LSM -29 MAY 1988 Living Standards Measurement Study Working Paper No. 29 The Distribution of Welfare in Cote d'Ivoire in 1985 PaMtG LSMS Working Papers No. 1 Living Standards Surveys in Developing Cou7ntries 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 Measuremnenit 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 Philippinies No. 6 Household Survey Experience in Africa No. 7 Measurement of Welfare: Theory and Practical Guidelines No. 8 Employment Datafor the Measuremenit of Living Standards No. 9 Income and Expenditure Surveys in Developing Counltries: Sample Design and Execution No. 10 Reflections on the LSMS Group Meeting No. 11 Thiree Essays on a Sri Lanka Household Survey No. 12 Thze ECIEL Study of Househiold Income and Consumption in Urbani Latin America: An Analytical History No. 13 Nutrition and Health Status Indicators: Suggestions for Surveys of the Standard of Living in Developinig Countries No. 14 Chiild Schooling and the Measuremenit 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 Measuremnentt Study No. 19 The Conceptual Basis of Measures of Household Welfare and Their Implied Survey Data Requirements No. 20 Statistical Experimentation for Household Strveys: Two Case Studies of Hong Kong 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: AnI Annotated Questionnlaire No. 25 The Demandfor 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 C6te d'Ivoire in 1985 No. 30 Quality, Quantity, and Spatial Variation of Price: Estimatinig Price Elasticities from Cross-Sectional Data No. 31 Financing the Health Sector in Peru No. 32 Informal Sector, Labor Markets, and Retumrs to Education in Peru No. 33 Wage Determinants in C6te d'Ivoire No. 34 Gutidelines for Adapting the LSMS Livinig Standards Questioninaires to Local Conditions No. 35 The Demanid for Medical Care in Developing Counltries: Quanitity Rationinlg in Rural Cdte d'lvoire (List continues on the inside back cover) The Distribution of Welfare in C6te d'Ivoire in 1985 The Living Standards Measurement Study 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 statistical offices in developing countries. Its goal is to foster increased use of household data as a basis for policy decisionmaking. 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 govemment 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 LSMS 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. LSMS Working Paper Number 29 The Distribution of Welfare in Cote d'lvoire in 1985 Paul Glewwe The World Bank Washington, D.C., USA. Copyright (© 1988 The International Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing May 1988 This is a working paper published informally by the World Bank. To present the results of research with the least possible delay, the typescript has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. Any maps that accompany the text have been prepared solely for the convenience of readers; the designations and presentation of material in them do not imply the expression of any opinion whatsoever on the part of the World Bank, its affiliates, or its Board or member countries concerning the legal status of any country, territory, city, or area or of the authorities thereof or concerning the delimitation of its boundaries or its national affiliation. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to Director, Publications Department at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to photocopy portions for classroom use is not required, though notification of such use having been made will be appreciated. The most recent World Bank publications are described in the catalog New Publications, a new edition of which is issued in the spring and fall of each year. The complete backlist of publications is shown in the annual Index of Publications, which contains an alphabetical title list and indexes of subjects, authors, and countries and regions; it is of value principally to libraries and institutional purchasers. The latest edition of each of these is available free of charge from the Publications Sales Unit, Department F, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'1ena, 75116 Paris, France. Paul Glewwe was an economist in the Welfare and Human Resources Division of the World Banks Population and Human Resources Department when this paper was written. Library of Congress Cataloging-in-Publication Data Glewwe, Paul, 1958- The distribution of welfare in Cote d'Ivoire in 1985 / Paul Glewwe. p. cm. -- (LSMS working paper, ISSN 0253-4517 ; no. 29) Bibliography: p. ISBN 0-8213-1053-4 1. Cost and standard of living--Ivory Coast. 2. Consumption (Economics)--Ivory Coast. 3. Households--Ivory Coast. I. Title. II. Series. HD7067.G55 1988 339.2'09666'8--dc19 88-14073 v ABSTRACT This paper examines the distribution of welfare in Cote d'Ivoire (Ivory Coast) in 1985 as measured by per capita consumption expenditures. The data employed are from the 1985 C8te d'Ivoire Living Standards Survey. While it is meant to be primarily descriptive in nature, possible explanations of particular patterns in the distribution of welfare are offered at several points. The major findings are: 1. The urban population in Cote d'Ivoire is substantially better off than the rural population, and the poor in Cote d'Ivoire are overwhelmingly found in agricultural pursuits in rural areas. 2. There is a strong association between education and welfare, which highlights the importance of educational policies, particularly those affecting the school attendance of children. 3. Household composition, more specifically the proportion of household members who are children or who are employed, does not explain why some households are poor - poor households have a larger proportion of working members and a lower proportion of children than do other households. - vi - ACUOVEDG This report is based on The C8te d'lvoire Living Standards Survey, which was conducted in 1985 by the World Bank's Living Standards Unit and the Direction de la Statistique, Ministere de l'Economie et des Finances of the Republic of CMte d'Ivoire. I would like to thank the entire staff from the Bureau of Statistics, with special thanks to Meite Nediembo, Director of Statistics; Bakary Daho, Project Director; Idrissa Quattara, Deputy Project Director; Roch Sopri, Assistant Deputy Project Director; and Kouakou Chia B16, Computer Programming Specialist. Finally I am indebted to Carmen Martinez for typing several drafts of this paper. - vii - Table of Contents Page I*Introduction .... ........................................................ 1 II. Consumption Data as an Indicator of Welfare........................3 III. The Distribution of Consumption in C8te d'Ivoire....8...............8 A. The Distribution of Consumption Expenditures by Population B. Characteristics of Households by Per Capita Expenditure Grus.......................................... ............ ........... 10 C. School Attendance, Housing and Ownership of Durables........... 19 IV. Inequality in Cote d'Ivoire ................................. ...... 27 A. Historical Comparison with Other Countries.*.....*.....*...*.*.27 B. Regional Inequality Decomposition ..............................28 C. Decomposition by Nationality and Ethnic Groups.................30 D. Decomposition by Educational Attainment............. .. ... .. .. . 33 V. Characteristics of the Poor in C8te d'Ivoiree ...........e........36 A. Poverty in C8te d'Ivoire - A National Profile.................36 B. Poverty in Urban C8te d'Ivoire .....................e........45 C. Poverty in Rural C8te d'Ivoire .53 VII. Conclusion ............................. 6 REFERENCES ...................... ........... ................ 65 Appendix A: The C8te d'Ivoire Living Standards Survey..................67 Appendix B: Food Consumption as a Welfare Measure......................75 Appendix C: Measurement of Inequality ..................................78 - viii - Tables Page Table 1: Welfare Indicators of C8te d'Ivoire and Sri Lanka...............7 Table 2: Distribution of Total Consumption by Welfare Deciles...........9 Table 3: Characteristics of Households by Quintiles ....................11 Table 4: Characteristics of Households by Regions,,*......................13 Table 5: Sex of Head of Households Within Regions, by Quintiles........ 16 Table 6: School Attendance by Welfare Quintiles ........................ 21 Table 7: Housing Characteristics in C8te d'Ivoire......................23 Table 8: Ownership of Durables by Quintiles and Regions................ 25 Table 9: Gini Coefficients on Per Capita Income .......................28 Table 10: Per Capita Expenditure Inequality-Decomposed by Region ....... 30 Table 11: Per Capita Expenditure Inequality-Decomposed by Nationality ....1E ....i ........i......r.....e..n.eualc...... 31 Table 12: Per Capita Expenditure Inequality-Decomposed by Ethnic Groups .,.**................................ . . . . oo..... **so..... ........ 32 Table 13: Per Capita Expenditure Inequality-Decomposed by Education Table 14: Household Size and Per Capita Consumption of the Poor........37 Table 15: Sex of Household Head Among the Poor .........................38 Table 16: Distribution of the Poor by Regions. ..........................38 Table 17: Distribution of the Poor by Ethnic Group.4 ...................39 Table 18: Distribution of the Poor by Employer and Occupations..........40 Table 19: Distribution of the Poor by Educational Level .............. .41 Table 20: Illness, School Attendance and Housing of the Poor........... 43 Table 21: Composition of Household Members Among the Poor ..............44 - ix - Table 22: Household Size and Per Capita Consumption of the Urban Poor...................................................... 46 Table 23: Sex of Household Head Among the Urban Poor...........e ..... 47 Table 24: Nationality and Ethnic Group Among the Urban Poor .............47 Table 25: Distribution of the Urban Poor by Employer and Occupation....49 Table 26: Education and Poverty in Urban Areas.........................50 Table 27: Housing Characteristics of the Urban Poor ...................51 Table 28: Household Composition Among the Urban Poor......... .... 52 Table 29: Household Size per Capita Consumption of the Rural Poor......54 Table 30: Sex of Household Head Among the Rural Poor..................... .55 Table 31: Nationality and Ethnic Group Among the Rural Poor..... .........56 Table 32: Distribution of the Rural Poor by Employer and Occupation....57 Table 33: Education and Poverty in Rural Areas......................... 58 Table 34: Housing Characteristics of the Rural Poor .....................59 Table 35: Household Composition Among the Rural Poor ...... 60 Table A.1. Regional Price Indices in C8te d'Ivoire, 1985........... .....71 Table A.2: Welfare Deciles According to Different Welfare Measures ...... 72 Table A.3: Composition of Total Consumption in C8te d'Ivoire, 1985 ...... 74 Table B.1: Distribution of Food Consumption by Food Consumption Decies ......................................................... 76 I. Introduction The distribution of the benefits of economic growth in developing countries is a subject which has received much attention since at least the early 1970's. It is generally accepted that the success of any economic policy is in part measured by the extent to which it promotes equity. However, the information necessary to measure the impact of recent policies on the distribution of goods and services is often lacking in the developing world. This is particularly the case in sub-Saharan Africa, which contains many of the poorest countries and which has suffered severe setbacks since the 1960's. This paper examines the distribution of welfare in The Republic of Cate d'Ivoire (Ivory Coast) in 1985. Although C8te d'Ivoire is not typical of African countries in that its economy has grown substantially since its independence in 1960, its success recommends its policies to other African countries and in this sense an analysis of the distribution of welfare under these policies is relevant to the future of all African countries that wish to emulate its prosperity. In addition, several characteristics of C8te d'Ivoire, such as a large agro-export industry, a substantial variation in agroclimatic zones within the country, and a mixture of different ethnic and religious groups, are common to many sub-Saharan African countries. Given the high quality of data from C8te d'Ivoire and the dearth of data from other African countries, an examination of the distribution of welfare in C8te d'Ivoire is a first step towards investigating the distribution of welfare in sub-Saharan Africa. It also has implications for the debate on growth vs. equity among development economists (see Bhagwati, 1985, and Chenery, 1974). -2- The distribution of welfare in Cote d'Ivoire is also of interest in its own right given recent developments in the Ivorian economy. After 20 years of rapid economic growth (see den Tuinder, 1978), total GDP began to decline in 1980. By 1985, private consumption expenditures per capita had fallen 26.0X relative to 1980. Although preliminary data point to future recovery, the distributional aspects of these recent developments have never been analyzed. An examination of the distribution of welfare in 1985 is important for understanding the welfare effects of C6te d'Ivoire's recent economic decline. The organization of this paper is as follows: Section II discusses the theoretical basis of consumption expenditures as a measure of welfare. The next section presents a detailed description of the distribution of consumption expenditures in C6te d'Ivoire. The fourth section analyzes the distribution of welfare using summary measures of inequality, and the fifth section provides a profile of the poor in Cote d'Ivoire. A final section concludes the paper. -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 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; 2) does not imply any particular cardinalization of the common individual utility function. This can be done using "money-metric" utility, which is the amount of money required (given a set of prices and the assumption of utility maximization) to attain a specified level of utility. The use of money-metric utility to compare different utility curves assumes that total consumption, as opposed to total income, represents the welfare levels of individuals.- 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,2/ 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 nonfood 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 if one examines only food expenditure data; and 4) Food expenditure data are thought to be more accurate than those on 1/ 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. 2/ 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 inkind income is given monetary value so that expenditures include consumption of these items. -5- non-food expenditures. This paper will 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 in the previous paragraph 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 this phenomenon is the estimation of "adult equivalent 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 equivalent scales for C8te d'Ivoire in 1985 is no small task and lies beyond the scope of this paper. However, - 6 - one can reasonably assume that adult equivalent scales 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 equivalents 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. For 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 Cote d'Ivoire, 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 two very different countries - Cote d'Ivoire and Sri Lanka. Table 1 presents some figures relevant for welfare comparisons from Cote d'Ivoire and Sri Lanka. They are given here for illustrative purposes -7- only. The figures show that Sri Lanka does relatively well in non-consumption measures of welfare, such as life expectancy, infant mortality and the adult literacy rate, while C6te d'Ivoire has a substantially higher GNP per capita, which is clearly a consumption-based measure. There is no clear method tomake welfare judgments across these two countries and no attempt will be made here. However, it should be borne in mind that the welfare comparisons in C6te d'Ivoire 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. Table 1: Welfare Indicators of Cote d'Ivoire and Sri Lanka Welfare Measure Cote d'Ivoire Sri Lanka CNP per Capita (1984) $610 $360 Life Expectancy (1984) 52 70 Infant Mortality Rate (1984) 106 37 Adult Literacy Rate (1980) 35% 85% Source: World Bank (1983, 1986). This completes the discussion of the framework used in this paper for examining the distribution of welfare in Cote d'Ivoire. 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 C6te d'Ivoire. -8- III. The Distribution of Consumption in C6te d'Ivoire This section provides descriptive data on the distribution of welfare in C6te d'Ivoire as measured by 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 regions in Cote d'Ivoire. The third subsection examines school attendance, housing and ownership of durable goods, by welfare levels. 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 lOX of the population (as measured by adjusted per capita consumption expenditures - see Appendix A for details), Decile 2 contains the next poorest 10%, etc., and Decile 10 contains the "wealthiest" 10% 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, but there is little difference in the distribution of welfare using either method. 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% of the population receive 14-15% of total consumption while the wealthiest 20% receive 49-51%. -9- Table 2: Distribution of Total Consumption by Welfare Deciles Mean Per Capita % of Total Food Share (Z) Annual Expenditures Expenditures of Total Expenditures (CFAF x 1000) in Cote d'Ivoire within each Decile Decile Adjusted Unadjusted Adjusted Unadjusted Adjusted Unadjusted 1 69.5 42.5 1.98 1.96 0.711 0.709 2 114.5 68.9 3.26 3.18 0.660 0.680 3 154.0 93.1 4.39 4.30 0.659 0.635 4 189.0 113.2 5.39 5.23 0.622 0.624 5 227.0 135.5 6.47 6.26 0.596 0.609 6 271.4 163.0 7.73 7.53 0.591 0.595 7 334.3 199.5 9.53 9.21 0.569 0.550 8 415.4 255.6 11.84 11.80 0.499 0.506 9 556.9 344.8 15.87 15.93 0.478 0.488 10 1177.0 749.3 33.55 34.60 0.352 0.348 All.C6te d'Ivoire 350.9 216.5 100.00 100.0 0.489 0.487 A comparison of the data in Table 2 with comparable data from other countries would provide information on differences in the distribution of welfare between C8te d'Ivoire and other countries. Unfortunately, comparable data are almost non-existent. However, data on income inequality are available and will be discussed in Section IV. - 10 - B. Characteristics of Households by Per Capita Expenditure Groups The previous subsection gives some information on the distribution of welfare in Cote d'Ivoire but it reveals almost nothing about the welfare levels of different population groups. In this subsection the welfare levels of different groups within the Ivorian population will be compared. Table 3 presents tabulations which divide the total population into five welfare groups (quintiles). The first quintile contains the poorest 20% of the population as defined by adjusted (i.e. smaller weights for children) per capita expenditures, the second contains the next poorest 20%, etc., and the fifth contains the wealthiest 20% of the Ivorian 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) Nationality of the head of household; 4) Ethnic group (tribe) of the household; 5) Type of employer of the head of household; 6) Occupation of the head of household; and 7) 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. The composition of quintiles according to regions reveals that Abidjan (the colonial capital and largest city) is home to relatively well off households. Other urban areas are also relatively wealthly. The poorest region is the Savannah, followed by the densely populated East Forest, region. The West Forest region, which has relatively recent settlements and is less heavily populated than the East Forest, contains households that are - 11 - Table 3: Characteristics of Households by Quintiles Breakdown within Quintiles All C8te Quintile Quintile Quintile Quintile Quintile Mean Expend. per Capita (CFAFx1000 per year) Characteristic d'lvoire 1 2 3 4 5 Adjusted Unadjusted Region Abidjan 18.8 3.3 5.2 13.4 29.2 42.8 633.8 402.3 Other Urban 22.4 7.0 18.1 28.2 27.1 31.8 412.7 253.3 West Forest 15.2 8.8 19.6 21.9 14.9 11.0 296.0 182.2 East Forest 24.7 35.2 35.4 22.5 19.9 10.6 246.2 146.5 Savannah 18.9 45.7 21.8 14.1 9.0 3.9 177.7 107.7 Sex of Head Male 90.0 93.7 90.1 89.0 88.2 89.1 347.7 212.7 Female 10.0 6.3 9.9 11.0 11.8 10.9 379.5 250.9 Nationality Ivorian 85.7 89.3 84.6 85.4 82.4 86.9 350.6 216.5 Burkino 6.2 5.6 '.2 7.8 6.8 3.7 287.8 177.6 Mali 4.0 2.5 5.5 2.9 6.0 3.1 359.3 203.5 Guinea 1.9 2.1 1.8 2.3 2.0 1.2 297.2 180.9 Ghana 0.2 0.0 0.4 0.2 0.4 0.2 423.7 374.8 Other African 1.8 0.5 0.2 1.3 2.4 4.5 582.6 374.3 Other 0.1 0.0 0.3 0.0 0.0 0.4 770.6 543.9 Ethnic Group Akan 38.1 37.6 42.6 34.7 35.0 40.9 354.6 217.2 Krou 13.9 7.9 11.6 20.3 16.9 12.8 367.5 228.5 Mande (North) 13.0 17.8 12.1 9.0 14.4 11.4 338.5 208.6 Mande (South) 11.2 6.2 11.3 13.8 9.9 14.6 388.9 244.2 Voltaic 9.5 21.0 6.5 8.8 5.6 5.7 244.7 153.0 Non-lvorian 14.1 9.2 15.4 13.4 18.2 14.1 378.8 231.5 Other 0.3 0.3 0.5 0.0 0.0 0.5 290.9 213.4 - 12 - Table 3: Characteristics of Households by Quintiles (Continued) Breakdown with Quintiles All C6te Quintile Quintile Quintile Quintile Quintile Mean Expend. per Capita (CFAFx1000 per year) Characteristic d'lvoire 1 2 3 4 5 Adjusted Unadjusted Employer of Head None 1.1 0.2 0.3 1.0 1.5 2.6 516.6 362.9 Government 11.4 1.1 3.0 9.0 13.6 30.3 665.0 409.8 Parastatal 1.3 0.0 0.0 1.0 1.9 3.4 550.3 346.8 Private 12.5 2.2 5.5 15.6 14.5 24.6 511.5 318.5 Self-employed 73.8 96.6 91.2 73.4 68.6 39.2 269.5 165.1 Occupation of Head None 0.9 0.2 0.3 0.5 1.5 2.2 553.4 391.8 Agricultural 60.3 90.1 81.1 60.8 46.2 23.3 232.4 142.1 Sales/Services 20.5 6.0 12.0 23.4 32.7 28.2 430.0 263.0 Constr./Mfg. 4.7 2.0 3.3 6.8 5.7 5.9 385.6 237.7 White Collar 11.7 0.5 3.0 7.5 11.9 35.8 773.8 485.4 Other 1.9 1.3 0.2 1.2 2.1 4.9 467.9 284.6 Highest Grade Complete by Head None 65.3 86.2 80.7 67.9 58.9 32.9 254.7 156.8 Elem. (1-7) 18.5 12.8 17.0 21.0 24.5 17.2 329.5 203.5 Middle (8-11) 9.7 0.9 2.3 8.2 13.2 24.1 590.8 350.8 HS (12-14) 4.0 0.0 0.0 2.3 3.1 14.8 828.4 526.4 University 2.4 0.2 0.0 0.5 0.4 11.1 1348.9 872.5 Note: 1. All figures are In terms of the percentage of the population living in households with the relevant characteristics. 2. If the head of the household was not employed, characteristics are taken from the household member who earned the highest income (money and in-kind) from a paid occupation. (Self-employment is a paid occupation if some of the product is sold). - 13 - Table 4: Characteristics of Households by Regions Breakdown by Regions All Cote Other West East Characteristic d'Ivoire Abidjan Urban Forest Forest Savannah Sex of Head Male 90.0 82.0 87.3 95.2 91.5 95.2 Female 10.0 18.0 12.7 4.8 8.6 4.8 Nationality Ivorian 85.7 17.2 84.5 81.0 86.9 98.1 Burkino 6.2 6.7 5.7 7.6 9.9 0.4 Mali 4.0 9.2 4.1 5.0 2.1 0.4 Guinea 1.9 2.4 1.6 4.9 0.4 1.2 Ghana 0.2 0.2 0.8 0.0 0.1 0.0 Other Africa 1.8 4.2 3.3 1.1 0.4 0.0 Other 0.3 0.1 0.0 0.4 0.3 0.0 Ethnic Group Akan 38.1 31.0 38.2 1.8 68.0 35.4 Krou 13.9 10.9 12.7 45.3 8.5 0.0 Mande (North) 13.0 14.8 16.5 7.5 3.5 23.7 Mande (South) 11.2 10.9 5.6 26.7 6.8 11.4 Votaic 9.5 7.9 11.3 0.7 0.6 27.9 Non-Ivorian 14.1 24.5 15.3 18.1 12.0 1.7 Other 0.3 0.0 0.4 0.0 0.7 0.0 Employer of Head None 1.1 2.4 1.2 0.8 0.5 0.8 Government 11.4 22.4 26.9 0.5 3.4 1.3 Parastatal 1.3 4.2 1.6 0.0 0.3 0.2 Private 12.5 39.0 15.9 2.8 2.9 2.3 Self-Employment 73.8 32.0 54.4 96.0 92.9 95.5 - 14 - Table 4: Characteristics of Households by Regions (Continued) Breakdown by Regions All CBte Other West East Characteristic dtIvoire Abidjan Urban Forest Forest Savannah Occupation of Head None 0.9 2.4 0.9 0.8 0.5 0.2 Agricultural 60.3 1.7 31.4 87.1 88.5 94.3 Sales/Services 20.5 45.4 39.5 8.4 5.1 2.9 Constr./Mfg. 4.7 13.2 6.5 1.1 2.1 0.7 White Col./Mgr. 11.7 35.2 17.5 0.6 3.4 1.5 Other 1.9 2.2 4.3 2.1 0.3 0.5 Highest Grade Completed by Head None 65.3 36.5 51.0 70.5 75.5 93.5 Elem. (1-7) 18.5 21.8 21.8 24.9 20.0 4.2 Middle (8-11) 9.7 23.7 14.8 4.6 3.4 2.3 H. Sch.(12-14) 4.0 10.1 8.2 0.0 1.1 0.1 University 2.4 8.0 4.2 0.0 0.0 0.0 Per Capita Expenditures (CFAFx1000) Adjusted 350.9 633.8 412.7 296.0 246.2 177.7 Unadjusted 216.5 402.3 253.3 182.2 146.5 107.7 Note: 1. All figures are in terms of the percentage of the population living in households with the relevant characteristics. 2. If the head of the household was not employed, characteristics are taken from the household member who earned the highest income (money and in-kind) from a paid occupation. (Self-employment is a paid occupation if some of the product is sold). - 15 - mostly in the middle quintiles. Of course, one should not infer that movement form rural to urban areas 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 that high levels of welfare in urban areas reflect their relatively well educated residents. Perhaps the most surprising information in Table 3 is that households headed by women have higher welfare levels than those headed by men. This is in contrast with many other developed and developing countries, where female- headed households are most often poorer than those headed by males. Part of the explanation of the situation in C8te d'Ivoire lies in the fact that female-headed households are disproportionately found in Abidjan and other urban areas and are relatively rare in the rural areas. If one examines quintiles within urban and rural areas female-headed households are slightly poorer than average in urban areas but not necessarily so in rural areas, particularly the East Forest region. This information is given in Table 5. In any case, it seems that female-headed households are not significantly worse off than those headed by males in C6te d'Ivoire. Differences in quintile composition by nationality show few patterns. Ivorians, who comprise 85.7% of the surveyed population, I, are fairly evenly spread over all welfare levels. Citizens of Burkino Faso, Mali and Guinea are over-represented in the lower quintiles while Ghanaians and other Africans (who are predominantly found in urban areas - see Table 4) are 1/ There is some evidence that the non-Ivorian population is under-sampled in the CILSS. Table 5: Sex of Head of Household Within Regions, by Quintiles Mean Exp. Per Capita Sex of Entire (CFAFxlOOO per year) Region Head Region Quintile I Quintile 2 Quintile 3 Quintile 4 quintile 5 Adjusted Unadjusted Urban Male 80.8 71.7 79.3 82.8 86.5 81.3 658.3 412.5 Female 19.2 19.2 20.7 17.2 13.5 18.7 522.6 355.7 Other Urban Male 86.2 86.0 80.8 86.6 87.5 88.2 419.7 255.1 Female 13.8 14.0 19.2 13.4 12.5 11.8 364.5 240.9 F' West Forest Male 93.7 90.9 95.2 91.7 98.0 92.5 296.0 181.2 Female 6.3 9.1 4.8 8.3 2.0 7.6 291.0 201.6 East Forest Male 90.3 93.9 98.5 87.0 90.8 83.7 241.5 143.4 Female 9.7 6.2 1.5 13.0 9.2 16.3 296.9 179.0 Savannah Male 92.7 90.2 90.9 91.5 94.0 96.1 178.8 108.2 Female 7.3 9.8 10.0 8.5 6.0 4.0 156.1 97.5 - 17 - better off than the typical Ivorian. Yet Chanaians and other Africans amount to only 2% of the total population of Cote d'Ivoire. The distribution of welfare by ethnic groups reveals that most tribal groups have roughly the same distribution within quintiles, except that Voltaic households clearly have lower welfare levels on average. Much of this is due to the fact that Voltaics are found predominantly in the Savannah area (see Table 4), which is the poorest region of the Cote d'Ivoire. Given that the Savannah area has low levels of educational attainment (Table 4) it is possible that the low living standards of Voltaics are in part a result of low educational levels. The correlation between schooling and welfare is discussed below. As in many other developing countries, working for the government or for a large private firm 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. The same is true to the lesser extent of those living in households where the head works for parastatal firms (government-owned corporations) or for private firms. On the other hand, households headed by self-employed workers (most of whom are farmers) are predominant in the lower quintiles. One can see in Table 4 that the type of employer is strongly correlated with different regions - government, parastatal and private sector jobs are common in urban areas while the self-employed compose nearly the entire rural labor force. An interesting observation from Table 3 is that individuals living in households where the head (and all other family members - see note 2 of Table 3) was not working have higher welfare levels than most Ivorian households. It seems that unemployment is more common among the well to do than among the poorer - 18 - households in C6te d'Ivoire. Of course, some of these households may be entirely composed of retired persons living on pensions or other sources of transfer income. 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 in agricultural pursuits, while households in which the head works in construction and manufacturing 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 agricultural while urban areas have many households in which the head was in sales and services, construction and manufacturing, 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 the Ivory Coast is no exception (van der Gaag and Vijverberg, 1986). Eighty-six percent of the poorest 20% of the population lived in households where the head had no education at all, while only 1.1% of these poor households had a head with a middle school, high school or university education. In contrast, only 32.9% of the wealthiest 20% of the population lived in "no education" households, which compares to a figure of 50.0 percent for households where the head had a middle school or higher education. One can see in Table 4 that households with better educated heads are found in urban areas, particularly Abidjan, while less educated heads predominate in rural areas, particularly the Savannah. - 19 - 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 in Table 3. In many households the head may be an older family member who is no longer the main 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 C6te d'Ivoire. This correlation between education and welfare has important implications for policy, particularly in terms of the distributional impact. The World Bank (1986) reports that 40% of Ivorian children of primary school age have no access to any form of education, while at the same time large amounts of funds are spent on the few students in higher education. A reorientation of government policies toward a more equitable educational system would seem to be in order if the benefits of economic growth are to be more evenly distributed. C. School Attendance, Housing and Ownership of Durables The information collected in the C6te d'Ivoire 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. - 20 - 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 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. This data is given in Table 6. Before discussing the data in Table 6 it is useful to make three observations. 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 a 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 patterns in school attendance. Second, the school attendance figures given in Table 6 are artificially low because children who were on summer vacation were not counted as attending school. Third, there is evidence in Cate d'Ivoire and other African countries that children begin school at a relatively late age (thus lowering the figure for ages 6-10) and often repeat grades (thus raising the figure for ages 11-15). In C6te d'Ivoire 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 almost always the case that the poorest quintile has lower school attendance than average, - 21 - Table 6: School Attendance by Welfare Quintiles Entire Quintiles Region 1 2 3 4 5 All COte drIvoire Ages 6-10 50.5 32.9 45.2 51.6 57.2 64.0 Ages 11-15 49.6 32.4 43.8 53.5 55.1 58.1 Abidjan Ages 6-10 54.0 40.0 42.5 60.3 63.8 64.2 Ages 11-15 49.3 35.0 51.2 51.5 48.9 56.8 Other Urban Ages 6-10 63.0 38.5 61.7 68.9 73.2 66.0 Ages 11-15 59.4 54.4 61.2 58.6 63.0 58.7 West Forest Ages 6-10 49.7 54.4 50.7 50.8 43.7 49.4 Ages 11-15 46.9 36.0 39.7 59.2 44.2 59.1 East Forest Ages 6-10 55.5 33.6 55.7 59.8 59.8 65.0 Ages 11-15 54.8 38.1 53.3 60.0 57.0 67.4 Savannah Ages 6-10 26.2 13.9 21.1 32.9 33.0 28.4 Ages 11-15 25.3 21.1 25.0 27.1 14.9 38.0 Note: Within each of the five sectors quintiles are defined with respect to those sectors (e.g. poorest 20% of all Abidjan residents, not the poorest 20% of all Ivorians who also happen to live in Abidjan). - 22 - while that of the wealthiest quintile is higher than average, the absolute differences are not dramatically large. 1- Differences in school attendance are much more apparent between regions, particularly the low school attendance in the Savannah. It is also curious that the Other Urban and East Forest regions have higher school attendance than does Abidjan. Given that school attendance is lowest in the Savannah, that the poorest people in C8te d'Ivoire are found in that region, and the findings in the previous subsection that education is strongly correlated with household welfare, the outlook for the welfare of Ivorians in the Savannah region is not good. 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 40% of primary school children do not have access to facilities and the low correlation between welfare levels and school attendance within regions suggest that the problems is a lack of good facilities. As pointed out above, this has direct policy implications for the Ivorian educational system. Housing conditions are important measures of welfare themselves, and they also have indirect implications for welfare, particularly in the area of health. Table 7 gives information on sources of drinking water, sources of lighting and the type of toilet used by welfare quintiles and by regions. In Cote d'Ivoire as a whole, there are a variety of sources of drinking water, some of which are clearly preferred by wealthier households, / It is possible that wealthier households send some children away to school, so that those remaining at home underestimate the fraction attending school. But this is most likely to occur among older children. - 23 - Table 7: Housing Characteristics in C6te d'Ivoire All Quintiles C6te d'Ivoire 1 2 3 4 5 Source of Drinking Water Indoor Faucet 14.1 1.6 1.3 7.8 15.3 44.5 Outdoor Faucet 10.2 0.8 6.0 12.1 14.5 17.6 Water Vendor 10.4 3.2 8.1 10.8 18.6 11.4 Pump Well 27.7 52.2 34.7 29.3 13.5 8.8 Other Well 26.3 26.9 32.8 29.1 30.1 12.8 Rain, Lake, River 10.8 15.0 16.8 10.4 7.5 4.3 Other 0.5 0.4 0.4 0.5 0.6 0.7 Source of Lighting Electric 41.6 13.2 21.3 43.0 55.8 74.8 Gas 0.2 0.3 0.0 0.2 0.4 0.0 Candles/Oil Lamps 58.2 86.5 78.7 56.8 43.9 25.2 Type of Toilet Flush 22.0 3.4 4.2 17.6 27.9 57.0 Pit 41.8 40.4 45.8 43.2 48.6 30.8 None 36.3 56.2 50.1 39.2 23.5 12.2 All Other West East C6te d'Ivoire Abidjan Urban Forest Forest Savannah Source of Drinking Water Indoor Faucet 14.1 46.9 21.7 0.0 1.2 0.6 Outdoor Faucet 10.2 8.6 27.6 0.0 8.7 1.3 Water Vendor 10.4 43.9 8.3 0.0 1.3 0.0 Pump Well 27.7 0.0 2.3 37.4 43.4 57.1 Other Well 26.3 0.7 38.6 48.6 28.6 16.4 Rain, Lake, River 10.8 0.0 0.8 14.0 15.5 24.7 Other 0.5 0.0 0.7 0.0 1.4 0.0 Source of Lighting Electric 41.6 84.8 78.7 0.9 20.9 14.6 Gas 0.2 0.2 0.2 0.0 0.3 0.2 Candles/Oil Lamps 58.2 15.0 21.1 99.1 78.8 85.2 Type of Toilet Flush 22.0 81.4 26.9 0.0 1.7 1.5 Pit 41.8 15.7 66.6 46.7 53.0 19.5 None 36.3 2.9 6.6 53.3 45.3 79.1 - 24 - such as indoor and outdoor faucets, and others which predominate among poorer households, such as well water and water from natural sources. In Abidjan indoor faucets and water vendors are the most common water sources, while outdoor faucets and (non-pump) wells are common in other urban areas. Residents of rural areas get most of their water from wells and natural sources. In many cases households may have little choice regarding their water supply, so that the patterns by quintiles may to a large extent reflect the region in which the household resides. The two main sources of lighting, electricity and candles or oil 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 candles and oil lamps. As with water, the type of lighting may often be out of the household's control. Finally, the type of toilet in the household displays expected patterns. Poorer households have a pit toilet or no toilet at all, while wealthier households often have flush toilets. Flush toilets are predominant in Abidjan, while pit toilets are the most common in other urban areas and either pit toilet or no toilets at all are 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 8. - 25 - Most of the durable goods in Table 8 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 exception is bicycles, which are found among the poorer households because they are most often found in rural areas, where the poorer households reside. One is to some extent surprised by the wide distribution of relatively expensive goods such as refrigerators, televisions, tape players Table 8: Ownership of Durables by Quintiles and Regions All Quintiles C8te d'Ivoire 1 2 3 4 5 Sewing Machine 28.3 13.2 23.1 27.2 38.1 40.2 Refrigerator 27.1 5.9 9.8 19.8 38.9 61.1 Fan 26.1 5.5 10.0 23.2 38.6 53.1 Radio 26.8 19.1 28.7 23.1 28.3 34.6 Tape Player 52.2 38.1 42.9 53.6 54.7 71.6 Television 34.5 9.3 15.4 29.9 59.2 68.7 Bicycle 36.3 55.7 49.1 31.2 25.3 20.4 Automobile 12.0 2.1 3.1 4.0 13.5 37.2 All Other West East Cote d'Ivoire Abidjan Urban Forest Forest Savannah Sewing Machine 28.3 40.5 44.6 12.4 29.8 7.8 Refrigerator 27.1 58.5 51.2 1.9 16.6 1.2 Fan 26.1 51.9 55.9 0.7 12.6 3.0 Radio 26.8 34.1 30.3 20.7 32.8 12.4 Tape Player 52.2 59.5 65.7 48.7 46.7 38.7 Television 34.5 74.8 61.7 9.9 15.7 6.6 Bicycle 36.3 4.6 24.0 28.2 48.5 73.3 Automobile 12.0 26.2 18.0 3.6 9.6 0.5 Note: All figures are the percentage of the total population living in households possessing the durable good. - 26 - and automobiles. In particular, tape players are much more common than radios, which are presumably less expensive. Even the poorest 20% of the population have purchased a substantial number of tape players, which seem to be quite popular even in the Savannah region. - 27 - IV. Inequality in C8te d'Ivoire 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 not deemed to be appropriate 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 C6te d'Ivoire relative to that of other developing countries. Data in Table 9 are taken from a variety of sources. The figures are not always comparable but they do convey an approximate notion of how C8te d'Ivoire compares with 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 9 suggest that Cote d'Ivoire had a relatively equitable distribution of income in 1959 compared to other African countries and an "average" level of equity relative to non-African countries. This is - 28 - Table 9: Gini Coefficients on Per Capita Income African Gini Cini Countries Coefficients Non-African Countries Coefficients Chad (1958) 0.3687 Costa Rica (1971) 0.4757 C6te d'Ivoire (1959) 0.4556 Dominican Republic (1969) 0.4550 Dahomey (1959) 0.4675 El Salvador (1969) 0.4653 Gabon (1960) 0.6899 Honduras (1967-68) 0.5658 Madagascar (1960) 0.5618 India (1964-65) 0.3957 Senegal (1960) 0.5874 Malaysia (1970) 0.5045 South Africa (1965) 0.5813 Peru (1961) 0.6664 Tanzania (1967) 0.5282 South Korea (1970) 0.4065 Tunisia (1961) 0.5094 Source: Jain (1975). consistent with other observers perceptions of C6te d'Ivoire (e.g., den Tuinder, 1978). However, these figures must be approached with caution because: 1. They are based on income rather than expenditure data; 2. They are from different years; 3. They do not have identical concepts of income. Despite these caveats, they are of some use in putting the following analysis into perspective. B. Regional Inequality Decompositions Measurement of expenditure inequality will be done using four comonly 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 using group-decomposable inequality measures is that they can be - 29 - 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 C8te d'Ivoire due to differences in average (adjusted) per capita expenditures among the five regions (Abidjan, Other Urban, West Forest, East Forest, and Savannah). This allows one to determine the potential effect on overall inequality of policies aimed at reducing differences among these regions. If the between-group component is small (say, less than 5%) 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 recommend from the equity perspective. On the other hand, relatively large between-group contributions (say 20% or higher) reveal possibilities for promoting greater equity in Cate d'Ivoire. In this section, groupings will be done according to regions, ethnic groups, nationality and the education of the head of household. Table 10 gives the decomposition of inequality in Cate d'Ivoire when it is divided into five 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. First, inequality in C6te d'Ivoire as a whole is greater than it is in any given region, which implies a substantial between-group contribution to overall inequality. Second, inequality is always highest in Abidjan, which is not surprising given its concentration of the wealthiest households in Cate d'Ivoire. Third, inequality is lowest in the West Forest region. This region has been recently "developed" and settled, so that differences in incomes that arise over time may not have been fully played out. - 30 - Table 10: Per Capita Expenditure Inequality-Decomposed by Region Mean % of Total Adjusted Exp. Log Region Population (CFAFx1000 yr.) Gini Theil T Theil L Variance Abidjan 18.8 633.8 0.4109 0.3108 0.2809 0.5035 Other Urban 22.4 412.7 0.3601 0.2297 0.2135 0.3997 West Forest 15.2 296.0 0.3172 0.1773 0.1703 0.3327 East Forest 24.7 246.2 0.3752 0.2625 0.2366 0.4341 Savannah 18.9 177.7 0.3800 0.2515 0.2424 0.4683 C6te d'Ivoire 100.0 350.9 0.4350 0.3530 0.3254 0.6079 Between-Group Contribution 0.0948 0.0946 0.1775 (Z) 26.9% 29.1% 29.2% The most important finding in Table 10 is that the between-group component is relatively large. This implies that policies aimed at equalizing mean incomes (and thus expenditures) among these different regions have significant potential to reduce overall inequality in C6te d'Ivoire. However, the form such policies may take is as yet unclear. This issue will be discussed below. C. Decomposition by Nationality and Ethnic Groups C6te d'Ivoire has a substantial population of expatriates from neighboring countries, other African countries, and Non-African countries (particularly France). Recent policies have adopted "Ivorianization," the replacement of foreign workers by Ivorians, as worthy goal. One can divide these foreign workers into two categories (see den Tuinder, 1978), highly skilled workers (primarily non-African) whose skills are rarely found among Ivorians, and low-skilled workers (Africans from neighboring countries) who - 31 - accept low-paid jobs which many Ivorians do not want. The 1985 CILSS does not include French and other non-African expatriates in its sample but does include all Africans living in C6te d'Ivoire. One can check to see whether the existence of other Africans has any impact on the distribution of welfare among the entire population, Ivorian and non-Ivorian. Table 11 presents inequality decompositions by nationality. Note that 85Z of the population is Ivorian which already implies a low between-group contribution to overall inequality since most of the population lies within this group. Table 11: Per Capita Expenditure Inequality - Decomposed by Nationality Mean % of Total Adjusted Exp. Log Nationality Population (CFAFxlOOO yr.) Theil T Theil L Variance Ivorian 85.7 350.6 0.3640 0.3361 0.6289 Burkino 6.2 287.8 0.2297 0.2154 0.4099 Mali 4.0 359.3 0.2825 0.2488 0.4335 Guinea 1.9 297.2 0.2316 0.2176 0.4107 Ghana 0.2 423.7 0.2893 0.2480 0.3908 Other 1.9 596.7 0.2478 0.2326 0.4447 C8te d'Ivoire 100.0 350.9 0.3530 0.3254 0.6079 Between- Group Cont. 0.0053 0.0047 0.0087 (Z) 1.5% 1.4% 1.4% It is clear from Table 11 that nationwide figures are very similar to those given for Ivorians only - mean adjusted per capita expenditure is virtually identical and inequality measures are quite similar. Of the non- Ivorian groups, immigrants from Burkino Faso, Mali and Guinea are (on average) poorer than Ivorians while those from Ghana and other African countries are - 32 - wealthier. These differences reflect the fact that the latter are predominantly found in urban areas while the former are more common in rural areas (see Table 4). In summary, one finds that differences in nationality contribute very little to overall inequality. One characteristic of Cote d'Ivoire which is common to most sub- Saharan African countries is that there are several distinct ethnic (or tribal) groups. Rivalries between these groups is common problem in Africa, and at times one ethnic group may be able to secure for itself a privileged position which is reflected by its level of economic welfare. Table 12 decomposes overall inequality in C8te d'Ivoire into its between- and within- group components when the population is divided along ethnic lines. Table 12: Per Capita Expenditure Inequality - Decomposed by Ethnic Groups Mean % of Pop- Adjusted Exp. Log Ethnic Group ulation (CFAFxlOOO yr.) Theil T Theil L Variance Akan 38.2 354.6 0.3740 0.3353 0.6070 Krou 13.9 367.5 0.3069 0.2530 0.4291 Mande (North) 13.0 338.5 0.3879 0.3701 0.7045 Mande (South) 11.2 388.9 0.2745 0.2612 0.4949 Voltaic 9.5 244.7 0.4147 0.3979 0.7537 Non-Ivorian 14.1 377.2 0.3119 0.2820 0.5074 C6te d'Ivoire 100.0 350.9 0.3531 0.3253 0.6077 Between-Group Contribution 0.0062 0.0068 0.0258 (Z) 1.8% 2.1% 4.2% Note: Four households could not be identified by ethnic group and were dropped from the sample, which explains why figures for C6te d'Ivoire differ slightly with those in previous tables. - 33 - As discussed in the previous section, the Akan group is primarily found in the southeastern quarter of Cote d'Ivoire (and thus predominates in the East Forest region), the Krou group is found in the Southwest (the West Forest region). The Mande groups are found in the Northwest areas, while the Voltaic group is found in the North and Northeast regions. Non-Ivorian groups are found throughout C6te d'Ivoire and are primarily composed of the non- Ivorian nationalities given in Table 11. The overall impression from Table 12 is that differences in per capita mean expenditures among the different ethnic groups do not account for a substantial amount of overall inequality. The between-group contribution is less than 5% for all three inequality measures. Except for the Voltaic group, whose low level of welfare is not surprising given its location in the northern Savannah areas of Cote d'Ivoire, differences in mean adjusted per capita expenditures are very small among these groups. D. 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 to 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 in Table 3 in the previous section, one could expect this to also be the case in C6te d'Ivoire. Table 13 gives inequality decompositions where households are grouped according to the educational level of the head of the household (or of the main earner if the head is not employed). - 34 - Table 13: Per Capita Expenditure Inequality-Decomposed by Education of Head Mean % of Pop- Adjusted Exp. Log Education of Head ulation (CFAFx1000 yr.) Theil T Theil L Variance None 65.3 254.7 0.2328 0.2318 0.4683 Primary (1-7) 18.5 329.5 0.1937 0.1961 0.3979 Middle (8-11) 9.7 590.8 0.2333 0.2153 0.3965 High School (12-14) 4.0 828.4 0.2030 0.1916 0.3599 University 2.4 1348.9 0.2289 0.2389 0.5060 C6te d'Ivoire 100.0 350.9 0.3528 0.3252 0.6077 Between Group Cont. 0.1299 0.1031 0.1629 (%) 36.8% 31.7% 26.8% Note: One household was dropped because no data were available on schooling attainment of its head. This has a small effect on the C8te d'Ivoire figures. The figures in Table 13 clearly demonstrate the important role played by education in determining the distribution of welfare in C6te d'Ivoire. Depending on the inequality measure used, differences in the educational attainment of the household head account for one-fourth to one-third of overall expenditure inequality. In fact, these figures may understate the contribution of education for two reasons. First, the educational levels of other household members are not considered, so that their contribution to differences in overall welfare is not taken into account. Second, as explained in Appendix C, the log-variance measure of inequality is not sensitive to Pigou-Dalton transfers at high income levels. This implies that it may understate the between-group contribution if one group (in this case those with a university education) has a very high mean expenditure level. Thus the 26.8% figure is probably misleadingly low. These results reinforce - 35 - the argument made in Section III that a reorientation of the educational system in Cate d'Ivoire to provide better access to primary education 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 Cate d'Ivoire. Substantial differences exist between the welfare levels among the five regions in Cate d'Ivoire, accounting for about 27-29% of overall inequality. In contrast, differences in nationality and ethnic group welfare levels are very small relative to differences within those groups. Finally, the proportion of overall inequality due to differences in welfare levels by education groups is quite high, which underscores the role played by education in determining the welfare levels of Ivorian households. - 36 - V. Characteristics of the Poor in C6te d'Ivoire Another method of describing the distribution of welfare in Cote d'Ivoire 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 is 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 Cote d'Ivoire, as a whole, and also within urban and rural areas. A. Poverty in C6te d'Ivoire - 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 10% 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 in which adjusted per capita expenditures are below 95,681 CFAF per - 37 - year. The larger poverty group has a corresponding poverty line of 170,766 CFAF per year. In the remainder of this subsection both poverty groups will be described in detail. Table 14: Household Size and Per Capita Consumption of the Poor Poorest 10% Poorest 30% All Ivorians Household Size 14.5 13.1 12.3 Adjusted Household Size 9.1 8.0 7.4 Food Consumption 32.2 47.1 105.5 Adjusted Food Consumption 49.8 75.5 171.7 Total Consumption 45.0 70.0 216.5 Adjusted Total Consumption 69.9 112.7 350.9 Notes: 1. Figures are per capita averages, not per household averages. 2. Consumption is measured in CFAFx1000 per year. Table 14 provides basic information on household size and per capita consumption for both poverty groups. Figures are also included for the population as whole. One can see that a typical poor person lives in a larger household than an average Ivorian, but the differences are not unusually large. In terms of per capita consumption expenditures, both adjusted and unadjusted by family composition, the poorest 30% have less than half of the food expenditures and only a third of the total expenditures of an average Ivorian. For the poorest 10% the corresponding fractions are less than one third and one fifth. Thus, both definitions of poverty represent populations whose consumption levels are much smaller than that of the typical Ivorian. - 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 15. In fact, male-headed households are more common among the poor than among the overall population. Table 15: Sex of Household Head Among the Poor Sex Poorest 10% Poorest 30% All Ivorians Male 93.5 93.8 90.0 Female 6.5 6.2 10.0 This is in part due to the disproportionate number of female-headed households in urban areas, as pointed out in Section III. The fact that the poor are most often found in rural areas is seen in Table 16, which gives the distribution of the poor by the five regions in C6te d'Ivoire. Table 16: Distribution of the Poor by Regions Region Poorest 10% Poorest 30% All Ivorians Abidjan 2.0 3.5 18.8 Other Urban 2.0 10.8 22.4 West Forest 8.1 11.2 15.2 East Forest 31.1 34.4 24.7 Savannah 56.8 40.1 18.9 - 39 - The distribution of the poor is quite uneven across the five regions. Poverty, particularly extreme poverty (poorest 10%) is relatively rare in urban areas. In the rural areas poverty is not evenly distributed but is found primarily in the East Forest and Savannah regions. In contrast, the West Forest region does relatively well. The majority of the extremely poor is located in the Savannah, which is consistent with the findings in Section III. In the next two subsections poverty will be analyzed within urban and rural areas, respectively. One may also be interested in the nationalities and ethnic groups which predominate among the poor. Eighty-eight percent of the poorest 30 percent and ninety-one percent of the poorest 10% were citizens of Cate d'Ivoire, which is not very different from the nationwide figure of eighty-six percent. The distribution of the various ethnic groups among the poor is given in Table 17. These figures reflect those given-in Table 16 in that the groups in the Savannah area (Northern Mande and Voltaic) are more likely to be poor than those found in other parts of the country. This relationship is stronger when examining the poorest 10% of the population. Table 17: Distribution of the Poor by Ethnic Group Ethnic Group Poorest 10% Poorest 30% All Ivorians Akan 34.0 38.4 38.1 Krou 5.2 9.9 13.9 Mande (North) 22.3 15.4 13.0 Mande (South) 4.8 7.6 11.2 Voltaic 26.5 16.8 9.5 Other 0.7 0.2 0.3 Non-Ivorian 6.5 11.6 14.1 - 40 - Perhaps the most useful information for policy analysis is the incidence of poverty among different types of workers. Table 18 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 almost always self-employed and they work primarily in agriculture. Thus raises in wages paid to government and parastatal workers will have virtually no effect on poverty and neither will legislation regarding minimum wages to be paid by private employers. It is clear that policies to reduce poverty must be aimed at self-employed workers in agriculture, particularly in the East Forest and Savannah regions. Raising agricultural incomes in Cote d'Ivoire is necessary to make significant reductions in poverty. Table 18: Distribution of the Poor by Employer and Occupation Poorest 10% Poorest 30% All Ivorians Employer None 0.3 0.1 1.1 Government 1.0 1.6 11.4 Parastatal 0.0 0.0 1.3 Private 1.0 3.2 12.5 Self-Employed 97.7 95.1 73.8 Occupation None 0.3 0.1 0.9 Agricultural 92.3 87.5 60.3 Sales/Services 5.1 7.8 20.5 Industrial 0.8 2.3 4.7 White Collar/ Management 0.0 1.5 11.7 Other 1.6 0.9 1.9 Note: Employer and Occupation refer to that of the head of household, or to that of the main earner if the head does not work. - 41 - The characteristic of Ivorians 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 C6te d'Ivoire. These figures demonstrate the importance of education in C6te d'Ivoire. Virtually all of the poor have little or no education, which implies that a moderate or high level of education is almost certain to remove one from the ranks of the poor. Of course, changes in the distribution of education come very slowly even when policies are working well, but in the long run the benefits of education, particularly at the lower levels, are difficult to overemphasize. 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 C6te d'Ivoire. -School attendance of children among the poorest 30% of the population is clearly lower than it is for all Ivorians. This is particularly true for the poorest 10%, whose children attend school at a much lower rate than even the poorest 30%. Of course, much of this reflects regional differences. School attendance of the poor with urban and rural regions will be discussed below. Table 19: Distribution of the Poor by Educational Level Educational Level Poorest 10% Poorest 30% All Ivorians None 88.6 84.6 65.3 Primary (1-7) 11.1 13.9 18.5 Middle (8-11) 0.0 1.4 9.7 High School (12-14) 0.0 0.0 4.0 University 0.3 0.1 2.4 Note: Education level is that of the head of household. - 42 - Another noteworthy phenomena is that schooling appears to be delayed for the poorest 10% since school attendance is substantially 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. What is probably happening is that poor people have a less restrictive definition of good health than wealthier people, so that they consider certain health problems as "normal" while a wealthier person may classify them as illnesses. Thus the data on illness is of little use since people at different welfare levels have different conceptions of health and illness. 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. Toilet facilities are commonly nonexistent, or at best are pit latrines. Finally, drinking water comes from wells or natural sources and almost never from faucets or water vendors. The implications of these toilet and drinking water facilities on health of the poor are probably negative since they 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 fewer children and more working members (as - 43 - Table 20: Illness, School Attendance and Housing of the Poor Region Poorest 10% Poorest 30% All Ivorians School Attendance Ages 6-10 23.0 36.1 50.5 Ages 11-15 30.0 37.2 49.6 Illness % People Sick 25.6 28.4 30.8 Days ill (x) 11.9 11.9 12.3 Days inactive (X) 7.3 7.4 7.3 Source of Drinking Water Indoor Faucet 1.1 1.4 14.1 Outdoor Faucet 0.0 1.6 10.2 Water Vendor 0.9 4.0 10.4 Pump Well 65.0 47.5 27.7 Other Well 21.1 28.9 26.3 Lakes, River, Rainwater 12.0 16.4 10.8 Other 0.0 0.3 0.5 Source of Lighting Electric 9.8 14.4 41.6 Gas 0.0 0.2 0.2 Candles/Lamps 90.2 85.4 58.2 Type of Toilet Flush Toilet 2.0 3.2 22.0 Pit Latrine 32.6 40.8 41.8 None 65.4 56.0 36.3 - 44 - Table 21: Composition of Household Members Among the Poor Household Member Poorest 10% Poorest 30% All Ivorians Children 0-6 yrs. 22.8 24.3 24.2 Children 7-12 yrs. 16.7 17.5 18.1 Children 13-17 yrs. 10.7 11.3 12.3 Workers 57.5 53.7 43.2 Note: Some workers may also be children, so the categories do not necessarily sum to 100%. a percentage of total members) than the typical Ivorian household.l/ This implies that it is not the lack of work 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 primary found in agriculture, is clearly needed, but is beyond the scope of this paper. To summarize the most important findings of this subsection, the poor are predominantly found in rural areas, particularly the East Forest and Savannah regions. They are almost always self-employed and are primarily in agricultural occupations. They have a very low level of education, and unfortunately school attendance among the poor is rather low. Although they come from somewhat larger households, the percentage of household members who are children is actually lower than average, and the percent of working 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 - household members is substantially higher than that of a typical Ivorian household. B. Poverty in Urban Cate d'Ivoire The description of poverty in the previous subsection does not account for differences between poverty in urban areas and that found in rural areas. For example, it is doubtful that the urban poor are workers in agriculture. This subsection will examine poverty in Abidjan and other urban areas, while the next subsection will do the same for rural areas in Cate d'Ivoire. As pointed out above, the poor are much less numerous in urban areas than in rural areas given a poverty line for C8te d'Ivoire as a whole (see Table 16). Yet poverty has a relative aspect, so that one could classify the urban poor as the poorest 10% or 30% 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 Abidjan and the poor in other urban areas. The poverty lines in Abidjan for the poorest 10% and poorest 30% are 213,482 CFA and 349,859 CFAF, respectively. The analogous poverty lines in other urban areas are 157,393 CPA and 239,741 CFAF. Table 22 presents basic information on household size and per capita consumption for the urban poor in C8te d'Ivoire. It also includes average figures for both Abidjan and other urban areas. As with C6te d'Ivoire as a whole, household size is slightly larger for the urban poor relative to urban residents in general, but the differences are small. For both Abidjan and other urban areas, the food consumption of the poorest 30% is about half of - 46 - Table 22: Household Size and Per Capita Consumption of the Urban Poor Abidjan Other Urban Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All Household Size Actual 10.8 10.3 10.2 14.1 13.4 12.5 Adjusted 7.5 6.6 6.4 8.9 8.3 7.4 Food Consumption Actual 50.4 69.8 136.7 34.7 52.3 101.8 Adjusted 71.8 108.1 215.4 55.7 83.7 167.2 Total Consumption Actual 110.5 154.4 402.3 76.3 107.7 253.3 Adjusted 157.4 237.6 633.8 121.4 172.9 412.7 Note: 1. Figures are per capita averages. 2. Consumption is measured in CFAFx1000 per year. what it is for an average urban resident. The corresponding fraction for the poorest 10% is slightly above one third. In terms of total consumption expenditures, the spending level of the poorest 30% is about 40% of the average level, while the figure for the poorest 10% is about 25-30%. Table 23 investigates whether female headed households are more common among the urban poor. In Abidjan, households headed by women are more likely to be found among the poor, but the differences are not very large. In other urban areas the poorest 30% are slightly more likely to live in households headed by women but the poorest 10% are less likely. This suggests that living in a household headed by a woman entails a small disadvantage in Abidjan but makes little difference in other urban areas. - 47 - Table 23: Sex of Household Head Among the Urban Poor Abidjan Other Urban Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All Male 76.4 78.6 82.0 91.5 85.1 87.3 Female 23.6 21.4 18.0 8.5 14.9 12.7 Do the urban poor belong to different nationalities or ethnic groups relative to the overall urban population? The answers to these questions are given in Table 24. As far as nationality is concerned, the poorer people in urban areas, particularly Abidjan, are more likely to be non-Ivorian than the Table 24: Nationality and Ethnic Group Among the Urban Poor Abidjan Other Urban Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All Nationality Ivorian 60.6 68.2 77.2 70.5 74.8 84.5 Burkino 10.8 11.5 6.7 8.5 11.7 5.7 Mali 20.9 14.5 9.2 13.9 8.5 4.1 Guinea 5.0 3.6 2.4 5.4 1.8 1.6 Other 2.7 2.2 4.5 1.7 3.2 4.1 Ethnic Group Akan 20.1 24.8 31.0 33.2 37.3 38.2 Krou 3.9 16.9 10.9 2.4 7.4 12.7 Mande (North) 27.0 15.8 14.8 17.3 15.3 16.5 Mande (South) 0.0 1.8 10.9 5.1 2.4 5.6 Voltaic 9.7 12.5 7.9 17.0 12.6 11.3 Other 0.0 0.0 0.0 0.0 0.0 0.4 Non-Ivorian 39.4 28.3 24.5 25.1 25.1 15.3 - 48 - urban population as a whole. This coincides with the observation that many low-paying jobs often go to African immigrants (den Tuinder, 1978). Of course this is not surprising since C6te d'Ivoire's success relative to its neighbors attracts immigrants from those countries who are quite willing to accept jobs that pay low wages by Ivorian standards. A similar phenomenon takes place among the ethnic groups in Abidjan. The Northern Mandes and Voltaics came from the relatively poor Savannah region and thus tend to be found among the poor in urban areas. However, there is little difference among the poor and the general population in other urban areas, except that the Krou ethnic group is under-represented among the poor. The previous subsection highlighted the fact that most poor people in C8te d'Ivoire (given a national poverty line) are farmers in rural areas. Of those who are relatively poor in the urban areas, what jobs do they take and whom do they work for? This information is given in Table 25. In Abidjan those who are relatively poor or more likely to live in households where the head is self-employed, usually in a sales or services occupation. This is consistent with the view that workers with little education and few skills (some of whom may be migrants) have difficulty finding formal employment and thus turn to self-employment. Another interpretation is that self-employment is more lucrative for such people and so they choose that rather than work at low wages for someone else. In addition to sales and services, the poor in Abidjan are also more likely to live in households in which the head is an industrial worker, and less likely to live in households where the head is in a white collar job and/or works for the government. - 49 - Table 25: Distribution of the Urban Poor by Employer and Occupation Abidjan Other Urban Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All Employer None 5.0 2.8 2.4 0.0 0.0 1.2 Government 11.6 12.7 22.4 8.5 10.5 26.9 Parastatal 0.0 0.7 4.2 0.0 0.0 1.6 Private 19.7 42.2 39.0 15.3 14.9 15.9 Self-Employed 63.7 41.5 32.0 76.3 74.6 54.4 Occupation None 5.0 2.8 2.4 0.0 0.0 0.9 Agricultural 1.2 2.8 1.7 53.2 54.7 31.4 Sales/Services 59.1 54.9 45.4 23.4 25.1 39.5 Industrial 23.9 20.4 13.2 9.8 7.0 6.5 White collar 5.4 14.1 35.2 13.6 11.9 17.5 Other 5.4 4.9 2.2 0.0 1.4 4.3 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 C6te d'Ivoire as a whole, poor people in other urban areas are more likely to be found in agriculture relative to the general population in other urban areas. Also, government employment of the head of household is negatively correlated with poverty. Yet in contrast with Abidjan, sales and services occupation are less common among the poor in these areas. 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 Abidjan the poor live in households where the head has a much lower educational level relative to all households in Abidjan. This is also the - 50 - case for other urban areas, yet in those areas the differences are not as large, and there is little difference between the figures for the poorest 10% and the poorest 30%. School attendance declines dramatically for those households in the poorest 10% of the population in Abidjan, but this is not the case in other urban areas, where school attendance does not decline substantially for the poorest. There may be some kind of "underclass" in Abidjan in the sense that the children of the poor do not obtain much education, but this cannot be determined until further research is done. Table 26: Education and Poverty in Urban Areas Abidjan Other Urban Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All Education of Head None 69.1 62.0 36.8 76.6 76.5 49.9 Primary (1-7) 25.9 24.5 19.8 17.0 16.3 19.8 Middle (8-11) 1.9 9.3 21.9 6.4 5.1 15.3 High School (12-14) 1.5 3.7 11.1 0.0 1.5 9.6 University 1.5 0.5 10.5 0.0 0.7 5.4 School Attendance Ages 6-10 23.5 44.0 54.0 45.0 50.4 63.0 Ages 11-15 21.9 43.3 49.3 57.1 54.0 59.4 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 toilet. In Abidjan the main source of water for the poor is water vendors. Little is known about either the - 51 - vendors or their water. Electric lighting is still the most common source of lighting for the poorest in Abidjan, though candles and/or lamps are nearly as common. Finally, flush toilets are still predominate even among the poor. All of this demonstrates that Abidjan is relatively modern and "westernized" compared to other large cities in Africa. In other urban areas the most common sourde of water is a (non-pump) well, though faucets and water vendors are also important. About half of the poorer population have electric lighting while the other half use candles or lamps. Latrines are the most common toilet among the poor and the population in general in other urban areas of C8te d'Ivoire. Table 27: Housing Characteristics of the Urban Poor Abidjan Other Urban Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All Source of drinking Water Indoor faucet 12.7 20.0 46.9 13.9 9.0 21.7 Outdoor faucet 3.5 6.4 8.6 12.5 14.5 27.6 Water vendor 79.9 72.3 43.9 12.2 16.0 8.3 Pump well 0.0 0.0 0.0 0.0 3.7 2.3 Other well 3.9 1.4 0.7 56.3 54.2 38.6 Lakes, Rivers, Rainwater 0.0 0.0 0.0 5.1 2.7 0.8 Other 0.0 0.0 0.0 0.0- 0.0 0.7 Source of Lighting Electric 62.5 72.0 84.8 41.4 56.9 78.7 Gas 0.0 0.7 0.2 0.0 0.7 0.2 Candles/lamps 37.5 27.3 15.0 58.6 42.5 21.1 Type of Toilet Flush toilet 61.4 71.7 81.4 15.9 10.7 26.9 Latrine 30.1 25.3 15.7 68.5 73.1 66.6 None 8.5 3.0 2.9 15.6 16.2 6.6 - 52 - The previous subsection made the point that the poor in C6te d'Ivoire 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 Abidjan and other urban areas. The same patterns that were found in Cote d'Ivoire as a whole are also found in urban areas - poor households have a larger proportion of working household members and a smaller proportion of children. However, it is important to note that the proportion working is much smaller in urban areas, especially Abidjan, than in Cote d'Ivoire as a whole. This is probably due to the relatively small degree of self-employment in urban areas. Self- employed heads of household, particularly farmers, have other family Table 28: Household Composition Among the Urban Poor Abidjan Other Urban Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All Children 0-6 yrs. 18.1 21.8 21.3 22.4 22.6 22.7 Children 7-12 yrs. 13.5 15.8 17.4 19.0 17.0 18.1 Children 13-17 yrs. 12.4 11.6 13.4 12.5 15.4 16.3 Workers 27.0 24.0 24.8 40.0 40.3 32.6 members 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 very small. - 53 - To summarize, the poor in urban areas of C8te d'Ivoire have substantially lower welfare levels relative to other urban residents, but many would not be considered poor if compared to the Ivorian population as a whole. Many of the urban poor are non-Ivorian, particularly the poorest 10%. A disproportionate number of the poor are self-employed, with sales and services being the most common occupation in Abidjan and agricultural occupations predominant in other urban areas. As one would expect, the education level of the heads of poor urban households are much lower than those of urban residents in general. While school attendance in other urban areas is not lower among the poor, the poorest 10% in Abidjan have substantially lower school attendance. Household amenities such as electric lighting and flush toilets are common even among the poorest in Abidjan, though less so in other urban areas. Finally, poor urban households have a lower proportion of children and a higher proportion of working members (which probably reflects self-employment) than typical urban households. C. Poverty in Rural C6te d'Ivoire In the previous subsection the poor in urban C6te d'Ivoire were defined in terms of the poorest 10% and the poorest 30% of the urban population. This subsection examines the poorest 10% and 30% of the rural population. Specifically, the three rural regions (West Forest, East Forest, and Savannah) are all examined using this definition of poverty within each region. Thus we examine the poorest 30% in the West Forest, then the poorest 30% in the East Forest, etc., and similarly for the poorest 10%, so that all three rural regions have different poverty lines. The poverty lines for the poorest 10% and 30% in the West Forest region are 120,674 CFAF and 198,168 - 54 - CFAF, respectively. The corresponding poverty lines for the East Forest region are 86,946 CFAF and 139,312 CFAF, and those for the Savannah are 59,509 and 95,681. It is important to recognize that the poorest 10% (30%) in the East Forest are significantly poorer than the poorest 10% (30%) in the West Forest, and the analogous groups in the Savannah are the poorest of all. Table 29 provides some useful data on household size and per capita consumption for the rural poor in C6te d'Ivoire. Two points should be emphasized. First, the poorest 30% and 10% 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 West Forest doing relatively well (even better than in other urban areas in terms of food consumption) while the Savannah region is the poorest of all. Table 29: Household Size per Capita Consumption of the Rural Poor West Forest East Forest Savannah Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Household Size Actual 12.8 11.2 10.8 16.4 15.1 15.2 16.9 15.1 11.6 Adjusted 8.0 6.9 6.6 9.4 8.9 8.8 10.7 9.6 7.1 Food Consumption Actual 36.2 60.7 125.5 25.8 40.0 89.4 21.4 34.5 83.6 Adjusted 56.5 96.9 204.1 43.5 65.6 148.9 31.9 52.7 137.6 Total Consumption Actual 56.8 86.9 182.2 39.9 60.2 146.5 29.5 44.5 107.7 Adjusted 89.8 139.2 296.0 67.3 98.8 246.2 44.3 68.1 177.7 Note: 1. Figures are per capita averages. 2. Consumption is measured in CFAFxlOOO per year. - 55 - 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 patterns in general. Poor households in rural areas are as likely to be headed by women as other households. Only the Savannah area shows a pattern, with poor households being slightly less likely to be headed by women than non-poor households. Thus poverty shows little relation to the sex of the head of household. Table 30: Sex of Household Head among the Rural Poor West Forest East Forest Savannah Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Male 95.5 95.5 95.2 88.0 94.2 91.4 97.7 96.8 95.2 Female 4.5 4.5 4.8 12.0 5.8 8.6 2.4 3.2 4.8 It was seen in the previous subsection that the poor in urban areas are more likely to be non-Ivorian than typical urban residents. This suggests that migrants from C6te d'Ivoire's poorer neighbors take low-paying urban jobs, but do they also take low-paying rural jobs? The relevant information is given in Table 31, which classifies the poor in rural areas by nationality and ethnic group. In rural areas there is no overall trend, but there are small differences among the three rural regions. In the West Forest region Ivorians are more likely to be poor than non-Ivorians, while in the East Forest the trend is slightly in the opposite direction. In the Savannah almost everyone is Ivorian so almost no nationality differences arise. In terms of ethnic group, there are few discernible patterns. Perhaps the most important finding - 56 - is that in the Savannah, which is the poorest region in Cote d'Ivoire, the Voltaic group is substantially over-represented among the poor. The reasons for this should be examined in future research. Table 31: National ity *ed Etelc o"op Among t1 Rural Por West Forest East Forest Savannah Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Nationality Ivorian 90.6 86.6 81.0 64.7 86.7 66.9 100.0 96.4 98.1 Burkino 2.5 5.0 7.6 15.3 12.7 9.9 0.0 0.0 0.4 Mali 2.0 0.7 5.0 0.0 0.6 2.1 0.0 0.5 0.4 Guinea 5.0 6.3 4.9 0.0 0.0 0.4 0.0 1.1 1.2 Other 0.0 1.3 1.5 0.0 0.0 0.7 0.0 0.0 0.0 Ethnic Group Akan 0.0 0.0 1.8 67.8 69.2 68.0 11.0 21.3 35.4 Krou 60.2 46.7 45.3 2.2 7.1 8.5 0.0 0.0 0.0 Mande (North) 4.0 3.7 7.5 7.4 6.3 3.5 27.3 32.5 23.7 Mande (South) 24.4 35.6 26.7 4.6 4.8 6.8 2.4 1.9 11.4 Voltaic 4.0 1.3 0.7 0.0 1,3 0.6 60.4 43.4 27.9 Other 0.0 0.0 0.0 2.8 0.9 0.7 0.0 0.0 0.0 Non-lvorian 7.5 12.7 18.1 15.3 10.5 12.0 0.0 0.9 1.7 Working in the formal sector (either for the government or a private employer) was shown to be more lucrative than self-employment, but since most of these jobs are in urban areas it may not be relevant to characterizing rural poverty. Table 32 presents statistics on the poor based on employers and occupational categories. It is quite clear that nearly everyone in rural areas lives in a household where the head is self-employed. The few who live in families where the head vorks for the governmnt or not at all are never - 57 - found among the poor. Agricultural occupations are most common for heads of households in rural areas, though sales and service occupations also exist. There is little occupational difference among poor and nonpoor households. 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 Cote d'Ivoire, and it is also useful to examine school attendance for children in rural areas. This can be seen in Table 33. Most people in rural areas live in households where the head had no education at all. In the East Forest and Savannah areas, though not in the West Forest, this is even more likely for poorer people. School attendance patterns are Table 32: Distribution of the Rural Poor by Employer and Occupation West Forest East Forest Savannah Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Employer None 0.0 0.0 0.8 0.0 0.0 0.5 0.0 0.0 0.8 Government 0.0 0.0 0.5 0.0 0.0 3.4 0.0 0.0 1.3 Parastatal 0.0 0.0 0.0 0.0 0.0 0.3 0.0 0.0 0.2 Private 7.0 4.0 2.8 0.0 1.6 2.9 0.0 0.4 2.3 Self-Employed 93.0 96.0 96.0 100.0 98.4 92.9 100.0 99.6 95.5 Occupation None 0.0 0.0 0.8 0.0 0.0 0.5 0.0 0.0 0.2 Agricultural' 83.6 91.3 87.1 89.9 93.3 88.5 96.5 97.8 94.3 Sales/Services 9.5 6.3 8.4 10.1 4.2 5.1 2.0 1.3 2.9 Industrial[ 0.0 0.0 1.1 0.0 2.5 2.1 0.0 0.0 0.7 White Collar 0.0 0.0 0.6 0.0 0.0 3.4 0.0 0.0 1.5 Other 7.0 2.3 2.1 0.0 0.0 0.3 1.6 0.9 0.5 - 58 - rather ambiguous. In the East Forest and Savannah poorer children aged 6-10 are less likely to attend school, but the opposite is true in the West Forest. In both East and West Forest poorer children aged 11-15 are less likely to attend school, yet this is not the case in the Savannah. As before these figures need to be treated with caution because poorer families may simply send children to school at a later age - this could explain the pattern found in the Savannah. It is also relevant to recall that 40% of Ivorian children of primary school age have no access to any kind of educational facilities, and it is almost certain that nearly all of them are found in rural areas. Table 33: Education and Poverty in Rural Areas West Forest East Forest Savannah Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Education of Head None 63.2 63.8 70.5 84.7 80.8 75.5 97.6 94.6 93.5 Primary (1-7) 30.9 29.4 24.9 15.3 19.2 20.0 2.4 5.4 4.2 Middle (8-11) 6.0 6.8 4.6 0.0 0.0 3.4 0.0 0.0 2.3 High School (12-14 0.0 0.0 0.0 0.0 0.0 1.1 0.0 0.0 0.1 University 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 School Attendance Ages 6-10 62.5 54.4 49.7 31.3 41.8 55.5 15.0 12.7 26.2 Ages 11-15 34.8 41.0 46.9 36.5 44.9 54.8 21.9 24.4 25.3 Housing is quite different in rural areas relative to urban areas. Differences among rich and poor households can be seen in Table 34. Since almost all water in rural areas comes from wells or natural sources, there is - 59 - little variation between rich and poor, and what variation does exist is difficult to interpret. When electric lighting is available, which is the case in the East Forest and Savannah regions, it is less common among the poor, who instead rely on candles and lamps. Finally, poorer people are less likely to use latrines (the alternative being no toilet at all) in the Table 34: Housing Characteristics of the Rural Pbor West Forest East Forest Savannah Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Source of Drinking Water Indoor faucet 0.0 0.0 0.0 0.0 0.0 1.2 0.0 0.0 0.6 Outdoor faucet 0.0 0.0 0.0 0.0 1.0 8.7 0.0 0.0 1.3 Water vendor 0.0 0.0 0.0 0.0 0.0 1.3 0.0 0.0 0.0 Pump well 34.8 37.4 37.4 73.3 57.1 43.4 77.3 67.6 57.1 Other well 58.7 47.1 48.6 15.6 22.9 28.6 14.5 18.9 16.4 Lakes, rivers, rain 6.5 15.5 14.0 11.0 18.0 15.5 8.2 13.5 24.7 Other 0.0 0.0 0.0 0.0 1.0 1.4 0.0 0.0 0.0 Source of Lighting Electric 0.0 0.0 0.9 11.0 10.6 20.9 0.0 8.6 14.6 Gas 0.0 0.0 0.0 0.0 0.8 0.3 0.0 0.0 0.2 Candles/lamps 100.0 100.0 99.1 89.0 88.6 78.8 100.0 91.4 85.2 Type of Toilet Flush 0.0 0.0 0.0 0.0 0.0 1.7 0.0 0.0 1.5 Latrine 61.7 59.1 46.7 65.3 60.1 53.0 2.8 13.4 19.5 None 38.3 40.9 53.3 34.7 39.9 45.3 97.3 86.6 79.1 Savannah, but the opposite is the case in the East and West Forest. Overall, the data in Table 34 show little systematic variation in these housing characteristics. - 60 - Finally one can examine household composition in rural C6te d'Ivoire. This is presented in Table 35. The same trends found in urban areas and in C8te d'Ivoire as a whole also hold in rural areas. Specifically, poorer households have a larger proportion of working members and a slightly smaller proportion of children than typical rural households in Cote d'Ivoire. Table 35: Household Composition Among the Rural Poor West Forest East Forest Savannah Poorest Poorest Poorest Poorest Poorest Poorest 10% 30% All 10% 30% All 10% 30% All Children: 0-6 yrs 21.9 23.2 25.2 24.8 25.2 25.7 19.2 23.1 26.2 7-12 yrs 18.4 17.9 18.3 18.4 17.7 19.4 18.0 15.9 16.9 13-17 yrs 9.5 11.4 9.5 13.8 12.6 12.0 10.6 9.5 9.2 Workers 51.7 51.3 49.7 56.1 55.0 52.1 61.2 60.1 57.3 To summarize, the poor in rural Cote d'Ivoire are in many ways typical of other rural residents. There is little difference in terms of the sex of the head of household and nationality. There is one important characteristic regarding ethnicity, which is that the poorest people in the poorest region, the Savannah, are disproportionately members of the Voltaic ethnic group. This needs more investigation. Self-employment in agricultural, and to some extent in sales and services, is the predominant occupational arrangement among both the poor and the population in general in rural areas. When ranked by the education level of the head of household, poor people tend to live in "less educated" households in the East Forest and Savannah areas, but this is not the case in - 61 - the West Forest region. School attendance of children is most often lower among the poor but this is not always the case. Housing characteristics seem to vary by only a small amount in rural Cote d'Ivoire. Finally, one again finds that poor households have more workers and fewer children than wealthier households, which again suggests that poverty is primarily a matter of low earnings among those who work. - 62 - VII. Conclusion The ability to make appropriate policy decisions in the areas of economic development is heavily dependent on one's knowledge of social and economic characteristics of the country in question. In particular, policies aimed at raising social welfare and reducing inequalities in the distribution of welfare cannot be articulated clearly until the characteristics of households at different welfare levels are known. This paper has provided a characterization of the distribution of welfare in C6te d'Ivoire in 1985. It provides useful information not only for policy choices in that country, but hopefully for other sub-Saharan african countries as well. The following findings are the most important for purposes of policy formulation. First, welfare levels are clearly higher in urban areas, particularly Abidjan, than in rural areas. Of the three rural areas, the West Forest has a higher level of welfare than the East Forest, while the northern Savannah areas have the lowest levels of welfare. That urban areas are better off than rural areas is no surprise, as this pattern is found in virtually every country in the world. Of course, this does not imply that a rural resident will be better off by moving to an urban area, since characteristics of rural residents differ from those of urban dwellers. Urban dwellers have higher levels of education and are more likely to be working for the government or a private employer, and those characteristics are strongly associated with high welfare levels. Second, differences in nationality and ethnic group (tribe) were not very important in determining welfare levels, and contributed little to overall inequality in C8te d'Ivoire. In most cases the welfare of different ethnic groups corresponded highly with those of the regions in which they - 63 - predominate. In particular, the Voltaic group, which is the poorest, predominates in the northern Savannah region. A related finding is that the sex of the head of household has little to do with welfare levels, particularly in rural areas. Third, given the strong association between education and welfare levels, it is important that school attendance among the poor is rather low. Much of this variation is due to the fact that school attendance in the Savannah, where the poor are most often found, is lower than in any other region of Cote d'Ivoire. Yet school attendance is low in other regions as well and 40% of primary school age children have no access to educational facilities. Clearly, policies are needed to raise school attendance throughout C8te d'Ivoire. Fourth, the poorest households in Cote d'Ivoire are overwhelmingly in agricultural pursuits, where they are self-employed cultivators. Any policy aimed at equalizing the distribution of welfare and reducing poverty should include efforts to raise agricultural productivity in general and among these workers in particular. Given past achievements, there is a precedent for raising productivity in agriculture in the future. Yet this must take place not only in the forest regions but most importantly in the Savannah area as well. Finally, household composition, in the sense of many dependents and few workers, does not explain why certain households are poor. In fact, the poorest households have a lower proportion of children and a higher proportion of workers than wealthy households. Thus the productivity of workers in agricultural pursuits is the key variable from the viewpoint of making policy - 64 - choices. An analysis of agricultural productivity would be most useful in formulating specific policy recommendations. This paper is primarily exploratory in nature, and thus raises many more questions than it answers. Yet, it constitutes an important first step by providing descriptive information about welfare levels in C6te d'Ivoire and providing initial direction on appropriate policies for consideration. The data available from the C6te d'Ivoire Living Standards Survey can be used to conduct thorough analyses of the questions asked but not answered in this paper. Two questions in particular ought to receive immediate attention: 1. What determines the school attendance of children and the availability of school facilities in rural areas? 2. What are the principal determinants of the incomes of agricultural households? The data now available remain to be exploited for answering these and many other questions concerning the distribution of welfare in C6te d'Ivoire. - 65 - REFERENCES Ainsworth, Martha and Juan Munoz. 1986. "The C6te d'Ivoire Living Standards Survey: Design and Implementation," Living Standards Measurement Study, Working Paper No. 26. The World Bank. Anand Sudhir, and Christopher Harris. 1985. "Living Standards in Sri Lanka, 1973-1981/82: A Partial Analysis of Consumer Finance Survey Data," Mimeo. Bhagwati, Jagdish. 1985. "Growth and Poverty." Center for Advanced Study of International Development, Michigan State University. Occasional Paper No. 5. Central Bank of Ceylon, 1984. Report on the Consumer Finances and Socioeconomic Survey 1981/82. Colombo. Sri Lanka. Chenery, Hollis, et al. 1974. Redistribution with Growth. Oxford University Press. 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. den Tuinder, Bastiaan A. 1978. Ivory Coast: The Challenge of Success. Johns Hopkins University Press. Jain, Shail. 1975. The Size Distribution of Income: A compilation of Data. The World Bank. Kozel, Valerie. 1986. Savings and Investment Behavior of Ivorian Households. Unpublished Ph.D. Thesis. Department of Urban Studies, M.I.T. 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. Thomas, Duncan. 1986. "Food Shares as a Welfare Measure." Unpublished Ph.D. Thesis. Department of Economics. Princeton University. - 66 - van der Caag, Jacques. 1982. "On Measuring the Cost of Children." Children and Youth Services Review. Vol. 4, No. 1, pp. 77-109. van der Caag, Jacques and Wim Vijverberg, 1986. "A Switching Regression Model for Wage Determinants in the Public and Private Sectors of a Developing Country." The World Bank. Mimeo. World Bank. 1983, 1986. World Development Report. Oxford University Press. World Bank. 1986. "The Ivory Coast in Transition: From Structural Adjustment to Self-Sustained Growth". Mimeo. Washington, D.C. - 67 - Appendix A: The C6te d'Ivoire Living Standards Survey The C6te d'Ivoire Living Standards Survey (CILSS) is a random sample of 1600 households interviewed from mid-February, 1985, to mid-February, 1986. 1! The data collected include information on food and nonfood 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 Ainsworth and Munioz (1986). The data pertinent to measurement of consumption are discussed in this appendix. Data collected in the CILSS which are relevant for the measurement of consumption include: 1) Daily expenses on regularly purchased nonfood items (such as fuel, cigarettes and personal health items) and food consumed outside the household within the last two weeks; 21; 2) Expenditures on clothing, household goods and maintenance, medicines and other irregular expenditures within both the last year and the last two weeks; 3) Remittances sent from the household to persons not residing in the household in the past year; 4) Food expenditures in the last year and the last two weeks; 5) Possession of durable goods, including present value and cost when purchased; 6) Value of food produced and consumed by the household. 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, 31 households have been dropped from the original 1600. 2/ Specifically, between the time the first and second interviews were conducted, which is generally two weeks. - 68 - Of the four types of explicit expenditure listed above (i.e. those numbered one through four) it is assumed that remittances are pure transfers which do not raise the welfare of the household members sending the transfers. Thus remittance expenditures are not included as part of household consumption expenditures. The other three expenditure categories clearly raise the welfare of households and the annual data are summed to give annual expenditure in the past year. 1/ 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 for the Abidjan area and for other urban areas in Cote d'Ivoire by Kozel (1986). 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 / The first type of expenditure listed in the preceding paragraph is annualized based on expenditure in the past two weeks. - 69 - than calculated for home owners, again after adjusting for sample selection bias. - Unfortunately, there is almost no rental market in the rural areas of C6te d'Ivoire, which prevents one from estimating imputed rents in rural areas. This forces one to impose imputed rents in rural areas. Based on the twenty-four households which report rents in rural areas, it is assumed that each rural household has an imputed rent of 3.3% of total expenditures. This may result in a small reduction in the estimate of overall variability in total consumption, but since the amount imputed is only 3.3% it is unlikely to significantly affect the results. For durable goods, the 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 forgone investment earnings. It is assumed that the foregone opportunity cost is 10% in real terms. This estimate may be rather high, but since the use value of durable goods only accounts for 4.0% of total consumption there is little cause for concern. Estimates of imputed rents and of the use-value of durables were added to the three categories of explicit expenditures (excluding remittances) ! Not all varibility in rents was captured in the hedonic rent equation since the R statistic was 0.646 in Abidjan and 0.776 in other urban areas. Thus predicted rents for home owners do not reflect all the variability in this component of total consumption. Since rents iimputed or actual) account for only 10-15% of total consumption and the R statistics were rather high, this does not seem to be a serious problem. - 70 - described above. The value of food produced and consumed in the household was then added to arrive at the value of total household consumption. A minor adjustment was made to exclude the purchase of durables and maintenance expenditure on housing and durables. The latter should be reflected in the value of housing and durable goods and thus are accounted for in estimates of imputed rent and the use value of durables. 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. An additional difficulty in comparing expenditures across different households is that prices may vary over different regions of the country and over different agricultural seasons. Since almost all expenditures are based on purchases within the past year, seasonal price variations should not significantly affect comparability across households. Regional differences, however, are important, and two price indices, one for food and one for total expenditures, have been calculated for five separate regional groups: the Abidjan area, other urban areas, the West Forest region, the East Forest region, and the Savannah area in the North. The food price index is based on prices from 17 items which account for over 70% of total food expenditures (including items produced and consumed at home). The nonfood price index presented difficulties due to the small number of items for which prices were available, and more seriously, the lack of comparability between items priced in Abidjan and those priced in the rest of C6te d'Ivoire. Because it is thought that price differences in nonfood items are primarily due to differences in transportation costs, it seemed advisable to use an item or set of items which is both relatively non-perishable and found - 71 - throughout the country. The only item which meets these criteria is tomato paste, which is sold in small cans. Thus the price of these cans of tomato paste are used to represent the price of nonfood items. Because there are no comparable data on the price of tomato paste in Abidjan it is assumed that it is the same as the price prevailing in other urban areas. The 1979 Ivorian budget survey found that tomato paste was slightly (7%) cheaper in other urban areas relative to Abidjan. The use of this item as an indicator of nonfood prices has some intuitive appeal - its price is lowest in urban areas and highest in the sparsely populated Savannah region in the North. Table A.1 presents the regional price indices used in this paper. Table A.1: Regional Price Indices in C8te d'Ivoire, 1985 Food Non-Food Total Price Index Price Index Price Index All Cote d'Ivoire 100.0 100.0 100.0 Abidjan 127.5 97.7 112.8 Other Urban 98.1 97.7 97.8 West Forest 84.3 100.8 92.3 East Forest 90.9 98.5 94.7 Savannah 83.3 108.9 96.0 At this point one should check to see whether these regional price indices and other adjustments to the welfare measure have a large impact on the overall distribution of welfare in C6te d'Ivoire. This is done in Table A.2, which demonstrates the impact of different definitions of or refinements - 72 - to the measure of welfare. Column 1 gives shares of total welfare going to each 10% of the 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% 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 tendancy to exaggerate inequality by adding an erroneous source of variation, which is evident in column 2 of Table A.2. If no regional price indices are used the distribution of per capita consumption (after adjustment for household composition) is that given in column 3. As was seen in the text, the Savannah is the poorest region, while Abidjan is the wealthiest. Some of this difference (in nominal terms) is due to higher prices in Abidjan and lower prices in the Savannah. If no adjustment is made for this, then inequality may be exaggerated. This is seen to be the case for C8te d'Ivoire since money-metric welfare is less evenly distributed in column 3 than in column 1. 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 mild. In any case, this adjustment does not have a large influence and so does not artificially "create" some of the results in the text. - 73 - Table A.2: Welfare Deciles According to Different Welfare Measures Adjusted No Food per Capita By No Price Composition Consumption Decile Consumption Households Adjustment Adjustment Per Capita 1 1.98 1.57 1.84 1.96 2.44 2 3.26 2.85 2.99 3.18 4.02 3 4.39 3.93 4.06 4.30 5.17 4 5.39 4.99 4.91 5.23 6.21 5 6.47 6.12 5.98 6.26 7.33 6 7.73 7.56 7.27 7.53 8.55 7 9.53 9.25 8.97 9.21 10.12 8 11.84 11.92 11.70 11.80 12.11 9 15.87 16.88 15.99 15.93 15.59 10 33.55 34.94 36.28 34.60 28.46 Finally, 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 Cote d'Ivoire. Table A.3 gives the breakdown of total consumption by various categories. In C8te d'Ivoire as a whole, slightly less than half of total consumption (including imputed rent and the value of durable services) is food consumption. Of nonfood consumption, rents (imputed or real) and durable services play a relatively minor role. When one examines consumption patterns by different areas of C6te d'Ivoire, one finds that the fraction of total expenditures devoted to food consumption is generally smaller in wealthier regions than in poorer regions. - 74 - Table A.3: Composition of Total Consumption in Cbte d'lvoire, 1985 All Other West East C8te d'ivoire Abidjan Urban Forest Forest Savannah Food purchased 31.9% 33.9% 33.5% 29.9% 28.0% 28.2% Food produced 11.8 0.1 4.0 27.6 27.5 37.1 Food eaten away from home 3.7 4.4 2.9 5.4 3.1 2.0 Rents 8.7 10.5 13.6 3.3 3.3 3.3 Water/Elec. Utilities 5.7 8,3 8.5 0.0 1.4 1.4 Durable services 3.8 4.7 4.6 1.7 2.9 2.0 Other nonfood 34.4 38,1 33.1 32.2 33.8 26.0 Total food 47.5 38.4 40.3 62.9 58.6 67.4 Total nonfood 52.6 61.6 59.7 37.2 41.5 32.7 Total value CFAFxlOOO per capita per year 216.5 402.3 253.3 182.2 146.5 107.7 Note: Figures have not been adjusted by regional price indices or household composition. - 75 - 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 proportion 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 C8te d'Ivoire. 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. - 76 - Table B.1: Distribution of Food Consumption by Food Consumption Deciles Mean Per Capita Food % of Food Expenditures Expenditures Food Share (Z) (CFAFxl000 per year) in C6te d'Ivoire within each Decile Decile Adjusted Unadjusted Adjusted Unadjusted Adjusted Unadjusted 1 41.9 25.5 2.44 2.41 0.471 0.478 2 69.0 41.4 4.02 3.93 0.472 0.483 3 88.8 53.5 5.17 5.08 0.508 0.504 4 106.7 63.9 6.21 6.06 0.512 0.501 5 125.8 75.2 7.33 7.13 0.456 0.465 6 146.9 88.9 8.55 8.43 0.463 0.475 7 173.8 104.7 10.12 9.93 0.483 0.471 8 207.9 124.8 12.11 11.83 0.494 0.500 9 267.8 167.5 15.59 15.89 0.500 0.513 10 488.8 309.1 28.46 29.32 0.497 0.480 All C6te d'Ivoire 171.7 105.5 100.00 100.00 0.489 0.487 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 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 - 77 - 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 susceptible 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.1, and if true it implies that per capita total expenditure is a better iAdicator of welfare than per capita food expenditure. - 78 - 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 characteristics 1/ which are highly desirable: 1. Mean Independence - inequality is unaffected by equiproportionate changes 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. - 79 - 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: - 80 - N Y Y. N Y Y Y. /Y 1) Theil (T) = Y In = Z {_-} Tj + Z { i21 ln Y= Y j .3 Y 3 N N. N. N./N 2) Theil (L) = NI ln {Y N =+ n i1 1 .3. .3 N 2 I N 3) Log Variance (LV) = [ l[n (Yi) - lnY] Z {j'} LV + Z [|n Y.- ln YJ i=1 1 .N .3 where Y = total income of the population, Yi = income of individual it 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 In Y. = mean of ln (Yi) over the population in group j. The termi 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 Lorens diagram. Its mathematical formula is: Gini (G) E E iY i2 Yl I where 2 where il and i2 simply correspond to the respective suimation signs. LSMS Working Papers (continued) No. 36 Labor Market Activity in C6te d'Ivoire 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 The Effects of Household and Community Characteristics on7 the Nutrition of Preschool Children: Evidence from Rural Cote d'Ivoire No. 41 Public-Private Sector Wage Differentials in Peru, 1985-86 No.42 The Distribution of Welfare in Peru in 1985-86 No. 43 Profits from Self-Employment: A Case Study of Cote d'Ivoire No. 44 The Living Standards Survey and Price Policy Reform: A Study of Cocoa and Coffee Production in Cote d'lvoire No.45 Measuring the Willingness to Pay for Social Services in Developing Countries No. 46 Noonagricultural Family Enterprises in C6te d'lvoire: A Descriptive Analysis No. 47 The Poor during Adjustment: A Case Study of Cote d'Ivoire No.48 Confronting Poverty in Developing Countries: Definitions, Infornation, and Policies No. 49 Sample Designs for the Living Standards Surveys in Ghana and Mauritania/Plans de sondage pour les enquetes sur le niveau de vie au Ghana et en Mauritanie No.50 Food Subsidies: A Case Study of Price Reform in Morocco (also in French, SOF) No.,51 Child Anthropometry in Cote d'Ivoire: Estimates from Two Surveys, 1985 and 1986 No. 52 Public-Private Sector Wage Comparisons and Moonlighting in Developing Counltries: Evidencefrom C6te d'Ivoire and Peru No. 53 Socioeconomic Determinants of Fertility in C6te d'Ivoire No. 54 The Willinigness to Pay for Education in Developing Countries: Evidence from Rural Peru No. 55 Rigidite des salaires: Donnees micro&conomiques et macro6conomiques sur l'ajustement du march' du travail dans le secteur moderne (in French only) No. 56 The Poor in Latin America during Adjustment: A Case Study of Peru No. 57 The Substitutability of Public and Private Health Carefor the Treatment df Children in Pakistan No. 58 Identifying the Poor: Is "Headship" a Useful Concept? No. 59 Labor Market Performance as a Determinant of Migration No. 60 The Relative Effectiveness of Private and Public Schools: Evidence from Two Developing Countries No. 61 Large Sample Distribution of Several Inequality Measures: With Application to C6te d'Ivoire No. 62 Testingfor Significance of Poverty Differences: With Application to Cote d'Ivoire No. 63 Poverty and Economic Growth: With Application to Cote d'Ivoire No. 64 Education and Earnings in Peru's Informal Nonfarm Family Enterprises No. 65 Formal and Informal Sector Wage Determination in Urban Low-Income Neighborhoods in Pakistan No.66 Testing for Labor Market Duality: The Private Wage Sector in C6te d'Ivoire No. 67 Does Education Pay in the Labor Market? The Labor Force Participation, Occupation, and Earnings of Peruvian Women No. 68 The Composition and Distribution of Income in C6te d'Ivoire No. 69 Price Elasticities from Survey Data: Extensions and Indonesian Results The World Bank Headquarters European Office Tokyo Office 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-cho'me Telephone: (202) 477-1234 Telephone: (1) 47.23.54.21 Chiyoda-ku, Tokyo 100, Japan Telex: WUI 64145 WORLDBANK Telex: 842-620628 Telephone: (03) 214-5001 RCA 248423 WORLDBK Telex: 781-26838 Cable Address: INTBAFRAD WASHINGTONDC