V- j 8705 SOCIAL DIMENSIONS OF ADJUSTMENT IN SUB-SAHARAN AFRICA _________ @0@ Malnutrition in Cote d'Ivoire Prevalence and Determinants David E. Sahn SDA Working Paper Series Editorial Board Chairman Ismail Serageldin Director, Occidental and Central Africa Department World Bank Members Ramesh Chander, Statistical Adviser, World Bank Dennis de Tray, Research Administrator, World Bank Yves Franchet, Director General, Statistical Office of the European Communities Ravi Kanbur, Editor of World Bank Economic Review and World Bank Research Observer and Senior Adviser, SDA Unit, World Bank Gabriel Kariisa, Chief Economist, African Development Bank F. J. C. Klinkenberg, Director, Directorate General for Development, Commission of European Communities Jacques Loup, Coordinator of Assistance to Developing Countries, United Nations Development Programme E. M. Morris-Hughes, Chief, Nutrition Planning, Assessment and Evaluation Service, Food and Agriculture Organization F. Stephen O'Brien, Chief Economist, Africa Region, World Bank Graham Pyatt, Professor of Economics, University of Warwick Paul P. Streeten, Director, World Development Institute, Boston University Victor E. Tokman, Director, Employment and Development Department, International Labour Office R. van der Hoeven, Senior Adviser, United Nations Children's Fund Editor Michel Noel Chief, SDA Unit World Bank Managing Editors Marco Ferroni Christiaan Grootaert Senior Economist, SDA Unit Senior Economist, SDA Unit World Bank World Bank SOCIAL DIMENSIONS OF ADJUSTMENT IN SUB-SAHARAN AFRICA WORKING PAPER NO. 4 Policy Analysis Malnutrition in Cote d'Ivoire Prevalence and Determinants David E. Sahn The World Bank Washington, D.C. Copyright 1990 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 1990 The findings, interpretations, and conclusions expressed in this paper are entirely those of the author 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. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. In order to present the results of research with the least possible delay, the manuscript has not been edited in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. 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 dissemnnation 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 complete backlist of publications from the World Bank is shown in the annual Index of Publications, which contains an alphabetical title list (with full ordering information) and indexes of subjects, authors, and countries and regions. The latest edition is available free of charge from 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'Iena, 75116 Paris, France. ISSN 1014-739X David E. Sahn is senior research associate and deputy director at the Cornell University Food and Nutrition Policy Program. Library of Congress Cataloging-in-Publication Data Sahn, David E. Malnutrition in C6te d'Ivoire. (Social dimensions of adjustment in Sub-Saharan Africa) (SDA working paper series) Includes bibliographical references (p. ) 1. Poor-Ivory Coast. 2. Malnutrition-Ivory Coast. I. Title. II. Series. III. Series: SDAworkingpaper) series. HC1025.Z9P62 1990 363.8'1'096668 90-12449 ISBN 0-8213-1556-0 SDA Working Paper Series Foreword Integration of social and poverty concerns in the struc- economic crisis in Africa on the one hand and the tural adjustment process in Sub-Saharan Africa is a adjustment response on the other hand affect the liv- major driving force behind the design of the World ing conditions of people. Empirically, major improve- Bank's adjustment lending program in the Region. To ments are needed in our knowledge of the social further the goal, the Social Dimensions of Adjustment dimensions of life in Africa, how they change, and (SDA) Project was launched in 1987, with the United whether all groups in society participate effectively in Nations Development Programme and the African the process of economic development. Gaining this Development Bank as partners. Since then many other knowledge will demand new efforts in data collection multilateral and bilateral agencies have supported the and policy oriented analysis of these data. Most im- project financially as well as with advice. The task portantly, policy actions are needed in the short term presents a formidable challenge because of the sever- to absorb undesirable side-shocks stemming from the ity of economic and social constraints in Africa and the adjustment process so that the poor and disadvan- intrinsic difficulty of tracing the links between eco- taged are not unduly hurt, and in the long term to nomic policies and social conditions and poverty. It is ensure that these groups fully participate in the newly essential to have a continuous professional dialogue generated growth. The SDA Project's mandate is to between all concerned parties, so that the best ideas operate, in a concerted way, in all three domains: get discussed by the best minds, and become, as concepts, data, actions. This working paper series will quickly as possible, available for implementation by report progress and experience in all three areas. I policymakers. This is the aim of the SDA working encourage every reader's active participation in the paper series. series and the work it reports on. It is meant to be a To fulfill its mission, the SDA Project operates on forum not only for exchange of ideas but even more different levels. Conceptually, contributions need to importantly to advance the cause of sustainable and be made which advance our understanding of how the equitable growth in Africa. Edward V.K. Jaycox Vice President, Africa Region iii l ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~ ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The Social Dimensions of Adjustment (SDA) Project Working Paper Series The SDA Project has been launched by the UNDP The Surveys and Statistics subseries focuses on the Regional Programme for Africa, the African Develop- data collection efforts undertaken by the SDA Project. ment Bank, and the World Bank in collaboration with As such, it will report on experiences gained and other multilateral and bilateral agencies. The objective methodological advances made in the undertaking of is to strengthen the capacity of governments in the household and community surveys in the participat- Sub-Saharan African Region to integrate social dimen- ing countries to ensure an effective cross-fertilization sions in the design of their structural adjustment pro- in the participating countries. The subseries would grams. The World Bank is the executing agency for also include "model" working documents to aid in the the Project. Since the Project was launched in July implementation of surveys, such as manuals for inter- 1987, 30 countries have formally requested to partici- viewers, supervisors, data processors, and the like, as pate in the Project. well as guidelines for the production of statistical The Project aims to respond to the dual concern in abstracts and reports. countries for immediate action and for long-term in- The Policy Analysis subseries will report on the stitutional development. In particular, priority action analytical studies undertaken on the basis of both programs are being implemented in parallel with ef- existing and newly collected data, on topics such as forts to strengthen the capacity of participating gov- poverty, the labor market, health, education, nutrition emments (a) to develop and maintain statistical data and food security, the position of women, and other bases on the social dimensions of adjustment, (b) to issues that are relevant for assessing the social dimen- carry out policy studies on the social dimensions of sions of adjustment. The subseries will also contain adjustment, and (c) to design and follow up social papers that develop analytical methodologies suitable policies and poverty alleviation programs and pro- for use in African countries. jects in conjunction with future structural adjustment Another subseries, Program Design and Implemen- operations. tation, will report on the development of the concep- The working paper series "'Social Dimensions of tual framework and the policy agenda for the project. Adjustment in Sub-Saharan Africa" aims to dissemi- It will contain papers on issues pertaining to policy nate in a quick and informal way the results and actions designed and undertaken in the context of the findings from the Project to policymakers in the coun- SDA Project in order to integrate the social dimensions tries and the international academic community of into structural adjustment programs. This includes economists, statisticians, and planners, as well as the the priority action programs implemented in partici- staff of the international agencies and donors associ- pating countries, as well as medium- and long-term ated with the Project. In the light of the three terrains poverty alleviation programs and efforts to integrate of action of the Project, the working paper series con- disadvantaged groups into the growth process. The sists of three subseries dealing with (a) surveys and focus will be on those design issues and experiences statistics, (b) policy analysis, and (c) program design which have a wide relevance for other countries as and implementation. well, such as issues of cost-effectiveness and ability to reach target groups. v Table of Contents Executive Summary I 1. Introduction 2 2. Data 5 3. Choice of Nutritional Indicators 6 4. Prevalence of Malnutrition and Associations with Socioeconomic Characteristics 8 5. Determinants of Nutritional Status 16 Results 16 Long-Term/Chronic Malnutrition 16 Current Malnutrition 20 6. Conclusions 23 Appendices A. Price Index 25 B. Variable Means and Standard Deviation 27 C. Reduced-form Human and Physical Capital Models 28 References 29 List of Tables 1. Mean Z-scores and percent malnourished, by region and age 9 2. Percent malnourished, by per capita expenditure category and by region 9 3. Indexes of depth of malnutrition and combined depth and head count measures of malnutrition, by per capita expenditure category 10 4. Percent malnourished, by share of land devoted to cash crops 11 5. Percent malnourished, by parents' education categories, per capita expenditures, and urban/rural 12 6. Percent malnourished, by father's and mother's height 13 7. Percent malnourished, by dependency ratio and per capita expenditure class 14 8. Long-term/chronic nutrition functions 17 9. Current nutrition functions 21 Appendix Tables B. Variable means and standard deviation 27 C. Second stage reduced-form consumption expenditure function 28 vii Acknowledgements This paper was prepared for the World Bank's Social Dimensions of Adjustment Project under the guidance of Michel Noel and Christiaan Grootaert. Their assistance in facilitating this research is greatly appreciated. Numerous other persons deserve recognition for their role in performing this work. First are Jong Kim and Jerry Shively who provided excellent research assistance. In addition, the cooperation of Jacques van der Gaag was essential to completing this paper in a timely fashion. Under his direction, Paul Glewwe, Valerie Kozel, and Hailu Mekonnen provided me valuable support in the form of access to data files and insights into the survey itself. Executive Summary Sixteen percent of the preschool-aged children in C6te trition. Children from households that allocate a larger d'Ivoire are stunted, an indicator of long-term (or share of theirland to producing export crops than food chronic) malnutrition. Regionally, the problem is most crops did not display more stunting or wasting. Moth- serious in the Savannah where 23 percent of the chil- ers with more education will be less likely to have dren are stunted, in contrast to only 11 percent being children who suffer from acute malnutrition, when stunted in the urban areas. Wasting, an indicator of controlling for income levels. The education of the current and acute episodes of malnutrition, was prev- father, however, does not confer the same positive alent among 7 percent of the preschool-aged children. benefits upon his children's nutritional welfare, except In order to explore the causes of malnutrition, this as mediated through higher earnings. Parental height, paper estimates reduced-form nutritional status func- especially of women, also has an important impact on tions that include a linear combination of independent long-term nutritional status. The characteristics of the variables, which explain the household's per capita village in which the household resides also plays an consumption expenditures. Instrumented consump- important role in determining levels of malnutrition. tion expenditures are considered a good proxy for In particular, the presence of a nurse and the proximity permanent income, and the findings indicate that they to a doctor are important determinants of malnutri- are an important determinant of long-term (or tion. Likewise, children suffer less malnutrition if they chronic) malnutrition. Income, however, does not reside in villages where dysentery and malaria are not have a significant effect on current (or acute) malnu- serious health problems. 1 1. Introduction Malnutrition is considered to be a problem approach- Even if there is considerable ambiguity as to the ing epidemic proportions in developing countries. effectiveness of raisingincomes to improve nutritional This is an especially serious problem in Sub-Saharan status, as mediated through raising calorie intake, a Africa where recent statistics indicate that the preva- second pathway that links income to nutrition is that lence of malnutrition is worsening in many countries more income raises the demand for other goods and (FAO, 1985). services that are inputs into health (Kennedy, 1989). Reducing malnutrition, both in order to eliminate Consequently, income-augmenting and related wage the human suffering that it causes and protect the and employment policies can be hypothesized to im- quality of human resources, clearly represents a pillar prove nutritional outcomes even if income elasticities of any development policy. However, the elemental of demand for calories are low. information required to formulate enlightened policy In an attempt to cast further light on this important and projects includes understanding the levels, char- debate, this paper will focus on the role of income as acteristics, and determinants of malnutrition, the sub- a determinant of preschool-aged malnutrition in Cote ject of this paper for C6te d'Ivoire. Therefore, the focus d'Ivoire. In order to avoid the potential bias that could of this paper is to explore, descriptively and econome- be introduced if income is jointly determined with trically, these issues employing data from the C6te nutritional outcomes, instrumented, rather than re- d'Ivoire Living Standards Survey. ported, income is employed. In addition, by regressing The most important issue to be addressed is anthropometric indicators directly on instrumented whether raising household consumption expendi- income, the potential of income-augmenting mea- tures (a proxy for income) reduces childhood malnu- sures in improving nutritional status is directly ex- trition? It is hypothesized that raising household plored. However, the issue of the relative importance incomes will increase household caloric intake as well of the mediating role of nutrient demand versus de- as expenditures on other goods and services, thereby mand for other goods and services in bringing about contributing to better nutritional status. improved anthropometric growth performance is not To amplify, there are two hypothesized pathways addressed. I that link income and nutritional status. The first is This paper, furthermore, focuses on the role of a simply that more income leads to higher calorie intake variety of other individual- and household-level char- and perforce, better nutrition. The second is that in- acteristics in determining preschool-aged malnutri- come improves nutrition through its effect on other tion. Among the potential household characteristics inputs, such as health care and practices. that one can include in the reduced-form function is While the biological literature is clear that raising household size and other household structure vari- calorie intake improves nutrition as indicated by an- ables.4 Doing so will also allow us to examine the thropometric indicators, new doubts have recently contention that children of larger households are more been raised as to the responsiveness of calorie intake vulnerable to malnutrition. Studies from Nigeria, (and consequently nutrition) to changes in incomes. Thailand, and India (World Bank, 1975), for example, This has led some to question whether raising incomes suggest a negative correlation between family size and will lead to improved nutrition, as measured through nutrition. In contrast, Sahn (1988) and Behrman and anthropometric outcomes. More specifically, it has Wolfe (1984) show that there are economies of scale of been argued that bias in estimation due to measure- nutrient intake in terms of household size, which ment errorl and the exclusion of variables correlated would tend to favor the nutritional status of children with both income and consumption2 have led to over- from larger families. Therefore, the effect of household stating the response of calories, and consequently nu- size, standardized with respect to its composition, is trition, to changes in income.3 explored. 2 3 Several other hypotheses related to household This study will expand upon the growing evidence structure and composition are explored. The first is that parental height is an important determinant of that older siblings (that is, six years old and above) nutritional status (Strauss, 1987; Thomas, Strauss, and who do not directly compete for nurturing time with Henriques, 1987; Horton, 1988; Kennedy and Cogill, a preschool-aged child and who are also able to pro- 1987). vide some child care assistance, will result in better The second area is the role of parental education, current nutritional status for the young child. Second above and beyond the expected impact of schooling is that the greater the number of adult women, holding on the productivity of market activities. Concerning constant household size and age composition of the this issue, previous research has provided ample evi- household, the better the nutritional status of the dence that education is an important determinant of youngchild.Thislatterissuearisesoutoftheliterature nutritional status (Horton, 1988; Behrman and from Africa on conflicts between males and females Deolalikar, 1988; Haddad, 1987). However, as a conse- on the choice of consumption expenditures (Jones, quence of relying on the conventional reduced-form 1984). model, previous studies were not able to distinguish In order to further explore gender issues, the finding whether education improves nutrition as mediated of Kennedy and Cogill (1987) that children living in through the effects of increased productivity in market households headed by women are better nourished is activities, and perforce, incomes, or through other also examined. However, this is done in urban areas independent channels. For example, these non-in- only, given the absence of women-headed households come channels through which education could im- in rural areas of C6te d'Ivoire. prove nutrition may include altering the household The other gender-specific issue included in this preference function, increasing productivity of house- study is determining whether malnutrition and hold activities, and adopting improved methods of growth retardation is worse among boys or girls. Des- child care.5 In terms of the issue of preferences, Behr- pite the arguments often heard that male children are man and Wolfe (1984) also hypothesize that the edu- favored, both in termsof the intrahousehold allocation cation of the mother is especially vital. This might be of food and other health and nutrition inputs, recent explained by the mother's preference ordering being studies in the Philippines by Haddad (1987) and more oriented toward child nutrition. By employing a Horton (1986) indicate to the contrary. linear combination of independent variables to ex- The effect of a child's birth order on nutritional plain income, which is included as a regressor in the outcomes is also addressed in this paper. The litera- reduced-form nutrition function along with educa- ture suggests that younger siblings are nutritionally tion, this paper explores explicitly the effects of edu- disadvantaged, especially when measured in terms of cation on nutrition, independent of increased long-term malnutrition (Horton, 1988). This is often productivity leading to higher earnings and profits. explained in terms of an additional child straining The remainder of this paper is organized as follows. household resources-that is, the per capita availabil- First, a brief discussion of the data is presented in ity of food and other nutrition inputs are less for the Chapter 2. This is followed in Chapter 3 by a discus- children of higher birth order. However, this issue of sion of the choice of nutritional indicators. Informa- more children spreading household resources more tion on the extent of current (that is, acute) and thinly is implicitly addressed in the models presented long-term (that is, chronic) malnutrition is presented in this paper through defining household expendi- in Chapter 4. These descriptive data are also disaggre- tures in per capita terms and controlling for the age gated by some important characteristics such as re- structure of household members. Nonetheless, other gion, income levels, and various other individual-, reasons have also been posited for an earlier sibling household-, and community-level characteristics. position presenting an advantage. These include bio- Chapter 5 expands upon the insights gained in doing logical explanations (for example, less maternal deple- the descriptive statistics, and models the long-term tion), cultural reasons (for example, the oldest son is and current nutritional status of preschool-aged chil- important in funeral rites), and socioeconomic reasons dren living in urban and rural areas of Cote d'Ivoire. (for example, parents are more dependent upon first The importance of going beyond these descriptive born children for their security in old age) (Horton, statistics is well-illustrated in this paper, as simple 1988). correlations (or the lack thereof) are not always sup- There are two final important areas of inquiry in this ported by the econometric analysis that controls for study. The first is the role of the combined effects of confounding variables. The paper concludes with a genotype and phenotype as captured by parental stat- discussion of the most salient findings and their policy ure in affecting growth performance of the offspring. implications in Chapter 6. 4 Notes Also, see Alderman (1989) for a critical discussion of the issues regarding this emerging debate. 1. See the discussion of Bouis and Haddad (1989) about the 4. The problem of doing so is that family size may be endoge- problemsofthepositivecorrelationbetweenmeasurementerrorsin nous, as the literature on the tradeoffs between the quality and the dependent andindependent variables, as well as for a discussion quantity of children has shown (Becker, 1975). Failure to account for of the large differences in parameter estimates when the model such endogeneity may result in biased estimators. However, iden- employs 24-hour food recall rather than food expenditures. tifying suitable instrumentsfor household size is considerablymore 2. See, for example, Behrman and Wolfe (1984), who stress the difficult than for income. This, coupled with the fact that its exclu- importance of including education in estimating equations; in con- sion introduces a potential bias in the income coefficient, which is trast, see Alderman (1987) who finds that using fixed effects does the focus of this study, suggests that it is reasonable to include not change markedly parameter estimates, thereby indicating that household size, acknowledging the possible bias. no serious bias is introduced as a result of missing variable bias. 5. Rosenzweig and Schultz (1983) argue the importance of edu- 3. See, for example, Bouis and Haddad (1989) and Alderman cation of women in their responsibilities for nurturing and caring (1986) who survey the range of estimates found in the literature. for children. 2. Data The data used in this survey are from the C6te d'Ivoire only 1,008 children under the age of six were mea- Living Standards Survey (CILSS). The data were col- sured. lected by the C6te d'Ivoire Direction de la Statistique The sample for the second round of data, collected in collaboration with the World Bank Living Stan- from March 1986 to March 1987, also consisted of dards Measurement Study. The data, described in de- approximately 1,500 households. A little less than half tail by Grootaert (1986) and Ainsworth and Munioz of those households were drawn randomly from the (1986), are from two survey rounds. The first round of set of households that were sampled during the first data were collected between February 1985 and Feb- round; the remainder were drawn randomly from the ruary 1986. It included approximately 1,500 house- larger national probability sample. Measurements on holds, selected on the basis of a national probability 2,315 children were available from the second round sample, representative of the general population. since the anthropometry was collected throughout. Since the survey did not begin gathering anthropo- Pooling the two rounds of data resulted in anthropo- metric data until the seventh month of the first round, metric data from 3,323 children. 5 3. Choice of Nutritional Indicators This paper employs anthropometric indicators to contrast to stunting that measures long-term nutri- measure nutritional status. Anthropometry is recog- tional well-being.7'8 nized as the technique of choice for deternining defi- Of course, nutritional indicators such as height-for- cits in food energy and protein that manifest age and weight-for-height of children only find mean- themselves in stunting (that is, slow linear growth) ing in terms of comparisons with normative and wasting (that is, being emaciated). In combina- standards. Once again in keepingwith convention, the tion, inadequate food energy and protein intake have National Center for Health Statistics (NCHS) reference been shown to be the most widespread and debilitat- population is used as the benchmark (U.S. Public ing nutritional problem in developing countries. Health Service, 1976).9 In addition, it is necessary to The assessment of nutritional status is based on two select a cutoff point, below which a child is classified indicators, in keeping with the conventional standards as being previously or currently malnourished. The of classification (WHO, 1983).The first, height-for-age, convention for doing so is to employ a value -2 Z- measures the degree to which linear growth is re- scores below the median value for the three indicators tarded due to inadequate nutrient intake and poor discussed above.10 The -2 Z-score cutoff point is con- health. The stunting of growth is a physiological re- structed by taking the median value of a reference sponse to nutrient deficits, and likely represents an population's weight-for-height or height-for-age of a attempt to maintain vital bodily functions, a normal given age cohort. Fifty percent of the children in the level of physical activity, and a proportionate weight reference population have measures greater than this for a given stature or length. When a child is stunted, value, and half of the children will be below this value. it is an indicator of chronic, long-term, and/or previ- Then a range, plus and minus two standard deviation ous malnutrition.6 The implication, of course, is that units, is set around that median. This includes 97.6 identifying the socioeconornic determinants of stunt- percent of the children in the population. A child ing or previous malnutrition is difficult in the absence whose weight-for-height or height-for-age falls within of lagged information on the circumstances of the this range is classified as being normally nourished. If child in the preceding months or years. It is necessary, a child falls two standard deviation units, alternatively therefore, to resort to the reasonable assumption that referred to as -2 Z-scores below the median, then the there is high correlation between the present socioeco- child is classified as tndernourished. In essence, using nomic environment and that which prevailed in pre- Z-scores allows the analyst to make a probability state- vious periods, in order to assess the determinants of ment based on a comparison of the measurements of long-term malnutrition. a child from Cote d'Ivoire with the healthy reference A child's weight for a given height is an indicator of population. Specifically, there is less than a 0.023 prob- leanness, or wasting. When a child's weight-for-height ability that a child with a height-for-age or weight-for- is below the normative standard, it is a sign of acute height of less than -2 Z-scores would be found in the malnutrition and physiological stress. Wasting is a healthy reference population. reflection of the current nutritional status of a child, in 6 7 Notes groups; and that there is no strong evidence to suggest that the genetic growth potential is different among children from different 6. For the remainder of this paper, chronic malnutrition, long- ethnic and racial groups. Thus, healthy preschool-aged children in term malnutrition, previous malnutrition, and stunting are used Africa should grow at the same rate as those in the U.S. interchangeably. 10. A Z-score is the standard deviation of a given indicator for 7. For the remainder of the paper, acute malnutrition, current an individual. Children with Z-scores less than -2 are classified as malnutrition, and wasting are used interchangeably. malnourished. Z-scores are calculated as follows: 8. A third classification of malnutrition, being concurrently stunted and wasted, is not employed, given that only around one Z = (Mo - Me)/SDe percent of the population falls in this category. 9. Martorell and Habicht (1986) and WHO (1983) discuss the whereMo is the observed height or weight of individuals in a given appropriateness of employing the NCHS growth standards, based age or height group, Me is the expected median height or weight of on the measurement of children from the United States, to assess that group of the reference population, and SDe is the standard nutritional status amongdeveloping countrypopulations. Thebasic deviation of the measurements for that group of the reference pop- argument is that the NCHS standards are based on a heterogenous ulation. population, randomly selected from different economic and ethnic 4. Prevalence of Malnutrition and Associations with Socioeconomic Characteristics At a national level, 16.2 percent of the preschool-aged as per capita expenditures rise, there is a decline in the population in Cote d'Ivoire are stunted, indicating percentage of children who suffer from long-term long-term or previous malnutrition; and 7.1 percent of malnutrition. Nationally, 19.3 percent of the children the preschoolers suffer from wasting, a measure of in the lowest expenditure group are chronically mal- current malnutrition (Table 1). This low incidence, nourished, 17.0 percent in the second quintile, and relative to those found in other poor African and Asian only 12.5 percent a,mong the highest 20 percent of the countries, is an initial indicator of a combination of a expenditure distribution.12 Across these expenditures relatively high level of household food security (that groups, the percentage of acutely malnourished chil- is, adequate access to food) and/or low levels of mor- dren declines modestly from 8.7 percent in the lowest bidity. quintile to 5.3 percent among the upper 20 percent of Regionally, stunting (that is, previous or long-term the expenditure distribution. malnutrition) is most prevalent in the poorest region In order to examine the percent of malnourished of Cote d'Ivoire, the Savannah. The probability of children by expenditure group, two sets of quintiles being stunted is lowest in Abidjan and other cities. The were constructed--one for urban areas and the other same regional pattern does not hold for acute (that is, for rural areas. This was necessitated by virtue of the current) malnutrition. In Abidjan, 9.4 percent of the fact that in the Savannah and Forest regions, 8.3 and children are classified as suffering from acute malnu- 11.1 percent, respectively, of the children reside in trition, a figure that is as low as 3.4 percent in the West households falling in the highest two quintiles of the Forest. national expenditure ranking; while in Abidjan less Wasting, which indicates a current episode of acute than 4.0 percent of the children fall in the lowest malnutrition, is primarily a problem among children quintile of the national expenditure ranking. Thus, if 12 to 23 months of age, with 14.6 percent of the chil- the national expenditure quintiles are employed, the dren being wasted in this cohort. The vulnerability of cell sizes in the upper groups are too small to infer children during this weaning period has been docu- anything meaningful in rural areas, while the same is mented extensively in other developing countries. true of the cell sizes in the urban areas in the lowest Also as observed elsewhere, the incidence of stunting, expenditure groups. which indicates previous or long-term malnutrition, In Abidjan, one finds a marked decline in the per- increases from 7.9 percent among children less than cent of stunted children between the lowest 60 and one year of age to 20.2 percent among children three upper 40 percent of the expenditure distribution. In years of age. Thereafter the level of chronic, long-term other urban centers, there is also a drop in the level of undernutrition declines to 15.6 percent among those long-term malnutrition that is pronounced between children 48 to 60 months of age. 1 the fourth and fifth quintile. Wasting, reflecting cur- Examining the anthropometric data by gender re- rent malnutrition, also shows a marked decline over vealed no evidence that either boys or girls are better the expenditure distribution in Abidjan, falling from nourished. This applies even when the gender com- 19.3 in the lowest quintile to 4.0 percent in the highest parisons are further disaggregated by age cohorts, and quintile. In the rural areas, there is no clear pattern the regions for which data are presented in Table 1. between expenditures and child malnutrition. This is Identifying the sociodemographic characteristics of the case for both stunting and wasting, in the Forest malnourished children and the households in which and Savannah regions. they reside is a prerequisite to any targeted effort to One limitation of the use of Z-score cutoff points is improve nutritional well-being. Table 2 indicates that that simply reporting the percentage of children that 8 9 Table 1. Mean Z-scores and percent malnourished, by region and age Region Age Group Indicator Abidjan Other Cities WestForest East Forest Savannah All (months) (percent) 0-11 Stunted 6.3 4.9 13.0 8.4 8.7 7.9 Wasted 12.5 11.3 5.9 12.9 9.5 10.7 N 96.0 142.0 85.0 155.0 127.0 605.0 12-23 Stunted 16.9 13.4 19.8 20.6 24.8 19.0 Wasted 18.1 12.4 7.0 17.9 17.3 14.6 N 83.0 5.0 86.0 112.0 93.0 479.0 24-35 Stunted 10.8 15.9 20.5 17.9 28.7 18.7 Wasted 5.9 6.1 1.2 10.6 5.2 6.1 N 102.0 132.0 83.0 123.0 115.0 555.0 36-47 Stunted 14.4 10.7 26.4 18.0 29.5 19.1 Wasted 7.8 3.1 1.4 3.7 0.9 3.4 N 90.0 131.0 72.0 161.0 112.0 566.0 48-60 Stunted 7.9 11.7 16.9 17.3 20.3 15.1 Wasted 3.7 1.9 2.3 4.3 0.9 2.7 N 164.0 265.0 172.0 300.0 217.0 1,118.0 ALL Stunted 11.4 11.3 18.7 16.9 23.2 16.2 Wasted 9.4 6.0 3.4 9.0 7.1 7.1 N 535.0 775.0 498.0 851.0 664.0 3,323.0 Notes: Stunted, or previously malnourished, is defined as height-for-age Z-score -2 Wasted, or currently malnourished, is defined as weight-for-age Z-score -2 Table 2. Percent malnourished, by per capita expenditure category and by region Per Capita Urban Areas Rural Areas Expenditure Other West East Category Abidjan Cities Forest Forest Savannah National (percent) 1 Stunted 12.5 12.6 14.6 17.5 31.3 19.3 Wasted 19.3 7.8 2.9 18.2 6.9 8.7 N 88.0 167.0 103.0 137.0 160.0 657.0 2 Stunted 17.1 10.7 23.9 16.4 19.1 17.0 Wasted 11.7 7.4 1.1 8.6 8.6 8.2 N 111.0 149.0 88.0 152.0 162.0 886.0 3 Stunted 16.0 11.8 12.4 16.9 17.6 15.3 Wasted 9.4 3.9 1.0 9.6 5.9 8.3 N 106.0 152.0 105.0 177.0 119.0 659.0 4 Stunted 7.5 13.9 21.4 15.2 22.1 17.5 Wasted 4.7 4.6 5.4 7.6 4.8 5.2 N 106.0 151.0 112.0 184.0 104.0 659.0 5 Stunted 4.8 6.7 22.2 19.4 24.4 12.5 Wasted 4.0 5.2 6.7 4.2 8.4 5.3 N 124.0 134.0 90.0 191.0 119.0 656.0 ALL Stunted 11.4 11.3 18.7 17.1 23.2 16.3 Wasted 9.3 5.8 3.4 9.2 7.1 7.1 N 535.0 753.0 498.0 841.0 664.0 3,291.0 Notes: Per capita expenditure categories in this table are different for urban and rural areas, and for the national figures as well. The quintiles for the urban areas were determined by ranking all children living in Abidjan and other cities, according to their per capita expenditures, deflated by a price index based on food prices. Therefore, 20 percent of the children in urban areas fall in each of the quintiles. In a similar vein, the quintiles for the rural areas were determined by ranking all children living in the East and West Forest, and Savannah, according to their per capital expenditures, deflated by a price index based on food prices. Twenty percent of the children in rural areas fall in each quintile. The national quintiles are based on ranking all the children measured, and ranking them according to the deflated per capita expenditures. The actual cutoff points employed, in terms of deflated annual per capita expenditures (in CFA) are as follows: Urban: (1) Less than 100,799; (2) 100,800-144,799; (3) 144,800-201,829; (4) 201,830-315,373; (5) greater than 315,374. Rural: (1) Less than 67,837; (2) 67,838-92,179; (3) 92,180-121,682; (4) 121,683-175,029; (5) greater than 175,030. National: (1) Less than 77,135; (2) 77,136-107,143; (3) 107,144-145,984; (4) 145,985-226,429; (5) greater than 226,430. 10 are malnourished does not provide any insight into Kanbur (1987), who employed the measure for in- the severity of wasting and stunting.'3 Therefore, in come-determined poverty lines in C6te d'Ivoire, this Table 3, additional measures of malnutrition that take indicator essentially raises the extent to which an into account the severity of the problem are presented. individual's weight-for-height or height-for-age fall To amplify, two further indicators are employed. short of the cutoff point to the second power, sum- The first and third columns measure the depth or ming these squared deviations over the malnourished severity of malnutrition, conditional upon being cate- children, and thereafter normalizes in terms of the gorized as such. The national data suggest that the numbers of the entire population of well- and mal- average gap between the cutoff point (that is, -2 Z- nourished children. scores) defining normal height-for-age or weight-for- Results at the national level once again do not indi- height and the actual mean height-for-age or cate any clear relationship between the value of this weight-for-height among the currently malnourished composite index for clhronic or acute malnutrition and population is stable throughout the expenditure expenditure quintile. When these same variables are groups. examined by urban and rural areas, one finds that in In the second and fourth columns, a measure that rural areas, none shovw any association with per capita combines the concern over the numbers of malnour- expenditures. In urban, however, all the variables ished, with the severity of malnutrition, according a show a marked decline with the level of expenditures. greater sensitivity to those who are more severely This indicates that among those children suffering malnourished, is presented (that is, M2). Following from long-term and current malnutrition, the severity from Foster, Greer, and Thorbecke (1984), who devel- is worse the lower the expenditure group. And like- oped such a concept for measuring poverty, and wise, the composite index of the prevalence and sever- Table 3. Indexes of depth of malnutrition and combined depth and head count measures of malnutrition, by per capita expenditure category Per Capita Height-for-Age Weight-for-Height Expenditure Category M1 M2 Ml M2 All Regions 1 0.493 0.782 0.322 0.152 2 0.516 0.919 0.368 0.205 3 0.382 0.442 0.390 0.282 4 0.398 0.511 0.261 0.092 5 0.432 0.448 0.377 0.159 Urban 1 0.487 0.468 0.383 0.315 2 0.406 0.518 0.293 0.146 3 0.352 0.305 0.255 0.168 4 0.367 0.316 0.262 0.044 5 0.245 0.069 0.249 0.046 Rural 1 0.494 0.907 0.313 0.160 2 0.521 0.996 0.339 0.143 3 0.444 0.622 0.423 0.216 4 0.402 0.588 0.438 0.240 5 0.478 0.908 0.407 0.228 Note: -2-z Ml = -2 M2 = - 2 r--) *10 ~-2 where: z = the mean Z-score among those classified as being currently or previously malnourished; zi = the Z-score of individual i; q = the number of children classified as being currently or previously malnourished; n = the total number of children in the population. 11 ity of currently and chronically malnourished children The probability of being malnourished is also exam- shows a strong negative correlation with the per capita ined in terms of the educational achievement of the expenditure level of the child's household. mother and father (see Table 5). In rural areas, 12.9 A variety of other household-level factors that may percent of the children of mothers who have received be related to nutritional outcomes were explored. some education are stunted or previously malnour- First, the nutritional status of children residingin rural ished, as opposed to 21.1 percent for mothers having areas was examined by both farmsize and farmsize per no education. The level of wasting or current malnu- capita (the latter is found in Table 4). There is no trition is not significantly lower among children of evidence that landholdings of the household is related mothers with some education. In rural areas, there is to nutritional status of children, even when normal- no significant difference in the long-term and current ized by the size of the household. In addition to exam- nutritional status of children of fathers with no, versus ining the effects of the quantity of landholdings on some, education. nutritional status, the relationship between the share Education is a major determinant of household ex- of the total land in cultivation devoted to the tradi- penditures.Y Amongchildrenfromurbanhouseholds tional export cash crops, coffee, cocoa, and cotton, and that fall in the lowest expenditure group, 71.5 percent nutrition was also explored. One can see that for the of their fathers had not completed the first grade; the sample of all households that reported cultivating comparable figure in the highest expenditure quintile land, there is no association between the share of land is less than 20.0 percent. Therefore, it is important to devoted to cash crops and nutritional status (see Table examine the effects of education on nutritional status, 4). When these shares are examined stratified by total disaggregated by expenditure class, in order to deter- land cultivated per capita, there is likewise no relation- mine whether there are any independent effects of ship between land use and the level of stunting and education, not mediated by returns to labor. wasting. Not shown inTable4, although also notewor- The data suggest that a child will be less likely to be thy, is that no association was observed between pre- chronically undernourished if her/his mother has re- school-aged nutrition and the share of land in cash ceived some level of education. This relationship is crops, when disaggregated by the per capita expendi- strongest for the lower three expenditure quintiles. ture categories. Note, for example, that in the rural areas within the Table 4. Percent malnourished, by share of land devoted to cash cropsa Per Capita Percent of Land Cultivated Devoted to Cash Crops Land Category < 5.6 5.7-25.91 25.92-42.02 42.03-5830 > 583 ALL (percent) 1 Stunted 21.4 17.9 20.3 11.8 30.0 20.2 Wasted 7.5 9.0 7.2 8.8 15.0 8.3 N 187.0 67.0 69.0 34.0 20.0 377.0 2 Stunted 22.1 25.8 19.5 14.6 18.7 20.6 Wasted 9.1 7.7 8.3 6.1 10.4 8.1 N 77.0 105.0 72.0 82.0 48.0 384.0 3 Stunted 25.6 10.9 20.2 17.8 20.5 18.9 Wasted 2.3 7.3 4.0 5.6 9.0 5.8 N 43.0 55.0 99.0 107.0 78.0 382.0 4 Stunted 25.0 21.0 17.9 10.7 15.6 17.2 Wasted 8.4 7.4 10.2 4.0 8.2 7.7 N 36.0 81.0 78.0 75.0 109.0 379.0 5 Stunted 22.5 25.3 12.7 25.3 23.5 22.2 Wasted 7.5 4.2 8.4 6.6 7.3 6.9 N 40.0 71.0 71.0 75.0 136.0 393.0 ALL Stunted 22.4 21.1 18.3 16.6 20.5 19.8 Wasted 7.3 7.1 7.5 5.9 8.7 7.3 N 383.0 379.0 389.0 373.0 391.0 1,915.0 a. Cash crops refer to the traditional export crops of coffee, cocoa, and cotton. b. Per capita land category corresponds to the following size landholdings (hectares): (1) 0.010-0.384; (2) 0.385-0.599; (3) 0.600-0.874; (4) 0.875-1.329; (5) > 1.330. Table 5. Percent malnourished, by parents' education categories, per capita expenditures, and urban/rural Per Father's Education Mother's Education Capita Urban Rural Urban Rural Expenditure S S ClaSsa b Some Some Some Some None Elenentary Elementary None Elemnentary None Elenentary Elenentary None Elerentary (percent) I Stunted 14.2 8.6 8.6 24.1 15.9 12.9 7.4 18.2 24.1 10.7 Wasted 10.8 11.4 20.0 10.0 7.3 9.2 22.2 36.4 9.3 12.5 N 176.0 35.0 35.0 311.0 82.0 217.0 27.0 11.0 344.0 56.0 2 Stunted 9.6 13.9 24.5 17.9 22.0 10.1 14.7 34.8 20.7 10.90 Wasted 7.0 5.6 18.9 7.8 4.6 9.5 7.4 13.0 7.1 6.32 N 157.0 36.0 53.0 268.0 132.0 168.0 68.0 23.0 338.0 64.0 3 Stunted 13.0 11.6 16.5 17.8 11.9 10.1 16.0 29.2 17.6 8.3 Wasted 6.9 7.0 5.0 5.2 7.9 7.6 4.0 4.2 6.4 5.6 N 131.0 43.0 79.0 270.0 126.0 158.0 75.0 24.0 329.0 72.0 4 Stunted 6.7 16.7 12.0 19.5 17.6 11.9 14.5 4.4 19.4 16.8 Wasted 11.2 0.0 1.0 6.1 6.9 6.7 2.9 2.2 6.3 6.3 N 89.0 48.0 100.0 261.0 131.0 134.0 69.0 45.0 304.0 95.0 5 Stunted 11.1 3.2 5.3 22.4 19.3 11.2 4.5 2.1 23.3 15.1 Wasted 13.3 3.2 3.3 6.6 5.2 9.0 0.0 4.1 7.6 0.0 N 45.0 31.0 152.0 259.0 135.0 89.0 67.0 97.0 314.0 86.0 ALL Stunted 11.4 11.4 11.7 20.5 17.5 11.4 12.1 10.5 21.1 12.9 Wasted 9.2 5.2 6.5 7.2 6.3 8.5 5.2 6.5 7.4 5.6 N 598.0 193.0 419.0 1,369.0 606.0 766.0 306.0 200.0 1,629.0 373.0 a. Per capita expenditure classes for urban and rural areas correspond to the figures in Table 2. b. None refers to persons who have failed to complete the first year of school. 13 first expenditure group, 24.1 percent of the children of where children of fathers who have gone beyond ele- mothers with no education are previously malnour- mentary school have a higher probability of being ished; the comparable figure for children of mothers stunted, are hard to explain; they likely represent a who have at least completed one year of schooling is structural problem with the data. More in keeping 10.7 percent. There is much less convincing evidence with expectations, among the highest expenditure of a relationship between the father having attended quintile, 11.2 percent of the children are previously school, and the probability of the child living in rural malnourished if the mother has no education, while areas being previously malnourished. For mothers the figures are only 4.5 and 2.1 percent if the mother and fathers, there is no relationship between their has attended secondary school, and exceeded second- education and wasting or current nutritional status of ary school, respectively. the child across all expenditure groups. And within The association between nutrition status and the expenditure groups, it is only in the fifth quintile that educational achievement of the head of the household the child is less likely to be wasted if the mother has (when different from the father or mother), as well as some education. the senior wife in polygamous families was also exam- In urban areas, it is possible to distinguish between ined. No clear patterns emerged. children of parents who have not completed the first The role of parental height, in relation to long-term year of schooling, those whose parents have attended nutritional status, is of considerable importance. In (although not necessarily completed) elementary Table 6, one can see that if both of a child's parents school, and those whose parents have gone beyond were in the lowest height quintile, the probability of elementary schooling. Across all expenditure groups, being previously malnourished is 38 percent, as op- educational achievement among mothers and fathers posed to 9 percent if both the child's parents are in the shows no statistically significant relationship to the tallest 20 percent of the population. percentage of children who are stunted or wasted. One would expect the genetic influence on stature Within a given expenditure quintile, there is likewise from the father and mother to be the same. Addition- no clear pattern to support that previous (that is, long- ally, for the mother, the environmental impact on term) or current malnutrition is associated with edu- genetic potential, the phenotype, is also expected to cational achievement. Some surprises, such as those in affect the offspring. This would, in theory, suggest that the second expenditure group in the urban areas the mother's stature should be more influential on the Table 6. Percent malnourished, by father's and mother's height Father's Height Mother's Height Category Category 1 2 3 4 5 ALL (percent) 1 Stunted 37.8 25.3 13.7 20.3 15.9 25.0 Wasted 8.9 8.5 8.2 2.9 11.1 8.1 N 135.0 95.0 73.0 69.0 63.0 435.0 2 Stunted 26.4 12.6 16.5 13.1 7.9 15.3 Wasted 1.4 3.4 4.1 7.9 2.6 4.0 N 72.0 87.0 121.0 76.0 76.0 432.0 3 Stunted 23.3 26.3 23.3 17.0 11.2 20.7 Wasted 6.5 6.1 3.5 3.0 14.6 6.9 N 107.0 99.0 86.0 69.0 89.0 450.0 4 Stunted 18.2 22.1 22.0 15.3 14.6 18.6 Wasted 11.7 9.3 20.0 8.2 3.7 10.9 N 77.0 86.0 100.0 85.0 82.0 430.0 5 Stunted 10.9 25.7 8.9 7.8 9.0 11.7 Wasted 8.1 13.6 6.9 3.9 8.2 7.8 N 37.0 74.0 101.0 103.0 122.0 437.0 ALL Stunted 26.4 22.4 16.9 14.1 11.3 18.2 Wasted 7.5 7.9 8.6 5.2 8.1 7.5 N 428.0 441.0 481.0 402.0 432.0 2,184.0 Notes: Father's height categories are (in meters): 1. less than 1.634; 2.1.634-1.670; 3. 1.671-1.702; 4. 1.703-1.745; 5. greater than 1.745. Mother's height categories are (in meters): 1. less than 1.539; 2.1.539-1.572; 3. 1.573-1.600; 4. 1.601-1.636; 5. greater than 1.636. 14 child's height-for-age than that of the father, an issue riety of factors, including expenditure groups, farm explored in greater detail in the next section of this size, and region. paper. The lack of such associations, however, are difficult Household size showed no relationship with nutri- to interpret. First, the observed behavior of house- tional indicators. Given that households are often holds (for example, the share of income from con- comprised of more than two adults, and offspring sumption of own production) is strongly associated from more than one set of parents, this absence of with other factors that affect nutrition; these cannot be correlation between the nutritional status of the child captured employing descriptive statistics. Second, and the size of her/his family was not surprising. outcomes such as the share of income from consump- The relationship between the number of children tion of own production is a manifestation of house- and the number of adults was explored to determine hold choices concerning factors such as how much whether childrenfromhouseholdswithhigher depen- time the household members work in their own fields dency ratios had a higher probability of being mal- versus hiring themselves out as wage labor; which nourished. There is little strong evidence that the crops to grow; and the share of farm output to market probability of being chronically or currently malnour- for cash income versus the share to store on the farm ished is related to the dependency ratio even when for future consumption, which is partly determined by disaggregated by expenditure group (see Table 7). the efficiency and competitiveness of local markets. Likewise, a variety of other potential associations be- But most important is that the same exogenous tween nutrition and a variety of other socioeconomnic factors that may result in a high proportion of house- factors were examined employing descriptive statis- hold income from home consumption, may also con- tics. Most are not reported given that no relationship tribute to a high prevalence rate for malnutrition. For was observed. One such avenue of exploration was the example, consider the farm household in a remote area issue of whether the source of income in rural areas of Cote d'Ivoire that is less likely to market their farm was associated with levels of malnutrition. In particu- output due to the thin and noncompetitive commod- lar, this involved determining whether the share of ity market. It is also likely that the opportunities for agricultural production that is consumed at home, the off-farm employment will be minimal. Consequently, proportion of food expenditures that is from own the share of income from home consumption may be production, and the share of income that is received quite high. These same remote spatial characteristics from the production of cash crops, are related to nu- may portend a lack of health infrastructure and envi- tritional outcomes. No statistically significant associa- ronmental sanitation fac ilities leading to high levels of tion was observed between such derived variables and malnutrition. Such a circumstance may result in a nutritional status, even when disaggregating by a va- positive correlation between the share of income from Table 7. Percent malnourished, by dependency ratio and per capita expenditure class Per Capita Expenditure Ratio of Household Members 14 to Household Members > 14 ClaSSa 0.6 < 0.61-0.8 0.81-1.0 1.1-1.5 > 15 ALL (percent) 1 Stunted 19.3 23.1 19.1 18.6 14.9 19.3 Wasted 8.4 10.8 11.0 4.0 8.1 8.7 N 166.0 130.0 163.0 124.0 74.0 657.0 2 Stunted 17.2 17.5 15.7 14.7 22.4 17.0 Wasted 12.7 6.4 6.4 10.5 3.5 8.2 N 134.0 126.0 172.0 143.0 85.0 660.0 3 Stunted 14.6 17.2 11.5 15.8 16.5 15.3 Wasted 8.8 10.3 8.9 5.8 7.0 8.4 N 137.0 174.0 113.0 120.0 115.0 659.0 4 Stunted 14.9 22.2 14.1 21.2 18.0 17.5 Wasted 5.6 3.7 4.3 6.1 6.3 5.2 N i61.0 108.0 163.0 99.0 128.0 659.0 5 Stunted 9.2 8.7 13.3 10.6 17.1 12.5 Wasted 4.1 5.4 5.6 5.6 5.5 5.3 N 98.0 92.0 143.0 142.0 181.0 656.0 ALL Stunted 15.5 18.1 15.0 15.8 17.7 16.3 Wasted 8.0 7.8 7,2 6.5 6.0 7.2 N 696.0 630.0 754.0 628.0 583.0 3,291.0 a. Per capita expenditure classes correspond to the national figures in Table 2. i5 home consumption and malnutrition. This, however, Notes is obviously not causal. 11. The fact that nutritional status varies markedly with the age These problems of simultaneity and choice are es- of the child admonishes the analyst to examine the age structure of pecially severe concerning the discussion of income the population before presenting descriptive statistics such as nutri- sources. However, the limitations of the descriptive tional status by region or by expenditures group. Doing so is neces- analysis presented throughout this section revolves sary to ensure that the age composition of the populations are around the fact that correlation, or the lack thereof, is roughly comparable, and do not bias the results. difficult to. interpre. One should becareful not to draw 12. Per capita expenditures were deflated across communities difficult to interpret. One should be and across survey rounds to account for spatial and temporal price any causal inferences from these relationships. Sim- differences. See Appendix A for a discussion of how the deflators ply, the direction and degree of causation is not deter- were constructed. minable without resorting to econometric analysis. 13. This issue has been discussed at some length in terms of Thus, the descriptive statistics provide only limited income-determinedpovertymeasures bySen (1976), Foster, Greer, and Thorbecke (1984), and most recently in context of C6te d'Ivoire, guidance into the consequences of various policies on by Kanbur (1987). nutritional status. The need for integrating these con- 14. The role of education in determnining household expendi- cerns into a more coherent model is manifest. tures is presented more precisely in the models in Appendix B. 5. Determinants of Nutritional Status This section of the paper presents the econometric In (2) and (3), we therefore estimate a reduced-form analysis of the determinants of nutritional status. nutrition function conditional upon expenditures.17 A Models are estimated separately for urban and rural problem with this forrnulation, however, is that con- areas, as well for the causation of previous and current sumption expenditures is endogenous to the model. malnutrition. In addition, it is a function of many of the same char- The model of nutritional status can be derived from acteristics that simultaneously determine child the maximization of the following household utility growth. Therefore, this paper includes a linear combi- function: nation of independent variables to explain household consumption expenditures. This procedure circum- U = u (X, L, Q) (1) vents the potential problem of joint endogeneity of expenditures and nutritional outcomes. By in- where X and L are the household's consumption of a strumenting per capita consumption expenditures, composite good and leisure, respectively, and Q is the one ensures that the expenditure parameter estimates quality of children, to be represented by their nutri- and the standard errors are unbiased. tional status. The composite consumption good is de- rived from the vectors of the consumption Results expenditures of individual household members.'5 The assumption is that good nutrition, as represented by The results are presented separately for the determi- the vector of nutritional status of preschool-aged chil- nants of long-term (that is, chronic) malnutrition, and dren, is desirable in its own right; and it is likewise current (that is, acute) malnutrition. For each, a base assumed that households make consumption deci- model is presented for urban and rural areas. Then, a sions on the basis of reasons other than nutrition (Pitt second set of models are shown that examine explic- and Rosenzweig, 1985). itly the non-linearities of education on nutrition and The utility function is maximized subject to several the interactions of education and income on nutri- technology constraints. In addition, the reduced form tional status, as well as the effect of community char- functions, conditional on per capita expenditures, for acteristics in affecting levels of malnutrition. In each individual in the household can be represented addition, a third set of models is presented for rural as follows: areas only that looks at the impact of landholding and export cropping on nutritional status. Li = L (X, Ai, Zi, Ci, ei) (2) Long-Tenn/Chronic Malnutrition Wi = W (X Ai, Zi, Ci, ei) (3) The results of the nutritional status functions are where Li is long-term nutritional status of child i, Wi is reported in Table 8 for urban and rural areas with short-term nutrition status, X is the household's per variable definitions, means, and standard deviation in capita consumption expenditures, Ai is the vector of Appendix Table B. The reduced-form models for con- the individual's i's personal attributes that may influ- sumption expenditures (that is, reduced-form human ence his/her nutrition, Zi and Ci are the vector of and physical capital functions) are reported in Appen- observable household and community characteristics, dix C. respectively, thatmay influence child i'snutrition, and First, and of greatestiimportance, is thatinbothrural ei is the child-specific random disturbance term.16 and urban areas, Model I indicates that income is an 16 Table 8. Long-term/chronic nutrition functions Dependent Variable: Height-for-Age Z-Score Urban - I Rural - I Urban - II Rural - II Rural - III Inzdependent Variables Parameter Standard Parameter Standard Parameter Standard Parameter Standard Parameter Standard Estimate Error Estimate Error Estimate Error Estimate Error Estimate Error Intercept -18.665 2.244 -9.736 2.026 -18.464 2.690 -8.795 2.254 -8.440 2.377 Mother's height 5.899 0.996 3.761 0.866 5.724 1.001 3.645 0.889 3.157 0.921 Father's height 2.672 0.877 0.331 0.790 2.606 0.881 0.208 0.816 0.494 0.842 Female head dummy 0.666 0.712 0.659 0.712 Child 0-6 month dummy 1.235 0.215 0.764 0.203 1.219 0.215 0.749 0.209 0.724 0.216 Child 7-12 month dummy 1.051 0.247 0.450 0.207 1.049 0.248 0.436 0.211 0.406 0.217 Child 25-36 month dummy -0.307 0.194 -0.321 0.176 -0.325 0.194 -0.407 0.182 -0.441 0.188 Child 37-48 month dummy -0.025 0.197 -0.107 0.181 -0.025 0.197 -0.130 0.186 -0.170 0.191 Child 49-60 month dummy 0.157 0.186 0.083 0.165 0.151 0.187 0.039 0.170 -0.027 0.175 January-March dummy -0.383 0.179 -0.027 0.142 -0.401 0.180 -0.339 0.167 -0.353 0.169 April-June dummy -0.239 0.188 0.150 0.166 -0.280 0.192 -0.074 0.174 -0.069 0.177 July-Septemberdummy -0.026 0.144 0.236 0.139 -0.044 0.144 -0.014 0.156 0.045 0.159 No. children 0-5 yrs -0.053 0.050 0.122 0.048 -0.050 0.050 0.142 0.049 0.158 0.051 No. children 6-14 yrs 0.105 0.057 0.083 0.056 0.111 0.057 0.106 0.058 0.115 0.059 No. female > 14 yrs 0.076 0.047 0.001 0.053 0.070 0.047 0.024 0.054 0.008 0.055 Ln expenditures per capita 0.387 0.107 0.189 0.089 0.408 0.153 0.218 0.112 0.220 0.122 ,,A Child gender dummy var. -0.237 0.111 -0.089 0.098 -0.249 0.111 -0.128 0.101 -0.142 0.104 '4 Abidjan dummy var. 0.124 0.117 0.121 0.118 1985 dummy var. 0.215 0.137 -0.129 0.119 0.230 0.138 -0.167 0.133 -0.164 0.139 Household size -0.005 0.033 0.012 0.036 -0.005 0.033 0.002 0.037 0.015 0.038 Mother's age -0,005 0.006 -0.001 0.004 -0.005 0.006 -0.003 0.004 -0.004 0.004 Birth order -0.072 0.049 -0.050 0.044 -0.070 0.049 -0.059 0.046 -0.070 0.047 West Forest dummy var. 0.064 0.128 -0.142 0.132 -0.211 .0.137 East Forest dummy var. -0.051 0.135 -0.017 0.142 -0.151 0.160 Mother's education 0.009 0.018 0.036 0.031 -0.566 0.312 0.061 0.709 -0.413 0.806 Father's education -0.026 0.015 -0.016 0.017 0.311 0.276 0.280 0.369 0.572 0.395 Mother's ed x Ln exp per capita 0.048 0.026 -0.021 0.060 0.021 0.067 Father's ed x Ln exp per capita -0.031 0.024 -0.031 0.032 -0.059 0.034 (Mother's ed) -0.003 0.003 0.030 0.015 0.030 0.019 (Father's ed) 0.004 0.003 0.008 0.005 0.011 0.005 Malaria -0.350 0.228 Dysentery -0.551 0.239 Chicken/smallpox/measles 0.406 0.152 Distance to doctor -0.221 0.080 Nurse dummy 4.856 2.832 Nurse x Ln per capita exp. -0.368 0.246 Land per capita 0.037 0.077 Export crop share -0.002 0.003 422 2 2 2 R =0.1842 R0.0697 R = 0.1897 R =0.0991 R = 0.1070 18 important determinant of long-term nutritional status. height coefficient is smaller than that observed in the This is shown by the positive and significant parame- cities, but nonetheless highly significant. In contrast, ter estimate indicating that as the expenditures of the in rural areas, the height of the father has no explana- household increase, so do the height-for-age Z-scores tory power in predicting a child's stature. of the preschool-aged children. In rural areas, the mag- The household size variable was not significant in nitude of the significant expenditure coefficient is at either the urban or rural areas. This suggests the ab- proximately half that in the urban areas. This indicates sence of any econormies-of-scale effect. It is interesting, that expenditures play a less important role in rural however, that some of the household composition areas in affecting nutritional outcomes. Among the variables proved significant. Specifically, the rural bottom quintile of the income distribution, elasticities model shows that the greater the number of children of height-for-age Z-scores with respect to income are less than five years of age for a household of a given 1.07 and 0.28 for children in urban and rural areas, size and number of adults, the higher the height-for- respectively. age Z-score. This finding may seem counter-intuitive, A second feature of these models is that once one given the expectation that there might be some com- controls for household income, mother's education petition for child care and attention between, say, a has no measurably significant impact on the nutri- one-year-old and a three-year-old child, or vice versa. tional status of the child, although the sign of the A plausible explanation for this finding is, however, coefficient is positive. In urban areas, increased that the greater the number of young children in any schooling of the father exerts a negative influence on given size household, the greater the value of expen- child nutrition, once again, controlling for the mediat- ditures per adult equivalent unit. This would imply a ing role of income. These findings suggest that the greater availability of food and nonfood goods per schooling of the mother does not have an independent child, and may contribute to improved long-term nu- effect on long-term nutritional status, either through tritional status. changing preferences or improving the productivity In both urban and rural areas, there is a positive of household activities. Meanwhile, there is evidence coefficient for the number of children 6-14 years of of a deleterious impact of raising father's education, age; although it is not significant at P < 0.1 in rural controlling for income. One plausible explanation for areas. This finding likely reflects the fact that the pres- this finding is that education proxies for income con- ence of older siblings that are capable of providing trol and bargaining power within the household, some level of child care and nurturing but are too which determines how resources are spent and allo- young to be working extensively in the labor market cated. If male preferences tend toward expenditures affords a degree of benefit to the preschoolers' long- on goods and services that are less beneficial to child term nutritional well-being. nutrition (e.g., entertainment, clothing), in contrast to In urban areas, but not rural areas, one finds that the the choices that would be made by women, then it greater the number of adult females for a household follows that increased male education, controlling for of a given size and composition, the greater the child's income, may adversely affect child nutrition. Of linear growth. It is hypothesized that this finding is course, these findings on the relative benefits of male partially a reflection of women exerting more control and female education do not imply that male educa- overexpenditures, andthatthe child'snutritional wel- tion should not be promoted. The increased house- fare is more important in the preference ordering of hold income that accompanies paternal education will urban women than men. Alternatively, however, this result in improved child nutrition. may simply reflect that women (like young children) Another noteworthy finding of the urban equations "cost" less than men in terms of their consumption of found in Table 8 is that the height of both the mother food, clothing, and so on. It is noteworthy that the and father affect the stature of their preschool-aged gender of the head of the household has no effect on offspring. While both parental stature coefficients are the linear growth of children in urban areas. significant, the parameter estimate for the mother's The age of the mother serves as a proxy for experi- height is more than twice the magnitude of the coeffi- ence in caring for a child. Maternal age is also expected cient for the father's height. This suggests that the to affect the birth outcome and birth weight, which is intergeneration link of stature is stronger for the correlated with a young child's length-for-age. In both mother than the father. This could be attributable to urban and rural areas, however, the coefficient proved the fact that the maternal influence incorporates fac- insignificant. tors relating to phenotype (for example, the environ- Concerming the effect of the child's age, in both ment within the womb before birth), as well as urban and rural areas, the coefficients on the dummy genotype, while only the father's genotype has any variables for children less than or equal to 6 months of influence on his offspring. In rural areas, the mother's age and 7-12 months are positive and significant. This 19 can be explained by the fact that the accumulation of sign from Model I, and is significant should not be episodes of nutritional and health stress are less likely interpreted as more education leading to worse nutri- to manifest themselves in terms of stunting during a tion. Rather, this switch simply reflects the inclusion child's first year of life than between the ages of 13 and of the interaction term. When the total derivative is 24 months. The negative coefficient found on the taken, it is positive over most of the relevant range of dummy variable for children aged 25-36 months indi- urban households, although the positive effect of ed- cates that Z-scores for height-for-age are lowest ucation becomes negligible among households at the among this age group. This reflects the acute periods bottom end of the income distribution. These female of malnutrition suffered previously during the wean- education and education interaction terms, however, ing period of 13-24 months of age, when infection and are not jointly significant at standard levels, although diarrheal disease are known to be most debilitating. the male education variables are jointly significant at In urban and rural equations, the male dummy the 0.15 level (F equal to 1.91). variable (that equals 1 if the child is a boy, and 0 if a In rural areas, the fit of the expanded model is an girl), has a negative coefficient. It is, however, only improvement. Among the community-level charac- significant in the urban equation. The fact that the teristics, the negative and significant malaria and dys- height-for-age Z-score, and thus the long-term nutri- entery variables indicate that children are likely to tional status, is lower for boys when controlling for suffer less stunting in villages where these diseases other socioeconomic variables, supports the conclu- were not reported to be serious health problems.20 The sions of Svedberg's (1988) recent review of the litera- surprising positive coefficient for the smallpox/chick- ture that finds little evidence of discrimination against enpox/measles variable is difficult to explain. One female children in Africa, unlike in Southeast Asia.18 potential reason may revolve around the fact that The effect of birth order on previous malnutrition as since the variable included smallpox, it may be that measured in terms of linear growth was also exam- those villages with greater health awareness are cog- ined. In both rural and urban areas, the parameter nizant of the potential threat that smallpox used to estimate is negative, but only significant at the 10 represent. An alternative explanation is that if chick- percent level in urban areas. Seasons also play a role enpox/smallpox/measlesare amongthe most serious in terms of predicting height-for-age Z-scores. This is health problems, by implication others are not. Com- in accordance with the evidence that linear growth pared with the chronic episodes of dysentery and does display seasonal cycles (Payne, 1989). malaria, one-off episodes of smallpox/chickenpox/ The basic model described above was expanded in measles may be less likely to affect the long-term our search for non-linearities and interactions be- nutritional status of preschool-aged children. tween variables, as well as the effect of community- Also in keeping with expectations, the negative and level variables on child nutrition. More specifically, a significant result for the distance to the doctor variable quadratic term for education and an interaction term indicates that the more distant such medical services, between education and income were included in a the lower the height-for-age Z-score. Similarly, the second model.19 Concerning the impact of community dummy variable for whether there is a nurse in the variables, these data were taken from a separate com- village has a positive coefficient, indicating that a munity-level questionnaire administered only in the nurse's presence might be expected to improve long- rural areas. They assist in providing some insight into term child nutrition. Interestingly, the negative inter- the effect of regional factors and community services action term between nurse and expenditures also on nutrition. suggests that for wealthier households, the presence The results of Model II indicate that the income of a nurse in the village is not as important as for terms in urban and rural areas remain significant. In households in the lower end of the income distribu- fact, the parameter estimates are slightly higher, yield- tion. The nurse and nurse multiplicative interaction ing elasticities of height-for-age Z-score with respect term with expenditures are jointly significant with a to income of 1.12 and 0.33 for urban and rural house- p-valuelessthan0.0001, asareall thecommunity-level holds, respectively, in the bottom income quintile. In variables combined. urban areas the interaction term between income and The addition of the quadratic schooling and interac- mother's education is significant and positive, imply- tion terms with education resulted in the quadratic ing that income brings higher retums in terms of term for mother's education being positive and signif- improved nutrition when the mother has more educa- icant at the 0.05 level. This suggests a non-linear rela- tion, and/or that the positive impact of education is tionship between mother's education and long-term greater amonghouseholds with more income. The fact malnutrition in rural areas where there is an increas- that the coefficient for mother's schooling changes ingly positive return to mother's school with more 20 education. However, the education and education in- and significant in rural areas. The indication that chil- teraction terms for both mothers and fathers were not dren of taller mothers/fathers will be more lean and jointly significant at standard levels. at greater likelihood of being characterized as cur- Next, a third model is run for rural areas only, rently malnourished needs to be carefully assessed. adding two more variables: per capita landholdings The question arises as to whether children are misclas- and the share of land devoted to the production of the sified as currently malnourished because of their ge- major export crops-coffee, cocoa, and tobacco-as netic propensity for being tall and lean, and in fact do well as the less important cola nut and tobacco. In the not suffer some functional impairment because of case of perennial crops, such as coffee and cocoa, the their relatively low proportionate weight-for-height. land allocated to these crops is not subject to year-to- This issue needs to be the focus for further research. year changes such as for field crops. Consequently, the As with the height-for-age regressions in both urban land use variable is considered exogenous. With the and rural areas, household size plays no role in affect- inclusion of the per capita landholding variable, ing current nutritional status when controlling for the which controls for farn size, it is possible to test household structure. There is also some tentative sup- whether there is a deleterious impact on nutritional port for the hypothesis that as with the height-for-age status of producing export, rather than domestically equations for a given household size, the greater the consumed food crops. number of women over the age of 14, and perforce the The results indicate that neither the landholding per fewer men, the higher the weight-for-age 7-score of capita nor the share of the household's land devoted the child. In addition, there is a positive and significant to the production of the major export crops are signif- coefficient in rural areas for the number of children icant in the long-term nutrition functions. This sup- between 6 and 14 years of age. This corresponds to ports the descriptive data presented above, indicating what was observed in the long-term nutrition produc- that the use of land for producing export crops, rather tion functions, further reinforcing the notion that older than food crops, will not affect the long-term nutri- siblings provide valuable child care for preschool- tional status of preschool-aged children. aged children. Concerning other household structure and demo- CurrentMalnutrition graphic variables, as with long-term nutrition, the age of the mother has no effect on a child's current nutri- The determinants of current nutritional status, or tion. And likewise, the dummy variable for female- wasting, among preschool-aged children are expected headed households is not significant in the urban to be more difficult to capture in a model because of current nutrition functions. the stochastic nature of episodes of stress and disease The age of a child is an important determinant of the that precipitate acute malnutrition. Nevertheless, cur- weight-for-height Z-score. In both urban and rural rent nutritional status functions are presented in areas, children 0-6 months and 2-4 years of age have Table 9. better current nutritional status than children between The results indicate that the income parameter, al- 13 and 24 months, the vulnerable weaning period. though positive, is not statistically significant in either Unlike for chronic malnutrition, the gender of the the urban or rural areas. The variable for mother's child does not play a role in affecting the level of schooling is positive and significant, indicating that wasting. The child's birth order variable is negative in unlike for long-term or previous malnutrition, the both urban and rural areas, although not significant in education of the mother enhances the nature of pref- urban areas. As in the case of long-term nutrition, this erences and decisions in such a way as to raise the reinforces that children in earlier sibling positions are child's current nutritional status in urban and rural advantaged, relative to brothers and sisters that are areas. This supports the hypothesis that the education born later. of women has a greater positive effect on the nutrition As with the long-term nutrition function, the ex- of children as mediated through their greater effi- panded models of current malnutrition are presented. ciency in household productive activities. The community-level variables indicate once again The higher parameter value in the rural equations the importance of malaria in the village as a determi- indicates that mother's education will have a stronger nant of preschool-aged malnutrition. The chicken- effect in rural than in urban areas. At the same time, pox/smallpox/measles indicator takes on the the education of the father shows no effect on child expected statistically significant negative sign, reflect- nutrition in either urban or rural areas. ing how the occurrence of such a disease in the com- Among other noteworthy findings is the influence munity may have deleterious short-term nutritional of mother's height, which is negative and significant impacts. The insignificant dysentery variable is diffi- in urban areas, and father's height, which is negative cult to explain. Once again the distance to the doctor Table 9. Current nutrition functions Dependent Variable: Weight-for-Height Z-Score Urban - I Rural - I Urban - II Rural -1I Rural - III Independent Variables Parameter Standard Parameter Standard Parameter Standard Parameter Standard Parameter Standard Estimate Error Estimate Error Estimate Error Estimate Error Estimate Error Intercept 0.838 1.639 0.900 1.432 1.081 1.966 1.245 1.615 1.693 1.701 Mother's height -1.277 0.727 0.100 0.612 -1.225 0.732 0.021 0.638 -0.153 0.660 Father's height -0.082 0.644 -1.093 0.559 -0.049 0.648 -0.905 0.585 -1.013 0.603 Female head dummy -0.244 0.519 -0.259 0.520 Child 0-6 months dummy 0.598 0.159 0.759 0.145 0.609 0.159 0.761 0.152 0.795 0.157 Child 7-12 months dummy 0.221 0.180 -0.143 0.146 0.214 0.181 -0.147 0.151 -0.144 0.155 Child 25-36 months dummy 0.272 0.142 0.571 0.124 0.283 0.142 0.547 0.130 0.586 0.135 Child 37-48 months dummy 0.400 0.143 0.551 0.128 0.402 0.144 0.539 0.133 0.569 0.137 Child 49-60 months dummy 0.292 0.136 0.482 0.117 0.292 0.136 0.475 0.122 0.497 0.125 January-March dummy 0.289 0.131 -0.063 0.100 0.281 0.132 -0.032 0.119 -0.019 0.121 April-Junedummy 0.074 0.137 -0.066 0.117 0.096 0.140 -0.006 0.124 0.048 0.127 July-Septemberdummy -0.003 0.105 -0.007 0.098 0.008 0.106 0.062 0.111 0.079 0.114 No. children 0-5 yrs -0.010 0.037 0.032 0.034 -0.014 0.037 0.034 0.035 0.023 0.036 No. children 5-14yrs -0.004 0.042 0.103 0.039 -0.009 0.042 0.101 0.041 0.114 0.042 No. female > 14 yrs 0.051 0.035 0.054 0.038 0.054 0.035 0.074 0.039 0.075 0.039 Ln expenditures per capita 0.073 0.078 0.029 0.063 0.036 0.112 0.023 0.080 0.025 0.087 * Child gender dummy var. 0.119 0.081 -0.066 0.069 0.123 0.081 -0.099 0.072 -0.125 0.074 '9 Abidjan dummy var. -0.296 0.086 -0.290 0.086 1985 dummy var. -0.136 0.100 0.395 0.084 -0.142 0.101 0.343 0.095 0.321 0.100 Household size -0.003 0.024 -0.031 0.026 -0.002 0.024 -0.031 0.027 -0.034 0.027 Mother's age -0.002 0.004 0.000 0.003 -0.002 0.004 0.002 0.003 0.003 0.003 Birth order -0.026 0.036 -0.063 0.031 -0.025 0.036 -0.068 0.033 -0.072 0.034 West Forest dummy var. -0.429 0.090 -0.473 0.095 -0.412 0.098 East Forest dummy var. -0.412 0.095 -0.386 0.102 -0.284 0.115 Mother's education 0.028 0.013 0.039 0.022 0.070 0.228 0.539 0.508 0.431 0.593 Father's education -0.005 0.011 0.003 0.012 -0.098 0.202 -0.185 0.264 -0.383 0.284 Mother's ed x Ln exp per capita -0.003 0.019 -0.032 0.043 -0.021 0.049 Father's ed x Ln exp per capita 0.011 0.017 0.013 0.023 0.029 0.025 (Mother's ed?2 -0.000 0.002 -0.020 0.011 -0.023 0.013 (Father's ed) -0.003 0.002 0.004 0.004 0.005 0.004 Malaria -0.569 0.164 Dysentery 0.238 0.171 Chicken/smallpox/measles -0.285 0.109 Distance to doctor -0.149 0.057 Nurse dummy 1.136 2.041 Nurse x Ln per capita exp. -0.082 0.177 Land per capita -0.120 0.055 Export crop share 0.002 0.002 R2 =0.0621 R2 0.1046 R = 0.0665 R =0.1347 R2 0.1440 22 is negative and significant, indicating the importance total hectarage cultivated will not have any impact on of access to medical care as a determinant of current nutritional status. However, the per capita landhold- nutritional status. The nurse dummy variable and ing variable was significant and negative in this model interaction term with income, although not signifi- of current malnutrition. This suggests that among cant, have the same signs as they did for the height- households with comparable income, the more land for-ageequation. Theyarejointlysignificantat the 0.15 under cultivation, the worse the current nutritional level. Intotal, thecommunitylevel variablesarejointly status of the child. While at first glance this may be significant at the 0.0001 level with an F-statistic equal considered counterintuitive, one can formulate a rea- to 6.93. sonable hypothesis explaining the negative parameter The interaction terms between education and in- estimate. A larger amount of land under cultivation come prove insignificant. The significant and negative undoubtedly involves more time spent in agricultural quadratic term for mother's education in rural areas activities, especially among women, and perforce, less implies that in contrast to the long-term nutrition func- time for child nurturing and related household pro- tions, the positive effect of mother's schooling on nu- duction activities such as food preparation. This may trition diminishes as the level of educational result in less or poorer quality child care, thereby achievement rises. The mother's education and educa- leading to more current malnutrition. An alternative tion interaction terms are jointly significant for explanation for this finding is that, controlling for weight-for-height at the 0.15 level (F equal to 1.89), income, households with a higher share from agricul- although this was not the case for the male education ture maybe at greater nutritional risk as a consequence variables. of the lumpy nature of agricultural incomes and the In rural areas, Model III, which includes landhold- difficulties in intertemporal savings. ing variables, indicates that the share of cash crops in Notes 18. To the extent that the NCHS standards are wrong (that is, if there is a systematic gender bias in the reference population but not 15. Consumption expenditures are defined as the sum of food in the observations), this finding would not be valid. However, the and nonfood expenditures, the value of goods produced and con- evidence as put forth in Habicht (1974) and Martorell and Habicht sumed by the household, the imputed value of durable goods (1986) provides some confidence in the use of the NCHS standards owned by the household, and the actual or imputed rents of urban for comparisons across ethnic groups. residents. In addition, consumption expenditures are in real terms, 19. It is noteworthy that an attempt was made to introduce a with nominal values deflated by the cluster-specific price index. quadratic and cubic income term into Models I and II. In both cases, Consumption expenditures are considered the best possible mea- it only resulted in increased multicollinearity and insignificant pa- sure of household permanent income, and similarly, the rameter estimates. household's unobservable utility. 20. The community-level questionnaire asked for a listing of the 16. The model does not indude commodity prices, which could fourmajorhealthproblems. If malaria wasgiven asthemostserious, cause a bias if the omitted prices are associated with any of the or second most serious health problem in the village, it was coded included variables. This is because many of the prices used to deflate as a 1, and otherwise 0. The same procedure was employed for expenditures were predicted, or derived from cell means, and not dysentery and smallpox/chickenpox/measles, the latter group of observed. While using such information as the basis for constructing diseases that were all coded together. deflators for the cross tabulationspresented earlieris reasonable, the 21. A similarmodel was estimated including an interaction term use of these predicted prices as regressors in an econometric model between the share of land devoted to export crops and per capita was not considered to be justified. Furthermore, the effect of prices expenditures. The variable proved insignificant. on nutrition would belimited to differences in relativepricesinduc- 22. An alternative explanation for such a finding is that women ing different patterns of expenditures. Given that rural and urban are able to exert more influence in the household decision making models are estimated separately, and both include seasonal, re- process since better education improves their intrahousehold bar- gional, and round dummy variables, it is likely that most of the gaining power. Of course, this is predicated on women having a relative price differences have been captured. In addition, the price different preference ordering from their spouse, one which tends to of leisure or the imputed wage rate is not induded in the models. favor the current nutrition of their children. As indicated by This despite that it was induded in the community-level question- Rosenzweig and Schultz (1983), it is difficult to distinguish between naire. However, it was felt asking one person in a village, at one the alternaive hypotheses. A recent paper by Thomas (1989) at- season, for the prevailing agricultural wage, results in a suspect tempts to look upon this issue by examining whether the effect of representation of the opportunity cost of time. Furthermore, only a income under the control of different households members is the couple percent of the rural working age population reported per- same in terms of nutrient intakes, fertlity, child survival, weight- forming wage labor, preduding estimatng a wage function, and for-height, and height-for-age. While his results reject the sugges- suggestingthatruralwagelabormarkets,if theyexist,areextremely tion that all income is pooled or allocated by a dictator, for all but thin. height-for-age, the results need to be interpreted with caution since 17. The theory of conditional demand functions in consumer there is a strong assumption that current unearned income does not behavior is discussed in Pollack (1969), where he shows that condi- reflect past labor supply decisions. tional demand functions meet all the requirements satisfied by ordinary demand functions. 6. Conclusions This paper has examined the extent of malnutrition in techniques, or prepare more nutritious weaning foods C6te d'Ivoire and the determinants of long-term and [Behrman, 19881). In addition, to the extent that edu- current nutritional status among preschool-aged chil- cation is a proxy for income control, and that women's dren. The most important finding of this work is that preferences are more oriented toward purchasing the levels of both long-term and current malnutrition goods and services that are inputs into good nutrition, are low relative to survey data from other Sub-Saharan this too could explain both the positive effect of ma- African nations. This is indicative of a combination of ternal education on current nutrition, and the fact that relatively good access to food (that is, household food father's education appears to be a risk factor for long- security) and low levels of debilitating disease. None- term malnutrition in urban areas, controlling for theless, there remains ample opportunity to improve household incomes. the nutritional status of the 16.2 percent of the children Exploring non-linearities in the role of education that are stunted and 7.1 percent of the children that are and the interactions between education and expendi- wasted. tures provided no generalizable patterns. For exam- Raising expenditure levels is the key element to any ple, increasing returns to education of women in rural effort to reduce chronic malnutrition. In the longer areas were noted for long-term nutrition, while just term, expenditure levels will be raised by increasing the opposite was found for current nutrition. Simi- the assets of the household (for example, land, equip- larly, for long-term nutritional status, evidence that ment, business assets), and raising their productivity mothers' education is more important for households through education. In the shorter term, whether it be with higher income was noted only in urban areas. through income transfer schemes or employment gen- In sum, it is therefore inferred that a higher educa- eration projects, targeted efforts to increase incomes of tional attainment of the mother improves the produc- poor households will also go a long way toward alle- tivity of household activities, such as child care, and viating chronic nutritional problems. improves the quality of decisions that fall in the While income is of considerable importance in re- women's domain concerning the choice of food and ducing long-term malnutrition, a number of other other inputs that contribute to good nutrition. Since factors contribute to the problem. In particular, the men do not assume primary responsibility for child results suggest that the education of women plays an care, improving their knowledge does not contribute important role in improving a child's long-term nutri- directly to better nutritional outcomes. A related po- tional status, as well as in equipping mothers to avoid tential explanation for the positive effect of the and/or cope with crises that may precipitate acute mother's, but not the father's, education is that educa- periods of nutritional stress (that is, current malnutri- tion proxies for authority over the expenditure of tion). These influences are over and above the fact that household resources. Consequently, if a woman is better-educated women (like men) have higher earn- better educated, she may exert more control over de- ings that tend to raise incomes and nutritional status. cisions in such a way as to favor nutrition, assuming Controlling for household incomes, the observed her preference function differs from that of her spouse. positive effect of the education of women on current This latter explanation finds some additional support nutritional status was not observed for men. This in the fact that many of the nutrition production func- likely reflects that the mother is the key household tions indicate that the larger the proportion of women member concerned with health-related decisions and in a household (of a given size and age composition), implementation. More education will consequently the better the nutritional status of the child. improve the mother's knowledge and health practices This being said, the evidence does indicate that and have a greater impact on child nutrition (e.g., there are positive effects of the father's and mother's because they are more likely to use oral rehydration education as mediated through income. In addition, 23 24 the observation that women's education is more ben- land to the production of major export crops, as an eficial than men's education when controlling for ex- alternative to producing food crops for domestic con- penditure levels must be tempered by the fact that the sumption, represents a short-term nutritional risk fac- gross retums to education among men are higher than tor. This suggests that liberalizing prices and for women (see Appendix B). Therefore, the vital role removing distortions that will encourage agricultural of raising the education of men in order to increase growth will improve nutritional status, even though household income levels must not be overlooked. such a move implies favoring the production of highly It is also important to emphasize that nonformal taxed export rather than food crops. While the objec- education, especially for women, will also likely con- tive of food self-sufficiency is still within the grasp of tribute to improved child nutrition. Experiences from C6te d'Ivoire, any effort to achieve that objective must other countries especially in Asia illustrate the benefits be carefully weighed against the economic costs of the of nutrition education messages, targeted to the poor sub-optimal use of valuable resources. Given that and designed to elicit a response to a carefully crafted households who choose to grow export crops do not and delivered message (Griffiths, 1985; Berg, 1987). manifest a higher probability of their children being This paper also amply illustrated the importance of malnourished, there is a strong argument for Cote community characteristics as determinants of malnu- d'Ivoire's following a development strategy that fo- trition. For example, both height-for-age and weight- cuses on growth in incomes and improving returns to for-height Z-scores are lower in communities where factors, even if it be at the cost of food self-sufficiency. malaria is a serious health problem, where doctors are The fact that the survey data did not include any a greater distance from the community, and where information on consumption and that the commodity nurses are absent. Therefore, investments in health price data were limited and of suspect quality, detracts infrastructure that reduce the prevalence of diseases from exploring the impact of price policy on nutrition, such as malaria and dysentery, and similarly improve either directly through effecting commodity choice, or the availability of health care personnel offers a fruit- indirectly through the effect on real earnings. Like- ful avenue for improving child nutrition. wise, the potential benefits of improving the efficiency While household size showed no effect on nutri- of commodity markets cannot be examined. Thus, the tional status, birth order was an important causal ele- role of a variety of policy instruments on nutrition, ment of malnutrition. The results reinforce the need such as real exchange rate depreciation that alters for family planning programs and fertility reduction relative prices, cannot be determined. Likewise, the measures, such as breastfeeding, to address the nutri- absence of consumption data precludes relating nutri- tional problems associated with later-born siblings. tional status to household food security concerns that And these positive effects do not include the fact that revolve around access to adequate food. higher per capita expenditure levels will likely also Despite these drawbacks, this paper generally sup- follow from fewer children. ports the contention that malnutrition is a problem of The role of intergenerational influences in deter- poverty; and any policies that either reduce the in- mining nutritional status was also highlighted. The comes of the poor or result in less availability of social observation that mother's height has a greater impact services, especially in the areas of primary health care on child nutrition indicates that it is not just genetics and primary school, will have immediate and long- but also the mother's phenotype and/or human capi- term deleterious impacts on the nutritional welfare of tal accumulation that condition nutritional outcomes. children. These intergenerational determinants of malnutrition This being said, waiting for economic growth to focus attention on the long-term consequences of ne- reduce malnutrition likely represents an unacceptably glecting human resources in the short term. long time horizon. Experiences from other countries This paper also explored whether allocating land to indicate that targeted programs, whether they be in export crops, rather than food crops that may either be the domain of food distribution, income augmenting marketing or consumed at home, has a deleterious measures, and/or primary health care, can raise nutri- impact on nutritional status of preschool-aged chil- tion levels of children even in the face of a stagnant dren. Doing so has important implications of a strat- economic environment. Therefore, more immediate egy of agricultural development based on exploiting targeted action is necessary. This, of course, requires comparative advantage through increasing incentives expanding institutional capacity to plan projects and for producing the dominant export crops. Controlling deliver services, a costly challenge in its own right. But for the total value of household income in cash and if such development is possible anywhere in Africa, in-kind, and per capita landholding, there was no the circumstances in C6te d'Ivoire are most amenable indication that the household's decision to allocate to such institution building. Appendix A Price Index In order to address the problem of comparing house- portation, (d) there was a passable road, (e) people hold expenditures across survey rounds (that is,years) migrate out of the community in search of work, and and regions, a price index was formulated. The defla- (f) people migrate into the village in search of work. In tors were constructed from price data collected by addition, seasonal and yearly dummy variables were enumerators in the form of a community question- included, as well as dummy variables for agricultural naire that was administered in the clusters in which zones. In combination, these variables reflected the the household survey was undertaken. The index took level of market integration and sectoral and temporal the following form: factors that are expected to affect commodity prices. This model was run for all the commodities in- Iijk ijk (Pijk * Wi) cluded in the index, both by individual year and across = (pji * Wi) the two rounds. In the latter case, an additional dummy variable for round was included. When the where Pi corresponds to the mean price of commodity models were significant, predicted values were com- i across all clusters and both survey rounds, Wi rep- pared with actual prices in those clusters in which resents the mean food share expended on that com- prices were recorded in the community questionnaire. modity across clusters and survey rounds, and Pijk is If the predictions were accurate, and the correlation the price paid for commodity i, in region j in round k. between predicted and actual prices high, the prices The community questionnaire included a list of 18 predicted from the model were used to fill in missing food commodities and 4 nonfood commodities on prices, although reported prices were used when which the surveyor was instructed to seek three rep- available. resentative retail prices. In many instances, however, In some cases, however, the prediction equations none, one, or two prices were recorded for a commod- were not significant and/or the comparison of pre- ity in a given community. When only one price was dicted and actual values raised some suspicion as to available from the community questionnaire for a the accuracy of the model. In such a case, the mean given commodity, it was used in constructing a price price for the region and round (that is, Abidjan, Other index. When two or three prices were recorded, the Cities, West Forest, East Forest, and Savannah) was mean value was employed. inserted to fill in missing values. For a number of commodities, in a number of clus- Owing to the small number of nonfood items (that ters in any given round, there were no prices reported is, cloth, enamel bowl, menthol, and sandals) included in the community questionnaire. In order to arrive at in the community questionnaire, coupled with the the most accurate index numbers, an econometric large number of missing values and large variance in model was employed to predict commodity prices reported values likely due to the nonhomogeneity of based on a variety of community-level and temporal the product, the deflators employed were based on characteristics. Included as regressors were the fol- food prices only. Food makes up approximately 55 lowing: population size; the level of concentration of percent of the average household expenditures. The hamlets; and a series of dummy variables that equaled absence of any reliable nonfood price data in the sur- 1 if (a) the community was served by a paved road, (b) vey has the obvious disadvantage of running the risk there was a market place, (c) there was public trans- of the deflators for urban areas being too high. How- 25 26 ever, food prices, in general, are not markedly higher using non-deflated values. The descriptive data in the urban market-that is likely because the flow of changed slightly, especially when comparing results goods produced in rural areas is often into market across sectors where deflators are most variable. Most centers, and then once again out to thinner rural mar- important were the changes in the Savannah where kets. food prices are markedly below the national average. To deterrnine whether the use of the deflators af- The econometric results presented in the paper were fected the results, sensitivity analysis was employed unaffected by the use of the deflated values. Appendix B Variable Means and Standard Deviation Appendix Table B. Variable means and standard deviation Urban Rural Standard Standard Independent Variables Meun Deviation Mean Deviation Weight-for-height -0.38 1.15 -0.22 1.31 Height-for-age -0.16 1.66 -0.51 1.83 Ln of per capita expenditure 12.11 0.67 11.60 0.59 Father's education 5.26 5.84 2.45 4.08 Mother's education 3.21 4.44 1.13 2.53 Father's height (meters) 1.70 0.06 1.68 0.07 Mother's height (meters) 1.60 0.06 1.59 0.06 Household size 11.83 6.50 11.79 5.39 No. children 0-5 yrs 2.97 1.86 3.29 1.73 No. children 6-14 yrs 3.28 2.60 3.14 2.22 No. females > 14 yrs 3.03 2.10 3.15 1.86 Birth order 6.18 3.95 5.99 3.29 Mother's age 28.48 10.75 29.29 13.39 Father's age 31.95 13.57 35.61 15.54 Ln of value of housing (CFA) 6.91 7.33 Ln of business assets (CFA) 7.20 6.35 3.13 5.07 Ln of value of land (CFA) 13.52 3.21 Ln of value of tools (CFA) 10.40 2.12 Ln of value of equipment (CFA) 2.98 4.88 Malaria 0.07 0.27 Dysentery 0.05 0.22 Chicken/smallpox/measles 0.19 0.22 Ln distance to doctor 3.08 0.74 Nurse 0.26 0.44 27 Appendix C Reduced-form Human and Physical Capital Models Appendix Table C contains the results of the second rate of increase in the marginal returns is also higher. stage estimates of the reduced-form human and phys- In rural areas, there is no evidence of positive returns ical capital functions where the dependent variable is to women's education, and once again the returns to the natural log of expenditures. They represent the education for men increase with additional schooling. second stage of a two-staged least squares procedure For men with only a couple of years of schooling, the used to generate the results found in Tables 1 and 2. marginal effect on income appears to be negative, The reduced-form models are similar in parameters to although this likely reflects the limnited curvature in the human welfare functions estimated for C6te the function estimated. d'Ivoire by Glewwe (1987). Another point is that the age coefficient, a proxy for Despite the fact that the purpose of this paper is not experience, was not significant. And finally concern- to address issues regarding the returns to physical ing assets, the returns to business assets are positive in assets and human capital, a few highlights are worth both urban and rural areas, with their being greater in emphasizing. First, the gross returns to education are rural areas; and there are positive returns to assets in positive for men and women in urban areas, and they the form of equipment and land in rural areas, but not increase with more schooling. Second, the returns are tools. greater for men than women in urban areas, and the Appendix Table C. Second stage reduced-form consumption expenditure function Dependent Variable: Log of Expenditures Urban Rural Independent Variables Paraneter Estinate Standard Error Paraneter Estimnate Standard Error Intercept 12.044 0.091 11.848 0.103 Mother's age 0.002 0.002 -0.002 0.001 Father's age 0.001 0.001 -0.000 0.001 Household size -0.050 0.058 0.068 0.059 Ivorian dummy -0.247 0.046 -0.073 0.046 Abidjan dummy -0.001 0.039 East Forest dummy -0.019 0.039 West Forest dummy -0.080 0.048 Mother's education 0.017 0.012 0.004 0.032 Father's education 0.031 0.010 -0.029 0.015 (Mother's ed4 0.001 0.001 0.000 0.005 (Father's ed) 0.002 0.001 0.004 0.001 Ln of value of housing 0.008 0.003 Ln of value of tools -0.057 0.014 Ln of business assets 0.014 0.003 0.026 0.003 Ln of value of equipment 0.017 0.003 Ln of value of land 0.045 0.010 No. children 0-5 yrs 0.008 0.059 -0.130 0.058 No. children 6-14 yrs 0.031 0.060 -0.078 0.059 No. males > 15 yrs 0.045 0.060 -0.086 0.061 No. females > 15 yrs 0.025 0.061 -0.079 0.062 1985 dummy -0.177 0.040 -0.051 0.037 R =0.4744 R2 =0.1633 28 References Ainsworth, Martha and Juan Mufioz (1986): "The C6te World Federation of Public Health Associations, d'Ivoire Living Standards Survey: Design and Im- Washington, D.C. plementation," Living Standards Measurement Grootaert, Christiaan (1986): "Measuring and Analyz- Study, Working Paper No. 26. World Bank, Wash- ing Levels of Living in Developing Countries: An ington, D.C. Annotated Questionnaire," Living Standards Mea- Behrman, Jere R. (1988): "Nutrition and Health and surement Study, Working Paper No. 24. World Their Relation to Economic Growth, Poverty Alle- Bank, Washington, D.C. viation and General Development," University of Habicht, Jean-Pierre, R. Martorell, C. Yarbrough, R. Pennsylvania, Philadelphia. Malina, and R. Klein (1974): "Height and Weight Behrman, Jere R. and Anil B. Deolalikar (1986): Standards for Pre-school Children: How Relevant "Health and Nutrition," in Handbook of Development are Ethnic Differences in Growth Potential?", Lancet Economics, edited by Hollis Chenery and T. N. 1:611-15. Srinivasan. North Holland Publishing Co. Haddad, Lawrence (1987): "Agricultural Household ________ (1987): "Will Developing Country Nutri- Modeling with Intrahousehold Food Distribution: tion Improve with Income? A Case Study for Rural A Case Study of Commercialization in a Southern South India," Journal of Political Economy 95 (3): 492- Philippine Province." Ph.D. dissertation, Stanford 507. University, Palo Alto, CA. _ _ (1988): "Seasonal Demands for Nutrient Heller,PeterandWilliamDrake (1976): "Malnutrition, Intakes and Health Status in Rural South India,' in Child Morbidity and the Family Decision Process," Seasonal Variability in Agriculture, Employment, and Journal of Development Economics 23: 161-76. Food Markets: Consequences for Food Security, edited Horton, Susan (1988): "Birth Order and Child Nutri- by David E. Sahn. Johns Hopkins University Press, tional Status: Evidence from the Philippines," Eco- Baltimore, MD, in process. nomic Development and Cultural Change (January): Behrman, Jere R. and Barbara L. Wolfe (1984): "'More 341-45. Evidence on Nutrition Demand," Journal of Develop- _ _ (1986): "Child Nutrition and Family Size ment Economics 14: 105-28. in the Philippines," Journal of Development Economics Berg, Alan (1987): "Malnutrition, What Canbe Done?" 23: 161-76. World Bank, Washington, D.C. Kanbur, S.M.R. (1987): "Poverty Alleviation Under Food and Agriculture Organization of the United Na- Structural Adjustment: A Conceptual Framework tions (1985): The Fifth World Food Survey, FAO, and Its Application to Cote D'Ivoire," World Bank, Rome. Washington, D.C. Foster, J., J. Greer, and E. Thorbecke (1984): "A Class Kennedy, Eileen (1989): "Health and Nutrition Effects of Decomposable Poverty Measures." Econometrica. of the Commercialization of Agriculture: A Com- Glewwe, Paul (1987): "Measuring the Determinants of parative Analysis," International Food Policy Re- Household Welfare in Cote d'Ivoire," World Bank, search Institute, Washington, D.C. Mimeo. Washington, D.C. Mimeo. Kennedy, Eileen and Bruce Cogill (1987): "Income and __________ (1986): "The Distribution of Welfare in the Nutritional Effect of the Commercialization of Ag- Republic of Cote d'lvoire," Living Standards Mea- riculture in Southwestern Kenya," Research Report, surement Study, Working Paper No. 29. World International Food Policy Research Institute, Wash- Bank, Washington, D.C. ington, D.C. Griffiths, Marcia (1985): "Growth Monitoring of Pre- Martorell, Reynaldo and Jean-Pierre Habicht (1986): school Children, Practical Considerations for Pri- "Growth in Early Childhood in Developing Coun- mary Health Care Projects," Prepared for UNICEF. tries," in Human Growth: A Comprehensive Treatise, 29 30 2nd ed. edited by J. Tanner and F. Falkner. Plenum, Improvements, edited by Per Pinstrup-Andersen. In- New York. temational Food Policy Research Institute, Wash- Payne, Phillip (1989): "Public Health and Functional ington, D.C. Consequences of Seasonal Hunger and Malnutri- Svedberg, Peter (1988): "Undernutrition in Sub- tion," in Seasonal Variability in Third World Agricul- Saharan Africa: Is There a Sex Bias?", Institute for ture: Consequences for Food Security, edited by David International Economics, Seminar Paper No. 421, E. Sahn. John Hopkins University Press, Baltimore, University of Stockholm, Stockholm. MD, in process. Thomas, Duncan (1989): "Intra-household Resource Pitt, Mark and Mark Rosenzweig (1985): "Health and Allocation: An Inferential Approach," Yale Univer- Nutrient Consumption Across and Within Farm sity, New Haven, CT. Mimeo. Households," Review of Economics and Statistics 67: Thomas, Duncan, John Strauss and Maria Helen 212-23. Henriques. (1987): "Child Survival and Nutrition Sahn, David E. (1988): "The Effect of Price and Income Status and Household Characteristics: Evidence Changes on Food Energy Intake in Sri Lanka," Eco- from Brazil," Yale University, New Haven, CT. nomic Development and Cultural Change (January): Mimeo. 315-40. United States Public Health Service (1976): "NCHS Sen, A.K. (1976): "Poverty: An Ordinal Approach to Growth Charts," Health Resources Administration, Measurement," Econometrica. Rockville, MD. Strauss, John (1987): "Household, Communities, and Wolfe, Barbara and Jere R. Behrman (1982): "Determi- Preschool Children's Nutritional Outcomes: Evi- nants of Child Mortality, Health, and Nutrition in a dence from Rural Cote D'Ivoire," Living Standards Developing Country," Journal of Development Eco- Measurement Study, Working Paper No. 40. World nomics 11: 163-93. Bank, Washington D.C. World Bank (1987): Poverty and Hunger. World Bank, (1985): "The Impact of Improved Nutri- Washington D.C. tion on Labor Productivity and Human Resource World Health Organization (1983): Measuring Change Development: An Economic Perspective," in The in Nutritional Status, World Health Organization, Political Economy of Food Consumption and Nutrition Geneva. Distributors of World Bank Publications ARGENTINA FINLAND MALAYSIA Fcrwbafrax crnaw Carlos Hirsch. SRL Akateeninm Kt*akauppa Univerity of Malya Cooperative Intentiaonal lbssipltin Saoser Gale,iacumem P.O. BSo 12 Bork}hop, Limited P.O. BSo 4109S Florida 165, 4th Floor-O, 453/465 SF4-tOI P.O. Box 1127. Jlan PNtnb Baru CrD*w 1333 Buea Airm Heldnki 10 Kuala Lumpur jd m r ber 2D24 AUSTRALIA. PAPUA NEW GUINEA. FRANCE MEXICO SPAIN FIRIL SOLOMON ISLANDSr World Bn'x Pbti,u m, bFFOTEC Mundi-Premcc LIby. SA. VANUATU, AND WESTERN SAMOA 66, *venued'Icra Apartado P." 224E0 Ca eill 37 DA. BSoks &joumals 75116 Pais I4060T IpaoMelcD.P. 2091 Madtid 648 Whitehorse Road Mitdram 3132 GERMANY, FEDERAL REPUBLIC OF MOROCCO Lirbs lehad-oA AEDOS Victoria UNO-Verlg Soeted'Ete dEhAMaketicgMcrM ie ConselldeCent,391 PApPelPdorfe, AlleeSS 12 rue Moa Sd. d'Anfa 08009Barcesna AUSTRIA D-539 Stowl I Casblanca Gerold nd Co. SRI LANKA AND THE MALDIVES GAboen 31 GREECE NETHERLANDS Lake House Bo4 ksh.p A-l0l1 Wiro KEME InOr-Pohlikhtesb.v. P.O. Box 244 24, Ippodamou Street PFatia Plastra P,O. BOx 14 100, Sir Chittmnplam A. Gardiner BAHIRAIN Afhens-1l635 7240 BA Lxchem Mhawtha Sabrain Reearch cnd Conauolaa.cy Calombe 2 A_ciater Ltd GUATEMALA NEW ZEALAND P0. Bo. 22103 Lib,e-iaaPied Sacntc Hill. Library cd Ir timahcerv.ci SWEDEN Marnama Town 317 C95tr Cultural Piedra San Privatelag PFni.8la fir 11 callr 6-RI cIru, I New Market FritPcd.cbo1u*etage BANGLADESH Goatereala Cty Auckbnd Regdingsgatan I2, Bc, 1o656 Miro Indutries Del.punl S-103 27eSteckhopme Asmace Society (MIDAS) HONG KONG, MACAO NIGERIA Hocre S, Road 16 Asia 20 Ltd. Unirverity Yree Limited F r l a Diauommodil/Area Mongkk Pc.I Office ThreCrownsBuioditgjenicho Wenne-grrn-WiDLhaeeAB Dhak. 1209 ButeStneetN. 37 PriveteMailS SC91 Bo" 3DD4 MongkokI Kowico Ibadcn S-10425Stdhelm 8mssh'fica Hrng Kceg I6 Nor Ahmned Sark NORWAY SWtTZERLAND Cilttagong 4D01 HUNGARY Naven Intfsmatier, Center Fc2sila Itbilk Kultuir Bertrand Naereos vei 2 lbrairie Pcyot 76, KDA. Avenue P.O, Bo. 139 P.O. Box 6125 Entad 6, rue Greaus Kulna 13E9 Budapest 62 N-O6G2slo 6 Cse F,pt 381 CH-12121 CGce- 11 BELGIUM INDIA OMAN Pobli.tircccdeaNgie.-Urdea Allied Poblhhm Private Lid. MEMRB InfermaltionService Fosubscri$simcMerdn: APubca.on 202 751 MsltRoad P.O. B- 1613, SebAirport [iteaie Peyct 1060 l or Madr - 600 002 Murc3t Secrice de Abononemnt C.ase peet 3312 BRAZIL B rrAri deo M PAKISTAN CH1epW2I3aoaeru PobllccseeeTe,,ica Inod5isIJ.N. H-rdic Marg Mire Seek Ag-r.q Ltdad Bon-rd Estate 65, Shahrhuaid-A,aun TANZANIA Rua P6-txc Gotnidc 229 Bomrbay .4009025 P.O. Soc N. 729 Oaford Universiy Fres. 01 409 S.. Paulo, SP Leh.. 3 PD0. Sox 5299 13/4 Asrf Ali RedL Dar e 53laum CANADA New Delhi t11022 PERU L.Dtlfueeur Editorial Desarrollo SA THAILAND CFP oS,159BlrueArnpre 17 Chittaranjn Avenue Apertade 3924 Central DepartmentStie Bod.trviDe% QOebec C.cl-tte-700072 Lima 306 Silom Road j4B 586 Bangkok Jcydeva Hortd BSildlng PHILIPPINES CHINA Sth Main Road Gdhinagr Natiotal Book Stre TRINIDAD & TOBAGO, ANTIGUA Chine Fineactal & Ecnomic Publihing Bangalore - S10 D9 701 Riecl Avenue BARBUDA BARBADOS, Iloousc3-5-1129KDrhigudaCror d P.O Bo 1934 DOMLNICA, GRENADA, GUYANA, S, Da Fe Si Dang jie 3.5-1129 Kddigod. Cross Reid MetreManfila JAMAICA, MONTSERRAT, ST. 3eijng Hyderabad - 50 027 KlTTS & NEVIS, ST. LUCIA, POLAND ST. VINCENT & GRENADINES COLOMBIA Pracmana ELa, 2ld Floer ORPAN Sytenmta Studies Unit rolace lid.. Near Thikore Bag, Navrangp-ra Patac Kultury i Nauki #9 Watt Street ApartAdoAero34270 Ahoedbad-380009 CI-901 Warsea Curepe Bogota D.E. Trinidad, West Indier PaaWc House PORTUGAL COTE D'IVOIRE 16-A Ashok Ma.3 U-rria Portuga; TURKEY Centre d'Edition et de Diffusn L -cknow - 226 0O1 Rua Do Carmo 7G 74 Haret Kitapevi, AS Afticai-e (CEDA) 1209 Liebon letildal Caddes No. 469 04 BSPF 541 INDONESIA 1epo de o Abidjo.04F Pteau Pt Indira Liited SAUDI ARABIA, QATAR Istanbul ,1. Sam Ratulat gi 37 J]Birw Sck Store CYPRUS P.O. Box 181 P.O. BoS 3196 UGANDA MEMERB Irforealln Sevices Jakarta Purat Riyadh 11471 Uganda Bokhop P.O. BOX 2298 P.O. Bo. 7145 Nicesia IRELANID MEMRB Infuertetieo Svimie Kampala TOG Publihers Brochftp DENMARK 2 LNeth Frederick St Al Ale" Strt UNITED ARAF EMIRATES Samftard.Litterarhr Dublint I Al Dahat Cente MEMRB Golf C.. EcsenneomoAl161l. First Fleer P.O. Sec 6097 DK-1970 Frederiksbg C ITALY Pi Se- 7185 SRaydh lioas, Coorieienria Sanscr SPA Riyadh DOMINICAN REPUBLIC Vt. Seeedetto Pertml, 120/10 UNITED KINGDOM Editor. TalTer, C. po A. CueM. sPF et 22 HajiAbd.uah AhretBuilding Mi-ofe Ltd. Reetararadene labeL a Catlit 309 50125 flrence King Khald Stret PO Boc 3 Apartado Postal 219B P.O. See 3969 Alton HamprireGU34 2PG Santo Dcdg. JAPAN Deananu England Eastern Sock Service EL SALVADOR 37-, Honge 3-Chome, BuSkyoku 113 33. Mohammed Haan Awd Street URUGUAY fonades Tokyo P.O. BoSee978 Inatituok Nctonal dd Ubre A-medc Manuael Enrique A-ajp 11130 Jedd.h SanoJeIll6t EdifidoeSISA, e. Piw KXENYA Montvdoc Sc. Saleder Africa Book Seevce(EA ) Ltd. SINGAPORE, TAIWAN, MYANMARM P.O. Se. 45245 BRUiNEI VENEZUELA EGYPT. ARAB RPANUBLIC OF Nairob Infermetion Publicattions libreh del Eate Al Abeam Prirate, lid. Aplde. 60337 Al GalIa Sreet KOREA, REPUBUC OF 02-06 lt Fl, Pei-Fu Industrial Cacac 1060-A Cairo Fam Korea BoSk Corporation Bldg. P.O. Se. 101, K.vagwhamo 24 New Industria Read YUGOSLAVIA TheMiddle Bot bserver PeOl Singapor w 2853 Jugslovenc KejIga S Chawarb Stre P.O. Boo 36 Cairo KUWAIT 301110 AFRICA, BOTSWANA Tag Rtepublike MEMRBS nformation Service Fareigle i OtiS. YUAll1 eBlgrede P.O. BS 5465 Wafrd Ueiveraty Preas Southem Afric,& P.O. BS 1141 Cape Town MM I The World Bank Headquarters European Office Tokyo Office 1818 H. Street, N.W. 66, avenue d'lena Kokusai Building Washington, D.C. 20433, U.S.A. 75116 Paris, France 1-1 Marunouchi 3-chome Chiyoda-ku, Tokyo 100, Japan Telephone: (202) 477-1234 Telephone: (1) 40.69.30.00 Facsimile: (202) 477-6391 Facsimile: (1) 47.20.19.66 Telephone: (3) 214-5001 Telex: wui 64145 WORLDBANK Telex: 842-620628 Facsimile: (3) 214-3657 RCA 248423 WORLDBK Telex: 781-26838 Cable Address: INTBAFRAD WASHINGTONDC 0 O) 0> (D Cyt8 H 0 Cover design by Bill Fraser 0-8213-1556-0