FILE COPY SWP601. Perspectives on Poverty and Income Inequality in Brazil. An Analysis of the Changes during the 1970s

Transcription

1 Public Disclosure Authorized FILE COPY Perspectives on Poverty and Income Inequality in Brazil An Analysis of the Changes during the 1970s SWP601 Public Disclosure Authorized David Denslow, Jr. William G. Tyler WORLD BANK STAFF WORKING PAPERS Number 601 Public Disclosure Authorized Public Disclosure Authorized

2

3 WORLD BANK STAFF WORKING PAPERS #r0, o Number 601 Perspectives on Poverty and Income Inequality in Brazil An Analysis of the Changes during the 1970s David Denslow, Jr. William G. Tyler INTERNATIONAL MONETARY FUND JOINT LIBRMY MAR I14TERNATIONAL BANK FOR IIECOISTiUCTION AND DEVELOPMENT WASHINGTON. D.C The World Bank Washington, D.C., U.S.A.

4 Copyright 1983 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, N.W. Washington, D.C , U.S.A. First printing July 1983 All rights reserved Manufactured in the United States of America This is a working document published informally by the World Bank. To present the results of research with the least possible delay, the typescript has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. The publication is supplied at a token charge to defray part of the cost of manufacture and distribution. The views and interpretations in this document are those of the author(s) and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting on their behalf. Any maps used have been prepared solely for the convenience of the readers; the denominations used and the boundaries shown do not imply, on the part of the World Bank and its affiliates, any judgment on the legal status of any territory or any endorsement or acceptance of such boundaries. The full range of World Bank publications is described in the Catalog of World Bank Publications; the continuing research program of the Bank is outlined in World Bank Research Program: Abstracts of Current Studies. Both booklets are updated annually; the most recent edition of each is available without charge from the Publications Distribution Unit of the Bank in Washington or from the European Office of the Bank, 66, avenue d'lena, Paris, France. David Denslow, Jr. is professor of economics at the University of Florida and a consultant to the World Bank; William G. Tyler is a senior economist in the Country Programs Department of the Bank's Latin America and Caribbean Regional Office. Library of Congress Cataloging in Publication Data Denslow, David, 1942-,, Perspectives on poverty and income inequality in Brazil. (World Bank staff working papers ; no. 601) Bibliography: p. 1. Income distribution--brazil. 2. Poor--Brazil. I. Tyler, William G. II. Title. III. Series. HCl90.I5D '6' ISBN

5 PERSPECTIVES ON POVERTY AND INCOME INEQUALITY IN BRAZIL: AN ANALYSIS OF THE CHANGES DURING THE 1970s ABSTRACT This paper provides some exploratory analysis of the advance tabulations of the 1980 Brazilian demographic census. A descriptive examination of various measures of poverty is undertaken with comparisons made between 1970 and Computations are also made of Gini coefficients, Theil indices, and income deciles. Changes in the overall inequality in income distribution are not clearly discernible between 1970 and 1980, although more disaggregated changes, in part offsetting each other, were identified. In general, real absolute incomes increased for all decile groups, suggesting a reduction in poverty over the decade of the 1970's. PERSPECTIVAS SOBRE POBREZA E DESIGUALDADE DE RENDA NO BRASIL: UMA ANALISE DAS MUDAN,AS OCORRIDAS DURANTE A DECADA DOS ANOS SETENTA RESUMO Este trabalho oferece uma analise exploratoria das tabulag8es avangadas do censo demografico, recentemente publicadas pelo IBGE. Um exame descritivo das diversas medidas de pobreza e feito com comparagoes entre 1970 e Estimativas tambem sao feitas de coeficientes de Gini, indices de Theil, e dos deciles de renda. Mudangas na desigualdade global da distribuigao da renda nao sao claramente identificadas nos dados entre 1970 e 1980, mesmo que mudancas mais desagregadas, em parte compensat 6 rias, fossem evidentes. Em geral, rendas reais aumentaram para todos os grupos dos deciles, sugerindo uma redugao em pobreza durante a decada de 70.

6 ACKNOWLEDGMENTS The authors wish to express gratitude to the World Bank for supporting their research, but must emphasize that the views expressed are entirely their own. The authors are also grateful to Hendrik van der Heijden, Rodolfo Hoffmann, Dennis Mahar, Guy Pfeffermann, Rubens Vaz da Costa, and Paulo Vieira da Cunha for useful comments on an earlier version, which was presented at the Fourth Meeting of the Brazilian Econometric Society in Aguas de Sao Pedro, Sao Paulo on December 7, Thanks also go to Jay Prag and Joan Ahern for programming and research assistance. The usual caveats apply.

7 TABLE OF CONTENTS Page No. I. Introduction 1 II. Non-Income Measures of Poverty 4 III. IV. Income Measures of Poverty and Income Concentration 10 Interpretations of Changes in Income Inequality 18 V. Concluding Remarks 29 REFERENCES 31 APPENDIX 33

8

9 I. INTRODUCTION In the past 20 years the Brazilian economy has experienced substantial growth. Income per capita grew at annual average rates of 5.0% and 4.9% during the periods and , respectively, reaching a level of about US$2,000 by The nature of this growth and the growth process itself have been controversial, with many critics contending that poverty and human misery have increased. At the other extreme, other individuals extol the virtues of the so-called Brazilian model of growth, pointing to the human benefits accompanying the growth process. This paper attempts to shed some limited new light on an old debate with an analysis of recently available data from the 1980 demographic census, undertaken by the Instituto Brasileiro de Geografia e Estatistica (IBGE).

10 - 2 - Over the past ten years much of the debate has centered on questions of inequality in the distribution of income. With the availability of the 1970 demographic census materials comparisons could be made with 1960, focusing on changes in relative poverty and the distribution of income over time. Virtually all of these studies 1 / demonstrated that relative income shares became more unequal between 1960 and The income benefits of economic growth in Brazil were seen to be unequally distributed, with upper income groups gaining disproportionately and the poor being left behind. While the real absolute incomes of the poorest 40% of income earners were seen to grow modestly, if at all, the upper income groups enjoyed substantial income increases. Considerable controversy has existed concerning the reasons for the observed increased income concentration. It can be noted, however, that the overall results are consistent with an interpretation of Brazilian growth following dualistic lines as envisaged in Lewis-Ranis-Fei type growth models. The question since the analysis of the 1970 data has been whether the patterns of income concentration and growth observed during the 1960's persisted during the high growth 1970's 2 /. In looking towards the 1980's a better perspective on trends and events during the 1970's appears useful. 1/ See, among others, Fishlow (1972), Langoni (1973), Hoffman and Duarte (1972), Fonseca (1978), and Fields (1977). Other important works, focusing on the 1970 data, include Lluch (1981, 1982), Costa (1977) and Fox (1982). 2/ Some analysis making use of the ENDEF (the national household budget survey) materials for has concluded that some fundamental changes in fact have occurred. See Pfefferman and Webb (1979,1982), Thomas (1982) and Knight (1981). These studies in general attest to rather substantial progress in poverty alleviation. Other work (e.g., Rossi, 1982) has employed income tax return information, showing increases in inequality measures over time presumably as the result of the inclusion of lower relative income recipients in the income tax system.

11 - 3 - The recent publication by the Brazilian Geographic and Statistical Institute (IBGE) of the advance tabulations from the 1980 demographic census permits a fresh look at changing patterns of poverty and income inequality in the country with a systematic focus for the first time on the 1970's 3 /. The principal question posed in this paper is what has happened to poverty and income concentration during the decade. Has growth brought poverty alleviation? Or has it simply continued to concentrate incomes? And what inferences, if any, can be made about changes in social welfare? The analysis undertaken here is viewed as preliminary in nature, with a large number of questions raised and left unanswered. The public use sample of the 1980 demographic census, scheduled to be available soon from IBGE, will enable other economists to explore more fully and thoroughly some of the questions addressed here. Nevertheless, on the basis of information so far made available to the public, the analysis possible, explained in this paper, produces some rather far reaching results and conclusions. Despite persistent poverty, during the 1970's substantial progress in reducing poverty and improving living standards occurred. While there is some limited and weak evidence of overall increased income concentration, growing overall relative inequality, if it did occur, was minor and could not be fully discerned by our measures. In addition, there were observed reductions in income inequality among regions and among sectors. The agricultural sector in particular, while witnessing growing income inequality within it, was characterized by rapidly growing average incomes which served to reduce the gap between agricultural and urban-based occupational incomes. 3/ The basic data source for the information analyzed in this paper is IBGE (Instituto Brasileiro de Geografia e Estatfstica), Tabulagcoes Avancadas do Censo Demografico, Brazil , Vol. 1, Tomo 2, (Rio de Janeiro: IBGE, 1981).

12 - 4 - In Section II we present descriptive evidence on non-income measures of poverty. Section III discusses income measures of poverty and income inequality and contains Gini and Theil indices estimated for 1970 and Absolute and relative comparisons of decile group incomes are also presented. Section IV provides interpretations of the results, and a final section offers concluding remarks. II. NON-INCOME MEASURES OF POVERTY The 1980 demographic census continues and extends a number of non-income measures of poverty and well-being instituted in earlier censuses. These socio-economic indicators include computable measures of literacy, schooling, enrollments, household size and composition, labor force participation, access to social services, dwelling characteristics, housing conditions, and durable consumer goods possession. Also, from the available information, estimates can be generated regarding life expectancy and infant mortality. Unequivocally, while one can be dissatisfied with existing levels, all of these measures show substantial improvements during the 1970's with real gains across the country, some reductions in the considerable regional disparities and, generally lessened urban-rural differences as well. Some of this information is described in the next few pages; other data are presented in the Appendix Tables. 4 / 4/ Additional statistical materials, on which this paper is based, can be obtained by writing either author at his institutional address.

13 Looking first at literacy, Table 1 demonstrates an increase in overall literacy rates from 59.4% in 1970 to 68.7% in 1980, despite a slight absolute increase in the number of illiterates. Each major region showed improvement, as did both rural and urban population groupings. 5 / Regional disparities were also reduced, in 1970 Northeastern literacy was 66% of the national average; by 1980 the comparable figure had increased to 69%. Some information on school enrollments is presented in Table 2. Enrollment rates at both primary and secondary school levels increased significantly. Again, the Northeast, while lagging considerably, is seen to have reduced the gap between its enrollment rates and those of the country as a whole. Comparing average years of schooling of the economically active population in 1970 and 1980 yields similar results. In addition, to be referred to later, the inequality of schooling across individuals has decreased markedly. (Appendix Table A6).. Table 3 presents basic information on the provision of water supply and sewerage services. Both relate in a fundamental way to health conditions. 6 / The pattern over the decade of the 1970's has been one of 5/ Similar results are apparent with "functional" literacy rate comparisons, where the "functional" literacy rate is defined as the proportion of the population aged 15 years and older reporting five or more years of formal schooling. One can note that literacy is rather loosely defined and schooling levels and enrollments not entirely reflective of actual educational attainment. Yet, unless these indicators themselves have deteriorated in their quality, improvement in the phenomena measured by the indicators over time represents a real improvement. 6/ One recent study (Merrick, 1982) has found a strong negative relationship between the provision of piped water and infant mortality across households.

14 - 6 - Table 1 AVERAGE LITERACY RATES BY REGION AND URBAN OR RURAL LOCATION, 1970 and 1980 a/ (Percentage) Region Urban Rural Total Northeast Southeast Frontier b/ BRAZIL Notes: a/ The mean is unweighted. Data based on population 5 years of age or older and total includes those persons not reporting literacy. b/ Excludes the Federal District, where the rate was 76 % and 40% in 1970 and 84 % and 68 % in 1980 for urban and rural areas respectively. Sources: Peter T. Knight et al., Brazil: Human Resources Special Report (Washington: The World Bank, October 1979) and IBGE, Censo Demogra?fico, Brasil , Volume 1 and Tabulacoes Avancadas do Censo Demografico, Brasil , Volume 1 - Tomo 2 (Rio de Janeiro: IBGE, 1973 and 1981).

15 - 7- Table 2 SCHOOL ENROLLMENT RATES, 1970 AND 1980 Grades Grades 1-8 1/ 9-122/ (%) (%) Region Northeast Southeast Frontier BRAZIL Notes: 1: Enrollment rate is defined as enrollment of population five years or older in grades 1-8 expressed as a percentage of population aged 7-14 years. 2: Enrollment rate is defined as enrollment of population five years or older in grades 9-12 expressed as percentage of population aged years. Source: IBGE, Censo Demografico, Brasil-1970, Vol.1 (Rio de Janeiro: IBGE, 1973) and IBGE, Tabulacoes Avangadas do Censo Demografico, Brasil , Vol. 1, Tomo 2, (Rio de Janeiro: IBGE, 1981).

16 - 8- Table 3 PIPED WATER SUPPLY AND SEWERAGE SERVICES, BY REGION, Sewerage Network Piped Water Supply Services or Septic Tank Annual Annual Percentage of Average Percentage of Average Households Rate of Households Rate of Supplied (%) Increase 1/ Possessing (%) Increase 1/ Region (%) (%) Northeast Urban Rural Southeast Urban Rural Frontier Urban Rural BRAZIL Urban Rural Note: 1/ Annual compounded rates for the number of households covered. Source: The 1970 data are taken from Peter T. Knight et al., Brazil: Human Resources Special Report (Washington: The World Bank, 1979). Imformation for 1980 was computed from IBGE, Tabulagoes Avangadas do Censo Demografico, Brasil 1980, Vol. 1, Tomo 2 (Rio de Janeiro: IBGE, 1981).

17 -9- significant expansion in the coverage of services to households. While the Northeast lags behind the rest of the country considerably, some modest catching up during the 1970's did occur. Material progress and reduced poverty during the course of the 1970s are also reflected in the expansion of the possession by households of durable consumer goods and services. 7 / Some comparisons between are presented in Appendix Table Al. Of particular interest is the expansion in the households provided with electricity. A total of 48% of Brazilian households in 1970 were provided with electricity; by 1980, the proportion had increased to 67%. While rural areas are seen to still lag considerably, substantial progress was made. for other durable consumer goods and services. Similar patterns are observed Dramatic increases in household possession are observed throughout the country, with some of the greatest proportional increases being witnessed in the Northeast. For Brazil as a whole by % of all households possessed a refrigerator, 76% a radio, 55% a television, and 22% an automobile. These figures, and their growth, attest to the expansion of European or American style middle class living standards in Brazil and/or the proletarization of durable consumer goods consumption. 7/ It can also be noted that the average size of household decreased from 5.1 persons in 1970 to 4.5 persons in Also, the number of persons per room fell from 2.5 to 2.2 between 1970 and The observed decline in average household size may be overstated because of a possible sampling bias towards smaller households in the administration of the long form census questionnaire, on which the advance tabulations are based. Although this possibility is minimized by IBGE, the effects of any such bias, if indeed existent, are difficult to assess in their impact on the income distribution measures reported in this paper. Subsequent to the release of the advance tabulations, IBGE has been engaged in analysis envisaging a reweighting prior to the release of the final census results. The effect of the bias on individual income estimates, and as a result the distribution of income by income

18 On the basis of the observed increase in literacy, the expansion and increased distributional equality of schooling, improved sanitation facilities, and increased household possession of consumer goods and services, one is tempted to draw some tentative conclusions about changes in living standards and welfare during the decade of the 1970's. 8 / The non-income evidence discussed so far and available from the census materials suggests some substantial improvements. At the minimum, it would appear that there is little factual support for the contention by some of deteriorating living conditions and growing poverty. Further evidence, suggesting improvement, is observed when incomes are introduced into the analysis. It is to such analysis that we now turn our attention. III. INCOME MEASURES OF POVERTY AND INCOME CONCENTRATION In presenting income measures of poverty and income concentration, our analysis is necessarily based upon individual incomes. Family incomes cannot be constructed from the published advance census tabulations. 9 / Although the family figures are superior for many purposes, we think that the information on individuals is also useful, both 8/ Some ultimate reflections of social welfare conditions are changes in infant mortality and life expectancy. Measures of these variables can be generated from demographic census materials, although this, to our knowledge, has only been done in preliminary form from the 1980 Brazilian demographic census information. Comparing these preliminary estimates for 1980 made by IBGE and reported in the press (Jornal do Brasil, November 14, 1982) with earlier estimates, improvement is seen to have continued. Infant mortality rates for Brazil as a whole, expressed per 1000 live births, fell from 123 in 1960, to 107 in 1970, and to 93 in Life expectancy at birth, again for the country as a whole, is estimated to have increased from 53.5 years in 1970 to 58.7 years in / Analyses of family income distributions for 1970 include Lluch (1981) and Fox (1981).

19 for insights regarding the labor market and, since few in the top 40% of the individual distribution are likely to be in the bottom 40% of the adult-equivalent per capita family distribution in a country characterized by great inequality, for conclusions about welfare. In constructing inequality estimates we have employed Gini coefficients, Theil indices, and deciles to summarize income distributions. The Gini and Theil indices buttress each other as single-valued measures because of their differing sensitivities. The Gini index is especially sensitive to income transfers near the mean. The Theil index, even relative to its generally greater overall variability, is more sensitive to transfers within the lower and higher ranges. In view of the uncertainties introduced by our approximation techniques, especially at the extremes of the spectrum, we feel more confident about our conclusions when the Gini and Theil measures change in the same direction. The Theil index is a superior measure for complete data and has useful decomposition properties. The Gini coefficient, besides having well-analyzed properties, complements the Theil index when data are grouped. The deciles provide more detailed information. The Gini coefficients are estimated from eight income intervals using Gastwirth's method I (1972) to establish bounds. indices are established using Theil's method (1967). Bounds on the Theil Single values are derived from these bounds with the simple interpolations advocated by Cowell and Mehta (1982). The eight income intervals are converted into deciles with the polynomial approximation to the Lorenz curve described by

20 Kakwani (1980, pp ). Especially for low-income groups, figures for the lower two deciles are suggestive only. The advance tabulations of the demographic census are constructed from a sample of the population. For most groups this sample is quite adequate, but in some cases, sectors within small states, for example, sampling error could have a noticeable effect on our measures. In such cases, we do not have true upper and lower bounds but merely consistent estimates of them. Nevertheless, we think this problem is minor for all regional and national aggregations. To make comparisons in income between 1970 and 1980 adjustments for inflation and necessary. Ideally, a cost-of-living index would be the most suitable deflator, despite a tendency to overstate inflation, in theory at least, because of the failure to adequately account for substitution in the market basket in response to relative price changes. In practice, however, owing to alleged irregularities in constructing the Rio de Janeiro cost-of-living index, particularly in 1973, its use appears to understate inflation and thereby to overstate real income growth. Since the general internal price index is a weighted average of different indices, including the Rio cost-of-living index, its use also reflects the apparent tampering that occurred in the mid-1970s. In view of these difficulties with the cost-of-living and general price indices, it was decided to use information from the national income accounts. Instead of simply employing the GDP implicit deflator, as Langoni did in his comparisons, we elected to construct an implicit deflator for consumption expenditures from the published FGV estimates of constant and current price

21 comsumption expenditures in GDP over time. 10 / This procedure has given us the most conservative, i.e, lower bound, estimates of real income increases. Most of our estimates are based on data for the economically active population (EAP) with positive incomes. In 1980, of the total EAP of 43,796,763, there were 3,294,659, or 7.5%, who had no cash income and 146,744, or 0.3%, who refused to report their incomes to the census takers. purposes. This latter group is small and may be ignored without harm for our More significantly, we have also chosen to exclude those with no cash income. This is partly to facilitate comparison with the 1970 decile group estimates of Langoni (1973) and partly because including zero-income individuals makes the inequality measures derived from data for individual recipients less useful as indicators of changes in the distribution of family income. This is suggested by the fact that in 1970 the zero-income members of EAP came almost as often as not from middle-income families as shown by Lluch (1981, p. 24). The determination of changes in the degree of income inequality between 1970 and 1980 in Brazil is complicated by the existence in 1970 of an upper coding limit on income. All incomes of Cr$ 9998 or more per mnth were coded as Cr$ 9998, a figure which after adjustment for inflation by the implicit consumption deflator corresponds to Cr$285,000 of 1980, or 10/ Fundacao Getu'lio Vargas, Conjuntura Econo'mica, Vol. 35, No. 12 (December 1981). For every year in the period, the change in the implicit consumption deflator exceeded the Rio cost-of-living index increase, with the greatest part of the overall difference being attributed to For that year the implicit consumption deflator was computed to rise 22.3% as compared to the 12.6% increase in the Rio cost-of-living index.

22 approximately US$5,215. It is about 30 times the mean income for This truncation of the upper tail of the income distribution reduces both the Gini coefficient and the Theil index. Adjustments to the 1970 estimates of Langoni (1973) to account for the limitations imposed by the 1970 coding limit raises the Gini and Theil measures considerably. 11/ Accordingly, the growth in inequality observed during the 1960s may have been understated. Including such adjustments for the 1970 coding limit problem, Table 4 presents the results of the basic overall 1970 and 1980 comparisons for income inequality. In the upper half of this table, comparison of the first and third columns gives the impression of a moderate increase in inequality during the 1970s. But, as indicated, Langoni's calculations for 1970 were hampered by the upper coding limit on income. The second column shows plausible values for the true inequality measures for that year. With the adjustments the picture becomes less clear. We emphasize that we are not advancing the figures in the second column as accurate. Our point is simply that any change in overall inequality over the decade was not large enough to be fully and unequivocally discerned by our measures. The lower half of Table 4 provides additional support for this conclusion of no significant deterioration of the income distribution of the total EAP, as distinct from the non-zero income EAP. With adjustment for the 1970 coding limit, the argument of no discernible deterioration becomes still stronger. The evidence does appear sufficient to reject the hypothesis of increased inequality. I1/ See the Appendix and Denslow (1982).

23 Table 4 BASIC OVERALL 1970 AND 1980 INCOME DISTRIBUTION COMPARISONS Langoni Adjusted Estimate Estimates Excluding Zero-Income Workers Gini Coefficient Theil Index Share of Lowest 40% 10.03% 9.74% 9.72% Share of Top Decile 46.47% 48.03% 47.89% Estimates Including Zero-Income Workers Gini Coefficient Theil Index Share of Lowest 40% 7.42% 7.17% 8.94% Share of Top Decile 51.66% 55.08% 49.34% Source: Using the formulae derived in the appendix, the Gini coefficient and the Theil indices are calculated from those derived with zero-income workers excluded. For 1970, the share of the lowest 40% is estimated as the midpoint of bounds set on the relevant point of the Lorenz curve, using data from Langoni (1982, p.21) and from Tabulacoes Avancadas do Censo Demografico 1970, Table 8. The bounds on the Langoni share are (7.32%, 7.52%) and on the share adjusted to allow for the upper coding limit (7.07%, 7.26%). The share of the top decile is estimated with a quadratic approximation technique using the same sources of data. The bounds on the Langoni share are (50.90%, 51.71%) and on the adjusted share (54.32%, 55.13%). Since the Tabulac6es Avangadas of the 1980 census provide both means and limits for income groups, we use the third-degree polynomial approximation suggested by Kakwani (1982, p. 103) to estimate the share of the lowest 40% and top decile in 1980.

24 The overall stability in distributional inequality can also be seen through an examination of the income decile groups. Table 5 presents estimates of the proportion of total income going to different decile groups for the country's total economically active population for 1970 and Little change can be discerned. Table 5 also presents estimates of the mean income for each decile group. 12 / The average income for all groups, as observed, increased by an estimated 49%.13/ Comparing 1970 and 1980, all decile groups participated in this increase. This feature of the 1970's contrasts markedly with the experience of the 1960's, for which disproportionate gains for the upper income groups were quite evident. One should note that, because of coding limits and the rather arbitrary assumptions necessary for estimating the tails of the distribution, the gains of the first and tenth deciles are not entirely reliable. Yet, it appears that the gains from rapid economic growth during the 1970's were widely distributed. Substantial real income gains were observed by all decile groups. This pattern is also evident when decile groups for 12/ The average exchange rate in August 1980 was Cr$54.65 per US dollar. Using this figure, which of course has a very loose relationship to the ratio of the cost-of-living in Brazil and the United States, suggests an average annual income of $2,622 per member of the economically active population, $639 for the bottom 40%, $12,600 for the top decile, $1,556 for the Northeast, and $1,462 for agriculture. For the bottom 40% of those with positive incomes in agriculture in the Northeast, the average annual income is $348, approximately a dollar a day on average for these two million workers. In all Brazil about four million workers earned cash incomes below a dollar a day. 13/ As noted above, the observed average increase of 49% represents a lower bound estimate. Using alternative indices for deflation yields average increase of 100%, 76% and 55% for the Rio de Janeiro cost-of-living index, the general price index, and the GDP implicit deflator, respectively. The magnitude of the average gains does not affect the conclusion that the gains from growth during 1970s did not appreciably change the income distribution. In any event, the average real income increases were substantial and were experienced by all decile groups. It should be noted, however, if the sampling error referred to in Footnote 7 resulted in a systematic exclusion of large families in outlying rural areas, it is likely that the gains in the bottom decile are overstated.

25 Table 5 RELATIVE AND ABSOLUTE INCOME MESASURES FOR INCOME DECILE GROUPS IN BRAZIL, ECONOMICALLY ACTIVE POPULATION, 1970 AND 1980 Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change , ,650 2, ,415 3, ,064 4, ,037 5, ,959 6, ,798 8, ,003 11, ,178 18, ,366 57, AVERAGE 8,040 11, Note: The figures for 1970 are based on Table 8 of the Tabula46es Avangadas do Censo Demogrdfico, The figures for 1980 are from Tables 5.2 and 5.3 of the Tabulagoees Avangadas do Censo Demografico, The eight income intervals reported are converted into deciles with the polynomial approximation described by Kakwani (1980, pp ). For both years the data describe members of the economically active population with positive incomes and reporting those incomes to census takers.

26 regions and urban and rural occupations are examined (Appendix Tables A7, A8, and A9). Consequently, despite continuing marked income inequality, the evidence suggests that the decade of the 1970's witnessed a reduction in poverty, seemingly on a considerable scale. IV. INTERPRETATIONS OF CHANGES IN INCOME INEQUALITY As is often the case, the evident aggregate stability in the income distribution covers sectoral turbulence. In Brazil during the 1970s two major developments offset each other. First, the difference between average incomes among sectors narrowed, thereby reducing inequality. On the other hand, inequality within the agricultural sector soared, raising inequality. These two developments, along with changes in regional disparities, can be seen in Table 6. The first panel of Table 6 suggests that regional disparities diminished in the 1970s, with average income per member of the economically active population rising more rapidly in the low-income Northeast than in the higher-income Southeast. Average income in the Northeast went from 46% to 51% of that in the Southeast. However, both rural and urban incomes rose less in the Northeast than elsewhere. The Northeast gained only because, with its labor force more heavily concentrated in agriculture, it benefited more than other regions from the exodus from agriculture and the more rapid growth of incomes in the primary sector. This greater rapidity of growth was due to a jump in incomes in the top two rural deciles coupled with above average rates of increase in the other eight. The proportion of the EAP in the Northeast in the bottom 40% of the income distribution of Brazil stayed at 67% from 1970 to 1980.

27 Table 6 MEASURES OF INEQUALITY FOR THREE REGIONS OF BRAZIL, 1970 AND 1980, EAP WITH POSITIVE INCOMES TOTAL Mean Income 1/ Gini Theil Region Change Southeast 9,746 13, Northeast 4,486 7, Frontier 6,678 10, BRAZIL 8,040 11, RURAL Mean Income Gini Theil Region Change Southeast 4,907 8, Northeast 2,681 4, Frontier 4,569 8, BRAZIL 3,965 6, URBAN Mean Income Gini Theil Region Change Southeast 11,976 16, Northeast 7,103 9, Frontier 9,276 13, BRAZIL 10,778 13, Notes: 1/ Mean income expressed is in cruzeiros of August 1980 per month, with adjustments made for the 1970 information using an implicit consumption expenditures deflator, as explained in text. 2/ Estimates for regions for 1970 are constructed as described in Denslow (1982). The Southeast consists of Minas Gerais, Esp rito Santo, Rio de Janeiro, Sgo Paulo, Parana', Santa Catarina, and Rio Grande do Sul. The Northeast includes Maranhao, Piauf, Cearl, Rio Grande do Norte, Para ba, Pernambuco, Alagoas, Sergipe, and Bahia. The Frontier region incorporates the remaining states and territories.

28 One way to examine sectoral changes is to partition the Theil index across sectors. We can write: 3 3 (1) T 2 -T 1 = (I2-Il)+ Yij(Ti2-Tii)+ Ti2(Yi2-Yi1) i=1 i=1 where Ti is the Theil index for year j, Ij is the among sectors inequality index for year j, Yij is the share of sector i in total income in year j, and Tij is the within-sector Theil index for sector i in year j. The first term on the right-hand side of Equation (1) may be called the among-sectors effect, the second term the within-sectors effect, and the third term the sectoral shift effect. Applying Equation (1) to the EAP in 1970 and 1980 in the primary, secondary, and tertiary sectors yields: Effect Among sectors Within sectors.1090 Primary.0665 Secondary.0074 Tertiary.0351 Sectoral Shift Total.0430 Interpreted as written, the results above show a decrease in the among sectors effect attributable to the rise in relative incomes in agriculture, more than offset by an increase in inequality within agriculture. A small but heavily weighted increase in the Theil index in the tertiary sector adds to the rise in the overall index. While these results make no allowance for the 1970 upper coding limit, similar

29 exercises making adjustments revealed no major differences. Two conclusions seem reasonably secure. First, inequality among sectors fell, and, second, inequality within agriculture rose. The cause of the reduction of inequality among sectors is readily identified: income per economically active person in agriculture climbed from 37% of the average for the secondary and tertiary sectors in 1970 to 50% in 1980, according to the census data. Other sources confirm this remarkable increase. For example, while the proportion of the EAP engaged in agriculture declined from 40% to 27%, the national income and product accounts show the agricultural share in the GDP roughly constant at 10%. This suggests that a major source of income growth in Brazil has been the transfer of labor out of the low productivity agricultural sector into higher productivity sectors. It is also true that relative price movements between agriculture and industry during the 1970s were partly responsible for narrowing the gap between rural and urban mean incomes. The internal terms of trade, defined as wholesale agricultural prices relative to wholesale industrial prices, moved in favor of agriculture, especially in An index number expressing the ratio of the two price indices increased from in 1970 to in For the agricultural sector itself a marked increase in income inequality can be discerned. All measures tell the same story. The Gini coefficient rose from 0.44 to 0.54, the Theil index soared from 0.43 to 0.80 (Appendix Table A3), the income share of the bottom 40% fell from 15.6% to 12.4%, and the share of the top decile went from 36% to 48% (Appendix Table A8).

30 As with the investigation of inequality among sectors, decomposition of the Theil index is a useful tool for preliminary analysis of the sources of this greater income disparity in the primary sector. The Theil index for 1970, TI, may be written for agriculture as (2) T 1 = R 1 +IyilTil i where R 1 is a regional inequality term, the yil1s are regional income shares, and the Til's are regional Theil indices. Using analogous notation, (3) T 2 = R 2 +EYi 2 Ti2 i is the same equation for In both Equations (2) and (3) i ranges from 1 to 10, corresponding to the ten regions of the 1970 census. To allocate the change in the index we can write: (4) T 2 - T 1 = (R 2 - R 1 ) + Tyil(Ti2 - Til) + (Yi2 - yil)ti2- The first expression on the right-hand side of Equation (4) is an effect reflecting changed disparities among regions, the second represents changes in inequality within regions, and the third reflects shifts in income shares to regions with greater or lesser degrees of inequality. Applying Equation (4) to data from Appendix Table A3, we find that 90% of the increase in the Theil index for agriculture is due to more

31 inequality within regions. Only 4% is attributable to higher income shares in regions with greater inequality and only 6% to greater disparity among the ten 1970 census regions. While the fact that of the ten regions three -- the Lower Northeast (chiefly Bahia), the Center (chiefly Minas Gerais), and Paranda -- with 37% of agricultural income account for 56% of the increase in inequality, may be worth exploring, clearly the phenomenon is general. For the agricultural sector, every measure shows greater inequality for every region. In each of the ten census regions, as may be seen from Appendix Table A3, the Gini coefficient rose, the Theil index rose, the income share of the bottom 40% fell, and the share of the top decile rose. In the urban sectors, in contrast, the various regions had differing experiences. Inequality remained roughly constant in the Northeast, declined in Sao Paulo, and rose elsewhere. It is useful here to briefly discuss three possible explanations for the observed increase in inequality within agriculture. First, in addition to lifting agricultural incomes in Brazil generally, the commodities boom of also most likely generated inequality through, for example, the surging production of particular export crops, such as cocoa in Bahia and soybeans elsewhere. Soybeans, cultivated on relatively large farms, probably exercised an especially severe impact on the distribution of earnings where this crop replaced more labor-intensive coffee, as happened in Parana. But these location-specific explanations do not explain why inequality within agriculture rose within every region of Brazil. More geographically diffused sources of change must be sought.

32 A second, and perhaps most important, explanation for increased income inequality within agriculture lies in the growth of the credit subsidies for agriculture during the decade. With fixed nominal interest rates, high inflation, and substantial credit allocation for agriculture the subsidies, both expressed as rates and in volume terms, became substantial. One study has estimated that these credit subsidies, expressed as a proportion, amounted to 21% of the total value of agricultural output in / While also presenting substantial macroeconomic difficulties, along with other, microeconomic distortionary effects, it seems incontestable that these subsidies, and their increase, have had the effect of increasing inequality within agriculture. First, only landowners, as contrasted to landless workers, qualify. Second, the amount of credit, and consequent credit subsidy bonanza, appears to be a positive function of the size of land holdings. Despite the considerable increases in agricultural credit ceilings through the Banco do Brasil in the late 1970s, credit rationing has still proven necessary in the face of the demand for the subsidies. One study (Ferreira, 1981) estimatea that only 4% of the total amount of agricultural credit in the Northeast in 1975 went to farms of less than 10 hectares. A third possible explanation for the observed increase of income inequality within agriculture can be found in the theory of human capital. In every region of Brazil average years of schooling rose for the EAP in agriculture during the 1970's. For all Brazil the composition of the EAP in agriculture by years of schooling was: 14/ Tyler (1981). A comparable study by Gerva'sio Resende and Milton da Mata (1981), employing a somewhat different procedure, presents an even higher estimate.

33 YEARS % 53% 1 10% 7% 2 12% 9% 3 10% 11% 4+ 11% 20% Total 100% 100% Human capital theory predicts that inequality rises with average educational attainment. For Brazil in 1970, Langoni (1973, pp ) confirms this through ginasio by finding the Gini coefficient to be for illiterates, for those completing primario, and for those completing ginasio. Human capital theory also predicts that inequality rises with greater inequality of educational attainment. We find the data for the rural sector of Brazil in accordance with both of these predictions, as shown by Equation (5), in which GINI is the Gini index of income equality for the rural sectors of each of the 23 census "states" in 1980, LAVED is the natural logarithm of the average years of schooling of the EAP, and GED is a Gini index of inequality of years of schooling of the EAP. (5) GINI = LAVED GED R2 = (.060) (.250)

34 The standard errors indicate both coefficients to be significant, and the R 2 shows that these two variables explain five-eighths of the variance in rural inequality across the states of Brazil. Appendix Table A5 displays more regressions of this type. These regressions do not prove causation. It may be that inequality of years of schooling results from income inequality rather than vice-versa, just as inequality in ownership of consumer durables results from, but probably does not cause, income inequality. Studies which suggest the returns to education in Brazil are quite high, particularly for the early years, lead us to think that educational inequality in fact does cause income inequality, especially if years of schooling is considered to include better health and other qualities usually associated with educational attainment. We also note that in the primary sector of Brazil in the 1970's income inequality and educational inequality moved in contrary directions, a further indication that educational inequality plays an independent role. - If the human capital interpretation of Equation (5) is accepted, its coeffcients explain around two-fifths of the observed increase in income inequality in the rural sector as measured by the Gini coefficient. Appendix Table A6 shows that a 34% increase in average years of schooling between 1970 and 1980 accompanied a decrease of GED from to The first change raises inequality but the second lowers it, leaving three-fifths of the increase to be explained. Perhaps the age structure of the EAP changed in a relevant way over the decade. We cannot tell from the advance tabulations. Perhaps the more detailed income questions in the

35 census elicited fuller responses from the upper deciles of the primary sector more than from others. so from the material at hand. Again, we cannot determine whether this is At present the greater inequality within agriculture remains partly a mystery. The results of our analysis of the 1980 demographic census materials can also be used to offer some interpretations regarding the labor market. The evolution of the labor market between 1970 and 1980 is broadly consistent with the hypothesis that, in terms of. a dualistic model, Brazil reached a turning point during the early 1970's. 1 5 / Over the decade real wages roughly doubled, and this gain was shared by the lower income groups. Income inequality may have improved slightly or may have worsened slightly; in either case the change was small. A large rightward shift in the demand for labor in the urban sectors pulled migrants from rural areas, so that the share of agriculture in the labor force plunged from 40% to 28%. This process pulled up rural wages from 37% to 50% of urban levels. Higher urban wages also drew more women into the labor force. For women 10 and over the participation rate rose from 19% to 27%. Apparently women's elasticity of supply was greater at the lower skill levels, with the result that the ratio of female to male incomes, after rising a bit in the 1960's, lost ground in the 1970's, going from 61% to 56% for the economically active population. The Brazilian labor market also changed in ways generally consistent with human capital theory. Associated with higher incomes was a 15/ Morley (1982) advanced this hypothesis in a recent study.

36 rise of 40% in the average years of schooling of the labor force. Illiteracy declined from 33% to 26% of the population 15 and over. At the same time, in accordance with the human capital interpretation, higher educational levels pushed income inequality up, and improved distribution of schooling pulled it back down. The Gini coefficient for inequality of years of schooling dropped from to over the decade (Appendix Table A6). But much of the picture of the development of the labor market in the 1970's remains unclear. One of the gaps is the distributional impact of the higher female participation rate. In a strictly accounting sense rising female participation had little effect on the distribution of individual incomes. In an experiment in which we construct a hypothetical 1980 distribution of income by keeping the male and female patterns as they were in 1980 but reducing the female participation rate to its 1970 level, we find changes in the inequality measures for individuals only in the third places after the decimals. Nonetheless, the fact that a fourth of the increase in the Brazilian labor force in the 1970's was due to the rising female labor force participation rate probably does affect the distribution of income noticeably through processes our methods fail to capture. First, families have more income earners than otherwise. Is the effect concentrated on part of the range of family incomes or spread evenly across it? Second, women compete with men for jobs, especially at the lower end of the spectrum of wages. Does this extra competition drive those wages down? Investigation of these matters requires individual rather than categorical data.

37 Another question concerns the limited integration of the national labor market in Brazil. Wages in the Northeast average less than half of those in the Southeast and have for decades. Why is this differential not eliminated by inter-regional factor flows? labor in the Northeast is less qualified. Part of the explanation is that The average number of years of schooling in the Northeast for the EAP is only half that in the Southeast. Moreover, some migration occurs. Had there been no inter-regional migration at all, the population of the Northeast would have been larger by a sixth in been greater still. The effect of migration on the labor force may have And there may be substantial regional differences in the cost of living, although they would be balanced by a greater availability of public services in higher-cost areas. All in all, the persistence of such enormous regional disparities remains a puzzle. V. CONCLUDING REMARKS The 1980 demographic census presents a wealth of information. The data made available so far, in the form of advance tabulations, have permitted us to undertake some exploratory analysis of poverty and income inequality for between 1970 and The data can then be used for some comparisons The results show that there has been substantial progress in improving living standards during the 1970's. The non-income measures of poverty attest to considerable progress, despite continued and pressing problems of poverty. Average real incomes have also increased substantially, even among the poorest 40% of the economically active population. Overall income inequality did not undergo appreciable change between 1970 and The relative stability of a high level of

38 inequality was due to two offsetting changes. On the one hand, there was observed a large rise in rural incomes relative to urban incomes. On the other hand, however, there was an increase in the inequality of incomes within agriculture. The overall picture has been one of some reduction in regional income disparities, and, in general, the available evidence points to a reduction in absolute poverty during the 1970s. This is not to suggest, however, that income inequality and poverty do not, or should not, remain as principal problem areas for concern for Brazilian policy-makers. One must also ponder whether, in view of the employment opportunities generated during the high growth 1970s, slower economic growth, and recession as observed in , will not exacerbate underlying socio-economic problems.

39 REFERENCES Ramonaval Augusto Costa, Distribuigao da Renda Pessoal no Brasil, 1970 (Rio de Janeiro: IBGE, 1977). Frank Cowell and Fatemah Mehta, "The Estimation and Interpolation of Inequality Measures," Review of Economic Studies, April 1981, pp David Denslow, Jr., "Income Inequality and Poverty in Brazil: Measures Derived from the Advance Tabulations of the 1980 Demographic Census," unpublished paper, Leo da Rocha Ferreira, "Desigualdades entre Diferentes Grupos Socio-Economicos na Agricultura do Nordeste," Textos para Discussao Interna No. 33, IPEA/INPES, Junho de Gary S. Fields, "Who Benefits from Economic Development: A Reexamination of the Brazilian Experience," American Economic Review, Vol. 67, No. 4 (September 1977), pp Marcos G. da Fonseca, "Um Raio-X da Distribuicao da Renda Brasileira: Uma Decomposicao do Coeficiente de Gini," Revista de Estudos Economicos (1980). Albert Fishlow, "Brazilian Size Distribution of Income," American Economic Review, Proceedings Vol. 62 (May 1972), pp M. Louise Fox, "Income Distribution Analysis in Brazil: Better Numbers and New Findings," unpublished paper, January J.L. Gastwirth, "The Estimation of the Lorenz Curve and the Gini Index," Review of Economics and Statistics (May 1972), pp Rodolfo Hoffman and J.C. Duarte, "A Distribuigao da Renda no Brasil," Revista de Administragco de Empresas, Vol.12, No.2 (March 1972), pp Nanak C. Kakwani, Income Inequality and Poverty: Methods of Estimation and Policy Analysis (New York: Oxford University Press, 1980). Peter T. Knight, "Brazilian Socioeconomic Development: Issues for the Eighties," World Development, Vol. 9, No. 11/12 (December 1981), pp Peter T. Knight et al., Brazil: Human Resources Special Report, World Bank Country Study (Washington: World Bank, October Thomas W. Merrick, "The Impact of Access to Piped Water on Infant Mortality in Urban Brazil, 1970 to 1976," World Bank, unpublished paper, August 1981.

40 Carlos Geraldo Langoni, Distribuicao da Renda e Desenvolvimento Economico do Brasil (Rio de Janeiro: Expressao e Cultura, 1973). Constantino Lluch, "On Poverty and Inequality in Brazil," unpublished paper, September Samuel Morley, Labor Market and Inequitable Growth: The Case of Authoritarian Capitalism in Brazil, (London: Cambridge University Press, 1983). Guy Pfefferman and Richard Webb, "The Distribution of Income in Brazil," World Bank Staff Working Paper No. 356 (September 1979)., "Poverty and Income Distribution in Brazil," Revista Brasileira de Economia, forthcoming in Gervasio Castro de Resende and Milton da Mata, "Cr6dito Agricola no Brasil," IPEA/INPES, unpublished paper, Jose W. Rossi, "0 Menor da Concentra'ao de Gini Aplicado a Dados de Distribuisao da Renda no Brasil," Revista de Estudos Economicos, forthcoming. Henri Theil, Economics and Information Theory Holland, 1967). (Amsterdam:North Vinod Thomas, "Differences in Income, Nutrition and Poverty within Brazil," World Bank Staff Working Paper No. 505 (February 1982). William G. Tyler, "Trade Policies and Industrial Incentives in Brazil, ," INPES/IPEA, unpublished report, August 1981.

41 APPENDIX Zero-Income Workers In this appendix we show that the Gini and Theil indices calculated excluding zero-income workers easily can be adjusted to include these people. We turn first to the Gini coefficient and then to the Theil index. The adjustment of the Gini can be developed intuitively in terms of the derivation of this coefficient from a Lorenz curve for the population excluding zero-income individuals which relates q, the accumulated percentage of income, to p the population percentile, written as the inverse function of the usual form: (Al) p = f(q). The Gini coefficient, G, for this population is (A2) G - 2f(p-l)dq = pjdq - I Let N be the number of people with positive incomes and No the number with zero incomes. Let z equal NO/N, the ratio of zero-income EAP to positive-income EAP. The inclusion of the zero-income groups shifts the Lorenz curve rightward so that a percentile pf on the new curve corresponding to the same q is given by

42 (A3) f p+z Hence the Gini coefficient, Gf, for the full group is (A4) Gf = 4jpfdq - I After substitution from (A3) and some manipulation this may be written Gf= 1 1+Iz (2 dg - jf 1) z 0 (A5) Gf = G + - (1 - G). 1+z Thus Gini coefficient for the full population is simply the Gini for the positive-income group plus the ratio of the zero-income individuals to the full population multiplied by the complement of the original Gini. The derivation of the adjustment to the Theil index is direct. The Theil index, T, for positive-income recipients may be written (A6) T =IyilnyiN

43 where yi is the income share of the ith person and "+" indicates summation over positive-income individuals. Defining z as before, the Theil index, Tf, for the full population is (A7) T= E yilnyin(l+z) + E yilnyin(1+z) 0 + where "o" indicates summation over those with zero incomes and each person is now one out of N(1+z) instead of N individuals. The first term on the right-hand side of (A7) goes to zero as the yi go to zero. Noting this and that (A8) Yi = 1 we can write (A7), substituting from (A6), as (A9) Tf = T + ln(1+z) Hence inclusion of a zero-income group augments both the Gini coefficient and the Theil 'index in simple ways. In summary, we have excluded zero-income members of the EAP from our calculations to aid comparison with Langoni's figures for 1970 and because their inclusion weakens the association between individual and family income deciles. Including zero-income recipients' raises the agricultural inequality measures markedly but does not change the general picture: during the 1970s, overall inequality remained roughly constant while rising sharply within agriculture.

44 The 1970 Upper Coding Limit As noted in the text, the determination of changes in the degree of inequality between 1970 and 1980 in Brazil is complicated by the existence in 1970 of an upper coding limit on income. All incomes of 9998 or more cruzeiros per month were coded as 9998, a figure which after adjustment for inflation by the Rio cost-of-living index corresponds to 285,000 cruzeiros of 1980, or approximately US$5,215. It is about 30 times the mean income for This truncation of the upper tail of the income distribution reduces both the Gini coefficient and the Theil index. The effect of this truncation on the Gini coefficient may be viewed through its impact on the Lorenz curve. Let V.Y 1 be the 1=1 amount of true income and,.~ * ŽYi be the amount of coded income, with G calculated from the distribution f(y) and G* calculated from the distribution f*(y*), where f*(y*) is identical to f(y) up to the upper coding limit. The respective Lorenz curves are L(q,p) and L*(q*,p), where p is the percentile of recipients while q and q* are the accumulated percentages of income. We define w to be the ratio of coded to actual total income: N * N (A10) w= Yi/ : i If wdia psh oli i=l pctei=l If we designate p as the population percentile at the upper

45 coding limit, then using an argument similar to that developed in the previous section -- with integration over p instead of q, and using limits -- it may be shown that (All) m 1 G = G* + (1-w)(1-G*). Equation (All) serves as a good approximation for our purposes since the area between the Lorenz curve and the 45-degree line to the right 2f p is quite small. A crude estimate of w, the ratio of coded to actual income, may be obtained from data presented by Lluch (1981). In a random selection of 16,310 families with reported incomes from the 1970 Public Use Sample, he found 19 coded at the upper limit, or 0.12%. Partly for simplicity let us suppose that for individuals the ratio would be slightly lower, say one per thousand. The average income for individuals was 282 cruzeiros per mnnth. As an approximation, take the upper coding limit to be 10,000 instead of 9,998. A person coded at the upper limit would have, out of a thousand typical individuals, a reported income equal to 3.55% of the total reported income of 282,000. Suppose this person's actual income were 20,000 (corresponding to a Pareto distribution with alpha equal to two). Then w would equal 282,000/292,000 or.966, and Langoni's 1970 Gini of.565 would be raised to.580, within the range of error of our estimate for 1980 of.590.

46 It is convenient to use this same numerical illustration to discuss the effect of the upper coding limit on the Theil index. The Theil index derived from coded incomes, T*, can be decomposed as (A12) T* = (272/282)ln[(272/282)/.999] + (10/282)ln[(10/282)/.001] + ( / )TL + (10/ )TH * * where TL and TH are the within-group Theil indices for those below the coding limit and those at or above it, respectively, based on coded incomes. Since all incomes in the upper group are coded at the same value, 10,000, TH is zero. Hence, using Langoni's value of.6629 for T, is found to be T L Now consider T, the Theil index based on actual incomes. If the average income of those in the upper group is 20,000, we have (A13) T = (272/292)ln[(272/292)/.999] + (20/292)ln[(20/292)/.001] + (272/292)(.5912) + (20/292)TH, where TH is the within-group Theil index for those at or above the coding limit. limit. The within-group index is unaffected by removal of the coding If the upper distribution is Pareto with alpha equal to two, then, using a formula derived by Theil (1967, p. 98), TH equals.3068 and T equals.796, which is higher than our estimate for 1980 of.704.

47 APPENDIX TABLE Al PERCENTAGE OF BRAZILIAN HOUSEHOLDS (PERMANENT) POSSESSING SELECTED DURABLE CONSUMERS GOODS AND SERVICES BY REGION AND LOCATION, 1970 AND Region Urban Rural Total Urban Rural Total BRAZIL a/ (Coal) Stove (Gas) (Wood) Telephone n.a. n.a. n.a Electricity Radio Refrigerator Television Automobile NORTHEAST b/ (Coal) n.a. n.a. n.a Stove (Gas) n.a. n.a (Wood) n.a. n.a Telephone n.a. n.a. n.a Electricity n.a. n.a Radio n.a. n.a Refrigerator n.a. n.a Television n.a. n.a Automobile n.a. n.a SOUTHEAST c/ (Coal) n.a. n.a. n.a. - - Stove (Gas) n.a. n.a (Wood) n.a. n.a Telephone n.a. n.a. n.a Electricity n.a. n.a Radio n.a. n.a Refrigerator n.a. n.a Television n.a. n.a Automobile n.a. n.a FRONTIER d/ (Coal) n.a. n.a. n.a Stove (Gas) n.a. n.a (Wood) n.a. n.a Telephone n.a. n.a. n.a Electricity n.a. n.a Radio n.a. n.a Refrigerator n.a. n.a Television n.a. n.a Automobile n.a. n.a Notes: a/ In 1970, 55.9% of the population in Brazil lived in urban areas. In 1980, the figure was 67.6%. b/ In 1970, 41.8% of the population in the Northeast lived in urban areas. In 1980, the figure was 50.4%. c/ In 1970, 64.4% of the population in the Southeast lived in urban areas. In 1980, the figure was 77.3%. d/ In 1970, 46.8% of the population in the Frontier lived in urban areas. In 1980, the figure was 60.7%. * sign -" means percentage of less than 0.1. SOURCES: IBGE, Censo Demografico, Brasil-1970, Volume 1 and Tabulacoes Avancadn s do Censo Demografico, Brasil-1980, Volume 1 - Tomo 2 (Rio de Janeiro: IBGE, 1973 and 1981)..

48 APPENDIX TABLE A2 ECONOMICALLY ACTIVE POPULATION IN URBAN OCCUPATIONS: POPULATION SHARES, INCOME SHARES, RELATIVE PER CAPITA INCOMES, GINI COEFFICIENTS AND THEIL INDICES FOR TEN REGIONS OF BRAZIL, 1970 and 1980 Index of Urban EAP Income Relative Per Gini Share (X) Share (X) Person Income Coefficient Theil Index Region I. North II. Upper NE III. Middle NE IV. Lower NE V. Center Vl. Rio VII. Sao Paulo VIII. Parana IX. Far South X. West BRAZIL Note: EAP stands for economically active population. Calculations in this table are limited to members of the EAP who received a positive income and stat ed the amount to the census-taker. NE stand for Norhteast. The regions are: I. Roraima, Acre, Amapa, Rond6nia, Para, and Amazonas; II. Maranhao and Piaui; III. Ceara, Rio Grande do Norte, Paraiba, Pernambuco, and Alagoas; IV. Sergipe and Bahia; V. Minas Gerais and Espirito Santo; VI. Rio de Janeiro; VII. Sao Paulo; VIII. Parana; IX. Santa Catarina and Rio Grande do Sul; X. Mato Grosso do Sul, Mato Grosso, Goias, and the Distrito iederal. The 1970 data are from Langoni (1973, pp ). Langoni's inequality measures are calculated from ILIuIvIouaJ aaea. ine itou figures are calculated from the Tabulacoes Avangadas of the 1980 demographic census of Brazil. For estimating inequality measures we used the number of people in each income interval from Table 5.2 and estimated the mean income for each income interval be weighting the interval means for urban males and females in Table 5.3 by male and female interval numbers in Table 5.2 state by state. Bounds on the Gini coefficients were calculated using Method I by Gastwirth (1972) and those on the Theil index with the technique described by Theil (1967). Point estimates of the Gini and Theil measures were derived from these bounds using the "one-third, two-thirds" rule adovcated by Cowell and Mehta (1982). The upper bound on coding the income data used by Langoni was approximately 35 times the mean income, affecting perhaps 1% of the population. We do not know if there was an upper bound on the 1980 data. If so it was at 70 or more times the mean income. For both 1970 and 1980 "urban" occupations are all but agriculture. For this group the mean income for Brazil was 378 cruzeiros per month in 1970 and 13,914 cruzeiros per month in Adjusted by the Rio cost-of-living index to August 1980 cruzeiros, the 1970 mean income in 8,023. The Gini coefficient for the EAP for urban populations for all Brazil in 1970 is calculated through the application of the Gastwirth method to data aggregated with a quadratic approximation technique. For comparison, the Theil index derived with these methods is.629, close to the value of.622 obtained with the decomposition formula for the Theil index.

49 APPENDIX TABLE A3 ECONOMICALLY ACTIVE POPULATION IN AGRICULTURE: POPULATION SHARES, INCOME SHARES, RELATIVE PER CAPITA INCOMES, GINI COEFFICIENTS AND THEIL INDICES FOR TEN REGIONS OF BRAZIL, 1970 and 1980 Agricultural Index of I EAP Income Relative Per Gini Share (X) Share (X) Person Income Coefficient Theil Index Region I. North II. Upper NE III. Middle NE IV. Lower NE V. Center VI. Rio VII. Sao Paulo IIII. Parana IX.. Far South X. West RAZIL iote: The regions, sources, and methods are the same as for Table 1, except the mean incomes for each incom interval are derived from those from the rural population of Table 5.3 for each state. For those receiving incomes from agriculture the mean income for Brazil was 139 cruzeiros per month in 1970 and 6,668 cruzeiros per month in Adjusted by the Rio cost-of-living index, the 1970 mean income is 2,950 cruzeiros per month.

50 APPENDIX TABLE A4 ECONOMICALLY ACTIVE POPULATION IN 1980 BY STATES OF BRAZIL. AVERAGE INCOMES, GINI COEFFICIENTS AND THEIL INDICES FUR AGRICULTURAL AND FOR URBAN OCCUPATIONS Population /a Mean Income /b Gini Theil State Urban Agricul. Urban Agricul. Urban Agricul. Urban Agricul. 1. Rondonia /c ,085 7, Amazonas ,964 7, Para ,674 7, Maranhao ,398 3, Piaui , Ceara ,335 3, Rio Grande do Norte ,891 3, Paraiba ,524 3, Pernambuco 1, ,768 3, Alagoas ,331 3, Sergipe ,099 4, Bahia 1,514 1,381 11,298 5, Minas Gerais 3,017 1,462 11,454 7, Espirtio Santo ,004 7, Rio de Janeiro 4, ,603 8, Sao Paulo 8,935 1,170 16,137 8, Parana 1, ,284 8, Santa Catarina ,238 9, IV. Rio urande do Sul 2, ,662 11,171.3bo.58b.b5l.13t.U. r4tc Uro^bo Qo b6l J19 6ib 12,6U4 11, b 1.u Mato Grosso ,329 8, Goias ,668 9, Distrito Federal ,421 17, BRAZIL 29,481 10,874 13,914 6, Note: The sources and methods are the same as those for 1980 in Tables I and 2. The heading "Agricul." refers tc agricultural occupations and "Urban" to all other. /a Population is in thousands and refers to members of the economically active population who are receiving a positive income and stated the amount to the census-taker. /b Mean income is in cruzeiros per month. At the time of the census, the exchange rate was equal to Cr$54.65 per US dollar. /c Rondonia, Acre, Roraima, and Amapa.

51 APPENDIX TABLE A5 REGRESSIONS OF INCOME INEQUALITY MEASURES ON THE LOGARITHM OF AVERAGE SCHOOLING AND A GINI INDEX OF SCHOOLING INEQUALITY, 1980 Dependent Coefficients Regression Year Observations Occupations Variable Constant LAVED GED R States All GINI (.0598).6029(.1592) States Rural GINI (.0595).5872(.2498) States Urban GINI (.0875).6038(.1842) States Pooled GINI (.0269).3191(.1335) States All TOP (.0338).3721(.0900) States All LOW (.0169) (.0405) Notes: The dependent variable GINI is the Gini Index of income inequality for each group. The dependent variable TOP40 is the income share of the highest 40% of individual earners and LOW40 is the lowest 40%. The regressors are LAVED, the natural logarithm of average years of schooling, and GED, a Gini index of inequality of years of schooling. These measures are calculated from Tabulacoes Avancadas do Censo Demografico, 1980, Table 5.2. Under the heading "Occupations" the word "pooled" refers- to-placing both urban and rural occupations in one regression as separate observations. Thus, in Regression 4, this amounts to having 23 rural plus 23 urban "regions. Standard errors are in parentheses. All variables are significantly different from zero at the 5% level.

52 APPENDIX TABLE A6 AVERAGE YEARS OF SCHOOLING AND INEQUALITY OF YEARS OF SCHOOLING, THREE REGIONS OF BRAZIL, 1970 AND 1980 Total Economically Active Population Average Years Gini Index of Schooling Region Southeast Northeast Frontier Brazil Rural Occupations Average Years Gini Index of Schooling Region Southeast * Northeast Frontier Brazil Urban Occupations Average Years Gini Index of Schooling Region Southeast Northeast Frontier Brazil Note: These measures are calculated from Tabulacoes Avancadas do Censo Demografico, 1970, Table 6, and Tabula,oes Avangadas do Censo Demografico, 1980, Table 5.2. They are approximations derived from interval data. The regions are defined in Table 7.

53 APPENDIX TABLE A7 INCOME DECILES FOR THREE REGIONS OF BRAZIL, ECONOMICALLY ACTIVE POPULATIONS, 1970 AND 1980 SOUTHEAST Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change ,284 1, ,226 3, ,938 4, ,137 5, ,220 6, ,734 8, ,388 10, ,298 14, ,719 21, ,503 64, AVERAGE 9,756 13, NORTHEAST Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change , ,113 1, ,568 2, ,969 2, ,310 3, ,653 3, ,281 4, ,137 6, ,848 9, ,452 35, AVERAGE 4,478 7, FRONTIER Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change ,484 1, ,254 2, ,681 3, ,423 4, ,137 4, ,394 5, ,505 7, ,333 10, ,813 15, ,756 53, AVERAGE 6,790 10, Note: The figures for 1970 are based on Table 8 of the Tabulacoes Avancadas do Censo Demografico, The methods for 1970 are described in Denslow (1982). Ths figures for 1980 are from Tables 5.2 and 5.3 of the Tabulacoes Avangadas do Censo Demografico, The methods are those of previous tables. For both years the data describe members of the economically active population with positive incomes and reporting those incomes to census takers.

54 APPENDIX TABLE A8 INCOME DECILES FOR RURAL OCCUPATIONS, REGIONS OF BRAZIL, 1970 AND 1980 THREE ALL BRAZIL Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change , ,392 1, ,712 2, ,180 2, ,623 3, ,910 3, ,614 4, ,360 5, ,477 8, ,308 31, AVERAGE 3,943 6, SOUTHEAST Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change ,027 1, ,740 2, ,168 3, ,511 3, ,824 4, ,708 4, ,251 5, ,021 7, ,133 11, ,457 42, AVERAGE 4,793 8, NORTHEAST Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change ,027 1, ,426 1, ,740 2, ,054 2, ,310 3, ,567 3, ,852 3, ,993 5, ,671 16, AVERAGE 2,710 4, FRONTIER Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change ,568 1, ,226 2, ,539 3, ,796 3, ,480 4, ,107 5, ,308 6, ,220 7, ,990 11, ,179 37, AVERAGE 4,536 8, Note: The figures for 1970 are based on Table 8 of the Tabulacoes Avangadas do Censo Demografico, The methods for 1970 are described in Denslow (1982). The figures for 1980 are from Tables 5.2 and 5.3 of the Tabulacoes Avancadas do Censo Demografico, The methods are those of previous tables. For both years the data describe members of the economically active population with positive incomes and reporting those incomes to census takers.

55 APPENDIX TABLE A9 INCOME DECILES FOR URBAN OCCUPATIONS, REGIONS OF BRAZIL, 1970 AND 1980 THREE ALL BRAZIL Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change ,133 1, ,333 2, ,636 3, ,645 4, ,453 5, ,194 7, ,561 9, ,522 13, ,498 20, ,815 62, AVERAGE 10, ,199 SOUTHEAST Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change ,484 1, ,882 3, ,422 4, ,220 5, ,677 7, ,159 9, ,784 11, ,495 15, ,171 23, ,489 67, AVERAGE 11,967 15,123 2 NORTHEAST Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change , ,312 1, ,997 2, ,653 3, ,537 4, ,137 5, ,306 *6, ,875 9, ,697 14, ,773 46, AVERAGE 7,189 9, FRONTIER Accumulated Percentage of Income Percentage of Income Average Income Percentage Decile Change ,341 1, ,425 2, ,395 3, ,165 4, ,764 5, ,762 6, ,131 8, ,670 11, ,522 18, , AVERAGE 9,271 12,3 r Note: The figures for 1970 are based on Table 8 of the Tabulazoes Ayanjadas do Censo Demografico, The methods for 1970 are described in Denslow figures for 1980 are from Tables 5.2 and 5.3 of the Tabulacoes Avangadas do Censo Demografico, The methods are those of previous tables. For both years the data describe members of the economically active population with positive incomes and reporting those incomes to census takers.

56

57 World Bank Publications of Related Interest NEW rhe Determinants of Labour income countries in terms of a num- Economic Growth and Earnings in Developing ber of socioeconomic indicators, it is Employment in China netropolises: Estimates noted that there are large regional Thomas G. Rawski disparities between the northeast and firom Bogota and Call, the southeast with the northeast far- Examines how China's agricultural Colombia ing about the same as most low- sector has made significant strides Rakesh Mohan income countries. toward full employment for a labor force of nearly half a billion persons. Examines the patterns of behavior World Bank Staff Working Paper No. Implicit in the variation of labor earn February pages (Includ- Oxford Uniuersity Press, ings in Bogota and Cali, Colombia. ing references, map). 208 pages (including maps, bibliogra- World Bank Staff Working Paper No. Stock Nlo. WP $3.00. phy, index) October pages (includ- LC ISBN , ing 2 appendixes, bibliography). $18.95 hardcover. ISBN , Stock No. WP $5.00. The Distribution of Income In Brazil $7.95 paperback. French: Croissance et emploi en Chine. Guy P. Pfeffermann and Economica, NEW Richard C. Webb ISBN , 39 francs. World Bank Staff Working Paper No. Differences in Income, 356. September pages Spanesh: Crecimiento economico y lmutrition, and Poverty (Including 2 appendixes). empleo en China. Editorial Tecnos, iinod Thoras Stock No. WP $5.00. ISBN , 600 pesetas. Regional disparities in living staniards in Brazil are examined and!stimates for real income and nutri- :onal levels and poverty for major irban and rural areas are provided. Ulthough the growth rate of the counry surpasses the average for most Jeveloping countries and the country,erforms as well as many middle- Employment Patterns and Income Growth Joseph J. Stern and Jeffrey D. Lewis World Bank Staff Working Paper No September pages (including bibliography, 2 appendixes). Stock No. WP $3.00.

58 ,mployment Policy in Household Income or Income Distribution Policy Weveloping Countries: Household Income in the Developing Coun- L Survey of Issues per Capita in Welfare tries: A Case Study of Korea mnd Evidence Comparisons Irma Adelman and yn Squire Gautam Datta and Sherman Robinson ow rates of growth in industrial Jacob Meerman Seeks to answer the question of how mployment, high rates of unemploy- World Bank Staff Working Paper No. much can actually be done to ient among new entrants to the 378. Bank Staff Work ing improve income distribution by rban labor market, and low levels of 378. March pages (including means of a dynamic general ibor productivity and remuneration 3 appendixes, references). equilibrium model for investigating re the three issues addressed in this Stock No. WP $3.00. the potential impact of standard tudy. The author identifies the impor- policy instruments and programs int determinants of labor demand intended to improve the relative and nd supply and the extent to which How Segmented is the absolute incomes of the poor. ie growth of labor demand has been Bogota Labor Market? Stanford University Press, Stanford, dvanced-by inappropriate policies. Gary S. Fields California 94305, U.S.A. (except the in the demand side, industrial trade World Bank Staff Working Paper No. United Kingdom and the Commonolicy, agricultural growth, and the 434. October pages (includ- wealth Market) Oxford University peration of capital markets are ing bibliography). Press (except Mexico, United States, iscussed; on the supply side, atten- Stock No WP-0434 $ and Canada) pages (includon is focused on population and ' ' ing 6 appendixes, index). ducation policy. LC ISBN xford University Press, Incidence of Poverty and (Stanford), $12.50 hardcover. ages (including bibliography, index). the Characteristics of ISBN X (Oxford), 7.50 C ISBN X, the Poor in Peninsular hardcover ( 11.50) hardcover, ISBN Malaysia, , $7.95 ( 5.00) Pravin Visaria Income Inequality and aperback. Examines the results of the Poverty: Methods of Estima- Household Expenditure and Income tion and Policy Applications irowth and i Equity in Semi- ~~~~~~~~Malaysia Survey conducted during 1973 in peninsular order to Nanak C. K~akwani ndustrialized Countries understand better the correlates Deals with income distribution oel Bergsman of poverty. methods and their economic 'orld Bank Staff Working Paper No. World Bank Staff Working Paper No. applications. 51. August ii pages 460. May viii pages Oxford University Press, ncluding references). (including statistical appendix). pages (including bibliography, tock No. WP $5.00. Stock No. WP $ indexes). LC ISBN , irowth with Equity: Income Distribution $12.95 paperback. 'he Taiwan Case and Poverty in Mexico ohn C. H. Fei, Gustav Ranis, Joel Bergsman NEW nd Shirley W. Y. Kuo World Bank Staff Working Paper No. itroduces a method for tracing 395. July pages (including Inequality and Poverty ie inequality of family income to references). in Malaysia: ie inequality of various kinds of Stock No. WP $3.00. Measurement and mctor income and the shares f factor income in family income. Sudhir Anand ixford University Press, An account of income inequalities 44 pages (including index). and poverty in Malaysia. The research C ISBN , is policy oriented and the findings, to ( 10.25) hardcover; which the author's statistical tech- ;BN , $12.95 nique is applied, are thoroughly C4.25) paperback. discussed. A range of issues is

59 covered, from data problems to con- regarding links between human Studies the evolution of the structure ceptual questions arising with respect development and labor productivity of consumer preferences as a functo measurement. and focuses on policy measures tion of economic development. Oxford Uniuersity Press aimed the poor at raising theodutvt o f Oxford University Preso, and promoting economic OxodUiestPrs pages (including bibliography, index). development. pages (including bibliography, LC ISBN , World Bank Staff Working Paper No. indexes). $27.50 hardcover October in + 70 pages LC ISBN , (includintobe iog +70hage $22.50 hardcover; (including bibliography). ISBN , $12.95 International Migrant Stock No. WP $3.00. paperback. Workers' Remittances: Issues and Prospects NEW Gurushri Swamy TePolicy interventions for A background study for A World background Manpower study for and International Developing Countries DevJelopment Report Puts together available data on interna- Labor Migration in the Mid- Charles Cooper tional migrant workers' remittances to dle East and North Africa World Bank Staff Working Paper No. developing countries and analyzes Ismail Serageldin, James A December pages the regional structure of and growth Socknat, Stace Birks, Bob Li, (including references). in these flows. Relates the flow of anclva.scai remittances to the level of and fluc- an ive. inclair Stock No. WP $3.00. tuations in economic activity and to This study of labor market trends in inflation in the host countries and the region from 1975 to 1985 attempts to forecast future demand analyzes the demand for and supply Poverty and Growth for migrant labor and the likely of manpower by country, sector, in Kenya growth in remittances. occupation, sex, educational level, Paul Collier and Deepak Lal and ethnic composition. It indicates World Bank Staff Working Paper No. the sources and destinations of World Bank Staff Working Paper No August pages (including workers and examines the social and 389. May pages (including 2 appendixes, bibliography). economic effects of migration for 2 appendixes, bibliography). Stock No. WP $3.00. both importers and exporters, par- Stock No. WP $5.00. ticularly the implications forstcnowp government policy and planning. Labor Migration Oxford University Press About Poverty and Unemployment from Bangladesh to 224 pages. in India: An Analysis of the Middle East ISBN $27.50 hardcover. Recent Evidence Syed Ashraf Ali, Pravin Visaria World ~Technological Innovation in Abu Ahmed Arif, A. u Md Habibulah, A Model for Income World Bank Staff Working Paper No. A. R. M. Anwar Hossain, Distribution, Employment October iii + 39 pages Rizwanul Islam, Wahiduddin and Growth: A Case Study (including statistical annex). Mahmud, S. R. Osmani, of Indonesia Stock No. WP $3.00. Q.M. Rahman, and Syamaprasad Gupta A.M.A.H. Siddiqui Presents a quantitative framework to Public Expenditure in Discusses the economic and social explore the implications of alternative Malaysia: Who Benefits implcatins laor mgraion rom policies for investment, employment, an Wh implcatins f laor mgraion rom income distribution, and riscalan Wh Bangladesh to the Middle East. Part of activity for medium-term and Jacob Meerman a research project that also includes a long-term growth. A methodologically novel study of study on Pakistan. The Johns Hlopkins University Press, the household distribution of public World Bank Staff Working Paper No. pages Uniuding s7er138 tvices in relation to incomes. The 454. April pages (including ipages (including study examines public spending for appendix). 5 appendixes, bibiography). education, medical care, agriculture, Stock No. WP $ LC ISBN , public utilities, and welfare transfers $5.50 ( 3.25) paperback. and the use of these services by households. NEW Patterns in Household Oxford University Press, Lar Demand and Saving pages (including map, bibliography, Labor Productivity: Un Tour Constantino Lluch, Alan A. index). d'horizon Pojtwhell, atnrjd RtosOs A. Williagmrs, LC ISBN , Susan Horton and Pwell, aondtribtoss A. Williams, $27.50 (V~.25) hardcover; Timothy Ksing withncontribtionsr byhoger,r ISBN , $12.50 (~.025) This study surveys the state of and Philip Musgrove paperback. knowledge and the research available

60 Rural Enterprise and Trends in Rural Savings and LC ISBN , Nonfarm Employment Private Capital Formation $20.00 ( 6.25) hardcover; Dennis Anderson and in India ISBN , $9.95 (S2.25) Mark Leiserson Raj Krishna and paperback. Examines the importance of rural G.S. Raychaudhuri nonfarm activities as a source of World Bank Staff Working Paper No. Women in the Urban Labor employment and income for the 382. April pages (including Markets of Africa: The poorest groups of the world's 2 tables, 3 appendixes, references). Case of Tanzania A World Bank Paper. January Stock No. WP $3.00. Nwanganga Shields pages (including 25 annexes). English, World Bank Staff Working Paper Nio. French, and Spanish. Unskilled Labor for 380. April pages (including Stock Nos. PP E, PP-7802-F, Development: 5 appendixes). PP-7802-S. $5.00. Its Economic Cost Stock No. WP $5.00. Orville John McDiarmid Size of Land Holding, Living Estimates the economic cost of REPRINTS Standards and Employment unskilled labor by examining market Income Distribution Effects of Techniin Rural Western India, conditions spread between that cause a significant cal Change: Some economic and Analytical Issues Hans P2 Binswanger 1 ' market wages. World Bank Reprint Series: Number 212. P'ravin Visaria TeJhsHpisUiritPes, Reprinted from South East Asian Economic Analyzes the relationship between the pages (including 5 appen- ReSiteo Nom. RP-2no. 3rDecember 1980)c size of land holding possessed by a dixes, bibliography). Stock No. RP Free of charge. household and the household's level of per capita expenditure in rural LC ISBN , Intra-Urban Location of Manufacturing areas of Gujarat and Maharashtra $14.00 (48.50) hardcover; Employment In Colombia states of India from October 1972 to ISBN , $5.50 ( 3.25) Kyu Sik Lee September 1973, on the basis of the paperback. World Bank Reprint Series: Number th Round of the Indian National Reprinted from Journal of Urban Economics, vol. Sample Survey. 9 (1981): World Bank Staff Working Paper No. The Urban Labor Market Stock No. RP Free of charge May pages (including and Income Distribution: Technological Change, Distributive statistical appendix). A Study of Malaysia Bias, and Labor Transfer In a Two- Stock No. WP $5.00. Dipak Mazumdar Sector Economy An integrated analysis of differences Ume Lele and John W. Mellor in income among urban households, World Bank Reprint Series: Number 205. the determinants of employment Reprinted from Oxford Economic Papers. uol. 33. NEW rates of household members, and no. 3 (November 1981): Some Aspects of Relative the major factors affecting personal earnings. Stock No. RP Free of charge. Poverty in Sri Lanka, Oxford University Press, Pravin Visaria 392 pages (including index). The major findings of an intensive LC ISBN , analysis of the data collected in the $26.00 hardcover; ISBN , Socioeconomic Survey of Sri Lanka $9.95 paperback. from November 1969 to October 1970 are presented. Highlights the importance of the criteria used to rank Wilo Beneflts from Goverhouseholds (per capita vs. total ment Expenditure? A Case household expenditure or income) for Study of Colombia the estimates of the level of inequal- Marcelo Selowsky ity, as well as for the measures of elasticity of demand and for different Clearly identifies which income categories of consumption items. groups benefit from public expenditures and attempts to separate the World Bank Staff Working Paper No. factors behind the supply of and 461. May xii pages (includ- demand for a public service as a ing 5 annexes). means of understanding the present Stock No. WP $ pattern of consumption. Oxford Uniuersity Press, pages (including statistical appendix, bibliography, index).