Macroeconomic Data. All macroeconomic indicators reported in this paper are based on original data from the archives of the Brazilian Statistical Institute (IBGE). GDP and GDP per capita figures reported in Section I come from the series shown below in Table Al. This series was constructed from the current GDP series (A), which was revised in 1995 and backdated to 1990; and from the old series (B), from 1976 to its final year: 1995. The series reported below comprises the values of series A from 1990 to 1996, and the values of series B scaled down by a factor of 0.977414 from 1976 to 1989. This factor is the simple average of the ratios A/B over the years from 1990 to 1995. The series is expressed in 1996 Reais, using, the IBGE GDP deflator. Table Al: Real GDP and GDP per capita, Brazil 1976-1996, annual (constant 1996 prices) Year GDP (reais) Population (,OOOs) GDP per capita (reais) 1976 434,059,220 107,452 4040 1977 455,477,123 110,117 4136 1978 478,113,823 112,849 4237 1979 510,432,394 115,649 4414 1980 562,395,141 118,563 4743 1981 538,474,976 121,213 4442 1982 542,971,306 123,885 4383 1983 527,054,370 126,573 4164 1984 555,515,747 129,273 4297 1985 599,129,793 131,978 4540 1986 644,002,821 134,653 4783 1987 666,708,887 137,268 4857 1988 666,304,312 139,819 4765 1989 687,391,828 142,307 4830 1990 651,627,236 144,091' 4522 1991 658,339,124 146,408 4497 1992 654,759,303 148,684 4404 1993 687,004,026 150,933 4552 1994 727,213,139 153,143 4749 1995 757,918,030 155,319 4880 1996 778,820,353 157,482 4945 The GDP per capita growth rates plotted in Figure 1 are derived from this series. Annual inflation and unemployment rates also come from the relevant IBGE series. 36 The PNAD data sets All of the distributional analysis performed in this paper is based on four data sets (1976, 1981, 1985, 1996) of Brazil's National Household Survey (Pesquisa Nacional por Amostra de Domicilios: PNAD), which is fielded annually by the IBGE. For the latter three years, the survey is nationally and regionally representative, except for the rural areas of the North region (minus the state of Tocantins) which are not surveyed. For 1976, rural areas were surveyed neither in the North nor in the Center-West regions. In this paper, we are concerned only with urban areas, which are defined by state-level legislative decrees. The urban proportions of the population in each year are given in Table 1. The PNAD sample sizes, as well as the proportion of missing income values, are given below in Table A2: Table A2: PNAD Sample Sizes and Missing or Zero Income* Proportions Year Number of Number of Proportion of Proportion of households Individuals individuals with individuals with RFPC missing RFPC = zero 1976 84660 385282 0.0052 0.0063 1981 110151 477607 0.0073 0.0141 1985 127128 520069 0.0073 0.0108 1996 91621 329434 0.0291 0.0313 Note: *: Income is Total Household Income per capita (RFPC). Each PNAD questionnaire contains a range of questions pertaining both to the household and to individuals within the household. Among the former, are questions about regional location, demographic composition, quality of the dwelling, ownership of durables, etc. The latter include age, gender, race, educational attainment, labor force status, sector of occupation and incomes, both in cash and kind, and from various sources. The main variables used in our analysis are the those related to incomes, education, the demographic structure of the household and labor force participation. Tables A6 - A9 summarize the main items in the questionnaire concerning these variables, and the changes from 1976 to 1996. Most importantly, the distributions analyzed in this paper (except where explicitly otherwise indicated) have as welfare concept total household income per capita (regionally deflated). It is constructed from summing all income sources for each individual within the household, and across all such individuals, except for lodgers or resident domestic servants. The latter two categories constitute separate households. Total nominal incomes are spatially deflated to compensate for differences in average cost-of-living across different areas in the country, according to the spatial price index given in Table A3 below: 37 Table A3: A Brazilian Spatial Price Index (RM Sao Paulo = 1.0) PNAD Region Spatial Price Deflator RM Fortaleza 1.014087 RM Recife 1.072469 RM Salvador 1.179934 Northeast (other urban areas) 1.032056 Northeast Rural 0.953879 RM Belo Horizonte 0.958839 RM Rio de Janeiro 1.002163 RM Sao Paulo 1.000000 Southeast (other urban areas) 0.904720 Southeast Rural 0.889700 RM Porto Alegre 0.987001 RM Curitiba 0.987001 South (other urban areas) 0.904720 South Rural 0.889700 RM Belem 1.088830 North (other urban areas) 1.032056 RM Brasilia 1.037915 Center West (other urban areas) 0.968388 Note: This regional price index is based on the consumption pattems and implicit prices from the PPV 1996 survey, for the Northeast and Southeast regions, and extrapolated to the rest of country according to a procedure specified in Ferreira et. al. (1999), where the exact derivation of the index is also discussed in detail. We assume, largely due to the lack of earlier comparable regional price information, that the structure of average regional cost-of-living described above remained constant over the period. Temporal deflation was undertaken on the basis of the Brazilian consumer price indices IGP-DI (for 1976), and INPC-R for the three subsequent years. For 1996, the INPC-R was upwardly adjusted by 1.2199, so as to compensate for the actual price increases which took place in the second half of June 1994, and which were not computed into July's index, since the latter was already computed in terms of the URV. This adjustment is becoming the standard deflation procedure at IPEA when comparing incomes across June/July 1994. (See Macrometrica, 1994). In order to center the indices on the first day of the month, which is the reference date for PNAD incomes, the geometric average of the index for a month and for the preceding month was used as that month's deflator. Once again, this is now best practice for price deflation in hyper- inflationary periods. Once the deflators were constructed in this way, the values to convert current incomes into 1996 Reais were as follows: Table A4: Brazilian Temporal Price Deflators (Selected Years) 1976 1 4.115 1 1981 1 49.512 1 1985 1 2257.294 1 1996 1.000 A final possible adjustment to the PNAD data concerns deviations between survey-based welfare indicators (such as mean household income per capita) and National Accounts-based prosperity indicators (such as GDP per capita). The international norm is that household survey means are lower than per capita GNP, both because the latter includes the value of public and publicly 38 provided goods and services, which are generally not imputed into the survey indicators, and because of possible under-reporting by respondents. Given that the levels of the two series are not expected to match exactly, analysts are usually concerned by deviant trends, which may indicate a problem with the survey instrument. On the other hand, it may plausibly be argued that National Accounts data have errors of their own, and that many of the 'correction' procedures applied to household data rely on reasonably strong assumptions, such as equiproportional under- reporting by source. In deciding whether to adjust the PNAD data with reference to the Brazilian National Accounts over this period, we examined the evolution of the ratios of GDP per capita to mean household incomes from the PNAD (for the entire country, and without regional price deflation, for comparability). As Table A5 below shows, these were remarkably stable. In particular, the ratios for the starting and end points of the period covered, which are of particular importance for our analysis, are almost identical. In this light, and since even the disparity with respect to 1981 and 1985 are reasonably small, we judged that the costs of making rough adjustments to the PNAD household incomes on the basis of the National Accounts outweighed the benefits. Table A5: Ratios of GDP per capita to PNAD mean household incomes, 1976-1996 Year GDP per capita (A) Mean PNAD income (A) / (B) (B) 1976 336.6 190.2 1.770 1981 370.2 187.3 1.976 1985 378.3 188.6 2.005 1996 412.1 233.0 1.769 Tables A6 - A9 summarize the main items in the questionnaire concerning these variables, and the changes from 1976 to 1996. 39 Table A6: Comparing Income Variables across the 1976 and 1996 PNADs 1976 1996 Variable Name Question Variable Name Question V2308 Reodimaata e.i. o Quanta ganba as gunbava meotalmenir aa ocupu5o V9532 Reud/Me.sal nes.s Qual era a -ecdim-rto mensal que gauhava -ormalnente, deelarada an quaita 4 (ocupac9o/ peartssuo que -seree em setembro de 1996, nesse traba1ho (pr-inipal - em dinheira)? a -u-ru durante mais tempo)? V2358 Re.dimento-Variiv-l V2359 Rendi..t. TV9535 Rend/Mens-l nense Qurt ens o enudimento mensal qu ganhava normalmente, Prod/Mee-ada em setembro de 1996, nesse trabalhb (principal am me.eadoesas oa valor dos produtos)? V2362 Outra Raida - Outer Ocupap9o Tem renda habitual alm do delarada no quesito 8 -V23 12 ? V9982 Rend/Mnnsal no tea Quai era o rendimerl a manual quo ganhava normalmeuse, em setembro do 1996, nes tsabalha ueuundario.(rm dieheio) 7 V9985 Rend/Mental no tea Qual era o rediamntu musesl sle gan-h-uanmrlmentu, em sesembro de 1996, ne-se trabalho seuundleio( em merecadarias oa valor dos prodntos)? V1022 Reed Mhs 9 Nouteas Qual .ean rendimento mental qu- gruhbva norm. lmente, em sesembro de 1996, not ate s teabalhos quo sisba na stmana do 22 a 28 de setembro do 1996. (rm di.ebine)? V1025 Rend MWs 9 Noutros Qual eea o eradimouta mossal que ganbava nermalmerte, em scttmbra de 1996, nos uutros tbabalhts qu- tinho or semana de 22 a 28 de setemb-o de 1996, ( -em morcdorias at valor dos p-edut-n)? V2365 Ontra Reada - Apase.tudoiri/Penuo Tem ren.da babit-a a16m do de1rdaa quetito 8 ,V2312 ? V1252 Valor I Reod apos Peru Rec Qual -ra o eredimento mensal qae retebia aarualmeetr em setembre de 1996, de apusatladoria dv instituto de previdn-eia ot do gaueno federal (em dinheiro)? V1255 V.I., I Rend Pent Prev Req Qual ra o rendimeuto measal que -e-cbia n-rmalmeute em sosembee d& 1996, do peuslo de instituta do previdencia on do goveno. federal (am diobeire)? V 1258 Valar I Read Outra up6s Re Qual era o rendimneto mensal que recebias ormalmunto em setembro do 1996, de oulva tip. de apasontadania (RS)7 V1261 Value I R-ctd Outer Pens Re Qual -ra orndimunto mental que -reebia narmalmeete em setembra do 1996, autro tip. de presto (RS)? V2363 Otra Reoda - Alugadis Tem reeda habital a16m da d-elurada no quesite 8 -V2312 7 V 1267 Rend Aleguel Rev Qual era o reudisuata ma. er. - qe eebia neemalmaule em untembeo d& 1996, de alug-el (RS)? V2364 Outer Renda - Doaq5o/Mesada Tem reuda habitual al1m da de"l-rada so quesite 8 -V2312 ? V1270 Rend Don0c0 Rec Na Qual er,o reudimenta masu - q ree-bia nanmaImente em setmb-ro de 1996, de doaa6o -rurbida do rIo morador (RS)7 V2366 Ostra Reoda - Outeas Tem r-uda hobitual al6m da deladno qu-sia 8 -V23 12 ? V1273 Rend Ilras Re-nbid Qnal ura redimento mens-I que -ecebia normalmenle em sntumbeo de 1996, deju.os do cadesneta de paupanqa a do ontrn aplicaqOes, divideudos a o e-o runubimentos (RS)? V1264 Valou I Rend Ab.na Prom Re Q.al era a nendimenta mensal que recebia uurmatmunme em ueteubra do 1996, de abano de peemauducia (RS)? V2956 R.mn..en..o Tudna Ocup. Vn957 Remu. re-gIo Ourt. Readl. A Som. de.sas dar.e Igual a Sams do tOdas as aurora Fnte: Constealda Com base -oa qsestion-Aias a dicianarias da Pesquisa Naciatal por Amestra e Damijilies (PNAD) de 1976 e 1996. 40 Table A7: Comparing Education Variables across the 1976 and 1996 PNADs 1976 1996 Variable Name Categones Variable Name Categories V2222 Sabe ler e escrever - menos de 5 anos V0601 sabe ler e escrever 0- parte ignorada I -Sim l-sim 2- Esqueceu 3- nao 3- Nao sabe 9- ignorado 9- Sem declara,cao -nao infomiado V2223 Onde aprendeu a ler e escrever - menos de 5 anos I - Escola regular 2- Outra forma 3- Nfo sabe ler e escrever 9- Sem declaracao V2224 Frequenta escola - serie - Nao aplicavel V0602 Frequenta escola ou creche 2- sim 0 - nao hi serie 4- nao I - l"serie 8- se v0601=1 ou 3 demais variaveis da par 2- 2 s&ie 9-ignorado 3- 30 serie - nao informado 4- 4 serie V0605 Qual a serie que trequenta 1- primeira 5- 50 srie 2- segunda 6- 60 serie 3- terceira 7- 70 serie 4- quarta 8- 8' serie 5- quinta 9 - sem declara,cao 6- sexta 7- s6tima 8- oitava 9- ignorado - nao informado V2225 Frequenta escola - grau - Nao aplicavel V0603 Qual o curso que frequenta 0-ignorado 0- nao ha s&ie 1-regular de 1° grau 1- primeiro grau 2- regular de 20 grau 2- segundo grau 3- supletivo de 10 grau 3- medio prim. Ciclo 4- supletivo de 20 grau 4- rmedio seg. ciclo 5- superior 5- superior 6- alfabetizagao de adultos 6- alfabetiza,ao de adultos 7- pre-escolar 7- admissao 8- pre-vestibular 8- supletivo 9- mestrado ou doutorado 9- Art.99 prim. Cielo -nao informado 10 - Art99 seg. ciclo I l-vestibular 99- sem declara,co V0604 curso e seriado7 2- sim 4- nao 9- ignorado -ngo informado 41 Table A7 (ctd): Comparing Education Variables across the 1976 and 1996 PNADS V2226 Nao frequenta escola - serie - Nao aplicavel V0606 Anteriorinente frequentou 2- sim 0 - nao ha serie escola ou creche? 4- nao I - I' serie V067 Qual tio o curso mais elevado que 0- ignorado ~ ~ 2-2 serie frequentou anteriormente? 1- elementar(primario) 3- 3' s6rie 2- medio primeiro ciclo(ginasial) 4- 4'serie 3- inmdio segundo ciclo 5- 5' serie 4- primeiro grai 6- 6 s6rie 5- segundo grau 7- 7as6rie 6- superior 8- 8'serie 7- mestrado ou doutorado 9 - sem declaragao 8- alfabetizagao de adultos V2227 Nao trequenta escola - grau - No aplicavel 9- pre-escolar 0- nao ha serie -nao informado 1- primeiro grau 2- segundo grau 3- medio prim. Ciclo 4- medio seg. ciclo 5- superior 9- sem declaragao V060g Este curso que trequentou 2- aim anteriormente era seriado 4- nao 9- ignorado -n8o informado V0609 roi aprovado pclo menos na 1-aim primeira s6rie deste curso que 3- nao frequentou anteriormente 9- ignorado - nao informado V0610 Qual toi a srice que concluiu I-primeira com aprovagao neste curso 2-segunda que frequentou anteriormente 3-terceira 4- quarta 5- quinta 6- sexta 7-setima 8- oitava 9- ignorada -n8o in formada V06 11 Concluiu ese curso que I - aim frequentou anteriormente 3- nao 9- ignorado - nao informado Fonte: Construida com base no questlonarios e dicionarios dia tesquisa Nacional por Amoatra e Domicihos (PNAD) de IY76 e 1996: 42 Table A8: Comparing Labour Market variables across the 1976 and 1996 PNADs 1976 1996 Variable Name Categories Variable Name Categories V2301 Qae fez na semana? -Menos de 10 anos V9001 Trabalhou de 22 a 2S/9/96? 0-Parte Ignorada 0-Sem ocupagao I-Sim I -Estava trabalhando 3-Nao 2- Tinha trabalho 9-Ignorado 3-Procur. Trabalho Nao informado 4-Proc. Trab. I vez 5-Afazeres dom6stico 6-Frequent. Escola 7-Aposenrt/Pesion 8-Vive de renda 9-Doente/invalido V9002 Trab Cultivo, Pesca, Criaplo 2-Sim 4-Nao 9-Ignorado Nao informado V9003 Trab. Constru,ao do prTprio uso? I-Sim 3-Nao 9-Ignorado Nao informado V9004 Afastado temporariamente 2-Sim 4-N,ao 9-Ignorado Nao informado V2307 PosicAo na Ocupaeao 0-Sem declaraeao V9008 Neste trabalho era? I-Empregado permanente I -Empregado 2-Empregado permanente agricultura 2-Conta-pr6pria 3-Empregado permanente outra atividade 3-Conta Prop NAo Est. 4-Empregado tempor6rio 4-Parceiro Emreg. 5-Conta-pr6pria nos servipos auxiliares 5-Parc. Conta Prop. 6-Conta-pr6pria na agricultura 6-Parc. Empregador 7-Conta-pr6pria em outra atividade 7-Membro da Familia 8-Empregador nos servicos auxiliares 9-Membro de Inst 9-Empregador na agricultura I o-Empregador em outra atividade I I-Trabalhador nao remunerado 12-Outro trabalhador nao remunerado 13-Trabalhador na producao 88-Tem ativ.agricola e nao inform. Pos ocup 99-Ignorado Nao informado 43 Table AS (ctd): Comparing Labour Market Variables across the 1976 and 1996 PNADS V9029 No Empr. tinha area p/pr Partic ]-Sim 3-Ngo 9-Ignorado Nao informado V9032 Este emprcgo era no setor? 2-PFivade 4-Publico 9-Ignorado Nao mfwormado V9033 esse emprego era noi-are Im 3 -E=td.al 5-M-niipa1 NSo iufornado V9034 Nesse emnprcgo ca osilitor?, 2-Sim 4-Nio 9-Ignorado NSo informado V9035 Nessenmprego era Func. Pub. Estatr I-Sim 3 Nao 9-1gnonado NA. mnformado V9042 Ncsse Empr tinha Cart. Trb. Ass. 2-Sin 4-N So 9-Ignor-do NSa inforrmado V2323 Meio p/consegairteabalho S-Se-declarago V9t IS Provideneio Tra na SeS. referacia? I.Sim I -AgZncia Publica 3-Nao 2-Agdncia Particalar S-Sem resposta nos quesitos de proc. Trab 3-Dirmito Empreg. 9-Ignorado 4-Amigos/Parentes Nao infarmado 5-Colegas Profiss 6 Anuncios V91 16 Provideaciou Trab. no mes rfernencia? 2-Sim 7-Recebeu proposta 4-Nao 8-Outra 9-4gnorado 9-Nada Fez Nao informado Fonte: Construida com base nos qaestionarios e dicionarios da Pcsquisa Nacional por Amostra e Domicilios (PNAD) de 1976 e 1996. 44 Table A9: Comparing Some Demographic Variables across the 1976 and 1996 PNADs 1976 1996 Vanable Name Categones Variable Name Categories V1004 Sltuacao 1-Urbana V4728 Situac3o 1-Urbana-area urbana 2-Rural 2-Urbana-area n0o urbana 3-Urbana-area isolada 4-Rural-extensao urbana 5-Rural-povoado 6-Rural-nuicleo 7-Rural-outros 8-Rural-exclusive -Nao informado V210/ Condic3 no domicilio I -Chete de tamiiia V0402 Condic3o na familia l -Pessoa de reterencia 2-Conjuge 2-Conjuge 3-Filho (a) / enteado 3-Filho 4-Pais ou sogros 4-Outro parentc 5-Outros parentes 5-Agregado 6-Agregado 6-Pensionista 7-Pensionista / Hospede 7-Empregado domestico 8-Empregado domcstico 8-Parente do empregado domestico 9-Individual dom. col nao informado V0403 Numero da famiiia V1401 Familia -N3o Aplicavel 1-Unica 2-Individual 3-Principal 4-Primeira secund. 5-Segunda secund. V1402 ispecie -Nao Aplicavel V0201 Especie de domicilio 1-Particular Permanente 1-Particular 3-Particular Improvisado 2-Coletivo 5-Coletivo 3-Improvisado Nao informado V1012 fipo de area i-Area metropolitana V4727 tIpo de area I-RegiAo metropolitana 2-Auto repreresentativa 2-Auto repreresentativa-nao metropolitana 3-N3o auto representativa 3-Nao auto representativa Fonte: Construida com base nos questionirios e dicionirios da Pesquisa Nacional por ATnostra e Doomicilios (PNAD) de 1976 c 1996. Nota: A variavel Niuiero da Familia (V0403) esta presente uo programa de 1996 e nao esta presente no prograuna de 1976, pois esta PNAD so esta disponivel a nivel de domicilio. 45 Appendix 2: Table A2.1: Evolution of mean income and inequality: a summary of the literature Ano 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 Household In come per capita Bonelli & Sedlaceck Gini Coefficient 0.561 0.550 0.542 0.549 Gini Coefficicnt (1) 0.583 0.588 0.584 0.589 0.592 Hoffman (2) Mean (3) 4.7 4.8 4.6 4.7 3.8 4.0 4.5 5.6 Gini Coefficient 0.588 0.597 0.584 0.587 0.589 0.588 0.592 0.586 Theil-T 0.523 0.536 0.519 0.520 0.523 0.526 0.529 0.519 Ferreira e Litchfield Mean (4) 143 126 125 150 213 166 166 196 164 Gini Coefficient 0.574 0.584 0.577 0.589 0.581 0.582 0.609 0.618 0.606 Theil - T 0.647 0.676 0.653 0.697 0.694 0.710 0.750 0.796 0.745 Total Individual Income (Active Pop.) Bonelli & Sedlaceck (5) Mean(6) 2241.8 2081.2 2264.0 2040.6 1835.6 2222.1 3112.8 Gini Coefficient 0.589 0.574 0.590 0.562 0.582 0.588 0.577 Hoffman (1) (7) Mean (8) 340.2 331.2 297.5 293.6 335.7 426.1 Gini Coefficient (9) 0.585 0.572 0.591 0.587 0.599 0.589 Lauro Ramos (10) Mean (11) 85.4 87.5 89.7 93.6 93.4 91.9 86.8 89.2 94.6 Gini Coefficient 0.564 0.543 0.531 0.530 0.514 0.520 0.534 0.536 0.545 Theil-L 0.556 0.511 0.488 0.486 0.457 0.465 0.496 0.498 0.521 Theil-T 0.709 0.607 0.571 0.560 0.513 0.527 0.565 0.558 0.584 Fontte: Hoffman (89) - Pesquisa Nacional por Anostra de Domicilios (PNAD) de 1979, 1981, 1982, 1983, 1994, 1985 e 1986, Censo Demografico 1980 e Anuio esatistico 1985 para os anos dc 1979, 1980, 1981, 1982 c 1983. Bonelli & Sedlaceck (89)- Pesquisa Nacional por Aniostra de Domicilios (PNAD) de 1976, 1979, 1981, 1983, 1985, 1986 e Censo Dernografico de 1980 Lauro Ramos (90)- Pesquisa Nacional por Amostra dc Domicilios (PNAD) de 1976, 1977, 1978, 1979, 1981, 1982, 1983, 1984 e 1985. Ferreira e 1,itchfield (96) E Pesquisa Nacional poT Amostra de Domicilios (PNAD) de 1981, 1983, 1984, 1985, 1986, 1987, 1988, 1989, 1990. Nota: (1) - Inclui as faTnilias sem renda. (2) - Para 1979: exclusive as popusla.6cs das zonas rurais da regiao norte, Matogrosso, Matogrosso do Sul e Goias e para 1981 a 1986: exclusive a populacao da zona rural da regiao norte. (3) - Valor real, en) salarios niinimos de agosto de 1980, deflacionado pelo ICV-DIFESE. (4) - US$ de 1990, (5) - Exclusive pessoas sem rendimentos ou sem declara,ao de rendimentos. Somente PEA. (7) - Somente PEA c/ rendimento positivo (6) - Deflator:IGP/IBGE. Precos de Cz$ 1000 de scVt86; exclusive 7ona rural da regiao norte (todos os anus), c zona rural de Matogrosso, Matogrosso do Sul c C3oiis (76 e 79). (8) - Valores em 1000 cruzeiros de setV84. Deflatores: INPC.IBGE, ate ago/85; ICVIDIEESE, entre set/85-setV86. (9) - Media ponderada dos valores minimo e maximo. (10) - Universo: homens entre 18 c 65 anios, paxticipasido da forca de trabalho, trabalbando nmais de 20 horas por semana e nmorando em irca uTbanas; ressda total (11) - Base: 1980=100. Appendix 3: The Estimation of the Model: Regression Results 9v 47 Table A 3.1: Dependent variable: participation of the household head Year 1976 1981 1985 1996 Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Occupied as employee versus unoccupied Education 0,008 0,010 0,444 0,025 0,008 0,002 -0,004 0,008 0,598 0,001 0,008 0,881 Education2 0,002 0,001 0,000 0,003 0,001 0,000 0,003 0,000 0,000 0,002 0,000 0,000 Age 0,039 0,007 0,000 0,097 0,005 0,000 0,104 0,005 0,000 0,153 0,005 0,000 Age2 -0,001 0,000 0,000 -0,002 0,000 0,000 -0,002 0,000 0,000 -0,002 0,000 0,000 Gender -1,833 0,040 0,000 -1,415 0,029 0,000 -1,291 0,027 0,000 40,922 0,025 0,000 Number ofmembers fromO to 14* -0,001 0,009 0,905 0,024 0,007 0,001 0,029 0,008 0,000 40,015 0,010 0,124 Number ofrmembers from 14 to 65* 40,052 0,011 0,000 40,050 0,009 0,000 40,049 0,009 0,000 -0,079 0,012 0,000 Number ofmembers older than 65* 0,076 0,049 0,121 0,000 0,041 0,991 0,001 0,040 0,982 0,006 0,045 0,892 Presence ofother members from 14 to 65 (dummy) -0,008 0,139 0,953 0,659 0,106 0,000 0,922 0,103 0,000 0,494 0,106 0,000 Mean education* -0,092 0,012 0,000 40,056 0,011 0,000 -0,083 0,010 0,000 0,009 0,011 0,410 Mean education2* 0,001 0,001 0,286 -0,001 0,001 0,316 -0,001 0,001 0,187 40,004 0,001 0,000 Mean age* 0,026 0,008 0,001 0,002 0,006 0,711 0,001 0,006 0,806 0,004 0,006 0,510 Mean age2* 0,000 0,000 0,000 0,000 0,000 0,014 0,000 0,000 0,003 0,000 0,000 0,014 Women proportion* 0,012 0,006 0,025 -0,010 0,004 0,018 0,003 0,004 0,508 0,000 0,004 0,940 Constant 2,465 0,167 0,000 0,403 0,107 0,000 0,518 0,103 0,000 -1,052 0,105 0,000 Occupied as self-employed versus unoccupied Education -0,063 0,011 0,000 -0,037 0,009 0,000 40,061 0,009 0,000 0,009 0,009 0,340 Education2 0,001 0,001 0,233 0,000 0,001 0,518 0,001 0,001 0,115 -0,002 0,001 0,000 Age 0,072 0,008 0,000 0,130 0,006 0,000 0,121 0,005 0,000 0,175 0,006 0,000 Age2 -0,001 0,000 0,000 40,002 0,000 0,000 -0,002 0,000 0,000 40,002 0,000 0,000 Gender -1,719 0,047 0,000 -1,452 0,035 0,000 -1,412 0,032 0,000 -1,479 0,033 0,000 Number ofmembers firomOto 14* 0,030 0,010 0,002 0,075 0,008 0,000 0,096 0,008 0,000 0,055 0,011 0,000 Number ofmembers from 14to65* 40,055 0,012 0,000 40,049 0,010 0,000 -0,071 0,010 0,000 -0,090 0,013 0,000 Number of members older than 65* 0,036 0,056 0,522 0,067 0,046 0,145 -0,061 0,046 0,184 -0,081 0,051 0,113 Presenceofother membersfroml4to6S(dumny) -0,015 0,159 0,925 0,689 0,124 0,000 0,909 0,118 0,000 0,469 0,126 0,000 Mean education* -0,090 0,014 0,000 -0,036 0,013 0,004 -0,055 0,012 0,000 0,039 0,013 0,002 Mean education2* 0,003 0,001 0,004 0,000 0,001 0,674 0,000 0,001 0,638 -0,003 0,001 0,000 Mean age* 0,016 0,009 0,074 -0,016 0,007 0,020 -0,013 0,007 0,040 -0,002 0,007 0,740 Mean age2* 0,000 0,000 0,001 0,000 0,000 0,700 0,000 0,000 0,484 0,000 0,000 0,248 Womenproportion* 0,007 0,007 0,287 -0,002 0,005 0,670 0,002 0,005 0,660 40,014 0,004 0,001 Constant 0,646 0,188 0,001 -1,513 0,129 0,000 -1,152 0,121 0,000 -2,860 0,131 0,000 Source: Based on "Pesquisa Nacional por Amnostra de Domicilios" (PNAD) ofthe 1976 and 1996. Note: * excluding the head. 48 Table A 3.2: Dependent variable: participation of other members Year 1976 1981 1985 1996 Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Occupied as employee versus unoccupied Education 0,207 0,006 0,000 0,122 0,006 0,000 0,117 0,006 0,000 0,073 0,006 0,000 Education2 -0,006 0,000 0,000 0,003 0,000 0,000 0,002 0,000 0,000 0,003 0,000 0,000 Age 0,333 0,004 0,000 0,314 0,003 0,000 0,315 0,003 0,000 0,303 0,003 0,000 Age2 -0,005 0,000 0,000 -0,005 0,000 0,000 -0,005 0,000 0,000 -0,004 0,000 0,000 Gender -1,399 0,018 0,000 -1,143 0,015 0,000 -1,152 0,014 0,000 -0,860 0,017 0,000 Number of members from 0 to 14* -0,105 0,005 0,000 -0,091 0,004 0,000 -0,101 0,004 0,000 -0,123 0,006 0,000 Number of members from 14 to 65* 0,229 0,005 0,000 0,205 0,004 0,000 0,220 0,004 0,000 0,158 0,006 0,000 Numberofnmesnbers olderthan 65* 0,195 0,021 0,000 0,090 0,018 0,000 0,119 0,018 0,000 -0,068 0,020 0,001 Presence ofother members from 14 to 65 (dummy) -1,111 0,132 0,000 0,072 0,106 0,497 0,276 0,099 0,006 0,294 0,111 0,008 Mean education* -0,313 0,007 0,000 -0,315 0,007 0,000 -0,328 0,006 0,000 -0,163 0,007 0,000 Mean education2* 0,006 0,000 0,000 0,005 0,000 0,000 0,006 0,000 0,000 -0,003 0,000 0,000 Mean age' 0,043 0,006 0,000 0,003 0,005 0,522 0,001 0,004 0,865 -0,011 0,005 0,029 Mean age2* 0,000 0,000 0,000 0,000 0,000 0,316 0,000 0,000 0,251 0,000 0,000 0,247 Womenproportion* 0,109 0,005 0,000 0,095 0,003 0,000 - 0,092 0,003 0,000 0,091 0,003 0,000 Self-employed head (dummy) -0,584 0,020 0,000 -0,420 0,016 0,000 -0,351 0,015 0,000 -0,280 0,017 0,000 Labor income of the head (if employee) 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 Constant -4,867 0,088 0,000 -5,055 0,070 0,000 -5,034 0,065 0,000 -5,191 0,073 0,000 Occupied as self-employed versus unoccupied Education 0,196 0,013 0,000 0,052 0,010 0,000 0,010 0,009 0,267 0,085 0,011 0,000 Education2 -0,011 0,001 0,000 -0,002 0,001 0,001 0,001 0,001 0,136 -0,002 0,001 0,000 Age 0,369 0,007 0,000 0,356 0,005 0,000 0,362 0,004 0,000 0,347 0,005 0,000 Age2 -0,005 0,000 0,000 -0,004 0,000 0,000 -0,004 0,000 0,000 -0,004 0,000 0,000 Gender -1,815 0,042 0,000 -1,428 0,030 0,000 -1,463 0,027 0,000 -1,343 0,030 0,000 NumberofmenmbersfromO to 14* -0,043 0,009 0,000 -0,010 0,007 0,151 0,002 0,007 0,785 -0,028 0,011 0,010 Number ofmembers from 14 to 65* 0,053 0,012 0,000 0,029 0,008 0,001 0,037 0,008 0,000 0,021 0,011 0,064 Number ofmembers older than 65* 0,224 0,039 0,000 0,025 0,031 0,422 0,083 0,028 0,003 -0,034 0,031 0,287 Presence ofother members from 14to65 (dummy) 0,199 0,230 0,387 0,943 0,165 0,000 0,769 0,150 0,000 0,898 0,173 0,000 Mean education -0,262 0,017 0,000 -0,203 0,012 0,000 -0,215 0,011 0,000 -0,114 0,013 0,000 Mean education2* 0,008 0,001 0,000 0,004 0,001 0,000 0,005 0,001 0,000 0,001 0,001 0,312 Mean age* 0,007 0,011 0.522 -0,021 0,008 0,006 -0,010 0,007 0,144 -0,036 0,008 0,000 Mean age2* 0,000 0,000 0,484 0,000 0,000 0,046 0,000 0,000 0,926 0,000 0,000 0,000 Women proportion* 0,055 0,011 0,000 0,058 0,007 0,000 0,061 0,006 0,000 0,061 0,006 0,000 Self-employed head (dummy) 0,187 0,036 0,000 0,141 0,026 0,000 0,160 0,023 0,000 0,512 0,026 0,000 Labor income of the head (if employee) 0,000 0,000 0,000 0,000 0.000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 Constant -7,942 0,150 0,000 -7,682 0,113 0,000 -7,389 0,099 0,000 -7,905 0,120 0,000 Source: Based on "Pcsquisa Nacional por Amostra de Domicilios" (PNAD) ofthe 1976 and 1996. Note: * excluding the head. 49 Table A3.3: Dependent variable: Education* Year 1976 1981 1985 1996 Coefficient Standard P-value Coethicent Standard P-value Coefficient Standard P-value Coefficient Standard P-value Intercept -0,675 0,051 0,0001 3,392 0,031 0,0001 3,307 0,031 0,0001 3,239 0,037 0,0001 Age 0,310 0,003 0,0001 0,156 0,002 0,0001 0,185 0,002 0,0001 0,226 0,002 0,0001 Age 2 -0,004 0,000 0,0001 -0,002 0,000 0,0001 -0,003 0,000 0,0001 -0,003 0,000 0,0001 Gender -0,115 0,024 0,0001 -0,110 0,014 0,0001 -0,043 0,014 0,0024 0,195 0,017 0,0001 North region -0,826 0,070 0,0001 -0,732 0,040 0,0001 -0,679 0,036 0,0001 -1,092 0,038 0,0001 Northeast region -1,293 0,030 0,0001 -1,247 0,018 0,0001 -1,339 0,018 0,0001 -1,372 0,021 0,0001 West-center region -0,822 0,055 0,0001 -0,552 0,030 0,0001 -0,417 0,029 0,0001 -0,569 0,034 0,0001 South region 0,061 0,033 0,0684 -0,107 0,021 0,0001 -0,166 0,021 0,0001 -0,152 0,025 0,0001 R2 0,119 0,085 0,099 0,115 Source: Based on "Pesquisa Nacional por Amostra de DomiCilios" (PNAD) of the 1976 and 1996. Note: * People older than 10 years Table A3.4: Dependent variable: Total members of households younger than 14 years Year 1976 1981 1985 1996 Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Intercept 0,231 0,061 0,000 0,659 0,043 0,000 0,832 0,037 0,000 1,580 0,033 0,000 Schooling of the head -0,085 0,005 0,000 -0,098 0,004 0,000 -0,086 0,003 0,000 -0,041 0,003 0,000 Schooling of the head 2 0,001 0,000 0,000 0,003 0,O00 0,000 0,003 0,000 0,000 0,001 0,000 0,000 Age of the head 0,106 0,003 0,000 0,079 0,002 0,000 0,065 0,002 0,000 0,009 0,001 0,000 Age of the head 2 -0,001 0,000 0,000 -0,001 0,000 0,000 -0,001 0,000 0,000 0,000 0,000 0,000 North region 0,715 0,040 0,000 0,691 0,027 0,000 0,595 0,022 0,000 0,368 0,017 0,000 Northeast region 0,501 0,017 0,000 0,483 0,012 0,000 0,392 0,011 0,000 0,230 0,010 0,000 West-centerTegion 0,374 0,032 0,000 0,308 0,020 0,000 0,232 0,017 0,000 0,047 0,015 0,002 South region 0,064 0,019 0,001 0,015 0,014 0,270 -0,026 0,012 0,032 -0,004 0,011 0,677 R2 0,173 0,173 0,000 0,167 Source: Based on "Pesquisa Nacional por Anmslr de Dormicilios" (PNAD) of the 1976 and 1996. 50 Table A3.5: Dependent variable: Total members of households with age between 14 to 65 years Year 1976 1981 1985 1996 Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Intecept -3,024 0,055 0,000 -2,854 0,041 0,000 -2,630 0,036 0,000 -1,958 0,037 0,000 Schooling of the head 0,024 0,004 0,000 0,027 0,003 0,000 0,013 0,003 0,000 0,005 0,003 0,111 Schooling of the head 2 -0,003 0,000 0,000 -0,003 0,000 0,000 -0,002 0,000 0,000 -0,002 0,000 0,000 Age ofthe head 0,258 0,002 0,000 0,247 0,002 0,000 0,236 0,002 0,000 0,205 0,001 0,000 Age of the head 2 -0,003 0,000 0,000 -0,003 0,000 0,000 -0,002 0,000 0,000 -0,002 0,000 0,000 North region 0,202 0,036 0,000 0,223 0,026 0,000 0,221 0,021 0,000 0,196 0,019 0,000 Northeastregion 0,032 0,015 0,041 0,094 0,012 0,000 0,127 0,011 0,000 0,117 0,011 0,000 West-center region 0,091 0,028 0,001 0,083 0,019 0,000 0,107 0,016 0,000 0,033 0,017 0,051 Southregion -0,027 0,017 0,109 -0,020 0,013 0,122 -0,059 0,012 0,000 -0,106 0,012 0,000 R2 0,185 0,199 0,000 0,217 Source: Based on "Pesquisa Nacional por Anistra de Donicilios" (PNAD) of the 1976 and 1996. Table A3.6: Dependent variable: Total members of households older than 65 years Year 1976 1981 1985 1996 Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Coefficient Standard P-value Intrcept 1,034 0,013 0,000 0,942 0,010 0,000 0,958 0,009 0,000 0,848 0,010 0,000 Schooling ofthe head 0,006 0,001 0,000 0,005 0,001 0,000 0,004 0,001 0,000 0,005 0,001 0,000 Schooling ofthehead2 0,000 0,000 0,001 0,000 0,000 0,001 0,000 0,000 0,335 0,000 0,000 0,033 Age of the head -0,060 0,001 0,000 -0,056 0,000 0,000 -0,057 0,000 0,000 -0,053 0,000 0,000 Age of the head 2 0,001 0,000 0,000 0,001 0,000 0,000 0,001 0,000 0,000 0,001 0,000 0,000 North region 0,010 0,009 0,263 0,003 0,006 0,597 0,003 0,005 0,554 0,002 0,005 0,776 Northeast region 0,008 0,004 0,025 0,006 0,003 0,050 0,003 0,003 0,199 -0,002 0,003 0,513 West-center region -0,016 0,007 0,021 -0,006 0,004 0,190 -0,007 0,004 0,097 -0,014 0,005 0,003 South region -0,009 0,004 0,025 -0,003 0,003 0,362 -0,004 0,003 0,155 -0,004 0,003 0,245 R2 0,510 0,532 0,556 0,578 Source: Based on 'Pesquisa Nacional por Amowsha de Donidlios" (PNAD) of the 1976 and 1996. 51 Appendix 4: Figures 52 Anos Inflag5o anual Taxa de crescimento do PIB per capita 1976 0.415807391 0.075882257 41.58 7.59 1977 0.357078624 0.023947561 35.71 2.39 1978 0.382317228 0.024286355 38.23 2.43 1979 0.619987319 0.041748207 62.00 4.17 1980 0.795342571 0.074721737 79.53 7.47 1981 0.844575498 -0.063465143 84.46 -6.35 1982 0.855317408 -0.013398372 85.53 -1.34 1983 1.575321858 -0.049928718 157.53 -4.99 1984 1.744341494 0.031986942 174.43 3.20 1985 1.896145314 0.056405921 189.61 5.64 1986 0.571154652 0.053543234 57.12 5.35 1987 3.192561351 0.015535717 319.26 1.55 1988 7.027903132 -0.01884077 702.79 -1.88 1989 11.07192169 0.013611828 1107.19 1.36 1990 15.08922874 -0.063766297 1508.92 -6.38 1991 3.495892866 -0.00568845 349.59 -0.57 1992 8.99721581 -0.020662048 899.72 -2.07 1993 17.0639523 0.033612223 1706.40 3.36 1994 13.28010288 0.043252631 1328.01 4.33 1995 0.201152293 0.027621274 20.12 2.76 1996 0.094643222 0.013464913 9.46 1.35 53 Figure la: Macroeconomic instability in Brazil: Inflation 1,800 - -1706- 1,600 - 1509 1,400 _ _ _ _ 1,200 1 -- -- - 1,000 -- _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 800 703_ _ _ __ _ _ _ _ _ _ _ _ _ _ _ __ _ - _ _ _ __ _ _ _ _ 600 - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 400 - ___ __-__ 350 ____ 420- 36 38 6158 174 190 42 2 841 57 20 9 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Years Source: Fundacao Getulio Vargas (1999) and IBGE (1999). 54 Figure lb: Macroeconomic instability in Brazil: Per capita GDP 10 8 7 8_ |~~~~~~~~~~~~ 6 4 1=&||3 3- g~~~~ 2IiIIII L,1:11 i (2) _ -1-- u ~~~~~~~~~~~~~~~~-2 -2 A4 (4) __ _ _ (6) -5 _ -6 -6 (8) - - _ _ 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Years Source: IBGE (1999). 9661-.- 5861 +- 1861-' 9L61-m- uolliodold IuoilIndod OAT lWnlunD LLL IIIIIII I t I I L I I fIL-LLIIJI Iffi ILIiL-L ffi4JffiJw-LkfiLJJLLILILJ LI10mao - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - - - - I - - - - - - - - - - - - - - - - - - - - - - I 6'o0 0* 9661-9L61 S zul Or :u? amli,. O 9661 - S861- 1861 -- 9L61 uo!vodoid uoi Indod sAInltunD Ob 6£ 8E LE 9C S£ b£ £C ZE 1£ Of 6Z 8Z LZ 9Z SZ bZ £Z ZZ IZ OZ 61 81 LI 91 S1 N El ZI 11 01 6 8 L 9 S tb £ Z I --- I I I I I I I --- -I I IIII IIII _| -- -- -- -- - -- -- -- - - ------ - -- -- - -- -- -- -- -- - -- 1- - - - -- - - - - - - - -- -- -- - --- - -- - ---- - - ---- - - -- -- -- --- -- - -- -- -- -- -- -- -- -- - Z ° ° C _ O -I 01-0 OU'O 9661-9L61 's;aA.tna zuaiorl p;ajuaunijL :qz amn2ll 9s 57 Figure 3: Truncated Pen Parades, 1976-1996 200 6~~~~~~~~.17 0* 198 .418 .1 4 0 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - -- - - - - - - - - 162 -------------------------------------------------------------------------------------------.----- 1 4 0 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -,- - - - - - - - - - - - - - - - - - - - - - 12 ------------------- v----------------------------------------------------------------------------- -------- III97II6 1981 +18 +19T 58 Figure 4: Plotted Quadratic Returns to Education (Wage Earners) 25 - 20 15 z 5 e 1>~ 20 I I-- -- I ----- - T 1 -- - - - - - r-- -- - - - r 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Years of Schooling -i- 1976 1981 . 1985 -+- 19961 59 Figure 5: Plotted Quadratic Returns to Experience (Wage Earners) 6 54| o .1 _r I I F 1 ~F1Y fY~1 VIITIYT-1TTfTII IIII I I I ITT . I I I I I1 I II11 _TIII (o 5 e 9 b~o e ,a °@ 0 Experience -u 1976 - 1981 _.-1985 +.-1996 PS.4... SJAA _ _ _ _ _ _ _ __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 0 0 00 ° 09 61 Figure 7: Price Effects Separately and Jointly (both sectors) 0.3 0.2 $ 0.1- a _ -o 0.1 - - -0.2 Percentiles -Joint price effect -Economic Growth - Returns to Education -Returns to Experience -Retums to being female Source: "Pesquisa Nacional por Amostra de Domicilios" (PNAD), 1976 and 1996. 62 Figure 8: Price Effects: A Summary 0.3 0.2 a,) o 0.1, 0 -02 0.3 -0.2- _ ^ ~5 Percentiles r".'-alphas and betas - alphas beta Source: "Pesquisa Nacional por Amostra de Domicilios" (PNAD), 1976 and 1996. '9661 pu:e 9L61 '((IVNd) ,,So!l!olwoU op etIsouw .od IsuoTosN tsinbsoa,, :oojnos srlu S s iiq 'sildl_ siuuiS 1F seq pue svqde Sa3!RU3OJ3d Co C° ,p A e 4°V \ , 9 -& 9 9 c i> ' o 9'1- ----- ~ ----- ------ -------------------- ------- --------- --------------------- -------C----D------------ l _ .... - - - -- ... -- --*--'' ,6 P~~~~--- - --- ------ - ----------------------- ----------- ----- 8 0- r 0 I n - 90- J -0 - VO- 0~~~~~~~~~C - - - - ~~~~~~~~~~- - - -- - - - - - -- - - - - - -- - - - - - -- - - - - - -- - - - - - -- - - - - - -- - - - ---- - - - Z 0 0.0 C9 64 Figure 11: Demographic Effects 0.4 - - - 01 ---------- ----- ------ ------- ------ --------- -- 0.0 = -0.4 0.2 - ----- --- ---- -------- - - - n) O .6 4 0 - - - - - . - - - - - - - - - .~~~~~~~~~~~~~~~~~~~~~~------- - -------- ------ 0 -1.2 _ __ _ - __ - - - - - - - - - _ _ _ __ _ c > °o Q,b O ,0 + > V , , OK Percentiles -mu (d) ~mu(d), aiphas, betas, gammas Source: "Pesquisa Nacional por Amostra de Domicilios" (PNAD), 1976 and 1996. *9661 PUB 9L61 '(fVNd) ,,soITIOuoa op elusos od PeUO! iN esinbsod,, :ooinos L stwuxIeR 's13q 'Lj4dlP '(p)nwU_ SrnwiS 'se,aq 'svqdlt,* slaq put, sLqdlt. soplXuo3jad ___ ___ ___ ___ _____ _______ ~~~~~~~~~~~~~~~~~9'1- ---------- ------------ ----- - ---- - -- - - - - - - - - -- - - - - -- - - - -- - - - - - -- -- -- - 80- , ~ . .. Z I ..- _____- ------ - -- --- -- ------- ------- ---------- --- --- -----_ ------------- --- --------- --- ---- --- -------- q9~~~~~~~~~~~~~~~~~90 - - : : _ : ~~~~~~~~~~~~~~~~~~~~~~~~~~~L SbO- ; -, ,,, ; , ,,, , ,~~~~~~~~~~~~~~~. . ,, ;_;.. ,...... - ------ - Z O O O uoi sodtuloaat jui iu v :ZI aantij W .1 0~~s 66 Figure 13: Shift in the Distribution of Education, 1976-1996 100 -.__ _ ____ _ . -_ _ __ _ _ _ __ _ _ 80/ - 70 0 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 Years of Schooling too ~ ~ ~ ~ ~ t176_19 9661 PUt' 9L61 '(cVld) ,sotploiuoa op iu4sotu Vod peuoileN usinbsoj,, :oinoS stuunuu 'sulaq 'suqdlt, '(o)ntu '(p)nwtu_ (a)ntu put, (p)nw.u_. (a)nwa._ -- - -------- ----------- - -- - - - - - ---------- -- ---- - O - CD 0 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~00 90 |i=--= .---.oo:. W~~~~~~--w.= . .-..- .- . .0Z saajj[ znqduiuotuat put juammopua uolwgnpg :VI amn.3t L9 68 Figure 15: A Complete Decomposition -0.5 o ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . ct:~~~~~~~~~~~~~~~p~ --- ------ -2.0 Percentiles -nUalphas and betas - alphas, betas, gammas -*-mu(d), alphas, betas, gammas - mu(d), mu(e), alphas, betas, gammas ---1996-1976 Source: "Pesquisa Nacional por Amostra de Domicilios" (PNAD), 1976 and 1996. 69 Figure 15a: A Complete Decomposition 0.0 ____ _-_ _ C)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~T 0 on 00 '.4-4 0 -2.0 % 43 \ b 0 ,° + ¢ oR{pc. c O @ @ <1 4 @ 4° e 4> 9 Percentiles Sp-alphas and beta 9961976 Source: "Pesquisa Nacional por Amostra de Domicilios" (PNAD), 1976 and 1996. 70 Figure 15b: A Complete Decomposition 0. -0.5 00 0 d.) -2.0 - 2.0 _ _ _ _ _ _ _ _, 1 _ _ _O _ _ _ _ _ _ _ _ _ _ _ _ _ _ Percentiles Hmaiphas and: betas alphas, betas, gammas 996- 6 Source: "Pesquisa Nacional por Amostra de Domicilios" (PNAD), 1976 and 1996. 71 Figure 15c: A Complete Decomposition 0.5 __ _ ___ ___ __ __ _ _ _ _ _ _ _ __ _ _ _ 0.0 -_ _ _ _ _ _ ._ _ _ _ _ _ _ _ _ _ 0 , ._.-w .rf f _ 0- -0.5 - 0.01 OU~ ~ ~~~~~-rl" ...i . .. . . . . .. . . .. . ...... .... .... .... I S5 - ---- --------- ------------------------------------------------------- --------------------------- ------------------- 0 4) 1-4 -2.0 - Percentiles ininalphas and betas-~alphas, betas, gammas -4mu(d), alphas, betas, gammas 1996-1976 Source: "Pesquisa Nacional por Amostra de Domicilios" (PNAD), 1976 and 1996. Policy Research Working Paper Series Contact Title Author Date for paper WPS2187 Who Determines Mexican Trade Jean-Marie Grether September 1999 L. Tabada Policy? Jaime de Melo 36896 WPS2188 Financial Liberalization and the Pedro Alba September 1999 R. Vo Capital Account: Thailand, 1988-97 Leonardo Hernandez 33722 Daniela Klingebiel WPS2189 Alternative Frameworks for Stijn Claessens September 1999 R. Vo Providing Financial Services Daniela Klingebiel 33722 WPS2190 The Credit Channel at Work: Lessons Giovanni Ferri September 1999 K. Labrie from the Republic of Korea's Financial Tae Soo Kang 31001 Crisis WPS2191 Can No Antitrust Policy Be Better Aaditya Mattoo September 1999 L. Tabada Than Some Antitrust Policy? 36896 WPS2192 Districts, Spillovers, and Government Reza Baqir September 1999 S. Devadas Overspending 87891 WPS2193 Children's Growth and Poverty in Michele Gragnolati September 1999 M. Gragnolati Rural Guatemala 85287 WPS2194 Does Democracy Facilitate the Jean-Jacques Dethier October 1999 H. Ghanem Economic Transition? An Empirical Hafez Ghanem 85557 Study of Central and Eastern Europe Edda Zoli and the Former Soviet Union WPS2195 Aggregating Governance Indicators Daniel Kaufmann October 1999 D. Bouvet Aart Kraay 35818 Pablo Zoido-Lobat6n WPS2196 Governance Matters Daniel Kaufmann October 1999 D. Bouvet Aart Kraay 35818 Pablo Zoido-Lobat6n WPS2197 Production Sharing in East Asia: Francis Ng October 1999 L. Tabada Who Does What for Whom, and Why? Alexander Yeats 36896 WPS2198 How the Chinese System of Charges Hua Wang October 1999 H. Wang and Subsidies Affects Pollution Ming Chen 33255 Control Efforts by China's Top Industrial Polluters WPS2199 Managing Risks of Capital Mobility Mansoor Dailami October 1999 W. Nedrow 31585 WPS2200 The Role of Trust in Financial Sector Biagio Bossone October 1999 E. Mekhova Development 85984 Policy Research Working Paper Series Contact Title Author Date for paper WPS2201 Financial Development and Industrial Biagio Bossone October 1999 E. Mekhova Capital Accumulation 85984 WPS2202 Specialization without Regret: Michael R. Carter October 1999 M. Fernandez Transfer Rights, Agricultural Yang Yao 33766 Productivity, and Investment in an Industrializing Economy WPS2203 Market versus Administrative Michael R. Carter October 1999 M. Fernandez Reallocation of Agricultural Land Yang Yao 33766 in a Period of Rapid Industrialization WPS2204 Corruption under Moral Hazard Gunnar S. Eskeland October 1999 H. Sladovich Henrik Thiele 37698 WPS2205 Foreign-Owned Capital and Marcelo Olarreaga October 1999 L. Tabada Endogenous Tariffs 36896 WPS2206 Household Childcare Choices and Michael M. Lokshin October 1999 P. Sader Women's Work Behavior in Russia 33902 WPS2207 Jamaica's Food Stamp Program: Kene Ezemenari October 1999 G. Peralta Impacts on Poverty and Welfare Kalanidhi Subbarao 37405 WPS2208 Ethnic Partition as a Solution to Nicholas Sambanis October 1999 H. Sladovich Ethnic War: An Empirical Critique 37698 of the Theoretical Literature WPS2209 Does Corruption Relieve Foreign Shang-Jin Wei October 1999 H. Sladovich Investors of the Burden of Taxes 37698 and Capital Controls?