Report No. 22425-BR Public Expenditures for Poverty Alleviation in Northeast Brazil Promoting Growth and Improving Services June 11, 2001 Brazil Country Management Unit Latin America and the Caribbean Region Document of the World Bank Currency Equivalents US$1= 1.20 Brazilian Real (as of December 14, 1998) Weights and Measures Metric System Fiscal Year January 1-December 31 Abbreviation and Acronyms CHESF Hydroelectricity Company of the Sao Francisco River DHS Demographic Health Surveys DNOCS Federal Agency of Works Against Droughts (formerly IFOCS) EEW Electric energy and water supply FGV Fundacao Getulio Vargas GDP Gross Domestic Product GIEWS Global Information and Early Warning System IBGE Instituto Brasileiro de Geografia Estatistica IPEA R Instituto de Pesquisa Economica Aplicada - Rio de Janeiro IPEA Instituto de Pesquisa Economica Aplicada IV Instrumental variable LPG Labor productivity growth MC Municipal Councils NB Banco do Nordeste do Brasil OLS Ordinary least squares PAPP Programa de Apoio ao Pequeno Produtor PME Pesquisa Mensal de Emprego PNAD Pesquisa Nacional Anual do Domicilios PNDS Pesquisa Nacional Demografia e Sauide PPV Pesquisa sobre Padroes de Vida SADC Southem African Development Community SUDEN Superintendencia do Desenvolvimento Econ6mico do Nordeste SUDENE Superintendencia de Desenvolvimento do Nordeste TC Transport and communication TFP Total factor productivity TFP Total factor productivity UNICEF United Nations International Children's Emergency Fund VAR Vector autoregression WFP World Food Programme Vice President, LCR: David de Ferranti Director, LCC5C: Gobind T. Nankani Lead Economist, LCC5C: Suman Bery/Joachim von Amsberg Task Managers: Jeffrey Hammer/Zmarak Shalizi Public Expenditures for Poverty Alleviation in Northeast Brazil: Promoting Growth and Improving Services Table of Contents Executive Summary ........................................................................i A. Regional Growth ........................................................................i B. Expanding Social Services ....................................................................... vi C. Targeting Transfers .........................................................................x D. Need for Selected Further Work to Refine these recommendations ........................ .......................... xii Chapter 1. Introduction ........................................................................1 A. An Update On Poverty ........................................................................2 B. Land Distribution and Rural Poverty ........................................................................5 Chapter 2. Broad-Based Regional Growth and Poverty Reduction ............................ ..........................7 A. Trends In Output Growth And Volatility ......................... ...............................................7 B. Determinants of Growth in the Northeast ....................................................................... 12 C. State Growth Rates and Poverty Reduction ................................... .................................... 13 D. The Importance of Macro and Fiscal Policies for Poverty Reduction in the Northeast .......... ........... 13 Chapter 3. Poverty Reduction Through Expanding Social Sector Expenditures ............................... 21 A. Health ....................................................................... 21 B. Education ....................................................................... 33 Chapter 4. Poverty Reduction Through Selected Targeted Interventions and Transfers ................. 40 A. Pensions ....................................................................... 40 B. Coping With Droughts ........................................................................ 42 Chapter 5. Conclusions ....................................................................... 46 Bibliography ....................................................................... 49 Appendix 1. Volatility of Income in the Northeast A. Possible Explanations for the Decline in Volatility of Regional Income ................ ....................... Al -I B. Patterns of Volatility in Agricultural Output Across States ............................................................ Al -3 C. Major Agricultural Products, Northeast Brazil ........................................................................ Al-8 D. The Trend in Share of Agricultural Output in All Northeastern States ................ ........................ Al-10 Appendix 2. Growth of Income in the Northeast A. Motivation for a "Determinants of Growth" Type Analysis ........................................................... A2- 1 B. Methodological and Data Limitations Precluding a Full "Determinants of Growth" Analysis for Northeast Brazil ....................................................................... A2-3 C. Rudimentary Analysis and Provisional Interpretations ................................................................... A2-5 D. Provisional Implications for Policy ........ ............................................................... A2- 10 Appendix 3. Macroeconomic Statistical Results A. Poverty and Inequality ....................................................................... A3- 1 B. State Policies and Deficits ....................................................................... A3-2 Appendix 4. Health Results A. Method ............................................................ A4-1 B. Statistical Analysis of Health and Health Care ............................................................ A4-2 C. Insurance ............................................................ A4-7 Appendix 5. Education Appendix 6. Drought and Poverty in the Northeast A. Minimizing the Effect of Drought on the Poor ............................................................ A6-1 B. The 1998 Drought ............................................................ A6-2 Text Figures ES- I National and Regional Cumulative Distribution Functions of Per Capita Consumption ii ES-2 Ratio of Per Capita GDP of the Northeast Region to the Rest of Brazil. iii ES-3 Structural Composition of the Northeast Economy .v ES-4 Lack of Access to Water and Sanitation Increases Child Mortality .vii ES-5 The Poor Have Less Access to Safe Water .viii ES-6 The Poor Have Less Access to Sanitation .viii ES-7 Percent of Income from Pensions .x ES-8 Average Pension in Reais .xi ES-9 Number of Households by Consumption Decile, Northeast Rural .xi 1 National and Regional Cumulative Distribution Functions of Per Capita Consumption . 4 2 Ratio of Per Capita GDP of the Northeast to Brazil .8 3 Regional Variations in Per Capita Income .9 4 Growth Rates of Per Capita Output .9 5 Structural Composition of the Northeast Economy .10 6 Administration Costs and Deficits .17 7 Deficit as Proportion of Revenues Ceara, 1986-1995 .17 8 Public Employment Rates By Region .18 9 Distribution of Expenditures With and Without Public Employment-Northeast .1 9 10 Child Deaths .22 11 Nationwide: The Poor Have Less Access to Sanitation .23 12 The Poor Have Less Access to Safe Water .24 13 Lack of Access to Water and Sanitation Increases Child Mortality .24 14(a) Coefficients in Census Study by Predicted Effect and Statistical Significance .25 14(b) Coefficients in Census Study by Predicted Effect and Statistical Significance ............................. 26 15 Likelihood of Seeking Care by Health Problem .28 16 From the National Perspective: The Distribution of Health Care Benefits in the Rural Northeast .29 17 Within the Rural Northeast: Distribution of Health Care Benefits .29 18 Health Care Visits by Decile .30 19 Coverage by Health Insurance .31 20 Value of Insurance by Household Per Capita Consumption .31 21 Value of Insurance by Cost of Insured Service .32 22 From the National Perspective: Distribution of Benefits to Public Primary Education .33 23 Within the Rural Northeast: Distribution of Benefits to Public Primary Education .34 24 Brazil 1996- Northeast Education Attainment for Ages 15-19 .34 25 Brazil 1996-Northeast Education Attainment for Ages 15-19 .35 26 Brazil 1996-Northeast Education Attainment Ever Enrolled for Ages 6-14 .35 27 Brazil 1996-Non-Northeast Ever Enrolled for Ages 6-14 .36 28 Brazil 1996- Northeast Education Attainment for Rural Sector .................................................... 37 29 Enrollment Rates by Decile, Northeast ............................................................. 38 30 Distribution of the Educational Level for Natives and Emigrant Natives of the Northeast 1995 .39 31 Percent of Income from Pensions ............................................... 40 32 Average Pension in Reais ............................................... 41 33 Number of Households by Consumption Decile, Northeast Rural ............................................... 42 Text Tables 1 Poverty Rates, 1990-95 . 2 2 Land Ownership by Consumption Deciles in the Northeast and Southeast . 5 3 Northeast Brazil: Output, Employment, and Labor Productivity .1 1 4 Impact of Safe Water and Sanitation on Mortality Rates . 26 Appendix Figures AL .1 Rates of Growth of Agriculture and Aggregate Output, Northeast Brazil . A-I A 1.2 Structural Composition of the Northeast Economy . Al-2 A 1.3 Maranhao: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 ................ A1-3 A1.4 Ceara: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 ....................... A1-4 A 1.5 Paraiba: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 .................... Al -4 A1.6 Piaui: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 ........................ Al -4 A1.7 Rio Grande do Norte: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96Al -4 A1.8 Sergipe: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 .................... Al-6 Al.9 Alagoas: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 .................... Al-6 A 1.10 Bahia: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 .. A1-6 Al .11 Pemambuco: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 .. Al -6 A 1.12 Alagoas: Relative Share of Agriculture and Industry in GDP, 1970-96 ...Al1.- 10 A1.13 Ceari: Relative Share of Agriculture and Industry in GDP, 1970-96 .. Al-10 A1.14 Maranhao: Relative Share of Agriculture and Industry in GDP, 1970-96 .. Al-10 A1.15 Rio Grande do Norte: Relative Share of Agriculture and Industry in GDP, 1970-96 .. Al-10 A 1.16 Paraiba: Relative Share of Agriculture and Industry in GDP, 1970-96 .. Al-l0 A1.17 Bahia: Relative Share of Agriculture and Industry in GDP, 1970-96 .. Al-11 Al.18 Pemambuco: Relative Share of Agriculture and Industry in GDP, 1970-96 .. Al-11 A1.19 Piaui: Relative Share of Agriculture and Industry in GDP, 1970-96 .. Al-11 A1.20 Sergipe: Relative Share of Agriculture and Industry in GDP, 1970-96 .. Al-11 A2.1 Private and Public Investment in Northeast Brazil, 1965-91 ... A2-8 A4.1 Value of Insurance by Cost of Insured Service .. A4-8 A4.2 From the National Perspective: Distribution of Benefits to Public Primary Education ... A4-8 Appendix Tables A 1.1 Change in Output Between Drought and Nondrought Years . ..A 1A-2 A 1.2 Extent of the Semiarid Region in Northeast States .. Al-3 Al1.3 Permanent Agricultural Products, Northeast Brazil .. A1-8 A1.4 Temporary Agricultural Products, Northeast Brazil .. A1-9 A2.1 Description of Variables Used in the Analysis .. A2-4 A2.2 Factors Influencing Output Growth in the Northeast: Estimates Using a Single Equation Model .. A2-6 A2.3 Summary of Empirical Estimates .. A2-7 A2.4 Factors Influencing Private Capital Formation in the Northeast .. . A2-9 A2.5 Association Between Public Capital and Output Across States in Northeast Brazil . . A2-9 A2.6 Effects of Water and Electric Infrastructure Investment on Output Across Northeast States...A2-10 A2.7 Northeast Brazil: Population in Major Urban Areas ................................................................. A2-11 A3.1 Determinants of Poverty Rates in Northeastern States, 1981-95 ................................................ A3-1 A3.2 Macroeconomic Effects on Poverty and Inequality .................................................................... A3-1 A3.3 Determinants of State Fiscal Deficits ...................................................................... A3-2 A3.4 Determinants of Hourly Wages ...................................................................... A3-2 A3.5 Percent Fall in Rural Wages from a 20 Percent Fall in Public Employment .............................. A3-3 A4.1 Determinants of Mortality Equations ...................................................................... A4-3 A4.2 Determinants of Self-Reported Morbidity ...................................................................... A4-4 A4.3 Logit of Infant Mortality Rates ...................................................................... A4-5 A4.4 Health Care Demand (Public versus Private) ...................................................................... A4-6 A5.1 Determinants of Earning ...................................................................... A5-2 A5.2 Education Demand ...................................................................... A5-3 This report was prepared by Jeffrey Hammer and Zmarak Shalizi of the Development Research Group and Antonio Magalhaes from the Resident Mission, Brasilia, with contributions from Martin Ravallion. Research assistance was provided by Somik Lall, John Voss, and Maru Bonilla-Chacin. Cynthia Bernardo, Anna Marailon and Roula Yazigi processed the manuscript. Most of the work leading to this report was undertaken in 1997 and 1998, based on data from the PPV household survey fielded in 1996/7. This report is based on information available at that time does not reflect changes that have occurred since then. In particular, major social policy reforms have been ongoing for the last few years and are not reflected in this report. Drafts of this report have been discussed with many institutions and individuals of the Brazilian Government, and their comments have been taken into account. However, the views expressed in this report are exclusively those of the World Bank and not necessarily those of the Brazilian Government. EXECUTIVE SUMMARY 1. This report addresses the role that public expenditure can play in the alleviation of poverty in the Brazilian northeast. The search for solutions to poverty in the northeast is as long-standing, in terms of years of concern and pages expended on analysis, as any topic of economic and social policy in Brazil. It should come as no surprise, then, that this report will not be able to suggest a simple formula, a "magic bullet," to alleviate poverty in the region that somehow escaped the attention of the contributors to that literature and policy debate. However, it does try to make two kinds of contribution. First, it will link public expenditures to an overall approach to poverty alleviation involving both regional growth and social services. Second, it brings to bear on the question several new or newly compiled sources of data. 2. The policy discussion is primarily from the perspective of policymakers in the region rather than that of the federal government. While some policies needed to combat poverty are primarily the responsibility of federal government and will be discussed here, the focus will be on those aspects of public expenditure which are under the control of the state governments. Ultimately, it is at this level that public expenditure policies are implemented and at this level that the poverty orientation of government can be made tangible. 3. A useful framework, commonly used in the World Bank for discussing policies to alleviate poverty, stands on two-and-a-half legs (World Bank 1990). The first is to encourage broad-based growth, usually understood as labor-intensive growth. The second is to ensure that social services reach the poor-usually by guaranteeing that they are universally available. And the half-leg is to use (sparingly) carefully targeted transfer programs to address special needs of particular subpopulations. This broad framework can be profitably applied to the role that public expenditures should play in the context of the Brazilian northeast with a few characteristically Brazilian twists. 4. One caveat should be mentioned at the outset. Even on the topics evaluated in this report no attempt is made at a complete analysis of the appropriate reform of public expenditures. Such an analysis would have to examine the effect of (and transactions costs associated with) public expenditures on two main categories of objectives: efficiency, in the form of correcting market failures in the various sectors, and equity, in the form of distributional effects of various policies. The following discussion is concerned only with the latter. All of the specific areas of expenditure that we discuss here-public infrastructure to encourage regional growth, public employment, health, education, pensions, and other transfers-all involve serious problems of efficiency as well as equity. Roads, hospitals and schools are not built solely for the purpose of poverty alleviation. And credit market problems that require public support for pensions do not hurt the poor exclusively. Therefore, the distributional impact of policies is not the sole basis for evaluating their efficacy. But it is still relevant to ask "how much do these policies benefit the poor?" and this is the focus of this report. A. Regional Growth 5. The northeast is still the poorest part of the country and poverty is still disproportionately rural. The most recent evidence on standards of living across the country is shown in figure ES- 1. The figure shows the fraction of people with total expenditures less than the indicated amount. The rural northeast is by far the most deprived. Figure ES-1. National and Regional Cumulative Distribution Functions 1 20 % 1 0 % 80% N atio n a I 60% I , - NE M etro . I, - - - -NE Urban v 1, 0 6/ / ' N E R ural - SE M etro SE U rban I' / % - SE Rural 40% '/ / 0 % 0 40 80 120 390 200 240 280 3O2 300 400 4 40 .0 520 500 ... eta0 sea ian io 0 Oa 30 880 9S 2 ar9 1000 Per capita consum ption (Reais per m onth) of Per Capita Consumption Source: Pesquisa Sobre Padr6es de Vida (PPV), Instituto Brasileiro de Geografia e Estatistica (IBGE). Long-Term Growth: Experience, Prospects and Implications for Public Expenditure 6. Notwithstanding relatively high growth in the past two years, the northeast lags behind the rest of the country and there does not appear to have been much convergence over time. As figure ES-2 shows, per capita GDP is just under 60 percent of the GDP of the rest of the country, approximately the same as in 1965. ii 80% - 60% - __ 60 - 50% -_ 40%- 30% I l l 1960 1965 1970 1975 1980 1985 1990 1995 2000 Figure ES-2. Ratio of Per Capita GDP of the Northeast Region to the Rest of Brazil Note: Data in 1995 US$. Source: 1965-95 SUDENE. 7. Therefore there is no room for complacency with regard to overall growth prospects. Over the long term, it is the growth in average incomes that will yield a sustainable reduction in regional poverty. In addition, as long as the per capita GDP gap between the northeast and the rest of Brazil persists at its current level there is likely to be continued pressure for net outmigration from the region. In conjunction with other observations in this paper it should be clear that a much bolder and more creative regional development strategy is required to close the per capita income gap between the northeast and the rest of Brazil-particularly the Southeast. The previous emphasis on creating an "autonomous center" of development in the northeast is, in principle, an appropriate approach. However, it was focused too narrowly on manufacturing expansion, and much of that in the form of capital-intensive, enclave investments. Instead the focus should be on creating more dynamic urban centers as engines of growth with a diversified economic base that is more directly linked to local comparative advantage. 8. How does public expenditure fit into this story? Using data generated by Instituto Brasileiro de Geografia e Estatistica (IBGE), Instituto de Pesquisa Economica Aplicada (IPEA) and Superintendencia do Desenvolvimento Econ6mico do Nordeste (SUDENE) a principal determinant of regional growth appears to be private investment. Public investments do not appear to effect growth directly, but only indirectly in so far as they stimulate private investment. Further, our analysis indicates that certain types of public infrastructure investments that lower local cost of production in the region (such as electric energy and water supply) are associated with increased private investment. Public infrastructure investments that increase the integration of the region with the rest of Brazil (such as transportation and communication) have mixed effects. They enable producers from iii outside the region to compete in the region just as they enable local producers to market these products in the rest of the economy. Local consumers benefit from the drop in the natural tariff barriers but sometimes local producers do not. Public investments in state enterprises as a whole do not encourage private investment. To a large extent, these results confirms a more general principle that public investment should be complementary to private investment by providing public goods or overcoming other market failures rather than substituting for private activities. 9. Any attempt to be more specific about the nature of public investments that will best encourage broad-based growth depends on the time horizon one wants to emphasize. In the long run, as the economy develops, the role of agriculture can be expected to decline both as a share of the regional output and of regional employment. Suggested strategies include the following: (1) The expansion and improvement of education are crucial. Ensuring that the goal of universal primary education becomes a reality is essential not only to reach the poorest with a valuable public service but also to facilitate the transition away from agriculture. Currently, there are large numbers of migrants from the rural northeast in other parts of the country. Their level of education is higher, on average, than people remaining in the rural areas. This is evidence that greater mobility and flexibility in the labor market is achieved by education. Furthermore, analysis in this report corroborates evidence in many studies that earnings are considerably higher for the better educated. The one exception to this rule is thait people operating their own farms do not appear to have higher incomes as a result of more education. Everyone else, however, benefits from greater opportunities and higher productivity. An important policy implication of higher earning capacity from education resulting from improved opportunities and mobility is that funding for education needs to have a substantial federal component. The results point to a possible interregional externality, i.e., that the returns to state investments in education may well accrue to other states. The country as a whole has an interest in maintaining educational levels in the region. (2) Infrastructure that supports the creation of off-farm employment opportunities will be necessary. The balance of investment may turn toward road and communication links and away from irrigation and other investments directed toward agriculture. (3) Infrastructure requirements for guiding urban development are also of high priority as the decline in agriculture leads to a further urbanization ofpoverty. One important aspect is the provision of safe water and sanitation. Analysis presented below shows enornous positive effects of access to these services on health, access that is currently denied only to the poorest. This is of highest priority given that it is relatively easier and cheaper to supply these services in urban areas. 10. On the other hand, in the short run agriculture will remain a prominent feature of the economy. Furthermore, even when the economy as a whole shifts away from agriculture, there will still be concentrations of the poor among small-landholders and the landless working in what remains of the sector. In this context, the policy options for alleviating poverty are more difficult. Suggested strategies include the following: (1) Public investments in support of agriculture are natural candidates. Continuing support for water development projects, rural roads to improve marketing opportunities, and the completion of rural electrification are all likely to be important. However, the impact on poverty from the development of agriculture is not clear. Increases in agricultural output that result from improved iv yields and productivity will only improve the condition of the poor to the extent that they increase wages for unskilled people or increase the profitability of land owned by the poor. The impact of public investments that increase productivity in agriculture on labor demand and, therefore, wages is not clear. Technical progress in agriculture varies in its effects on labor demand but does not usually increase demand for unskilled labor. So, while there is a role for public investment on efficiency grounds, the effect on poverty is more dependent on the ownership of improved land by the poor rather than employment. (2) Land reform becomes impossible to avoid. If labor demand cannot be expected to grow substantially, the only way that poor people who remain in agriculture can increase their income is to own more productive assets. As argued above, technical progress in the sector will only help the poor if they benefit from the extra profitability of farming through ownership of land. And in the absence of technical progress, more equitable ownership of land is even more critical. Whether this requires much extra public expenditure is unclear. Previous policies on land reform have wasted substantial sums of money purchasing or rather expropriating low-quality land. More market-oriented approaches will place the government in the role of facilitator rather than purchaser of land. (3) Education remains important but will require changes in emphasis. As noted above, the analysis in this report casts some doubt on the effectiveness of current forms of education on the productivity of farmers. One consequence of improved technology in agriculture, however, will be a need for better farm management practices and, therefore, of more educated farmers. There will be a need to reorient education to satisfy this need. (4) The greater and more varied demands of infrastructure, requiring knowledge of local conditions, suggests one particular form of expenditure. Social funds relying on community support and participation have been a successful contributor to rural society and economies. Ceara's Sao Jose program and similar schemes in other states (sometimes supported by the World Bank's PAPP program) have been able to mobilize community resources, identify investments which respond to felt local needs and have brought about substantial changes in the way communities see the future.' This conclusion is not based solely on impressions but has some support from research. At least in the area of rural water projects, community participation has been shown to be a critical factor in their success. (Narayan and Pritchett 1996). Investments requiring cooperation among beneficiaries for their operation or routine maintenance appear to gain most from the community's participation in their selection, planning, and construction. Water projects fit this description, as do construction of schools and clinics. Ceara's experience with health services is further evidence. 11. Some types of public expenditures are required regardless of the time horizon. This includes infrastructure needs common to agriculture and nonagricultural activities, as well as education (including the reorientation of the curriculum for farmers). However, how much public resources should be directed to "agriculture-specific" investments requires a better fix on the likely role agriculture will play in the future. Unfortunately, this is hard to predict ' Two comments heard on a visit to a Sao Jose village are instructive. One, by someone originally from rural Ceara but unfamiliar with the program, was that the village members seemed to have a clear set of priorities for future investments, knew what the next steps were and how to go about doing them. This contrasted starkly with traditional characterizations of the rural poor as fatalistic and passive. She thinks this is a genuine change from the program (and not a shortcoming of the traditional characterization). The second, by someone much more familiar with the program, was that there have been second-round social effects of the program. It seems that a common sight in the halls of state government are representatives from Sao Jose villages demanding, for example, that the schools they built with Sao Jose funds be staffed, as is their right, by Department of Education teachers. Such demands are also common in public works, health and other agencies. This public monitoring of government, too, was unheard of before the introduction of the program. v based on readily available macro data. Figure ES-3 shows the composition of output since 1965. 80% 70% __ _ ~~~~~~~~~~~~~~High inflation_ 60% - Services __.___ 50% - >< _______ E 40% - ~~~Industry_ _ 30% 20%- AgricuIture 0% N 00 0 0 o g C C g - I - ° x - - - - c r- - - - -g - - - Figure ES-3. Structural Composition of the Northeast Economy Note: Nominal data used for calculating shares. Source: 1965-95 SUDENE; 1996 World Bank Mission. 12. While there had been a secular decline from 1965 to the early 1980s, this trend seems to have stopped since the onset of the original debt crisis. In the absence of any productivity growth in agriculture in the past 20-25 years the share of agriculture in total employment followed the pattern of agricultural output in total GDP. Whether the declining trend continues with the resumption of stability or stays at its current levels remains to be seen. Any improvements in agricultural productivity will cause employment trends to lay behind output trends. Macroeconomic StabUity 13. There are places in the world where the issues of long-run growth can be discussed separately from the issue of short-run macroeconomic stabilization. Brazil is not one of them. Prospects for growth in the country depend on maintaining the price stability gained from the reforms begun in 1994. 14. Beyond its contribution to long-run growth, however, price stability has a more direct and immediate impact on the poor. A substantial amount of research, much of it originating from IPEA- Rio de Janeiro, has identified a close link between inflation and unemployment on the one hand and poverty rates on the other. This relation is as true in the northeast as vi anywhere else in Brazil. High inflationary periods boosted poverty rates and the recent reform program brought them down again. 15. While the control of the macroeconomy is primarily the responsibility of the federal government, the states can contribute to the effort by maintaining political support for national stabilization efforts and by reducing their own deficits. States can help their poor by maintaining fiscal balance since they are more likely to have the freedom to engage in "pro- active" policies than they would if fiscal discipline is enforced by, say, current treasury rules. 16. The analysis in this report indicates two areas in particular where deficit reduction is possible and desirable. The first is the reduction of administrative costs. There is a close relation between the share of state budgets going to administrative costs and the overall level of deficits. States that have able to keep their administrative costs in check have been more successful in maintaining low deficits. 17. Streamlining public employment (though related to administrative costs) is the other main area in which expenditures can be reduced. While employment opportunities in public service have been justified as an "antipoverty" device, their effectiveness in this regard is very doubtful. The difference between earnings in the public and private sector is likely to be small, given the characteristics of people employed in the public sector, such as their education and experience. Furthermore, those employed in the public sector do not share characteristics of the poor. Finally, the share of family income (not simply a single wage- earner's income) that a wage differential between public and private employment represents is also quite small. In combination, public employment contributes very little to poverty alleviation. If public employment is to be used to combat poverty, it should be used as a guaranteed employment scheme to protect the poorest. Such a scheme can also help in protecting the poor in times of drought. Regular public employment should be thought of as necessary to produce services for the poor rather than to produce jobs. B. Expanding Social Services 18. The poverty assessment done jointly between the Bank and Brazilian researchers and policymakers in 1996 identified a major gap in knowledge concerning the use of public services by the poor. It recommended using a multipurpose survey that could fill this gap. As a result, IBGE undertook a new survey, the Pesquisa sobre Padroes de Vida (PPV) similar to those done by the World Bank in other countries. This data is relied upon heavily in this report. Health 19. There are two questions concerning the effect of public health spending on the poor. The first is, how much of the health budget reaches the poor? The other is what kinds of spending will most improve the health of the poor? 20. How effectively the health budget transfers income to the poor depends to a large extent on the level of government paying for the services. The users of public health services vii in the northeast are disproportionately poor by national standards. People in the northeast who are among the poorest 30 percent of Brazil as a whole make up 39 percent of visitors to public facilities in the northeast in contrast to only 22 percent in the south. If payments for health services come from the average taxpayer nationwide, this represents a significant transfer from the relatively well-off to the poor. 21. Within the region, however, the scope for redistribution to the very poor is more limited It is the relatively well off within the region that disproportionately use services. Whereas the users of health services in the northeast are relatively poor in comparison to the nation as a whole, they are not relatively poor in comparison to other residents of the northeast. The poorest 30 percent of people living in the rural areas of the northeast account for only 14 percent of visits to public facilities. The poorest 30 percent of people living in the region as a whole account for 25 percent of visits. So, if health services are supported by the average taxpayer from a state in the northeast (and even more so from those at more local levels of government) the degree of transfer is very limited. 22. What kind of spending most improves the health of the poor? On this question there are some very clear answers. The results of four separate studies confirm that the main determinant of various measures of health status are income, parents' education, and differences in the quality of water and sanitation facilities. When measured, the quality of or access to medical care are not found to contribute to improved health. While we would hope that income and educational opportunities will improve for the poor, the extension of cleaner living conditions either by public investment in water and sanitation or by focused education on hygiene can be done immediately and can be expected to have an effect in the short term. 23. Figures ES-4, ES-5 and ES-6 illustrate both that access to water and sanitation are extremely effective in improving child survival and that extension of access to safe water and sanitation is extremely well targeted to the poor. . - ,4 -tI1F 2- 0 Figure ES-4. Lack of Access to Water and Sanitation Increases Child Mortality Source: (PNDS), and calculations based on appendix 6. viii 80- 1 2 3 4 5 6 7 8 9 10 Consumption deciles Figure ES-S. The Poor Have Less Access to Safe Water Source: PNDS. 70- ~60/ 30 - .20 1 2 3 4 5 6 7 8 9 10 Consumption deciles Figure ES-6. The Poor Have Less Access to Sanitation Source: PNDS. 24. More generally, traditional public health interventions for combating communlicable disease are also highly pro-poor. These kinds of interventions: health education, promotion of immunization, disease-bearing pest controls, are also desirable on pure efficiency grounds since they address clear externalities and are sometimes pure public goods. There is a great deal of complementarity between equity and efficiency in the use of these interventions. 70-~~~~~~~~~i 25. Does the absence of a health effect and the question of the degree of targeting to the poor with state funds reduce the value of medical services provided in rural areas? This is too hard a question to deal with on the basis of distributional effects alone. However, a further important benefit of subsidized medical care is the amelioration of anxiety over potential financial losses from costly illness. This is the main reason for health insurance-a commodity not common in the rural areas. Estimates based on costs relative to incomes and the probability of disease lead to estimates of the value in welfare terms of between 2 and 30 percent of the costs of supplying services to the poor. While this range is too wide to be used as a basis for policy, the high end of the range argues for serious consideration of further work to identify a benefit of policy that is usually ignored in analysis. Education 26. The problems of the education system in the northeast are well documented. A recent report done cooperatively by the Ministry of Education/Northeast Basic Education Project, World Bank, and United Nations International Children's Emergency Fund (lUNICEF) covers the educational issues in great detail. The problems of the sector go well beyond failure to serve the poor and would require fixing regardless of their distributional impact. 27. From the perspective of addressing poverty, however, two questions very similar to those in health apply. Does the education budget reach the poor? And will improvements in education make a difference in the earnings prospects of the poor? 28. From both the national and regional perspectives, current enrollments are highly equitable. Over 50 percent of the children in classrooms in the northeast are below the poverty line. Even from the region's perspective, over 30 percent of school children are from the poorest 30 percent of children in the region. So the educational system is potentially an effective means of at least reaching the poor. Enrollment is not the same as expenditure, however, and it is hard to assess whether resources spent also reach the poor. 29. As far as the second question is concerned: as mentioned in the discussion of the future of agriculture, with the sole possible exception of people working their own farm, education makes a major contribution to earnings capacity. No strategy for alleviating poverty will succeed without expanding and improving education. However, due to the long lag in effect and the still significant numbers of the non-poor who have yet to attend school and may well be among the next set of children to enroll, education should not be the sole instrument, the "magic bullet" for poverty alleviation. 30. That there is progress to be made in education is clear. The attendance pattern of children in the rest of Brazil looks very much like that of Latin America. Enrollments are virtually universal to grade 4 even arnong the poor. Within the northeast, however, the enrollment rate looks like countries at much lower levels of income. Children start school late and leave before finishing lower primary grades. 31. What would bring more children to school? The Ministry of Education, World Bank, and UNICEF study pointed to quality improvements. Further evidence from the PPV indicates that proximity to schools also raises attendance so there may be difficult quality versus quantity problems to address in future. The evidence presented here also points to x other determinants that relate to children's "opportunity cost" in terms of work as well as "out-of-pocket" costs associated with travel and materials. The responsiveness to monetary barriers indicate, however, that experiments in Brasilia with direct payments to students for school attendance may well be replicable in the northeast as both the "opportunity cost" of children's time as well as the "out-of-pocket" costs of schooling contribute to low enrollment rates. A payment that compensates for a wage of 1 Real per day is estimated to increase school attendance by as much as 20 percent. In the long run, expansion of enrollments will have a self-sustaining effect since parental education level is a major determinant of attendance, an effect that is discernible at low levels of education of the parents. C. Targeting Transfers With a wide variety of specific transfer programs in the country, we do not pretend to be able to address them all (or any of them in great detail). There are though two kinds of transfer programs that are of particular concern here: Pensions A common perception of the impact of preferential treatment of rural residents of the northeast in the allocation of pensions is that the pension system results in a large inflow of money to the region. This is simply not true. The germ of truth behind it is that for each income group, measured in deciles of per capita consumption, pensions constitute a somewhat higher fraction of income in the northeast than they do in other parts of the country (see figures ES-7, ES-8, ES-9). Figure ES-7. Percent of Income from Pensions 30 0NE-rural _ 25 _ 20 SE-rual 1 15 10 5 1 2 3 4 5 6 7 8 9 10 Consumption dedles Source: PPV. xi 32. However, the overall per capita transfer to the region is considerably smaller than to the southeast for two reasons: (1) there are many more people in the poorest three deciles of the national income distribution in the rural northeast (these people receive less, per capita, than those in the middle class); and (2) there is a much lower percentage of people of retirement age in the region. 33. However, for those in the rural northeast who do receive pensions, there is a significant effect in protecting people from falling into poverty, a much stronger effect than elsewhere in the country. Many people are kept out of the bottom income groups due to the receipt of this money since there is a fair amount of clustering of the recipients just above the poverty line when full, pension inclusive, income is calculated. 180- 0 _ Northeast Source: Authors' calculations based on PPV. xii 2. FIGURE ES-9. NUMBER OF HOUSEHOLDS BY CONSUMPTION DECILE, NORTHEAST 160 140 0 Total Expenditure 0_ Expenditure minus transfers V120 ioo 80 6 100 40 20 r t 1 2 3 4 5 6 7 8 9 10 Consumption deciles RURAL Source: PPV, and authors' calculations. 34. This result is true for a wide variety of assumptions concerning the incidence of taxes used to finance the program. In sum, the pension program is not a major transfer of money to the northeast but what money is involved has a substantial impact on poverty. Other transfer programs are much smaller as are private remittances from family members living elsewhere (a point which also merits more careful study. See discussion of education). 35. It is important to note that calculations of the financial flows implicit in the pension program (as would be true, one would speculate, with all straight money transfers) have virtually all of their aggregate effect as a result of the receipts of the richest two or three deciles of the population. This reflects the extremely large disparities of income across social classes. Thus the amounts of money paid to the poor are tiny compared to the overall cost, but are very large relative to their own incomes. This indicates that a wide variety of possible effects on poverty are allowed for by any particular aggregate cost of such programs. Changes in design, which would scarcely show up in aggregate analyses, can be very important in their impact on poverty. This is true for transfers within the northeast as well as interregionally. Government's Role in Ameliorating Risk 36. Several of the policies mentioned in this report have a common characteristic: drought relief, other means of cushioning the poor in bad times, such as unemployment insurance temporary transfers, and health care in the absence of health insurance and all means of poverty people from risks. There are good reasons on efficiency grounds why government needs to help people handle the risks they face. Insurance markets are often quite limited for systematic reasons and there are many opportunities for public policy to improve matters. Besides efficiency concerns, it is frequently the poor who face the greatest exposure to risk. Wealthy people have diversified earning opportunities, better access to credit, and are xiii the least likely to be rationed out of insurance markets. This report discusses some of the problems of this sort that poor people face but, in general, this area has not been adequately explored and more effort on the measurement of the value of risk reducing policies is needed. Unfinished Business: Drought Relief 37. A specific case is the government's role in ameliorating the consequences of drought. Despite substantial successes in avoiding famines, the cycles in production and consumption associated with drought persist. In particular, early warning and timeliness of interventions, the way in which the needy are targeted and covered, and coordination between drought and nondrought related efforts all need improvement. D. Need for Selected Further Work to Refine These Recommendations 38. Agricultural employment. The analysis of the agricultural sector is severely hampered by lack of data. There was a gap of eleven years between agricultural censuses (this is in contrast, for example, to the monthly labor surveys in metropolitan areas). New census data should make it possible to explore issues of changes in land ownership, cropping patterns, and agricultural incomes and employment. In addition, determining the weight to be given to public sector investments, particularly infrastructure (to manage water resources and support growth) in the agricultural sector will require a better assessment of the long-term competitive potential for "labor-intensive" versus "nonlabor-intensive" CrOpS in individual states as well as across the region. A similar study will be needed to determine the potential of "off-farm" employment opportunities for the poor in rural areas versus in small towns and metropolitan peripheries. 39. Educating farmers. The results of the impact of education on owner-operated farm income is based on a very small sample from the new PPV. Since the issue of education is central to that of poverty alleviation, the connection between the two needs additional careful examination. xiv Chapter 1. Introduction 1.1 This report addresses the role that public expenditures can play in the alleviation of poverty in the Brazilian northeast. The search for the solution to poverty in the northeast is as long-standing, in terms of years of concern and pages expended on analysis, as any topic of economic and social policy in Brazil. It should come as no surprise, then, that this paper will not be able to suggest a simple formula, a "magic bullet," that somehow escaped the attention of the contributors to that literature and policy debate. It does try to make two kinds of contributions. First, it will link public expenditures with an overall approach to poverty alleviation involving both regional growth and social services. Second, it brings to bear several new or newly compiled sources of data on the question. 1.2 The policy discussion is primarily from the perspective of policymakers in the region rather than that of the federal government. While some policies needed to combat poverty are primarily the responsibility of central government and will be discussed here, the focus will be on those aspects of public expenditure which are under the control of the state governments. Ultimately it is at this level that expenditure policies are implemented and at this level that the poverty orientation of government can be made tangible. 1.3 A useful framework, commonly used in the World Bank for discussing policies to alleviate poverty, stands on two-and-a-half legs (World Bank 1990). The first is to encourage broad-based growth, usually understood as labor intensive growth. The second is to ensure that social services reach the poor-usually by guaranteeing that they are universally available. And the half-leg is to use (sparingly) carefully targeted transfer programs to address special needs of particular subpopulations. This broad framework can be profitably applied to the role that public expenditures can play in the context of the Brazilian northeast with a few characteristically Brazilian twists. 1.4 After a brief discussion of recent trends and new information concerning poverty in the northeast, we discuss the growth experience of the region and changes in the structure of the economy. In the context of northeastern Brazil, the type of growth we are talking about is that of broad-based growth. It is not really possible to identify policies that would encourage labor-intensive growth particularly, nor should state governments try to fine tune growth in this way. High rates of poverty in the rural northeast mean that broad-based growth prospects require attention to the gains that those currently working in agriculture can expect from it. Appropriate expenditure policies for poverty alleviation depend on assessments of the future composition of the economy, particularly the role of agriculture. 1.5 Given the macroeconomic experience of Brazil, it is difficult to discuss long-term growth prospects without addressing issues arising from short-term stabilization efforts. In this regard, we next present evidence on the connection between the recent reform period and poverty in the region. While primarily a responsibility of the federal government, we will argue that the overall budget deficits of the states can contribute to the national effort. In this connection, we show that control over administrative costs are important in deficit reduction and the determination of the overall budget envelope. We also assess the direct risks that reductions in public employment might pose for poverty alleviation. 1.6 The subsequent section discusses public expenditures on social services, focusing particularly on health and education. The poverty assessment done jointly between the Bank and Brazilian researchers and policymakers in 1996 identified a major gap in knowledge concerning the use of public services by the poor. It recommended using a multipurpose survey that could fill this gap. As a result, IBGE undertook a new survey, the Pesquisa sobre Padroes de Vida (PPV) similar to those done by the World Bank in other countries. This data is relied upon heavily in this report. 1 1.7 As for the "half-leg" of antipoverty strategies, we focus on one existing scheme as well as a class of policy options that might be considered in future. The existing program is the provision of old age pensions. The last section starts with a discussion of the contribution that scheme makes to poverty in the region. The remainder of that section discusses the role the government can play in mitigating several kinds of risks that poor people, particularly in rural areas, face. The particular problem of drought is discussed in this regard. 1.8 One caveat should be mentioned at the outset. A complete treatment of the appropriate reform of expenditures in all of these areas is not possible in this paper. Such an analysis would have to examine the effect of public expenditures on two main categories of objectives: efficiency in the form of correcting market failures in the various sectors, and equity. The following discussion is concerned only with some of the distributional effects of various policies. All of the specific areas of expenditure that we discuss here: public infrastructure to encourage regional growth, public employment, health, education, pensions and other transfers, all involve serious problems of efficiency as well as equity. Roads, hospitals and schools are not built solely for the purpose of poverty alleviation. And credit market problems that require public support for pensions do not hurt the poor exclusively. Therefore, the distributional impact of policies is not the determining factor. But it is still relevant to ask "how much do these policies benefit the poor?" and this is the focus of this report. A. An Update on Poverty 1.9 Table 1 shows the percentage and number of people below the poverty line for metropolitan, urban and rural areas in the northeast and the rest of Brazil (excluding the north) for the years 1990 to 1995. The data comes from the Pesquisa Nacional Anual do Domicilios (PNAD) and is based on income. Table 1. Poverty Rates, 1990-95 1990 1993 1995 Poverty Number of Poverty Number of Poverty Number of rate poor rate poor rate poor Region (percent) (millions) (percent) (millions) (percent) (millions) Northeast 45.8 18.90 45.6 18.85 31.0 13.49 Fortaleza 41.5 .93 38.9 .92 25.7 .64 Recife 48.5 1.34 51.5 1.43 34.0 .98 Salvador 39.2 .92 44.7 1.13 34.9 .91 Urban 43.7 7.73 42.6 7.90 29.8 5.90 Rural 49.2 7.99 49.4 7.47 32.2 5.05 Rest ofthe Countrya 18.9 20.85 22.3 21.11 14.4 14.11 Metropolitan 25.3 8.68 29.1 10.26 17.2 6.23 Urban 14.9 7.93 17.5 8.03 11.8 5.65 Rural 18.3 4.24 20.4 2.82 16.3 2.24 Brazilb 30.3 41.97 30.4 43.28 20.6 30.44 Note: a. Excluding north. b. Including north. Source: PNAD calculations based on Rocha (1997). 1.10 The table shows several general points. First, poverty rates in the northeast remain consistently higher than elsewhere. Rural areas in the northeast have approximately double the poverty rate of several areas in the center-south in all three years. There seems to be a slight narrowing of the differential in 1995, but barely enough to note. Second, within the northeast there has been a shift in poverty such that rural areas have the highest rates in 1990 but the metropolitan areas of Recife and Salvador have higher rates following the macroeconomic crisis. Again, the 2 change is quite modest. Third, and most striking, is the dramatic fall in poverty rates everywhere between 1993 and 1995. 1.11 Two aspects of these raw facts are worth noting and will motivate some of the main themes of this report. First, the improvement between 1993 and 1995, of course, corresponds to the period of macroeconomic stabilization. It is clear that the reform program has been good for poor people. While poverty reduction is rightly considered a long-term, structural issue, it is clear that short-term effects can be of major importance. How macroeconomic reform affects the poor and what state governments can do about it is discussed in chapter 2. 1.12 Second, implicit in the data is a relatively large shift of people from rural to urban areas, nationwide. While the differential rates of poverty between urban and rural areas in the northeast have remained roughly equal, the total number of people in poverty has fallen much faster in rural areas than in urban. This indicates a larger base in the urban areas on which the rates are calculated and hence, migration to urban areas. Another main theme of this report is that the nature of the appropriate policy response to poverty should hinge on the assessment of the future evolution of the size, economic activities, and poverty rates in rural areas in the northeast. The difficulty is predicting that evolution as there are numerous contradictory indications. Table 1, for example, not only shows a general shift from rural to urban areas, perhaps arguing for less concern for rural poverty, but also shows a much more rapid shift in the rest of Brazil than in the northeast. For the South, the rural population fell by 11 percent over the 5-year period while in the northeast, it fell by only 3 percent. Perhaps poverty nationwide will become much more urban in the years to come, but in the northeast this may not happen as rapidly. Making the issue even murkier, areas that are classified as urban seem to be quite different between the northeast and elsewhere. In the northeast, PPV data indicate that people in the urban areas look a great deal more like people in rural areas than they do like people in metropolitan areas in terms of occupational status, education, and other characteristics. In the south and southeast, it is the opposite: people in urban areas look more like residents of the metropolitan areas than they do of rural residents. Maybe the small cities of the northeast are really rural by southern standards. If we look at the population of non-metropolitan areas of the northeast, we find it to have grown by 4.7 percent over the five years. So, is the northeastern population that relies on agriculture for its livelihood, either directly or indirectly, shrinking or growing? Or is it just dispersing spatially? It should be possible to answer these questions once the recently completed agricultural census data is analyzed. 1.13 Poverty in the PNAD data is defined in terms of income. A criticism that is frequently made of income as an indicator of welfare is that it can be very variable between years, especially in agricultural households. Consumption or total expenditure levels will be more indicative of long-term prospects as judged by the household. These measures incorporate the effect that borrowing or saving can have in smoothing erratic income flows that will, in general, improve the well-being for the family. Because of these considerations, the PPV survey relies more heavily on calculations of total expenditure than on income. Since we are going to use both kinds of data in the following analyses, it is important to look at how consistent the two measures are in the picture of poverty they paint. 1.14 Figure 1 shows the cumulative distribution functions for consumption from the PPV data. The data have been adjusted for cost of living differentials across the country and so should reflect equivalent standards of living. Poverty rates for each region could be determined by choosing a poverty line on the consumption axis (common to all regions due to the cost of living adjustment) and finding the level of the curve at that point. 3 Figure 1. National and Regional Cumulative Distribution Functions of Per Capita Consumption 120% 100% .7~~~~~~~~~~~~~~~~~ National : 60% 1. -/ - ---NE Metro ---- /NE Urban NE Rural - SEMetro SE Urban - SE Rural [i / 0% 0 40 80 120 160 200 240 280 320 380 400 440 480 520 560 600 640 680 720 760 800 840 850 920 960 1000 Per capita consumption (Reais per month) Source: PPV. 1.15 It is not possible to define a poverty line that would be directly comparable to that used by Instituto de Pesquisa Econ6mico-Social Aplicada (IPEA) on the PNAD surveys since the PPV uses consumption and the PNAD surveys use income as the measure of household standards of living. In spite of this limitation, several differences between table 1 and figure 1 are noteworthy. First of all, the distribution function for the rural northeast lies everywhere to the left of the others (with the exception of a crossing or two with southeastern rural areas at higher levels of consumption). For any poverty line that would define fewer than 85 percent of the residents of the rural northeast as poor, the poverty rates in this region are higher than in any other region by a considerable margin. This 4 contrasts with the PNAD study for 1995 that would have poverty rates in the metropolitan northeast higher than rural. In the PPV, poverty in both rural and urban areas of the northeast is higher than anywhere else for all plausible levels of the poverty line (i.e., all lines yielding poverty rates less than 60 percent since that is where the rural south line crosses the northeastern urban line). And except for the rural south, the differences are quite large. Since there is no PNAD study for 1997 available, it is not possible to tell if this is a change over the two years or the difference in measures. One would suspect a difference in measures except that the equivalent figure for income defined in the PPV is consistent with uniformly higher poverty rates for the rural northeast. Income and consumption in the PPV show considerably higher correlation with each other than is typical of such measures and differ from each other in a way that is expected from theory. The variance of the consumption measure is lower than that of income reflecting the smoothing effect of borrowing and saving. 1.16 If the argument that the urban northeast is similar to rural areas is accepted, figure I indicates that all rural areas are much poorer than (i.e., have distribution functions that lie well to the left ot) any of the truly urban areas. Again, there is little consistency across measures on the relative poverty of urban and rural areas. 1.17 A striking, if not unexpected, feature of figure 1 is the extraordinary spread between incomes across the country. The median level of consumption in metropolitan southeast is two-and-a-half times the median of the rural northeast. For aggregate groupings of this size-the rural northeast accounts for over 10 percent of the national population-this is a very large difference. Put another way, a person at the thirtieth percentile in the metropolitan southeast consumes more than 80 percent of the people in the rural northeast. This differential shows up in various guises throughout the report. An implication of this disparity is that small percentage changes in incomes of a relatively small fraction of the rich translate into large changes in the incomes of a fairly large proportion of the poor. That is, tax or subsidy changes with small proportional impacts on relatively well-off people generate revenue changes that are very large relative to the entire income of the poor. There is great scope for redistribution with a very small degree of sacrifice among the well-off segments of society. B. Land Distribution and Rural Poverty 1.18 While there is substantial potential for income transfers, it may also be important to take a second look at the distribution of rural assets. Table 2 shows the distribution of land ownership by consumption groups in the PPV survey. Because only families with farning as their main occupation are in the sample, groups were defined to have enough observations for making legitimate inferences. Table 2. Land Distribution by Consumption Group and Consumption Deciles Consumption decile Land owned Poorest 2 3 4 and 5 6 and higher Northeast <0.5 Ha 77% 68% 43% 43% 34% >0.5 Ha 23% 32% 57% 57% 66% Number of obs. 74 65 42 96 68 Southeast <0.5 Ha 71% 58% 42% 40% 38% >0.5 Ha 29% 42% 58% 60% 62% Number of obs. 24 24 26 50 82 1.19 The table reinforces the notion that the pattern of land ownership is extremely skewed throughout the country and slightly more in the northeast. More than three-quarters of the farmers 5 who are in the poorest decile nationwide farm less than one-half of a hectare. While being a part of the poorest decile of the whole population, farmers in this group make up over one-fifth of the northeast farm families. At the other end (actually the other half) of the income distribution, two- thirds of the people in the top one-half of the national income distribution farm more than one-half a hectare. Though the sample becomes too small to make precise inferences, no family in the top quintile farms less than one-half a hectare in the sample. 1.20 How this pattern links with the overall land distribution needs more careful matching with agricultural census data. However, it is clear that farmers relying on plots this small for their income cannot hope to escape poverty. Since it is also the case that land quality is lower in the northeast, many poor farms are relatively large. This effect magnifies the extent of deprivation of northeast farmers suggested by these figures. 1.21 The next chapter looks at the evolution of income and the structure of the NE economy in more detail. We are looking for systematic changes in the NE economy that both (i) have implications for the level and type of poverty we might expect in future and (ii) that can be influenced by public expenditure policies. 6 Chapter 2. Broad-Based Regional Growth and Poverty Reduction 2.1 Until the latter part of the nineteenth century the northeast of Brazil-especially the areas around Recife and Salvador-were among the most developed in Brazil. The economic base was centered on sugarcane. The twentieth century saw a major shift in economic activity and prosperity to the southeast of the country. Even though Brazil's national development plans did not explicitly contain spatial/regional goals in the 1940s and the 1950s, sectoral investments in transportation, energy, and industrial development strongly favored the development of the southeast. These programs, coupled with recurrent droughts, contributed to increasing disparities between the northeast and the rest of Brazil and a substantial population migration from the northeast to the southeast. It is estimated that today as many as 4 million residents of Greater Sao Paulo are Nordestinos. The growing disparities coupled with a severe drought in 1958 stimulated the Government of Brazil into formulating explicit policies for the northeast (Baer 1995). 2.2 The main regional development strategy was to establish an autonomous center of manufacturing expansion in the northeast by attracting "dynamic" and high-growth industries (World Bank 1987), such as those in metallurgy, machinery, electrical equipment and paper products. Instruments like fiscal incentives, transfers, and direct expenditures in the form of industrial land and infrastructure were widely used to facilitate regional growth in the northeast (Goldsmith and Wilson 1991, Markusen 1994, World Bank 1987 and 1996a). 2.3 SUDENE was formed in 1959 to coordinate and oversee this effort (in part to stem the flood of migration to the south). More explicit regional development objectives, policies, strategies, and programs were developed in the mid- to late 1960s. A central regional development objective was to reduce the inequality of income between residents of the northeast/drought polygon and their counterparts in the center-south (Government of Brazil 1968). Some development objectives were in the form of specific targets, e.g., to increase the regional share of GDP to 26 percent by the year 2000, or to reduce the rate of unemployment and underemployment to 5 percent by the year 2000 (World Bank 1987). 2.4 This section briefly summarizes how the economy has actually performed relative to the explicit objectives of regional policy in the northeast and the factors that might explain the results. A. Trends in Output Growth and Volatility 2.5 The two primary concerns that were to be addressed by a regional development strategy were (i) a reduction in the gap between per capita incomes in the northeast and the rest of Brazil, and (ii) a reduction in the volatility of the economy in the northeast. Has there been a convergence in per capita incomes between the northeast and the rest of Brazil? 2.6 Many studies have been done regarding the convergence of the regional economy in Brazil (Azzani 1995, Schwartzman 1996). The careful recent studies confirm the following analysis and show that the large gap in per capita incomes between the northeast and the rest of the nation has not been reduced. Per capita incomes in the northeast were only 56 percent of the nation in 1965. Figure 2 indicates that this ratio actually worsened in the 1970s when the rest of Brazil was booming. Some of the lost ground was recovered in the 1980s, but even as recently as 1995, the ratio of per capita income in the northeast (approximately 53 percent) had not regained its 1965 level relative to that of the nation. In the absence of a good analysis of the regional dynamics of the Brazilian economy (which is beyond the scope of this study) it is not possible to attribute the failure to attain this stated 7 goal as an indictment of past regional development programs since it is possible that without them the situation in the northeast may have even worsened further. Figure 2. Ratio of Per Capita GDP of the Northeast to Brazil 80% - 70% - _~~ ~~~~_ __ - __ _ ___ .___ ____ 70% - _ _ _ 60% - 50% - _ _ _ _ _ _ __ 40% - __ --- 30% i l l 1960 1965 1970 1975 1980 1985 1990 1995 2000 Note: Data in 1995 US$. Source: 1965-95 SUDENE. 2.7 Comparing the northeast with the rest of Brazil (inclusive of the northeast) does not really address the extent of regional inequality (see figure 3). Despite decades of regional policy, the northeast continues to lag the prosperous "triangle" of industrializing and exporting regions in the southeast (Sao Paulo, Rio de Janeiro, and Belo Horizonte). Per capita income differences between these regional groupings (for which statistical data is compiled) is relatively large and surprisingly stable over very long periods. Per capita income in the Southeast was 2.9 times that of the northeast in 1939 and 2.8 times in 1992 (see appendix 3a, table I for details). The scale of the spatial units that are compared can mask the extent of per capita income disparities across space. At a smaller spatial scale regional inequalities (i.e., the per capita income differences) are much more pronounced-with per capita incomes in Sao Paulo (the wealthiest Southeastern state) 7.2 times that of Piaui (the poorest northeastern state). Regional inequalities in per capita income in Brazil are way out of line with regional inequalities in other parts of the world-whether the countries are poorer or richer than Brazil on average. For example, per capita income in the prosperous coastal region of China was 1.8 times that of the poorer Western region in 1980, with some deterioration to 1.9 times in 1990 following liberalization in the economy (Yang and Wei 1996). In 1997 the per capita income in the richest state in the United States was roughly double that of the poorest state (Connecticut at $36,263 vs. Mississippi at $18,272, Bureau of Economic Analysis, U.S. Dept. of Commerce). 8 Figure 3. Regional Variations in Per Capita Income 4500 4000 --+Notheast -U--Southeast 3500 - -Brazfil X3000-__/ 4g 2500- __ __ ____ ____ __ _ ______ __ _______ 250 200 c 2000 - ~--------- -- _--- _ - -___ ___ __ __________ 1500 _ _ _ _ _ 1000 1960 1965 1970 1975 1980 1985 1986 1987 1988 1989 1990 1991 1992 Source: IBGE; Fundacao Getulio Vargas (FGV); data series adapted from Azzoni (1996). Has the volatility of the economy in the northeast been reduced? 2.8 In addition to the persistent large gap in incomes, the northeastern GDP has been very volatile. Both overall gross regional product and per capita gross regional product in the northeast has been more volatile relative to that of the Brazilian economy as a whole (see figure 4 and the high standard deviations in table 2 and 3, appendix 3a). However, this volatility has decreased over time as evident in the recent decline in the standard deviations shown in those tables. A further characteristic is that its growth pattern is more closely correlates with the rest of the nation. Thus, while the volatility in the northeast economy has decreased over time, it is still more volatile than the national economy whichever definition of volatility one wants to use. Figure 4. Growth Rates of Per Capita Output 25% 20% P re-hyperinflation Hyperinflati 5% Northeast 4 15% Brazi% -lo% =U5% 0% -5% -10% Note: Data in 1995 US$. Source: SUDENE (1996), and Boletim Conjuntural, pp. 363-73. 9 What explains this drop in volatility? Does the drop in volatility suggest that despite the lack of convergence the previous regional development strategy may have chalked up some successes? There are three plausible explanations for the decline in volatility: (1) It is an artifact of the data. The hyperinflationary period introduced so much noise into the relative price data that price deflators for the northeast are not clearly distinguishable from those of the nation as a whole. (2) It reflects a reduction in the volatility of agriculture, the most volatility component of the economy, due to improved infrastructure investments (e.g., water storage and distribution to mitigate the effects of drought, or transport and communications to better integrate and deliver farm output to markets). (3) It reflects a structural change in the economy. The economy is now less dependent on a volatile agriculture sector. 2.9 Analysis in appendix la shows that the apparent decline in the volatility of aggregate output cannot be attributed either to statistical aberrations or to a reduction in the volatility of agriculture. Rather, the decline in the volatility of regional output is a direct result of the drop in the share of agriculture in regional GDP. Figure 5 illustrates this point. However, since the northeast economy is not fully drought resistant and the agriculture sector is highly susceptible to climatic variations, the problem of high volatility in agriculture is still a source of concern from the prospective of poverty alleviation-as 40 percent of the northeastern labor force is employed in this sector and the rural poor remain heavily dependent on farm incomes. Figure 5. Structural Composition of the Northeast Economy 7 0 70t/. _ .......... _ ^~~~~~~~~~~~~. H y p ein fl a tio n 6 0 __ _____ S e rv i ce s z 2 0 . rn I n~~1 d u st-r _ A ig , i V f Note: Nominal data used for calculating shares. Source: 1965-95 SUDENE, 1996 World Bank Mission. 10 2.10 As noted above the share of agriculture in northeast output has decreased over time. This decrease is neither uniform across states nor over time. Equally important is the fact that over the long term-1960 to 1990-the share of agricultural employment has declined from 69 percent of total employment to 38 percent. The secular decline in agricultural employment has been absorbed by the services sector (see table 3 below). Most of this has been in urban areas. The continued volatility in agricultural output has employment effects, mostly in the form of seasonal (or drought-cycle related) shifts to service sector and construction activities in the urban periphery. However, the most important feature of the agricultural sector in the northeast is the low productivity of labor. In 1990, agriculture employed 38 percent of the northeastern workforce while contributing only 13.3 percent of aggregate output. What is more problematic for the future is that labor productivity in the sector appears to have been stagnant over the past 30 years. The output-labor ratio for agriculture has even declined slightly from 942.8 Reais/worker in 1960 to 883.5 Reais/worker in 1990. Labor productivity in the services sector to which most freed-up agricultural labor migrates has also remained stagnant- though at a level more than three times that of agriculture. In comparison, overall labor productivity over the 30-year period has increased from 1,588 Reais/worker in 1960 to 2,512 Reais/worker in 1990-primarily due to a doubling of labor productivity in the industrial sector from 2,380 Reais/worker in 1960 to 4,536 Reais/worker in 1990. Table 3. Northeast Brazil: Output, Employment, and Labor Productivity Output Output Employment labor ratio Output in 1994 Share of GDP Laborforce Share of labor R$ (billions) (percent) (millions) force (percent) (R$/worker) 1960 1990 1960 1990 1960 1990 1960 1990 1960 1990 Agriculture 5.4 6.7 41.0 13.3 5.73 7.66 69.0 37.9 942.8 883.5 Industry 1.5 14.4 12.0 28.5 0.66 3.19 8.0 15.8 2380.2 4536.2 Services 6.1 29.5 47.0 58.2 1.91 9.35 23.0 46.3 3243.9 3158.2 Total 13.1 50.8 100.0 100.0 8.30 20.20 100.0 100.0 1588.9 2512.0 Source: 1960 GDP and Employment Share (BNB); 1960 Labor Force, World Bank; 1990 Employment Share, PNAD; 1990 Labor Force, IBGE; 1990 GDP and share, SUDENE. 2.11 A more thorough assessment of the future role of agriculture in the northeast is required to answer the following question: which is more likely, that agriculture's share will resume its decline of 1965-83 period or remain relatively stable at its low level in the more recent period? This question is more important for agricultural employment than it is for agricultural output. There may be some scope for increasing agricultural output either extensively (by expanding the area cultivated through irrigation), or intensively (by increasing yields through improved crops and crops rotation), or both. However, given the relatively stagnant and low productivity of labor in agriculture it is more likely that increased agricultural output will be obtained by increasing agricultural labor productivity rather than by expanding agricultural employment.1 In this case, poverty reduction in rural areas will most likely require expansion of off-farm employment activities in rural areas or the facilitation of migration to growing and dynamic urban areas. This leads to a second question: does the high covariance observed between agriculture and services in figure 5 indicate integration of rural and urban areas-calling for an integrated approach to overall poverty, both urban and rural, or will it be sufficient for a regional poverty alleviation strategy to focus on rural growth? l As noted in the previous chapter, the current distribution of land limits the potential for alleviating poverty not just among landless laborers but also independent farmers relying on very small plots. 11 B. Determinants of Growth in the Northeast 2.12 The regional effects of national program will often not be uniform as different parts of the nation have different constraints and comparative advantages. The purpose of regional analysis is to identify these subnational differences and evaluate policies and programs designed to address region- specific issues. One approach to tackling this problem is to do an analysis of the determinants of output growth in the region-specifically the impact of public and private capital formation on growth. In the case of the public sector the focus is on the role of public infrastructure investments. 2.13 There are major weaknesses in the disaggregated data series for the region and its constituent states, which preclude the analysis of complex models or the use of sophisticated estimation procedures to derive robust conclusions. These problems, as well as, the results of the more rudimentary estimation procedures used here are described in some detail in appendix 2. In this section we summarize the main results. 2.14 Private capital has had a positive effect on output growth in the northeast. By contrast, public investment did not have a direct effect on output increases. However, it is not possible to dismiss the role of public investments in increasing output growth in the northeast because it is positively related to private capital and thus has an indirect relationship with output growth. 2.15 Even though private capital is positively influenced by overall availability of public capital the effects are not uniform across types of public expenditures. Investments in state enterprises (e.g., financial services or industry) had a weak negative effect on private capital, whereas fiscal incentives and investments in infrastructure had a weak positive effect. It is not possible to determine which types of private capital were sensitive to fiscal incentives vs. investments in infrastructure. 2.16 Within public infrastructure the results were mixed. Investments in infrastructure that reduce local costs of production (whether or not market size is increased), such as electric energy and water supply, have a positive effect on private capital (they are also positively associated with output increases across the nine states). However, the data were not sufficiently disaggregated to enable us to distinguish between the effects of electric energy vis-a-vis the effects of water supply. It is also not possible to determine the effect of investment in irrigation on private capital (or overall output) as irrigation is amalgamated with all of agriculture. The latter has a weak positive effect, which is not statistically significant. 2.17 The results for other public infrastructure investments, which expand markets linking them more effectively, such as transport and communications, are indeterminate. For example, results in appendix table A2.4 show that transport and communication have a negative effect on output. This may partly be due to the short lags specified between transport investments and output changes (more appropriate lags could not be tested given data limitations). However, it is also possible that the negative association is picking up a real effect. Transport and communication investment improves linkages with the rest of the country and effectively reduces a natural tariff barrier. This may have a negative effect on regional output and private investment in the short run as transport improvements may open up the northeast markets to more efficient producers from outside the region. Larger firms serving larger markets (such as the southeast) and benefiting from economies of scale will have lower unit costs of production and can more easily expand into new markets in competition with local producers when unit costs of distribution between two points are lowered. Therefore, it is possible that in some, or many cases, producers from other regions will crowd out local producers to the benefit of local consumers but not of local production or employment. Without further analysis it is difficult to comment on the exact nature of the effect of these investments. 12 2.18 The ambiguity of the latter results do have some policy implications: in general investments in transport and communication infrastructure should be evaluated primarily in terms of their contribution to increasing the national efficiency of production and distribution, not region-specific improvements. However, with additional analysis of disaggregated data (for example, the impact of specific projects, such as regional airports and sea ports) it should be possible to identify a subset of transport and communication infrastructure investments whose benefits are more region specific but realized with a substantial lag and only in the context of a package of complementary policies. C. State Growth Rates and Poverty Reduction 2.19 The previous section pointed out that the nature of growth and the composition of the economy vary substantially. It is also the case that growth will have quite different effects on poverty depending on this variation and on other distributionally sensitive characteristics of economic change. An analysis of state level data was made by looking at annual (or intersurvey) changes in the headcount indices of poverty from PNAD data for the northeast states from 1981 to 1995 and relates these figures to income growth rates and changes in the Gini coefficient measuring the dispersion of the income distribution within the state. The results of this analysis shows that both variables, growth in state income and Gini coefficient, are significant and that characteristics of the income distribution, at least, are an important factor in explaining poverty rates in the region. They indicate that a one percentage point change in a state's growth rate can be expected to reduce poverty by .067 percentage points. More powerful, however, is the effect of the distribution of income for a given mean income. A 1 percent increase in the Gini coefficient (reflecting more inequality) will increase poverty by 1.14 percent and tracks this quite closely. See appendix 3, table 1. Therefore, all growth is not created equal and those growth strategies with better distributional characteristics will have substantially better success at poverty alleviation. These results are consistent with those found by Ravallion and Datt (1992) looking at aggregate data for the country as a whole throughout the 1980s. 2.20 This does not imply a necessary tradeoff between growth and poverty alleviation. There does not appear in this data to be any connection between the rate of growth of the state economy and the change in its Gini coefficient. This represents one more entry into the long-standing debate in Brazil concerning the relation, or lack thereof, between inequality and growth. Our results reinforce those of Li, Squire and Zou (1998) which show no correlation between income growth and distribution in a cross section of countries. Long-run growth is important to poverty alleviation. Its net impact depends on how it affects the distribution of income. The evidence suggests that an improved distribution of income does not imply a reduction in growth. D. The Importance of Macro and Fiscal Policies for Poverty Reduction in the Northeast The Macroeconomy and Poverty in the northeast 2.21 There are places in the world where the issues of long-run growth and its benefits can be discussed separately from the issue of short-run, macroeconomic stabilization. Brazil is not one of them. Prospects for growth in the country are tied to maintaining the price stability gained from the reforms begun in 1994. Besides enabling the country to get on a growth path of any sort, the maintenance of a stable macroeconomy has a more direct and immediate impact on the poor. 2.22 This section documents the connection between poverty rates and both short-term inflation and unemployment rates. It argues that state fiscal performance can contribute to national efforts in 13 stabilization by controlling state deficits and provides evidence that the deficits can in fact be controlled. 2.23 Inflation, unemployment, and poverty in the northeast. There is a long literature (Paes de Barros, et al. 1998, and Ferreira and Litchfield 1997) demonstrating the connection between inflation, unemployment and poverty rates in Brazil as a whole. Both inflation and unemployment increase poverty. The effect of unemployment on the headcount-ratio the fraction of people with income below the poverty line-is clear; some fraction of wage earners near the poverty line are at risk of falling below it with the loss of wages. 2.24 The effect of inflation is not clear from a theoretical viewpoint. It is the real wage that determines the bulk of income of the poor and new poor. Inflation does not necessarily change real wages. However during periods of rapid inflation nominal wage adjustments lag, at least for lower- skilled workers, and the poor are less likely to have means of protection from inflationary pressures. This is well established for the country as a whole and the evidence suggests that macroeconomic conditions affect urban poverty in the region to approximately the same extent as elsewhere in the country (See appendix 3a, table A3.2, and discussion). 2.25 Both unemployment and inflation seem to have approximately the same effect on poverty in the northeast as they do elsewhere; there is no reason to believe that the poor in metropolitan areas of the northeast are immune from the damage done by macroeconomic instability. 2.26 One might hypothesize that rural areas, particularly those with little marketed surplus would be relatively less exposed to changes in their real incomes than are wage laborers. To test this requires the use of PNAD (Pesquisa Nacional Anual de Domicilios) data since the PME is restricted to major cites. Unfortunately, results derived from PNAD are more ambiguous. Stabilization and State Level Fiscal Deficits 2.27 The next step in the argument is that state fiscal performance contributes to national macroeconomic performance. There are two parts to this: the relation between state fiscal balance and national stabilization, and the relation between state efforts at fiscal balance and their success at achieving it. 2.28 Normally subnational levels of government can only influence the scale, composition or growth of economic activity in their jurisdictions through administrative or fiscal levers (in the latter case either through revenue policy or expenditure policy, or some combination of both). Other levers for influencing economic activity within a geographic boundary such as monetary tools (credit creation, setting of interest rates, setting of exchange rates), or trade tariffs and quotas are reserved for sovereign entities such as national governments. This difference in policymaking options restricts the degrees of freedom at the subnational level while putting a premium on sound fiscal management. 2.29 Until the de facto default and rescheduling of state debts in 1993, this traditional distinction was not fully operative in Brazil as states often spent in excess of their revenue collections without financing the excess through normal borrowing against future revenues (i.e., bonds). Instead they de facto created credit by encouraging the state banks to lend to government preferred private projects with state guarantees which were called and these in turn were financed by federal financial institutions, who eventually had to clear their accounts with the central bank. Thus, in the absence of a hard budget constraint, states in Brazil had little incentive to improve fiscal management and almost all states ran large deficits. 14 2.30 This was a contributing factor to the hyperinflation in Brazil in the late 1980s and early 1990s. (Hyperinflation also destroyed the signaling potential of relative prices within and across states. In fact the ability to evaluate the performance of the real economy in subnationaljurisdictions is confounded by the lack of adequate sectorally and spatially disaggregated price indices). 2.31 Stabilization of the currency, even as a one-time adjustment, has two benefits for the states. It leads to the more effiient allocation of resources (as a result of more meaningful relative prices) within and across states, and it also leads to the better distribution of income within states, as discussed above. From this perspective, enlightened self-interest would suggest that states contribute effectively to the stabilization program by not generating fiscal deficits that cannot be financed through borrowing against their own future revenues. 2.32 Given the danger of free-riding in voluntary programs it is understandable that the recent fiscal adjustment and competitiveness measures in the Real Plan included changing the rules of the game for the monetization of state debts.2 These measures prohibit financing of state obligations by federal financial institutions for states that have not signed a contract with the federal government for a fiscal adjustment program. This step could impose a more stringent budget constraint on state expenditures.3 However, past experience with enforcement of such rules suggests that the incentive for states to seek federal relief is not removed just that the venue is shifted from the fiscal domain to different parts of the financial system.4 2.33 When there is a sudden change in economic regime (i.e., a change in the rules of the game) and the once soft budget constraint is transformed into a genuinely hard budget constraint it should not come as a surprise that some states will enter the equivalent of bankruptcy (such as Alagoas). How to stimulate growth in a fiscally prudent way will be a major challenge for the states in the northeast following the introduction of the fiscal adjustment and competitiveness measures in the Real Plan. States with hard budget constraints will need to restructure fiscal performance of the public sector by avoiding excessive spending. And this is feasible as evidenced by the success of Ceara and Bahia in managing their fiscal deficits. 2.34 However, a corollary of the observation above requires careful qualification: States that manage their fiscal position more effectively-by keeping their annual fiscal deficits within their intertemporal budget constraint (i.e., linked to their revenue-generating capacity over time)-are more likely to have the freedom to engage in pro-active economic policies than states who mismanage their fiscal deficits. Strictly speaking this is correct only if a hard budget constraint is enforced. If not, there is no incentive to improve fiscal management (this is fully analogous to the case of state enterprises that might have recourse to the budget to cover their deficits). And herein lies a danger that can undermine the sustainability of collective action. 2.35 If some states are treated as if they are too big or too important to fail and if these states successfully get the federal government to pay for excessive spending they are likely to be better off than their more fiscally prudent "competitors" in the short run (if the additional expenditures are on 2 See Finance Ministry Web Site: http://www.fazenda.gov.br/inyles/welc.html for a detailed description. 3 According to the Finance Ministry: "19 States have now accepted the Federal Government's Program of Support for the Restructuring and Fiscal Adjustment of the States. The success of this initiative will mark the beginning of a new pattern of fiscal relations between the Federal Government and the governments of the States, whereby the latter will be expected to make a substantial contribution toward improving the fiscal performance of the public sector as a whole during the current year." 4 See Dillenger, W. (1997), page 20. 15 wages and employment), as well as in the long run (if the additional expenditures are in the form of investment in public goods, infrastructure, etc., which are more likely to improve their long-term locational comparative advantage).' The principal disadvantages of redistribution through "clearing accounts" from prudent states to imprudent states is that the size of the transfer is not transparent, and the scale and direction of the transfer may be potentially inconsistent with explicit policy goals (e.g., on poverty reduction). As will be discussed below, many services such as health and education are important to the poor in the northeast and are of greatest value if they are funded from the federal government. To the extent that covering state deficits cuts into federal assistance in these areas, these services will have worse net redistributive characteristics. Fiscal Responsibility and Deficits 2.36 The view that states can control their own expenditures may seem self-evident but is sometimes questioned on the ground that while some expenditures are under the direct control of state governments, large parts of these expenditures are determined by general economic conditions. For example, recessionary periods trigger social transfers, unemployment insurance, and possibly extra services such as health care. Furthermore, the revenue side of fiscal management is heavily influenced by economic activity via the tax base. It is possible that state finances merely reflect macroeconomic conditions but are not themselves to blame. 2.37 Figure 6 shows the key results of an analysis of state deficits in the northeast from 1980 to 95. The analysis is based on a panel of states, that is, data from each state for each year. Important determinants of the deficit, determinants that are more consistently related to deficits than income growth itself, are directly related to the administrative performance of the state governments. The fraction of state expenditures that go to the central administrative functions is closely related to deficits in the period. Not quite as strong but still clear is the relation between the share of government expenditures going to wages and the overall deficit. In contrast, states seem to be able to absorb changes in social spending. The composition of expenditures (and not just their absolute level) has an important independent component in affecting the overall deficit. See appendix 3b, table 3 for the main results of this analysis. 5 See Wildasin, D.E. (1998) for a discussion of this issue in the broader context of the undermining of fiscal discipline by state govermments when fiscal responsibilities can be shifted from the state to the central level. 16 Figure 6. Administration Costs and Dericits 0.6 -- 0.5 - _ _ _ _ __ _ _ __ * 0.4 1 0.3 ~* * ~ * * _ _ _ _ _ _ _ _ _ 0.2. - -0.1 0 0.1 0.2 023 0.4 0.5 0.6 Administration share oftotal expenditure Source: Data from Finangas do Brazil and statistical analysis in appendix 3b, table A3.3. 2.38 Other than the statistical connection between deficits and uses of expenditures, the case of Ceara is also instructive. Figure 7 shows the evolution of the deficit over the same period as covered in the previous table. In the early years, Ceara maintained very high deficits, much higher, on average, compared to the region as a whole. With the refonms beginning in 1987, there is a clear reduction in the size and the variability of deficit spending. Deliberate policy actions are a major part of state fiscal status and not simply a passive respondent to larger economic conditions. The interpretation of this result is not that administrative costs enter the deficit but other expenses do not. Rather, it is likely that both the overall deficit and the administrative load reflect general administrative responsibility. Figure 7. Deficit as Proportion of Revenues for Cear4, 198695 047 -0.1 cL 0.0102032. . . 0.1~ ~ ~ ~~~Ai nsrtonsaeoTtta xedtr Sore0aafo ia~sd rzladsaitclaayi napni b al 33 2.8 Ote ta testtstclconcto btee efct and9uses of exenitre,th cseo Cer sas ntutv. iue7sosteeouto ftedfctoe tesm eida oee intepeiu ale nteeryyar,0a.mi2iedvr ihdfcts uhhge,o avrae,copaedtoth rginasa hoe.Wih hereors einin i 187 ter i acla 2.39 To the extent that state finances complicate the problem of the federal treasury in maintaining fiscal control, control that is very important in the evolution of poverty in the region, state fiscal responsibility is a high priority. Private investment is also likely to be dependent on the overall sense of fiscal responsibility in the states where investments are planned. This could well have further implications for the effect of the deficit on growth. Public Employment 2.40 Clearly intertwined with the control of fiscal deficits is control of public employment. Whether poverty alleviation ever was the true motivation for increasing the public payroll or not, it has always been justified as such. In this section, the role that public employment plays in reducing poverty is examined and found to be of very little use. Most people benefiting from public employment have opportunities in the private sector sufficient to help them out of poverty. Therefore, the contribution to personal incomes near the poverty line is small and does not justify the large fiscal burden that excessive public employment imposes. This section builds on previous work by the World Bank and IPEA (Carneiro and Gill 1997), adding extra evidence from the PPV in support. 2.41 There is a germ of truth in the belief that public employment contributes to relieving poverty in the northeast. Figure 8 shows the percentage of income that public employment represents in the region. It is, in fact, higher than in other parts of the country. Public employment, therefore, does add income to the region. Figure 8. Public Employment Rates by Region 30% 25% ONortheast 3 % 110 Southeast Z 20% - 15% * 10%o:t ;00. ;S E 5% 0% Municipal Urban Rural Region Source: PPV. 2.42 However, a closer analysis of who benefits from this extra income and by how much leads to the conclusion that the poverty-reducing effects are minimal. How much public employment adds to peoples' incomes is very sensitive to assumptions of what this income would look like without public sector involvement. On these grounds, there are two kinds of comparisons possible. The first looks at the characteristics of employees and tries to assess what their wages would be in private employment. This is done by estimating an equation for wages to see what the effect of public employment is for different kinds of employees. 2.43 A second approach is to estimate what the overall impact on wages would be given the entire market for labor. The first method is from the perspective of the individual worker, the second is from the perspective of the market. Table A3.4 in appendix 3b shows the results of the regression analysis 18 from the PPV data set on determinants of wages. The results show that at higher income levels, the public sector employees are paid less than their private counterparts. The impact of public employment is higher only at the low ends of the wage scales. 2.44 To assess the effect that employment has on poverty alleviation, figure 9 presents the results of a simulation of the effect of public employment on the distribution of consumption. It shows the current distribution as well as the distribution that we would predict with people having private sector jobs rather than public sector jobs. On the assumption that no one is worse off by taking a public sector job than a private job (since it is not forced upon them), the effect on the distribution of consumption is quite small. Basically, a small fraction of the population have public employment, the difference between what such people would get in private and public employment is also small and the job categories which have the largest effects are not near the poverty line. The net effect is a result of multiplying these small effects together, leading to a very small impact. Figure 9. Distribution of Expenditures With and Without Public Employment, Northeast 100 90 Exponditure DExpenditure - Sector Premium 80 ___ _ _ X _ = * 70 :1 -__hX_ 60 .0 40 30 2~ 0 .0 0 39 - 1 2 3 4 5 6 7 8 9 10 Consumption deciles Source: PPV and authors' calculations. 2.45 At a regional level, the effects depend critically on the nature of the labor market in the northeast. The relevant characteristics are the elasticities of demand and supply of labor, the degree of mobility between urban and rural areas and the share of the public sector in total employment After analyzing the current levels of public employment, we were able to show that the effect of a 20 percent drop in public employment could reduce wages by as much as 6 percent but only under the most extreme assumptions: no supply response to wages, very low elasticity of demand for labor, and perfect mobility between urban and rural areas. See appendix 3b, table A3.5. In more realistic cases reflecting current technology and behavior (Lopez and Valdez 1997), the effect on wages of even so large a fall in employment would be in the range of 1.2 percent. Migration possibilities to the rest of the country reduce this figure further. Effects on particular submarkets may be larger; for example for workers with skills specific for public service, but people at risk of falling beneath poverty line do not, in general, have such skills. Therefore, it is highly unlikely that the effect of public employment on poverty is important. Given the impact on the state budgets, employment fares poorly as a poverty alleviation strategy. 2.46 For public employment to be a mechanism for poverty alleviation, its design should be more focussed on that objective. One type of program that has been used in places as different as Chile and Maharashtra State in India is a guaranteed employment scheme, in which the government stands 19 ready to employ anyone in the region at a wage somewhat below minimum wage. This sort of policy has two features, which would make it attractive in the rural areas of the northeast. First, its benefits tend to be highly concentrated among the truly needy, without any administrative mechanism to identify the poor. Only those with very poor eamings prospects would make use of the program. Studies in India have shown that a greater proportion of money spent goes to poor people in the employment scheme, without any screening of participants, than does a credit subsidy with an explicit requirement that participants be below the poverty line (Datt and Ravallion 1994). It is simply too hard and too expensive to identify the poor by administrative discretion relative to letting the poor identify themselves merely by their willingness to participate. 2.47 The other feature of a guaranteed employment scheme well suited to the northeast is its ability to serve as insurance against adverse agricultural conditions. Income variability in agriculture is high and as discussed above, has not fallen over time. The prospect of guaranteed income when incomes are low serves as an important source of security. A further benefit may be to reduce migration in years of crop failures. Migration to cities is common in bad years and, while this is insufficiently studied, much of this may be "involuntary" in the sense that the decision is regretted and would not have been taken had there been another way to get through the year. This form of public employment would have much greater effects on poverty than the present one. 2.48 For regular employment in the public sector, it might be better to consider the output of the employment-the actual services delivered to the poor from public workers-rather than the inputs to the services. A more direct and sustained focus on serving the poor would be better than diffuse effects in labor markets. This leads us to the second main leg of the antipoverty strategy: social sector expenditures. 20 Chapter 3. Poverty Reduction Through Expanding Social Sector Expenditures A. Health 3.1 No discussion of poverty alleviation strategies can be complete without including the health sector. There is a two-way relationship between health and poverty. First of all, poor health and the premature death that often results are arguably the most tragic manifestations of poverty. The causal link between income and health is well established. In the other direction, maintenance of good health is an important way to conserve human capital in the only resource that poor people ultimately own: their own labor. While the degree to which ill health is a cause of poverty is unknown for the northeast, there is some evidence elsewhere that has identified it as an important contributing factor (Gertler and Molyreaux 1996). Whether ill health directly lowers earning capacity of the poor or not, it imposes a serious cost on the poor which government action can ameliorate. 3.2 It is not common to find data that include measures for both health and economic well-being. This section relies on three main sources of data. First the PPV has extensive information on household income and consumption. An important motivation for the study was to assess the use of public services. As a result, there is detailed information on the use of health facilities, both public and private. A drawback of this data as far as assessing health status is that it is too small at 5,000 households, to capture relatively rare events such as infant or child deaths. For this, other sources are used. The second source of data is Pesquisa Nacional Demografia e Sauide (PNDS) of 1996, which has extensive information on a variety of measures of health status, and, at 12,000 households is large enough to get reliable estimates of health outcomes. A drawback of this study is that it does not include questions directly measuring income or total expenditure. What it does have is a battery of questions on ownership of various consumer durables. These questions were used to form a composite index of wealth, or, ownership patterns for such items as radios, bicycles, building materials for houses, numbers of rooms in the house (per person), among others. Appendix 4a discusses the construction of this index. 3.3 The third source of data was assembled and analyzed by IPEA-RJ by creating a panel of districts for the census years 1980 and 1991. This data provides a very accurate measure of community infant mortality rates along with other demographic and economic variables. 3.4 That the health of the poor is worse than that of others should come as no surprise. In Brazil, however, the difference between rich and poor on this dimension of well-being is extraordinary and further reflects the severity of the gap between them. Figure 10 shows the relationship between income and the probability of death of children. This information is derived from the PNDS of 1996 and the indicator variable is a little unusual. It represents the fraction of children ever born to women under 40, at the time of the survey, who have died from any cause. It is not the same as the more common indicator of child mortality (the probability of dying between the ages of one and five), but allows for a much larger sample to be used in its calculation. 21 Figure 10. Child Deaths PSercent 1 2 1 2 3 4 5 6 7 8 9 1 0 Consumption deciles Source: PNDS. 3.5 The difference between the richest and poorest decile of the wealth measure used from the survey is a factor of five. Children born to the poorest families are five times more likely to die before their mothers reach an age of 40 than are children in the top 10 percent of families. For comparison, in India the same mortality rate doubles between the poorest and the richest. 3.6 What are the poor dying from? A large proportion of children die from communicable and environmentally determined diseases. Diarrhea and acute respiratory infections are common among the poorest. Other infectious diseases are still common among the poor even though they are quite rare among the better-off. Among older people, the noncommunicable diseases are responsible for more deaths and are increasing in prevalence. Indeed, poor people suffer from the noncommunicable diseases more than do richer people (Murray, Yang, and Qiao 1992). However, the relative incidence of communicable and environmentally determined diseases between rich and poor is much higher. Therefore, a much higher proportion of money spent on communicable diseases will reach the poor than the same amount on noncommunicable diseases. 3.7 The questions for public policy to answer are first, how much of public money spent on health actually reaches the poor in terms of more or better services, and second, what kinds of policy are most effective in improving their health or well-being? The first question depends on the fraction of users of publicly provided health services who come from the poorer groups. The second depends on understanding the determinants of disease and ill-health. In particular, we need to identify the main determinants that are influenced by public policy. This chapter deals with the latter question first and then returns to the issue of health care. 3.8 There is a tremendous degree of complementarity between the pursuit of equity and efficiency in the health sector when policies are directed at communicable diseases. Public goods, externality-correcting policies and basic health information are all of high priority in the traditional public health agenda and are well justified in terms of standard efficiency arguments: they directly address market failures. At the same time, pursuing them is particularly pro-poor in their incidence. Policies exhibiting such a degree of complementarity are not common and provide an unusual opportunity. 22 3.9 Several different studies investigate the relationships between public policy and health status. While there are a few differences among them, they tell a remarkably consistent story. 3.10 The first study relies on the study and tries to explain the pattern of deaths of children shown in figure 10. The variables chosen to explain this pattern are (1) a set of variables that measure household characteristics, (2) a set of determinants that are not part of the health sector but which are expected to have important health effects, especially the education of mothers, and (3) determinants that are part of traditional public health programs, including access to water and sanitation services. 3.11 By decomposing the mortality numbers into these sets of determinants, we can make an assessment of the relative importance of different policies in improving the health of the poor. By comparing the effect of these variables with the differences between rich and poor, we can attribute such differences to their constituent parts. 3.12 Appendix 4b, table A4.1 shows the equation that results from regressing the mortality variable on these hypothesized deterrninants. Expenditure and education play important roles in the relation as in many other studies of this type. Of particular note, however, is the effect that the source of water and sanitation facilities have on mortality. 3.13 Not only are sanitation and water supply important as determinants of health, they are also services which sharply distinguish the poor from the non-poor. Figure 11 and figure 12 show the relationships between expenditure and access to sanitation facilities and the source of drinking water. The figures paint a stark picture. In the poorest decile of the population, disproportionately in the northeast, almost 70 percent are without any sanitation facilities. Furthermore, the rate at which this proportion falls is dramatic. By the third decile, still poor by national standards, this number drops to less than 15 percent. By the middle of the consumption distribution, sanitation facilities are basically universal. Figure 11. Nationwide: The Poor Have Less Access to Sanitation 7 ,0 - = 4 0 COnSUm ptiou deciles Source: PNDS. 23 Figure 12. The Poor Have Less Access to Safe Water 90- 70 60 2 60 I 40 30 Consu m p iot de cile s 3.14 The quality of drinking water is almost as revealing. The survey did not actually measure contaminants in the water. It merely indicated the source. However, some of the sources are clearly good and others are clearly bad . The relatively good sources are piped into the household and bottled (which is fairly common in the highest decile). The clearly bad sources were from surface ponds, streams, etc. The remaining sources were basically wells and community pumps. These could be good or bad; since we cannot distinguish between them in the data they are not included here. On the basis of these results, one could almost define a poor person as someone who does not have piped water or sanitation services, the correlation with wealth being that close. 3.15 Combining the results of the regression equation with the evidence concerning differential access, yields figure 13. This picture takes the household profile of a poor, northeastern family with average values for education, expenditure and so forth, and simulates the effect of replacing the probability of having sanitation and safe water sources of poor people with the probabilities of the top decile (for sanitation to a 100 percent probability). This is done for water and sanitation separately and jointly. The proportion of children born to these people who subsequently die is cut in half by the combined effect of better water and sanitation, particularly of the former. Figure 10 showed a gap between the mortality rates of rich and poor to be 8 percentage points - 10 percent for the poor, 2 percent for the rich. The results in figure 13 shows that 5 percentage points of these 8 are attributable to differential access to water and sanitation. This is a tremendously important concern. Figure 13. Lack of Access to Water and Sanitation Increases Child Mortality 10 X_X__ _ 9 = 7 - ~. 6 E 5 3 4- *. J = J3 2 0 Source: PNDS and calculations based on appendix 6. 24 3.16 This result is confirmed by other studies. First, the PPV gives evidence concerning the effect of water and sanitation on morbidity. The question asked was whether a child had diarrhea in the past two weeks. Ordinarily, self-reported morbidity is not a particularly reliable source of information. The difficulty is that people differ in what they consider a problem and it is common to find that richer people have higher morbidity than poorer by this measure. Rich people seem to have less tolerance for even small problems where poor people take them as normal. So, variables correlated with self-reported morbidity could be related to either the morbidity itself or the probability of reporting it. For the current purposes, however, it seems quite likely that the source of drinking water or sanitation facilities are related to the reality of being sick, but are not at all likely to be related to the probability of reporting conditional on being sick, with income and education being separately controlled. 3.17 With this assumption, table B4.2 in appendix 4b shows the results of the analysis of PPV data. The water and sanitation variables are highly significant in deternining the occurrence of diarrhea. 3.18 Further analysis based on census data also reinforces the importance of water and sanitation. The data for this study was based on districts in the 1980 and 1991 censuses. It includes infant mortality rates, income, numbers of publicly employed medical personnel, and the mean number of years of training of publicly employed nursing staff. The medical personnel, variables capture the density of coverage of the pubic health care system. Errors in measurement are possible if residence and employment are in different districts. The data also included the proportion of people covered by safe water with access to sanitation facilities. The regressions were run by year, region of the country, and urban/rural status for a total of eighteen different regressions. Selected results are in Appendix 4b table B4.3. The results concerning water and sanitation are summarized in figure 14. Fifteen of the eighteen coefficients on water supply are significant and the other three are of the correct sign, even if not significant. This is strong corroborating evidence on the importance of water. The results on sanitation were not as clear cut, as was true with the survey data. Here the coefficients vary in sign and significance. There is a slight predominance of correct over incorrect (six versus three coefficients being significant in the predicted versus the perverse case, respectively), but not nearly as convincing as for water. The correlation between having safe water and proper sanitation is very high, however, so the effect of sanitation may be captured by water itself. Figure 14. Coefficients in Census Study by Predicted Effect and Statistical Significance (a) 15 - r Water 13Sanitation 10 0 U 4. O~~~~~ / _- 44 0 .. 4 Source: Barros, da Costa, and Mendonca (1998). 25 Figure 14b. Coefficients in Census Study by Predicted Effect and Statistical Significance (b) 1 8, 16- 1 2- 10- 0~~~~~~~ 6 o ~~~4- z ~~~~~2 W~~~~~~C 8 /l }1 - = C _ 10 0-~~~~~~ e, ° -> ~z 5 ! aE, M ed ical care in p u ts Source: Paes de Barros et al. (1998). 3.19 As for public medical inputs, the results are discouraging. Figure 14b shows the distribution of coefficients on the numbers of public sector doctors and nurses as well as on the training of nurses. Here the results are striking. There are considerably more perverse effects that are significant than effects that go in the "right" direction. Sixteen coefficients are significantly negative and only seven are positive. Furthermore, of the ones that are positive, most are related to the training of nurses and not the coverage of the population of either doctors or nurses. It is possible that this reflects the importance of maintaining quality in public facilities or the contribution of relatively less well-trained professionals (nurses vs. doctors), as long as standards are met. It is also possible that this merely reflects higher earning capabilities or general educational attainment for women in the district which is reflected in higher wages and higher standards for employment. In any case, it is impossible to infer any effect of publicly supplied medical inputs on infant mortality. 3.20 Finally, a recent study of water and sanitation was done by the World Bank (1998b). This also examined census data for 1990 for four states: Minas Gerais, Pernambuco, Rio de Janeiro, and Siao Paulo. The relevant results are presented in table 4. Once again, water and sanitation appear highly significant. The results are so consistent across studies with differing data sets, researchers, and methodologies that it is hard to believe they are spurious. Table 4. Impact of Safe Water and Sanitation on Mortality Rates Mortality Infant Less than 5 years old Average Mortality rate 39.40 8.80 Change due to a 10 percentage rise in Urban access to piped water 0.80 0.25 Urban access to sewers 0.60 0.15 Source: World Bank (1998). 3.21 What are the policy implications of the results on water and sanitation? The jump from empirical findings and policy is not straightforward. Figure 11 and figure 12 directly imply a pair of policies based on distributional concerns alone. The government has in its control two instruments: the extent of the distribution system for water and sanitation via direct investments-the quantity dimension-as well as the proportion of costs that are subsidized by charging fees, which are too 26 low-the price dimension. Since current possession of water and sanitation services are closely related to income, or, at least, the poorest do not benefit much from them at all, a progressive policy option is to cut current subsidies by raising fees. "Average" incidence of the policy of subsidization is not good. Very little of these subsidies reach the poor so there is not much of an equity benefit from the policy and they tend to distort the use of water, encourage wasteful practices, and thus compromise efficiency. On the other hand, since all of the non-poor already have sanitation facilities and good sources of water, any expansion of the system to those not currently enjoying these services will be progressive. The "marginal" incidence of expansion is very pro-poor. So, for any level of subsidy chosen, expansion is progressive and for any level of coverage (up to the point of universal coverage) raising fees is progressive as well. 3.22 This leads to an unambiguous answer to the question of whether the pursuit of better health for the poor via extension of services would be terribly expensive. With appropriate pricing schemes, extending the system to include the poor may have little or no fiscal impact at all. The World Bank study of Brazil cited above, as well as many other studies around the world, have found very high willingness-to-pay by people for good, reliable drinking water as well as sanitation systems that remove wastes from the house. The fees that people appear willing to pay are almost always higher than average costs of provision in urban areas and certainly higher than marginal costs. To the extent that very poor people in urban areas may not be willing to cover all costs, the amount of subsidy needed to reach them may be very small indeed. Neither water nor sanitation are public goods though they may have extemalities justifying small subsidies. On the other hand, the private sector has not been much of an actor in the sector. Why this is so may be worthy of further study but given that it is so, this presents an opportunity for the public sector to improve efficiency and the health of the poor at the same time. It simply does not necessarily require large outlays of money. 3.23 The results of the analysis of the PNDS data show the actual, physical source of water to be the important factor. Depending on the mechanism by which better water leads to better health, the types of policies appropriate to rural areas will differ. Increasing coverage will not necessarily entail hook-ups to houses for waste disposal. Many of the policies in these cases may rely instead on public information and health education campaigns. Promoting basic hygienic practice (not something the private sector does) is a standard public health role which may have great payoffs in terms of improved health of the poor. 3.24 Turning to medical care, why is it so difficult to find a correlation between measures of access to publicly supplied medical care and health status? There are several possible explanations. Appendix 4b, table B4.4 presents the results of an analysis of the PPV data in which examines people's likelihood of seeking care once sick in either the public or the private sector. Several conclusions are worth noting. First, figure 15 shows that the likelihood of seeking care is closely related to the type of illness experienced. People find a way to handle relatively serious illnesses such as accidents and heart attacks, but may not bother to seek medical attention for colds or general pains. This might help explain why it is hard to find an effect of medical care on such gross measures of health status as mortality rates. When people are faced with life threatening situations, they find some way to cope: so it would only be in the less severe cases or in financial protection that we might see the benefit of medical services, not in mortality. 3.25 Two results from the analysis of facility use are also interesting from the point of view of distributional effects of health spending. First, it is very clear that the higher a person's expenditure, the more likely they will seek care in the private sector and not the public sector. Second, the effect of having insurance is similar to expenditure: insurance leads to a large shift from public to private facilities. Tax benefits for insurance come only with formal employment and higher expenditures, 27 and indicates a substantial benefit to relatively well-off people, which does not show up in the health budget. Figure 15. Likelihood of Seeking Care by Health Problem Accidents Heart attack 1 Dental'III StomachI General pain Cold 30 4 0 0 10 20 30 40 50 60 Likelihood of seeking care Source: PNDS and calculations based on appendix 6. 3.26 Of course, since our only measure of publicly supplied medical care is from personnel numbers, there could be many other ways in which public action can benefit health status. How public personnel spend their time is not captured in this analysis and may be of critical importance in affecting outcomes. There is ample evidence, worldwide, that high immunization rates can significantly reduce infant and child mortality and that such rates are due in large part to public action. The effect that clinical, patient-demanded services covered by the public sector have on health is more problematic (Filmer et al., 2000). 3.27 Regardless of the direct effect on health status, how does the health subsidy look in terms of redistribution to the poor? This depends on who is paying. 3.28 Figure 16 shows the usage of public facilities in the northeast by consumption decile. The deciles used are based on the national distribution of consumption. Note that there are very few people in the highest deciles that use public facilities. This is because there are very few people in the northeast who are in these deciles. The usage pattern is very progressive from this perspective. To the extent that usage of public facilities can be translated into subsidies (discussed below), very little of the public subsidy goes to people who are wealthy relative to the Brazilian population, and over 50 percent of these subsidies accrues to people in the poorest 30 percent of the consumption distribution (the lowest 30 percent corresponds roughly to those in poverty). If the budget for health were paid from federal sources, i.e., from taxes collected from people on average richer than those in the northeast, it would be an excellent means of redistributing income. 28 Figure 16. From the National Perspective: The Distribution of Health Care in the Northeast 25- =,@ 20 i 5 1 lo- Ua 5 0 ...... ....... 1 2 3 4 5 6 7 8 9 10 Consumption deciles (national) Source: PPV. 3.29 However, if funding were to switch to the states themselves, the distributional impact becomes less clear. Figure 17 shows the proportion of visits to facilities from people in different deciles of the consumption distribution defined within the northeast. That is, people in the lowest decile in figure 17 are the poorest 10 percent of the northeast population not the poorest 10 percent of the national population who live in the northeast. Figure 17. Within the Rural Northeast: Distribution of Health Care Benefits 14- .1 1112-Z *~10- 6- 2- 0 1 2 3 4 5 6 7 8 9 10 Comsumption deciles (northeast rural) Source: PPV. 3.30 If visits correspond to use of public money and if the money needs to be generated within local communities, then the pattem of use in figure 17 represents the beneficiaries of subsidies relative to the population paying for them. In this case, the distributional impact is not nearly so progressive. It is not the poorest of the poor in the northeast who use health services 29 disproportionately, but the well-off relative to others in the region, though not relative to the country as a whole. If the people in the northeast have to support health services from their own tax revenues, these services will lose their ability to serve a redistributive function. 3.31 Since the data in figures 16 and 17 are visits and not money, it is not possible to assess the incidence of subsidies. Should it be the case that richer people can bargain for better and costlier services, figure 16 would be flatter and figure 17 would be steeper. If the government is successful in targeting its subsidies and gives better (or at least more expensive) services to the poor, then the reverse is true and using health as a redistributive device is even better than it looks here. International experience suggests things are worse than they look. 3.32 The distributional effect still depends sensitively on whether money is coming from within the region or from the federal government. The states can choose to add to federal transfers to different degrees. CearA, for example, added about US$23 per capita to its budget on top of its federal transfer of US$43 whereas Bahia added US$16 to its allocation of US$32 (World Bank 1998). The distribution of income of beneficiaries relative to the taxpayers footing the bill can differ substantially. 3.33 Health services do not appear to make much difference in terms of mortality, and, if paid for from local sources, may not be a very good vehicle for redistribution. On the other hand, the poor don't have much choice. Figure 18 shows the use of different services by consumption group (going back to national definitions). Less than 15 percent of visits to health-care providers by the poorest 30 percent are to private facilities, leaving the large majority to rely on public services. This leaves us with a cruel dilemma. The services may not be very good and they may not go primarily to the poor, but the poor have no alternative. This is another manifestation of the extreme skewness of the consumption distribution. So little of the expense in aggregate terms helps the poor, but that small amount represents a large contribution relative to the poor's resources. Figure 18. Health Care Visits by Decile 1.00- 0.90 OPublic Hospital OPublic Clinic 0.80 OPri vate 60.70 0.60 0 .5 2 0.40 ~-0.30 0.20 0.10 D e cile Source: PPV. 3.34 Furthermore, the experience in Ceara shows that public services can be made effective. Tendler and Freedheim (1994) assess a major health campaign in which infant mortality fell 36 percent over only a few years. The key elements of success appeared to be (1) a merit hiring system and large advertising campaign that created a "sensed mission" for the program and respect for its workers, (2) flexibility in job descriptions that allowed workers to take on tasks, unusual for health 30 workers but which increased trust, and (3) rejected job candidates were educated about what to expect from the health workers, turning them into "informed public monitors." 3.35 Improving the delivery of public services is not easy. Civil service rules, the preferences of professionals to be in or close to cities, and bureaucratic constraints on a personal service all combine to undercut the quality of services that can be promised to the rural poor. Increased decentralization of services and their administration may also run into the constraint that municipal level employers often have less training than those employed at high levels of governance. As the program in Ceara matured, factors such as the guarantee of regular pay (a frequent problem in some public programs) and continued training complemented the recruitment rules resulted in a sustainable program. Creative programs such as that in Ceara provide encouraging examples of government's ability to overcome these constraints. 3.36 A further benefit of medical services to the poor also relates to insurance. Very few poor people have formal insurance (Figure 19). For these people, the prospect of free or heavily subsidized care through public facilities serves a function similar to insurance. It frees them from worry about financial vulnerability. How much this is worth can be approximated by calculating the "risk premium" associated with assurances against financial loss. Appendix 4c presents a description of this concept and its calculation. Figures 20 and 21 show the size of this benefit as a fraction of the costs associated with providing the services themselves as a function of the cost of services and of the income of the recipient. Figure 19. Coverage by Health Insurance N~ ~ ~ ONER C30razii|_ 60 - - 450 3 40 20 l 2 3 4 5 6 7 S 9 10 Co.suan ption deciles Source: PPV. Figure 20. Value of Insurance by Household Per Capita Consumption 60 - - -Low cost servce "-.Hghos evc 4,50 - _ _ _ _ _ _ _ _ _ _ _ 10 __ . __ O- -- - - -- - - -- j 1 - - -_ _ _ 0 J I 100 200 300 400 500 600 700 800 900 1000 Ho.sehold consumption Figure 21. Value of Insurance by Cost of Insured Service 250 250 -| Low income households - - High icome households - 200 0 5 e 8 100 _, 50- 10 10 35 60 85 110 135 160 185 210 235 260 285 Cost of service (Reais) 3.37 Several points are noteworthiy. First, the insurance value of providing free service can be a very high fraction of the cost of the services themselves. Second, the value of this benefit goes up wvith the cost of the service. It is more valuable to insure against expensive procedures rather than relatively cheap ones. This implies that the role of primary health centers as a source of referrals to hospital treatment may be more important than its role in providing free care. People may be willing and able to pay for routine curative care and subsidies at that level may not be worth much. Finally, the insurance benefit of the services is higher for poor people than for rich at every level of expense. Therefore, this could offer an important means of helping the poor relative to the rich. 3.38 However, there is an imnportant caveat to this conclusion. Free hospital services for the poor, will require that access to the services be based either on income itself or on medical need. Since the determnination of a person's income is often difficuft, the better prospect is to assure that the referral system operate fairly. If the poor had the same probability of receiving care once sick as other people, their higher disease incidence would be enough to make the system pro-poor. In conjunction wvith the higher insurance benefit per unit lost to poor people than rich from expensive hospital care, the potential is very large. The reality depends on how referrals are administered. 3.39 In summary, there are policy options that are immediately available that can alleviate some of the worse manifestations of poverty in the northeast. These include expansion of access to safe water and sanitation. In urban areas, this can be done by construction of facilities with appropriate charges for the services such that the net cost to government is low. In rural areas, different policy options might be discussed ranging from increased health education for basic hygiene and support for community water supply projects. 3.40 The scope and quality of medical services need to be further examnined. As currently organized, they do not appear to reduce infant and child mortality. However, because of (1) the reliance that the rural poor, in particular, place on the public system, (2) the large, potential if currently speculative, welfare gains due to the implicit role as insurance which the public system represents, and (3) the precedent of marked improvement in the quality of public services it is best to emphasize improvement of the system rather than allocation away from it. 32 B. Education 3.41 The problems of education in the northeast have been documented in great detail in recent joint work by the Bank, the Ministry of Education, and UNICEF. Here we try to assess the role that education can play in poverty alleviation. The basic conclusions are first, that previous analyses are correct in that education will have to be the centerpiece of long-term efforts to reduce poverty. Second, there are clear policy-related measures that have not been exploited that should yield improvements in educational attainment. Third, the role of education will vary depending on the degree to which the long-term development of the region is expected to rely on agriculture. To the extent that development will depend on people moving out of agricultural occupations, education is the key to further growth. To the extent that agriculture will remain the principal occupation of the poor, the role of education is not as straightforward and complementary strategies will have to be attempted. Finally, the effects of education are long term. As important as education is, we should not ignore the numerous antipoverty programs that can make substantial progress in the near term. 3.42 Expenditures on primary education in the northeast are potentially very progressive. Figure 22 shows the distribution of children in primary school in the whole region and in the rural areas in particular. It is clear that poor people benefit greatly from current enrollment patterns. As with health, we cannot equate service use (kids in school) with amounts spent. Expenditure per student can vary greatly between states and between schools within a state. To the extent that service use is correlated with expenditure, however, expenditures on primary education by the federal government represent as well targeted an expenditure item as one could imagine. Further, in contrast to the analysis of health care, even if the whole of the education budget were financed from within the northeast, the distribution of beneficiaries is still quite equal (figure 23), that is, educational facilities are used by the poorest of the poor, the poorest 10 percent of people in the northeast, to at least as great an extent as any other group in the region. Figure 22. From the National Perspective: Distribution of Benefits to Public Primary Education b 25 n ~ ~ ~ N4taist ru! . 0- 1 2 3 4 5 6 7 8 9 10 Cbmtionle (namonl) Source: PPV. 33 Figure 23. Within the Rural Northeast: Distribution of Benefits to Public Primary Education 12 - 8 - 6 - 4- 2 0 r- - .- r- - - I- 1 2 3 4 5 6 7 8 9 1 0 Consumption decile (rural northeast) Source: PPV. 3.43 However, as has been documented before, the region has further to go in improving access to education for the poor. One reason why education expenditures go disproportionately to the poor is that they have more children. The equality of enrollment numbers does indicate some shortfall in reaching the poorest since, on the basis of number of children, they should be receiving an even higher fraction of expenditure. 3.44 Figure 24 and figure 25 show a distinct difference between the northeast and the rest of the country concerning education attainment. In the rest of the country, even the poorest people get their children to start school and keep them there for four years. This is not true in the northeast. While almost everyone starts school, poor people in the region drop out quite soon after and only 50 percent reach grade 4. More specifically it is boys that show the greatest deficit in education and rural boys worse still. This is in contrast to urban areas in the region and in contrast to the poor in the rest of the country. In fact, the rest of the country illustrates a quite characteristic pattern of Latin American countries as a whole (Filmer and Pritchett 1998), which shows that all children stay in school through grade 4 before disparities between the income groups emerge. The northeast displays a pattern that is characteristic of much poorer countries. Figure 24. Brazil 1996-Northeast Education Attainment for Ages 15-19 1.2 Richest 30% A 1 : _ C =, 4Middle 30% 0.6 0.4 1 2 3 4 5 6 7 8 9 Grades Source: PNDS. 34 Figure 25. Brazil 1996-Non-Northeast Education Attainment for Ages 15-19 1.2 | Richest 30% T-Middle 30% 1 "---_oes 30 :2 0.8 0.6 ~.0.4 0.2 0 - - - - - ---ll 1 2 3 4 5 6 7 8 9 Grades Source: PNDS. 3.45 A second dimension in which the poor of the northeast differ from the poor in the rest of the country is in the age at which children first start school. Figure 26 and figure 27 show that in the rest of the country, the great majority of even the poorest people tend to start school by the age of 8. In the northeast, however, there are still a lot of children who don't start school until much later. Again, this is characteristic of poorer countries and not even of the poor in the rest of Brazil. These two observations may well be linked. Older children, especially boys, in rural areas can provide useful labor in agriculture and will tend to drop out as the alternative uses of their time become more and more valuable. If they have not begun school until they are nine or ten years old, they will not progress very far before these other claims on their time arise. Figure 26. Brazil 1996-Northeast, Ever Enrolled for Ages 6-14 1.2 0.8 C .a ~v~A +Richest 30% ', 0.6 -U- Middle 30% o Poorest 40% e 0.4 0.2 0 6 7 8 9 10 11 12 13 14 Age of household member 35 Figure 27. Brazil 1996-Non-Northeast Ever Enrolled for Ages 6-14 0.8 - -: 4-Richest 30% 5 06 -U-Middle 30% X 0 4 - ,i * i - Poorest 40% 04 o II I I 6 7 8 9 10 I1 12 13 14 Age of household memb er Source: PNDS. 3.46 Increased enrollments can be expected to improve earnings. Appendix 5 table A5. 1 present results an of econometric investigation of the determinants of earnings of different categories of workers. With only one exception, the results show strong effects of increased education. For all people living in rural areas, increasing levels of education leads to substantial increases in earnings. For farm workers, this effect is even stronger.6 These results reinforce prior research that show large effects of education. The only exception to this rule is the effect on earnings of farmers working their own land. In this case, there is no discernable effect of education on earnings. This result is consistent with the literature in that such effects are most consistently found in places undergoing rapid changes in technology (Foster and Rosenzweig 1995). The main effects of education may be, therefore, in the change of employment from farm-work to other rural (and urban) employment rather than in agriculture itself. We return to this issue below. 3.47 Having identified a gap in enrollment between the children of the poor and the non-poor as well as a strong correlation between education and earnings, the next question is: how to overcome the gap? The recent work by the Government of Brazil, The World Bank, and UNICEF go into this issue in great detail, looking at ways to improve the efficiency, coverage, and quality of the educational system. This report adds a little more detail on the demand for education that can be recovered from the PPV. 3.48 Table A5.2 in appendix 5 shows an estimate of the demand for schooling. They indicate a few constraints on the effort to increase enrollments, but also imply some policy measures that can help. First the constraints. The amount of parental education is an important determinant of the demand for education. These are not things that can be changed in the short run so the demand for education among the poor will clearly be lower, other things being equal, than among the non-poor for some time to come. An interesting result is that the father's education appears to be somewhat more influential than the mothers in determining children's enrollments. This differs from the results found in some other countries but seems consistent with the data in figure 28, which shows that boys are less likely to attend school than girls. It is possible that the father provides the relevant role model 6 This result may seem odd since it is not clear how higher education would improve earnings of farm workers. Evidently the statistics do not distinguish between increased productivity in a given job and choices between jobs. Only very high pay would induce a college-educated person to be a farm worker and then the role is likely to be managerial. 36 for boys and it is the enrollment of boys that is the most variable for young children. There is not much, however, that policy can do about this. Figure 28. Brazil 1996-Northeast Education Attainment for Rural Sector 0 8 067 0.6- Fe a d 0.4 o- SE 0.4-l 0 1 0 2 3 4 5 6 7 S 9 G rades Source: PNDS. 3.49 What does open the possibility of large policy effects is the strong and consistent findings that enrollments respond to costs of schooling: out of pocket costs, travel time, and the opportunity costs of children's time. Out of pocket costs for uniforms, books, and fees reduce attendance. Even more striking, however, is the effect of the amount of time it takes to get to school. Here is one area in which public policy can affect the decision to attend. Expanding the network of schools, bringing more schools closer to children appears to have substantial potential for increasing enrollments. 3.50 Two factors affect how children respond to the opportunity cost of having alternative work. First, the nature of work that the family does plays an important role in determining whether a child goes to school. In particular, a distinction can be made between the children of people who own their own farm and those who are farm workers. The latter, who generally work for hourly wages or for piece work during supervised work periods, face a direct opportunity cost for every hour taken off for school. They cannot make up time spent in school by rearranging schedules. When families own the farm, however, there is no reduction in children's enrollment. Work time is more flexible. 3.51 The stronger evidence for the effect of opportunity costs on attendance, however, is that the wage of the child is an important factor in the decision to attend school. The higher the wage, the less likely it is for the child to attend school. While the wages foregone by children appear to be outside of the control of policymakers, the fact that children (and their parents) are responding to the financial consequences of attendance opens the possibility that programs such as the "bolsa escola" being implemented in Brasilia may overcome the lost earnings of kids in school. The degree to which such payments will affect enrollment varies substantially by age, earning ability, and initial enrollment rate. Young children from poor families in the northeast-precisely those who we see dropping out at rates much higher than average-would have the strongest response to such a scheme, representing perhaps as much as a 20 percent increase for a payment of I real per day. Such program may also have a significant effect due to its impact on family income, the income elasticity of demand for education being quite high. Income transfer schemes tied the enrollment affects both the income and substitution effects for increasing enrollment 3.52 These results point quite clearly to education as the lynchpin to increased incomes for the poor. It would be a mistake, however, to assume that the answer to poverty in the northeast, at least 37 for a fairly long time to come, is solely a matter of improved education. There are three sources of concern. 3.53 First, will an expansion of education necessarily reach the poor? Figure 29 shows the enrollment rate of each of the income groups in the northeast. Whereas, the poor have the longest way to go to attain universal enrollments, there are quite a few children in the region who are not in school. Is it possible that, short of the ultimate attainment of universal education, expansion of the system may affect the middle income groups in the region first? In contrast to the distribution of sanitation facilities in the previous chapter, it is not the case that the marginal beneficiary is t certain to be poor. This does not argue against expanding the system, only to suggest patience in anticipating an effect on the very poor. Figure 29. Enrollment Rates by Decile, Northeast 100 a= 90 2>j 9 50 80 - 30- 2 0- 96 10 1 2 3 4 67 910 Consum ption deciles Source: PPV. 3.54 Second, increased education is likely to lead to increased migration out of the region. Figure 30 compares the educational attainments of people in the northeast to migrants from the region. Migrants are distinctly better educated. It appears that one of the beneficial effects (from the individual's point of view) of education is that it opens up opportunities in a larger labor market. People benefit from a national market for relatively well educated people. From a national welfare point of view, there is nothing wrong with this. From the point of view of the taxpayers living in the northeast, however, it is more problematic. Benefits from state expenditures on education are very likely to accrue to people who end up leaving the region and to their employers outside the region. As noted above, the redistributional effect of education expenditures is higher when funding comes from outside the region. The "regional externality" identified here adds a further argument for the maintenance of educational expenditures as a federal responsibility. As a whole, the country benefits from the education of migrants whereas the region may not. 38 Figure 30. Distribution of the Educational Level for Natives and Emigrant Natives of the Northeast, 1995 5 0 cm 40 20I 20 IL 1 0 == 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Years of schooling °Natives *Emigrant natives Source: IPEA, 1998. 3.55 Third, the earnings functions described above identified a strong effect of education for everyone except people who farm their own land. It is possible that this lack of an effect might be overcome with curriculum changes specifically geared to agricultural production. It is also possible that the production effect of more education simply isn't very strong. Certainly more research should be done to establish the connection between education and agricultural production but these results, and those like them in other countries should raise some doubts as to the effect of education as long as agricultural activities constitute a substantial part of the earnings of the relatively poor in the foreseeable future. If the tendency of the region to shift from agriculture to other activities to continues into the future, education will play the central role. To the extent that poor people choose to remain farmers, other policies will be needed to address poverty. 3.56 Finally, the effects of education on poverty take place over a long time. Even if agriculture continues to decline, current education will still take a generation to affect the poverty profile of the region. The promise of education should not slow the search for more immediate actions to help the poor. 39 Chapter 4. Poverty Reduction Through Selected Targeted Interventions and Transfers 4.1 The "half-leg" of the poverty-alleviation strategy deals with direct money transfers. The range of possible programs, used or imagined, that fall under this rubric is very broad. Here we concentrate on one existing program, the pension system, and the problems of risk reduction and drought relief, which have great potential to help the poor. A. Pensions 4.2 The national pension system is currently undergoing substantial reform to improve the targeting performance of the program. In this section, some misconceptions concerning the old program are highlighted and its potential for poverty alleviation in the northeast is identified. It is too early to evaluate the post-reform effects, although the main changes appear to improve the redistributive impact. 4.3 A common misconception has been that the system of pensions results in a major transfer of resources from the south and southeast to the northeast regions of Brazil. The rules of the game are such that pensioners in the northeast are entitled to payments of one minimum wage whether or not they have contributed to the system. The PPV (Pesquisa Sobre Padr6es de Vida) data show this impression to be mistaken, but also provides a plausible explanation for the misconception. 4.4 Figure 31 shows that there is a germ of truth to the opinion that pensions transfer money to the region. It indicates that at each level of income pensions represent a larger fraction of income for people in the northeast than they do for people in the rest of the country. Therefore, if people made their judgments about who gets what in different parts of the country and had a particular type of household in mind for the comparison, they might make the inference that transfers went from south to north. However, this inference is untrue as seen in figure 32. When added up across all income groups within a region we find that the average transfer is 69, 89 and 132 Reais in the rural, urban and metropolitan northeast respectively and 38, 124 and 178 in the South and Southeast. Figure 31. Percent of Income from Pensions 30 o 13 ~~~~~~~ ~ ~ ~~~NE-rurall X 25 _ a~~~~~~~~~~1SE-rural _ E0 2 -5 - - - -i 1 ~ 25 7 8 9 1 E 0 1 1 15 -_ a. 0 1 2 3 4 5 6 7 8 9 1 0 Income deciles Source: PPV. 40 4.5 The reason for the discrepancy is twofold. First, the age distribution of the northeast is quite different from that of the south. A smaller fraction of the population is of pensionable age. People in the northeast don't live as long as people in the rest of the country. Second, as noted above, the distributions of income are much different. While a larger percentage of peoples' incomes consist of pensions, the fact that incomes in the north are so much smaller than those in the south means that the total transfer is quite a bit less. Figure 32. Average Pension in Reais 1 8 0 - 1 64 0 -_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 140-/l_ 8 20 Wt_ 8 100- W _______________| ; 80 -*vN_ -- . _ __ |Southeast 6 0 20 - 0 M'etro. Urban Rural Source: Authors' calculations based on PPV. 4.6 Fully accounting for the interregional transfer is more complicated than this calculation. The net effects need to be seen in relation to the source of funding and prior contributions (if considered by the public as a funded scheme). Clearly, the first generation to receive unfunded benefits clearly receives an advantage. Assessing this fully requires an intergenerational accounting scheme with very demanding data requirements. To the extent that the program is generally a pay-as-you-go scheme, however, the net impact on regions can be assessed through contemporaneous tax payments. According to this assumption, a very progressive income tax as a tax base turns the results around and shows a net flow to the northeast. However, under a wide variety of more realistic assumptions concerning tax incidence, the results of the direct transfer calculation stand. 4.7 Even though it appears that the pension system is not particularly well targeted to people in the northeast paradoxically it appears to be an important source of income support to the poor. Figure 33 shows the distribution of consumption in the northeast states with and without the contribution of pensions. Because there are a large number of people with consumption after transfers that are in the fourth decile or so (again, the boundary between the third and fourth corresponds roughly with the poverty line), the pensions appear to keep a large fraction of people from dropping into poverty. This is represented by the large spike that appears in the bottom decile of the "without" transfer distribution. Similarly Delgado and Cardoso (2000) find an increase in income of about 16% in households receiving pension. 41 Figure 33. Number of Households by Consumption Decile, Northeast Rural 160- ,,> 1 4 o 140 rTotal expenditure o 12 P OExpenditure minus transfers I o l 50 80 4- 60 E 40 z 2 01 1 2 3 4 5 6 7 8 9 10 Incom e decile Source: PPV and authors' calculations. 4.8 The analysis implicit in figure 33 is too simplistic to capture reality in two ways. First it assumes that people would behave the same way with and without the pension system; that is, they would not depend on family ties for transfers and would not work longer or harder into their retirement years. Fully accounting for such changes would reduce the estimate of the amount many recipients would net. However, the welfare of people depends on more than money and includes the extra work and obligations to family members that a reduced pension would require. The calculation of the direct monetary effect is a relevant measure of welfare, and it is clear that the money transferred makes a very large contribution to the amount of income available to very poor people. 4.9 The paradox is that the program, which is not particularly well targeted to the poor in that the majority of funds do not reach them, still substantially ameliorates poverty. This result is, once again, due to the extreme skewness of the income distribution in Brazil. Very small percentage changes at the high end of the consumption distribution represent very large percentage changes at the low end. Reforms that do not affect the richest people in Brazil cannot hope to make significant amounts of money available for redistribution. The distribution is so skewed that this is the only source of money. Policy reform must grapple with the need for sacrifice from the richest and most politically powerful groups. A more optimistic interpretation is that relatively large changes in the design of existing programs, as far as the ability to change the amounts transferred to the very poor, can be made within the same budget as current programs-small changes in total costs translate into larger redistributive effects. This dependence of total fiscal effect on the incidence of the program on the rich would appear in all programs based on cash payments. 4.10 In summary, the pension system has been an important aid in ameliorating poverty in the northeast, but the poverty-reducing effects have neither led to a major fiscal drain nor a net transfer of income to the region. Current reforms are to be welcomed, but the pension system cannot replace social services or other methods of helping the poor. B. Coping With Droughts 4.11 Even though droughts are a regular feature in the northeast, the rural economy is slill notfully drought resistant either in production or consumption Agricultural output between drought and non- 42 drought years is still volatile (see table 3). Furthernore, droughts do not just result in a drop in local output due to climatic variations but continue to generate significant socioeconomic dislocations, particularly for the rural poor. At a more general level, a reliable system of coping with droughts involves of three elements: (1) Investment in infrastructure and techniques for managing the rural economy to avoid or reduce the impact of droughts on rural output. This often requires coordinated effort on a regional scale across political and administrative jurisdictions. (2) An intelligent and well-planned intervention program based on how the entitlements of vulnerable sections of the population may be threatened and can be protected. The focus of attention in transfer and distribution programs shifts from a geographic region to a household as the appropriate unit of analysis and targeting. (3) A mechanism to ensure that an early decision to act is taken by responsible authorities in the event of a crisis. This part of the system has an important political dimension. 4.12 The recent drought in northeast Brazil was the worst in 15 years, affecting more than 10 million people in eight northeastern states. A massive relief effort is underway. The three main components are (i) targeted food distribution, (ii) a workfare program in which participants must work or attend training to obtain benefits, and (iii) a subsidized credit scheme. 4.13 The data currently available do not permit a rigorous evaluation of the drought relief program. Some key determinants of the gains to drought-affected families simply cannot be measured. The information system for some key components of the drought relief effort is thus less than ideal, reflecting in part the urgency of the situation and the logistic complexity of the operation. What follows is a partial assessment of the current relief effort (for more detail on each point see appendix 6). Targeted Food Distribution 4.14 No data are available for assessing how well-targeted the food distribution is, although there is nothing to suggest serious misallocation. The provisions of basic necessities, unprocessed or in the form ofprepared meals are unlikely to appeal to the non-poor. Workfare 4.15 The choice of the wage rate is critical to the success of workfare programs. The wage rate determines who wants to participate, and (with the budget allocation) how many can actually be accommodated on the program. A good rule of thumb is that the daily wage rate should be no higher than the market wage for similar work at normal times (about R$5 per day in northeast Brazil versus the statutory minimum wage of approximately R$5.50 per day) This helps the program reach the poor. It will also protect work incentives. Note that the workfare wage is probably well above the shadow wage rate during the drought, implying sizable transfer benefits to participating families. (In 7As Sen (1981) points out, when an individual household's "entitlement" (that is, its ability to acquire food through the legal means available in a society) is eroded because of a fall in asset ownership (crops, livestock, property, job, etc.) its members will, if not protected by some form of social security, face starvation no matter what the prevailing food situation may be. 43 one drought municipality, Pao de Azucar in Alagoas, the reservation wage rate was less than R$2.50 per day at the margin, less than one-half of the workfare wage.) 4.16 Given the available budget, the workfare jobs have to be rationed at the wage rate currently paid. Drought-affected municipalities are identified on the basis of rainfall data. Current coverage is probably about 60-70 percent of the number of workers in drought-affected areas who want work at this wage rate. The current eligibility rules are undoubtedly excluding non-negligible numbers of poor families in need of assistance (for types of groups possibly excluded see appendix 6). In terms of the overall efficacy of the drought-relief operation, there is a strong case for relaxing current restrictions on eligibility and geographic targeting to assure wider coverage to those needing help. Subsidized Credit 4.17 Banco do Nordeste do Brasil (BNB's) drought loans operate under similar rules to its general lending, which continues during the drought, but drought loans have more favorable terms. The program is underfunded, in part, because of the generous terms and there is a large unmet demand for credit at this time. An expansion in aggregate credit availability would help. The BNB reports that 60 percent of the drought loans go to "family farms." It is clearly implausible that all "family farms," as defined by BNB, are poor by Brazilian standards. And very few of those among the other 40 percent of borrowers are likely to be considered poor. This would not be a concern if the loans were not so heavily subsidized. A better allocation of creditfor drought relief would be possible by putting a ceiling on the loans, at (say) the current average loan size, reducing the subsidy, and expanding the lending volume. Together these steps could help BNB meet the demandfor credit during a drought. Unfinished Business for the Future 4.18 While the 1998 drought relief effort has been exemplary in a number of respects, and has undoubtedly saved many families in Brazil's northeast from destitution or worse, there are three main areas where the relief effort could be improved. (1) Preparedness and speed of response. This is often a weak point of drought relief efforts and this drought was no exception. (2) Coverage of the affected population. The overall coverage of the 1998 drought relief effort is less than ideal. (3) Concerns coordination of drought relief efforts with antipoverty policy in normal years. There have been some attempts to coordinate the drought relief with other pre-existing programs. However, these efforts have been ad hoc and partial. 4.19 More systematic coordination efforts must start from the realization that drought is intimately connected to the problems of rural development more generally: high risk, credit and insurance market failures, underinvestment in local public goods, and often weak local institutions. 4.20 A permanent safety net program in the northeast could help deal with all three problems. This would extend the coverage of the workfare component of the current drought relief effort to include nondrought times (when demand would be much lower, but almost certainly not zero). It would also relax the current eligibility restrictions on relief work. Such a program could thus combine the best features of a low-wage employment guarantee scheme with current social funds for supporting labor- 44 intensive community projects in poor rural areas. To cover the variability in disbursements, a central safety-net fund could be established by a center that makes regular payments sufficient to cover a normal sequence of good and bad years in agriculture, as well as the likely demand in nondrought years. 4.21 Brazil-particularly the northeast-has a rich portfolio of experiments and experiences in trying to mitigate the worst consequences of droughts such as famines. Major infrastructure projects have been undertaken in the past to improve water distribution and access, and a number of large projects are currently under development, for example in Ceara and Bahia, to connect different watersheds and improve water management. However, as noted above, the early warning systems, the preparedness to respond more smoothly to an unfolding crisis, and the ability to better target recipients of emergency aid (e.g., through means-tested programs) need improvement. It would be useful to reexamine Brazil's own experience and to revisit some international experience (see appendix 6) to identify overlooked or potentially better interventions. Continued vulnerability of the economy and the rural poor in the northeast to droughts will adversely effect the perceptions of the investors and educated laborforce the region is trying to attract and retain. 45 Chapter 5. Conclusions 5.1 Notwithstanding relatively high growth in the past two years, the northeast lags behind the rest of the country and there does not appear to have been much convergence over time. Therefore there is no room for complacency with regard to overall growth prospects. Over the long term, it is the growth in average incomes that will yield a sustainable reduction in regional poverty. In addition, as long as the per capita GDP gap between the northeast and the rest of Brazil persists at its current level there is likely to be continued pressure for net outmigration from the region. In conjunction with other observations in this paper it should be clear that a much bolder and more creative regional development strategy is required to close the per capita income gap between the northeast and the rest of Brazil, particularly the Southeast. The previous emphasis on creating an "autonomous center" of development in the northeast is, in principle, an appropriate approach. However, it was focussed too narrowly on manufacturing expansion, and much of that in the form of capital-intensive, enclave investments. Instead the focus should be on creating more dynamic urban centers as engines of growth with a diversified economic base that is more directly linked to local comparative advantage. 5.2 How does public expenditure fit into this story? Using data generated by Instituto Brasileiro de Geografia Estatistica (IBGE), Instituto de Pesguisa Economico-Social Alicada (IPEA) and Superintendencia do Desenvolvimento Econ6mico do Nordeste (SUDENE) a principal determinant of regional growth appears to be private investment. Public investments do not appear to effect growth directly, but only indirectly in so far as they stimulate private investment. Further, our analysis indicates that certain types of public infrastructure investments that lower local cost of production in the region (such as electric energy and water supply) are associated with increased private investment. Public infrastructure investments that increase the integration of the region with the rest of Brazil (such as transportation and communication) have mixed effects. They enable producers from outside the region to compete in the region just as they enable local producers to market in the rest of the economy. Local consumers benefit from the drop in the natural tariff barriers but sometimes local producers do not. Public investments in state enterprises as a whole do not encourage private investment. To a large extent, these results confirms a more general principle that public investment should be complementary to private investment by providing public goods or overcoming other market failures rather than substitutingfor private activities. 5.3 Any attempt to be more specific about the nature of public investments that will best encourage broad-based growth depends on the time horizon one wants to emphasize. In the long run, as the economy develops, the role of agriculture can be expected to decline both as a share of the regional output and of regional employment. Suggested strategies include the following. (1) The expansion and improvement of education are crucial. Ensuring that the goal of universal primary education becomes a reality is essential not only to reach the poorest with a valuable public service but also to facilitate the transition away from agriculture. Currently, there are large numbers of migrants from the rural northeast in other parts of the country. Their level of education is higher, on average, than people remaining in the rural areas. This is evidence that greater mobility and flexibility in the labor market is achieved by education. Furthermore, analysis in this report corroborates evidence in many studies that earnings are considerably higher for the better educated. The one exception to this rule is that people operating their own farms do not appear to have higher incomes as a result of more education. Everyone else, however, benefits from greater opportunities and higher productivity. An important policy implication of higher earning capacity from education resulting from improved opportunities and mobility is that funding for education needs to have a substantial federal component. The results point to a possible interregional externality, i.e., that the returns to state investments in education may well accrue to other states. The country as a whole has an interest in maintaining educational levels in the region. 46 (2) Infrastructure that supports the creation of off-farm employment opportunities will be necessary. The balance of investment may turn toward road and communication links and away from irrigation and other investments directed toward agriculture. (3) Infrastructure requirements for guiding urban development are also of high priority as the decline in agriculture leads to a further urbanization ofpoverty. One important aspect is the provision of safe water and sanitation. Analysis presented below shows enormous positive effects of access to these services on health, access that is currently denied only to the poorest. This is of highest priority given that it is relatively easier and cheaper to supply these services in urban areas. 5.4 On the other hand, in the short run agriculture will remain a prominent feature of the economy. Further, even when the economy as a whole shifts away from agriculture, there will still be concentrations of the poor among small-holders and the landless working in what remains of the sector. In this context, the policy options for alleviating poverty are more difficult. Suggested strategies include the following: (1) Public investments in support of agriculture are natural candidates. Continuing support for water development projects, rural roads to improve marketing opportunities, and the completion of rural electrification are all likely to be important. However, the impact on poverty from the development of agriculture is not clear. Increases in agricultural output that result from improved yields and productivity will only improve the condition of the poor to the extent that they increase wages for unskilled people or increase the profitability of land owned by the poor. The impact of public investments that increase productivity in agriculture on labor demand and, therefore, wages is not clear. Technical progress in agriculture varies in its effects on labor demand but does not usually increase demand for unskilled labor. So, while there is a role for public investment on efficiency grounds, the effect on poverty is more dependent on the ownership of improved land by the poor rather than employment. (2) Land reform becomes impossible to avoid. If labor demand cannot be expected to grow substantially, the only way that poor people who remain in agriculture can increase their income is to own more productive assets. As argued above, technical progress in the sector will only help the poor if they benefit from the extra profitability of farning through ownership of land. And in the absence of technical progress, more equitable ownership of land is even more critical. Whether this requires much extra public expenditure is unclear. Previous policies on land reform have wasted substantial sums of money purchasing or rather expropriating low-quality land. More market-oriented approaches will place the government in the role of facilitator rather than purchaser of land. The emphases should be on the appraisal and choice of parcels by the prospective owner who will be in a better position to know the details of land quality and be a better bargainer than will a bureaucracy. (3) Education remains important but will require changes in emphasis. As noted above, the analysis in this report casts some doubt on the effectiveness of current forms of education on the productivity of farmers. One consequence of improved technology in agriculture, however, will be a need for better farm management practices and, therefore, of more educated farmers. There will be a need to reorient education to satisfy this need. (4) The greater and more varied demands of infrastructure, requiring knowledge of local conditions, suggests one particular form of expenditure. Social funds relying on community support and participation have been a successful contributor to rural society and economies. Ceara's Sao Jose program and similar schemes in other states (sometimes supported by the World Bank's PAPP program) have been able to mobilize community resources, identify investments which respond to felt 47 local needs and have brought about substantial changes in the way communities see the future.8 This conclusion is not based solely on impressions but has some support from research. At least in the area of rural water projects, community participation has been shown to be a critical factor in their success. (Narayan and Pritchett 1996). Investments requiring cooperation among beneficiaries for their operation or routine maintenance appear to gain most from the community's participation in their selection, planning, and construction. Water projects fit this description, as do construction of schools and clinics. Ceara's experience with health services is further evidence. 5.5 Some types of public expenditures are required regardless of the time horizon. This includes infrastructure needs common to agriculture and nonagricultural activities, as well as education (including the reorientation of the curriculum for farmers). However, how much public resources should be directed to agriculture-specific investments requires a better fix on the likely role agriculture will play in the future. 5.6 Regardless of the direction that agriculture in the region takes in the future, there are immediate measures that help now. Streamlining state administration costs will reduce deficits and support national fiscal policy. Reorienting public programs away from ensuring employment and toward delivery of effective services would help. Finding out what is happening among the poor, especially in regular monitoring of living conditions and the impact of public health policies would also help. In fact, that may be the crux of the matter. This report has highlighted several important areas, but any particular suggestion may work or fail to work. The most important goal is to reduce poverty and guide public expenditures to their most effective use. This requires regular feedback about the evolving effectiveness of different programs and the willingness to act on this information. 8 Two comments heard on a visit to a Sao Jose village are instructive. One, by someone originally from rural Ceara but unfamiliar with the program, was that the village members seemed to have a clear set of priorities for future investments, knew what the next steps were and how to go about doing them. This contrasted starkly with traditional characterizations of the rural poor as fatalistic and passive. She thinks this is a genuine change from the program (and not a shortcoming of the traditional characterization). The second, by someone much more familiar with the program, was that there have been second-round social effects of the program. It seems that a common sight in the halls of state government are representatives from Sao Jose villages demanding, for example, that the schools they built with Sao Jose funds be staffed, as is their right, by Department of Education teachers. Such demands are also common in public works, health and other agencies. This public monitoring of government, too, was unheard of before the introduction of the program. 48 Bibliography The word "processed" describes infornally reproduced works that may not be commonly available through library systems. Arraes, R. A., and I. Castelar. 1989. "The Effect of Drought on Public Finances in the State of Ceara." In A. R. Magalhaes, and E. B. Neto, eds., Socioeconomic Impacts of Climatic Variations and Policy Responses in Brazil. United Nations Environment Programme/SEPLAN-CE, Fortaleza. Ayub, M. A., and U. Keffner. 1994. "Water Management in the Maghreb." Finance and Development 31(2, June):28-30. Azzoni, C. R. 1996a. "Economic Growth and Regional Income Inequalities in Brazil: 1939-92." Texto para Discussao Interna No. 06/96. Programa de Seminarios Academicos, FEAIUSP. Azzoni, C.R. 1996b. "Recent Trends in Regional Competitiveness and the Future of Industrial Concentration in Brazil." Processed. Baer, Werner. 1995. The Brazilian Economy: Growth and Development. 4th ed. Westport: Praeger. Barro, R. and Sala-I-Martin, X. 1995. Economic Growth. New York: McGraw-Hill. Batt, Rosemary. 1990. "Organizational Strategies for Development: A Case Study of Bahia, Brazil." World Bank Department of Urban Studies and Planning, MIT for the Studies and Training Design Department, Economic Development Institute, Washington, D.C. (July). Benabou, R. 1996. "Heterogeneity, Stratification, and Growth: Macroeconomic Implications of Community Structure and School Finance." American Economic Review 86(3):584-600. Binswanger, H. P. 1978. "Attitudes Towards Risk: Experimental Measurement Evidence in Rural India." American Journal of Agricultural Economics 62(3, August):395-407. Borsch-Supan, A. 1989. "The Role of Education: Mobility Increasing or Mobility Impeding? " National Bureau of Economic Research, Working Paper No. 2329, Cambridge, Mass. Baumol, W., S. Blackman, and E. Wolff. 1989. Productivity and American Leadership. Cambridge: MIT Press. Carneiro, F., and l. S. Gill. 1997. "Effectiveness and Financial Costs of Voluntary Separation Programs in Brazil: 1995-1997." Economic Note No. 25. World Bank, Latin America and the Caribbean Region, Country Department I, Washington, D.C. Chen, E. 1997. "The total factor productivity debate: determinants of economic growth in East Asia." Asian Pacific Economic Literature 11(l):18-38. Collins, S., and B. Bosworth. 1996. "Economic Growth in East Asia: Accumulation versus Assimilation." Brookings Papers on Economic Activity 2:135-203. Clements, B. J. 1991. "Growth Strategies, Employment, and Income Distribution in Brazil: An Input Output Assessment." International Monetary Fund Working Paper WP/91/122. Washington, D.C. Datt, G., and M. Ravallion. 1994. "Transfer Benefits from Public Works Employment: Evidence for Rural India." Economic Journal 104(427):1346-69. Delgado, G., and J.C. Cardoso Jr. 2000. "Principais resultados da pesquisa domiciliar sobre a previdencia rural na regiao sul do Brasil: projeto avaliacao socioeconomica da previdencia social rural." TEXTO Para Discussao No. 734. June. Rio de Janeiro, IPEA. Deverajan, Shanta, and Jeffrey Hammer. 1998. "Risk Reduction and Public Spending." Policy Research Working Paper No. 1869. World Bank, Washington, D.C. Dillinger, William. 1997. "Brazil's State Debt Crisis: Lessons Leamed." Country Department I Latin American and the Caribbean Region Economics Notes No. 14. World Bank, Washington, D.C. Diniz, C. C. 1994. "Polugonized Development in Brazil: Neither Decentralization nor Continued Polarization. International Journal of Urban and Regional Research 18:293-314. Diniz, C.C., and M. Razavi. 1993. "Emergence of New Industrial Districts in Brazil: Sao Jose dos Campos and Campinas Cases." Working Paper. Universidade Federal de Minas Gerais, CEDEPLAR, Belo Horizonte, Brazil. 49 Dreze, Jean. 1995. "Famine Prevention in India." In Dreze, Sen, Hussain, eds., The Political Economy of Hunger: Selected Essays. Oxford: Clarendon. Easterly, William Russell, and Deborah Wetzel. "Policy Determinants of Growth: Survey of Theory and Evidence." Policy, Planning, and Research Working Paper No. 343. Country Economics Department, World Bank, Washington, D.C. Fernandes, Ana C. 1993. "Stabilisation, Exports and Regional Development in Brazil: The Northeast in the 1980s." Geography Laboratory, University of Sussex, Brighton, England. Processed. Ferreira, F. H. G., and J. A. Litchfield. 1997. "Education or Inflation: The Roles of Structural Factors and Macroeconomic Stability in Explaining Brazilian Inequality in the 1980's." Processed. Filmer, D., J. S. Hammer, and L. Pritchett. 1998. "Health Policy in Poor Countries: Weak Links in the Chain." Policy Research Working Paper No. 1874. World Bank, Development Economics Research Group, Washington, D.C. Food and Agriculture Organization (FAO). 1994. Early Warning Prevents Famine in Drought-Stricken Southern Africa and Horn of Africa. FAO's Emergency Activities. http://ftp.fao.or2,/FOCUS/E/disaster/casestud/C5horn.htm). Foster, Andrew, and M. Rosenzweig. 1995. "Learning by Doing and Learning from Others: Human Capital and Technical Change in Agriculture." Journal of Political Economy 103(6):1176-1209. Friedman, Milton. 1957. "A Theory of Consumption Function." General series, No. 63. A study by the National Bureau of Economic Research, New York. Frischtak, C., D. Leipziger, and F. Normand. 1996. "Industrial Policy in MERCOSUR: Issues and Lessons." Economic Note No. 10. World Bank, Latin American and the Caribbean Region, Country Department 1, Washington, D.C. Glantz, M. H. 1996. "Are Famines so Difficult to Predict?" Internet Journal of African Studies, No. I (April). http://www.brad.ac.uk/research/iias/ijas.htm Goldsmith, W. W., and R. Wilson. 1991. "Poverty and Distorted Industrialization in the Northeast." World Bank, World Development Report 19(5):435-55. Washington, D.C. Government of Brazil. 1968. "Terceiro Plano Diretor de Desenvolvimento Econ6mico e Social do Nordeste, 1966- 1968." Superintendencia do Desenvolvimento do Nordeste. Recife. Hamer, A.M. 1985. "Decentralized Urban Development and Industrial Location Behavior in Sao Paulo, Brazil: A synthesis of Research issues and Conclusions." World Bank, Washington, D.C. Nijkamp, Peter, ed. 1986. Handbook of Urban and Regional Economics. 1986. Vols I and 11. New York: North- Holland. Hansen, N. 1972. "Criteria for a Growth Centre Policy." In Antoni Kuklinkski, ed., Growth Poles and Growth Centers in Regional Planning. The Hauge: Mouton. Higgins, B. H., and D. J. Savoie. 1995. Regional Development Theories and Their Application. New Brunswick: Transaction Publishers. IPEA. 1998. "Migration from the northeast: Who is Leaving? Jalan, Jyotsna, and Martin Ravallion. 1998. "Transfer Benefits from Workfare: A Matching Estimate for Argentina's Trabajar Program." World Bank, Development Research Group, Washington, D.C. Processed. Li, H., L. Squire, and H. Zou. 1998. "Explaining International and Intertemporal Variations in Income Inequality." The Economic Journal 108(January):26-43. Lopez, R., and A. Valdes. 1997. "Rural Poverty in Latin America: Analytics, New Empirical Evidence, and Policy." Report No. 16792-LAC. World Bank, Latin America and the Caribbean Region, Technical Department, Washington, D.C. Lynde, C., and J. Richmond. 1993. "Public capital and Long Run Costs in UK Manufacturing." The Economic Journal 103(July):880-93. Magalhges, A. R. 1988. "The Effects of Climatic Variations on Agriculture in Northeast Brazil." In M. L. Parry, T. R. Carter, N.T. Konijn, eds., The Impact of Climatic Variations on Agriculture. Vol. 2: Assessments in Semiarid Regions. 50 Magalhaes, Antonio R., and Michael H. Glantz. 1992. "Socioeconomic Impacts of Climate Variations and Policy Responses in Brazil." UNEP, Secretariat of Planning of Ceara, Esquel Foundation. Brasilia. Markusen, Ann. 1994. "Interaction Between Regional and Industrial Policies: Evidence from Four Countries." Proceedings of the World Bank Annual Conference on Development Economics, Washington, D.C. Murray, C., G. Yang, and X. Qiao. 1992. "Adult Mortality: Levels, Patterns and Causes." In R. Richard, A. Feacham, T. Kjellstrom, C. J. L. Murray, M. Over, and M. A. Phillips, eds., The Health ofAdults in the Developing World. Musgrove, P., and 0. Galindo. 1988. "Do the Poor Pay More? Retail Food Prices in Northeast Brazil." Economic Development and Cultural Change 37(October):91-109. Paes de Barros, R., C. Corseuil, and Rosane Mendoca. 1998. "Poverty, Inequality and Macroeconomic Instability." IPEA, Serie Seminarias No. 03/98. -----. I.S. Gill, M. Foguel, and Rosane Mendoca. 1997. "Labor Market Prospects of Public Employees in Brazil: An Empirical Evaluation." Economic Note No. 24. World Bank. Latin America and the Caribbean Region, Country Department 1, Washington, D.C. Pritchett, Lant. 1997. "Divergence, Big Time." Journal of Economic Perspectives I l(summer):3-17. Ravallion, Martin. 1998. "Appraising Workfare Programs." Policy Research Working Paper 1955. World Bank, Development Research Group. Washington D.C. Ravallion, Martin, and Gaurav Datt. 1992. "Growth and redistribution components of changes in poverty measures." Journal of Development Economics 38: 275-95. Richardson, H. W., and P.M. Townroe. 1986. "Regional Policies in Developing Countries." In Peter Nijkanp, ed., Handbook of Regional and Urban Economics, Volume 1. Amsterdam: Elsevier. Roberts, J. T. 1995. "Trickling Down and Scrambling Up: The Informal Sector, Food Provisioning and Local Benefits of the Carajas Mining Growth Pole in the Brazilian Amazon." World Bank, World Development Report, Washington, D.C. Rocha, S. 1992. "Pobreza Metropolitana: Balanco de Uma Decada." In Perspectivas da Economia Brasileira. Rio de Janeiro: INPES, Instituto de Pesquisas. ---------. 1997. Crise, estabilizacao e pobreza - 1990 a 1995. Conjuntura Econdmica, Janeiro. Romer, David. 1996. Advanced Macroeconomics. New York: McGraw Hill. Sakashita, N. 1967. "Regional Allocation of Public Investment." The Regional Science Association Papers 19: 161-182. Savedoff, William D. 1995. "Wages, Labour and Regional Development in Brazil." Vermont: Avebury." Sen, A.K. 1981. Poverty and Famines. Oxford: Clarendon. Shah, Anwar. 1991. "The New Fiscal Federalism in Brazil." World Bank Discussion Paper 124. Washington, D. C. Shell, K., ed. 1967. Essays on the Theory of Optimal Economic Growth. Cambridge: MIT Press. Shome, Parthasarathi, and Paul Bemd Spahn. 1996. "Brazil: Fiscal Federalism and Value Added Tax Reform." NIPFP Working Paper No. 96/11. National Institute of Public Finance and Policy, New Delhi. Squire, L., and H. F. Zou. 1998. "Explaining International and Intertemporal Variations in Income Inequality." The Economic Journal 108(446):26-43. SUDENE (Superintendencia do Desenvolvimento Econ6mico do Nordeste). 1996. "Agregados Econ6micos Regionais: Nordeste do Brasil, 1965-95." Recife Tendler, J. and S. Freedheim. 1994. "Trust in a Rent-Seeking World: Health and Government Transformed in Northeast Brazil." World Development 22(12):1771-91. Tendler, Judith. 1993. "New Lessons From Old Projects: The Workings of Rural Development in Northeast Brazil." World Bank, Operations Evaluation Department, Washington, D.C. Processed. Thomas, Vinod. 1987. "Differences in Income and Poverty within Brazil. World Development Report 15(2):263- 73. Washington, D.C. 51 Walker, P. 1989. Famine Early Warning Systems. London: Earthscan. Wexler, R. 1994. "Regional Development and Unrestrained Urbanization: Two Case Studies from the Brazilian Amazon." LBJ School of Public Affairs, University of Texas at Austin. Wildasin, D.E. 1998. "Fiscal Aspects of Evolving Federations: Issues for Policy and Research." Policy Research Working Paper 1884, World Bank. Wood, Charles H., and Jose Alberto Magno de Carvalho. 1988. "The Demography of Inequality in Brazil." Cambridge University Press. World Bank. 1987. "Brazil: Industrial Development Issues of the Northeast " World Bank Economic and Sector Report. Washington, D.C. -. 1990. World Development Report: Poverty. Washington, D.C. ---------. 1991. "Dynamics of rural development in Northeast Brazil: New Lessons From Old Projects." Operations Evaluation Department Study 10183. Washington, D.C. ---------. 1993. The East Asian Miracle. New York: Oxford University Press. ---------. 1996a."The Changing Fortunes of Subnational Regions: Regional Development Policies and Performance in Eight Major Countries Around the World." East Asia Department, Washington, D.C. Processed. ---------. 1996b. "Brazil Poverty Assessment." Report No. Br- 14326. Latin America and the Caribbean Regional Office, Brazil Country Management Unit, Washington, D.C. ---------.1998. "Brazil Social Spending in Selected States." Report No. Br- 17763. Latin America and the Caribbean Regional Office, Brazil Country Management Unit, Washington, D.C. ----------. 1998b. "Brazil: Managing Pollution Problems-The Brown Environmental Agenda." Report No. 16635- BR. Washington, D.C. Yang, D. L., and H. Wei. 1996. "Rising Sectionalism in China?" Journal of International Affairs. Young, A. 1992. A Tale of Two Cities: Factor Accumulation and Technical Change in Hong Kong and Singapore. NBER Macroeconomics Annual. Cambridge: MIT Press. Young, A. 1994. "Lessons from the East Asian NICs: A Contrarian View. " European Economic Review 38:964-73. Young, A. 1995. "The tyranny of numbers: confronting the statistical realities of the East Asian growth experience." Quarterly Journal of Economics 1 10(3):641-80. Zellner, A. 1962. "An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias." Journal of the American Statistical Association 57:348-68. World Bank Projects Rural Poverty Alleviation Program in the Northeast World Bank. 1996. Brazil: Northeast Rural Poverty Alleviation Program: Rural Poverty Alleviation Project- Pernambuco. SAR 15935-BR, 11/20. World Bank. 1996. Brazil: Northeast Rural Poverty Alleviation Program: Rural Poverty Alleviation Project- Piaui. SAR 15934-BR, 11/20. World Bank. 1996. Brazil: Northeast Rural Poverty Alleviation Program: Rural Poverty Alleviation Project - Rio Grand. SAR 15933-BR, 11/20. World Bank. 1996. Brazil: Northeast Rural Poverty Alleviation Program: Rural Poverty Alleviation Project - Rio Grande do Norte. Memo of the President P6979-BR, 11/20. World Bank. 1995. Brazil: Northeast Rural Poverty Alleviation Program: Rural Poverty Alleviation Project - Bahia. SAR 14390-BR, 6/6. World Bank. 1995. Brazil: Northeast Rural Poverty Alleviation Program: Rural Poverty Alleviation Project - Ceara. SAR 14395-BR, 6/6. 52 World Bank. 1995. Brazil: Northeast Rural Poverty Alleviation Program: Rural Poverty Alleviation Project - Sergipe. SAR 14394-BR, 6/6. Rural Development Projects World Bank. 1987. Brazil: Northeast (Alagoas) Rural Development Project. SAR 6668-BR. 5/26. World Bank. 1987. Brazil: Northeast (Minas Gerais) Rural Development Project. SAR 6669-BR. 5/26. World Bank. 1987. Brazil: Northeast (Paraiba) Rural Development Project. SAR 6670-BR. 5/26. World Bank. 1987. Brazil: Northeast (Maranhao) Rural Development Project. SAR 6136-BR. 5/26. World Bank. 1986. Brazil: Fourth Northeast (Piaui) Rural Development Project. SAR 6117-BR. 9/26. World Bank. 1986. Brazil: Northeast (Ceara) Rural Development Project. SAR 6135-BR. 9/26. World Bank. 1986. Brazil: Fifth Northeast (Pernambuco) Rural Development Project. SAR 6118-BR. 5/22. Irrigation World Bank. 1990. National Irrigation Program: First Northeast Irrigation Project. SAR 8038-BR. 1/19. Land Tenure World Bank. 1992. Brazil: National LandAdministration Program: Northeast Regional Land Tenure Improvement Project. PCR 10710-BR. 6/8. Education World Bank. 1994. Brazil: Third Northeast Basic Education Project. SAR 11959-BR. 10/29. World Bank. 1993. Brazil: Second Northeast Basic Education Project. SAR 11298-BR. 4/20. World Bank. 1993. Federative Republic of Brazil, Third Northeast Basic Education Project. SAR 11959-BR. Health World Bank. 1989. Brazil: Second Northeast Basic Health Services Project. SAR 7159-BR. 10/25. World Bank. 1988. Northeast Endemic Disease Control Project. SAR 6837-BR. 1/19. World Bank. 1986. Brazil: Northeast Basic Health Services Project. SAR 6088-BR. 4/30. Water Supply and Sanitation World Bank. 1987. Northeast Water Supply and Sewerage Project. SAR 6968-BR. 6/29. World Bank. 1991. Brazil: Northeast Urban Flood Reconstruction Project. PCR 9491-BR. 4/12. World Bank. 1995. Republic of Tunisia-Poverty Alleviation: Preserving Progress While Preparing for the Future. Vols. I and II. Washington D.C. 53 Appendix 1. Volatility of Income in the Northeast A. Possible Explanations For the Decline in Volatility of Regional Income Chapter 2 argues that the observed fall in volatility in regional income could be explained in one of three ways: (1) as an artifact of the data, (2) as a reduction in the volatility of agriculture, or (3) as a result of a structural shift away from agriculture in the economy. (1) Statistical aberration? Whereas there are some quality concerns associated with the data for subnational levels of government, they do not support the hypothesis that the observed decline in volatility is an artifact of the data. The series obtained from different sources (various publications of SUDENE, IBGE, BNB, and IPEA) have been compared for accuracy when possible. In fact during the hyperinflationary period northeastern price deflators are considerably different from the national price deflators, contrary to the hypothesis. Thus, the decline in and greater synchronization in the relative volatility of the northeastern economy is likely to reflect developments in the real economy. (2) Reduction in agricultural volatility? The agriculture sector has historically been the most volatile component of GDP in the northeast. The volatility of agriculture reflects the region's historic dependence on crops that are susceptible to droughts and other climatic variations rather than commodity price fluctuations on domestic or international markets. Volatility in agriculture has been much higher than that of aggregate output in the northeast. Even though agricultural volatility was lower from 1975 to 1985 it has reappeared and even increased since then (see figure Al.1). Figure Al.1 Rates of Growth of Agriculture and Aggregate Output, Northeast Brazil 80% 60% 40% 20% 0 % Note: Data in 1980 Reais, shaded areas represent drought years. Source: SUDENE. The data in table ALI suggest that agricultural output in the northeast is not yet drought resistant and that it still affects aggregate output on some if not all occurrences. Al-I Table A1.1 Change in Output Between Drought and Nondrought Years Change in Change in Change in Drought and Change in GDP agriculture Output GDP agriculture nondroughtyear (US$ billions) (US$ billions) (percent) (percent) 1978-1983 1.2 -0.36 19.5 -21.0 1986-1987 -8.9 -1.64 -32.9 -40.5 1992-1993 -3.1 -1.12 -15.1 -36.1 Note: Data in 1992 U.S. dollars. Source: SUDENE. Agricultural volatility exists in all the nine states in the northeast even though it is not uniform across states. Roughly four different patterns of volatility are observed (see appendix IB). These patterns are not strictly correlated with the proportion of semiarid land in each state (see table A1.2) but seem to be a function of the cropping pattern. This is an area requiring more careful analysis. (3) A reduction in the share of agriculture in regional output? The regional economy in the northeast has undergone a major structural shift over the past three decades. The contribution of agriculture to aggregate output in the northeast declined from 29 percent in 1965 to 13 percent in 1983, and has remained more or less at the same level since then, being 16 percent in 1996. Agricultural volatility, however, seems to have increased. This pattern is shared by most individual states (see appendix Id for the changing share of agriculture in GDP in each of the nine northeast states). The share of industrial output has also declined more or less steadily since 1983. As a result the service sector today accounts for two-thirds of regional output (up from less than one-half in the mid-1960s, see figure Al .2). This increased role of services should not be confused with the fact that during drought years, there are significant increases in the size of the services sector created mainly by government responses that increase off-farm incomes and purchase of inputs used in drought- related emergency programs (Arraes and Castelar 1989). Figure A1.2 Structural Composition of the Northeast Economy 80 1' IHyperinflation 60% - >;40% - I d sr Aegricult / 10 0% Al-2 Note: Nominal data used for calculating shares. Source: 1965-1995 SUDENE, 1996 World Bank Mission. Two-thirds of the net increase in the share of services in the past 30 years (or 12 percentage points) is due to the boom in financial services (from 29 percent to 38 percent of the service sector) and the growth in commerce (from 34 percent to 37 percent of the service sector). These two sectors alone now contribute about 75 percent of service output and about 42 percent of overall output in the northeast (i.e., twice the share of agriculture). These figures are surprisingly large for a relatively underdeveloped economy. More disaggregated analysis is required to differentiate between informal and formal service activities, public sector (and particularly administrative services) vs. private sector activities, and a clearer delineation of the share of other sectors in services (communication, education, health, tourism, transport, etc.). However, for our purposes it is sufficient to note the structural decline in the role of agriculture. In summary, the apparent decline in the volatility of aggregate output cannot be attributed either to statistical aberration or to a reduction in the volatility of the agricultural sector itself Rather, the decline in the volatility of regional output is a direct result of the drop in the share of agriculture in regional GDP. B. Patterns of Volatility in Agricultural Output Across States The nine states in the northeast can be grouped into three groups based on the extent of the semiarid land in each state (see table Al .2). Table A1.2 Extent of the Semiarid Region in Northeast States Land area (percent) States 0-9 Maranhlo 10-59 Alagoas, Sergipe 60-100 Bahia, Ceara, Paraiba, Pemnambuco, Piaui, Rio Grande do Norte However, in the following figures, roughly four different patterns of volatility are discernible which are not strictly correlated with the proportion of semiarid land in each state. The first pattern-regular volatility, which has actually increased since the early 1980s-is exhibited by Maranhao. Figure A1.3 Maranhao: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 50% 40% |=~~~~..Agricultuue 20% 10% 0% - -20% .30% -40% Source: SUDENE (1996). AgTegados Econamicos Regionas: Nordeste do BraTIl, 1965-95. Recif Al-3 Maranhao is the only state that is not part of the semiarid drought polygon in the northeast. Agricultural volatility has actually increased since 1980. This is the only case where volatility was higher during the hyperinflationary period (between 1987 and 1993) than before or after, suggesting some sensitivity to price regimes. The agricultural economy is dominated by extensive livestock enterprises and the production of traditional food crops (beans, cassava, maize, and rice) and commercial crops (babassu nuts and bananas). Rice is the principal crop, making Maranhao the second largest producer in the country (the state produces about 5 percent of Brazil's rice production, but our examination of crop value data between 1990 and 1995 does not include rice as having a high output value). Rice production is dispersed among a large number of small productive units. Another subsistence crop of relative importance is corn, of which 47.8 percent of total volume is produced in 10 percent of the municipalities. Manioc, a traditional crop in the state, has been losing importance. The production of soybean, which has been growing at high rates in Maranhao, is concentrated in the southern part of the state and destined mainly for export. With the exception of soybean, the production of food crops has been very volatile. The large increase in agriculture production between 1993 and 1994 is mainly the result of high levels of banana (US$2 million in 1993 to US$12 million in 1994) and cassava (US$ 22 million in 1993 to US$ 38 million in 1994) output. The second pattern-three major and broadly spaced spikes-is exhibited by the states of Ceara, Paraiba, Piaui, and Rio Grande do Norte. FigureAl.5faibw Anmal Rate of GrowthofAgcWuae nd Aeg3Outpot1971-96 Figure AlA Ceara: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 I~~ * n1 l _ _ _ _ _ _ _ _ _ __ V l SV 'I S027-0 SUDENE (1996) Agregodos Econonscos Regionos: Nordeste do Brai. 1965-95 RedSe - S006E1S7 ,0fr.O 5&07e. 5l07e r Figure AL7 Rio Grande do Nte Anul Rate of Cauwth of Figure Al.6 Piaui: Annual Rate of Growth of Agriculture and A, and AMgete Output 971-96 Aggregate Output, 1971-96 4030 W%~~~~~~~~~~~~~~~~~~~~~~3 20%~~~~~~~~~~~~~~~~~~~~~~~0 0%~~~~~~~~~~~~~~~~~~~~0 41X Source: SUDENE (1996). Agegados Es-,onoioo Reganr. Nodesle do Brsil, 1965-95. Rdrde soaue sLiENE (1ES). Agadbs E Dcos Plgonas: Nudeste do Basl, 1965-95. Al-4 These four states have more than 60 percent of their land in the semiarid region. Cotton is grown on a large part of the agriculture land in these states. However, due to the low quality of the northeast crop, the high proportion of land used to grow cotton does not translate into high value of output. IBGE data from 1990 to 1995 indicate high volatility in the value of another crop, sugarcane, during the drought of 1992/93 and a phenomenal recovery in the nondrought year of 1993/94. Individual states differ in the structure of their agricultural sector. Piaui is the poorest state in the northeast with the majority of its land being in the "drought polygon." There are large variations in the state's geography with its southeastern part being enveloped by the semiarid region with average rainfall of only 600 millimeters (mm) per year. The western part of the state consists of savannah vegetation and soils with average annual rainfall varying between 800 and 1400 mm. The agricultural sector is the largest employer in the state, occupying about one-half of the economically active population, yet it generates less than one-quarter of the state's GDP. Persistent droughts and the uneven distribution of rainfall, particularly in the semiarid region of the state, make the agricultural sector vulnerable to erratic climatic conditions and cause significant fluctuations in income of many small farms. The primary crops are of the subsistence type: beans, corn, cotton, manioc, and rice, with soybeans being the main cash crop. The entire state of Ceara is classified as part of the "drought polygon" with the limited humid areas being limited to the few fertile highlands. Soil types of medium and low fertility are found in the semiarid and infertile coastal areas. Principal agricultural products include beans, cashew, corn, cotton, and sugarcane. The state has a long coastline that supports a growing tourist industry and extensive fishing industry. The decline in agriculture production in Ceara during 1992-93 is mainly due to a large decline of approximately US$50 million in the production of beans, maize, and sugarcane. Similar to Ceara, most of the land in Paraiba is within the semiarid "drought polygon." The semiarid region covers about 77 percent of the State, and has an average annual rainfall of 500 mm, which is irregularly distributed, with a long dry season of seven to ten months. The coastal part of the state has a more humid climate and fertile soil. Between these two regions lie an elevated highland of fertile soils and a range of transitional zones. Agriculture in the humid coastal zones has been dominated by sugar cane. In the semiarid regions, the larger farmers engage in extensive livestock grazing, while their tenants and owners of small farms grow cotton and basic food crops (beans, corn, manioc, and rice). Consistent with regional patterns, crop yields fluctuate greatly due to factors like the impact of erratic climate, the poor natural resource base, inability to store water, intensive cropping on unsuitable lands, and low levels of technology. Rio Grande do Norte, spatially adjacent to the states of Paraiba and Ceara has 90 percent of its land within the semiarid "drought polygon." Only about 20 percent of the state is suitable for cropping purposes. The semiarid conditions and the incidence of drought have had a powerful impact on the state's economy. There has been an expansion of the industrial and service sectors, and a relative decline in the participation of the rural sector in the state GDP, from almost 19 percent in 1970 to about 6 percent in 1994. The agricultural economy is based primarily on crops and livestock, along with fisheries, forestry and extractive industries such as charcoal. Principal crops are bananas, beans, cashew, coconut, cotton, manioc, pineapple, and sugarcane; the importance of livestock production is now almost on a par with cropping due to improved breeding and health and sanitary practices. Large declines in beans, maize, and sugarcane production of over US$26 million were instrumental in the large drop of agriculture production by 48 percent in Rio Grande do Norte Al-5 between 1992 and 1993. Paraiba also had a large decline in agriculture production during this drought period, with production falling by almost 48 percent between 1992 and 1993. Large production losses were in maize, sugarcane, and beans, which together declined by approximately US$45 million from the previous year. The third pattern-three spikes in the last decade-is exhibited by Sergipe. Figure A1.8 Sergipe: Annual Rate of Gwwth of Agriculture and Aggregate Output, 1971-96 40% Vww~0~'mnen Sergipe is the smallest Brazilian state and has three broad zones: (1) the coast with moderate temperatures and rainfall and plentiful, good quality underground water; (2) the inland semihumid area with variable rainfall, and (3) the semiarid area with highly variable rain and prolonged dry periods. About 47 percent of the total land area in Sergipe constitutes part of the "drought polygon" and subsistence farming dominates in this semiarid region. The coastline has infertile soil with no original forests, and the coastal agriculture is dependent on fishing and coconut. The production of oranges makes up about 30 percent of the value of farm agriculture output. 'he fourth pattern-irregular but frequent fluctuations-is exhibited by Alagoas, Bahia, and Pemambuco. Figure A1.9 Alagoas: Annual Rate of Growth of Agriculture and Aggregate Output, 1971-96 Figure Al.10 Bahia: Annual Rate of Growth Figure Al.ll Pernambuco Annual Rate of Growth of of Agtiulture aud Aggregate Output, 1971-96 Agriaulture and Aggregate Output, 1971-96 Al -6 These three states differ from the others in that they all include a significant Atlantic Coast tropical rainfall climatic zone, which helps support sugarcane and cocoa production. Bahia is the largest state in the northeast and contains the most extensive semiarid zone in the region. Nearly 70 percent of the state area falls within the northeast "drought polygon." The eastern and southern sections of the state, with higher rainfall and fertile soils, are the main agricultural and livestock areas. The rural poor are primarily smaliholders, sharecroppers and wageworkers, who depend on a diverse strategy of income-generating activities in subsistence production of beans, maize, manioc, rice, and small livestock. Their well being fluctuates with changes in the agroclimatic conditions. For example, during the drought of 1991-92, value of maize production fell from US$31 million to US$23 million. Cocoa is the most important farm agriculture product in Bahia. In fact, Bahia is Brazil's leading producer of cocoa, accounting for more than 85 percent of cocoa production in the country. The states of Alagoas and Pernambuco are spatially adjacent to each other and both have a coastal region along with part of their land in the semi arid "drought polygon." Growth in agriculture has displayed similar trends in these states. In Pernambuco, the coastal region has abundant rainfall up to 2,000 mm per year, and with fertile soils and it is the main crop-growing region. The Agreste (20 percent of the state) has rainfall of 650 to 900 mm per year with a dry period of up to seven months and the semiarid "drought polygon" includes 70 percent of the state with annual rainfall below 500 mm. The agricultural economy is mainly based on cotton, sugarcane, and livestock. Export crops including grapes, melons, tomatoes, and basic food crops including beans, corn, manioc, and sweet potato. The large decline in agriculture production of about 45 percent during 1992-1993 is mainly due to the volatility of sugarcane, when the value of output fell by US$100 million during the same period. Sugarcane is also the most important crop in Alagoas contributing around 70 percent of agriculture output. The large decline in sugarcane production from US$123 million in 1992 to US$63 million in 1993 led to a decrease of 43 percent in agriculture output. It appears that cropping patterns are very important in explaining agriculture volatility in the northeast. The cropping patterns and output are a function of soil quality and climatic variations but also have been largely influenced by the low level of technical inputs and the skewed distribution of land ownership in the northeast. However, this requires more careful analysis at a disaggregated level There are several data issues that limit our analysis of the contribution of cropping patterns to agriculture volatility in the northeast. Cropping data from 1990 to 1995 were obtained from lBGE and the Agriculture Census. The data are reported either in physical units or in various currency units. For purposes of aggregation and the comparison of these aggregates across states and over time, it was necessary to convert the data to a common unit, constant 1992$ (see appendix IC ). However, even then the cyclical patterns did not correspond to those provided by IBGE. This may be due to the lack of data on the value of livestock and other agriculture products. We also reviewed World Bank Staff Appraisal Reports on the recent generation of poverty reduction projects under the "Northeast Rural Poverty Alleviation Program"(such projects are being implemented in almost every northeast state). In general, these reports start off with a description of rural poverty in Brazil, followed by the northeast, and finally in the state, the projects being implemented. Each report includes an extensive socioeconomic profile of a state. This profile confers broad economic trends and the role of agriculture over time. In short, these profiles identify sources of agricultural employment and the quantity of agriculture output. While these analyses are based on data from the agriculture census (they use employment and socioeconomic information), explicit reference to crop patterns and values are missing. A possible reason for this is that the project analysts faced the same difficulties we did. More work is needed to confirm whether cropping patterns explain the variation in volatility of the agriculture sector in the different northeast states. Al -7 Appendix 1C. Major Aaricultural Products, Northeast Brazil Table A1.3 Permianent Agriculture Products o1n -1201 .1 1t1 l1 199A Cocoa 2,610 5.913 8.444 6.165 *38.641 18.171 Cashew chestnut 5.774 10.776 6.228 1.188 12.024 14.636 Banana 2.908 3.008 4.119 1.923 847 8.532 Sleeve 1.449 1.570 1.872 1.076 5.422 4.727 Orange 316 373 644 462 2.578 1.870 Papaya 212 166 606 775 2.118 1.800 Total 14.994 23.390 22.897 12.190 67.363 53.919 Sergipe Orange 17.968 18.061 20.442 49.928 67,715 78.697 Maracui6 4.462 3.228 1.721 3.717 18.727 33.953 Cocna 3.856 8.314 4.921 3.735 40.236 18.209 Banana 2.580 1834 1.605 1.403 10.195 13.004 Sleeve 823 552 273 323 1.927 4.049 Panava 229 231 310 417 2.094 2.071 Total 30.302 32.452 29.900 59.697 144.053 150.451 Piasui Banana 3.307 2.953 2.706 2.446 11.876 17.862 Cashew 3.420 6.675 3.937 7.000 14.870 17.337 Sleeve 3.251 3.238 3.238 3.691 13.940 12.899 Orange 1.706 1.916 1.671 1.407 7.085 8.803 Cocoa 208 259 237 347 1.172 1.396 Total 14.043 16.264 12.351 15.117 50.777 59.947 Pernamnbuco Fanana 21.165 12.752 10.440 11.903 60.005 70.080 Goiaha 1.112 3.338 1.112 1.028 3.956 6.600 Coffee 1744 1.206 599 1.408 4.759 5.297 Ornaae 1.146 727 846 676 3.288 3.720 Tan.erina 1 555 335 290 2.271 1.294 Cashew 1.096 830 972 572 1.694 1.183 Total 241.689 244.050 262.570 122.527 545.376 705.785 Paraiba 1995 Banana 11.409 5.231 6.490 9.468 71.013 113.922 Coana 2.084 1.556 1.034 2.179 8.464 10.361 Sleeve 1.767 1.684 1.491 1.817 6 9.362 Panava 412 46 0 1.161 6.430 6.311 Sisal Fiber 4.209 2.781 2.131 736 6.597 6.016 Oran-e 1.565 713 1.055 1.219 1.970 2.814 Total 23.835 14.664 14.602 18.612 108.898 155.319 Maranhao Baanana 3.573 3.505 3.756 2-365 12.124 19.282 Orange 2.558 3.914 3.108 1.370 9.670 11.072 CoVcoa 581 612 529 326 2.565 3.710 Cashew 637 454 789 589 2.375 2.487 Manga 412 414 630 559 2.467 2.084 Total 8.880 9.577 9.805 5.620 33.658 42.984 Ceara Banana 9,377 10.637 8.051 5.034 41.111 44.890 Cocoa 6.921 9.885 7.608 4.301 50 42.668 Cashew 14-125 14.333 12.987 7.597 34.495 33.670 Maracuia 910 2.582 1.690 699 4.808 9.867 Coffee 1.290 1.465 1.339 532 8.609 6,908 Total 39.881 47.719 39.164 21.311 109.613 153.669 Bahia Cocoa 111.272 122.616 89.728 70.991 300.516 274.607 Cnffee 27.652 36.806 26.697 10.175 156.968 121.479 Panava 20.327 5.896 12.910 52.391 99.719 120.522 Orange 56.456 20.722 26.458 12.887 169.744 105.617 Banana 20.935 14.785 15.659 12.262 71.661 97.129 Total 286.856 269.543 237.599 217.856 985.886 1.010.401 Alagoas C(co-&da-haia 6.921 3.149 2.960 4.157 11.943 13.340 Banana 3.818 1.738 753 2.047 5 4.685 Maracuia 73 0 105 85 4.139 3.428 Oranae 547 546 452 536 933 1.472 Manaa 530 428 414 471 813 978 Total 11.944 5.861 4.694 7.313 17.836 23.908 Total (1997. JSS) 672,424,753 663,520,491 633,581,917 480,242,008 2,063,459,745 2,356,382,596 Nfote: Data in thousands of 1992 US$. Total (above) resresents total value all aericulture Droducts in the Northeast. Source: IBGE - Producao Ae icola MuniciDal. Sistema IBGE de Recuoeracfo Automntica - SIDRA 97 Al-8 Table A1.4 Temuorarv AMiculture Products PooL,,, 1Q# 1Q07 10790Q 100 . 6o 7004 WaleTnielon 3.789 4.022 4.S06 5M954 23.276 59.141 Suwarcane 16.073 20.193 15.818 6.951 4.139 51.414 Beans 3.523 13.516 12.019 619 44.686 41.485 Cassava 20.953 29.3% 15.607 14.257 2S.632 21.164 Pineannle 3.148 2.979 4.385 2.887 24.268 16.101 Maize 894 5.206 4.313 271 15.893 13.509 Total 53.261 84.096 62.221 3Z361 168.295 223.473 Sergipe Cassava 4.578 &771 13.821 9.591 29.511 26.459 Sowarcane 14.905 9.729 12.029 21.425 26.568 23.880 Beans 4.310 4.310 4.106 1.821 26.947 18.473 Poato 1.627 791 1.160 673 7425 10.803 Maize 2011 3.435 1.479 740 1Z942 10.494 Tobaco 1.391 924 823 2 9.808 5.671 Total 31.247 31.694 37.164 37.551 127.974 105-870 Piaui Cassava 26.340 19.343 20.254 16.502 37 83.372 Rice 13561 25.095 9.454 11280 92.168 71.839 Maize 7.250 13.388 3.833 4.136 48.624 50.466 Beans 14.545 13.118 6.194 5.694 39 36.487 Suearcane 11.539 10.203 8.005 8.489 25.671 23.775 Cotton 1 1.427 749 265 14.729 11.420 Pereamawhuo Suearcane 159.326 156.966 17&712 62.848 210.057 362.652 Watenoelon 1.095 1.073 863 806 4.468 89.463 Bearns 18269 15.078 13.688 5.020 91.463 71.594 Tanato 16.738 18.647 14.482 18.584 100.229 67.193 Cassava 20.279 27.897 34.330 21.564 57.953 47.005 Maize 6.802 8&657 5.239 294 37.947 32.192 Total 41.556 29.080 21298 28.209 95.358 90.914 Paraiba Suaarcane 81.538 43.849 56.967 20.066 80.842 178.412 Pineannle 35.214 17.847 8.079 11.319 80.790 102.940 TMato 2.498 1.593 1A93 783 10.279 40.673 Beans 12198 14.825 17.546 2 5Z.076 39.997 Maize 3.705 7.651 10.831 720 29.773 24.944 Cxoton 1.744 1.952 2.380 362 9.391 9.994 Total 153.390 105.753 113.794 51.036 304.664 425.464 Mlaranhao Cassava 40.786 50.666 42.699 22.747 98.143 169.765 Maize 9.469 16.727 10.912 S.397 48.023 52.253 Suzarcane 18191 10.293 14.681 7.014 30.891 37.149 Beans 9.944 8.371 5-994 5.654 24.838 30.179 .Sov 254 0 2.127 3.001 33.733 26.729 Total 121.604 150.793 109.437 76.592 425.895 332.421 Ceam Beans 20.971 23.525 22.538 8.295 112188 104.517 Maize 9.047 18.149 11.464 1.965 67.536 78.225 SuararCane 17.164 18.130 31.721 12.521 60.542 55.070 Rice 13.443 131957 14.538 8.614 49.299 45.210 (assava 12.524 15.373 22.289 7.440 31.426 44.900 Total 84.402 103.403 118.835 46.799 380.371 372.614 Bahia Mandioca 123.689 112.480 154.418 75.700 364.935 363.374 Smc 13.352 28.200 28.766 29.281 167.906 187.391 Sgnarcane 44.181 55072 35.531 29.065 148272 178.363 Beans 54.714 63.830 81.358 66.718 197.950 142.270 Maize 11.355 31.041 23.490 27.532 102.583 90.432 Total 316.559 357.093 395.316 297671 1.286.792 1.165.267 Alagoas Siuar Cane 132054 118692 123.322 64.158 436 443.239 Bean t2.058 6.607 6.970 3.291 40.919 28.264 (Csasava 8.078 5.912 10269 2.375 13.500 25.140 Tobacco 46.886 1.725 2.037 3.501 11 8.646 Maize 1.762 1649 1.727 347 7038 6&366 Total 207.660 138590 147.803 78.321 72475 526.969 Tntf1J 1SQQ~ jig BeIar&t 1 tUC jf7 n&7Ag 4616 M O6 hue 103 KQei413 343 3n4s 370 60A 1n?3797.677 Note: Doat in thotwands of 1992 iSS. Total (above) reoresents total value all azricutdure rouducts in the northeast Source: IBOE - Producjo Amicola Muicivoal: Sisteoma lBGEde Recutrwao AutonAhca - SIDRA 97. Al -9 D. The Trend in the Share of Agricultural Output in All Northeastern States The states where the share of agricultural output declined significantly in the past 25 years. Figure A1.12 Alagoas: Relative Share of Agriculture and Figure A1.13 CearA: Relative Share of Agriculture and Industry in GDP, 1970-96 Industry in GDP, 1970-96 351F 4091 3C 30 * o~o-0 ,~0% 25sl:__ 150 4 450 C' _ Soooce.SUOENE (l0O0) Agregdros00 Eo,ono.mroOs OdgoOroo. 00-0 Rod oteofBssi,16,5 FigurF Al.14 Paranba : Relative Share of Agiculture Fande -lnd ~ ~ ~ ~ ~ ~ ~ ~ inu Inunr in? GD,19970Arcltr-nd96uyinG t1909 591 1~~~~~~~~~~4 I4 '4 '4g ;o 41 , 1° 4 41 1.1 51.1 1 '.1 1'1 *1e U1'0 14 eq' 1.' 14 1* 1- 1,e 14o .* lcg le .tZF ,.ip 000, 0000(09) Asg,ro ooo.ojo, ~on 0000,000.0.105-OSE19d 0,01, mrRfmu:odeldEulld-9 m S=-c SUDENE (1996b Aogd m RefoE d dS B 195R ~~~~~~~~~~~Figure ALM1 Maranhao: Relative Share of Agriculture FiueA.5RoGande oNre eaieSaeo and ~ 40 Industry in GDP, 1970- giutueadIdutyi DP 909 35'Ai ~ ~ ~ ~ ~ 0 l\ Agg-ud e 30% % t~~~~~~~0 a W~~~~~~~~~0 104~~~~~~~~~0 5%~~~~~~~~~~~~~~~~~~5 ~~~~~~1 1 1 l1 e *1e ', pt? pP l t t1qo.# e 0% p8 q Source ~ ~ ~ ~ ~ ~~ WDN SUDEN (l993 &.-,-d Egi mcomi B-1 E.-9 Rd. ods3 oEaU 96 sR FgrA11PArab1-10 tv SaeofArcltr n The states where the share of agricultural output has declined-but not significantly-in the past 25 years. Figure At17 Balhia Figure A1.18 Penuambuco Relative Share of Agrculture and Industry in GDP, 1970. Relative Share of Agriculture aid Industry in GDP, 197096 23% 25% 20~~~~~~~~~~~~~~~~~~~~~~'% 20%~~~~~~~~~~~~~~~~~~~~~~~~~0 70%~~~~~~~~~~~~~~~~~~~~~~~~5 15%~~~~~~~~~~~~~~~~~~~~~~5 5%~~~~~~~~~~~~~~~~~~~~~~5 5% 0% 0% s-006 SUDENE (1966) A) Armdd6. s R .R. i- N56 - 9596il-S 661.95 4ir Figure Al.19 Piaui Figure A1.20 Sergipe Relative Share of Agriculture and Industry in GDP, 1970-96 Relative Share of Agriculture and Industry in GDP, 1970-96 10%~~~~~~~~~~~~~~~~~~~~~~~~~~0 69 0 '^1~~~~~~~~~4 41P i , S- SUDENE (969) A Ec R6p: N J 6i 6.9 e, [es95 6.i66 Sgg: SUDENE ((996) AXqd 56066,6. Re&gs:. No60* B* -hit 1965.95 60*0. Al-I l Appendix 2. Growth of Income in the Northeast In this appendix, we provide a brief review of the purpose of a "determinants of growth type" analysis and the type of policy implications that could be drawn at the regional level, if such an analysis were performed correctly. This is followed by a more limited analysis for northeast Brazil, given data limitations, as well as some provisional results. A. Motivation for a "Determinants of Growth" Type Analysis Most "determinants of growth" studies have been performed at the national level. The first generation of empirical work focussed on the stability of factor shares within a single economy. The more recent empirical literature on growth focuses on growth across countries. These studies have usually been motivated by a need to explain persistent disparities in aggregate growth rates across countries (disparities that have often translated into large differences in per capita welfare). Reviews by Barro and Sala-I-Martin (1995), Pritchett (1997), and Romer (1996), among others provide succinct reviews of this body of literature. (1) Theoretical growth models are based on an aggregate production function with the quantity of labor input, physical capital, and technology as parameters in examining changes in output. The stock of human capital is often embodied in the effective labor force. Technology is introduced in the relationship as either labor augmenting or Hicks neutral (a Hicks neutral process is one where the ratio of the marginal product of capital to the marginal product of labor remains the same with shifts in the production function). There has been limited discussion on the role of technical progress (factor productivity) vis-a- vis factor expansion in the growth process. The commonly used measure of economic efficiency (as distinct from economic welfare) is total factor productivity (TFP) or multifactor productivity. The difference between growth per capita and the growth of a composite measure of per capita factor inputs is TFP, i.e., the "technical progress" or efficiency gain that cannot be explained by factor accumulation alone. The use and limitations of alternative productivity yardsticks is discussed in some detail by Baumol, et al. (1989, chapter 11). They argue that TFP must not be construed as a better alternative to labor productivity, since they measure different things: "while TFP is undoubtedly the better index of efficiency of input use . . labor productivity is the more illuminating measure of the result of the process for its human participants." The measures should be treated as complimentary indices that estimate different things. Taking a purely empirical approach, TFPG can be estimated without any direct reference to a production function. This is the approach adopted in the growth accounting literature (Chen 1997). A weighted sum of the contributions of capital and labor (fractional growth rates K and L ) are subtracted from the growth rate output Q to yield TFPG as an unaccounted for "growth residual." TFPG = Q - a.K - b.L Here a and b are shares of capital and labor in national income; if all income is divided up between capital and labor then a and b necessarily sum to 1. In this TFPG is the difference between labor productivity growth and the growth rate of the capital-labor ratio k scaled down by the output elasticity of capital. It is easily shown that if a and b are assumed constant the growth accounting approach is equivalent to the production function approach (Chen 1997). A2-4 A comprehensive summing up of the problems associated with measuring TFP, going back to the earliest efforts, is given in Chen. He identifies the problem of measuring capital input as the greatest problem. In the standard growth accounting framework the capital growth term subtracts the contribution to overall growth made by technology embodied in capital. Since this embodied technology is prepackaged and paid for, it is not taken as contributing to technological change in the entity (firm or country) that imports it. Chen concludes that disembodied, Hicks-neutral, technical change as measured by the TFP residual is an arbitrary concept, highly sensitive to the growth accounting procedure. In short while the index of overall labor productivity growth can be measured with reasonable accuracy and taken as an indicator of overall technological change, it is very difficult to separate out disembodied technical change from that embodied in the inputs. As Chen suggests, confusion arises from the fixation of analysts with the former and their inability to see the latter as the major contributor to technological advances, as in East Asia. The above arguments are borne out by the empirics, in particular the variation in TFP estimates found in World Bank (1993), Young (1992, 1994, 1995), Collins and Bosworth (1996) and others listed in Chen (1997). (2) Empirical estimation of growth models rely on more readily available operational variables such as per capita income (real and not nominal values are needed to calculate rates of growth), domestic investment and savings as a share of GDP, government consumption as a share of GDP, exports as a share of GDP, percent of population of working age (labor force), share of labor in agriculture and industry, and human capital or education variables like school enrollment, high school/ college graduates as percentage of population, urbanization rate, among others. Cross-country regressions have also used levels of democratization and other political variables in analyzing growth. Such estimates can be useful for determining policy priorities. For example, the rate of savings was considered a critical parameter in the early literature on development, whereas more recently (as in the frequently cited work "Policy Determinants of Growth: Survey of Theory and Evidence" by Easterly and Wetzel 1989) it has been suggested that the efficiency of investment is very important in determining growth performance across countries not just the level of investment. Such growth models have been adapted to regional analysis. (See Handbook of Urban and Regional Economics 1986, Sakashita 1967, Shell 1969). The neoclassical growth models-with exogenous technical progress-have been used most often in empirical analysis of regional growth. In the neoclassical framework (built on the assumption of decreasing returns to reproducible factors) disparities arising from differences in regional capital-labor ratios diminish over time-both trade and factor flows tend to equalize factor prices. A richer set of explanations of the persistence of differences in growth is provided for by the "endogenous growth' literature, which show growth rates also varying with initial conditions. Both capital and labor flow to richer regions, which have the advantages of lower unit costs, higher wages and larger market sizes (market imperfections complicate the analysis). Altering factor mobility remains critical for the prospects of lagging regions in both types of analysis. Policy interventions arising from neoclassical models focus on enhancing efficiency of resource allocation and removal of supply-side constraints. Whereas policy interventions arising from endogenous growth models focus on increasing returns (often seen as arising from technological or pecuniary agglomeration externalities) which can generate a process of circular, cumulative causation. A2-2 B. Methodological and Data Limitations Precluding a Full "Determinants of Growth" Analysis for Northeast Brazil A "determinants of growth" analysis for the northeast would help identify the various factors that have led to changes in output and productivity in the region. However, a full-blown analysis could not be performed due to several methodological and data limitations noted here (for example, the analysis presented does not include some of the finer empirical tests that have been used in the growth literature). The results provided from the more limited analysis conducted in this section of factors leading to changes in output and income in the northeast therefore remain provisional. Methodologies and specifications similar to those used in the "determinants of growth" literature required more data at the subnational level than was available for the northeast. To test the factors contributing to increases in output as well as increases in productivity would have required subnational data not just for the northeast but also the various states that included output, private investment, local savings, changes in labor force, as well as various categories of public expenditures. Broadly there were four different types of limitations to conducting a more sophisticated analyses of the determinants of growth in the northeast. First, the limited time span of the data available from various sources in Brazil does not permit the analysis of structural changes that may have occurred in the northeast economy. Our complete data series for the northeast (for variables such as output, population, private investment, and public investment) is from 1970 to 1991, even though some variables are available for a longer time period. The literature suggests that a test of causality (for example, the Granger causality test) is not reliable with only 22 years of data. The second methodological limitation is related to the first. Specification of appropriate lags to examine the effects of different types of investment were not possible due to the limited time span of data. For example, where networks are sparsely developed transport and communication infrastructure investment is likely to be lumpy initially and to have a longer gestation period before impacts can be identified. With a relatively short time series it would not be possible to empirically test the output effects of these investments. Deriving policy implications based on short-term relationships without appropriate lags between transport investment and economic growth would be biased and possibly misleading. If a longer time series data were available, we could have tested the appropriate lag specification (for example, with a vector autoregression (VAR) methodology) and reported the effects of these investments on growth. The third methodological problem was the inability to examine variations in growth patterns across states in the northeast. This type of analysis at the subnational level would be analogous to the new economic growth literature at the national level, focussing on identifying the sources that lead to cross-county differences in income and productivity. Such an approach would permit the inclusion of unobservable state-specific heterogeneity in growth regressions. The lack of data on factors like investment (private) precluded such analyses. The fourth methodological problem was the inability to identify whether growth in the northeast was driven by increases in total factor productivity or was just a function of factor expansion. Due to limitations with the lag specifications and the use of an exogenous technical progress parameter, we were not able to test for retums to scale that may indicate increases in total factor productivity. Our analyses to examine the role of factors of production (private capital, public capital, and labor) in the growth of the northeast were performed at two levels. First, at the regional level, we used data for all three categories to examine the role of these factors in increasing output. Second, at the A2-3 state-level, we tested the relationship between public capital (and its sectoral categories) and state level economic growth. Economic growth was measured by changes in state GDP. The lack of private capital data at the state level limited the use of the panel of nine states in estimating a production function type of analysis. There are some other data issues that are germane to the analysis actually conducted. The variables used in examining the factors influencing growth in northeast Brazil's growth are presented in table A2. 1. All expenditure and output data used in the study have been converted to constant 1980 figures to control for the effects of high inflation in Brazil. Time-series data from 1970 to 1991 were used in this study. This particular time period was used primarily because private investment data were only available for this period. The data for GDP and investment were obtained from SUDENE and IBGE. The data were carefully converted from their original paper form to electronic format. Fiscal incentive data were obtained from IBGE sources. Table A2.1 Description of Variables Used in the Analysis Variable Description Source Gross domestic product Annual data in 1980 R$ SUDENE (1995); Government of Brazil Private investment Annual data in 1980 CR$ IBGE, Brazil Fiscal incentives Annual data in 1980 CR$ IBGE, Brazil Public capital Annual data in 1980 CR$ IBGE, Brazil Transport and communication Annual data in 1980 CR$ SUDENE (1995); Government of Brazil investment Electric energy and water supply Annual data in 1980 CR$ SUDENE (1995); Government of Brazil Population Annual population IBGE, Brazil; International Financial Statistics, IMF. In some cases, we had to use proxies in the analysis, as the actual data series were not available. For example, we did not have annual labor force data, so we used population (and several of its variations) as a proxy, even though that is likely to introduce estimation problems. In the analysis, an attempt has been made to control for simultaneity and other related issues that could generate biased estimates. For example, as the public expenditures and population series were highly correlated, we estimated a two-stage process to address the simultaneity between these two data series. The level of aggregation in the public investment series makes it difficult to draw out relevant policy conclusions about specific expenditure categories. For example, it was not possible to estimate the effect of irrigation on output (or even on agricultural output) as irrigation expenditures are lumped with agriculture. Similarly it was not possible to estimate the effect of water supply investments on private capital or on output because water supply was aggregated with electric energy investments. A2-4 C. Rudimentary Analysis and Provisional Interpretations The results presented here summarize the important findings related to factors influencing economic output changes in the northeast. We have estimated three types of models to examine the effects of public infrastructure capital, private capital, and labor growth on changes in output. Analysis at the Regional Levelfor the Northeast Many national programs and policies will have effects throughout the country. However, these effects will often not be uniform as different parts of the nation have different constraints and comparative advantages. The purpose of regional analysis is to identify these subnational differences and evaluate policies and programs designed to address region-specific issues. One approach to tackling this problem is to do an analysis of the determinants of output growth in the region, specifically the impact of public and private capital formation on growth. In the case of the public sector the focus is on the role of public infrastructure investments. In the simplest framework, we estimate a modified Cobb-Douglas production function in which infrastructure is treated as an input to production. This methodology has been widely used in the infrastructure literature (Aschauer 1989, Munnell 1990, Holtz-Eakin 1992) where the stock of public capital is included in an aggregate production function. The production function without public capital can be written as: Qt = ,80L,"fi'K 2e6t (1) Qt= output in the northeast in time (t), Lt= regional population in time t; Kt = private capital in time t. The government capital stock (G) is included in the relationship by modifying (1). Qt = Pl0L'1K 2G!)3ee6 (2) In our analysis, public infrastructure is disaggregated into the two categories: transport and communication (TC) and electric energy and water supply (EEW), as these are the only disaggregated infrastructure series available.] We treat for the possibility of simultaneity between population and (total) public investment in the single equation model by estimating a two-stage least squares model. In this framework, we first regressed total public investment against several instrumental variables, and then used the estimated value of public investment to estimate the final model. Summary results are presented in table A2.2. We also dis-aggregated public capital into the two infrastructure categories (table A2.2, column B). The results presented in table A2.3 suggest that private capital is the driving force behind output changes in the northeast. The elasticity for private capital is positive and significant. For example, the results in column B show output elasticity of capital at 0.3 meaning that a 10 percent increase in private capital will increase output by 3 percent in the northeast. In the same equation, 1 One could argue that there is some economic rationale to using these two series (to separate the differing roles of different types of infrastructures in the economic growth process: transport and communication networks establish regional linkages and enhance connectivity, i.e., create larger markets for producers, whereas electric energy and water supply are more important at the local level in reducing costs of production, whether or not larger markets are served. A2-5 public infrastructure has negative elasticities, with the effect of transport and communication being statistically significant. These results are obtained partly due to the short time lag specified between output and transport investments. Investments in transport have long gestation periods and could be reducing consumption and output in the short run. The results for the hyperinflation dummy and the time trend are consistent with a priori expectations.2 Table A2.2 Factors Influencing Output Growth in the Northeast: Estimates Using a Single Equation Model Column A Column B Explanatory variable TSLS Log (output) OLS Log (output) C -102.60* -75.10 Log (Labor) 1.007 0.57 Log (Private Capital (-2)) 0.62* 0.30** Log (Public Capital (-2)) -0.66 [Estimated] Log(EEW(-2)) -0.07 Log(TAC(-2)) 0. 13* Hyperinflation A.08* -0.05* Time Trend 0.05* 0.03* F-Statistic 97.04*** 95.55*** Adjusted R2 0.92 0.95 Number of Observations 20 20 Values in parentheses following variable names represent lags; *** significant at I percent, ** significant at 5 percent, * significant at 10 percent. In addition to the single equation model, we also tested the feedback loop that exists between public infrastructure and economic growth. We estimated a recursive model to include the effects of output on public investments as well as the role of public infrastructure in stimulating private capital formation. Zellner's (1962) seemingly unrelated regression method (SUR) was used to estimate the recursive system. (The SUR is a recursive model that consists of a series of endogenous variables considered as a group because they are conceptually linked with each other.) In the SUR procedure, each equation has an endogenous variable on the left side and only exogenous variables on the right side. As in the standard regression case, the disturbances are assumed to be uncorrelated with the exogenous variables. Each equation of this kind of a system could be estimated by regression, equation by equation. However, if the disturbances of the equations are correlated, the SUR estimator is more efficient, because it takes account of the entire correlation matrix for all of the equations. 2 The only problem with the model is that returns to labor are not statistically significant (even at the 50 percent level of confidence). It is possible that the low labor to output ratio observed in the various "snapshots" translates into these insignificant estimates. A2-6 Table A2.3 Summary of Empirical Estimates Column A Column B Column C Column D Explanatory variable Log (EEW) Log (TAC) Log(PVK) Log(output) Log(Output(-1)) 1.18*** 0.37*** 0.14 Log(PVK(-I)) 0.71*** Log(PVK(-2)) -0.25 0.40*** Log(TAC(-2)) -0.02 0.13** Log(EEW(-2)) 0.12* 0.05 Log(Labor(- 1)) -1.00 HYPDUM -0.08** Time Trend 0.07*** Constant 4.6*** 3.80*** 2.55*** -138.1*** Adjusted R2 0.63 0.26 0.86 0.95 Number of Observations 21 21 20 20 Values in parentheses following variable names represent lags; *** significant at I percent, ** significant at 5 percent, * significant at 10 percent. The results from the recursive model (table A2.3) reinforce the results presented in table A2.2 that show that private capital is the driver of regional economic growth in the northeast. Again public investments in transport have a negative effect on growth. In addition to the long time horizon for the transport investments to be put in place, there is another important issue related to these investments. New transport and communication networks open the lagging northeast markets to more efficient producers in the south, were the unit cost of production may be much lower. As a result, the producers in the northeast may be disadvantaged with the introduction of new access routes, thus leading to a decline in output. The negative estimates of transport in the context of private capital (column C) also suggest the same point, but this may require more analysis. As noted in chapter 2, the ambiguity of the latter results do have, however, some policy implications: in general investments in transport and communication infrastructure should be evaluated primarily in terms of their contribution to increasing the national efficiency of production and distribution by expanding and integrating markets, not region-specific improvements. However, with additional analysis of disaggregated data (for example, the impact of specific projects, such as regional airports and sea ports) it should be possible to identify a subset of transport and communication infrastructure investments whose benefits are more region specific but realized with a substantial lag and only in the context of a package of complementary policies. On the other hand, water and energy investments attract private investment by lowering the unit cost of production in the region. As a result, they are likely to stimulate private investment as well as increase economic output. Due to the shorter gestation period of water investments, their effect on private capital and output can be appropriately estimated with a shorter lag specification. In addition to estimating the contribution of the factors of production to aggregate output, we tested whether public expenditures had any effect on private capital formation in the northeast. Results presented in table A2.4 (column A) suggest that public capital and separately, fiscal incentives, positively influence private capital.3 3 It should be noted that the lag specification of one year has not been tested as the appropriate lag, but has been used because of data limitations. A2-7 Figure A2.1 Private and Public Investment in Northeast Brazil, 1965-91 300 20 50 - __ ~~~~~~~Public investm ent 50 o' I IO I 4 I 4 I I I1\~ I I I I I I I I,- 1_ I- I4~ I-`; Source N ortheast D eveloom ent Bank (B N B N 1SUDENE. Data in 1980 Cruzeros. The effect of public capital on private capital can be observed in figure A2. 1, where there is a close association between the levels of public and private capital.4 However, it is possible that some types of public investments (for example, roads and education) are complementary to private investment and therefore likely to crowd-in private investment, while others (for example, investments in industry and state-owned enterprises) are substitutes and likely to crowd-out private investment. To test for this we replaced aggregate public investment by major (and important infrastructure) categories, and the effects do vary across sectors. Infrastructure investments like transport and communication (TAC) and electric energy and water supply (EEW) have positive effects on changes in private investment. However, public investments in financial services appear to have a negative effect on private investment.5 These results are shown in table A2.4 (column B). We also tested the effects of other categories of public investments on private capital but did not find the relationships to be significant. These categories included industry, mining, agriculture, and other miscellaneous services. Other important categories such as education and public health, could not even be tested as data were not available for these categories. 4 While the two series are correlated, it is anomalous that they are so similar in levels and trends (including peaks). This requires further exploration as multiplier effects and countercyclical effects are not efficient in the current results (if anything, private investment appears to lead public investment). 5 Exclusively state-owned banks and fnancial institutions. A2-8 Table A2.4 Factors Influencing Private Capital Formation in the Northeast Column A Column B Explanatory variable Private capital Private capital C 2.65** 3.12*** Log (Output(-l)) 0.23 0.68*** Log (Public Capital(-l)) 0.44** Log (Fiscal Incentives(-1)) 0.18** Log (TAC(- 1)) 0.32** Log (EEW(-1)) 0.18* Log (Fin. Services (-1)) -0.12** F-Statistic 32.05*** 36.62*** Adjusted R2 0.82 0.88 Number of Observations 21 21 * ** significant at I percent, ** significant at 5 percent, * significant at 10 percent. The estimates are interpreted as elasticities due to the log-log form of the function. EEW-Electric energy and water supply, TAC-Transport and communication, FSV-Financial services. Analysis at the State Level At the state level, we could not investigate whether public investments crowd-in or crowd-out private investment, due to the absence of adequate data on private capital for the states. However, we examined a reduced form of this relationship by estimating the effect of public investment on growth of output. The panel data set is used in lieu of just the time series data (as more units of observation are available for the panel data set). The results from the empirical analysis are presented in table A2.5. These results suggest that public investments in electric energy and water supply6 are more closely associated with output changes across the northeast states in comparison to other development inputs like public investments in transport and communication, and agriculture (column A). Investments in transport and agriculture do not have significant output effects. Table A2.5 Association Between Public Capital and Output Across States in Northeast Brazil Column A Column B Estimate Estimate (with cross- Variable Interpretation (OLS Fixed effects) section effects for EEW) Year Time trend 0.058* 0.055* Hypdum Hyper inflationary dummy -0.095** -0.059 Log (Agr (-1) Logged value of public investments in -0.017 -0.014 agriculture Log(TAC (-1)) Logged value of public investments in -0.03 -0.002 transport and communication Log(EEW(-1)) Logged value of public investments in 0.18*** electric energy and water supply F-Statistic 570.31 *** Adjusted R2 0.869 0.865 Observations 171 171 * ** significant at I percent, ** significant at 5 percent, * significant at 10 percent. The estimates are interpreted as elasticities due to the log-log form of the function. EEW-Electric energy and water supply, TAC-Transport and communication, FSV-Financial services. 6 The effect of recent privatizations of public utilities on the scale of investment in these sectors and the associated impact on the growth of output cannot be evaluated at this stage with the available data. A2-9 As electric energy and water supply infrastructure has a positive effect on output across the nine northeast states, we estimated the differences in the output elasticity of this infrastructure across states. The main results are presented in table A2.5, column B, and the regional specific values are presented in table A2.6. The results in table A2.6 indicate that the output effects of electric energy and water supply are positive in all states with the exception of Bahia and Ceara (where they are not significant). Table A2.6 Effects of Water and Electric Infrastructure Investment on Output Across Northeast States State Elasticity of electric energy and water supply Alapoas 0.10*** Bahia -0.13*** CearA 0.05 Maranhao 0.18*** Paraiba 0.07*** Pernambuco 0.10** Piaui 0.05** Rio Grande do Norte 0.11 * Sergipe 0.09*** *** significant at I percent, ** significant at 5 percent, * significant at 10 percent. D. Provisional Implications for Policy In general, our results indicate that private capital is influenced by the availability of public capital, and is more closely associated with sectors like infrastructure (transport and electric energy, water supply) than investment in state enterprises (financial services or industry).7 Fiscal incentives have had a positive effect on stimulating private capital in the northeast. Private capital has a positive effect on output growth in the northeast. While public investment does not consistently have a positive effect on output increases, it is positively related to private capital, thus implying an indirect relationship with output growth. Thus, we cannot dismiss the role of public investments in increasing northeast output growth. Methodological and data problems indicated earlier did not allow the us to test for increasing returns to scale that may signify increases in factor productivity. For example, a better-specified public investment equation would allow us to comment on its exact role in the growth process and test for returns to scale. Investment in electric energy and water supply have a positive effect on private capital and are also positively associated with output increases across the nine states. The positive effects of these investments consistently show up across various estimation procedures and methodologies, leading us to believe that these results are robust. However, the level of aggregation does not facilitate the distinction between effects of electric energy vis-a-vis effects of water supply. The effects of the former include reducing the unit cost of production for northeast producers, thus contributing to increases in output. The effects of the latter (water supply) are primarily on health and improve the quality of life of the residents.8 7 We performed these analyses with data on several public investment categories. We have only reported the significant results in this section. A listing of all the empirical tests is available from the authors. 8 The effect of irrigation cannot be derived from these estimates as it is grouped with agriculture investment. The effect of agricultural investments was positive but not significant. A2-10 The health benefits of water supply and sanitation have been discussed at length in the section on health in this report. These results are consistent with the findings of a recent World Bank (1998) study on managing pollution problems in Brazil, where the lack of access to clean water and adequate sanitation services are identified as a very serious public health problem. Pollutants and pathogens in sewage systems are a major source of mortality and morbidity, especially in young children. By reducing exposure to pathogens, clean water and sewage collection systems can reduce the incidence of mortality and morbidity. There are large disparities in access to clean water across Brazilian states. For example, only 46.3 percent of households in Bahia and 40.3 percent in Ceara have access to clean water. This is in contrast to 95.6 percent in Sao Paulo and 91.3 percent in Rio de Janeiro. The cost per life saved (CPLS) for water and sewer connections is five to ten times higher in the southeast and the south than in the northeast and the north of the country. The importance of water supply and sanitation in urban areas is heightened by the rapid urbanization in the northeast. Table A2.7 presents some data on urbanization trends in the northeast. Large northeastern cities like Salvador and Fortaleza are growing at rates that are almost 60 percent faster than overall population growth rate in the northeast. In 1991, population in these nine urban areas was 8.5 million, which accounted for 20 percent of the northeastern population. The percentage of residents in the main urban areas is increasing in almost every northeastern state. Population growth rates of most major northeastern cities have been higher than southeastern cities like Belo Horizonte (13.4 percent), Rio de Janeiro (7.6 percent), and Sao Paulo (13.5 percent) between 1980 and 1991. Table A2.7 Northeast Brazil: Population in Major Urban Areas Population in capital city State State Capital population State Growth population population in capital Population Population Percentage rate 1980 1980 citv 1980 1991 1991 of state 1980-91 State Capital (million) (million) (percent) (million) (million) 1991 (percent) Northeast 34.80 42.50 Alagoas Maceio 2.00 0.399 19.95 2.5 0.629 25.16 57.64 Bahia Salvador 9.50 1.500 15.79 11.90 2.075 17.44 3S.33 Ceara Fortaleza 5.30 1.300 24.53 6.40 1.768 27.63 36.00 Maranhao Sao Luis 4.00 0.450 11.25 4.90 0.696 14.20 54.67 Paraiba Joao 2.80 0.329 11.75 3.20 0.497 15.53 51.06 Pessao Pernambuco Recife 6.10 1.200 19.67 7.10 1.298 18.28 8.17 Piaui Teresina 2.10 0.377 17.95 2.60 0.599 23.04 58.89 Rio Grande Natal 1.90 0.416 21.89 2.40 0.606 25.25 45.67 do Norte Sergipe Araciu 1.10 0.292 26.55 1.50 0.402 26.80 37.67 Source: IBGE, Directoria de Pesquises, Population Census. In addition to the rapid urbanization in the northeast, the World Bank (1998) report (annex 2) indicated that urban access to piped water and sanitation has a significant effect on reducing infant and child mortality, whereas rural access may have little or no such effect in Brazil. The analysis showed that the level of urbanization was significantly related to infant and those less than 5 years of age mortality rates (higher in urban than rural areas), in all other factors held constant. As the cost of providing water and sanitation infrastructure is lower in urban areas than in rural areas, it can be concluded that the returns to urban infrastructure (particularly water and sanitation) is high. The results for other public infrastructure investments like transport and communications are indeterminate and we cannot conclusively comment on the effect of these investments. For example, A2-1 I results in table A2.4 show that transport and communication have a negative effect on output. This may partly be due to the short lag specification between transport investment and output changes. However, there may be some negative short-run effects where transport improvements may open up the northeast markets to external producers. Larger firms serving larger markets (such as the southeast) and benefiting from economies of scale will have lower unit costs of production and can more easily expand into new markets in competition with local producers when unit cost of distribution between two points is lowered. Transport and communication investment improves linkages with the rest of the country and may implicitly reduce a natural tariff barrier. It is possible that in many cases producers from other regions will crowd out local producers to the benefit of local consumers, but not of local production or employment. Thus, it is difficult to comment on the exact nature of these investments without further analysis. E. The Importance of Better Data for Policy Analysis Availability of better labor data will be important in estimating the contribution of labor to the growth process. A large body of existing research on the northeast has indicated that the northeast's workforce has relatively little education, fewer skills and poor health, along with little investment in social infrastructure of schools and teachers, public facilities, communication networks, and other aids to economic performance (Azzoni 1996, Goldsmith and Wilson 1991, Markusen, 1994). It would be desirable to test for the effects of these characteristics on output and productivity in the region. In the end, it is valuable to ask if the availability of these data would enhance the quality of analysis and provide any relevant tools for regional policy and decisionmaking? We believe that the additional data would help in at least three ways. First, the availability of data would facilitate better analysis and help in understanding the structure of the economy as well as the factors that contribute to the growth of output and incomes in the region. Second, in an era of shrinking public budgets, it becomes critical to identify the sources of public intervention that have the highest welfare impacts. Availability of better data would help in prioritizing various expenditure and investment categories based on their returns on various social and economic indicators. Third, availability of better and longer time-series data would permit the testing of various policy instruments. For example, we could use different parameter estimates system to examine the effectiveness of changes in public investments on output. This will be helpful in analyzing the effects of investment expansions on output and incomes in the region. Thus it the availability of better data will not be a wasted exercise but will contribute to a better understanding of the determinants of growth in the northeast. A2-12 Appendix 3. Macroeconomic Statistical Results A. Poverty and Inequality Table A3.1 shows the results of an analysis of state level data in the northeast. It looks at annual (or intersurvey) changes in the head count indices of poverty from PNAD data for the northeast states from 1981 to 1995 and relates these figures to income growth rates and changes in the Gini coefficient measuring the dispersion of the income distribution within the sate. As shown in the table, a one percentage point change in a state's growth rate can be expected to reduce poverty by 0.067 percentage points. In the case of the effect of the distribution of income for a given mean income, a one percentage increase in the Gini coefficient will increase poverty by 1.14 percent. Table A3.1 Determinants of Poverty Rates in Northeastern States 1981-95 Variable Regression coefficient Standard error Growth in state income -0.067 0.035 Log change of Gini coefficient 1.136 0.195 Adj. R-squared 0.255 Number of observations 108 108 Table A3.2 shows the impact of inflation and unemployment on urban poverty as measured by the PME (Pesquisa Mensual de Emprego) for the metropolitan areas of Salvador and Recife in contrast to the rest of metropolitan Brazil. For the country as a whole each extra percentage point in inflation increases the poverty gap ratio by 0.08 percent, the poverty gap ratio reflecting both the increase in the number of poor, as well as the average shortfall from the poverty line. For the northeastem cities covered by the survey, the poverty gap ratio in Salvador is slightly more sensitive to inflation and in Recife it is slightly less sensitive than in the rest of Brazil. The head count ratios in the two cities change in approximately the same way as the poverty gap ratios indicating simply that more workers in Salvador cluster near the poverty line than do workers in Recife. Table A3.2 Macroeconomic Effects on Poverty and Inequality Unemployment Inflation Poverty - Mean income gap Recife 1.51 0.026 Salvador 2.08 0.071 Brazil 1.60 0.080 Inequality - Theil index Recife 0.865 0.086 Salvador 2.550 0.186 Brazil 1.200 0.150 Note: Entries are percentage changes in the indicated poverty and inequality measures resulting from a one percentage point change in unemployment and inflation. Source: IPEA, Paes de Barros, Corseuil and Mendoqa (1998). The impact of an extra percentage point in the unemployment rate is to increase the poverty gap ratio by 2.08, 1.51, and 1.60 percent in Salvador, Recife, and the rest of metropolitan Brazil respectively. The theoretical connection between the poverty gap and unemployment is not as obvious since the newly poor may still have incomes higher than those of the original poor. If those who were poor originally did not experience much of a drop in income, the poverty gap could be smaller. However, empirically the already poor suffer as well. A3-1 B. State Policies and Deficits Table A3.3 shows the results of the analysis of state deficits in the northeast over the period, 1980-95. As it can be observed, the administrative performance of state governments has a direct effect on state deficits. The administration share of expenditures is the variable that most greatly affects the deficit. Not quite as strong, but still clear, is the relationship between the share of government expenditure going to wages and the overall deficits. It is also important to point out, that following these results, states seem to be able to absorb changes in social spending. These results indicate a close relationship between administrative responsibility and overall deficit reduction but not between programmatic differences among the states. Responsible governments can maintain low deficits regardless of the substantive policies pursued in terms of social sector expenditures. Table A3.3 Determinants of State Fiscal Deficits Coefficient Standard error Administration share of expenditures 0.4287 0.1628 Wage share 0.028 0.021 Social sector share -0.001 0.003 + state effects Adj. R-squared =.283 Table A3.4 shows the results of the regression analysis from the PPV data set on determinants of wages. The results confirm those found in the earlier studies by the Bank and IPEA, Paes de Barros, et al. (1997) and Carneiro and Gill (1997) that the impact of public employment is in fact, larger on relatively low-skill workers. In fact, the results appear to show that at higher income levels, the public sector employees are paid less than their private sector counterparts. The explanation for this might be that current wages do not capture the full benefit of public employment. In particular, the relative generosity of public pension plans as well as considerably greater job security in the public sector may still make public employment attractive relative to private. Indeed, one might even use the wage differential as a measure of the value of these nonwage benefits on the assumption that people move between jobs until they are equally attractive. At the lower end of the wage scales, however, the pension and security benefits are certainly higher than they would be in the private sector so that the wage differential is clearly a net benefit. Table A3.4 Determinants of Hourly Wages Determinants of hourly wages Coefficient Standard error Years worked -0.0466 0.023 Age 0.3163 0.049 Age 2 -0.0022 0.000 Completed primary education 2.4155 1.879 Completed secondary education 4.3094 1.697 Completed tertiary education 16.8730 1.744 Interaction term between sector dummy and prim variable -0.5479 0.977 Interaction term between sector dummy and sec variable -0.6184 0.907 Interaction term between sector dummy and ter variable -4.8534 0.969 Dummy I if working in public 0.7718 0.762 Constant -6.9969 1.736 Number of obs. adj. 4728 R-squared 0.1645 Source: PPV. A3-2 Table A3.5 presents the effect on rural wages of a 20 percent fall in the current level of public employment under a variety of assumptions concerning the labor market. As can be observed in the table, the effect of a 20 percent decrease in public employment could reduce wages as much as 6 percent but only under the most extreme assumptions: perfect mobility between urban and rural areas, no supply response to wages, and a very low elasticity of demand of-0.5. Table A3.5 Percent Fall in Rural Wages from a 20 Percent Fall in Public Employment 1. Perfect mobility Elasticity of demandfor labor Elasticity -0.5 -1.0 -1.5 -2.0 of supply 0 6.78 3.39 2.26 1.70 of labor 0.5 3.13 2.14 1.63 1.31 1.0 2.03 1.56 1.27 1.07 1.5 1.50 1.23 1.04 0.90 2.0 1.19 1.02 0.88 0.78 2. No mobility Elasticity of demand 0.5 -1.0 -1.5 -2.0 Elasticity 0 3.48 1.74 1.16 0.87 of supply 0.5 1.67 1.13 0.85 0.68 of labor 1.0 1.10 0.83 0.67 0.56 1.5 0.82 0.66 0.56 0.48 2.0 0.65 0.55 0.47 0.42 A3-3 Appendix 4. Health Results A. Method The Pesquisa Nacional Demografia e Sauide (PNDS) survey is a rich source of information concerning health, family planning, and education. Economists have generally overlooked it, however, because of the absence of questions concerning income or total consumption. Clearly, for an analysis of poverty in particular, this is a serious drawback. It is possible, however to use information about the ownership of certain consumer durables and assets to construct an index of wealth for the families in the sample. This appendix explains the construction of the index and provides a justification for its use. The method takes in account all variables in the survey related to ownership or to the standard of living; it takes the first principal component of the correlation matrix and uses the elements of that component as coefficients to obtain a weighted average of the initial variables. Since the actual score (the weighted average) has no real meaning, that is, the units that this score is measured in do not correspond to observable quantities the resulting score is divided into deciles. Deciles, being defined ordinally, do not depend on these units and are conventional for the analysis of standards of living. The variables used were the following: * Building materials of floors, walls, and roofs * Number of residents per room * Household ownership of a bicycle, automobile, radio * Number of maids * Whether household has electricity and if so, * Whether household owns a refrigerator, television, VCR - Number of color televisions. The first component of the principal component analysis "explains" 58 percent of the variation in the variance-covariance matrix of the variables for assets and consumer durables. The second component's contribution drops rapidly to 16 percent. This indicates that the first component is picking up the lion's share of the information captured by the ownership of possessions. A further indication that the resulting score reflects general wealth is that the weights on all variables are positive while those for the second is made up of varying signs appearing to be identifying households with electricity and electric appliances independently of other possessions. Indeed, the coefficients on the variables accord with reasonable prior notions of their income elasticity. So, those goods (such as number of color televisions!), which one would expect to be most highly sensitive to increases in income, also tend to be heavily weighted in the constructed variables. The resulting measure is a completely mechanical construct. What relation does it have to more conventional notions and measures of well being such as income or total consumption? The theoretical construction and its performance in practice for empirical work can be found in Filmer and Pritchett (1998). On theoretical grounds, it is not clear what the relevant concept of well being should be. For long-term concepts such as health status (serious enough to affect infant mortality) or lifetime plans for education, the relevant measure would be a subjective estimate of long-term earning capacity. For this reason, measures of consumption are frequently preferred to income, since consumption incorporates intertemporal smoothing of people's income over periods of seasonal or annual variation. Consumption would be governed by expectations of future earnings and prices. Such effects are particularly important in agriculture where such variation is induced by weather conditions. Even consumption, however, may be an imperfect measure of long-term living conditions. Any number of constraints on this smoothing behavior can interfere with the translation of "permanent income" into consumption. Credit constraints, where people may be able to save for the future in good years but not able to borrow against it in bad years, can disrupt the connection between the two. Further, judging the adequacy of the measures should be based in part on the degree of recall error in the different measures. Consumption is frequently measured on the basis of questions on food intake over an extended period (such as two weeks). The A4-1 degree of error in measurement is unclear, but for people with any variety in diet, it is likely to be substantial. For ownership of durables or of house construction, the degree of error should be quite small. The first solution for dealing with mis-measurement of permanent income was proposed by Friedman (1957) who also suggested the concept to begin with. This involved a choice of instruments for the true concept to be used for measured income. The instruments he used were, in fact, consumer durables, which he hypothesized to be closely related to judgement of long-term earning capacity and not related to transitory errors. These are the same sort of variables included in this measure. Empirically, Filmer and Pritchett assess the proposed measure against the more conventional measures of consumption. For several countries-mostly in South Asia-they examined survey data including both sorts of information. The test they performed was to use each variable in a particular regression both in ordinary least squares (OLS) as well as in instrumental variable (IV) form, the instruments being the alternative measure (the index of assets as an instrument for consumption and vice versa). The estimates should change and be larger in absolute value when there is a significant degree of measurement error. Their results show that using the asset index for consumption did just that, estimates were much larger. In the reverse direction, the asset index performed almost identically in both OLS and IV forms. This would indicate that the index is a better variable to capture the effect of long-term welfare than the conventional measure. B. Statistical Analysis of Health and Health Care Table A4. 1 shows the result of the regression of the mortality equation using DHS data. As can be observed, the education of the mother plays a significant role in the reduction of child mortality. All three dummy variables representing the highest educational level attained by the mother (primary, secondary, or tertiary) have negative coefficients and are significant even at a 1 percent level. Also, the variables representing the expenditure deciles have a significant effect in reducing child mortality. As discussed in the text, both the absence of toilet facilities and a source of water designated as "other" is strongly related to child mortality. "Other" water sources are mostly surface streams and other natural bodies. Table A4.2 shows the result of the regression of the determinants of self-reported morbidity, which in this case referred to the incidence of diarrhea. As can be observed, access to water and sanitation facilities significantly reduces the probability of having diarrhea. More specifically, access to piped water and sanitation carry negative coefficients and both are significantly different from zero. The results are consistent with a commonly found, if seemingly, perverse result. Higher per capita consumption is associated with more reported illness rather than less. The reason for this result is that the dependent variable is made up of two components-the probability of being sick times the probability of saying_that one is sick conditional on being sick. The probability of reporting an illness might be rising faster with income than the probability of being ill falls with income. This result appears in many contexts, for example, reported illness is higher in the United States than in India. Rich people have no tolerance for discomfort while poor people think it is normal. The interpretation made here, i.e., that sanitation and water affects mortality depends on the assumption that the type of facilities a person has will affect the actual illness experienced but will be unrelated to the likelihood of reporting it. Table A4.3 on this appendix reinforces the importance of the access to water and sanitation facilities on the health statues of the northeast population. The table shows the results of a study on different models for explaining infant mortality rates using data from the 1980 and 1991 census. In all the models used the variables that measured the water supply and sewage system adequacy have a significant role in reducing the probability of infant mortality in the northeast. This pattern is less significant in the south and southeast regions of Brazil. A4-2 Table A4.4 presents the results of an analysis of the PPV data in which the likelihood of people seeking care once sick, in either the public or private sector, is examined. Table A4.3 on this appendix reinforces the importance of the access to water and sanitation facilities on the health statues of the northeast population. The table shows the results of a study on different models for explaining infant mortality rates using data from the 1980 and 1991 census. In all the models used the variables that measured the water supply and sewage system adequacy have a significant role in reducing the probability of infant mortality in the northeast. This pattern is less significant in the south and southeast regions of Brazil. Table A4.4 presents the results of an analysis of the PPV data in which the likelihood of people seeking care once sick, in either the public or private sector, is examined. Table A4.1 Determinants of Mortality Equation Mother Coefficient Std Err. t-value P>t [95% Confidence interval] Decile dummy 2 -0.0040166 0.0075714 -0.53 0.596 -0.0188584 0.0108253 Decile dummy 3 -0.0077275 0.0079714 -0.969 0.332 -0.0233535 0.0078985 Decile dummy 4 -0.0310853 0.0079893 -3.891 0.000 -0.0467463 -0.0154243 Decile dummy 5 -0.0200478 0.0082202 -2.439 0.015 -0.0361615 -0.0039341 Decile dummy 6 -0.0378757 0.0083334 -4.545 0.000 -0.0542113 -0.0215402 Decile dummy 7 -0.042490 0.0084449 -5.031 0.000 -0.590437 -0.0259353 Decile dummy 8 -0.035941 0.0084995 -4.229 0.000 -0.0526022 -0.0192798 Decile dummy 9 -0.0479828 0.0086036 -5.577 0.000 -0.0648481 -0.0311174 Decile dummy 10 -0.054310 0.009198 -5.905 0.000 -0.0723405 -0.0362795 Water from well 0.005456 0.0044535 1.225 0.221 -0.0032737 0.0141861 Bottled water 0.008478 0.0086868 0.976 0.329 -0.008550 0.0255061 Other sources of water 0.038836 0.0092982 4.177 0.000 0.020609 0.0570624 Presence of toilet pit 0.001979 0.0038015 0.521 0.603 -0.0054729 0.0094311 Absence of toilet 0.013220 0.006657 1.986 0.047 0.000171 0.0262692 Age -0.0005834 0.0014171 -0.412 0.681 -0.0033612 0.0021944 Age squared 0.00004222 0.000021 2.013 0.044 0.00000111 0.0000834 Completed primary -0.0458366 0.0060272 -7.605 0.000 -0.0576515 -0.0340218 education Completed secondary -0.0608184 0.006459 -9.416 0.000 -0.0734796 -0.0481572 education Completed higher education -0.07053 0.0098187 -7.183 0.000 -0.0897771 -0.0512829 Urban 0.0131939 0.0050463 2.615 0.009 0.0033018 0.023086 Regional dummies 2 0.0104622 0.0078667 1.33 0.184 -0.0049585 0.0258828 Regional dummies 3 0.0023951 0.0077517 0.309 0.757 -0.0128003 0.0175904 Regional dummies 4 0.0060041 0.0079625 0.754 0.451 -0.0096044 0.0216126 Regional dummies 5 0.025085 0.007173 3.497 0.000 0.0110241 0.039146 Regional dummies 6 0.0139553 0.0083479 1.672 0.095 -0.0024087 0.0303192 Regional dummies 7 0.0179452 0.0078923 2.274 0.023 0.0024743 0.033416 Constant 0.0745753 0.0253781 2.939 0.003 0.0248278 0.1243228 Number of observations 8196 Adjusted R-squared 0.0745 A4-3 Table A4.2 Determinants of Self-Reported Morbidity Morbidity Coefficient Std. Err. z P>z Log real per capita expenditure 0.1937 0.0254 7.637 0 Autonomous farmer 0.1777 0.1183 1.502 0.133 Farm worker 0.2637 0.0525 5.025 0 Age -0.0375 0.0028 -13.289 0 Age squared 0.0005 4E-05 14.342 0 Region Northeast municipal 0.3123 0.0715 4.37 0 Northeast urban 0.1596 0.0854 1.868 0.062 Northeast rural -0.0014 0.0931 -0.015 0.988 Southeast municipal -0.0487 0.071 -0.685 0.493 Southeast rural -0.162 0.0961 -1.685 0.092 Water source Piped system -0.4273 0.181 -2.361 0.018 Own well -0.2358 0.1872 -1.259 0.208 Public well -0.2646 0.1939 -1.365 0.172 Pipe truck -0.4598 0.2853 -1.612 0.107 Other -0.3962 0.1888 -2.098 0.036 Sanitation Piped system -0.379 0.0908 -4.175 0 Septic pool -0.0934 0.0953 -0.98 0.327 Pool -0.1265 0.0865 -1.463 0.143 Pit -0.1639 0.1246 -1.315 0.188 None 0.0191 0.0957 0.2 0.842 Race Black 0.1591 0.0746 2.133 0.033 Mixed 0.2351 0.0394 5.971 0 Male -0.1544 0.0351 -4.403 0 Constant -1.3606 0.2352 -5.786 0 Number of obs. 19,409 Pseudo R2 0.0229 A4-4 Table A4.3 LQ nit oflitv Rate Census 1980 Census 1991 South Northeast Southeast South Northeast Southeast Coef t-stat P-val Coef t-stat P-val Coef t-stat P-val Coef t-stat P-val Coef t-stat P-val Coef t-stat P-val Model 1 Avg. adult schooling -0.27 -11.4 0 -0.06 -2.21 0.03 0 -0.09 0.93 -0.29 -9.89 0.00 -0.11 -5.78 0 -0.03 -1.94 0.05 Lnincome-Theil L -0.13 -2.01 0.04 -0.16 -5.35 0.00 -0.19 -6.94 0 -0.02 -0.22 0.83 0.06 1.52 0.13 -0.15 -4.93 0 Watersupplyadequacy -0.09 -1.04 0.3 -0.47 -4.73 0.00 -0.25 -4.26 0 -0.71 -4.6 0.00 -0.22 -2.68 0.01 -0.43 -6.13 0 Sewage system adequacy 0.31 3.43 0 -0.18 -1.57 0.12 -0.02 -0.62 0.54 0.07 0.91 0.36 -0.22 -2.94 0 -0.08 -1.79 0.07 PhysicianVpopulation 0.04 1.35 0.18 0.06 1.23 0.22 -0.01 -0.41 0.69 0.04 1.5 0.13 0.13 -0.68 0.5 -0.01 -0.93 0.35 Nurse dummy -0.04 -0.78 0.43 0.13 3.7 0.00 0.05 2.05 0.04 0.04 0.59 0.55 0 3.05 0 0.08 2.18 0.03 Nurse education 0.01 1.29 0.2 -0.01 -2.46 0.01 0 -1.4 0.16 0 -0.32 0.75 0.48 -0.68 0.5 -0.01 -1.75 0.08 Urbanization rate 0.57 9.53 0 0.69 8.95 0.00 0.39 9.3 0 1.04 11.69 0.00 -2.26 6.68 0 0.39 6.93 0 Constant -2.5 -2.26 -2.93 -2.51 0.08 -3.31 R-squared 0.36 0.08 0.15 0.3 1353 0.22 Number of obs. 637 1353 1402 657 1402 Model 2 Avg. adult schooling -0.27 -11.2 0.00 -0.08 -3.12 0.00 -0.02 -1.43 0.15 -0.28 -9.26 0.0 -0.14 -7.43 0 -0.03 -4.44 0.01 Lnincome-Theil L -0.13 -1.94 0.05 -0.15 -5.14 0.00 -0.18 -6.7 0.00 -0.02 -0.24 0.81 0.08 1.81 0.07 -0.15 -4.92 0 Watersupplyadequacy -0.11 -1.23 0.22 -0.46 -4.69 0.00 -0.24 -4.12 0.00 -0.72 -4.63 0 -0.2 -2.47 0.01 -0.42 -6.05 0 Sewage system adequacy 0.3 3.37 0.00 -0.2 -1.79 0.07 -0.04 -0.94 0.35 0.06 0.81 0.42 -0.2 -2.7 0.01 -0.08 -1.88 0.06 (Phy.+nurse)/population 0.01 1.38 0.17 0.05 3.31 0.00 0.02 3.22 0.00 0 0.16 0.87 0.02 4.28 0 0 -0.04 0.97 Nursedummy -0.05 -0.98 0.33 0.09 2.45 0.01 0.03 1.06 0.29 0.04 0.47 0.64 0.09 2.03 0.04 0.08 2.14 0.03 Nurse education 0.01 1.20 0.23 -0.01 -2.38 0.02 0.00 -1.29 0.20 0 -0.22 0.83 0 -0.62 0.54 -0.01 -1.74 0.08 Urbanization rate 0.57 9.31 0.00 0.65 8.51 0.00 0.38 9.06 0.00 1.05 11.74 0 0.45 6.33 0 0.39 6.99 0 Constant -2.48 -2.22 -2.89 -2.54 -2.2 -3.3 R-squared 0.36 0.09 0.16 0.3 0.09 0.22 Model 3 Avg. adult schooling -0.29 -13.2 0 -0.06 -2.26 0.02 0 -0.33 0.74 -0.29 -10.6 0 -0.12 -5.85 0 -0.04 -2.67 0.01 Avg. income gap -0.11 -0.49 0.63 0.48 5.42 0 0.73 8.43 0 0.13 0.49 0.63 -0.25 -1.78 0.08 0.58 5.41 0 Watersupplyadequacy -0.18 -2.01 0.05 -0.47 -4.73 0 -0.22 -3.74 0 -0.68 -4.32 0 -0.22 -2.75 0.01 -0.35 -4.67 0 Sewage system adequacy 0.27 2.99 0 -0.19 -1.62 0.11 -0.05 -1.18 0.24 0.07 0.97 0.33 -0.22 -2.93 0 -0.09 -2.03 0.04 Physician/population 0.04 1.36 0.17 0.05 1.07 0.28 -0.02 -1.22 0.22 0.04 1.47 0.14 -0.02 -0.6 0.55 -0.02 -1.25 0.21 Nurse dummy -0.03 -0.6 0.55 0.13 3.72 0 0.04 1.68 0.09 0.04 0.6 0.55 0.13 3.04 0 0.08 2.15 0.03 Nurse education 0.01 1.26 0.21 -0.01 -2.49 0.01 0 -1.18 0.24 0 -0.33 0.74 0 -0.69 0.49 -0.01 -1.66 0.1 Urbanization rate 0.53 8.84 0 0.69 8.96 0 0.4 9.57 0 1.05 11.75 0 0.48 6.65 0 0.4 7.06 0 Constant -2.29 -2.26 -2.97 -2.56 -2.23 -3.37 R-squared 0.35 0.08 0.16 0.3 0.08 0.22 Model 4 Avg. adult schooling -0.29 -13.1 0 -0.08 -3.22 0 -0.03 -1.88 0.06 -0.28 -9.83 0 -0.14 -7.53 0 -0.05 -3.36 0 Avg. income gap -0.1 0.45 0.65 0.46 5.23 0 0.68 7.94 0 0.16 0.57 0.57 -0.29 -2.08 0.04 0.57 5.33 0 Watersupplyadequacy -0.19 -2.15 0.03 -0.46 -4.7 0 -0.21 -3.61 0 -0.69 -4.31 0 -0.2 -2.54 0.01 -0.34 -4.53 0 Sewage system adequacy 0.26 2.95 0 -0.21 -1.85 0.06 -0.06 -1.52 0.13 0.06 0.89 0.38 -0.2 -2.67 0.01 -0.09 -2.15 0.03 (Phy.+nurse)/population 0.02 1.48 0.14 0.04 3.28 0 0.01 2.7 0.01 0 0.15 0.88 0.03 4.29 0 0 -0.1 0.92 Nurse dummy -0.04 -0.83 0.41 0.09 2.49 0.01 0.02 0.83 0.41 0.04 0.48 0.63 0.09 2.01 0.04 0.08 2.12 0.03 Nurse education 0.01 1.17 0.24 -0.01 -2.42 0.02 0 -1.09 0.28 0 -0.23 0.82 0 -0.63 0.53 -0.01 -1.64 0.1 Urbanization rate 0.52 8.68 0 0.66 8.54 0 0.39 9.32 0 1.06 11.83 0 0.45 6.29 0 0.4 7.12 0 Constant -2.27 -2.22 -2.91 -2.6 -2.16 -3.35 R-squared 0.35 0.09 0.17 0.3 0.09 0.22 A4-5 Table A4.4 Health Care Demand (Public versus Private) Care Coef Std. Err. t-value P>t [95% Confidence interval] Public Farm owner -0.0199852 0.2944908 -0.068 0.946 -0.5971766 0.5572062 Farm worker 0.3273990 0.1191373 2.748 0.006 0.0938941 0.5609039 Age -0.0218384 0.0064062 -3.409 0.001 -0.3439440 -0.0092824 Age squared 0.0001351 0.0000805 1.679 0.093 -0.0000226 0.0002929 Tots 0.1168220 0.0519150 2.250 0.024 0.0150740 0.2185735 Kids -0.0501500 0.0328458 -1.527 0.127 -0.1145266 0.0142266 White -0.0468692 0.0902556 -0.519 0.604 -0.2237669 0.1300284 Black -0.0088995 0.1564139 -0.057 0.955 0.3154652 0.2976662 Female 0.2061166 0.0795834 2.590 0.010 0.0501360 0.3620973 Logarithm of real per capita -0.5068027 0.1189731 -4.260 0.000 -0.7399857 -0.2736197 Cold -1.4105070 0.1260840 -11.187 0.000 -1.6576270 -1.1633870 Infect -0.6951180 0.1414836 -4.913 0.000 -0.9724209 -0.4178152 Accident 0.3779541 0.2094724 1.804 0.071 -0.0326042 0.7885125 Digest -0.6934580 0.2016282 -3.439 0.001 -1.0886420 -0.2982740 Pain -0.8106541 0.1374108 -5.899 0.000 -1.0799740 -0.5413338 Dentist -0.8846974 0.3055567 -2.895 0.004 -1.4835780 -0.2858172 Whether or not the person is -1.1807280 0.1417008 -8.333 0.000 -1.4584560 -0.9029994 insured Travel timeto health care 0.3162158 0.1505236 2.101 0.036 0.0211949 0.6112367 facility Expenditure per visit 0.8651918 0.1871139 4.624 0.000 0.4984552 1.2319280 Constant 0.0163899 0.5449318 0.030 0.976 -1.0516570 1.0844370 Private Farm owner 0.1081426 0.3737276 0.289 0.772 -0.6243501 0.8406352 Farm worker 0.3751327 0.1571149 2.388 0.017 0.0671931 0.6830722 Age -0.0203233 0.0076758 -2.648 0.008 -0.0353676 -0.0052791 Age squared 0.0001396 0.0000965 1.446 0.148 -0.0000496 0.0003287 Tots 0.1158484 0.0729040 1.589 0.112 -0.0270409 0.2587376 Kids -0.0314917 0.0469339 -0.671 0.502 -0.1234805 0.0604971 White -0.0642483 0.1088582 -0.590 0.555 -0.2776064 0.1491098 Black 0.0275279 0.2128346 0.129 0.897 -0.3896202 0.4446760 Female 0.1295324 0.0960417 1.349 0.177 -0.0587059 0.3177707 Logarithm of real per capita 0.4064922 0.1390631 2.923 0.003 0.1339334 0.6790510 Cold -1.6188270 0.1523855 -10.623 0.000 -1.9174970 -1.3201570 Infect -0.6864549 0.1659824 -4.136 0.000 -1.0117740 -0.3611353 Accident -0.2151360 0.2531375 -0.850 0.395 -0.7112763 0.2810043 Digest -0.5994287 0.2390350 -2.508 0.012 -1.0679290 -.0.1309286 Pain -0.9608640 0.1734729 -5.539 0.000 -1.3008650 -0.6208635 Dentist 0.0447332 0.2866440 0.156 0.876 -0.5170787 0.6065450 Insure 1.3879330 0.1121632 12.374 0.000 1.1680970 1.6077680 Travel time to health care 0.1222434 0.2090150 0.585 0.559 -0.2874184 0.5319052 facility Expenditure per visit 0.4818344 0.2299665 2.095 0.036 0.0311084 0.9325604 Constant -4.2440820 0.7240204 -5.862 0.000 -5.6631360 -2.8250280 Number of obs. 4507 Pseudo R2 0.176 A4-6 C. Insurance Chapter 3 makes the argument that public provision of health services serves as insurance as well as a health delivery function. The reasoning is that private demand for some health services will not emerge, or will be extremely limited, without insurance being available. Many services are simply too expensive for people to cover out-of-pocket, but they could afford actuarially fair insurance for those same services; that is, they could afford an annual payment of the cost of the service times the probability of needing that service in a given year. Further, they would be willing to make that payment for the peace-of-mind or protection from financial risk that such insurance provides. For quite standard reasons,' insurance markets may be very limited or even fail to exist. In this case the lack of ability to buy a desired product imposes a real loss to an individual's welfare. The government can improve welfare by overcoming this lack of an insurance market. It might provide insurance itself. However, if managing or regulating insurance is too difficult, it may choose to circumvent the explicit provision of insurance by offering free (or highly subsidized) care as another way to eliminate financial risk. The value of this improvement in welfare can, in principle, be measured by examining the willingness-to-pay for insurance. However, this is rarely, if ever done. An approximation can be made by appealing to the concept of a "risk premium." This concept is illustrated in figure A4.2 and measures the amount of money someone would be willing to pay to avoid sudden losses of purchasing power, in this case from medical expenses. The figure shows a utility function, or, the relationship between disposable income and the satisfaction one derives from it. It exhibits decreasing returns in that it starts very steep and gets progressively flatter; an extra dollar to a poor person is worth a lot more than it is to a rich person. Without insurance again an illness, a person's income will be Y if he or she stays healthy, but will be Y-C if he or she becomes ill. For a given probability of illness (p) expected utility is found by adding the utility when ill times the probability of being ill to the utility when healthy times the probability of staying healthy. In symbols, this is: pU(Y-c) + (1-p)U(y). Note that this same level of utility could be obtained by making a payment of V+pC each period independently of health status, thereby avoiding any uncertainty of income at all. The quantity pC covers the actual, expected cost of the service and the quantity V is the "risk premium," the extra value attributable to reducing risk itself. It is also equivalent to the maximum one would be willing to pay for insurance against this illness or the amount that makes a person indifferent between being insured and not being insured. Mathematically, it is expressed as: V= Y- U-'(pU(Y-C) + (l-p)U(Y)). Since the utility function is unknown, one has to be assumed. The conventional choice is U(X) X(l'x)/( I -a) where X = Y or Y-C depending on whether the person is sick. The parameter oa is called the measure of relative risk aversion and basically governs the curvature of the utility function in A4. 1. The higher is oc, the more curved the utility function is, the more risk averse the person is said to be and the more valuable is insurance. The value of insurance, therefore, depends on four terms: Y, C, a and p and is specific to both the person and the condition. To construct figures A4. 1 and A4.2 (repeated below), the relevant parameters ' There are two such reasons. The first, moral hazard, refers to the effect that being insured will have on people's behavior that could affect the cost of providing the insurance. In health care, insurance might lead to excessive demand for treatments or amenities that drive up the cost of provision but which are only demanded because the price facing the patient is low or zero. The second, adverse selection, refers to the different likelihood of applying for insurance given people's own assessment of their need for it. Low-risk people may not buy insurance at a cost that reflects the average probability of needing the service. If they don't buy insurance, the average probability, based on those who do buy it will, in turn, rise. This may lead people of moderate risk to stop buying and the whole process repeats itself. In the end, it is possible that no one would be insured even though they would all want insurance at the expected cost based on their own probability. A4-7 were taken from the PPV survey. Income and costs vary in the figures but the parameters a and p were chosen to be I and .02 respectively. The value for risk aversion is conventional and follows Binswanger (1978) work in India. The value for the likelihood of disease was assumed to be .02 but the results are not very sensitive to the probability assumed. Figure A4.1 Value of Insurance by Household Per Capita Figure A4.2 Value of Insurance by Cost of Insured Service Consumption 60 1 _ __ _ 2sH .... h.Id.... -p n(.I..) providing the service, pC. This*norwaizto allowsssetosse -he gmn inb wiehlfaefrmrplcn insurance alone relative to the cos of provgh Incomelhousehopsp o I_~~~~~~~~~~HlghBOOsteMce * . o ~ ~~~~~~~~~~~ ~ ~~~~ ~ ~ ~ ~ 72005- _____.____/__ two figures it can alo beod oshmtown thaRtel vleoas,rne safacino expeteotst lsvvriscit p and e, falling with the former and rising with the latter. This perspective yields several conclusions about priorities for public spending. Public spending should achieve the highest degree of gain in welfare relative to the status quo without public spending, i.e., relative to what the private market would provide. If the private insurance market does not exist, then public spending is worth V, over and above the value of the service itself. Taking only the insurance effect into account, then priorities for public spending should be higher for high cost items (since V/pC rises with C). Note this effect is the exact opposite of measures of "cost-effectiveness" sometimes proposed as an allocation criterion. Also, the insurance value per Reais of public expenditure goes down with p. This means that greater welfare gains per Reais of public expenditure can be achieved by guaranteeing coverage of the rarer diseases. Note this effect is the opposite of measures of "burden of disease" which are also sometimes proposed as an allocation criterion. The insurance problem is only one of several in the health sector and so does not provide a complete ranking of priorities. It is important enough, however, to cast doubts on the usefulness of methods of priority making which both ignore this effect and give answers so much at odds with it. A4-8 Appendix 5. Education Table A5.1 shows several earnings functions estimated from PPV data. The dependent variable in each case is the logarithm of earnings from different types of water-autonomous farmers (operating their own farms), farm workers and all other living in rural areas. The results show strong relationships between education and earnings for both farm workers and others in rural areas. There does not appear to be a correlation between education and earnings on one's own farm. Farm size and, interesting from a policy perspective, electricity dominate these results. This sort of analysis is limited in several ways. Finding a relation between education and output does not necessary imply and increase in productivity as a result of education. Without modeling the alternative labor-market options, the correlation could be a result of reverse causation. Only this opportunity (farms which are particularly lucrative will be able to attract people with education from other work for which they are qualified). For example, farm work may not improve education but if someone is educated, only better paying farm work, such as managerial position, would even be considered. Since we don't know what is included in the category of "farm work" this remains a possibility. The result is to bias the coefficient upwards. The same effect should operate for autonomous farmers, however, making the absence of a significant effect even more surprising. Table A5.2 shows the estimates of the demand for schooling. The equation is from a logistic regression of school attendance. Notable in the results are the effect of time to get to school (strongly negative), parents' education and opportunities to work on farms (also strongly negative). A5-1 Table A5.1 Determinants of Earning Earnings function for autonomous farmers Number of Observations 684 Adj. R-squared 0.1158 Income from crops sold plus value of home producedfood Coefficient Standard error t-value Log of age 4.1284020 7.0066380 0.589 Log of age squared -0.4904251 0.9371855 -0.523 Log of hectares owned 0.5150498 0.0613723 8.392 Completed primary education 0.4359721 0.3823280 1.140 Completed secondary education 0.1532105 0.3913517 0.391 Completed tertiary education -0.0620838 0.4863322 -0.128 Household has electricity 0.7843949 0.2028026 3.868 White -0.2979175 0.2793564 -1.066 Black 1.9128910 0.7034074 2.719 Female 0.0361827 0.3074206 0.118 Constant -5.3205420 13.0442000 -0.408 Earningsfunctionforfarm workers Number of observations 2456 Adj. R-squared 0.1952 Log of monthly income Coefficient Standard error t-value Log of age 6.6506070 0.6091099 10.919 Log of age squared -0.8433777 0.0837276 -10.073 W.Mhite 0.2312619 0.0374219 6.180 Black 0.0572902 0.0693103 0.827 Female -0.5355521 0.0406725 -13.167 Completed primary education 0.3292831 0.0512244 6.428 Completed secondary education 0.8482490 0.0917140 9.249 Completed tertiary education 1.7724560 0.1858148 9.539 Constant -7.7267100 1.1028720 -7.006 Earnings function for individuals living in rural areas Number of Observations 875 Adj. R-squared 0.2172 Log of monthly income Coefficient Standard error t-value Log of age 6.4683790 0.9622953 6.722 Log of age squared -0.8993225 0.1420759 6.330 White 0.2468350 0.0547983 4.504 Black 0.0728258 0.1019291 0.714 Female -0.5867659 0.0587815 -9.982 Completed primary education 0.0931594 0.0657015 1.418 Completed secondary education 0.8054444 0.1027631 7.838 Completed tertiary education 0.3681424 0.1058406 3.478 Constant -6.4565640 1.6137280 -4.001 A5-2 Table AS.2 Education Demand Standard In school Coefficient error z P>z 195% Conf interval] Log real per capita 0.8460796 0.0701682 12.058 0.000 0.7085524 0.9836069 expenditure Log time to school -0.3283226 0.0558048 -5.883 0.000 -0.4376979 -0.2189473 Autonomous farmer -0.3020260 0.3432934 -0.880 0.379 -0.9748686 0.3708166 Farm worker -0.6798987 0.1360775 -4.996 0.000 -0.9466056 -0.4131917 Region Northeast municipal 0.2248967 0.1903704 1.181 0.237 -0.1482224 0.5980157 Northeast urban 0.2174919 0.2131069 1.021 0.307 -0.2001900 0.6351738 Northeast rural -0.1468069 0.2049124 -0.716 0.474 -0.5484278 0.2548139 Southeast municipal -0.4268159 0.1895592 -2.252 0.024 -0.7983452 -0.0552866 Southeast rural -0.6824656 0.2038744 -3.347 0.001 -1.0820520 -0.2828790 Race Black 0.0190682 0.1779069 0.107 0.915 -0.3296229 0.3677593 Mixed -0.1229431 0.0935116 -1.315 0.189 -0.3062225 0.0603362 Gender Male -0.0157104 0.0827463 -0.190 0.849 -0.1778902 0.1464693 Mother's Mother completed 0.3263114 0.1099325 2.968 0.003 0.1108477 0.5417751 education preprimary Mother completed primary 0.1973571 0.1243410 1.587 0.112 0.0463468 0.4410609 Mother completed 0.7137726 0.2096842 3.404 0.001 0.3027992 1.1247460 secondary Mothercompletedtertiary 0.9236553 0.3721798 2.482 0.013 0.1941963 1.6531140 Father's Father completed 0.4897136 0.1038152 4.717 0.000 0.2862396 0.6931876 education preprimary Father completedprimary 0.1260243 0.1261653 0.999 0.318 -0.1212551 0.3733037 Father completed 0.2780432 0.1915162 1.452 0.147 -0.0973215 0.6534080 secondary Father completed tertiary 0.4613455 0.3706711 1.245 0.213 -0.2651566 1.1878480 Age 5 years old -2.0818110 0.2233150 -9.322 0.000 -2.5195000 -1.6441210 6 years old -0.9171299 0.2259310 -4.059 0.000 -1.3599460 -0.4743134 8 years old 0.4291867 0.2662375 1.612 0.107 -0.0926291 0.9510026 9 years old 0.6588413 0.2857140 2.306 0.021 0.0988522 1.2188300 10 years old 0.7056835 0.2843435 2.482 0.013 0.1483805 1.2629860 11 yearsold 0.9711607 0.2999934 3.237 0.001 0.3831844 1.5591370 12 years old 0.8975675 0.2954805 3.038 0.002 0.3184362 1.4766990 13 years old 0.0561038 0.2469040 0.227 0.820 -0.4278192 0.5400268 14 years old -0.2421556 0.2382470 -1.016 0.309 -0.709111 0.2247999 15 years old -0.7205688 0.2286364 -3.152 0.002 -1.168688 -0.2724496 16 years old -1.1626230 0.2270591 -5.120 0.000 -1.60765 -0.7175950 17 years old -1.8290690 0.2205277 -8.294 0.000 -2.261295 -1.3968420 18 years old -2.5642110 0.2210562 11.600 0.000 -2.997473 -2.1309490 Constant -1.0826710 0.4306258 -2.514 0.012 -1.926682 -0.2386598 A5-3 Appendix 6. Drought and Poverty in the Northeast A. Minimizing the Effect of Droughts on the Poor' The states of the northeast, from Piaui to Bahia and down to the north of Minas Gerais, share in their interior a large semiarid region of approximately 1 million square kilometers, or approximately 60 percent of the total northeast area. This is a highly drought-prone area. The coastal forest zone that goes through six states from Rio Grande do Norte to Bahia is less subject to droughts, but the transition zone between the two, known as the Agreste, is highly affected. Rainfall in the northeast is highly variable. Recurrent interannual droughts are related to the El Ninio and the Atlantic Sea Surface Temperatures (SST) and, to a lesser degree, to local geographical conditions. The occurrence of an El Niflo event like the 1997-98 one carries a high probability of droughts in the semiarid northeast, especially if coupled with low SST in the South Atlantic, as in 1998. The impacts of droughts in the northeast economy and society have been recorded since historical times. (See Magalhaes et al. 1988 and Magalhaes and Glantz 1992). Subsistence rainfed agriculture is highly affected. Small landholders and shareholders who depend on this activity for their living usually practice this kind of agriculture. These two categories of farmers, plus the rural landless workers (they comprise the majority of the rural population, the actual core of poverty in the rural northeast), are heavily affected by droughts. In the absence of government relief programs, their major survival strategy is migration. This kind of migration strategy to the northeast coastal zone and later on to other Brazilian regions-the Amazon, Rio de Janeiro, Sao Paulo, and Brasilia-have been initially adopted by the indigenous peoples that inhabited the northeast prior to the European occupation, and later on by the descendents of the new settlers. Historically, major droughts like the one in 1877 caused a massive exodus leading to death and starvation of hundreds of thousands of people along the way. Organized government responses to the drought problem started in the second half of last century, always triggered by an extreme drought event. In 1909, the Government of Brazil created the Federal Agency of Works Against Droughts (IFOCS), with objectives similar to those of the U.S. Bureau of Reclamation. IFOCS (now DNOCS) developed a large program of water and infrastructure development, and also helped to create a knowledge basis of the region's geography and environment. DNOCS would also manage relief programs during droughts. In 1945 the creation of Hydroelectricity Company of the Sao Francisco River (CHESF) initiated the production of hydroelectric energy, which then supported urbanization and industrialization. In 1951 the Bank of the Northeast was created to finance private sector projects. And in 1959, the Superintendency of the Development of the Northeast (SUDENE) was created to coordinate development plans for the whole region. ' The following material is based on an Aide Memoire by Martin Ravallion on a mission by Joachim Von Amsberg, Antonio Magalhaes and Martin Ravallion to northeast Brazil in the first week of September 1998, at the invitation of SUDENE. The mission's aim was to obtain an overview of the immediate drought-relief effort, and how it might be strengthened in the future. The mission's mandate did not include the govemment's longer- term efforts to reduce drought sensitivity. The mission talked at length with the head and various staff of SUDENE, the Secretaries for Planning, Rural Development, Water Control and other senior officials of the state governments of Pemambuco and Ceara. The mission also visited drought-affected municipalities in Pemnambuco and Ceara, as well as in the neighboring states of Alagoas and Paraiba. About 10 workfare project sites in these municipalities were visited. Informal interviews were held with some of the participating workers. A6-1 After SUDENE, from the 1960s to the 1990s, a series of development plans and programs were implemented, aiming at industrialization, irrigation, infrastructure development, rural development, and human resources development, all with one objective-to increase regional resistance to droughts. At the same time, institutional development at the level of the states strengthened their capability to plan and implement their own policies and eventually to take over more responsibility for federal programs. During all that time, government policies related to the drought problem in the northeast have addressed two main objectives: (1) increase regional resiliency to future droughts through less climate-dependent economic and social development ("permanent" policies), and (2) provide emergency relief to the stricken population during drought events. The "permanent" policies and programs have reduced the dependency of the regional economy on climate variability. The economic share of infrastructure projects (transports, communications, energy, water resources), industrial development, and urban services has increased. Water infrastructure spread all over the semiarid area has facilitated the supply of water during drought and nondrought years. The economic impact of a drought has decreased as the share of agriculture in GDP has decreased. But it is still high in small urban communities that depend on subsistence agriculture. Thus, at the aggregate level, the economic impact of present droughts is smaller than it used to be in the past. However, a vast majority of the rural populations still depend solely on agriculture and thus continue to be directly impacted by droughts. In the northeast droughts are a serious social as well as economic problem affecting millions of poor living in rural areas and at the periphery of small and medium cities in the semiarid areas. On the other hand, the experience related to relief programs-to provide jobs, food, and water to the drought affected people-is also very rich and varied. Many lessons, good and bad, can be drawn. All kinds of strategies, works (large infrastructure or small community projects), institutional arrangements (federal-state-local relations, community participation) have been implemented, and many problems (the "drought industry," for instance) have been resolved. Overall, these programs were successful in that they provided relief for millions of peoples (in 1983, for instance, more than 2.5 million people were hired in the workfare projects). Salary payments were sufficient to keep the urban economy of the interior northeast working and to slow down the migration process. The presence of the World Bank in the northeast started in mid-1970s and so far has supported "permanent" policies in poverty alleviation, water development (water policy), infrastructure development and human capital development (education and health). The Bank has not yet addressed relief programs. For the first time, a World Bank mission has assessed the drought situation and government responses underway during the 1998 drought. The findings of that mission are reported below. B. The 1998 Drought The current drought in northeast Brazil is among the worst in 15 years, affecting more than 10 million people in 8 northeastern states. A massive relief effort is underway. The three main components are (1) targeted food handouts, (2) a workfare program in which participants must work or attend training to obtain benefits, and (3) a subsidized credit scheme. (Other components include drinking water distribution, not discussed here.) Three million people are receiving a food handout. A further 1.2 million people were employed on the workfare component up to September 1998 at a cost of about R$ 110 million per month (about R$100 million). And about 100,000 families are receiving subsidized credit amounting to about R$450 million. A6-2 The data currently available do not permit a rigorous evaluation of the drought-relief program. There are some key determinants of gains to drought-affected families that simply cannot be measured. The annual national sample survey (PNAD) of households will be done soon. PNAD is not believed to measure rural incomes well, but it may still help in assessing income and other effects of the drought. However, the data will not be available until late 1999. Nor is it currently possible to provide a full accounting of what is being spent on the relief effort, given that other public programs are also being used to help mobilize resources, such as covering the nonwage costs of relief work projects. The information system for some key components of the drought-relief effort is thus less than ideal, reflecting in part the urgency of the situation and the logistic complexity of the operation. What follows is a first, partial assessment of the current relief effort. Food Handouts This is conceptually the simplest of the three components. It is basically an expansion of the existing food distribution system under the Community Solidarity program, which is permanent in poor municipalities. The coverage has been extended to include all designated drought-affected municipalities. Temporary Municipal Councils (MCs), set up specifically for the drought, decide who should receive the food handout in drought-affected municipalities. No data are available for assessing how well targeted the food handouts are, though there is nothing to suggest a serious misallocation. The handouts are basic necessities, unprocessed, or in the form of prepared meals, which are unlikely to be of much appeal to the non-poor. The Municipal Council also has the task of determining what workfare projects will be done, and who will get the work (discussed further below). The MCs are crucial to the success of the drought-relief operation, though a systematic assessment of how well they are doing their job in distributing food and work is impossible to determine from the data available. The MC was introduced (initially in Ceara in the 1987 drought) in an attempt to get around past problems of corruption in drought relief, whereby some local mayors and large landowners were known to be diverting the resources to other purposes; a profitable "drought industry" had evolved. In Ceari in 1987, the local agricultural extension worker chaired the MC. In northeast Brazil as a whole, in 1998, it is typically the local mayor. The composition of the MC is stipulated by central regulations, which severely constrain the ability of mayors to manipulate the MC. The broad membership of the MCs provides a very important additional check on the many ways in which funds for drought relief could otherwise miss their target. This is reinforced by public information and disclosure practices. Public information and press coverage clearly help in constraining misappropriation. There have been public denunciations of mayors found to be manipulating assistance to the poor for other purposes. Though these checks have helped, it is clear that the success of the MCs has varied. Workfare Programs This component provides work on various community projects and training. The MCs propose the projects. The selection of projects is made by state coordinating committees. In previous droughts the emphasis was on water-related projects, but this has now broadened to include a wider range of community projects and training. Examples of projects covered include underground dams and similar small-scale irrigation projects, water and sanitation projects, building and maintaining community facilities (schools, health clinics, parks), rural road construction, and training projects. The latter entails a combination of basic literacy skills and knowledge about droughts and how to dampen their impact. The training component varies from one to 2.5 days per week. The choice of the wage rate is critical to the success of workfare programs. The wage rate determines who wants to participate, and (with the budget allocation) how many can actually be A6-3 accommodated on the program. A good rule of thumb is that the daily wage rate should be no higher than the market wage for similar work at normal times. This helps the program reach the poor by both assuring that a higher proportion of those participating are among the poor and by assuring wider coverage of the poor, in that more workers will be able to get work. Wide coverage, ideally reaching as many workers as possible at this wage rate, also brings risk benefits to the poor by making it more likely that they will get work when needed, and make them less vulnerable to local administrative discretion in allocating relief. It will also protect work incentives by reducing dependency on the scheme in that workfare should not be more attractive than regular work when it becomes available. In the case of the relief work provided during Brazil's drought, the center pays the bulk of the wages plus a contribution to the nonwage costs. The center's contribution to the wage is R$65 per month for a 27-hour working week. The center also pays 20 percent of this amount for nonwage costs (tools and materials), representing R$15.6 per person per month. The states top up the center's contributions. For the wage bill, this amounts to an extra R$ 15 in most states. The exceptions are the states of Ceara, which pays an extra R$25 to bring the wage up to R$90, and Piaui, which pays nothing extra. The states also contribute R$3.12 per month per worker to nonwage costs. The daily equivalent of the average workfare wage is roughly the same as the average wage rate for casual wage labor in a normal year, which is about R$5 per day in northeast Brazil. The statutory minimum wage is R$130 per month for a 5.5 day week, or about R$5.50 per day which is also close to the implicit daily wage on workfare. Allowing for the training component (which workers clearly prefer), workfare is almost certainly preferred to the wage work available in a normal year. The workfare wage is probably well above the shadow wage rate during the drought, implying sizable transfer benefits to participating families. In the drought-affected municipalities visited, current wages for unskilled manual labor (when available) were reported to be 30 to 50 percent below the level in the predrought year. One indication of the transfer benefit was possible from the experience in one drought municipality (Pao de Azucar in Alagoas). The mayor there gave workers the choice between the normal workfare wage (R$80 per month for a 3.5-day work week) and a wage of R$95 for a five-day work week. Workers unanimously chose the latter, indicating that their reservation wage rate was less than R$2.50 per day at the margin, less than one-half of the workfare wage. Given the available budget, the workfare jobs have to be rationed at the wage rate currently paid. MCs select beneficiaries based on their incomes and losses due to the drought. Only one participant is allowed per family (though an extra participant is often allowed for large families), and the family can have no other income source (though it is unclear how well the latter criterion can be implemented.) State and local governments retain power over the geographic allocation. Drought- affected municipalities are identified on the basis of rainfall data. The program is confined to rural areas. The data currently available do not permit an assessment of how well the relief is targeted geographically within the municipalities declared drought affected. In one municipality, it was clear that the geographic allocation could be improved. (One village was receiving relief work when there was ample work available in stone cutting, which was unaffected by the drought.) There are also clearly families hurt by the drought in municipalities not declared as being drought-affected. The absolute rainfall data used for declaring a drought-affected municipality does not reflect the importance of relative rainfall to livelihoods; rainfall might be well below normal, with serious welfare consequences, but still above that in many other areas. It is also clear that there are people in A6-4 small towns and suburban areas (where many farm workers live) who are hurting from the drought but are not getting relief. Some states, such as Ceari, appear to be meeting the bulk of demand within the set of eligible participants. (Ceara added 40,000 workfare slots to those funded by SUDENE, the federal development agency for the northeast that is in charge of the relief operation, so that the state could meet what it estimated to be the demand for this work within the set of eligible participants.) However, by juggling the eligibility criteria one can assure good coverage of those deemed eligible. (Though some municipalities may well not be managed well enough to even achieve this goal.) Current coverage is probably about 60 to 70 percent of the number of workers in drought- affected areas willing to work at this wage rate. (The rural labor union estimates that there are two million able-bodied adults who need the work.) There also appears to be a sizable unmet demand for this work outside the set of households deemed to be eligible. As noted above, some of this unmet demand is found in small towns and the urban periphery, where non-negligible numbers of farm workers live. It is undoubtedly also found among poor large families, who need more than one slot. No doubt there are also families with one (nonrelief) wage job for which (given family size) the earnings are too low to make ends meet, particularly since the wage rates for the work that is available appear to have fallen sharply during the drought. So the current eligibility rules are undoubtedly excluding non-negligible numbers of poor families in need of assistance. In terms of the overall efficacy of the drought-relief operation, there is a strong case for relaxing the current restrictions on eligibility and geographic targeting to assure wider coverage of those needing help. The average labor intensity (share of wages in cost) is a key factor in the cost-effectiveness of workfare programs relative to alternative transfer schemes using the same gross budget. The labor intensity of the workfare projects is claimed to be 81 percent (80/(80+15.6+3.1)). This is very high even for a workfare scheme. However, the calculation is deceptive since it was clear from interviews with a number of local mayors that they top up the nonwage cost to allow a wider range of projects consistent with their development plans in the area. It is not possible with the current information system to calculate the actual labor intensity. With a full accounting of nonwage costs, the share of wages in the projects visited would probably be about 75 percent. Effort is made to assure that the labor intensity is high; for example, construction workers make bricks using local clay and makeshift kilns. Subsidized Credit The Bank of the Northeast (BNB) runs the credit program in response to the drought. There was a significant change in BNB's overall approach in 1995, when the regional bank became more client oriented, with a more demand driven, "bottom-up" approach. One implication is the greater emphasis given to training as a complement to credit. Another is a greater field presence, with the bank represented in all 1,800 northeast municipalities. This change has helped them respond more effectively to the drought. For example, BNB was able to disburse funds very quickly once the federal government made the credit available in May 1998; by July all money was committed. BNB's drought loans operate under similar rules to its general lending, which continues during the drought. The main difference is that the drought loans have more favorable terms. Like all BNB loans, the interest rate is well below the market rate. For investments in farm capital the normal interest rate is 8 to 9 percent; for the drought loans it is 6 percent. For working capital the drought loans are at an interest rate of 3 percent, as compared to 6.5 percent for other similar BNB loans. The repayment period is about the same for the drought loans. However, there is a grace period of up to 4 years (2 years for working capital) to reflect the impact of the drought on farm revenues. Also there is A6-5 a 50 percent rebate on principle and interest for drought loans. The federal government covers the rebate. The drought also put pressure on other BNB's operations, including its new microcredit program started in October 1997. This program provides small loans (R$100-1,000) to urban microenterprises at an interest rate of 5 percent, to be repaid in three months. The drought-induced migration to urban areas increased demand for these loans. The volume of lending has been increasing in response to the drought. BNB's drought-relief operation is clearly underfunded given the demand. Lack of credit was mentioned often in field trips to drought affected areas. In one municipality only 35 out of 300 applications for BNB's drought credit were approved. In another, only 100 were approved out of 10,000 applications. In another case, the mission came across a large "town-hall meeting" of fanners protesting that their application was turned down. Demand is undoubtedly high given the generous terms. Nonetheless, it is clear that there is a large unmet demand for credit at this time. An expansion in aggregate credit availability would help. The program is not sufficiently focused on reaching smallholders in most pressing need, or sufficiently linked to the drought-relief operations more generally. BNB reports that 60 percent of the drought loans go to "family farms." (A finer breakdown of the borrowers was not available.) A family farm, as defined by BNB, has no more than two employees, no more than four times the land area needed for subsistence, and no more than R$27,000 annual gross revenue. It is not known what proportion of all farms are family farms. (The definition is not easily interpreted in terms of the landholding data available in the Living Standards Measurement Study for Brazil.) However, it is clearly implausible that all "family farms," as defined above, are poor by Brazilian standards. And very few of those among the other 40 percent of borrowers are likely to be considered poor. This would not be a concern if the loans were not so heavily subsidized; but given the subsidy, one should ask whether it is reaching the poor. It might be argued that second round employment gains to the poor will accrue from loans to relatively well off landowners. However, this will take time, and is not of obvious relevance for short-term drought relief. BNB offices have no formal contact with the MCs for drought relief, though they do consult with Municipal Credit Commissions. A reorientation and redesign of BNB's drought-relief loans should be considered. A better allocation of credit for drought relief would be possible by putting a ceiling on the loans, at (say) the current average loan size. Another step would be to cut the rebate. Together with an expansion of lending volume, these steps could help BNB meet the demand for credit during a drought. Integration of the Three Components Under present rules, a farmer receiving a drought-relief credit cannot also receive either the food handout or workfare. With the present extent and distribution of the credit component this restriction is unlikely to bind; most of those who receive the credit are unlikely to be attracted by the workfare or food handouts. However, with a more pro-poor allocation of credit, the restriction should be reconsidered. There are potential complementarities between these components of drought relief, which should be exploited. For example, access to the food supplement may help a smallholder make better use of the credit, by freeing up the labor needed to work with the extra capital (financed by the credit). This argument is less compelling for workfare, but here too one should not presume that a restriction on joint access is desirable; sending one family member to the workfare program may well assure food security for the whole family, so that other workers can make the best use of the credit. A6-6 At the same time, the present drought-relief operation does allow a family to receive both the food and the work. This is also questionable. There is a well-recognized case for combining these two elements of drought relief such that workfare is aimed at the able-bodied poor, while the food handout is intended for the nonable-bodied poor. It is quite possible that restricting a family to just one of these two components would make more sense. Implications for the Future While the 1998 drought-relief effort has been exemplary in a number of respects, and has undoubtedly saved a great many families in Brazil's northeast from destitution or worse, there are three main areas where the relief effort could be improved. (1) Preparedness and speed of response. This is often a weak point of drought-relief efforts and this drought was no exception. The federal funding response was slow; civic meetings to mobilize action were held as early as August 1997 in one state (Pemambuco), and the signals were clear since the beginning of 1998. Yet it was not until May 1998 that the center went into action (though Ceara had started its program in December 1997). (2) Coverage of the affected population. The overall coverage of the 1998 drought-relief effort is less than ideal. In the aggregate, a higher funding level is needed. There is also scope for improving efficiency in reaching those in greatest need. (3) Coordination of drought-relief efforts with antipoverty policy in normal years. There have been some attempts to coordinate the drought relief with other pre-existing programs (in Ceara, for example, a World Bank poverty alleviation program was accelerated in response to the drought). However, these efforts have been ad hoc and partial. More systematic coordination efforts must start from the realization that drought is intimately connected to the problems of rural development more generally: high risk, credit and insurance market failures, underinvestment in local public goods, and often weak local institutions. A permanent safety net program in the northeast could help deal with all three problems. This would extend the coverage of the workfare component of the current drought-relief effort to include nondrought times (at which demand would be much lower, but almost certainly not zero). It would also relax the current eligibility restrictions on relief work. Such a program could thus combine the best features of a low-wage employment guarantee scheme with current social funds for supporting labor intensive community projects in poor rural areas. Under such a program, the federal govemment could guarantee to finance (say) 15 days a month of work on community projects and training to any adult. To assure that the work reaches those in need, and protects incentives to take up regular work when available, the wage rate should be set no higher than the local wage rate for unskilled agricultural labor in a normal year, and somewhat lower than this for training. The work would have to be on a technically feasible project proposed by a bona fide MC. As in the present drought-relief program, the center could pay a small amount extra for nonwage costs, though this will often be inadequate and further funding of the nonwage costs may have to be secured from other public programs at state or federal level, or by cost-recovery from non- poor beneficiaries (Ravallion 1998). The current drought-relief MCs would then become a permanent institution dealing simultaneously with droughts and the problems of poverty, risk, and rural underdevelopment in nondrought years. The MCs would maintain a shelf of useful projects in the community. With wide public knowledge of the existence of a federal employment guarantee on community work, and the A6-7 permanent MCs ready with a shelf of such projects, the basis for a rapid response would be generated from the bottom up, rather than relying on administrative discretion from the top down. The budgetary outlay on such a safety net would no doubt be highly variable over time, though possibly no more so than under the present arrangement in which large sums of money have to be injected into an overdue drought-relief operation. To cover the variability in disbursements, a central safety-netfund could be established into which the center makes regular payments sufficient to cover a normal sequence of good and bad years in agriculture, as well as the likely demand in nondrought years. However, this recommendation requires more careful assessment to ensure it does not generate the typical drawbacks common to earmarked funds. A more thorough review of drought prevention and relief programs, which critically re-examines Brazil's own rich experience, as well as international experience, may be helpful in identifying overlooked or better interventions that are politically and technologically more feasible today than they were in the past. Some examples are cited below which highlight a variety of relatively successful interventions, some emphasizing better coordination of public responses (South Africa), others incorporating markets and private participation in public programs (India), some emphasizing means-testing (Tunisia), and others self-targeting (Argentina). None are meant to be adopted in toto or without modification to Brazilian circumstances. Some International Experiences Worth Revisiting South Africa Early warning systems, coupled with a well-conceived intervention program, is best illustrated by the case of Southern Africa in the 1990s. Southern Africa faced severe droughts in 1991/92 and 1994/95, and seriously decreasing food production on both occasions. In 1991/92, cereal production was reduced by one-half, leaving subsistence farmers and their families in dire need with approximately 18 million people facing starvation (FAO 1994). However, effective early warning, rapid regional coordination, and adequate international support resulted in a successful relief effort that overcame widespread food shortages and the threat of famine. As a result of the experience of the 1991/92 crisis, the impact of the 1994/95 drought emergency, while serious, was less devastating. In the latter case, although the seasonal rains began on schedule in October 1994, allowing planting, the rains soon tapered off to a halt, leaving withered plants in parched fields. The Global Information and Early Warning System (GIEWS) and other FAO-assisted national and regional early warning systems in the Southern African Development Community (SADC) issued their initial warnings of impending drought in December. Soon after the first warnings were issued, governments, donors and UN agencies began meeting together to make plans for moving large-scale relief aid into the region should it be necessary. These preparatory arrangements were critical. In southern Africa, as elsewhere, drought areas are typically isolated and difficult to reach and importing relief food from overseas or distant locations, often via poor roads and inadequate ports, takes time. In January and February 1995 the impending onset of food difficulties as a result of drought- induced crop losses in most countries of the subregion was confirmed. In March and April, FAO and World Food Programme (WFP) experts arrived to assess the situation. The agencies jointly approved an emergency operation to cover urgent relief needs. FAO issued a special alert, asking the international community for relief assistance. In June, donors reacted to an appeal for help by A6-8 providing I million tonnes of food aid and other assistance, which averted malnutrition and starvation. A key lesson from this experience is that setting up an early warning system is not sufficient. In particular, it may not address the role of political institutions that might block efforts at famine prevention or capitalize on it for political ends. Most early warning systems are set up after a country has experienced a major drought. This is in part due to the availability of drought-relief aid at the time (which is often targeted towards the establishment of warning systems), and in part to the government's attempts to reassure the general public that it has the situation under control and will not again be caught off-guard. The establishment of an early warning system in such circumstances can become counterproductive if once the crisis is over it lulls decisionmakers into a false sense of security and diverts attention from the primary determinants of malnutrition and starvation: persistent underdevelopment, householdfood insecurity, and vulnerability. India2 One of the better-documented cases of averting major socioeconomic dislocations (such as famines or mass migrations) when faced with drought comes from the experience in India. While per capita food production has not increased greatly since India's independence in 1947, sources of livelihood for the rural population have diversified, and the general increase in the productivity of agriculture has substantially raised the general living standards and reduced the insecurity of rural incomes. When the public distribution system was unable to fill the initial gap between food availability and requirements, the income generation program enabled the affected population in rural areas to augment the publicly distributed food through purchases on the market. Private traders actively facilitated timely interstate food movements. In addition, the government's food policy has largely succeeded in stabilizing food prices (through the accumulation of large buffer stocks) and in insulating consumption from the fluctuations in production. Even though food consumption in drought years was lower than other years, large-scale dissatisfaction was averted because the food deficit was well distributed among various socioeconomic groups. Tunisia3 The programs introduced in the semiarid areas in the Maghreb region of Tunisia have many similarities to successful efforts in the northeast of Brazil to lessen the damaging effects of droughts through preventive actions, such as investing heavily in water development projects to avert a water crisis associated with highly variable and irregular rainfall and limited water resources. 4 However, the real story in delinking production shortfalls from aggravating the consumption of the poor was the ability of the public sector to successfully target assistance to the poor through means testing programs. Located on the southern rim of the Mediterranean Sea, Tunisia is a semiarid country, much like most of northeast Brazil. Water resources are scarce and in some areas, particularly in the south, water demand exceeds locally available fresh water resources. In the past two decades, the country has been faced with numerous droughts but has managed to avert serious food crises. 2 See Dreze (1995). 3World Bank (1995). 4 See Ayub and Kuffner (1994) A6-9 Periodic significant declines in agricultural production continued during drought years; however, these declines did not have a large effect on consumption due to government initiatives such as public works employment, and a well-functioning public distribution system. In particular, the poorest sections of the population were protected by means-tested, centrally administered programs for public income transfers. About 60 percent of the needy families benefiting from this program lived in rural areas. The targeted assistance to the poor using means-tested programs were not without problems (such as complex administration and dated eligibility lists), but they have been reasonably successful in providing assistance to the poorest households. Most of these programs were run by the central government thereby reducing the possibility for decentralized political misappropriation. However, since 1993, in part to protect against central government mismanagement, civil society has also been very active in monitoring and providing targeted assistance. Argentina Argentina's recent experience in helping the poor cope with a sharp increase in unemployment is also relevant to the situation in the northeast of Brazil. A recent World Bank supported project in Argentina, the Trabajar 2 Program, offered low wage work on community projects in poor areas as a means of helping poor unemployed workers and poor communities. By paying careful attention to the setting of the wage rate it was possible to assure excellent targeting of the benefits. Independent evaluations based on surveys found that two-thirds of the program participants came from the poorest income decile nationally (Jalan and Ravallion 1998). There were also improvements (relative to past programs) in how well the indirect benefits from the assets created were targeted to poor areas. A6-10