The distribution of welfare in Peru in 1985-86

LESEfl1S LSM - 42 Living Standards DEC. 1987 Measurement Studv Working Paper No. 42 The Distribution of Welfare in Peru in 1985-86 Paul Glewwe  LSMS WORKING PAPER SERIES No. 1. Living Standards Surveys in Developing Countries. No. 2. Poverty and Living Standards in Asia: An Overview of the Main Results and Lessons of Selected Household Surveys. No. 3. Measuring Levels of Living in Latin America: An Overview of Main Problems. No. 4. Towards More Effective Measurement of Levels of Living, and Review of Work of the United Nations Statistical Office (UNSO) Related to Statistics of Levels of Living. No. 5. Conducting Surveys in Developing Countries: Practical Problems and Experience in Brazil, Malaysia, and the Philippines. No. 6. Household Survey Experience in Africa. No. 7. Measurement of Welfare: Theory and Practical Guidelines. No. 8. Employment Data for the Measurement of Living Standards. No. 9. Income and Expenditure Surveys in Developing Countries: Sample Design and Execution. No. 10. Reflections on the LSMS Group Meeting. No. 11. Three Essays on a Sri Lanka Household Survey. No. 12.The ECIEL Study of Household Income and Consumption in Urban Latin America: An Analytical History. No. 13.Nutrition and Health Status Indicators: Suggestions for Surveys of the Standard of Living in Developing Countries. No. 14. Child Schooling and the Measurement of Living Standards. No. 15. Measuring Health as a Component of Living Standards. No. 16. Procedures for Collecting and Analyzing Mortality Data in LSMS. No. 17. The Labor Market and Social Accounting: A Framework of Data Presentation. No. 18. Time Use Data and the Living Standards Measurement Study. No. 19. The Conceptual Basis of Measures of Household Welfare and Their Implied Survey Data Requirements. No. 20. Statistical Experimentation for Household Surveys: Two Case Studies of Hong Kong. (List continues on the inside of the back cover)  LSMS Working Paper Number 42 THE DISTRIBUTION OF WELFARE D Pmu IN 1985-86 Paul Glewe Population and Human Resources Department The.World Bank Washington, D.C. 20433, USA  - ii - This is a working document published informally by the Population and Human Resources Department of The World Bank. The World Bank does not accept responsiblity for the views expressed herein, which are those of the author and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported by the Bank; they do not necessarily represent official policy of the Bank. The designations employed, the presentation of materials, and any maps used in this document are solely for the convenience of the reader and do not imply the expression of any opinion whatsoever on the part of the World Bank or its affiliates concerning the legal status of any country, territory, city, area, or of its authorities, or concerning the delimitation of its boundaries, or national affiliation. The LSMS working paper series may be obtained from the Welfare and Human Resources Division, Population and Human Resources Department, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A. Paul Glewwe is an economist working in the Welfare and Human Resources Division. December 1987  - iii - PREFACE The Living Standards Measurement Study (LSMS) was established by the World Bank in 1980 to explore ways of improving the type and quality of household data collected by Third World statistical offices. Its goal is to foster increased use of household data as a basis for policy decision making. Specifically, the LSMS is working to develop new methods to monitor progress in raising levels of living, to identify the consequences for households of past and proposed government policies, and to improve communications between survey statisticians, analysts, and policy makers. The LSMS Working Paper series was started to disseminate intermediate products from the LSMS. Publications in the series include critical surveys covering different aspects of the LSKS data collection program and reports on improved methodologies for using Living Standards Survey (LSS) data. More recent publications recommend specific survey, questionnaire and data processing designs, and demonstrate the breadth of policy analysis that can be carried out using LSS data.  - iv - ACKMOWILEDCHKENTS This report is based on the Peru Living Standards Survey which was conducted in 1985-86 by the World Bank's Living Standards Unit and the Peruvian Instituto Nacional de Estadistica (INE). I would like to thank the entire staff of INE for their contributions at various stages of this work. I am also indebted to Jorge Castillo for competent computer programming and to Brenda Rosa for typing several drafts of this paper.  - v - ABSTRACT This paper examines the distribution of welfare in Peru in 1985-86 as measured by (adjusted) per capita consumption expenditures. The data employed are from the 1985-86 Peru Living Stnadards Survey. It is primarily descriptive in nature, but possible explanations of patterns observed are given at several points. The major findings are: l.There is a strong correlation between education and welfare, which highlights the importance of education policies. 2.The urban population is clearly better off than the rural population, and the majority of the poor are found in rural areas, most often as self-employed agriculturalists. 3.Poor households have as many, if not more, working members as a fraction of total members, as a typical Peruvian household. 4.Within urban areas, heads of poor households are primarily sales and service workers industrial and craft workers, both self-employed and employed by private firms.  - vi - TABLE OF CONTENTS .I. Introduction ............................... II. Consumption Data as an Indicator of Welfare....................3 III. The Distribution of Consumption inPeru.......................8 A. The Distribution of Consumption Expenditures by Population Deciles ............ .. . . ........ . 8 .... B. Characteristics of Households per Capita Expenditure Groups.... . ......... .. .. .. ......o.. . .. . .10 C. Ownership of Agricultural Land ......oo................18 D. School Attendance, Housing and Ownership of Durables*.*i.9*..*.o..... ... *..*$... 4.%*......................22 IV. Inequality in Peru...... .... ...... . 30 A. Historical Comparison with other Countries............30 B. Regional Inequality Decompositions. ......... .......... 31 D. Decomposition by Educational Attainment.o.o.o...o....33 V. Characteristics of the Poor in Peru.......,, ........ .36 A. Poverty in Peru - A National Profile..................36 B. Poverty in Urban Peru..................................... 46 C. Poverty in Rural Peru..................................56 VIo Conclusion ....o. o*... ..*.... .#... o * ..** .o ..*o .e o ..............o...66 Referencess ... -0.00.0.... ..O... 0... ............... .0 .... o..o.o..68 Appendix A: Peru Living Standards Survey.*...*o*..oo..*o.....70 Appendix B: Food Consumption as a Welfare Measureo...o........78 Appendix C: Measurement of Inequality ..............................81 continued  - vii - TABLE OF CONTENTS-continued LIST OF TABLES Table 1. Welfare Indicators of Peru and Sri Lanka ...... ...........*7 Table 2. Distribution of Total Consumption by Welfare Deciles.....9 Table 3. Characteristics of Households by Quintiles... ......12-13 Table 4. Characteristics of Households by Regions ............ 14 Table 5. Agricultural Land Ownership in Peru .................l19 Table 6. Distribution of Agricultural Land in Peru ............... 22 Table 7. School Attendance by Welfare Quintiles .................. 25 Table 8. Housing Characteristics in Peru............ .......... ....26 Table 9. Ownership of Durables by Quintiles and Regions ...........29 Table 10. Cini Coefficients on Per Capita Income ..................31 Table 11. Per Capita Expenditures Inequality-Decomposed by Region o .......... o.... ..... .......... .**...... 33 Table 12. Per Capita Expenditures Inequality-Decomposed by Education of Head ....... . .. .......... . ...... . 34 Table 13. Household Size and Per Capita Consumption of the Poor...37 Table 14. Sex of Household Head Among the Poor ............. .......38 Table 15. Distribution of the Poor by Regions.....................38 Table 16. Distribution of the Poor by Language of Interview ....... 39 Table 17. Distribution of the Poor by Employer and Occupation of Head of Household. ..... .... Table 18. Agricultural Land Ownership Among the Poor.....*........ 41 Table 19. Distribution of the Poor by Educational Level ........... 42 Table 20. Illness, School Attendance and Housing of the Poor......44 Table 21. Composition of Household Members Among the Poor ..o*....45 continued  - viii - TABLE OF CONTENTS-continued Table 22. Household Size and Per Capita Consumption of the Urban Poor ..................o.......................48 Table 23. Sex of Household Head Among the Urban Poor ..............49 Table 24. Distribution of the Urban Poor by Employer and Occupation ......... *....................... . ................... 50 Table 25. Agricultural Land Ownership Among Urban Poor ............51 Table 26. Education and Poverty in Urban Areas ....................52 Table 27. Housing Characteristics of the Urban Poor...............54 Table 28. Household Composition Among the Urban Poor..............55 Table 29. Household Size and Per Capita Consumption of the Rural Poor ......... .....e.. .**..9. 57 Table 30. Sex of Household Head Among the Rural Poor............58 Table 31. Language of Interview Among the Rural Poor .............. 59 Table 32. Distribution of the Rural Poor by Employer and Occupation of the Head of Household ..................o61 Table 33. Agricultural Land Ownership Among the Rural Poor ........61 Table 34. Education and Poverty in Rural Areas ....................62 Table 35. Housing Characteristics of the Rural Poor................63 Table 36. Household Composition Among the Rural Poor .......... .......64 Table Al. Inter-regional Price Izdices (National = 100) for Peru, 1985-86 ....... .... . ............................... 7 Table A2. Welfare Deciles According to Different Welfare Measures*.*.**..o..e..*........................ ..... ..*....77 Table A3. Composition of Total Consumption in Peru, 1985-86....... 78 Table Bi. Distribution of Food Consumption by Food Consumption  - 1 - I. INTRODUCTION Latin American countries generally have higher levels of economic welfare, as measured by income per capita, than developing countries in most of Africa and a large part of Asia. Yet it is often asserted that the distribution of welfare in Latin American is more inequitable than in other areas of the world. Unfortunately, reliable evidence on the distribution of economic welfare is often lacking, both in Latin America and in other developing countries. Furthermore, the data that are available are often of dubious quality and suffer from comparability problems both over time and across different countries. This paper examines the distribution of welfare in Peru in 1985-86. Such an examination is particularly timely given economic events in that country in the late 1970s and early 1980s. Peru suffers from many problems common to Latin American countries, such as stagnant economic growth, high unemployment, high inflation, severe debt service problems, political and social unrest, widespread poverty and malnutrition, and budget cuts in the areas of education and health. A combination of ineffective policies and events beyond the control of the government (such as the worldwide recession and unfavorable weather) reduced average real household incomes by 24 percent from 1980 to 1983. Evidence also suggests that the distribution of income has become more unequal since the early 1970s. It is also asserted that the quality of educational and health care services has deteriorated in recent years.ll 11 This synopsis is taken from World Bank (1985).  -2- The government which assumed power in July 1985 is attempting new policy initiatives to address all of these problems. The effects of structural- adjustments now underway on the distribution of welfare are difficult to determine a priori, but an analysis of the data collected from the Peruvian Living Standards Survey (PLSS), which was administered from mid- July 1985 to mid-July 1986, is a first step in determining what those effects might be. In order to formulate effective policies for raising the living standards of Peruvians in the future, an accurate assessment of the present situation is needed. The organization of this paper is as follows. Section II discusses the theoretical basis of consumption expenditures as a measure of economic welfare. The third section presents a detailed description of consumption expenditures in Peru. Section IV analyzes the distribution of welfare using summary measures of inequality, and the fifth section provides a profile of the poor in Peru. A final section summarizes the paper and makes some tentative policy recommendations.  - 3 - II. CONSUMPTION DATA AS AN INDICATOR OF WELFARE Most empirical work on the distribution of welfare is done using either expenditure or income data recorded in household surveys. This is intuitively appealing but it is important to review the theoretical framework which allows one to draw the link between the distribution of income or expenditures and the distribution of welfare. This is found in the branch of economic theory known as welfare economics. The starting point of applied welfare economics is a social welfare function which takes the Bergson-Samuelson form (i.e. social welfare is a function of the utility levels of individuals). It is further assumed that individuals possess the same utility function. If one examines households rather than individuals, it is also assumed the households possess the same utility function, which has among its arguments the compositional characteristics of the household (e.g. number and ages of household members). Because individual utility functions are observationally equivalent under any monotonic transformation, it is necessary to find a method of labeling indifference curves which: 1) allows one to distinguish between individuals at different levels of utility given observable data; and 2) does not imply any particular cardinalization of the common individual utility function. This can be done using cost functions, which specify the amount of money required (given a set of prices and the assumption of utility maximization) to attain a particular level of utility (cf. Deaton and Muellbauer, 1980). The use of cost functions to compare different utility curves assumes that total consumption, as opposed to total income, represents the  welfare levels of individuals.2/ Using consumption rather than income data is also supported by the argument that the former is a better indicator of life- cycle welfare than the latter because income may fluctuate over short periods of time while consumption is allocated more evenly ("smoothed") over time. Furthermore, consumption data are likely to be more reliable than income data because the former are less sensitive information from the perspective of the survey respondents. Finally, consumption data are preferable because it is difficult to measure the income of self-employed workers. Once the choice has been made to compare the welfare of households by comparing their total consumption expenditures,3/ new issues arise which must be addressed. Specifically, some have argued that food expenditure data are better for measurement of welfare than total expenditure data (Anand and Harris, 1985). Several reasons can be offered: 1) Consumption of food is less susceptible to economies of scale within households than consumption of non-food items; 2) Construction of price indices is easier for food expenditures; 3) Imputation of use values to durable goods and owner- occupied housing is not necessary; and 4) Food expenditure data are thought to be more accurate than those on non-food expenditures. This paper will 2/ This would not be the case if savings (i.e. income minus consumption) is an argument in the utility function. In this paper the standard practice of considering savings to be delayed consumption, so that it does not enter into the utility function as a separate argument, will be followed. 3/ In the remainder of this paper the terms "expenditures" and "consumption" are used interchangeably. For goods which are purchased intermittently (i.e. durable goods and housing), consumption and expenditures are not simultaneous. The expenditure variable used here "spreads out" such expenditures so that they do represent actual consumption (see Appendix A). Also, consumption of home-grown food and other in-kind income is given a monetary value so that expenditures include consumption of these items.  examine total expenditures rather than food expenditures. The reasoning for this choice is given in Appendix B, but the general argument is that use of food expenditures to rank households implies that better off households have the same food shares (percentage of expenditures spent on food) as poor households, which is difficult to accept. However, the arguments raised here must be answered. The steps taken to deal with these problems are discussed in Appendix A, with the exception of the first argument, which will now be considered. It has often been asserted that additional household members, particularly children, are less "costly", in the sense of requiring additional expenditures to maintain the welfare levels of original household members, relative to the initial cost of attaining that level of welfare by a household composed of a single person or a childless couple. This assertion is supported by both common experience and economic reasoning. Clothing and other items can be handed down from older to younger children, durable goods such as radios and refrigerators can be enjoyed by additional household members at no extra cost, and even in the case of food, children consume less food than adults. Indeed, empirical work has consistently shown that the welfare of larger households (given a fixed level of household expenditures) is not strictly inversely related to household size (van der Gaag, 1982). The method of adjusting for these scale economies in consumption is the estimation of "adult equivalence scales", which measure the "cost" of additional household members in terms of fractions of adults (cf. Deaton and Muellbauer, 1980, Ch. 8). A rigorous construction of adult equivalence scales for Peru in 1985- 1986 is no small task and lies beyond the scope of this paper. However, one  -6- can reasonably assume that they lie between zero and unity, i.e. that an additional member will present increased costs to other household members but not to the point where household welfare is measured by per capita expenditure. In this paper children will be given smaller weight than adults; in terms of adult equivalence units, children less than seven years old will have a weight of 0.2, children between the ages of seven and twelve will have a weight of 0.3, and children between the ages of thirteen and seventeen wilL receive weights of 0.5. These weights are consistent with those estimated for Sri Lanka and Indonesia by Deaton and Muellbauer (1986). The sum of these weights for each household is used to divide household expenditures to arrive at a measure of household welfare. Due to lack of data on the intra-household distribution of goods and services, it is assumed that each individual has the same level of welfare as the household as a whole. All comparisons will be made between individuals (not households) after their welfare level has been assigned. Before examining the data in Peru, it should be pointed out that welfare measurement using consumption data may not cover all aspects of one's idea of what welfare is. Welfare rankings based on adjusted per capita expenditures omit some important but difficult to measure components of welfare, particularly the health status of individuals. In a given country the health of individuals may be roughly comparable, but comparison between different countries may not be possible without strong disclaimers. This point can be illustrated by comparing Peru with Costa Rica and Sri Lanka. Table 1 presents some figures relevant for welfare comparisons among Peru, Costa Rica and Sri Lanka. They are given here for illustrative purposes only. The figures show that Sri Lanka does relatively well in non-consumption  - 7 - measures of welfare, such as life expectancy, infant mortality and the adult literacy rate, while Peru has a substantially higher GNP per capita, which is clearly a consumption-based measure. Costa Rica also does much better than Peru in terms of life expectancy and infant mortality, though its income level is only slightly higher than that of Peru. There is no clear method to make welfare judgments across these three countries and no attempt will be made here. However, it should be borne in mind that the welfare comparisons among Peruvians made in this paper are essentially based on consumption measures of welfare and thus ignore aspects of living standards which are much more difficult to measure at the household level. Table 1: Welfare Indicators of Peru and Sri Lanka Welfare Measure Peru Costa Rica Sri Lanka GNP per Capita (1984) $1000 $1190 $360 Life Expectancy (1984) 59 73 70 Infant Mortality Rate (1984) 95 19 37 Adult Literacy Rate (1980) 80% 78% 85% Source: World Bank (1983, 1986). This completes the discussion of the framework used in this paper for examining the distribution of welfare in Peru. Appendix A describes the data in detail and shows how they are modified to correct for the problems discussed above. The following section examines the distribution of welfare (as measured by consumption) in Peru.  -8- III. THE DISTRIBUTION OF CONSUMPTION I>T PERU This section provides descriptive data on the distribution of welfare in Peru as measured by (adjusted) per capita total consumption expenditures. The first subsection examines the shares of total consumption going to population deciles ranked according to their welfare levels. The second subsection provides information on the characteristics of households by consumption quintiles and by different geographic areas of Peru. The following subsection examines the distribution of agricultural land, and the last examines school attendance, housing and ownership of durable goods. A. The Distribution of Consumption Expenditures by Population Deciles This subsection examines the distribution of total consumption by population deciles, where Decile 1 contains the poorest 10 percent of the population (as measured by adjusted per capita consumption expenditures at June 1985 prices - see Appendix A for details), Decile 2 contains the next poorest 10 percent, etc., and Decile 10 contains the "wealthiest" 10 percent of the population. Table 2 gives data on per capita consumption expenditures for all ten population deciles as given by household expenditures divided by household size (unadjusted) and also by household consumption divided by adult equivalents (adjusted). Of course, weighting children as fractions of adults leads to higher means for the latter as compared with the former. There is some difference in the distribution of welfare between the two methods - after adjusting for household size the distribution becomes more equal. The data in Table 2 indicate that the use of per capita total consumption expenditures as a welfare indicator results in broad rankings consistent with Engel's law - as the welfare of individuals rises the percentage of total consumption devoted to food declines. They also reveal the degree of inequality - the poorest 40 percent of the population receive 14-15 percent of total consumption while the wealthiest 20 percent receive 49-51 percent.  -9- Table 2: Distribution of Total Consumption by Welfare Deciles Mean Per Capita % of Total Food Share (Z) Monthly Expenditures Expenditures of Total Expenditures (Intis) in Peru within each Decile Decile Adjusted Unadjusted Adjusted Unadjusted Adjusted Unadjusted 1 111.8 69.3 2.01 1.80 0.717 0.727 2 188.9 118.1 3.39 3.07 0.717 0.721 3 248.2 157.7 4.46 4.10 0.695 0.708 4 307.2 195.4 5.52 5.08 0.674 0.681 5 365.4 236.3 6.57 6.14 0.664 0.659 6 431.3 289.0 7.75 7.51 0.629 0.645 7 524.4 355.9 9.42 9.25 0.618 0.593 8 656.8 449.9 11.80 11.70 0.574 0.576 9 874.3 615.4 15.71 16.00 0.524 0.505 10 1858.0 1360.0 33.38 35.36 0.404 0.395 Peru 556.6 384.7 100.00 100.0 0.542 0.530 Note: Intis figures are at June 1985 prices. A comparison of the data in Table 2 with data from other countries would provide information on differences in the distribution of welfare between Peru and other countries. Unfortunately, comparable data are almost non-existent. The only exception is a similar survey conducted in C8te d'Ivoire (Ivory Coast) in 1985, which shows a very similar distribution of  - 10 - expenditures by deciles (see Glewwe, 1987). Data on income inequality are available from other countries, though differences in survey structure and income definitions present major problems. These data will be discussed in Section IV. B. Characteristics of Households by Per Capita Expenditure Groups The previous subsection gives some information on the distribution of welfare in Peru but it reveals almost nothing about the welfare levels of different population groups. In this subsection the welfare levels of different groups within Peru will be compared. Table 3 presents tabulations which divide the total population into five welfare groups (quintiles). The first quintile contains the poorest 20 percent of the population as defined by adjusted (i.e. smaller weights for children) per capita expenditures, the second contains the next poorest 20 percent, etc., and the fifth contains the wealthiest 20 percent of the Peruvian population. Each quintile is then characterized by the proportion of people who belong to different population groups within that quintile. The groupings within quintiles are: 1) Region in which the household resides; 2) Sex of the head of household; 3) Language of interview; 4) Type of employer of the head of household; 5) Occupation of the head of household; and 6) Highest grade completed by the head of household. In addition, mean per capita consumption expenditures (both adjusted and not adjusted by adult equivalent scales) are given for each group. Table 4 provides the breakdowns of these groupings according to regions, which is useful in interpreting the data in Table 3. Peru can be divided into four distinct regions: the metropolitan area around the capital city of Lima, the coastal areas along the Pacific Ocean,  - 11 - the mountainous Andes region known as the Sierra, and the jungle area to the east of these mountains known as the Selva (jungle). The SeLva region is relatively less populated despite its large geographic size. Its location in the upper Amazon basin isolates it from the rest of Peru. In this paper the last three regions will be divided into urban and rural areas. The composition of quintiles according to regions reveals that Lima is home to relatively well off households. Urban areas on the coast and in the Sierra are also relatively wealthy but not as wealthy as Lima. The urban Selva is slightly more wealthy than Lima but only 3 percent of Peruvians live there.41 The rural regions are relatively poor, with the rural Sierra the poorest of all. One should not infer from the relative wealth of urban areas that rural to urban migration will result in a poor person becoming rich. As will be seen below, the educational level of the head of household is strongly correlated with welfare levels, so What high levels of welfare in urban areas reflect their relatively well educated residents. 41 Only 144 of the households in the sample are from the urban Selva. This low sample size reduces the accuracy of this paper's statistics on that region.  - 12 - Table 3: Characteristics of Hobuseolds by qulntiles Breakdown within Quintiles Quintile Quintile Quintile Quintile Quintile Mean Expend. per Capita All (Intis per month) Characteristic Peru 1 2 3 4 5 Adjusted Unadjusted Region Lima 26.8 6.0 18.2 28.8 35.4 45.5 770.9 569.2 Coastal Urban 15.2 11.1 14.7 17.6 15.4 17.2 569.8 390.7 Coastal Rural 7.2 8.8 9.8 7.2 6.8 3.5 421.3 265.5 Sierra Urban 11.0 9.0 9.6 10.2 11.5 14.8 649.9 442.0 Sierra Rural 30.5 52.8 38.5 28.1 22.9 10.4 366,8 235.4 Selva Urban 3.0 2.1 2.8 2.3 3.0 4.7 792.0 553.7 Selva Rural 6.3 10.3 6.5 5.8 5.1 3.9 413.5 264.8 Sex of Head Male 86.7 85.3 86.0 85.1 87.6 89.4 565.0 387.1 Female 13.3 14.7 14.0 14.9 12.4 10.6 500.1 367.5 Language Spanish 93.7 85.9 90.5 95.5 97.4 99.3 575.3 398.3 Quechua 5.1 12.1 7.9 3.5 1.7 0.5 263.9 170.3 Aymara 1.1 1.8 1.5 0.8 0.9 0.3 341.3 236.0 Other 0.1 0.2 0.1 0.2 0.0 0.0 306.9 176.9 Continued  - 13 - Table 3 continued Breakdown with Quintiles Quintile Quintile Quintile Quintile Quintile Mean Expend. per Capita All (Intis per month) Characteristic Peru 1 2 3 4 5 Adjusted Unadjusted Employer of Head None 6.0 5.1 5.8 6.0 6.4 6.6 582.4 478.8 Government 9.0 2.7 5.4 7.5 11.8 17.8 801.8 549.3 Parastatal 2.6 0.4 0.6 1.9 3.0 7.2 960.7 655.8 Private 20.6 15.0 21.0 23.0 21.0 23.0 599.2 415.8 Private Home 2.0 2.7 3.5 2.4 1.4 0.2 320.0 202.8 Self-employed 59.8 74.2 63.6 59.3 56.5 45.2 492.3 333.8 Occupation of Head None 5.9 5.1 5.7 6.0 6.1 6.6 583.7 480.9 Agricultural 40.0 64.7 47.3 38.5 31.8 17.5 389.4 255.3 Sales/Services 22.3 10.5 21.3 24.8 25.7 29.0 655.6 453.2 Industry/Craft 20.0 17.5 21.8 22.1 22.2 16.6 510.4 345.7 White Collar 11.8 2.2 3.9 8.5 14.2 30.2 1000.6 711.1 Other 0.0 0.0 0.0 0.1 0.0 0.0 375.4 195.2 Highest Grade Complete by Head None 13.8 28.1 17.2 12.4 7.4 3.8 329.0 223.6 Elementary 52.4 62.1 63.5 57.4 50.0 29.0 436.8 295.7 Secondary (some) 11.1 5.9 9.7 13.1 14.2 12.6 576.0 392.7 i (degree) 12.2 2.9 7.0 11.2 16.8 22.9 760.6 542.7 Post-Secondary 2.7 0.5 1.2 2.4 3.6 5.9 841.4 581.0 University 7.8 0.5 1.4 3.5 8.0 25.5 1298.8 926.6 Other 0.1 0.0 0.0 0.0 0.1 0.3 2935.6 2039.2 Note: 1. All figures are in terms of the percentage of the population living in households with the relevant characteristics. 2. Figures in Intis are at June 1985 prices.  - 14 - Table 4: Characteristics of Households by Regions Breakdown by Regions Coastal Coastal Sierra Sierra Selva Selva Characteristic All Peru Lima Urban Rural Urban Rural Urban Rural Sex of Head Male 86,7 86.2 83,2 89.9 86.7 87.6 83.0 91.0 Female 13.3 13.9 16.8 10.1 13.4 12.4 17.0 9.0 Language Spanish 93.7 100.0 99.9 100.0 99.2 86.6 100.0 90.9 Quechua 5.1 0.0 0.0 0.0 0.6 14.7 0.0 9.1 Aymara 1.1 0.0 0.1 0.0 0.2 3.4 0.0 0.0 Other 0.1 0.0 0.0 0.0 0.0 0.3 0.0 0.0 Employer of Head None 6.0 8.6 8.8 3.1 6.7 3.9 3.3 1.4 Government 9.0 15.2 10.3 1.4 17.2 2.6 16.4 2.4 Parastatal 2.6 3.4 5.4 0.4 4.7 0O8 2.4 0.0 Private 20.6 33.8 24.9 22.1 15.9 12.0 13.2 5.9 Private home 2.0 1.0 2.0 4.3 1.5 2.5 0.0 3.4 Self-Employment 59.8 38.1 48.7 68.7 54.0 78.2 64.8 86.9 Occupation of Head None 5.9 8.4 8.8 3.1 6.7 3.8 3.3 1.4 Agricultural 40.0 4.2 13.6 76.3 15.4 76.8 19.0 87.9 Sales/Services 22.3 33.9 38.7 11.8 30.8 5.8 35.5 4.3 Industry/Craft 20.1 30.6 25.2 7.2 29.2 10.5 25.8 5.6 White Col./Mgr. 11.8 22.9 13.8 1.5 17.7 3.1 16.5 0.9 Other 0,0 0.0 0.0 0.0 0.2 0.0 0.0 0.0 Highest Grade Completed by Head None 13.8 2.2 6.6 22.8 7.7 27.2 7.3 18.9 Elementary 52.4 37.2 53.7 63.9 44.0 61.9 47.9 71.7 Secondary (some) 11.1 16.6 14.1 5.2 13.5 6.0 17.5 5.0 Secondary (degree) 12.2 23.7 16.0 6.3 14.9 2.4 12.1 3.1 Post-Secondary 2.7 3.8 3.3 0.8 5.4 1.1 4.0 1.1 University (some) 7.8 16.5 6.2 1.3 14.2 1.4 11.2 0.3 Other 0.1 0.0 0.1 0.0 0.2 0.1 0.0 0.0 Note: 1. All figures are in terms of the percentage of the population living in households with the relevant characteristics.  - 15 - Households headed by women in Peru have lower welfare levels than those headed by men. This is a common phenomenon in developing countries. Yet the difference is not unusually large. The fraction of female-headed households declines somewhat as one moves from poorer to wealthier households, but in all quintiles the fraction is between 10 and 15 percent. As can be seen in Table 4, female-headed households are found in similar proportions throughout Peru, the only evident pattern is that they are somewhat less common in rural areas. Peru has a long history of antagonism between the Indian population and the descendants of the Spanish colonialists. The present population can be divided into roughly three ethnic groups, whites (those with no Indian blood), mestizos (those of mixed blood), and Indians. The whites form a small fraction of the population, less than 10 percent, while the remaining population is roughly evenly split between mestizos and Indians. Unfortunately, there is no clear definition of these groups (see Weeks- Vagliani, 1985, and Kluck, 1981). The only way to distinguish between them in the survey is to use the language of interview, since all Quechua and Aymara speakers are clearly Indians. Because many Indians are bilingual and the questionnaire administered was in Spanish, few interviews were done in these two languages - only 5.1 percent of the interviews were conducted in Quechua and even fewer, 1.1 percent, were done in Aymara. It is widely known in Peru that Indians are the poorest group, and this is clearly supported by the data in Table 3. Quechua and Aymara speakers are mainly found in the poorest deciles and have much lower expenditure levels than those who were interviewed in Spanish.  - 16 - As in many other developing countries, working for the government or a government-owned corporation (parastatal) is strongly correlated with welfare levels. Individuals living in households in which the head works for the government are most often found in quintiles three, four and five. People living in households where the head works for a private firm are slightly better off than the general population (the All-Peru figures are in Table 2). On the other hand, households headed by self-employed workers, most of whom are farmers, are over-represented in the lower quintiles. One can see in Table 4 that the type of employer is strongly correlated with different regions - government and parastatal jobs are more common in urban areas while self-employment predominates in the rural areas. Private sector employment is most common in Lima and the Coastal region (urban and rural). An interesting observation from Table 3 is that individuals living in households where the head was not working have higher welfare levels than the average Peruvian household. This suggests that unemployment is not strongly associated with poverty in Peru. Of course, some of these households may be entirely composed of retired persons living on pensions or other sources of transfer income. More information on employment and unemployment in Peru (and C6te d'Ivoire) can be found in Newman (1987). Welfare differences by the occupation of the head of household also display patterns commonly found in developing countries. Households with low levels of welfare are predominantly engaged in agricultural pursuits, while households in which the head works in production (industry/crafts) or sales and services have higher welfare levels, and those in white collar or management occupations have the highest of all. One can see from Table 4 that these patterns are related by region - rural areas are predominantly  - 17 - agricultural while urban areas have many households in which the head worked in sales and services, industry or crafts, or a white collar or management occupation. The last grouping is potentially the most informative, since education is virtually always found to have strong explanatory power in the earnings of workers and Peru is no exception (SteLcner, Arriagada and Moock, 1987). Over ninety percent of the poorest 20 percent of the population lived in households where the head had an elementary level of education or none at all, while only 0.5 percent of these poor households had a head with a university education. In contrast, only 32.8 percent of the wealthiest 20 percent of the population lived in "elementary or less" households, which compares to a figure of 25.5 percent for households where the head had a university education. One can see in Table 4 that households with better educated heads are found in urban areas, particularly Lima and the urban Sierra, while less educated heads predominate in rural areas. The extent to which education brings about (though strict causality is not demonstrated in this paper) differences in the welfare of households is to some extent underestimated by the figures given, since the educational levels of household members other than the head are not accounted for in the figures given in Table 3. In many households the head may be an older family member who is no longex the maii, contributor to household income, and thus his or her educational level may not be the most relevant one for judging the effect of the education of household members on levels of welfare. In any case, there is little doubt that education is strongly correlated with (and almost certainly a major determinant of) the welfare levels of households in Peru. This correlation between education and welfare  - 18 - has important implications for policy, particularly in terms of the distributional impact. The World Bank (1985) reports that economic difficulties since 1975 have led to austerity measures which entailed reductions in planned educational expenditures while other types of government expenditures received higher priority.5/ A further problem is that schooling in rural areas (and perhaps poor urban areas) is generally of inferior quality (see discussion of school attendance below). A reorientation of government policies toward a more equitable educational system may be in order if the benefits of future economic growth are to be more evenly distributed. C. Ownership of Agricultural Land The unequal distribution of agricultural land in Latin American countries is often cited as a major factor in explaining inequality in those countries (cf. de Janvry, 1981). This situation also prevailed in Peru until 1968, when a leftist military coup overturned the government. Since that time a large amount of land has been redistributed by the government, which often set up large cooperatives jointly owned by peasants.6 There has been much criticism of the efficacy of agrarian policies in Peru since 1968 (cf. de Janvry, 1981, and Eglin, 1981), but it is clear that a major redistribution of land took place. The aftermath of this is seen in Table 5, which presents data on land ownership from the Peru Living Standards Survey. 5/ The share of total Central Government expenditures devoted to education was 18.8 percent in 1970, 12.8 percent in 1980 and 9.6 percent in 1985. 6/ These cooperatives, as well as other similar schemes, divide all the cooperative land among the member households, so that nearly all agricultural land in Peru explicitly belongs (though ownership may be through the cooperative) to individual households.  - 19 - Table 5: Agricultural Land Ownership in Peru All Peru Quintiles 1 2 3 4 5 All Land Percentage who own land 44.6 69.7 52.5 43.3 33.9 23.7 Land/capita: All households 0.63 0.40 0.48 0.45 0.87 0.97 Landowning households 1.42 0.58 0.91 1.05 2.55 4.09 Irrigated Land Only Percentage who own land 22.5 31.4 25.5 24.3 17.7 13.4 Land/capita: All households 0.10 0.08 0.09 0.09 0.10 0.11 Landowning households 0.42 0.26 0.36 0.36 0.58 0.84 Coastal Coastal Sierra Sierra Selva Selva Lima Urban Rural Urban Rural Urban Rural All Land Percentage who own land 4.7 9.5 68.1 29.2 89.1 27.9 91.5 Land/capita: All households 0.01 0.06 0.33 0.56 0.84 0.50 4.65 Landowning households 0.28 0.61 0.48 1.92 0.75 1.80 5.08 Irrigated Land Only Percentage who own land 4.1 9.1 60.5 16.3 41.1 8.6 16.3 Land/capita: All households 0.01 0.06 0.27 0.04 0.37 0.10 0.18 Landowning households 0.18 0.61 0.44 0.26 0.15 1.18 1.09 Notes: 1. Land per capita is measured in hectares per capita. 2. All figures are weighted by persons, not households. Approximately 44.6 percent of all Peruvians live in households which own land and 22.5 percent live in households with irrigated land. These percentage figures are highest for poorer households and lowest for wealthy households, which reflects the fact that wealthier households are more likely to live in urban areas. The lack of landozmership in Lima and other urban  - 20 - areas is seen in the bottom half of Table 5. However, poorer households that do own land have much smaller plots than wealthy landowning households, On average, those landowning households which are among the wealthiest 20 percent of the population have over seven times as much land (in per capita terms) as those among the poorest 20 percent. The distribution of land among welfare quintiles is easily obtained from the figures on "land per capita: all households" since each quintile contains one fifth of the population. For all types of agricultural land the top 2 quintiles each have about twice as much land as any of the bottom 3 quintiles. One can easily calculate that the top 20 percent of the population owns about 30 percent of all arable land while the bottom 20 percent owns about 13 percent of that land. While this is an unequal distribution of land, it is not extremely skewed. Yet extremely skewed distributions can easily be shown if households are ranked according to land owned per capita. Since only 44.6 percent of the Peruvian population live in households that own land, the "bottom half" (not in terms of welfare, but in terms of land ownership) own no land whatsoever. Decile distributions of land ownership are given in Table 6 for all of Peru and for rural Peru alone. It is clear that land ownership in Peru is extremely skewed in that 86.3 percent of all agricultural land is owned by only 10 percent of the population and 95 percent is owned by only 20 percent of the population. But is is important to realize that the 20 percent who own 95 percent of the land are not the wealthiest 20 percent of Peruvians (though some may belong to this group), since it was shown in the previous paragraph that the wealthiest 20 percent of Peruvians own about 30 percent of Peru's agricultural land. This clarifies an important point which is often  - 21 - overlooked in discussions of land distribution in developing countries - those households which control the most land are not always the wealthiest households in the country. An additional point is also of equal importance - the quality of land is not measured in Tables 5 and 6 except to the extent that irrigated land is more valuable than unirrigated land. If the biggest landowners own mostly pasture land which is unsuitable for growing crops, then the data in Table 6 over-estimate inequality in the distribution of the value of land. Unfortunately the PLSS data do not distinguish between pasture land and crop land. On the other hand, the poorest households in Peru, who tend to have smaller plots, may also have low quality agricultural land, so the data in Table 6 may underestimate inequality in the distribution of land value. Despite the fact that Table 5 does not give data on land quality (except in terms of irrigated land) it is much more useful than Table 6 because it controls for welfare levels in examining the distribution of land among Peruvians.  - 22 - Table 6: Distribution of Agricultural Land in Peru All Peru Rural Peru Only Decile % Land Owned Decile X Land Owned 1 0.0 1 0.0 2 0.0 2 0.0 3 0.0 3 0.4 4 0.0 4 0.9 5 0.0 5 1.6 6 0.1 6 2.4 7 1.2 7 3.5 8 3.8 8 5.4 9 8.7 9 9.9 10 86.3 10 76.0 Note: 1. Decile rankings are in terms of land per capita. 2. Each decile includes 10 percent of the population, not 10 percent of households. D. School Attendance, Housing and Ownership of Durables The information collected in the Peru Living Standards Survey allows for some interesting observations to be made on the decisions households make, particularly the differences seen among households at different levels of welfare. In this section differences in school attendance, housing and ownership of durables by welfare quintiles will be examined. One of the most important questions concerning the nature of poverty in any country is whether the poor constitute the same group of people over long periods of time or whether there is a large amount of entry in and exit from the ranks of the poor over the years. An important aspect of this is whether children who come from poor families are likely to be poor when they become adults and have their own families. Given the strong positive  - 23 - correlation found between education and levels of welfare in the previous section, the relationship between welfare levels and school attendance of children deserves serious attention. These data are given in Table 7. Before discussing the data in Table 7 it is useful to make two obse.rvations. First, school attendance can be thought of as an interaction of supply and demand. In other words, low school attendance is in part due to family decisions based on the opportunity cost of schooling (demand for schooling) and in part on the availability and quality of school facilities (supply of schooling). Neither side should be neglected when analyzing school attendance patterns. Second, there is evidence in many developing countries that some children begin school at a relatively late age (thus lowering the participation rate for ages, 6-10) and often repeat grades (thus raising the figure for ages 11-15). In Peru as a whole, school attendance rises as one moves from poorer to wealthier households for both age groups. Yet to some extent this may be an artifact of differences in welfare levels among regions. In fact, within regions the relationship between welfare levels and school attendance is weak, showing no clear monotonic pattern. Even though it is usually the case that the poorest quintile has lower school attendance than average while that of the wealthiest quintile is higher than average, the absolute differences are not dramatically large. Differences in school attendance are much more apparent between regions, particularly the lower school attendance in rural areas. Given that school attendance is lowest in the rural Sierra, that the poorest people in Peru are found in that region, and the findings in the previous subsection that education is strongly correlated with household  - 24 - welfare, there appears to be a need for improvement of educational opportunities in that region. One cannot determine with certainty whether low school attendance is primarily due to some opportunity cost of schooling or the lack of good school facilities without an extended discussion beyond the scope of this paper. Yet the fact that 20-25 percent of school age children are not attending school and the low correlation between welfare levels and school attendance within regions suggest that the problem is a lack of good facilities. This is confirmed by data on quality of schools from the PLSS - in urban areas about 70-80 percent of enrolled children aged 6 to 16 attended schools where both mathematics and language books were available, while the corresponding figure for rural areas is 50-53 percent. Stelcner, Arriagada and Moock (1987) find that such measures of school quality have a strong positive effect on educational achievement, particularly in rural areas. Housing conditions are important measures of welfare themselves, and they also have indirect implications for welfare, particularly in the area of health. Table 8 gives information on sources of drinking water, sources of lighting and type of sewerage by welfare quintiles and by regions.  - 25 - Table 7: School Attendance by Welfare Quintiles Quintiles Entire Region 1 2 3 4 5 All Peru Ages 6-10 86.5 77.1 85.1 86.3 89.5 94.7 Ages 11-15 88.2 79.6 87.9 91.0 91.6 91.5 LimaI Ages 6-10 95.2 92.6 95.8 92.7 96.8 98.0 Ages 11-15 95.2 94.9 97.2 95.7 93.7 94.0 Coastal Urban Ages 6-10 94.3 93.5 89.6 93.7 98.3 95.9 Ages 11-15 94.4 95.4 93.4 95.5 94.0 93.5 Coastal Rural Ages 6-10 87.7 85.5 88.3 90.5 83.1 91.2 Ages 11-15 84.9 83.6 86.8 82.8 83.3 87.1 Sierra Urban Ages 6-10 92.6 89.7 94.0 91.3 90.7 96.4 Ages 11-15 93.0 93.3 90.9 95.8 93.8 90.7 Sierra Rural Ages 6-10 76.1 69.1 71.4 80.7 75.0 82.9 Ages 11-15 79.8 73.5 76.4 81.8 81.9 85.5 Selva Urban Ages 6-10 88.8 76.9 92.0 88.9 95.5 93.8 Ages 11-15 91.5 100.0 87.0 91.7 80.0 96.3 Selva Rural Ages 6-10 80.2 78.8 77.8 73.1 90.7 82.7 Ages 11-15 82.0 75.0 68.2 85.7 89.8 92.7 Note: Within each of the five sectors quintiles are defined with respect to those sectors (e.g. poorest 20 percent of all Lima residents, not the poorest 20 percent of all Peruvians who also happen to live in Lima).  - 26 - Table 8: Housing Characteristics in Peru All Peru Quintiles 1 2 3 4 5 Source of Drinking Water Public: Inside Dwelling 46.0 22.8 36.5 46.5 52.3 11.9 Inside Building 5.5 3.6 4.2 6.6 7.4 5.8 Outside Building 5.2 4.7 5.0 6.5 6.3 3.7 Well 9.8 15.5 12.2 7.6 9.3 4.4 River/Spring 27.0 47.7 3.4 6.2 18.5 9.2 Water Truck 3.5 2.1 5.2 3.9 3.6 2.9 Other 3.0 3.7 3.6 2.8 2.8 2.1 Source of Lighting Electric 53.9 22.8 41.3 58.1 64.7 82.5 Kerosene/Oil 42.1 71.5 54.3 38.3 31.1 15.4 Candles 3.9 5.1 4.5 3.6 4.2 2.0 None 0.2 0.7 0.0 0.1 0.0 0.1 Sewerage Public Service 40.9 12.3 28.7 42.3 51.4 70.0 Well-Septic 1.6 1.6 1.5 1.8 1.1 2.2 Cesspool 8.1 5.7 9.4 8.8 8.8 7.7 None 49.4 80.4 60.5 47.1 38.7 20.1 All: Coastal Coastal Sierra Sierra Selva Selva Peru Lima Urban Rural Urban Rural Urban Rural Source of Drinking Water Public: Inside Dwelling 46.0 78.5 71.5 14.5 60.5 13.1 54.2 12.9 Inside Building 5.5 9.7 5.2 0.0 12.2 2.0 5.5 0.3 Outside Building 5.2 3.2 7.8 4.1 9.5 5.2 5.2 2.2 Well 9.8 2.0 0.6 30.7 9.0 16.8 16.6 5.2 River/Spring 27.0 0.0 0.1 41.2 0.7 62.0 0.0 78.8 Water Truck 3.5 5.1 7.9 8.4 2.9 0.0 0.8 0.0 Other 3.0 1.6 7.0 1.1 5.2 0.9 17.7 0.6 Source of Lighting Electric 53.9 95.8 73.0 19.7 80.2 13.9 72.5 8.3 Kerosene/Oil 42.1 2.8 25.3 76.9 12.6 79.1 24.5 90.2 Candles 3.9 1.4 1.7 3.5 7.0 6.6 3.0 1.6 None 0.2 0.1 0.0 0.0 0.2 0.4 0.0 0.0 Sewerage Public Service 40.9 83.9 61.0 2.1 54.5 5.4 43.9 2.2 Well-Septic 1.6 1.4 3.2 1.2 2.9 0.5 1.6 3.0 Cesspool 8.1 9.5 11.8 17.8 6.2 2.8 11.1 9.3 None 49.4 5.3 24.1 78.9 36.4 91.3 43.4 85.5  - 27 - In Peru as a whole, there are a variety of sources of drinking water, some of which are clearly preferred by wealthier households, particularly publicly supplied water inside the dwelling, while others predominate among poorer households, such as well water and water from rivers and springs. In Lima publicly supplied water is by far the most common source. This is also true, though to a lesser degree, in other urban areas. Residents of rural areas get most of their water from wells, rivers and springs. In many cases households may have little choice regarding their water supply, so that patterns by quintiles may to a large extent reflect the region in which the household resides. The two main sources of lighting, electricity and oil (or kerosene) lamps, are clearly inversely related by welfare levels - wealthier households use the former while poorer households use the latter. The breakdown by regions is quite dramatic - urban areas overwhelmingly depend on electric lighting while rural areas are dependent on oil or kerosene lamps. As with water, the type of lighting may often be out of the household's control. Finally, the type of sewerage in the household displays expected patterns. Poorer households have no sewerage at all, while wealthier households often have public service. Public service predominates in Lima, and to some extent in other urban areas, while no sewerage at all is typical of rural areas. A better understanding of choices made by households is given in data on the ownership of durables. Ownership patterns over different welfare levels can be used to identify goods as necessities or luxuries, and to a certain extent may reveal pairs of goods to be complements or substitutes. A full analysis of expenditure patterns is beyond the scope of this paper, but a rough idea is given by the data in Table 9.  - 28 - Virtually all the durable goods in Table 9 are luxuries in the sense that wealthier households are more likely to possess them (of course this does not account for the fraction of outlay spent on a particular good). The sole possible exception is radios, which are found fairly evenly across all households. Certain expensive goods, such as automobiles, refrigerators and color televisions, are very sensitive to welfare levels. Finally, it should be noted that ownership patterns by regions may reflect regional characteristics - televisions are of little use in rural areas without broadcasting stations and the lack of roads in the Rural Selva makes automobiles and bicycles relatively useless.  - 29 - Table 9: Ownership of Durables by Quintiles and Regions Quintiles All Peru 1 2 3 4 5 Sewing Machine 46.0 28.3 39.6 49.8 51.2 60.8 Refrigerator 32.7 5.9 19.9 30.9 41.6 65.4 Phonograph/Stereo 33.2 14.2 23.5 28.9 39.1 60.0 Radio 66.2 58.5 64.4 70.1 68.7 69.3 Television - b/w 41.6 14.9 34.1 46.0 51.6 61.4 Television - color 14.1 0.2 2.3 8.4 17.2 42.6 Bicycle 17.2 7.8 11.0 16.4 21.5 29.4 Automobile 7.6 0.2 1.2 3.0 7.4 26.2 All Coastal Coastal Sierra Sierra Selva Selva Peru Lima Urban Rural Urban Rural Urban Rural Sewing Machine 46.0 52.7 51.6 46.3 58.4 33.7 61.4 34.0 Refrigerator 32.7 67.6 46.7 12.4 36.7 2.3 9.4 43.9 Phonograph/Stereo 33.2 50.8 29.1 21.8 43.8 21.1 36.6 19.3 Radio 66.2 62.5 64.8 63.4 80.8 65.2 70.3 66.0 Television - b/w 41.6 73.1 65.6 37.0 52.4 8.3 29.9 3.6 Television - color 14.1 33.3 12.8 2.7 8.2 1.1 23.9 0.7 Bicycle 17.2 21.4 15.6 18.5 23.1 15.0 13.8 3.9 Automobile 7.6 17.0 8.0 2.1 9.1 1.5 7.7 0.0 Note: All figures are the percentage of the total population living in households possessing the durable good.  - 30 - IV. INEQUALITY IN PERU In this section inequality will be analyzed using summary measures of expenditure inequality. A few points should be emphasized at the outset. First, a measure of expenditure inequality is a single number which describes the spread of an entire distribution of income or expenditures, thus it is only a first approximation to describing the nature of inequality. Second, because income data are deemed to be inappropriate for measuring welfare, inequality can only be decomposed by groups and not by income sources (see Appendix C). Third, the actual expenditure levels used are adjusted by household composition by counting children as fractions of adults, so that all the analysis in this section is based on adjusted per capita consumption expenditures. A. Historical Comparison with Other Countries Before examining the distribution of adjusted per capita expenditures in detail it is useful to compare the level of inequality in Peru relative to that of other developing countries. Data in Table 10 are taken from a variety of sources. The figures are not always comparable but they do convey an approximate notion of the degree of inequality in Peru relative to other countries. The Gini coefficient is perhaps the most commonly used measure of inequality. It varies from 0 (complete equality) to 1 (complete inequality) and is further described in Appendix C. The data in Table 10 suggest that Peru had an extremely unequal distribution of income in 1961 compared both with other Latin American countries and with Asian and African countries. However, these figures must be approached with caution because: 1) They are based on income rather than  - 31 - expenditure data; 2) They are from different years; and 3) They do not have identical definitions of income. Despite these caveats, they are of some use in putting the following analysis into perspective. Table 10: Gini Coefficients on Per Capita Income Latin American Gini Gini Countries Coefficients Other Countries Coefficients Costa Rica (1971) 0.4757 C6te d'Ivoire (1959) 0.4556 Dominican Republic (1969) 0.4550 India (1964-65) 0.3957 El Salvador (1969) 0.4653 Malaysia (1970) 0.5045 Honduras (1967-68) 0.5658 Senegal (1960) 0.5874 Peru (1961) 0.6664 South Korea (1970) 0.4065 Tanzania (1967) 0.5282 Tunisia (1961) 0.5094 Source: Jain (1975). B. Regional Inequality Decompositions Measurement of expenditure inequality will be done using four commonly used inequality measures, the Gini coefficient, the Theil entropy measure (T), an alternative entropy measure proposed by Theil (L), and the log variance of income (LV). The definitions of these measures and the justification for their use are given in Appendix C. For present purposes, it is relevant that the last three measures are group-decomposable. The advantage of group-decomposable inequality measures is that they can be used to divide overall inequality into inequality within different groups and inequality among those groups. For example, one can determine the proportion of overall inequality in Peru due to differences in average (adjusted) per  - 32 - capita expenditures among the seven regions (Lima, CoasLal Urban, Coastal Rural, Sierra Urban, Sierra Rural, Selva Urban and Selva Rural). This allows one to determine the potential effect on overall inequality of policies aimed at reducing differences among these regions. If the between-group cotnponent is small (say, less than 5 percent) policies whose sole objective is to reduce differences in welfare among these groups will have only a small effect on the overall distribution of welfare and thus may offer little to recomnend from the equity perspective. On the other hand, relatively large between-group contributions (say 20 percent or higher) reveal possibilities for promoting greater equity in Peru. In this section, groupings will be done according to regions and the education of the head of household. Table 11 gives the decomposition of inequality in Peru when it is divided into seven regions. Gini coefficients are given for purposes of comparison only since that inequality measure is not group decomposable. Although the different inequality measures do not always give the same rank to different regions, they do display broad agreement in many ways. Inequality is highest in the Selva Urban region, followed by Sierra Urban and Selva Rural, respectively. Thus Lima and the Coast areas (urban and rural) have a relatively low level of inequality. The high levels of inequality in the Sierra and Selva regions imply that the low mean incomes of the populations in the rural areas masks the even lower incomes of the poor in those regions. This will be examined in detail in Section V. Finally, it is worth noting that differences in mean incomes in the seven regions account for about 13-16 percent of overall inequality, which implies that the largest contribution to inequality comes about through inequality within each of the  - 33 - seven regions. Thus policies to improve equity should give at least as much attention to reducing inequality within regions as amg them. Table 11: Per Capita Expenditure Inequality-Decomposed by Region Mean ! % of Total Adjusted Exp. Log Region Population (Intis/month) Gini Theil T Theil L Variance Lima 26.8 770.9 0.3925 0.3077 0.2552 0.4310 Coastal Urban 15.2 569.8 0.3813 0.2584 0.2402 0.4470 Coastal Rural 7.2 421.3 0.3730 0.2550 0.2327 0.4302 Sierra Urban 11.0 649.9 0.4386 0.3543 0.3303 0.6209 Sierra Rural 30.5 366.8 0.3938 0.2968 0.2730 0.5293 Selva Urban 3.0 792.0 0.4852 0.4990 0.4036 0.7251 Selva Rural 6.3 413.5 0.4155 0.3139 0.2939 0.5562 Peru 100.0 556.6 0.4299 0.3534 0.3194 0.5967 Between-Group Contribution 0.0463 0.0476 0.0959 (X) 13.1% 14.9% 16.1% Note: Figures in Intis are at June 1985 prices. C. Decomposition by Educational Attainment Measuring the between-group contribution to overall inequality is most useful when the division of the total population into different groups has an underlying causal significance. In other words, dividing the population according to an underlying "exogenous" variable gives a first approximation of the contribution of that variable to overall inequality. In developing countries, perhaps the most important determinant of income (and hence expenditures) is educational level. Given the differences in welfare  - 34 - among educational groups in Table 3, one could also expect this to be the case in Peru. Table 12 gives inequality decompositions where households are grouped according to the educational level of the head of the household. Table 12: Per Capita Expenditure Inequality-Decomposed by Education of Head Mean % of Pop- Adjusted Exp. Log Education of Head ulation (Intis/month) Theil T Theil L Variance None 13.8 329.3 0.3123 0.2826 0.5507 Elementary 52.4 436.6 0.2614 0.2394 0.4518 Secondary (some) 11.1 576.0 0.2351 0.2059 0.3800 Secondary (degree) 12.2 760.6 0.2299 0.2151 0.4168 Post-Second. Training 2.7 841.4 0.1943 0.1939 0.3799 University 7.8 1298.8 0.3080 0.2709 0.4818 Other 0.1 2935.6 0.2116 0.2793 0.7025 Peru 100.0 556.6 0.3535 0.3195 0.5968 Between Group Cont. 0.0906 0.0795 0.1430 (Z) 24.8% 24.9% 24.0% Note: Figures in Intis are at June 1985 prices. The figures in Table 12 clearly demonstrate the important role played by education in determining the distribution of welfare in Peru. Differences in the educational attainment of the household head account for about one-fourth of overall expenditure inequality. In fact, these figures may understate the contribution of education because the educational levels of other household members are not considered, so that their contribution to differences in overall welfare is not taken into account. These results reinforce the argument made in Section III that a reorientation of the education system in  - 35 - Peru to provide better quality education in rural areas could have a large effect on reducing inequality in the long run. To summarize this section, group decomposable measures of inequality have been used to characterize the distribution of welfare in Peru. Differences between the welfare levels among the seven regions in Peru account for about 13-16 percent of overall inequality. The proportion of overall inequality due to differences in welfare levels by education groups is relatively high, which underscores the role played by education in determining the welfare levels of Peruvian households.  - 36 - V. CHARACTERISTICS OF THE POOR IN PERU Another method of describing the distribution of welfare in Peru or any other country is to focus on the population that falls below a given poverty line. The conditions faced by the poorest people in any country are presumably of particular importance to policymakers and others concerned with social welfare. The standard definition of poverty is to choose a "poverty line" , some level of expenditure (or income) which is assumed to be the minimum amount required for a "decent" standard of living, and classify all those whose expenditures fall below this poverty line as being in poverty. Once the poor have been classified, their characteristics can be ascertained, which is very important for choosing policies aimed at reducing poverty. This section will present the characteristics of the poor in Peru, as a whole, and also within urban and rural areas. A. Poverty in Peru - A National Profile Selection of a poverty line involves an element of arbitrary choice. It is best to try different poverty lines and then see whether they give a similar characterization of the poor. It is also convenient to choose a poverty line which classifies a certain percentage of the population as poor, so that one always knows the number of people under consideration. For these reasons two poverty lines have been chosen which allow us to examine the poorest lOX and poorest 30% of the population. The former group can be considered to be in extreme poverty while the latter group is defined by a more generous poverty criterion. In terms of adjusted per capita consumption expenditures, the extreme poverty group is defined as the population living in households for which adjusted per capita expenditures are below 155 Intis per  - 37 - month (June 1985 prices). The larger poverty group has a corresponding poverty line of 279.2 Intis per month. In the remainder of this subsection both poverty groups will be described in detail. Table 13: Household Size and Per Capita Consumption of the Poor Poorest 10Z Poorest 30X All Peruvians Household Size 7.3 7.3 6.7 Adjusted Household Size 4.7 4.7 4.4 Food Consumption 54.0 85.9 204.2 Adjusted Food Consumption 82.1 129.6 302.1 Total Consumption 75.5 122.4 384.5 Adjusted Total Consumption 112.2 183.2 556.4 Notes: 1. Figures are per capita averages, not per household averages. 2. Consumption is measured in Intis per month at June 1985 prices. Table 13 provides basic information on household size and per capita consumption for both poverty groups. Figures are also included for the population as a whole. One can see that a typical poor person lives in a larger household than the average Peruvian, but the differences are not very large. In terms of per capita consumption expenditures, both adjusted and unadjusted by family composition, the poorest 30 percent have less than half of the food expenditures and only a third of the total expenditures of an average Peruvian. For the poorest 10 percent the corresponding fractions are about one fourth and one fifth. Thus, both definitions of poverty represent populations whose consumption levels are much lower than that of the typical Peruvian.  - 38 - In Section III it was noted that female-headed households do not seem to be at a significantly lower level of welfare than those headed by a male. This is also the case with both poverty groups, as seen in Table 14. There is very little difference for the sex of the head of household among the poor when compared to the overall population. It may be the case that households headed by women receive some kind of remittance income from husbands or other relatives who are working in other areas of Peru. Table 14: Sex of Household Head Among the Poor Sex Poorest 10% Poorest 30% All Peruvians Male 86.5 85.3 86.7 Female 13.5 14.7 13.3 The fact that the poor are most often found in rural areas is seen in Table 15, which gives the distribution of the poor by the seven regions in Peru. The distribution of the poor is quite uneven across these regions. Poverty, particularly extreme poverty (poorest 10 percent), is relatively rare in urban areas. Although 56.0 percent of the Peruvian population resides in Table 15: Distribution of the Poor by Regions Region Poorest 10% Poorest 30% All Peruvians Lima 3.4 8.5 26.8 Coastal Urban 6.0 11.7 15.2 Coastal Rural 9.3 9.6 7.2 Sierra Urban 8.8 9.4 11.0 Sierra Rural 59.7 48.7 30.5 Selva Urban 2.3 2.2 3.0 Selva Rural 10.6 10.1 6.3  - 39 - urban areas, only 31.8 percent of the poorest 30 percent and 20.5 percent of the poorest 10 percent are found in urban areas. In rural areas poverty is most commonly found in the Sierra - 59.7 percent of the extremely poor are located there. In the next two subsections poverty will be analyzed within urban and rural areas, respectively. Given the relatively disadvantaged position of Indians in Peru, one could examine the poor by language of interview. This is presented in Table 16. Recall that many Indians speak Spanish and thus one cannot identify them as Indians using this criterion. Eighty-seven percent of the poorest 30 percent and 85.0 percent of the poorest 10 percent were interviewed in Spanish, which compares with 93.7 percent for the entire sample. Although this is a rough indicator, it is consistent with the hypothesis that Indians are much more likely to be poor than mestizos or whites. Table 16: Distribution of the Poor by Language of Interview Ethnic Group Poorest 10% Poorest 30% All Peruvians Spanish 85.0 86.8 93.7 Quechua 12.5 11.4 5.1 Aymara 2.5 1.7 1.1 Other 0.0 0.1 0.1 Perhaps the most useful information for policy analysis is the incidence of poverty among different types of workers. Table 17 gives the information in two forms, first classifying heads of households according to their employer and then by their occupation. As one would expect given the findings in Section III, the poor are most often self-employed and they work  -40 - primarily in agriculture. Thus raises in wages paid to government and parastatal workers will have almost no effect on poverty and legislation regarding minimum wages to be paid by private employers will affect only a fraction of the poor. It is clear that policies to reduce poverty must be aimed at self-employed workers in agriculture, particularly in the Sierra and Selva regions. Raising agricultural incomes is the key to significantly reducing poverty in Peru. Table 17: Distribution of the Poor by Employer and Occupation of Head of Household Poorest 10% Poorest 30% All Peruvians Employer None 5.5 4.7 6.0 Government 1.3 3.2 9.0 Parastatal 0.0 0.4 2.6 Private 12.3 16.9 20.6 Private Home 3.1 3.2 2.0 Self-Employed 77.9 71.6 59.8 Occupation None 5.5 4.7 5.9 Agricultural 70.8 61.2 40.0 Sales/Services 7.4 13.6 22.3 Industrial/Craft 15.1 18.3 20.1 White Collar/ Management 1.2 2.2 11.8 Other 0.0 0.0 0.0 One key determinant of agricultural incomes is land ownership. Data on land ownership among the poorest 10 percent and poorest 30 percent of Peruvians are given in Table 18. Although the poor are more likely to own land than the Peruvian population as a whole, the amount owned by landowning  - 41 - households is substantially smaller among the poor; landowning households among the poorest 30 percent own only about half as much land as the average landowning households in Peru, and for the poorest 10 percent the fraction is about one third. A detailed analysis of the contribution of land ownership to the distribution of welfare is beyond the scope of this paper, but it clearly has an important role to play. Table 18: Agricultural Land Ownership Among the Poor I ~~~~~~~~~~~~~~~~All Poorest 10% Poorest 30% Peruvians All --and Percentage who own land 77.1% 66.1% 44.6% Land/capita: All households 0.41 0.49 0.63 Landowning households 0.53 0.75 1.42 Irrigated Land Percentage who own land 33.6 30.6 22.5 Land/capita: All households 0.07 0.09 0.10 Landowning households 0.22 0.31 0.42  - 42 - The characteristic of Peruvians that has perhaps the strongest causal relationship to poverty is the educational levels of workers, as discussed above. Table 19 gives the distribution of education among the poor in Peru. These figures demonstrate the importance of education in Latin America. The great majority of the poor have only an elementary level of education or no education at all, which implies that a moderate or high level of education is very likely to remove one from the ranks of the poor. Of course, changes in the distribution of education come very slowly even under policies that are working well, but in the long run the benefits of education, particularly at the secondary level, are difficult to over-emphasize. Table 19: Distribution of the Poor by Educational Level Educational Level Poorest 10% Poorest 30% All Peruvians None 31.6 25.4 13.8 Elementary 61.7 62.9 52.4 Secondary (some) 4.1 7.0 11.1 Secondary (degree) 2.2 3.8 12.2 Post-Second. Training 0.0 0.4 2.7 University 0.4 0.5 7.8 Other 0.0 0.0 0.1 Note: Education level is that of the head of household. The importance of schooling leads one to ask whether children in poor families are less likely to attend school. The answer is found in Table 20, which gives information on school attendance, self-reported illness and housing conditions among the poor in Peru. School attendance of children among the poorest 30% of the population is clearly lower than it is for all  - 43 - Peruvians. Of course, much of this reflects regional differences. School attendance of the poor with urban and rural regions will be discussed below. Another noteworthy phenomena is that schooling appears to be delayed for the poor (and to a lesser extent among the non-poor) as school attendance is higher for those aged 11-15 than for those aged 6-10. Table 20 also contains data on self-reported illness and housing characteristics. It is rather paradoxical that poor people are healthier than wealthier people, especially when one examines their toilet facilities and sources of drinking water. The most likely explanation is that poor people have a less restrictive definition of good health than wealthier people, so that they consider certain health problems as "normal" which wealthier people may classify as illnesses. Thus the data on illness is of little use since people at different welfare levels have different conceptions of health and illness. Perhaps the only useful information is that the poor are more likely to be inactive due to illness than the general population. The information on housing given in Table 20 is similar to that given in Section III. Poor people have less access to electric lighting and so depend on candles and lamps. Sewerage is most often nonexistent. Finally, drinking water comes from wells or natural sources. The implications of these drinking water and sewerage facilities on health of the poor are probably negative as their relatively poor quality may promote the spread of infectious diseases. However, little more can be said without a detailed analysis. Poverty is often thought to be associated with the composition of poor households. Large numbers of children and small numbers of working household members may provide at least a partial explanation of why particular households are poor. However, the data given in Table 21 do not support this hypothesis. Poor households have about the same number of children and  - 44 - Table 20: Illness, School Attendance and Housing of the Poor Region Poorest 10% 1Poorest 30% All Peruvians School Attendance Ages 6-10 75.2 78.9 86.5 Ages 11-15 77.9 82.5 88.2 Illness % People Sick 38.2 39.7 42.8 Days ill (x) 14.1 13.9 14.3 Days inactive (X) 3.7 3.5 3.4 Source of Drinking Water Public: Inside Dwelling 18.8 26.0 46.0 Inside Building 2.4 3.6 5.5 Outside Building 3.4 4.7 5.2 Well 17.8 14.6 9.8 River/Spring 52.7 45.1 27.0 Water Truck 1.5 2.3 3.5 Other 3.5 3.7 3.0 Source of Lighting Electric 17.8 27.4 53.9 Kerosene/Oil 76.9 67.2 42.1 Candles 4.5 5.0 3.9 None 0.9 0.5 0.2 Sewerage Public Service 8.6 16.1 40.9 Well-Septic 1.6 1.5 1.6 Cesspool 4.0 6.9 8.1 None 85.8 75.5 49.4 Note: Days ill and days inactive refer to percentage over past 4 weeks. clearly more working members (as a percentage of total members) than the typical Peruvian household.7/ This implies that it is not the lack of work 71 Workers are any household members who, in either the last week or the last year, worked: 1. For an employer, 2. On a plot of land, or 3. In a family-owned business.  - 45 - activities per se or the large number of dependents that cause households to be poor but the income from the occupations that poor people have. This could be due to the low income per time unit of work or to the seasonal or part-time nature of the work (or both). Investigation of employment opportunities among the poor, who are primarily found in agriculture, is clearly needed, but is beyond the scope of this paper. Table 21: Composition of Household Members Among the Poor Household Member Poorest 10% Poorest 30% All Peruvians Children 0-6 yrs. 19.2 19.0 18.3 Children 7-12 yrs. 15.8 16.3 16.4 Children 13-17 yrs. 13.0 13.0 12.0 Workers 60.6 58.4 52.8 Note: Some workers may also be children, so the categories do not necessarily sum to 100%. To summarize the most important findings of this subsection, the poor are predominantly found in rural areas, particularly the rural Sierra. They are most often self-employed and are primarily in agricultural occupations. Although they are more likely to own land than non-poor households, they own much smaller plots on a per capita basis. They have a relatively low level of education, and school attendance among the poor is lower than average. Although they come from somewhat larger households, the percentage of household members who are children does not differ substantially from the average, and the percent of working household members is substantially higher than that of a typical Peruvian household.  - 46 - B. Poverty in Urban Peru The description of poverty in the previous subsection does not account for differences between poverty in urban and in rural areas. For example, it is doubtful that the urban poor are workers in agriculture. This subsection will examine poverty in Lima and other urban areas, while the next subsection will do the same for rural areas of Peru. As pointed out above, the poor are much less numerous in urban areas than in rural areas given a poverty line for Peru as a whole (see Table 15). Yet poverty has a relative aspect, so that one could classify the urban poor as the poorest 10 percent or 30 percent of urban residents. Of course, this implies a higher poverty line as measured by adjusted consumption expenditures. In this subsection the urban poor will be divided into the poor in Lima, in the Coastal urban region, in the urban Sierra and in the urban Selva. The poverty lines in Lima for the poorest 10 percent and poorest 30 percent are 283.5 Intis and 411.3 Intis per month, respectively. The analogous poverty lines in Coastal urban areas are 200.5 Intis and 315.9 Intis per month. Those for Sierra urban areas are 173.5 Intis and 307.4 Intis, while those for Selva urban areas are 192.4 Intis and 330.6 Intis. Table 22 presents basic information on household size and per capita consumption for the urban poor in Peru. It also includes average figures for all four urban areas. As with Peru as a whole, household size is usually somewhat larger for the urban poor relative to urban residents in general, the one exception being the urban Sierra. For Lima and the other urban areas, the food consumption of the poorest 30 percent is between 40 percent and 50 percent of what it is for an average urban resident. The corresponding  - 47 - fraction for the poorest 10 percent ranges from one fourth to one third. In terms of total consumption expenditures the spending level of the poorest 30 percent is slightly under 40 percent of the average level in Lima and the Coastal urban areas, while the figure for the Sierra and Selva urban areas is about 30 percent. For the poorest 10 percent of the population the corresponding percentages are 28 percent for Lima and the Coastal urban areas and about 19 percent for the urban Sierra and Selva. It is noteworthy that Lima is significantly better off than the other urban areas. In terms of total expenditures, the poorest 30 percent in urban areas other than Lima are at about the same welfare level as the poorest 10 percent in Lima.  - 48 - Table 22: Household Size and Per Capita Consumption of the Urban Poor Lima Coastal Urban Sierra Urban Selva Urban Poorest Poorest Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All 10% 30% All Household Size Actual 8.4 7.7 6.6 7.8 8.0 6.9 6.5 7.3 6.8 9.2 8.3 6.8 Adjusted 6.1 5.5 4.7 5.3 5.4 4.7 4.4 4.B 4.5 5.8 5.2 4.4 Food Consumption Actual 101.1 132.4 270.7 69.6 98.5 191.4 57.5 88.9 212.3 47.7 71.0 176.3 Adjusted 136.1 186.7 376.1 99.8 140.5 283.6 83.5 134.0 313.3 75,9 112.3 258.3 Total Consumption Actual 166.3 220.0 569.2 112.2 158.5 390.7 90.4 136.8 442.0 94.7 140.4 553.7 Adjusted 221.9 307.1 770.9 159.3 224.2 569.8 129.9 204.2 649.9 147.8 221.8 792.0 Note: 1. Figures are per capita averages. 2. Consumption is measured in Intis per month at June 1985 prices. Table 23 investigates whether female-headed households are more common among the urban poor. In all urban regions except the Selva, households headed by women are more likely to be found among the poor. This is particularly true of Coastal urban areas, where over one third of the poorest 10 percent of the population live in female-headed households. This suggests that living in a household headed by a woman entails a disadvantage in the Coastal urban areas of Peru, and to a lesser extent in the urban Sierra. The previous subsection highlighted the fact that most poor people in Peru (given a national poverty line) are self-employed farmers in rural areas. Of those who are relatively poor in the urban areas, what jobs do they take and for whom do they work? This information is given in Table 24.  - 49 - Table 23: Sex of Household Head Among the Urban Poor Lima Coastal Urban Sierra Urban Selva Urban Poorest Poorest Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All 10% 30% All Male 83.5 82.5 86.2 65.6 76.2 83.2 79.9 85.3 86.7 91.0 77.1 83.0 Female 16.5 17.5 13.9 34.5 23.8 16.8 20.1 14.7 13.4 9.0 22.9 17.0 In Lima those who are relatively poor are more likely to live in households where the head works for a private employer or is self-employed, but the difference is not very large. Even though heads of poor families are less likely to work for the government, a sizeable proportion still do even among the poorest 10 percent. The differen,ces with respect to occupation are not large, except that poor people are more likely to work in industrial or craft jobs and less likely to have white collar employment. In other urban areas, self-employment is even more important among the poor due to the fact that many people work in agriculture even though they are classified as living in urban areas. As one would suspect, given the findings in Peru as a whole, poor people in other urban areas are more likely to be found in agriculture relative to the general population in those areas. Government employment of the head of household is negatively correlated with poverty, as are white collar occupations.  - 50 - Table 24: Distribution of the Urban Poor by Employer and Occupation Lima Coastal Urban Sierra Urban Selva Urban Poorest Poorest Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All 10% 30% All Employer None 7.2 5.4 8.6 6.7 11.8 8.8 1.6 4.8 6.7 0.0 5.4 3.3 Government 8.3 10.2 15.2 2.9 3.7 10.3 10.5 10.7 17.2 0.0 10.4 16.4 Parastatal 2.1 1.1 3.4 0.0 0.3 5.4 0.0 0.0 4.7 0.0 3.3 2.4 Private 39.6 37.8 33.8 28.6 28.6 24.9 16.8 19.4 15.9 15.7 17.9 13.2 Private Home 1.1 2.4 1.0 6.9 3.4 2.0 4.9 3.8 1.5 0.0 0.0 0.0 Self-Employed 41.7 43.2 38.1 54.8 52.2 48.7 66.1 61.2 54.0 84.3 62.9 64.8 Occupation None 7.2 5.4 8.4 6.7 11.8 8.8 1.6 4.8 6.7 0.0 5.4 3.3 Agricultural 1.5 3.3 4.2 25.8 20.5 13.6 30.3 21.4 15.4 49.4 27.9 19.0 Sales/Services 32.7 39.2 33.9 36.1 34.3 38.7 27.0 30.2 30.8 5.6 24.2 35.5 Industry/Craft 51.0 40.1 30.6 28.2 29.1 25.2 32.6 38.3 29.2 44.9 40.0 25.8 White collar 7.6 11.3 22.9 3.1 4.3 13.8 8.6 5.3 17.7 0.0 2.5 16.5 Other 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.0 0.0 0.0 Since many poor in urban areas other than Lima work in agriculture, it is worthwhile to examine land ownership among the urban poor in Peru. This information is given in Table 25. Land ownership is rare among the poor and the population as a whole in Lima and the Coastal urban areas, but about one half of the poorest 10 percent and one third of the poorest 30 percent own land in the Sierra and Selva urban areas. As in Peru as a whole, the poorer landowning households have much smaller parcels of land than the average landowning household. In the urban Sierra the average plot of land owned among the poorest 10 percent of the population is only about one sixth of the average plot held by urban Sierra landowners. The respective figure for the urban Selva is about one third. It is likely that land ownership is an important determinant of welfare in the Urban Sierra and Selva.  - 51 - Table 25: AgrIcultural Land Owrship Among Urban Poor Lima Coastal Urban Sierra Urban Selva Urban Poorest Poorest Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All 10% 30% All All Land Percentage who own land 5.3 5.0 4.7 7.9 11.8 9.5 45.1 30.7 29.2 49.4 38.3 27.9 Land/capIta: All households 0.01 0.02 0.01 0.04 0.05 0.06 0.14 0.11 0.56 0.26 0.64 0.50 Landowning households 0.24 0.38 0.28 0.55 0.39 0.61 0.31 0.35 1.92 0.52 1.68 1.80 Irrigated Land Percentage who own land 3.0 4.0 4.1 7.9 11.8 9.1 24.3 16.1 16.3 23.6 8.8 8.6 Land/capita: All households 0.00 0.00 0.01 0.04 0.05 0.06 0.08 0.05 0.04 0.18 0.07 0.10 Landowning households 0.01 0.02 0.18 0.55 0.39 0.61 0.32 0.30 0.26 0.76 0.76 1.18 Table 26 gives the distribution of the urban poor by the educational level of the head of household, and also gives rates of school attendance. In Lima the poor live in households where the head has a much lower educational level relative to all households in Lima. This is also the case for other urban areas. School attendance does not vary dramatically among the poor and non-poor in any of the urban areas except that attendance among children aged 6-10 is somewhat lower among the poorest 10 percent (with the exception of the poorest 10 percent in Coastal urban areas). It seems that children from poor families start school at a later age than urban children in general.  - 52 - Table 26: Education and Pbverty In Urban Areas Lima Coastal Urban Sierra Urban Selva Urban Poorest Poorest Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All 10% 30% All Education of Head None 3.5 4.4 2.2 12.0 11.3 6.6 22.0 13.5 7.7 12.4 12.5 7.3 Elementary 63.7 56.0 37.2 73.7 64.6 53.7 54.9 59.8 44.0 68.5 70.4 47.9 Secondary (some) 15.2 17.0 16.6 10.3 13.9 14.1 10.2 13.6 13.5 7.9 10.4 17.5 Secondary (degree) 15.4 16.4 23.7 4.1 7.0 16.0 11.2 9.3 14.9 11.2 4.2 12.1 Post-Secondary 0.0 2.7 3.8 0.0 2.1 3.3 0.0 1.6 5.4 0.0 2.5 4.0 University 2.2 3.6 16.5 0.0 1.1 6.2 1.6 2.3 14.2 0.0 0.0 11.2 School Attendance Ages 6-10 85.7 94.3 95.2 98.0 91.8 94.3 86.1 92.5 92.6 75.0 85.7 88.8 Ages 11-15 96.0 95.5 95.2 96.4 94.9 94.4 94.3 93.4 93.0 100.0 96.3 91.5 In the previous subsection, certain household amenities were less common among the poor, but much of that may have been due to differences between urban and rural areas. Table 27 gives information on sources of drinking water and lighting, and type of sewage disposal. In Lima there are few differences between the poor and the general population - the main source of water for the poor is public waterworks. Electric lighting is still the most common source of lighting for the poorest in Lima, although candles and lamps are also used. Finally, public service sewage disposal is predominate even among the poor. In the Coastal and Sierra urban areas the most common source of water for the poor and non-poor is again public waterworks, while in the urban Selva this is only true of the general population - wells and other sources are more common among the poor. Electric lighting is predominant for the general population in urban areas outside of Lima, but the poor often depend on  - 53 - candles or lamps. Public service sewage disposal is the most common form among the poor in Coastal urban areas. It is only secondary, however, in the urban Sierra, and completely unavailable to the poor in the urban Selva. In the last two areas the poor most often have no sewage disposal at all.  - 54 - Table 27: Housing Chiracteristics of th Ueban Poor Lima Coastal Urban Sierra Urban Selva Urban Poorest Poorest Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All 10% 30% All Source of drinking Water Public: Inside Dwelling 71.8 71.8 78.5 59.0 65.6 71.5 44.7 52.9 60.5 15.7 26.7 54.2 Inside Building 12.6 11.3 9.7 10.3 6.7 5.2 15.1 8.7 12.2 0.0 1.3 5.5 Outside Building 1.8 4.8 3.2 13.8 9.0 7.8 3.0 7.2 9.5 11.2 8.8 5.2 Well 3.4 2.6 2.0 0.0 0.9 0.6 17.4 13.6 9.0 49.4 32.1 16.6 River/Spring 0.0 0.0 0.0 0.0 0.0 0.1 2.6 1.5 0.7 0.0 0.0 0.0 Water truck 7.2 7.3 5.1 10,0 9.2 7.9 2.0 5.3 2.9 0.0 0.0 0.8 Other 2,9 2.3 1.6 6.9 8.7 7.0 15.1 10.9 5.2 23.6 31.3 17.7 Source of Lighting Electricity 91.7 93.3 95.8 58,2 64,4 73.0 53.6 61.0 80.2 44.9 46.7 72.5 Kerosene/Oil 5.8 5.0 2.8 40.6 33,6 25.3 27.0 21.7 12.6 55.1 51.3 24.5 Candles 2.5 1.7 1.4 1.2 2.0 1.7 19.4 16.6 7.0 0.0 2.1 3.0 None 0.0 0.0 0.1 0.0 0.1 0.0 0.0 0.7 0.2 0.0 0.0 0.0 Sewerage Public Service 72.2 75.2 83.9 42.0 54.1 61,0 27.3 30.1 54.5 0.0 5.8 43.9 Well - Septic 2.2 1.9 1.4 4.3 3,5 3.2 5.6 4.6 2.9 0.0 5.4 1.6 Cesspool. 15,4 14,6 9.5 19.8 14,6 11.8 4.3 7.4 6.2 22.5 15.8 11.1 None 10.3 8.4 5.3 33.9 27.9 24,1 62.8 57.9 36.4 77.5 72.9 43.4 The previous subsection made the point that the poor in Peru belong to households that have relatively few children and more workers than the typical household. Is this the case in both urban and rural areas? Table 28 provides the relevant information for Lima and other urban areas. The same patterns that were found in Peru as a whole are generally found in urban areas - Poof households have a larger proportion of working household members (with the exception of urban Coastal areas) and roughly the same proportion of children. However, it is important to note that the proportion working is smaller in urban areas than in Peru a whole (cf. Table 21). This is probably  - 55 - due to the relatively small degree of self-employment in L2ban areas. Self- employed heads of household, particularly farmers, have other family members Table 28: Household Composition Paong the Urban Poor Lima Coastal Urban Sierra Urban Selva Urban Poorest Poorest Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All 10% 30% All Children 0-6 yrs. 11.6 13.9 14.2 14.6 15.1 16.1 16.4 18.2 17.6 25.8 20.4 18.1 Children 7-12 yrs. 13.0 14.6 14.1 14.6 14.1 16.2 16.8 16.2 15.9 14.6 19.2 16.6 ChIldren 13-17 yrs. 13.7 13.3 11.1 14.6 13.9 13.0 10.2 13.1 13.0 6.7 11.7 13.0 Workers 48.3 47.0 46.1 40.6 41.9 43.1 53.9 52.8 48.0 46.1 38.8 42.0 working with them, which would raise the percentage of workers among family members. Since self-employment is much less lucrative than formal employment, it is not surprising that this percentage is negatively correlated with welfare levels. It simply indicates that incomes from self-employment, even with other household members providing labor, are relatively small. To summarize, the poor in urban areas of Peru have substantially lower welfare levels as measured by expenditure levels, relative to other urban residents, but many would not be considered poor if compared to the Peruvian population as a whole. The poor in urban areas are less likely to work for the government - they are more often self-employed or employees in the private sector. White collar work is relatively uncommon among the urban poor. Landownership is relatively rare in Lima and the Coastal urban areas but is likely to play an important role in determining welfare levels in the  - 56 - urban Sierra and Selva since poorer landowning households have much smaller plots than their wealthier counterparts. As one would expect, the education level of the heads of poor urban households are lower than those of urban residents in general. However, school attendance in urban areas is not substantially lower among the poor and is generally quite high. Household amenities such as electric lighting and flush toilets are common even among the poorest in Lima, though less so in other urban areas. Finally, in most cases poor urban households have about the same proportion of children and a higher proportion of working members than typical urban households. C. Poverty in Rural Peru In the previous subsection the poor in urban Peru were defined in terms of the poorest 10 percent and the poorest 30 percent of the urban population. This subsection examines the poorest 10 percent and 30 percent of the rural population. Specifically, the three rural regions (Coastal rural, rural Sierra and rural Selva) are all examined using this definition of poverty within each region. Thus we examine the poorest 30 percent in the Coastal rural region, then the poorest 30 percent in the rural Sierra, etc., and similarly for the poorest 10 percent, so that all three rural regions have different poverty lines. The poverty lines for the poorest 10 percent and 30 percent in the rural Coastal region are 144.5 Intis per month and 230.3 Intis per mohth, respectively. The corresponding poverty lines for the rural Sierra are 113.3 and 199.5 Intis per month, and those for the rural Selva are 123.0 and 194.5. It is important to recognize that the poorest 10 percent (30 percent) in the rural Coastal region are somewhat better off than the poorest 10 percent (30 percent) in the other two rural regions.  - 57 - Table 29 provides some useful data on household size and per capita consumption for the rural poor in Peru. Two points should be emphasized. First, the poorest 30 percent and 10 percent of the population in rural areas are much poorer than the analogous population in urban areas (cf. Table 22). Second, within rural areas there is a definite ranking, with the Coastal rural areas doing relatively well. Table 29: Household Size and Per Capita Consumption of the Rural Poor Coastal Rural Sierra Rural Selva Rural Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Household Size Actual 8.5 8.0 7.3 7.3 6.9 6.4 8.1 7.9 6.7 Adjusted 5.4 5.2 4.6 4.6 4.4 3.9 4.8 4.6 4.0 Food Consumption Actual 57.2 80.5 173.3 38.7 67.4 169.7 33.9 55.2 154.6 Adjusted 87.7 121.2 276.0 58.3 102.9 265.2 55.0 90.4 241.4 Total Consumption Actual 75.4 111.1 265.5 53.4 87.0 235.4 56.5 84.0 264.8 Adjusted 115.6 167.0 421.3 80.7 132.5 366.8 91.6 138,1 413,5 Note: 1. Figures are per capita averages. 2. Consumption is measured in Intis per month at June 1985 prices The proportion of poor who live in households headed by women is given in Table 30 for each of the three rural regions. There are no distinct general patterns. Poor households in the Coastal and Selva rural areas are somewhat less likely to be headed by women, though the Selva figures are based on a small sample (79 households for the poorest 30 percent, 25 for the poorest 10 percent). Only the rural Sierra area shows that poor households  - 58 - are more likely to be headed by women than non-poor households. Thus the relation between poverty and the sex of the head of household is somewhat ambiguous. Table 30: Sex of Household Head Among the Rural Poor Coastal Rural Sierra Rural Selva Rural Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Male 93.1 90.2 89.9 83.4 86.8 87.6 99.4 95.7 91.0 Female 6.9 9.8 10.1 16.6 13.2 12.4 0.6 4.3 9.0 The correlation between poverty and language of interview is given in Table 31. In Coastal rural areas all interviews were done in Spanish, while 8.5 percent and 21.6 percent of the interviews in the Selva and Sierra rural areas, respectively, were done in other languages. In the Sierra the poorest 30 percent are somewhat less likely to speak Spanish than the general population, but the opposite is true of the poorest 10 percent. In the Selva rural areas, Spanish is more likely to be spoken by the poor. It is thought that most rural Sierra residents and many rural Selva residents are Indians, so perhaps the use of language of interview to distinguish between Indians and non-Indians does not work very well in these areas.  - 59 - Table 31: Language of Interview Among the Rural Poor Coastal Rural Sierra Rural Selva Rural Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Language of Interview Spanish 100.0 100.0 100.0 82.9 76.2 78.4 100.0 92.9 91.5 Quechua 0.0 0.0 0.0 13.4 20.0 16.4 0.0 7.1 8.5 Aymara 0.0 0.0 0.0 3.7 3.5 4.9 0.0 0.0 0.0 Other 0.0 0.0 0.0 0.0 0.2 0.3 0.0 0.0 0.0 Working in the government was shown to be more lucrative than private sector work or self-employment, but since most of these jobs are in urban areas it may not be relevant for characterizing rural poverty. Table 32 presents statistics on the poor based on employers and occupational categories of heads of household. Most people in all three rural areas live in a households where the head is self-employed. Private sector employment is mainly important in the rural Coastal areas, where large agricultural operations employ local residents and also migrants from the Sierra. It seems that working for private employers is more common among the poor in the Coastal rural areas than among the general population. Agricultural occupations are most common for heads of households in all rural areas, though sales and service occupations and industrial or craft work also exist. There is little occupational difference among poor and non-poor households except that in the rural Selva the poor are disproportionately found in agricultural work.  - 60 - The incomes earned by the rural poor depend in part on the land they own. Table 33 provides data on land ownership among the rural poor. Unlike Peru as a whole, there is little difference in terms of the percentage of poor and non-poor households who own land. The most important differences are in the quantity (in per capita terms) of land owned. In Coastal rural areas the poorest 10 percent of landowning households had slightly more than half the land of the average landowning household in that region. In the rural Sierra the difference was relatively small. Finally, in the rural Selva there is a huge difference in land ownership between the poor and non-poor - landowning households among the poorest 10 percent had only about one fifteenth of the land owned by an average landowning household in that region. It seems that the land reform efforts had little effect on the distribution of land in the rural Selva area. In Section III it was seen that most educated people reside in urban areas. Yet it is still important to quantify the educational composition of the poor in rural Peru, and it is also useful to examine school attendance for children in rural areas. This can be seen in Table 34. Most people in rural areas live in households where the head had no education at all or only an elementary level of education. In all three rural areas this is even more likely among poor households. School attendance patterns are somewhat similar to those found in urban areas. In the rural areas along the coast and in the  - 61 - Table 32: Distribution of the Rural Poor by Employer and Occupation of the Head of Household Coastal Rural Sierra Rural Selva Rural Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Employer None 7.4 4.8 0.8 5.9 5.6 3.9 1.8 1.6 1.4 Government 0.0 0.0 1.4 2.1 1.3 2.6 0.0 0.0 2.4 Parastatal 0.0 0.0 0.4 0.0 0.0 0.8 0.0 0.0 0.0 Private 31.4 27.8 22.1 6.0 9.3 12.0 3.5 5.5 5.9 Private Home 12.8 5.0 4.3 3.0 2.4 2.5 0.0 0.0 3.4 Self-Employed 48.5 62.4 68.7 83.0 81.5 78.2 94.7 92.9 86.9 Occupation None 7.4 4.8 3.1 5.9 5.6 3.8 1.8 1.6 1.4 Agricultural 73.0 73.8 76.3 80.8 81.9 76.8 97.7 90.0 87.9 Sales/Services 7.4 11.0 11.8 3.7 2.6 5.8 0.6 0.6 4.3 Industry/Craft 12.3 10.5 7.2 9.6 8.8 10.5 0.0 7.9 5.6 White Collar 0.0 0.0 1.5 0.0 1.1 3.1 0.0 0.0 0.9 Other 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Table 33: Agricultural Land Ownership Among the Rural Poor Coastal Rural Sierra Rural Selva Rural I~~~~~~ Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All All Land Percentage who own land 71.6 66.9 68.1 89.1 91.8 89.1 97.6 94.8 91.5 Land/capita: All households 0.19 0.26 0.33 0.53 0.46 0.75 0.30 1.24 4.65 Landowning households 0.26 0.38 0.48 0.60 0.51 0.84 0.30 1.31 5.08 Irrigated Land Percentage who own land 53.4 50.9 60.5 38.1 42.1 41.1 12.6 11.1 16.3 Land/capita: All households 0.12 0.13 0.27 0.08 0.10 0.37 0.01 0.02 0.18 Landowning households 0.22 0.25 0.44 0.21 0.24 0.15 0.09 0.18 1.09  I - 62 - Sierra poorer children aged 6-10 are less likely to attend school, but the opposite is true in the rural Selva. In the first two rural areas poorer children aged 11-15 are as likely to attend school as children from the general population, yet such attendance is lower among the rural poor in the Selva. As before the Selva figures need to be treated with caution because of the low sample size. In the rural areas along the Coast and in the Sierra the patterns found are consistent with the hypothesis that children from poorer families begin attending school at a relatively late age. It may be that this late start manifests itself in generally lower levels of education when poor children leave school. Table 34: Education and Poverty in Rural Areas Coastal Rural Sierra Rural Selva Rural Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Education of Head None 40.2 34.5 22.8 42.7 37.5 27.1 20.0 24.9 18.9 Elementary 59.8 61.4 63.9 54.5 57.9 61.9 80.0 72.2 71.7 Secondary (somne) 0.0 1.5 5.2 1.4 3.8 6.0 0.0 2.0 5.0 Secondary (degree) 0.0 2.6 6.3 1.5 0.7 2.4 0.0 1.0 3.1 Post-Secondary 0.0 0.0 0.8 0.0 0.2 1.1 0.0 0.0 1.1 University 0.0 0.0 1.1 0.0 0.0 1.4 0.0 0.0 0.3 School Attendance Ages 6-10 75.0 86.0 87.7 65.0 68.3 76.1 85.2 77.5 80.2 Ages 11-15 84.4 83.1 84.9 78.6 72.8 79.8 73.1 71.8 82.0 Housing is quite different in rural areas relative to urban areas. Differences among rich and poor households can be seen in Table 35.  - 63 - Most water in rural areas comes from wells or natural sources, but such sources tend to be even more common among the poor. Lamps are more common than electric lighting in all three rural areas, but this is more true of the poor, with the exception of the Selva region. Finally, poorer people are less likely to have access to some type of sewage disposal system, although this is true to a lesser extent for the entire population. Table 35: Housing Characteristics of the Rural Poor Coastal Rural Sierra Rural Selva Rural Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Source of Drinking Water Public: Inside Dwelling 0.0 6.5 14.5 13.6 11.7 13.1 4.7 8.3 12.9 Outside Dwelling 0.0 0.0 0.0 1.5 1.8 2.0 0.0 0.0 0.3 Outside Building 2.5 3.1 4.1 4,6 3.7 5.2 0.0 1.4 2.2 Well 64.7 50.6 30.7 12.3 13.7 16.8 0.0 8.3 5.2 River/Spring 31.9 36.9 41.2 67.0 68.6 62.0 94.7 81.7 78.8 Water Truck 0.0 2.6 8.4 0.0 0.0 0.0 0.0 0.0 0.0 Other 1.0 0.3 1.1 1.0 0.7 0.9 0.6 0.4 0.6 Source of Lighting Electric 8.3 10.1 19.7 7.2 6.6 13.9 8.2 10.0 8.3 Kerosene/Oil 90.7 85.1 76.9 87.2 87.4 79.1 87.7 86.6 90.2 Candles 1.0 4.8 3.5 3.4 4.7 6.6 4.1 3.3 1.6 None 0.0 0.0 0.0 2.2 1.2 0.4 0.0 0.0 0.0 Sewerage Public Service 0.0 0.0 2.1 2.3 1.3 5.4 0.0 0.0 2.2 Well-Septic 0.0 0.0 1.2 0.0 0.4 0.5 0.0 3.9 3.0 Cesspool 3.4 11.8 17.8 0.0 0.5 2.8 10.6 9.0 9.3 None 96.6 88.2 78.9 97.7 97.9 91.3 89.4 87.0 85.5 One can also examine household composition in rural Peru. This is presented in Table 36. The same trends found in urban areas and in Peru as a whole also hold in the Sierra and Selva rural areas but not in the rural  - 64 - Coastal areas. Specifically, poorer households in the first two rural areas have a larger proportion of working members and about the same proportion of children than typical rural households in Peru, while poor households on the rural Coast have fewer workers as a percentage of household members. Table 36: Household Composition Among the Rural Poor Coastal Rural Sierra Rural Selva Rural Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Children: 0-6 yrs 18.1 17.8 20.3 19.5 20.4 21.8 23.5 25.1 23.4 7-12 yrs 17.6 16.1 18.4 16.4 16.5 17.9 17.6 16.1 17.6 13-17 yrs 16.2 16.0 13.6 13.5 12.4 11.5 14.7 15.3 12.4 Workers 53.4 56.8 57.4 66.4 66.3 64.2 57.6 59.3 57.4 To summarize, the poor in rural Peru are in some ways typical of other rural residents. There is not much difference in terms of the sex of the head of household and language of interview. Self-employment in agriculture, and to some extent in sales and services, is the predominant occupational arrangement among the poor and, to a lesser extent, among the general population. However, private employment is more common among the poor than the non-poor in Coastal rural areas. Land ownership is relatively equal in the rural Sierra, less equal in the rural Coastal areas, and very unequal in the rural Selva. When ranked by the education level of the head of household, poor people tend to live in "less educated" households in all rural areas. School attendance of children is generally lower among poor children aged 6-10 except in the Selva rural areas. School attendance for children  - 65 - aged 11-15 is also roughly constant over poor and non-poor rural households, but again the Selva areas show the opposite trend. In general it appears that poor children start school at a later age. Housing characteristics do not vary dramatically in rural Peru. Finally, one again finds that poor households have more workers and about the same number of children relative to wealthier households, which is consistent with the hypothesis that poverty is primarily a matter of low earnings among those who work. The exception to this is the Coastal rural areas, where poorer families have a slightly smaller proportion of workers relative to the Coastal rural population as a whole.  - 66 - VI. CONCLUSION Peru faces a difficult task in the coming years - it must revive economic growth while preventing its already large inequities from becoming larger. This paper presents a variety of information on the distribution of economic welfare in Peru in 1985-86 that sheds light on the nature, and to a lesser extent, the causes, of inequality in Peru. Household welfare levels are measured by (equivalence scale.adjusted) per capita consumption expenditures, as explained in Section II. The data are taken from the Peruvian Living Standards Survey (PLSS). There are several important lessons for policy makers regarding the distribution of welfare in Peru. First, education is highly correlated with welfare levels - households in which the head has a university education have consumption levels three to four times higher than those whose head has no education. Promotion of education among the poor would very likely have an equalizing effect on the distribution of welfare in the long run, and is also likely to be conducive to economic growth given an appropriate policy framework. High levels of school attendance, even among the poor, are a positive sign, yet recent reductions in the fraction of government expenditures devoted to educatiod must be carefully scrutinized. The lower quality of schooling in rural areas, where most of the poor are found, is also a matter of concern. Second, poverty in Peru is more severe in rural areas than in urban areas. Only 46 percent of all Peruvians are found in rural areas, yet 79.5 percent of the poorest 10 percent of the population reside there. The rural Sierra has the biggest pocket of poverty - 59.7 percent of the poorest 10 percent live there. Of the poorest 10 percent of all Peruvians, 70.8 percent live in households in which the head is employed in agriculture, usually self- employed. Since almost none of these households have heads working for the  - 67 - government or a government-owned corporation, raising wage rates received by those workers would have very little effect on poverty. Third, poor households have as many, if not more, working household members as a fraction of total members, than does the typical Peruvian household. This suggests that poverty is not a matter of unemployment, but of low incomes in the jobs the poor have. In addition, it does not appear that poor households have substantially more children than do Peruvian households in general, which again illustrates that poor households are not characterized by large numbers of dependents relative to employed members. Fourth, within urban areas the heads of poor households are primarily sales and service workers or industrial or craft workers. These compose the bulk of the poor outside of those found in agricultural pursuits, and they are both self-employed and employees in private firms. Distinct policies must be aimed at both these households as well as those in which the head works in agriculture. This paper is primarily descriptive in nature, but it does raise questions which should be further explored using the PLSS data. Three such questions are: (1) What are the determinants of educational achievement, and what role does the quality of school facilities play? (2) What determines the incomes of agricultural workers, especially the self-employed? (3) How is it that some workers obtained government jobs while others do not? The first issue hAs been examined briefly by Stelcner, Arriagada and Moock (1987), but all three stand in need of thorough research. Investigations of these questions will provide much needed information to guide Peru's policy choices as it attempts to restore economic growth and raise the welfare levels of its citizens.  - 68 - REFERENCES Anand Sudhir, and Christopher Harris. 1985. "Living Standards in Sri Lanka, 1973-1981/82: A Partial Analysis of Consumer Finance Survey Data," Mimeo. Deaton, Angus, and John Muellbauer. 1980. Economics and Consumer Behavior, Cambridge University Press. Deaton, Angus, and John Muellbauer. 1986. "On Measuring Child Costs: with Applications to Poor Countries," Journal of Political Economy. Vol. 94, No. 4, pp. 720-44. Eglin, Darrel R. 1981. "The Economy", in Nyrop (ed.) Peru: A Country Study. U.S. Government Printing Office. Glewwe, Paul. 1987. "The Distribution of Welfare in the Republic of C6te d'Ivoire in 1985". LSMS Working Paper No. 29, The World Bank. de Janvry, Alain. 1981. The Agrarian Question and Reformism in Latin America. Johns Hopkins University Press. Grootaert, Christiaan, and Ana-Maria Arriagada. 1986. "The Peruvian Living Standards Survey: An Annotated Questionnaire: Education and Training Department, The World Bank. Mimeo. Jain, Shail. 1975. The Size Distribution of Income: A compilation of Data. The World Bank. Kluck, Patricia. 1981. "The Society and its Environment", in Nyrop (ed.) Peru: A Country Study. U.S.,Covernment Printing Office. Newman, John L. 1987. "Labor Market Activity in C6te d'Ivoire and Peru". LSMS Working Paper No. 36, The World Bank. Sen, A.K., 1973. On Economic Inequality. Clarendon Press, Oxford University. Shorrocks, Anthony. 1980. "The Class of Additively Decomposable Inequality Measures," Econometrica Vol. 48, No. 3, pp. 613-25. Shorrocks, Anthony. 1982. "Inequality Decomposition by Factor Components," Econometrica Vol. 50, No. 1, pp. 183-211. Shorrocks, Anthony. 1984. "Inequality decomposition by Population Subgroups, "Econometrica Vol. 52, No. 6, pp. 1369-85. Stelcner, Morton, Ana-Maria Arriagada and Peter Moock. 1987. "Wage Determinants and School Attainment Among Men in Peru". LSMS Working Paper No. 38, The World Bank.  - 69 - Thomas, Duncan. 1986. "Food Shares as a Welfare Measure." Unpublished Ph.D. Thesis. Department of Economics. Princeton University. van der Gaag, Jacques. 1982. "On Measuring the Cost of Children." Children and Youth Services Review. Vol. 4, No. 1, pp. 77-109. Weeks-Vagliani, Winifred. 1985. Actors and Institutions in the Food Chain: The Case of Peru. OECD, Paris. World Bank. 1983, 1986. World Development Report. Oxford University Press. World Bank. 1985. "Peru Country Economic Memorandum." Washington, D.C. Mimeo.  - 70 - Appendix A: The Peru Living Standards Survey The Peru Living Standards Survey (PLSS) is a random sample of 5000 households interviewed from mid-July 1985, to mid-July 1986.!' The data collected include information on food and non-food expenditures, agricultural production and consumption of food produced, income from different sources, health and educational status of household members, employment and other productive activities, migration, housing conditions, and a variety of other subjects. The survey is thoroughly described in Grootaert and Arriagada (1986). The data pertinent to measurement of consumption are discussed in this appendix. Data collected in the PLSS which are relevant for the measurement of consumption include: 1) Expenditures tn regularly purchased non-food items (such as fuel, cigarettes and personal health items) and food consumed outside the household within the last two weeks;2- 2) Expenditures on clothing, household goods and maintenance, medicines and other irregular expenditures within the last three months; 3) Food expenditures in the last two weeks; 4) Possession of durable goods, including present value and cost when purchased; 5) Value of food produced and consumed by the household in the last three months; and 6) Value of payments in kind (food and non-food) received by household members. Finally, data were collected on rents paid by households who were renters, which is useful in estimating imputed rents for owner- occupied housing. / Due to some missing information, 5 households have been dropped from the original 5000. 21 Specifically, between the time the first and second interviews were conducted, which is generally two weeks.  - 71 - The first four expenditure categories listed above clearly raise the welfare of households; these data are converted into monthly figures and summed to give monthly expenditures for each household. However, the use value of durable goods and the value of owner-occupied housing must also be calculated to get an accurate measure of consumption. The enjoyment of housing and durable goods does not take place at the time they are built or purchased, but instead extends over a long period of time (usually several years) during which they are used. Thus the welfare received from such goods must be based on estimated rental values of owning the good. For housing, the best approach is to estimate hedonic rent equations (i.e. to predict the rental value of housing based on the characteristics of the dwelling) for those households which are renters. Thus imputed rents can be calculated for households which own their dwellings based on the characteristics of those dwellings. This has been done separately for Lima and other urban areas. After using appropriate methods to correct for sample selection bias, rents were estimated for renters as a function of household characteristics such as floor area, type of dwelling, source of water and source of lighting. Imputed rents were than calculated for home owners, again after adjusting for sample selection bias. 31 31 Not all variibility in rents was captured in the hedonic rent equation since the R statistic was 0.514 in Lima and 0.559 in other urban areas. The estimated rents given by home owners could be predicted more easily (see next paragraph), the R statistics for Lima, other urban, and rural areas were 0.692, 0.608 and 0.576, respectively. Thus predicted rents for home owners do not reflect all the variability in this component of total consumption. Since rents (imputld or actual) account for only 5- 10 percent of total consumption and the R statistics were rather high, this does not seem to be a serious problem.  - 72 - Unfortunately, there is almost no r-ental market in the rural areas of Peru, which prevents one from estimating imputed rents in rural areas. This forces one to use estimaLed rerntal valtue as given by non-renting survey respondents. For ruiral households which did nIot give an adequate response, these estimated rents were regressed on various housinig characteristics in order to provide an econometric estimate of irnputed rents. For durable goods, tthe rental value can be estimated based on depreciation in the real value of those goods over time. The effective rental price of a durable good is its depreciation in value over the year in question (which can be calculated from the data on estimated present value and on cost when purchased) and the opportunity cost of owning the good in terms of foregone investment earnings. Estimates of imputed rents and of the use-value of durables were added to the four categories of expenditures described above. The food expenditure variable used in this paper is simply the sum of food expenditures, the value of food produced and consumed within the household, and the cost of food eaten outside the household (including free or subsidized food received from employers). All of the above deals with nominal rather than real incomes, yet Peru has experienced high rates of inflation in recent years. During the time of the Peru Living Standards Survey (July 1985 to July 1986) the consumer price index rose by over 70 percent. In examining the distribution of welfare in Peru it is necessary to adjust for this by deflating reported expenditure levels according to a monthly price index. In addition, it is necessary to adjust for differences in prices across regions at any given point in time.  - 73 - Monthly price indices are available in Peru (Ilustituto Nacional de Estadistica) from each of 13 cities. The 250 sampling areas from the PLSS were all assigned to the nearest of these 13 cities. This allows one to deflate nominal expenditures within the 13 "regions," so that the problem of price differences over time can be easily handled. However, price differences across these 13 regions are not evident in these price indices, which are all indexed at 100 for June 1985. Prices were obtained from the government of Peru for 7 food items and 5 non-food items. Nominal expenditures for the entire country as estimated in the PLSS were used to obtain weights for the consVruction of an inter-regional price index. These weights were used with the .rice data, adjusted for regional population, to create inter-regional price indices for food and non-food items. These indices are given in Table A.1. One comment needs to be made with respect to Iquitos. This city is in a very isolated region of the jungle (Selva) near the Amazon river close to the Brazilian border. Maps of Peru indicate that no roads connect this city with either the rest of Peru or Brazil, and only 5.1 percent of Peru's population live in this region. In addition to transportation difficulties, differences in diet may also result in high food prices for items consumed at lower prices elsewhere in Peru. There is little that can be done to correct for this, but it may be that deflated food expenditure data from Iquitos under-estimate welfare in that region. Since food expenditures are a substantial portion of total expenditures, deflated total expenditure figures may suffer from the same problem. This fear is partially put to rest by the fact that the urban Selva has the highest level of welfare in terms of deflated per capita expenditures.  - 74 - Table A.1: Inter-regional Price Indices (National = 100) for Peru, 1985-86 Region Food Non-Food Total Arequipa 88.9 100.3 95.8 Cajamarca 91.3 92.4 91.9 Cusco 87.2 80.6 83.2 Chiclayo 94.7 96.5 95.8 Chimbote 100.4 104.1 102.6 Huancayo 95.6 98.0 97.0 Ica 95.2 90.7 92.5 Iquitos 188.2 85.1 126.3 Lima 99.3 110.3 105.9 Piura 94.7 106.1 101.5 Puno 86.9 74.7 79.6 Tacna 96.6 105.7 102.0 Trujillo 94.2 110.9 104.2 Real expenditures for each household are calculated by first deflating the household's food and non-food nominal expenditure levels by the respective (food and non-food) regional deflators, after which the regional monthly deflators are used to obtain food and non-food purchases in real terms. All expenditures in the text are in constant June 1985 Intis. The non-food monthly deflator calculated for each region is based on nominal expenditure shares of food and non-food items and on the food and total expenditure deflators provided by the Peruvian government. At this point one can check to see whether different methods of defining the overall distribution of welfare in Peru have a substantial effect on the level of inequality. This is done in Table A.2, which demonstrates the impact of different definitions of or refinements to the measure of welfare. Column 1 gives shares of total welfare going to each 10 percent of the  - 75 - population (from poorest to wealthiest) as given by per capita total consumption after adjusting for regional price differences and household composition (i.e. smaller weights for children). This, of course, is the welfare measure used in this paper. Column 2 gives decile shares of total welfare going to each 10 percent of households. This can be misleading since households vary in number of members. In particular, small households are more likely to be classified as poor simply because they are small, while large households are more likely to be classified as rich because they are large. This has a tendency to exaggerate inequality by adding an erroneous source of variation, which is evident in column 2 of Table A.2. What if no adjustments are made to account for the fact that additional household members, particularly children, are less costly due to certain economies of scale? If poorer families have more children, than inequality may be exaggerated. There is some evidence for this in Table A.2 but the impact is not very strong. In any case, adjustment for "economies to scale" should not be overlooked when comparing the welfare levels of households.  - 76 - Table A.2: Welfare Deciles According to Different Welfare Measures Food Consump- Adjusted Per No Composition tion Per Decile Capita Consumption By Households Adjustment Capita 1 2.01 1.53 1.80 2.31 2 3.39 2.84 3.07 4.03 3 4.46 3.93 4.10 5.26 4 5.52 5.08 5.08 6.34 5 6.57 6.30 6.14 7.45 6 7.75 7.68 7.51 8.69 7 9.42 9.45 9.25 10.14 8 11.80 11.93 11.70 12.10 9 15.71 16.15 16.00 15.21 10 33.38 35.13 35.36 28.48 Next, one can compare food consumption with total consumption. The former is more equitably distributed than the latter, which is not surprising given that wealthier households are likely to spend proportionately less on food than poor households. As a welfare measure in itself, food consumption has some intuitive appeal. However, for reasons given in Appendix B, it is not used in this paper. Finally, it is useful to give some summary data on consumption in Peru. Table A.3 gives the breakdown of total consumption by various categories. In Peru as a whole, slightly more than half of total consumption (including imputed rent and the value of durable services) is food consumption. Of non-food consumption, rents (imputed or real) and durable services play a relatively minor role. When one examines consumption patterns by different areas of Peru, one finds that the fraction of total expenditures devoted to food consumption is generally smaller in wealthier regions than in poorer regions.  - 77 - Table A.3: Composition of Total Consu ptlon In Peru, 1985-8C All Coastal Coastal Sierra Sierra Selva Selva Peru Lima Urban Rural Urban Rural Urban Rural Food purchased 32.7% 36.2% 36.6% 38.5% 33.4% 24.6% 21.2 24.1 Food produced 14.6 4.2 6.6 22.9 9.9 42.9 5.2 29.4 Food eaten away from home 3.9 5.0 3.7 2.8 3.4 2.7 3.0 3.7 Food provided by employer 1.8 2.1 1.8 1.1 1.3 1.7 2.5 1.2 Rents 6.1 6.5 8.6 4.1 6.2 2.9 9.8 4.9 Water/Elec. Utilities 0.3 0,4 0.4 0.2 0.4 0.1 0.4 0.1 Durable services 6.2 6.9 7.7 2.9 8.2 2.6 11.2 3.5 Other non-food 34.4 38.6 34.5 27.6 37.2 22.5 46.8 33.2 Total food 53.0 47.6 48.7 65.3 48.0 71.9 31.8 58.4 Total non-food 47.0 52.4 51.3 34.7 52.0 28.1 68.2 41.6 Total value Intis per capita per month 384.7 569.2 390.7 265.5 442.0 235.4 553.7 264.8 Note: 1. Figures have not been adjusted by household composition. 2. lntis figures are at June 1985 prices.  - 78 - Appendix B: Food Consumption as a Welfare Measure As pointed out in Section II, one could use either total expenditures or food expenditures to determine the welfare levels of households. Both methods have advantages and disadvantages, but it would be useful to find a method of comparing the rankings to judge the relative merits of both methods. One possible standard is Engel's law, which states that total expenditures increase, the proportibn of total expenditures spent on food declines. The argument here is that the expenditure elasticity of food consumption is less than one because food is considered to be a "necessity" rather than a "luxury." Engel's law has received a large amount of empirical support in both developed and developing countries. Thomas (1986) has done a careful examination of the accuracy of Engel's law using a large number of data sets and finds that it holds in general, though perhaps not for the poorest people in some countries. Since total expenditures, ceterus paribus, should be monotonically related to welfare, this implies that the proportion of total expenditures allocated to food should decline as welfare rises. Given this reasoning, an accurate welfare ranking ought to show a decline in the fraction of total expenditures spent on food as one moves from people at low welfare levels to people at higher levels. Table B.1 gives the relevant figures for households ranked by per capita food expenditures. As can be easily seen, food shares do not decline as welfare levels rise when welfare is based on per capita food consumption (either adjusted or unadjusted). This indicates that such a welfare measure should not be used for Peru. In contrast, food shares do decline as welfare increases when total expenditure is used as a welfare ranking, as seen in Table 2 in the text.  - 79 - Table B.1: Distribution of Food Consumption by Food Consumption Deciles Mean Per Capita Food % of Food Expenditures Expenditures Food Share (Z) (Intis par month) in Peru within each Decile Decile Adjusted Unadjusted Adjusted Unadjusted Adjusted Unadjusted 1 69.6 44.2 2.31 2.17 0.446 0.474 2 121.5 78.5 4.03 3.84 0.547 0.557 3 158.8 101.7 5.26 4.98 0.562 0.579 4 191.5 123.8 6.34 6.07 0.563 0.563 5 225.0 147.1 7.45 7.21 0.545 0.537 6 262.1 174.3 8.69 8.54 0.541 0.574 7 306.0 206.1 10.14 10.10 0.553 0.544 8 365.1 247.8 12.10 12.14 0.530 0.513 9 458.9 310.9 15.21 15.24 0.513 0.487 10 859.5 606.2 28.48 29.71 0.562 0.532 All Peru 301.8 204.1 100.00 100.00 0.542 0.530 Note: Figures in Intis are at June 1985 prices. It is not hard to imagine why welfare ranking based on per capita food expenditures may be misleading. Leaving aside the difficulties encountered with differences in household composition, suppose one has a set of individuals among whom total expenditures are unequally distributed. Under  - 80 - the assumption that Engel's law holds, one would expect relatively wealthy individuals to spend, on average, a smaller percentage of total. outlay on food, even though random fluctuations in food expenditures may lead to variations among individual people. An accurate measure of total expenditure is an unambiguous indicator of welfare rankings with which to compare individuals, but if food expenditures are relatively inelastic with respect to total outlay (i.e. the expenditure elasticity is greater than zero but less than one) and are ;usceptible to random shocks, then a negative shock to food expenditures for a given individual will mistakenly lead one to classify him as having a lower level of welfare than is indicated by total expenditures. In addition such a negative shock will reduce the individual's food share (food expenditures as a fraction of total expenditures). By an analogous argument, a positive shock results in (mistakenly) raising an individual's welfare ranking while simultaneously raising his food share. Given a sufficiently large random term in the determination of food expenditures one can obtain welfare rankings based on food shares which do not conform to Engel's law even though it is assumed to hold on average. This hypothesis is consistent with the data in Table 2 in the text and Table B.l, and if true it implies that per capita total expenditure is a better indicator of welfare than per capita food expenditure.  - 81 - Appendix C: Measurement of Inequality Given a measure of welfare of individuals, an aggregate statistic which records the level of inequality among these individuals can be selected. Perhaps the best strategy is to specify characteristics which one would like an inequality measure to have and then use all proposed measures which satisfy those criteria. There are four characteristicsl! which are highly desirable: 1. Mean Independence - inequality is unaffected by equi- proportionate charnges in everyone's income; 2. Population-Size Independence - the same distribution of income over a larger or smaller population does not affect measured inequality; 3. Symmetry - exchanging income levels among different people does not affect inequality; and 4. Pigou-Dalton Transfer Sensitivity a transfer of income from a wealthy person to a poor person reduces measured inequality. Virtually all proposed inequality measures are population-size independent and symmetric and most are mean-independent (though variance is not) and sensitive to Pigou-Dalton transfers (though variance of the logarithm of income is not for high incomes). For detailed discussions of measurement of inequality see Sen (1973), Shorrocks (1980, 1982, 1984) and the references cited by both authors. Many suggested measures are eliminated by the following characteristics which are desirable, but not necessary, for a measure of inequality: 5. Decomposability - total inequality can be additively broken down by population groups or income sources; 6. Statistical Testability - one can test whether differences in inequality over time or between groups are Although these properties are described in terms of income, their essential nature is unchanged when expenditure data (adjusted or unadjusted) are used to measure the distribution of welfare.  - 82 - statistically significant. It turns out that decomposability by income sources (where total inequality is assumed to be a covariance-weighted sum of measured inequality from each income source) is, given generally acceptable axioms, independent of the measure of inequality chosen (Shorrocks, 1982), so that income source decomposability does not reduce one's choice of inequality measures as long as they meet the first four criteria. However, group decomposability (where total inequality is the weighted sum of inequality measured within each group plus inequality between the mean incomes of the different groups) limits one to the two entropy measures proposed by Theil (Shorrocks, 1980, 1984). The variance of the logarithm of income is also group decomposable but unfortunately it does not satisfy Pigou-Dalton transfer sensitivity for large incomes. Yet one may want to use this measure because, given the assumption that expenditures follow a lognormal distribution, one can test whether the difference in inequality between two different distributions is statistically significant. Future work on inequality measurement should focus on whether other inequality measures, particularly the two Theil measures, are amenable to statistical tests. Confining the analysis to the distribution of expenditure, and not income, means that income source decompositions cannot be used. This puts more weight on judicious use of group-decomposable measures of inequality for interpreting overall levels of inequality. Given the above discussion on the ability of inequality measures to meet particular axioms, we will use the three group-decomposable measures. The Gini coefficient will also be calculated for comparability with inequality studies of other countries. The three group decomposable measures are defined as follows:  - 83 - N Y. Y.N Y. Y. Y./Y 1) Theil (T) = E - In { I E= I J T + i { J} ln N 1 y}N. N. N./N 2) Theil (L) = N n Y JN) NL WIN 2) j~~~=i 1 . 3. N (Y)N. N, 2 3) Log Variance (LV) = E [ln i YLV= {N LV j N [T j Y TYI i=1 NN where Y = total income of the population, Yi = income of individual i, Yj = total income of group j, Nj = number of people in group j, N = total population, ln Y = mean of ln (Yi) over the entire population, and ln Y. = mean of ln (Yi) over the population in group j. The terms to the right of the inequality sign in each formula depict the decomposable properties of the respective measures - the first term is a weighted average of the inequality found within each group (henceforth referred to as the within-group component) and the second term is the level of inequality that would prevail if each individual had the mean income (or mean of the log income in the case of the LV measure) of his or her respective group (the between-group component). The Gini coefficient can be graphically depicted as the area lying above the Lorenz curve divided by the entire area in the Lorenz diagram. Its mathematical formula is: 2NY i i 2 i 12 where il and i2 simply correspond to the respective summation signs.  LSMS WORKING PAPER SERIES (continued) No. 21. The Collection of Price Data for the Measurement of Living Standards. No. 22. Household Expenditure Surveys: Some Methodological Issues. No. 23. Collecting Panel Data in Developing Countries: Does it Make Sense? No. 24. Measuring and Analyzing Levels of Living in Developing Countries: An Annotated Questionnaire. No. 25. The Demand for Urban Housing in the Ivory Coast. No. 26. The C6te d'lvoire Living Standards Survey: Design and Implementation. No. 27. The Role of Employment and Earnings in Analyzing Levels of Living: A General Methodology with Applications to Malaysia and Thailand. No. 28. Analysis of Household Expenditures. No. 29. The Distribution of Welfare in the Republic of C6te d'lIvoire in 1985. No. 30.Quality, Quantity and Spatial Variation of Price: Estimating Price Elasticities from Cross- sectional Data. No. 31. Financing the Health Sector in Peru. No. 32. Peru Informal Sector, Labor Markets, and Returns to Education. No. 33. Wage Determinants in C6te d'lvoire. No. 34. Guidelines for Adapting the LSMS Living Standards Questionnaires to Local Conditions. No. 35. The Demand for Medical Care in Developing Countries: Quantity Rationing in Rural C6te dtivoire. No. 36. Labor Market Activity in C6te dtlivoire and Peru. No. 37. Health Care Financing and the Demand for Medical Care. No. 38. Wage Determinants and School Attainment Among Men in Peru. No. 39. The Allocation of Goods within the Household: Adults, Children, and Gender. No. 40.Households, Communities and Preschool Children's Nutrition Outcomes: Evidence from Rural Cote d'lvoire. No. 41. Public-Private Sector Wage Differentials in Peru: 1985-86. No. 42. The Distribution of Welfare in Peru in 1985-86.  The World Bank Headquarters European Office Tokyo Office U 1818 H Street, N.W. 66, avenue d'Iena Kokusai Building Washington, D.C. 20433, U.S.A. 75116 Paris, France 1-1, Marunouchi 3-chome Chiyoda-ku, Tokyo 100, Japan Telephone: (202) 477-1234 Telephone: (1) 40.69.30.00 Facsimile: (202) 477-6391 Facsimile: (1) 40.69.30.66 Telephone: (3) 3214-5001 Telex: WUI 64145 WORLDBANK Telex: 640651 Facsimile: (3) 3214-3657 RCA 248423 WORLDBK Telex: 26838 Cable Address: INTBAFRAD WASHINGTONDC