University of Southern California Dissertations and Theses

Weight variations among preschoolers: An analysis of evidence from rural Tamilnadu

(USC Thesis Other)

Weight variations among preschoolers: An analysis of evidence from rural Tamilnadu

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content WEIGHT VARIATIONS AMONG PRESCHOOLERS: AN ANALYSIS OF EVIDENCE FROM RURAL 1AMILNADU by Anuradha Khati Rajivan A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Economics) August 1991 Copyright 1991 Anuradha Khati Rajivan UMI Number: D P23382 All rights reserved INFORMATION TO ALL U SER S The quality of this reproduction is d ep en d en t upon the quality of the copy subm itted. In the unlikely event that th e author did not se n d a com plete m anuscript and there are m issing p ag es, th e s e will be noted. Also, if m aterial had to be rem oved, a note will indicate th e deletion. Disswtafcn PuWisMng UMI D P23382 Published by P roQ uest LLC (2014). Copyright in the D issertation held by the Author. Microform Edition © P roQ uest LLC. All rights reserved. This work is protected against unauthorized copying under Title 17, United S ta te s C ode P roQ uest LLC. 789 E ast E isenhow er Parkw ay P.O. Box 1346 Ann Arbor, Ml 48106 - 1346 UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90089-4015 This dissertation, written by Anuradha Khati Rajivan under the direction of hSX. Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillm ent of re quirements for the degree of ?h.h Be '91 RIG I D O C TO R OF PHILOSOPH Y Dean of Graduate Studies D a te ........ DISSERTATION COMMITTEE C hairperson ..... ACKNOWLEDGEMENTS This research benefitted from the helpful suggestions and support of many. The numbers make it impossible to thank everyone individually and special mention is made of only some. In India: Conversations with Rey Martorell, and also with Sudarsanam from the Project Co-ordination office started me thinking that this study might be worth undertaking. I owe a debt to the Governments of Tamilnadu and India for granting me leave of absence from my regular duties. My thanks are also due to the TINP field workers, mothers and others from the sample villages whose comments and insights continuously inspired me during field research. In California: I acknowledge with gratitude the research training I received from Professor Richard Easterlin, my advisor, under whose guidance this dissertation was written. I particularly acknowledge the encouragement and suggestions of my other committee members, Professors Jeffrey Nugent, Timur Kuran and David Heer — they helped me make many improvements in the draft. Lee Lillard’s Econometrics class gave me confidence to get my hands dirty with real data. He also gave me useful tips for the field research and computerizing of data. Professor Cheng Hsiao, and John Strauss at the Rand Corporation carefully read through the econometric results and early versions of the draft and unhesitatingly gave me their time and suggestions. Very special thanks to Rajivan who was (and is) my sounding board, and to little Shyamali who kept me sane during the research and writing of this dissertation. It goes without saying that the faults that remain are all my own. CONTENTS Page ACKNOWLEDGMENTS LIST OF TABLES LIST OF FIGURES vrn u v Chapter I. INTRODUCTION 1 1.1 Genesis of the Study 1.2 The Questions 1.3 The Data 1.4 Organization of the Study II. WELFARE, HEALTH AND NUTRITION, AND ANTHROPOMETRY 13 2.1 Three Approaches to Welfare 2.2 Health and Nutritional Status 2.3 Socioeconomic Factors and Anthropometric Outcomes 2.4 Anthropometry and Choice of Weight as a Measure 2.5 Anthropometry and Welfare 3.1 Program Description 3.2 Demand Factors Inhibiting Program Impact 3.3 Concluding Remarks IV. SURVEY DESIGN, SAMPLE SELECTION AND DATA 59 III. THE TINP PROGRAM 47 4.1 The Survey Design 4.2 Sample Selection 4.3 The Data 4.4 Data Sources m V. CHILD WEIGHTS: LEVELS AND TRENDS 79 5.1 Some Individual Growth Curves: TINP Sample 5.2 Mean Weights Based on the TINP Sample 5.3 Comparison With Weights Based on Other Indian Data 5.4 Summary VI. METHODOLOGY 105 6.1 Choice of Variables 6.2 Variable Description 6.3 Econometric Procedure 6.3.1 Parameter Estimation 6.3.2 Hypothesis Testing VII. REGRESSION RESULTS: GROUP EFFECTS AND EFFECTS OF OTHER COVARIATES ON CHILD WEIGHTS 122 7.1 Group Effects 7.2 Effects of Other Covariates 7.3 Conclusion VIII. REGRESSION RESULTS: SEX AND BIRTH ORDER INTERACTIONS 8.1 Sex Interactions 8.2 Birth Order Interactions 8.3 Conclusion 164 IX. CONCLUSION 187 9.1 Summary 9.2 Policy Implications 9.3 Limitations APPENDIX REFERENCES 200 207 LIST OF TABLES Page 1.1 Percentage Literate in Tamilnadu Among Population Aged 15 Years and Above, 1981 2 1.2 Incidence of Poverty in Tamilnadu 3 2.1 Energy Intake of New Guinea Adults 44 2.2 Relative Values of Physical Measurements Under Field Conditions 45 2.3 Numbers of Children Out of a Population of 2019 Who Would Have Been Identified and Treated for Malnutrition Using Weight-for-age 46 4.1 The Survey Design 63 4.2 % Area Sown More Than Once to Total Cropped Area for Selected Districts, 1983 - 84 & 1984 - 85 65 4.3 Normal Annual Rainfall 66 4.4 Mean Total And 6-36 Months Old Population Per Block For Selected Districts 66 4.5 Population Profiles of the Selected Blocks, April 1988 68 4.6 Composition of the TINP Sample 71 5.1 50th Percentile Values of Weights of Well-to-do Hyderabad Boys and Girls 104 5.2 50th Centile Values of Weights of Male and Female Indian Children of a High Socio-Economic Group 104 v 7.1 Reduced Form Regressions for Weight: Base, Sibling and Current Data 7.2 Reduced Form Regressions for Standardized Weight: Base, Sibling and Current Data 7.3 Reduced Form Regressions for Weight With Dummies for Age Groups: Base, Sibs and Current 7.4 Reduced Form Regressions for Standardized Weight With Age Group Dummies: Base, Sibs and Current 7.5 Reduced Form Regressions for Weight and Standardized Weight With Dummies for Water and Pucca Road 7.6 Least Square Mean Standard Weight by Categories of Land Operated 7.7 Least Square Mean Standard Weight by Levels of Maternal Education 7.8 Least Square Mean Weight and Standard Weight by Sex 7.9 Least Square Mean Standard Weight by Whether SC 7.10 Least Square Mean Standard Weight by Whether Father is HH Head 7.11 Least Square Mean Standard Weight by Development Block 7.12 Least Square Mean Standard Weight by Water Quality and Pucca Road 7.13a Least Square Mean Standard Weight by Age Groups: Sibs and Current 7.13b Least Square Mean Standard Weight by Age Groups: Base and Current 8.1 Reduced Form Regressions for Standardized Weight With Sex Interactions: Base, Sibs and Current 8.2 Least Square Mean Standard Weight by Sex From HHs Operating Different Categories of Land: Base Data 8.3 Least Square Mean Standard Weight by Sex From HHs Operating Different Categories of Land: Sibling Data 8.4 Least Square Mean Standard Weight by Sex From HHs Operating Different Categories of Land: Current Data 8.5 Least Square Mean Standard Weight by Sex From SC Versus Non-SC Households 8.6 Reduced Form Regressions for Standardized Weight With Birth Order Interactions: Base, Sibs and Current 8.7 Regression Coefficients Computed for Birth Order Interactions APPENDIX 1 Mean Weights and SD by Age, Pooled TINP Sample 2 Female Mean Weights and SD by Age, TINP Data 3 Male Mean Weights and SD by Age, TINP Data 4 Female and Male Mean Weights by Age for Years Between 1974 & 1982: NNMB Data, Tamilnadu 5 Mean Weights by Yearly Age Groups: Pooled TINP Data, 1980 - 1989 6 Female and Male Mean Weights by Yearly Age Groups: TINP Data, 1980 - 1989 2.1 Growth Chart of an Actual Central American Child 2.2 Jasmine’s Percentile Chart of Her Weight: A Case of Enquiry Into Child Abuse 2.3 Growth Chart of an Actual TINP Child, Tamilnadu 2.4 Variations in Height at Age 7 by Socioeconomic and Ethnic Groups 5.1 Growth Curves of Two Sisters From TINP Sample 5.2 Growth Curves of Two Siblings From TINP Sample 53 Mean Weights: Current, Sibs and Base Children 5.4 Mean Weights: Currently Enrolled Males and Females 5.5 Mean Weights: Male and Female Siblings 5.6 Mean Weights: Base Line Males and Females 5.7 Female Mean Weights in Kgs., 1974-1982 5.8 Male Mean Weights in Kgs., 1974-1982 5.9 Male and Female Mean Weights, 1980 -1982: TINP Sample and NNMB Data 5.10 Mean Child Weights in Kgs., 1980-1989 5.11 Mean Child Weights in Kgs., 1980-1989 5.12 Mean Child Weights in Kgs., 1980-1989 100 5.13 Male and Female Mean Weights, 1980-1989 101 5.14 Male and Female Mean Weights, 1980-1989 102 5.15 Male and Female Mean Weights, 1980-1989 103 7.1 Least Squared Mean Standard Weights: Base, Sibling and Current Data 145 ix Chapter I 1 INTRODUCTION The aim of this research is to examine and explain the patterns of weight variations observed among preschool children under three years of age, associated with individual, household and community level variables that are largely exogenous to the households from which the children come. These variables, in combination, can be thought of as making up a large part of the environment in which children grow - the micro household and the macro community. Human weights at a given age (and, in fact, other auxological data like the tempo of maturation, heights, etc.) are the outcome of both genetic and environmental factors. Among environmental factors the adequacy of nutrition and the state of hygiene constitute important elements of the determinants of child weights. To a certain extent data on children’s weight can serve as indicators of aspects of well being and of socioeconomic inequalities within and between population groups, at a point in time and also over time. This is especially true when genetic factors are largely controlled. To quote from Bielicki (1986), "These social uses of auxological data form a research area that Tanner (1981) has called epidemiological auxology. It lies on the borderline between human biology and the social sciences, perhaps even more within the sphere of the latter than of the former, for its subject matter is not so much the study of processes of growth as such, as rather the utilization of growth data as an instrument for monitoring the social and economic situation of human communities, and particularly of economically disadvantaged groups." The evidence for this study comes from rural Tamilnadu, a state in south eastern India. Tamilnadu’s population was 48.41 million in 1981 and is projected to be 56.41 million in 1991. This size is similar to countries such as France, Italy, Egypt, Iran, 2 Turkey and Thailand, and larger than countries like Colombia, Ethiopia and Morocco. Of the total population, 67.02% live in rural areas. This relatively large rural population, low literacy rates, with percentage of population literate ranging between 23% to 51% among different population groups (table 1.1), the high incidence of poverty (table 1.2), a relatively high infant mortality rate (IMR) of 93 overall and 103 in rural areas, provides an opportunity to draw conclusions which would be of relevance to other less developed countries as well. A health and nutrition program, the Tamilnadu Integrated Nutrition Project (TINP), has been in operation in rural Tamilnadu since late 1980. Apart from providing a range of health and nutrition inputs to the preschool age population, and education inputs to the comm unify, the program regularly monitors the growth of children between 6 to 36 months of age on the basis of their weights: hence monthly weight records for a high percentage of the target population are maintained. Data from ongoing and past program records, collected by the author in the field, were an important source of information for the present study. Table 1.1 X LITERATE IN TAMILNADU A M O N G POPULATION A G ED 15 YEARS A N D ABOVE, 1981 Category of Population Percentage A ll persons Rural persons Females Rural females 51 A O 35 23 Source: Census of India 1981, paper 2 of 1983. 3 Table 1.2 INCIDENCE O F PO VERTY IN TAMILNADU, 1983 In d icator* Rural Urban 1. Head count ra tio 39.94% 40.45% 2 . No. of persons in poverty (m illio n s ) 13.24 20.18 ( * ) Represents the % and number of people below a poverty lin e level of expenditure. Source: Madras In s titu te of Development Studies, 1988. 1.1 Genesis of the Study The use of anthropometry - the systematic measurement of the human body - has been of interest to researchers in many fields. Heights, weights, mid-upper-arm- drcumference, skin fold over triceps, head circumference, etc., have been used by nutritionists, doctors and public health agencies to assess the growth, health, nutritional status and general standard of living of individuals and groups. Among the various anthropometric measures, height and weight are more commonly used, and are also more commonly understood. Systematic differences in height and weight of individuals have been recognized as reflecting their different health and nutritional status, especially after controlling for genetic factors (Eveleth and Tanner, 1976; Field et al., 1981; Chen et al., 1982; Martorell et al., 1984; Pacey and Payne, 1985; Floud et al., draft). In recent years economists have also used weight and height data to draw inferences about health, 1 The poverty line level of expenditure for these figures is the same as that adopted by Ahluwalia et al., (1979) in arriving at poverty estimates for other developing countries, and by the Planning Commission in India as well. It roughly corresponds to an expenditure level at which food consumption, on an average, provides 2250 calories per person per day. While the sanctity of this, or any other fixed norm is debatable, it must be admitted that those below even this level are absolutely and relatively deprived. 4 nutrition and, hence, aspects of welfare (Sen and Sengupta, 1983; Behrman and Deolalikar, 1985; Sen, 1985a; Horton, 1986; Barrera, 1987; Strauss, 1990). Among economic historians and those who study historical demography as well, there is an increasing interest in exploring anthropometric measurements and their relationship to living standards (Floud and Watcher, 1982; Fogel, 1986a, 1986b, 1987; Floud et al., draft). Anthropometric data have been used for comparisons between population groups and individuals, and to make inferences about the health, mortality levels, the environment and general standards of living. The preschool age is the period of an individual’s life when growth is the most rapid, apart from the adolescent growth spurt. At times when human growth is normally the fastest, it appears to be particularly sensitive to outside influences: for example, inadequacy or irregularity in food consumption, infections, and even child abuse. An individual child’ s growth can be seen responding, both, to deprivation, and then to the removal of threat to its welfare. As observed in Floud et al. (draft), "Weights will normally be first affected, but height growth and change in other anthropometric indicators will soon follow. Similarly, if the danger recedes, weight, height and other signs of growth will return; indeed the child will grow faster than before, a phenomenon known as ’catch-up growth’, to return to the growth path which it was previously following." Early childhood is, perhaps, the most vulnerable period and sustained deprivations experienced then, resulting in growth impairment, can lead to permanent stunting (lower heights) of those individuals as adults (Martorell and Habicht, 1986). Removal of deprivations at older ages does not result in ’catch-up growth’ later. Using more than one anthropometric indicator is undoubtedly better than relying on any one of them alone, as the different measures are not equally sensitive to 5 deprivations such as inadequacy of food or disease, and capture slightly different aspects of human growth, health and nutritional status. Weight is recognized as a short term indicator and height a more long term one. While a more detailed comparison of height and weight based on existing literature is carried out in chapter II, suffice it is to say here that, ideally, it would have been better to analyze both heights and weights, but the nature of the data constrained choice to weight. However, two practical reasons suggest that in the case of rural preschool children from LDCs weight could be a preferred alternative if a choice had to be exercised under field conditions. First, weight measurements are easier to record accurately, given the inability of preschoolers to stand erect, even when the child is fretful and uncooperative. Because of considerable measurement error problems, especially among children under 2 years of age, the taking of heights was discontinued in TINP program areas. Second, unlike among adults in DCs or the well off urban populations in LDCs, conscious decisions to reduce weight are not common for preschool children. In LDCs obesity among children from poor households is not known to occur, and especially not in rural areas (Eveleth and Tanner, 1976). Both height and weight deficits among children in developing countries are considered as being caused by similar factors, though on a different time scale (Floud et al., draft). Hence, the use of weight data for preschoolers under 3 years of age appears reasonable for the present study. Tamilnadu is chosen because of the operation of the program there, with detailed and uniform record maintenance for currently enrolled and past participants, making it an inviting situation for research. Weight records available for this study extend over the period October 1980 to April 1989, i.e., from the time of program inception to the time field research was undertaken by the author. 6 12 The Questions The study focuses on analyzing weight of preschoolers. The specific questions posed are as follows: (a) What are the levels of weights for different ages observed in the TINP data set and are there any changes in the means over time? How do TINP weights and trends compare with those observed in other Indian data? (b) How does child weight vary with individual level variables like child age, sex and birth order; with household level variables like economic status, maternal education, maternal height, caste and household structure; and community level variables like region, distance to a health facility or hospital, kind of water supply and ’ pucca’ 2 road connections? (c) If, as observed in countries of south Asia, there exist weight differences between males and females, are there significant interaction effects between the sex of the child on the one hand, and household economic and social status on the other? In other words, does the male-female gap differ significantly across households of varying economic status, and between ’ scheduled caste’ versus other households? (d) If, as often observed, birth order has a negative effect on weight, to what extent does mother’ s education and household economic status counter this? 13 The TINP Data The data set was put together on the basis of 6 months of field research by the author in Tamilnadu, during February to July 1989. The TINP data set consists of data from 854 rural households spread over 16 villages in 4 development blocks chosen from the 2 earliest program districts in Tamilnadu. Given resources, an attempt was made to capture as much variability in the data as possible, in order to relate child weights to many of the exogenous circumstances with which they might be expected to vary. Since the 2 Strong or firm: a road with black topping or concrete, suitable for vehicular traffic, as against a path or dirt road. 7 concern was with preschool children, only households that, at a minimum, contained adult- child pairs, with the child having participated in the program, were considered. The data were gathered by the author from 3 sources: (a) ongoing and past program records; (b) a survey conducted among households of the sample children; and (c) actual measurements of mothers’ heights. Belonging to the Indian Administrative Service and having been associated with the program as its Project Co-ordinator, the author was fortunate in having personal rapport with the program staff, easy access to records and official documents, and personal knowledge about the program and the areas in which it operated - a combination of circumstances that helped minimize problems normally faced by other researchers. Ongoing and past program records provided information on a number of individual level variables like date of birth, sex, monthly weight observed, the nutritional grade in which the child was placed at each weighing, etc Fortunately, the village nutrition centers have kept the old records on children who participated right from the early program months. These provide data for the earliest program months in each block - the base line data for the present study. Data on current participants was readily available in the centers, and data on their older siblings were drawn from past records. A survey was conducted by the author among households of the sample children, using pretested questionnaires translated into the local language, to obtain socioeconomic and other household and community level information. Further, extensive travelling in the sample villages and numerous (though haphazard) conversations with mothers, village elders, local leaders and other curious persons in the countryside suggested some useful insights. 8 It was considered important to obtain information on mother’s heights in order to capture genetic and other family background effects. Hence actual measurements of mother’s heights were taken in each case the mother was available in the village. Mothers who had left the village or died had to be excluded. A second visit was made in all cases where the mother was temporarily unavailable. This procedure yielded height data for a little over 91% of the mothers. Height measurements were taken with the help of village level workers who knew the mothers well and were already trained in measuring and recording child weights. Father’ s heights, while no less relevant, were not obtained. Local enquiries indicated that fathers would be much less willing to cooperate in being measured and were less likely to be available in the villages during daylight hours. The resulting data set includes monthly weight observations for currently enrolled preschool children, for older siblings of the currently enrolled, and for children enrolled when the program had just been introduced in each of the sample villages, taken, in each case, when they were 6-36 months of age. The data set also contains information on a number of individual level variables (age, sex, birth order, etc.), household level variables (land, kind of dwelling unit, maternal education, maternal height, caste, identity of household head, etc.), and community level variables (region, distances to a health facility and hospital, kind of water supply, etc.). These variables, largely exogenous to the household, at least in the short run, are used as regressors. Data on a number of other variables that may come under the purview of household choice, like duration of breast feeding, utilization of health care and hospital facilities, the number of children, etc., are available, but as explained in chapter VI, not used, since they are expected to be determined simultaneously with weight outcomes. 1.4 Organization of the Study 9 Chapter II contains a review of the various approaches to well being in economics in order to identify an approach that allows one to look within a household and examine individual well being separately from household welfare. Such an approach permits a distinct examination of individual well being allowing one to focus on preschoolers even as they typically live in households, exercise little choice on their own, and whose well being is affected by household decisions about resource allocation including food consumption and health care. This review suggests that behavioral models of a homogeneous household with a single utility function may not be appropriate for the problem at hand. Sen’s capabilities to function approach (Sen, 1984, and 1985a), together with the concept of physical functional capacities of individuals as described in Pacey and Payne (1985), seem more appropriate. Based on this approach inferences about welfare are made using data on direct observations of the physical states of individuals. The chapter also contains a discussion of the interrelationships between health and nutritional status and how socioeconomic and other environmental factors influence growth, health and nutritional status. While recognizing that weight (and height) outcomes are always determined by a combination of both genetic and environmental factors, it is nevertheless possible to ascribe most weight variations to environmental factors when genetic factors are controlled for. Further, differences in body size due to environmental factors within ethnic groups have been observed to be much greater than differences across ethnic groups for comparable socioeconomic classes (Martorell and Habicht, 1986). This is especially true in childhood. Therefore this study will be mainly concerned with the influence of environmental factors. Data on height and weight have 10 been used in the literature to draw inferences about growth, health and nutritional status. The relative strengths of indicators based on height and weight are examined and the reasons for choosing weight are explained. Chapter m contains a detailed description of the program, the Tamilnadu Integrated Nutrition Project (TINP). The program started in late 1980 and was introduced in a phased manner in 6 districts of Tamilnadu. The 3 earliest phases of the program, which are the source of data for the present study, cover 6 to 8 1/2 years of program operation until the survey was completed in April 1989. The program provides a number of health and nutrition related services to children under 3 years of age: growth monitoring, diarrhoea management, deworming, vitamin A administration, supplementary nutrition, related education to mothers, etc. Most of the program services are provided at the village level in local community nutrition centers. In most cases it is possible simply to walk to the center, making the costs of participation minimal (in terms of time spent or wages foregone). This is probably an important contributory factor for the high participation rates observed. The program m aintains uniform and detailed records on current and past participants, making available a rich data source. Chapters IV contains a discussion of the survey design and sample selection procedures adopted in the course of field research by the author. It also contains a description of the sample data and its sources. Keeping in mind the aims of this study, an attempt was made to fulfill the following conditions as completely as possible: (a) the sample had to be representative of the situation prevalent in rural Tamilnadu and also to provide sufficient variability; (b) it had to be selected cost efficiently, yielding maximum possible information for given resources; and (c) it could not be allowed to be so scattered 11 as to make supervision, quality control and uniformity of interpretation by the interviewers difficult to ensure. A problem with the sample is that all information relates to children participating in the program. This usually leads to a selectivity bias. However, since participation rates in the program are high, this lessens the seriousness of the problem. The average percentage of children under 3 years that participated ranged between 70% - 80% in the early months in each of the program’s phases, with the percentage rising to over 90% soon after (Office of the Project Coordinator, TINP). An issue related to using information on only participating children is that it becomes difficult to have a good control group to evaluate program effects. Even though program evaluation is not the primary concern here, some tentative conclusions about program effects may nevertheless be drawn by comparing trends in mean weights for Tamilnadu as a whole (based on other data) with the trends observed in the sample data from TINP areas. Chapter V looks at question (a) posed earlier. The levels and trends in weights in the TINP sample areas for different ages are examined for the period 1980 to 1989. These are compared with average weights observed in other Indian data sets. As against a lack of trend in other data, an upward trend in TENP weights for different ages is observed. Further, weights for males are found to be significantly higher than for females of corresponding ages. Before going on to the remaining questions, chapter VI contains a discussion on some methodological issues including choice and description of variables. Chapters VII and VIII address the remaining questions (b) to (d) posed earlier - to examine and attempt to explain the weight variations observed between different groups and those associated with other variables. Regressions for weight and age- 12 sex standardized weight scores are run to estimate the effects of individual, household and community variables. Some of the more interesting results are the negative effects of higher birth orders, being female and belonging to a nuclear household, and the positive effects of maternal education, belonging to landed households and better water quality. With respect to the sex of the child, being a female starts out as being relatively disadvantageous, with the relative disadvantage diminishing over time. When male and female weights by household economic and social status are examined, it is seen that the sex bias operates largely among the better landed households. It does not operate among the landless. Further, it is less in evidence in the socially disadvantaged scheduled caste households. Tentative explanations for such results are given. Birth order has a significant negative effect on weight. Maternal education has a positive and significant effect. Interactions between maternal education and birth order suggest that maternal education partly compensates for negative birth order effects. In contrast, interactions between land operated and birth order indicate that better household economic status does not systematically compensate for negativg birth order effects. Possible explanations for these results are examined. In chapter IX, after a summary of the results, some implications for policy are suggested and limitations of this study are pointed out, together with suggestions for further research. 13 Chapter n WELFARE, HEALTH AND NUTRITION, AND ANTHROPOMETRY Examining well being is of wide relevance in economics. It is central to welfare economics, important for any study of poverty, for inequality, for measuring living standards and assessing economic development. Keeping in mind that preschoolers are the focus of the present study, section 2.1 reviews three approaches to welfare in order to search for an approach that allows one to look within the household and examine separately the well being of some of its members, i.e., children. The approach chosen must also be capable of permitting the examination of inequalities between children within households. Using the physical functional capabilities approach, section 2.2 contains a discussion of the concept of nutritional status and its interrelationship with health based upon existing literature. In section 23 how socioeconomic factors influence health, nutritional status and growth as measured by anthropometric outcomes is examined. Anthropometric measures based upon height and weight are among those more commonly used to make inferences about child growth, health and nutritional status. These are described in section 2.4. The strengths and drawbacks of these indicators and the reasons for choosing weight are also discussed. Finally, in section 2.5, the use of anthropometric evidence to draw inferences about physical functional capacities is justified. 14 2.1 Three Approaches to Welfare Three broad approaches to welfare can be distinguished in economics. While the approaches are not unrelated, they focus on different aspects of welfare. Each of them is briefly examined below and the reasons for choosing the third approach, that of ’functioning capabilities’ of individuals, are explained. First, a strong tradition in economics has been to be concerned with the concept of utility, used in the sense of satisfaction, desire fulfillment1 or even that which a person maximizes2. While the utility based approach does offer insights useful with reference to particular problems, it is less useful when the well being of children needs to be examined separately from the households to which they belong. Traditional models of a homogenous household with a single utility function may not be appropriate, especially so when there could be trade offs between the interests of adults and children, and also between the children themselves, eg., males versus females, or the earlier versus the later born (Gopalan, 1979; Chen et al., 1981; Kynch and Sen, 1983; Horton, 1988; etc.). Moreover, in poof communities where poor health and nutritional status among children are widespread and accepted as part of life, parents could derive no less utility (in the sense of satisfaction or happiness) as compared with other societies where living standards, command over commodities and services and, therefore, expectations are higher. The utility approach, thus, also suffers from ’physical condition neglect’ (Sen, 1985a). For the present study, therefore, it is necessary to look elsewhere. 1 As in classical or much of modem utilitarianism. 2 See Sen (1982) for a critique. 15 Second, well being has also been thought of as the total availability of goods and services. Command over commodities reflects how well off a person is and can have a significant impact upon household and individual well being. Market purchase data or incomes have been used as evidence for this approach, and, in fact, even for the utility approach3. In the present context enhancing household economic status or improving the quantity and nutrient quality of food at the household level may, no doubt, improve the entitlements of children as well, but it may not be sufficient. One reason is when inequalities in distribution within the household are expected. Individual members, preschoolers in particular, do not independently purchase goods and services, especially food, shelter, health care, etc. It is extremely difficult to obtain reasonably accurate evidence on individual consumption of the various commodities purchased by the household. While the observation of individual food consumption has been attempted through diet surveys, the presence of an observer itself has been known to influence intakes. Evidence also shows that differences in well being in terms of living standards (mortality, literacy, life expectancy, etc.) can be much greater than can be explained by incomes. At a macro level Sri Lanka is a case in point with low incomes but a relatively high life expectancy, literacy and a low infant mortality rate (World Development Report, 1984). Moreover, external, non-market factors might influence well being independently of economic status - electricity supply, water quality, waste disposal facilities in the community can influence childhood mortality and morbidity. Like the utility approach, this 3 Conclusions about utility are deduced indirectly from market purchase data in consumer analysis. Economists have, by and large, preferred to avoid direct questioning on matters of satisfaction or happiness. Some notable exceptions have been works from the ’ Leyden school’ and Easterlin (1974). 16 approach also suffers from physical condition neglect. Thus, the command over commodities approach is also not appropriate when children are the focus of interest. Third, well being has been judged in terms of ’functioning capabilities’ of individuals as evidenced by direct observations of their physical states. "A functioning is an achievement of a person: what he or she manages to do or be. It reflects, as it were, a part of the ’state’ of that person. It has to be distinguished from the commodities which are used to achieve those functionings." (Sen, 1985a). Payne has described functional capacity to specifically mean all aspects of behavior in response to the environment - physical and mental activity, response to stress, resisting and recovering from disease, etc. (Payne, 1982). This is very dose to Sen’s concept of ’functioning capabilities’ which flow from individual entitlements. The latter concept is, perhaps, wider. Capabilities to function are outcomes; commodities can be used to affect them, though they result from not only commodity use, but also several non-choice factors like the environment, an individual’s metabolic rate, etc. The basic capabilities of survival and sustenance are still relevant in many LDCs, spedally among preschool children - an age when considerable irreversible damage can take place in the presence of sustained deprivations. In the development literature outcomes like ’ basic needs’ fulfillment, ’quality of life’ indices, literacy, data on mortality in general and infant mortality rates in particular, have been much discussed (Lipton, 1968; Adelman and Morris, 1973; ILO, 1976, 1984; Gwatkin, 1979; Kakwani, 1980; Guhan, 1981, etc.). All these come under this, third, functioning capabilities approach using non-market observations of the living conditions and states of persons as evidence for well being. Because of the problems of using food intake data (apart from the difficulties of observation and measurement mentioned before, there are variable 17 individual requirements and individual adaptability to nutritional inadequacies) Pacey and Payne (1985) use the idea of functional capacity of an individual to maintain adequate4 performance with respect to physical and mental activity, response to stress, resisting and recovering from disease, etc., in defining nutritional status and individual welfare. This is very close to Sen’s concept which, however, is applicable in a wider range of situations. Apart from using evidence on mortality, life expectancy and literacy, anthropometric data are being used as indicators of individual well being and for comparisons between the well being of groups, not only by pediatricians, nutritionists and in the public health literature (Gopalan, 1979, 1983; Pacey and Payne, 1985; Martorell and Habicht, 1986), but also among economists (Sen and Sengupta, 1983; Kynch and Sen, 1983; Vaidyanathan, 1984; etc.) and for historical comparisons (Floud and Watcher, 1982; Fogel et al., 1982). For the present study, then, since the focus is on children, as distinct from the household, with possible inequalities within the household, the third approach is chosen. Preschooler weights are used as evidence of physical well being. This allows an examination of inequalities in the state of individual children within a household and differences between groups. Unlike the utility and the command over commodities approach, the problem of physical condition neglect is better taken care of. Moreover, since child weight in LDCs is influenced not only by household command over commodities, but also by external non-market factors like water quality and sanitation, the 4 In the ultimate analysis what is ’adequate performance’ is a subjective decision, with no unique answer. For health and nutritional status it is now increasingly accepted that there is a continuum of states ranging from obviously inadequate with visible starvation, frequent infectious episodes and breakdown of functioning, through intermediate states with varying risks of breakdown, to adequate functionings. 18 influence of these external factors on the well being of preschoolers is not ignored. The shortcomings of this measure are discussed in section 2.4 below. 2.2 Health and Nutritional Status It is well recognized today that nutritional status cannot be viewed 1 independently of health status and that there are complex biomedical relationships j between an individual’s food intake, nutrient absorption and utilization by the body, | A individual activity levels and the incidence of disease. The concept of nutritional status has undergone considerable change in the last two decades. Nutritional status is no longer understood simply as the outcome of deficiencies or excesses of specific nutrients. Jelliffe (1966) regarded malnutrition as a "pathological state resulting from a relative or absolute deficiency or excess of one or more essential nutrients, this state being clinically manifested or detected only by biochemical, I anthropometric or physiological tests." In many LDCs where the presence of frequent ' infectious episodes is higher (diarrhoea, upper and lower respiratory infections), nutrient absorption and utilization by the body are less efficiently carried out. Malabsorption can i \ result from infestation with intestinal parasites. But, perhaps, diarrhoea is the most \ common reason causing food to pass through the intestine too quickly to be absorbed \ \ I (Scrimshaw, 1977; Chandra, 1980; Martorell, 1980; Martorell and Habicht, 1986). Infection | causes nutritional status to deteriorate, but at the same time undernutrition decreases \ \ resistance to infection - a synergistic relationship (Reddy et al., 1976; Kielman et al., 1976; f * Rowland et al., 1977; Rowland 1984). Thus, while the term ’nutrition’ brings to mind food and nutrients, the term ’nutritional status’ is used, at the very least, to describe an outcome 19 of several biomedical processes interacting over time (Martorell, 1980; Beaton and Ghassemi, 1982; Kielman et al, 1983; FAO World Food Survey, 1983). Simultaneously, in recent years, two distinct concepts of malnutrition, have gained currency. They have been called the ’fixed genetic potential view* and the ’individual adaptability view* (Pacey and Payne, 1985). Both are quite different from what was accepted two decades ago. The fixed genetic potential view operates on the premise that there is a preferred state of health and nutrition, fixed for each individual, determined by the individual’s genetic potential for growth, resistance to disease, longevity, and so on. In order to attain well being, it is assumed that everyone should achieve their full genetic potential. Malnutrition sets in when there is any departure from the preferred state. Since genetic potential cannot be measured, it is inferred from ’ standards’ of body size and food intake in ’ well fed’ populations. Based on this view data on the number of people with food intakes below prescribed levels or body weights below prescribed standards are used to estimate the prevalence of undernourishment in given populations. Under the ’individual adaptability view* nutritionists accept that human beings have the "capacity to adapt to a fairly wide range of dietary situations, and that only when that adaptive capacity is stretched beyond its limits does the body fail to maintain its functional capacity and malnutrition ensues." (Payne, 1982). Information on the number of people with food intakes or anthropometric measurements below standards are no doubt likely to indude an important portion of the ’ malnourished’, but they would not measure the extent of malnutrition in the second sense, which recognizes that people adapt to a great variety of dietary and work regimes, and this sometimes involves modifications in body size, levels of physical activity, and metabolic changes. A small body size would not necessarily indicate malnutrition unless the physical functional capadty of the body is 20 impaired to the point where the individual can no longer maintain ’adequate performance’ with reference to growth, physical or mental work, pregnancy, lactation, resisting and recovering from disease, etc. Malnutrition is a symptom that certain processes are regularly occurring in the lives of people which, if disregarded, will result in impaired functioning. At the very least ’adequate performance’ would entail the achievement of a sustainable mode of existence. Consider the example of two New Guinea communities where there was a wide gap in energy intakes of coastal and highland people (table 2.1). Yet both communities were successful from the view point of survival and health and showed no signs of malnutrition in the sense of impaired functioning. Moreover, body weights in the two communities were not significantly different. The explanation probably lies in the different lifestyles of the two communities. The coastal community needed to spend only a limited amount of time at work on gardens while the highland community spent much longer horns in heavy labor on mountain slopes. Here ’adaptation’ rather than undernutrition would be the appropriate description (Pacey and Payne, 1985). On the other hand, in the case of the Central American child whose growth is checked by several episodes of illness from which recovery is inhibited by low food intake, ’undemutrition’ would seem a more relevant description (figure 2.1). There is virtually no weight gain between the ages of 6 to 18 months with weights being well below the norm after the age of 6 months. Clearly, in this case performance is far from adequate. In intermediate cases it is harder to say with absolute certainty whether a particular weight loss is a part of some normal seasonal cycle, the result of a temporary illness or indicates the presence of malnutrition. 21 Apart from adaptability, human requirements for food vary for a variety of reasons. For adults, if work intensity is high, greater energy intakes than otherwise are needed to keep body weight from falling. This is less important for preschoolers when work requirements are not a consideration since the major physical activity of this age is play and exploration. Requirements can vary with seasons. If a good season follows a bad one, the low consumption of the bad may be compensated by later improved food consumption with little impact on functioning capacity. Successive bad seasons, however, may stretch the body’s adaptive mechanisms far enough to increase the probability of susceptibility to disease or lower activity levels. Among adult females pregnancy or lactation can increase intake requirements. So can age. Contrary to common impressions, young children require relatively large amounts of dietary energy for their body weight. See for example FAO/WHO report (1973) and Martorell (1980). "A one year old child is one-fifth the weight of an adult, but his or her energy intake will be about half the adult level." (Pacey and Payne, 1985). Illness can alter requirements temporarily. Illness commonly results in anorexia and reduced intakes. If there is decreased intake during illness, the recovery period will require supplementation. This is particularly true for children who, in addition to their needs for growth, have smaller reserves. Thus we see that health and nutritional status are interrelated and there is, in fact, a synergistic relationship between infections and undernutrition. Further, in view of the body’s capacity to adapt to a range of diets and situations of stress, below normal anthropometric outcomes, indicating less than normal growth, do not automatically indicate the existence of problems with health and nutrition. They may be due to adaptation taking place. However, as a precaution against underestimating the extent of deprivation, it needs to be kept in mind that the fact of adaptation itself indicates a 22 situation of stress. Poor health and nutritional status may be the next step if the stress is sustained. 2.3 Socioeconomic Factors and Anthropometric Outcomes This section examines the question of how socioeconomic factors operate relative to anthropometric outcomes. In so doing the causes of variations in growth, health and nutritional status are also discussed. Socioeconomic factors, together with community sanitation, water quality etc., make up the environment in which a child grows. When systematic variations in the patterns of growth of different populations are observed, one is naturally led to enquire into what causes these variations. The causes have been separated into two broad categories: heredity and the environment. In the case of adults the first systematic observations of height differences among recruits to the French army were made in 1829 by a French doctor, L.R. Villerme, who explained the differences he observed as follows (translated by Tanner, 1981): "Human height becomes greater and growth takes place more rapidly, other things being equal, in proportion as the country is richer, comfort more general, houses, clothes and nourishment better and labor, fatigue and privation during infancy and youth less; in other words, the circumstances which accompany poverty delay the age at which complete stature is reached and stunt adult height." Thus, a clear causal relationship has been postulated between human growth and a combination of poverty, ill-health and deprivation, i.e., socioeconomic and environmental factors, as early as in the 19th century. But Villerme was criticized for having ignored the more ’commonsensical’ racial and hereditary factors that could contribute to systematic differences in height. For example, in Britain the Anglo-Saxons were taller than the Celtic people (Tanner, 1981). Children of taller parents tend to be taller than those of shorter parents. Today it is recognized that, singly, neither heredity nor environment determine the growth pattern of individuals or groups. It is always a combination of both, interacting with each other, that can explain within and between group differences (Thoday, 1965; Fischbein, 1977; Bergman and Goracy, 1984; Floud et al., draft). Of course, in a particular situation when trying to explain variations in growth, it is possible that genetic factors are more important (for example, when environmental factors are similar), and in another situation environmental factors are more important (for example, when genetic factors have been controlled for). In fact, data collected by the International Biological Program suggest that variations in growth within very large racial groupings can primarily be attributed to the effects of the environment. This is summed up by Eveleth and Tanner (1976): "A child’s growth rate reflects, better than any other single index, his state of health and nutrition, and often indeed his psychological situation also. Similarly, the average value of children’s heights and weights reflect accurately the state of a nation’s public health and the average nutritional status of its citizens, when appropriate allowance is made for differences, if any, in genetic potential. This is especially so in developing and disintegrating countries." The impact of deprivation (eg., inadequacy of food), disease and other factors is observed to be the strongest during periods of most rapid growth - during the neonatal period, in infancy and also during the adolescent growth spurt. It is, therefore, common for pediatricians, health and social workers to examine growth charts of individual children to see if evidence of deprivation exists (eg., figure 2.1). Similar charts have also been used to look for evidence of child abuse (figure 2.2). Under foster care a 24 spurt in growth is noticed, whereas during the period of ’home on trial’ there is a decline in weight followed by death. A child’s growth pattern reflects several aspects of the environment in which it is brought up. To quote from Floud et al., (draft), "... a child requires an adequate5 intake of nutrients to maintain its body, to undertake physical activity and to grow. In addition, it must combat disease and other forms of stress ... Studies of children show that, if normal balance between the supply of nutrients to the body and the demand on those nutrients is upset, for example, by a decline in food intake or by a need to combat disease, then the child’s growth will rapidly be affected. It seems as if the child’ s body attempts to protect itself by sacrificing growth. Weight will normally be first affected, but height growth and change in other anthropometric indicators of growth will soon follow. Similarly, if the danger recedes, weight, heights and other signs of growth will return; indeed, the child will grow faster than before, a phenomenon known as ’catch-up growth’, to return to the growth path which it was previously following." While what is an ’adequate’ intake of nutrients is debatable, the impact of overall entitlements on physical states of preschoolers is unmistakable as is seen from actual growth charts (figures 2.1 and 2.3) and as will be seen when econometric analysis of TINP weight data is carried out in later chapters. Martorell and Habicht (1986) and others have shown that in the 20th century at least, differences between socioeconomic classes cause greater differences in growth of children within a country than differences due to ethnic factors across countries. This is illustrated in figure 2.4 which allows a comparison of the relative influence of ethnic and environmental factors. The gaps between socioeconomic classes within a country are measured by the vertical distance between a dark and a light circle. The gaps 5 What constitutes ’adequate’ food intake or the specification of calorific nutritional norms has been questioned because of interpersonal variability of nutritional requirements and the existence of ’adaptive mechanisms’ operating over time. See Sukhatme (1978) and Srinivasan (1979). 25 across ethnic groups, controlling for socioeconomic status, are the level differences between two or more dark circles or between two or more light circles. The figure shows that differences in body sizes between well nourished children across ethnic groups are minor as compared with differences across social classes within a broad ethnic group. Similar conclusions for weights have been obtained by the National Nutrition Monitoring Bureau, Hyderabad, with data from different states in India over the 1977-1982 period. Among the environmental factors contributing to growth variations adequacy of nutrition and/or the incidence and severity of infections are identified as two immediate or proximate factors whose contribution is most significant. It is interactions between these two (quite complex in themselves) that finally determine nutrient availability at the cellular level, an important outcome of which is growth. But it is now well recognized that the causes of growth failure and inadequate health and nutritional status cannot be found only in the immediate interrelationships between bio-medical factors alone, like food intake, nutrient absorption and utilization by the body, individual activity levels and the incidence of infections and disease. There are important underlying social and economic and community considerations that make up the environment, which ultimately influence child weight and height through their effects on the bio-medical factors (Marchione and Prior, 1980; Pollit and Leibel, 1980; Pacey and Payne, 1985; Bielicki, 1986; Martorell and Habicht, 1986). This is especially true in developing countries. In fact the very significant correlations between socioeconomic status, living standards and human physical states (height, weight and even signs of puberty like menarcheal age) has led to the social uses of auxological data (Tanner, 1981). Such studies while combining human biology and the social sciences, are perhaps more within the sphere of the latter, for their focus is less on the study of the process of growth as such, and more on the 26 utilization of growth data as an instrument for monitoring the social and economic situation of human communities, and particularly of economically disadvantaged groups (Bielicki, 1986).6 Environmental factors can operate at the community or the household level. While the immediate antecedents of growth faltering may be bio-medical, considerations of the following kind point to the underlying influence of the socioeconomic environment in which a child grows: (a) The level of income and fluctuations in it are relevant. Higher and steadier incomes allow higher and steadier entitlement possibilities for each member in the household. Lower nutrient intakes, other things being the same, have not only been directly related to lower heights and weights, but because of the resulting lower resistance to infections, have also been found to interact with disease incidence, further adversely affecting growth (eg., Rona et al., 1978, for England; Prokopec and Lipkova, 1978, for Czechoslovakia; Janes, 1970, for Nigeria; Bielicki, 1986, for Poland, etc.). (b) If alternate demands are heavier on a woman’s time or on the time of whoever the care giver is (i.e., demands other than child rearing), the less attention she will be able to devote to food distribution for very young children and to breast feeding. To the extent younger siblings are left in the care of older ones or largely left to fend for themselves, they are likely to be able to eat less adequately (Kumar, 1983; Pinstrup- Andersen, 1983; Katona-Apte, 1983; Piwoz and Viteri, 1984). 6 There are some societies in which class differences in growth have either disappeared or are rapidly disappearing. No significant differences in heights and weights of Swedish urban children across socioeconomic strata were found during 1964 - 1973 (Lindgren, 1976). Similar results were obtained among Oslo school children in 1975 (Brundtland et al., 1980). In these societies, while classes have not disappeared, environmental deficits do not seem to inhibit children’s physical growth and maturation in lower and upper classes. 27 (c) Lower perceived needs of children in the community resulting in poorer diets have been known to result in lower height and weight. This has been explained as a mechanism of cultural adaptation in a stressful environment. Children are less important to the survival of the household unit than working adults (Marchione, 1981). Sometimes cultural beliefs and food taboos like withholding protein rich foods from young children under the presumption that they cause flatulence, have been associated with lower height and weight outcomes. (d) Environmental factors, like the kind of water supply available for drinking (protected or not) have been known, through their impact on the incidence of disease, to afreet growth. When the incidence and severity of infectious episodes is higher (diarrhoea, upper or lower respiratory infections, etc.), lower rates of growth have been observed (Puffer and Serrano, 1973; Chen and Scrimshaw, 1983). Apart from anorexia associated with illness causing lower food intakes, the quick passing of food through the intestines results in lower absorption by the body. Lower nutritional status, in turn, increases the susceptibility to infections, resulting in adverse effects on weights at first, and then on other indicators. (e) In the case of pre-school aged children infant feeding practices (breast or bottle feeding, introduction of weaning foods) assume considerable relevance. Given the immunologic and bacteriostatic superiority of breast over bottle feeding, and the facts that in less developed countries sanitation and sterilization levels tend to be lower, and the load of infections is much higher, the advantage of breast milk is higher as compared with developed countries (Greene, 1980, Stini et al., 1980). Weaning foods are a major vehicle for transmission of diarrhoeal infections through contaminated water and food. Practices like allowing food to stand, especially in the absence of refrigeration, causes bacterial 28 proliferation. Recent work in Gambia and Bangladesh points to weaning as a major factor for transmission of fecal pathogens during infancy (Barrell and Rowland, 1979; Black et al., 1982). The focus of this research is to examine weight variations among preschoolers due largely to environmental factors. While the importance of genetic factors is recognized, the data used for analysis are all from rural Tamilnadu and so weight variations due to ethnic differences are expected to be less than those due to the environment in which a child grows. Nevertheless, in order to capture some of the variations due to genetic factors, mother’s height is used as one of the explanatory variables in the regression analysis. 2.4 Anthropometry and Choice of Weight as a Measure Some anthropometric measures based upon height and weight, used to make inferences about child growth, health and nutritional status, are described in this section. The strengths and drawbacks of these measures and the reasons for choosing weight are also discussed. Human growth has always been of interest. At a household level parents look at the heights and weights of their children to make judgments about their growth, health and nutritional status. Children outgrow their clothes and become harder to carry. At a more macro level governments of many developing countries and their public health agencies use weights and heights to assess the health and nutritional status of population groups perceived to be ’at risk’. Growth is the multiplication of body cells over time, a process which starts even before birth. Different tissues of the body grow at different 29 rates. For example skin cells are continuously produced to replace those which have died. But certain other tissues (nerve and muscle) formed during childhood growth cannot later be replaced (Tanner, J.M., 1978). Human growth is rapid in early childhood especially in preschool ages. While the process of human growth is recognized as very complex and imperfectly understood (Tanner, 1978; Floud et al., draft), the pattern of growth is now well documented. This is done with the help of anthropometry - the systematic measurement of the body. Several indicators are used, for example, height, weight, upper- mid-arm-circumference, skin fold over triceps and head circumference etc. The ’normal’ and ’non-normal’ values for these indicators are well documented on the basis of large numbers of observations in actual populations. For pre-schoolers weight and height are more commonly used not only by doctors and health practitioners, but are also more commonly understood by the general population. Growth charts, such as those in figures 2.1 to 2.3, are compiled to serve three purposes: (1) to establish the growth pattern of a particular population; (2) to allow for the identification of abnormalities in the growth of an individual which can aid in taking corrective action; (3) to serve as educational tool for parents especially, say, in rural areas of LDCs where the effectiveness of a health and nutrition program could possibly be increased if parents ’ watch’ how their child grows. This is an explicitly stated purpose in TINP areas. A typical weight growth chart from birth to five years of age shows sharp weight increases each month in the early months of a child’s life, with the curve getting flatter as the child ages. No consecutive months have the same weight and therefore, in 30 the case of any particular child, if the same weight is observed over successive months, or if there is an actual loss of weight, it is taken as a signal of growth faltering. Heights Versus Weights. Both heights and weights are used to measure human growth in childhood, each measuring a slightly different aspect. Height is a linear measurement made up of the sum of four components: legs, pelvis, spine and skull. It varies unidirectionally - it only increases or remains constant with age. It is usually measured by making the subject stand erect, without footwear, with feet parallel and heels, buttocks, shoulders and back of the head touching the vertical plane on which the measurements are marked. This automatically implies that children who cannot stand yet (e.g., those less than one year of age) or those who cannot follow instructions about the posture necessary while undertaking measurement (e.g., those under two years of age) are very difficult to measure accurately except in hospital or laboratory conditions. It is due to the considerable measurement error problems among very young children that horizontal supine length instead of vertical height is usually a preferred measure for those under two years of age. But it must be recognized that "There is a systematic difference between the two types, erect height being shorter by 13 cm than supine length." (Martorell, 1986). In measuring length an anthropometer can be used on which the child is ’stretched’ before measurement. This is considered too complicated, expensive and time consuming for rural field conditions where a large number of children have to be accurately and regularly measured. The measurement problems are known to increase when the child is uncooperative, crying or refuses to be pulled away from its parent for measurement purposes. If some children are measured in the supine position (say, those under two) and others in an erect position (say, the older ones), there is also 31 the additional problem of combining length data with height data. Height data can be very usefully employed for school age children and adults, where the problems of measurement are less serious. Weight is a measurement of body mass. Body mass automatically takes into account height to a certain extent in that the taller a person is, the greater is the body mass. Weight can, both, increase and decrease, though not with equal likelihood. A decrease in weight occurs only in abnormal periods, for example as a result of illness. But such a decrease is followed by a sharp weight increase during recovery (assuming an adequate diet) called ’catch-up’ growth by human biologists (Tanner, 1978). Weight is thus more sensitive to the environment in which a child grows in the short run, while height is a longer term indicator. A platform scale can be used for weighing older children. But for the age group of concern to the present research, a hanging bar or the Salter type scale is hung freely from a hook along with the "trouser-bag" in which the child is placed. The needle/pointer is brought to zero by adjusting the screw and the child is then placed in the "trouser bag" with minimum clothing on and the weight is read nearest 0.10 kg. Accurate measurements can be obtained even when the child is moving or uncooperative. Therefore, this method is often preferred in field conditions when frequent periodic measurements have to be taken for a large number of children aged three years or under. The question that we now examine is: how have anthropometric measurements on weight and height of preschoolers been used to observe child growth and draw conclusions about their health and nutritional status? Nutritionists, health practitioners and economists have used anthropometric data on height and weight to do this (Field et al., 1981; Martorell and Habicht, 1986; Vijayraghavan et al., mimeo, undated; 32 Chambers, 1982; Sen and Sengupta, 1983). Some of the commonly used indicators for preschool children are described below, together with their strengths and drawbacks. Heigfat-for-aee is used to measure one aspect of child growth - linear length. The extent of height deficit in relation to age, as compared with predetermined reference heights, has been used as a measure of past malnutrition. Height data have been summarized in various ways, two of which are as indicated below. (1) Waterlow’s classification: < 85% of expected height-for-age ... Severe malnutrition 80 to 90% of expected height-for-age ... Moderate malnutrition 90 to 95% of expected height-for-age ... Marginal malnutrition > 95% of expected height-for-age ... Normal (2) Mclaren’s classification: < 80% of standard ... Dwarf 80 to 93% of standard ... Short 93 to 100% of standard ... Normal. One problem with using height is the higher error percentages in measuring very young preschoolers as was pointed out earlier. This problem can be largely overcome in hospital or laboratory type conditions (where an anthropometer can be more easily used) or when the number of subjects is small so that each can be measured twice to ensure correctness (Gopaldas and Seshadri, 1987). In field conditions, however, and when the number of subjects is high, this is usually not done. The other difficulty arises from combining data on lengths with data on heights. Weight-for-age is also an indicator of child growth, measuring as it does body mass. Weight deficiency has been taken to represent current malnutrition in children. In LDCs it is also considered to be an indicator of protein-energy malnutrition (Gopaldas and Seshadri, 1987). Two ways in which weight data have been summarized to draw conclusions about nutritional status are indicated below. 33 (1) Gomez classification: < 60% of standard weight-for-age ... Grade HI malnutrition 61 to 75% of standard weight-for-age ... Grade II malnutrition 76 to 90% of standard weight-for-age ... Grade I malnutrition > 90% of expected height-for-age ... Normal (2) Indian Academy of Paediatrics classification: < = 50% of standard weight-for-age ... Grade IV malnutrition 51 to 60% of standard weight-for-age ... Grade HI malnutrition 61 to 70% of standard weight-for-age ... Grade II malnutrition 71 to 80% of standard weight-for-age ... Grade I malnutrition > 80% of expected height-for-age ... Normal. The problem with weights is that they are subject to greater fluctuations due to, say, a temporary bout of illness. Relying on a single low weight observation for a particular child, therefore, will not necessarily indicate a low nutritional status. This problem of fluctuations can be largely overcome if weight observations are available for a large number of children or several observations are available for any individual child, as is the case with sample data from TINP areas. On an average, in the sample 21 observations per child at monthly intervals are available for 1,227 children. It must be kept in mind that weights indicate short term growth. Weight-for-height is especially useful in communities where child age is not accurately known. This gives us a nearly age independent measure. It also has the advantage of combining weight and height information. For this, both, height and weight measurements of a child have to be taken at the same time. Then the weight of each subject for the corresponding height is compared with the standard weight for the same height. Weight-for-height data are used to measure short term malnutrition and conclusions about nutritional status have been drawn based on classifications such as: 34 < 75% of weight-for-height ... Severe malnutrition 75 to 84% of weight-for-height ... Moderate malnutrition 85 to 90% of weight-for-height ... Marginal malnutrition > 90% of weight-for-height ... Normal. The problem with this measure, especially in rural areas of LDCs, is that if a child is simultaneously both stunted (too short) and wasted (too light), it is possible that such a child may be reckoned as normal under the above classification - an odd and damaging conclusion for the child if low nutritional status is used to selectively provide benefits such as nutritional supplements. It appears, then, that an age dependent measure among rural, LDC children probably gives more information about their situation provided age estimation is not a problem. Height or weight for age, or both, can be usefully employed. Similar classifications are available in the literature about other anthropometric indicators like mid-upper-arm-drcumference, skin fold over triceps, etc., and a combination of one or more indicators. One last issue regarding weights versus heights needs to be discussed. Both weights and heights measure child growth as we have seen earlier. However, whether both can measure the same aspect of health and nutritional status is a more complex issue. Researchers analyzing data from developing countries consider weight deficits (wasting) and height deficits (stunting) as being caused by similar factors, though on a different time scale. Weight is a relatively short term indicator while height a long term one. In developed countries, however, more consideration needs to be given to the question since anthropometric indices based on weight and height do not always respond in similar ways, at least in modern times. Children from lower socioeconomic households in DCs tend to have lower heights than those from higher socioeconomic households, but they have also been known to have higher weight-for-height as compared with their peers from better off 35 households (Martorell and Habicht, 1986; Floud et al., draft). Obesity is not uncommon among these children. In LDCs, and particularly in South Asia, obesity is not known to occur among children from poor households. This is especially true in rural areas. The probability of coming up with a combination of low height but a high weight-for-age is likely to be negligible since "obesity is only common among well-off urban groups." (Eveleth and Tanner, 1976). Based upon experiences of field workers, the relative values of the 4 common anthropometric indicators used in growth monitoring of preschoolers are summarized in table 2.2, adapted from Swaminathan (1987). From this table it appears that in field conditions if a single anthropometric indicator for growth monitoring has to be chosen, then weight seems to have the highest overall value. A combination of indicators would naturally give a better picture. Ultimately, however, data availability largely determines what researchers have to use. To evaluate program impact in TINP areas, in addition to weights, heights were also available for a sample of TINP participants. However, due to high error percentages the height data were not ultimately used by the evaluation agency. Martorell’s comments in this connection are relevant: "Both length and height were measured with a tape, techniques which are difficult to standardize and which are error prone ... The height data may not be useful and heavy investment in resources to analyze these data would be inadvisable." Keeping in mind the above considerations, using weight alone can be considered reasonable for the following reasons: (a) the children are all three years of age or less - an age when accurate height/length measurements are difficult to ensure except in hospital or laboratory conditions, and there 36 is the added complication of combining height and length data; (b) the children come from rural areas of an LDC where obesity is not prevalent, so height and weight deficits are caused largely by similar factors; and (c) potential mothers are identified during pregnancy itself, so accurate records of birth dates are available because of which age estimation is not a problem. 2.5 Anthropometry and Welfare In this section the purpose of analyzing weight data is explained - to draw inferences about welfare based on the third approach discussed in section 2.1. We have seen that height and weight data on preschoolers allow us to measure child growth. We have also seen that they have been used to draw inferences about health and nutritional status. In as much as these are considered desirable for the well being of preschoolers, such data allow us to make inferences about their well being. Under the fixed genetic potential model, in fact, anthropometric measurements have been identified with a specific degree of malnutrition. If, however, considerations of individual adaptability are allowed, then there are some difficulties in exactly identifying height or weight data with the individual’s health and nutritional condition. Wherever the dividing line is drawn on the basis of anthropometric data, it is always possible that some not at risk will be identified as malnourished, while others who are at risk will not. This does not, however, mean that such data are of little value in attempts to measure individual functional capabilities. This is for the following reasons: First, it is possible to use anthropometric data as the basis for identifying problems with growth, health and nutrition, without malnutrition being identified exactly 37 with the percentage of individuals below some fixed percentage of reference weight (say, 60% or 75%). Child weight can be used as an early signal of possible growth faltering. Data can be interpreted in a probabilistic sense. Consider, for example, a group of children who weigh less than 60% of some reference weight for each age group and another group who weigh less than 90% at a particular time period. The probability of functional failure induding mortality would be higher in the first group than in the second. Such a conclusion would be strengthened if we had data for several points in time and found that the former group continued to manifest lower weights-for-age. Thus, in the study by Chen et al., (1980) in Bangladesh the overall death rate was much greater in the group of children with weights-for-age less than 60% of reference weight as compared with the group with less than 75% of reference weight (table 2.3). Similarly, with lower rates of height and weight growth there is an increased risk of infection and mortality (Reddy, et al., 1976, Rowland et al., 1977, Kielman et al., 1976 with Indian data). Second, recognizing that the human body adapts to stress, first by lowering weight and later by inhibiting height growth, and that lower weight or height does not automatically imply undernutrition, the fact that certain groups of children are in situations where they continually have to adapt more than others is itself of significance. Those having to adapt more frequently can be considered as facing greater and more sustained deprivations than others. Third, it would not be invalid to use these statistics as a basis to assess the adequacy of an individual child’ s ability to have access to food for consumption even though no reference is made to the extent of malnutrition. In other words, such data can be used to draw conclusions about an individual’s food entitlement. 38 Fourth, anthropometric data can be used for policy purposes to screen populations for identifying the worst cases in order to plan for interventions. Children below 60% of the norm cannot be equated with those below 75% of the norm since the probabilities of failure of functioning capabilities would be different in the two groups. If resources are limited, a more discriminating view of need may enable planners to concentrate on the worst cases rather than concluding that the problem is too large (eg., more than half the children are malnourished) for the resources available. Depending on resource availability higher percentages can be brought under policy interventions and, in fact, the norm itself can be revised upwards over time. The percentages of child population which come under the less than ’ normal’ categories have often been high in rural areas of countries like India. While it has been argued that identifying over 60% of the pre-school children as malnourished, and sometimes even over 90%, may be overstating the problem by insufficiently discriminating between the probabilities of functional failure among the different grades, one must also guard against going to the other extreme. If only the ’severely malnourished’ cases are considered, the percentages tend to be much lower - around 10%. However, seasonal and year to year fluctuations can alter the meaning of such an average percentage in the severe category in two ways: (a) Individuals are not at equal risk in all seasons. "Proportions with inadequate weight for age, with scanty breast milk and presenting at hospital with kwashiorkor or marasmus, all vary seasonally, tending to rise in the later wet season in most cases." (Lipton, 1983; Chambers, 1982). Similarly, greater proportions of such cases are observed during periods of unexpected stress induced by crop failure, disease, etc. This 39 means that averages will substantially understate the proportion and intensity of periodic severe functional failures. (b) It is not the same individuals who, at different times, suffer severe undemutrition. While at any one time approximately 10% or even less may be at severe nutritional risk, a much larger percentage moves in and out of this ’at risk’ category. Micro studies seem to indicate that problems of health, malnutrition and growth are more spread out than indicated by the percentages in the severe categories alone. If most children are found to be in the same condition over several rounds of observations, the problem is much more severe for them, but the ’target’ requiring intervention would be smaller. On the other hand, because of the fluctuating category, the problem is wider. Anthropometric data can, therefore, be used as a basis for inferences about well being according to the capacity to function approach. Looking at evidence on physical states of children is considered a relatively direct way to draw inferences about an important aspect of their individual welfare. It is proposed to analyze weight data from a sample of 1,227 children, 6 -3 6 months of age from rural Tamilnadu, together with information on a number of individual, household and community level variables expected to influence weights. 40 Figure 2.1 GROWTH CHART OF AN ACTUAL CENTRAL AMERICAN CHILD The chan below plots the growth o f an actual child in a poor Central American community and tells a story typical o f the childhood o f millions in the developing world For the first six months o f life, breastfeeding keeps the child growing normally. Thereafter, as weaning begins, malnutrition increases the risk o f infection and infection exacerbates the malnutrition. Together, they attack the child's rate o f growth so vigorously that there is no weight gain at all between the age o fsix months and eighteen months Often, the flattening o f the child's weight curve is a slow and invisible process — especially i f it is also happening to a majority ofthe children in the community. But regular m onthly weighing and the use ofa simple growth chart— kept at home by the mother— is an early warning system which makes malnutrition visible and can be a vita! aid to the mother in maintaining her child's growth and health. The chart shown here is based on studies by L~J. Mata, J.J. Urrutia, and A. Lechtigfor the Institute o f Nutrition o f Central America and Panama (INCAP) I i Source: J.P. Grant, UNICEF, "The State of the World's Children 1982-83." 15- 14- 1 3 - 12 - 11 — «o 10- £- 2 9- (9 O - — I I X (9 7- 6 - 5 — KEY 0 Diarrhoea BC Bronchitis BN Bronchopneumonia CEL Cellulitis CONJ Conjunctivitis FUO Fever of unknown origin 1 Impetigo M Measles S Stomatitis T Oral Thrush URI Upper Respiratory Infection CONJ Norm D D URI O D D URI M BCD ijRI D FUO BC UR D D D D BC BC URI u r , _D URI BN D O J L. 1 * i « I I I 12 15 18 21 AGE IN M ONTHS 24 27 30 33 36 i 41 Figure 2.2 JA S M IN E 'S PERCENTILE CHART OF HER WEIGHT: A CASE OF ENQUIRY INTO CHILD ABUSE Narne D ate of Sirth rit& fiffi I W eight HOME I I 0*4 T R IA L CARE ( H ^ f i l l l l i n i w i i u n M i l l HOME (14. IX. 19811 i jape RIBTH JT * > » . 1979: (Reproduced from Floud et al., draft.) Weight i n KGs. Figure 2.3 GROWTH CHART OF AN ACTUAL T IN P C H ILD , TAMILNADU 42 III “ = = = = IV i 25 2 8 27" 2 8 29 3 0 31 32 39 34 35 36 13 15 15 17 18 19 20 21 22 23 Y e a r 2 January 1987 - April 1989 11 12 1 2 - * * o r ■ ' 'm-mrn0m I N • to m 40 to S O 43 Figure 2.4 VARIATIONS IN HEIGHT AT AGE 7 BY SOCIOECONOMIC AND ETHNIC GROUPS 123 121 119 117 . E u 115 113 111 109 r s l NCHS PERCENTILES — 50th 25th 10th — 5th Mean heights of 7-year-old boys of high (e) and low (o) socioeconomic status. (Adapted from Martorell and Habicht, 1988) T a b le 2 .1 ENERGY INTAKE OF NEW GUINEA ADULTS Village Body Weight Energy Intake Per Day (Kgs.) Kaul Males 56 1940 (Coastal) Females 47 1420 Lafa Males 57 2520 (Highland) Females 51 2100 Source: Norgan, N.G., et al, 1974. T a b le 2 .2 RELATIVE VALUES OF PHYSICAL MEASUREMENTS UNDER FIELD CONDITIONS Skill Time Organization Cost Errors Ove: Vali Weight Less Less Less Low Less I Height More More More Low More III Mid-upper- arm-circumference Less Less Less Low Less II Skin fat fold High High High Low High IV Source: Adapted from Swaminathan, 1987. T ab le 2 .3 NUMBERS OF CHILDREN OUT OF A POPULATION OF 2019 WHO WOULD HAVE BEEN IDENTIFIED AND TREATED FOR MALNUTRITION, USING WEIGHT-FOR-AGE Weight- No. identified No. of prevent- No. of false Deaths prevented for-age as in need of able deaths positives per treatment treatment treated (1) (2) (3) (4) (5) < 60% of reference 427 48 379 0.112 < 75% of reference 1473 92 1381 0.062 Note: (1) Col. (3) consists of the number who actually died. (2) Col. (4) consists of the number who survived. Source: Chen et al, 1982a. Also in Pacey and Payne eds., 1985. ft C hapter HI 47 THE TINP PROGRAM In this chapter section 3.1 contains a description of the Tamilnadu Integrated Nutrition Project, TINP, which was started with the objective of improving the health and nutritional status of preschool children. Since the program is a potential determinant of weight, section 3.2 contains a discussion of demand factors that might inhibit the program’s impact. In the concluding section 3.3 the implications of supply and demand factors for what can be expected in the TINP data are examined. 3.1 Program Description The TINP was started in October 1980 on a pilot basis, with assistance from the World Bank, additional areas coming under the program in a phased m anner. The program currently covers 6 districts in Tamilnadu1 , districts identified by the Tamilnadu government as having relatively serious nutrition problems on the basis of surveys carried out earlier. The program’s primary objective is to improve the health and nutritional status of children under three - a section of the population very difficult to reach2. This group accounts for an estimated 90% of preschool mortality in the state and 1 Three of the program districts have since been subdivided because of which TINP is in operation in 10 out of 20 districts in the state. a Children three years and under are unlike school going children who assemble at a fixed place and for a fixed period of time every day, making them a relatively easy population group to target for. 48 malnourishment was found to be a leading or associated cause for 75% of those deaths (Chidambaram, 1989). The focus on this age group has been justified on three counts. First, during this period of rapid human growth, recognizing the vulnerability of the child, and the vulnerability of the household as a whole when children are relatively young, assistance is justified on welfare and equity considerations. Second, on efficiency grounds, intervening at a critical period of human development is viewed as enhancing the stock of human capital. This is expected to increase the future flow of goods and services. Third, health and nutritional status are considered as desirable ends in themselves.The program’s inputs are also extended to another high risk group - pregnant and nursing women - in order to monitor child growth from its prenatal stage. The program tries to integrate its 4 major components: (i) nutrition delivery services, (ii) health delivery services, (iii) communications and education, and (iv) monitoring and evaluation. The primary functions and inputs supplied by the 4 components are described below. (i) Nutrition Delivery Services. The growth of each child is monitored on the basis of weights taken each month, to identify early, cases of growth faltering and severe weight deficits. Since weights are the first to respond to health and nutritional deprivations (with height and other anthropometric measures following if the stress continues), specific criteria based on these monthly child weights have been set up to screen the children and select a subset of them for nutrition supplementation. Weights of each child are also plotted on a weight chart each month. Figure 2.3 contains a weight chart used in TINP areas. Children are divided into 5 grades, each grade being separated by a curve. The upper uppermost curve is used to separate the ’normal’ from the ’ mildly malnourished’ (or grade I cases) and represents minimum weights that a child should have for each corresponding age on the horizontal axis. Similarly, the other curves separate the 49 other grades of malnutrition. Grades I, II, III and IV represent increasing degrees of malnutrition. Monthly weights of each child are compared with the ’norms’. If the weight for a particular beneficiary falls in grades I or II, the child is considered to be mild to moderately malnourished. Beneficiaries falling in grades III and IV are considered severely, and very severely malnourished respectively. Since this a period of rapid human growth, weight over successive months should only increase; children failing to gain adequate weight, or those actually losing weight are selected for food supplementation3. A child with a single low weight observation is not selected since it could be the result of a diarrhoeal episode or other temporary illness. Two to four weights taken at monthly intervals are observed before food supplementation is started. However, all children with severe weight deficits (grades III and IV) are selected for supplementary feeding. The supplementation is continued for a minimum period of 90 days or as long as required for adequate growth recovery4. Thus not all children weighed are fed. These negative selection criteria serve two purposes. One, there is an economy of resources, additional food inputs being sharply focussed only on the worst off cases (less than 50% of all children weighed are also fed). Since the food supplement alone constitutes approximately half of the program’s operating costs, this cost consciousness leads to substantial savings. Selective feeding (as against universal feeding) is 3 The criteria for selection are: (a) for those between 6 -1 2 months, if weight gain is less than 300 gm over 2 successive months, the child is put under special observation. If weight gain is less than 300 gm. in the next month also, the child is enrolled for supplementation; (b) for those between 12-36 months, if weight gain is less than 300 gm. over 4 months, the child is enrolled for supplementation (Office of the Project Co-ordinator, TINP, Madras). 4 A weight gain of not less than 500 gm. over 90 days must be observed before the child is ’ graduated’ from supplementary feeding. 50 expected to contribute to the reduction in the number of ’ false positives’ treated (i.e., those not ’at risk’ ), and in the number of children ’at risk’ but not treated, since the program actively looks for cases of growth faltering (inadequate weight gain) and severe weight deficits (children in grades III and IV). See, for example, table 2.3 where only the extreme risk of death is considered for a sample of children in Bangladesh (Chen et al, 1982a). Ideally, it would be efficient to be able to eliminate completely both types of errors, given resource constraints. But since the cutoff points cannot be considered as uniquely dividing children into the ’normal’ and ’ non-normal’ categories, the TINP approach can only be considered as an attempt to reduce the incidence of such errors. Child weights are used to screen the population, compare risks of future functional failure, and allocate resources accordingly. Household food entitlements are allowed to prevail in the case of children demonstrating reasonable growth (based on the program’s criteria) on home food and the health care provided by the program. Two, observing each child’ s growth chart, followed by selective feeding of some children, is expected to serve as an educational tool. When only some children are fed, mothers question the program staff, see their child’s growth curves, and become aware of the idea of looking at weight deficits, motivating at least a proportion of them to follow the program’s messages. This awareness of the flattening of a child’s growth curve as compared with ’norms’ is an early warning system which makes poor growth visible to the mother - a process otherwise invisible, especially if it is also happening to a majority of the children in the community. Two doctors who visited some TINP villages in Madurai district to study growth monitoring have remarked,"... the interpretation of growth lines by mothers is accurate and impressive. They can identify growth faltering, malnutrition and relapse with considerable skill and accuracy.” (Bhan and Ghosh, 1986). 51 (ii) Health Delivery Services. Apart from growth monitoring and selective supplementary feeding, a number of health related inputs are provided to aU children, recognizing the synergistic relationship between nutrition and infectious diseases. There is no selection here. The broad aim is to improve the pre-project maternal and child health care system. The choice of specific health inputs was made on the basis of some empirical research in Tamilnadu. To quote from an evaluation study (Chidambaram, 1989), "Important findings of earlier researches were: average birth weight of children in Tamilnadu was low; rural and urban children had comparable weight at one year but as age advanced urban children were perceptibly heavier; in rural Tamilnadu nutritional morbidity accounted for about 38 percent of cases treated; the percentage age distribution by diseases (for rural Madurai district) showed a very high incidence in children under 5 years for marasmus (88%), kwashiorkor (86%) and vitamin ’ A’ deficiency was widespread frequently resulting in blindness and the intake of iron by preschool children and pregnant women was below ICMR’s5 recommended daily allowances." In addition to diarrhoea management, deworming of children to eliminate intestinal parasites, prophylaxis against vitamin A deficiency, iron and folic acid supplementation, a network of health sub centers (HSC) was established in stages (1 HSC serving approximately 5000 population, i.e., 4 to 5 village CNCs). Pregnant and lactating women are also covered through antenatal and postnatal checkups, immunizations, iron and folic acid supplements, etc. Treatment for minor ailments and referral services to doctors of primary health centers are also provided. (iii) The Communication Component. For longer term impact, spreading education in the community about growth, health and nutritional status of preschoolers is considered important by the project management. The population is divided into 2 groups for purposes of communication: (a) the core group consisting of the immediate family of 5 Indian Council of Medical Research. 52 the child - mothers, mothers-in-law, fathers, siblings; (b) the peripheral group, which is the rest of the community. Different communication approaches are used for the two groups. Direct interpersonal contacts by the program staff with the core group (home visits, encouraging the family to look at growth charts, diarrhoea management, teaching the importance of colostrum, etc.), and using mass media directed at the entire community to disseminate information about the program’s activities and purposes, and specific child care messages, in order to promote program acceptance. Behavior changes within the household are sought to be encouraged resulting in enhanced food entitlements for children - introducing solid foods early and not waiting until the child is 10 to 12 months old, not withholding foods and liquids during diarrhoea or other illness, oral rehydration, the importance of greens and other vegetables, increase in the frequency of feeding, etc. Weight charts maintained for each child are found useful not only as a screening device to select children for supplementation, but also as an educational tool. The communication component also helped in overcoming the early hesitation in acceptance of the program. In view of the low literacy levels in the rural communities, this component is considered especially important by the project management. (iv) Monitoring and Evaluation. A continuous monitoring of the program’s inputs and outcomes is carried out with data flowing from the village level operational units, the CNCs, through the different tiers of project control, to the project management. This enables timely interventions to rectify local problems like exhaustion of supplies in a specific area, etc. Evaluation of the project’s activities and impact are entrusted to an outside department of the government of Tamilnadu to ensure objectivity. Evaluation 53 results have pointed out some of the weak areas of the program and also its strengths, providing useful information to the management. 3.2 Demand Factors Inhibiting Program Impact The program is expected to be a potential determinant of child weight. However, the effectiveness of the various program inputs in terms of impact upon the intended beneficiaries, does not automatically follow. The existence of a program, assuming that necessary services are available without interruption, takes care of supply factors. The nutrition program that is the concern of this study is different from the environmental or mandatory health related programs which by their nature do not require households to choose how to respond (eg. malaria eradication or small pox vaccination). Environmental programs do not deal directly with individuals or households. Mandatory vaccinations tend to leave little room for household choice. Such programs have been documented as having had significant impacts and have been relatively successful in meeting their objectives. On the other hand, effectiveness of health and nutrition interventions that allow room for household choice tends to be a relatively complicated issue because of the need to allow for the response of the household to the program. The household is regarded as the basic decision making unit resulting in specific entitlements for its individual members. The establishment of an intervention program is intended to alter the within household allocation in favor of its more vulnerable members. As a result, first, "... an effective demand for, or the utilization of program services cannot be guaranteed; 54 second, even when adequate demand exists, it may stem from private objectives that are not congruent with program or social objectives (Chernichovsky, 1979)." Families could adjust their intra-family allocations after the program selects and benefits some of their members. It is possible for there to exist a kind of inertia or resistance to change pre-existing behavior patterns resulting in an inherent bias against observable impacts of such programs. This can be due to differences in preferences, economic factors, sociocultural factors, or some combination thereof. Factors that affect the demand for the program's inputs have to be examined. Four broad categories are identified, and the extent to which the program can cope with each of them is briefly discussed. (1) The extent of divergence between overall household well being that the household tries to achieve, and the well being of children that the program focuses upon is a relevant factor. The greater the divergence between the two, the more likely it is that a household will try to convert nutrition supplements to other uses, resulting in a reallocation of program resources in the process of the household’s attempts to achieve its objectives. If take home food is allowed, sharing of food among other members at home cannot be ruled out, and it becomes harder to ensure that the intended beneficiary receives the nutrition supplement. TINP partly overcomes this problem by insisting upon on-site feeding. However, reallocation at home still cannot be ruled out and nutrition supplements can be converted into substitutes. For example, there may be substitution at home away from the children who benefit from the program and in favor of older siblings and adults who do not. The intended nutrition supplement then gets converted into a substitute for a home meal. The problem of altering intra-family allocations becomes more complex and merely arranging for the supply of nutrition supplements may not be 55 sufficient. This is where the education factor conies in - reliance is placed upon education of the mother and other members of the child’s immediate family through the program’s communication component. (2) The extent to which the program draws upon household resources is an important consideration. Even if there is no direct monetary price for the program inputs, it does make demands on the parents’ time. Given that the child beneficiaries under consideration are under five years, they cannot come to nutrition centers unless one of the parents actually physically brings them, interrupting their agricultural or other activities. The opportunity cost of time (due to work interruption resulting in foregone income which would benefit the entire household) may be considered greater than the benefit to a particular member, the infant who, in any case, is less capable of making demands. In addition, feeding and hygiene practices suggested under such programs like the early introduction of solid foods, administering more frequent, if smaller, feeds during illness, boiling water before drinking, etc., may be resisted because the women simply have not got the time for them. Utilization of program inputs is likely to be inversely related to the alternate demands on the parents’ time. By locating the CNC in each village, largely within walking distance, the problem is partly overcome. This becomes evident if hamlets located away from the CNC are studied. Children living further off are much less likely to be brought in for supplementary feeding, weighing, health care, etc. (3) The influence of pre-existing institutions, including customs and practices which tend to result in an age bias against children, and between different children within the household, may be significant. Preferences of households may be based upon lower perceived needs of children resulting in relatively poorer diets. Anthropologists have explained this as a mechanism of cultural adaptation in a stressful environment 56 (Marchione, 1981). It is observed that children are less important to the survival of the household unit than working adults. This has also been explained in terms of ’ benign neglect’ of children (Cassidy, 1980), where, paradoxically, while parents otherwise demonstrate signs of affection and concern for children, they simultaneously practice customs which produce deleterious results. Food restrictions on items considered too ’strong*, ’indigestible’, or ’ worm producing5 for children have been known to result in withdrawal of many nutrient dense foods, including peas, meat,and eggs (Cassidy, 1980; Schofield, 1979). The reduction of food intake and liquids during diarrhea has effects contrary to what is intended. Socialization is accomplished through requiring females to eat later or eat ’inferior’ foods in order to enhance their social normalcy. If customs are perceived as promoting the well being of children, they are likely to be resistant to change. Since nutrition programs tend to select beneficiaries largely on the basis of degree of health and nutritional deprivations, it seems likely that they would not be biased against later bora children, or against females. In fact, to the extent that female children or the later bom tend to be more malnourished, there may be a greater percentage of these categories among the beneficiaries. But here again the question of parental motivation comes in, which may work in the opposite direction. The program’s communication component tries to deal with existing customs and practices. (4) Public perception of the association of this program with other, less popular, interventions within the household like family planning programs could inhibit acceptability. If the organization structure overlaps at the village level with the same worker promoting sterilization as the preferred means of birth control and also child health, the worker may less acceptable to the village population. In the case of TINP there is no overlap at the village level. The CNW’s role is does not include the direct promotion 57 of birth control. Child spacing is recommended in the interest of the health of the mother and child, but not as an end in itself. Further, no ’sterilization targets’ are assigned to the CNW. In the administrative hierarchy the CNW reports to the block and district level nutrition officers under the control of the Social Welfare Directorate at the state level, and not the directorate in charge of family planning. (5) The relative attractiveness of perceived substitutes can be significant. For example, traditional health and feeding practices may compete with program inputs. The suggestions of influential members of the community or households, nearness or accessibility of other health facilities, would be some relevant considerations here. To the extent there are less perceived substitutes, adoption of program inputs is likely to be greater. 33 Concluding Remarks When examining and interpreting TINP data, and especially when comparisons are made over time, it proves useful to keep in mind the simultaneous operation of supply and demand factors. Over time, as the program gets better established, is better accepted by the community, training of staff is completed, and health sub-centers are in place, an improvement in supply can be expected. This, by itself, may or may not result in significant effects. Over, perhaps, a longer time period, with increased knowledge about the program, factors which inhibit demand can be expected to have lesser impact. More significant program effects can then be expected. The time period for this to occur may vary in different situations, and is an empirical issue. In fact, a priori, there is no 58 reason to even suppose that supply must always precede demand, as is the case in TINP areas. These considerations lead us to expect that as compared with children enrolled in the early program period (base line data), weight data on children enrolled in later years may first show little improvement, followed by a period of more significant program effects. On the other hand, as the area of operation gets wider and more diverse, supply factors tend to get harder to control with the same uniformity as before, which could work in the opposite direction. Chapter IV 59 SURVEY DESIGN, SAMPLE SELECTION AND DATA In this chapter the considerations that went into and the procedures followed in the finalization of the survey design and sample selection based upon field research by the author are discussed in sections 4.1 and 4.2. A brief description of the sample data is contained in section 4.3, and of the data sources in section 4.4. 4.1 The Survey Design Since the study involved field research and primary data collection which had to be completed within a fixed time period, it was considered essential to lay out as completely as possible a prearranged program for collecting information from the Tamilnadu Integrated Nutrition Project (TINP). It became clear very soon that there could be no ’ ideal’ survey design even when the goals of research are fairly precisely defined. The specification of research goals tempered with an evaluation of the resources available (time, personnel and funds) were two important considerations in determining the survey design and sample size. The survey design had to be such as to yield data capable of addressing as many of the following issues as possible: (i) To identify levels and trends in weights among TINP preschool children, and possible program effects. (ii) To measure the extent of deviations from existing weight norms used by the program authorities prevailing among preschool children, i.e., the distribution of children in the normal category, and those belonging to grades I, n, in and IV. 60 (iii) To examine weight variations between certain groups - children from better off versus worse off households, children of more educated versus less educated mothers, males and females, scheduled caste children and others, and regional differences. (iv) To study the effects of birth order, age and certain community variables like the hygienic condition of water, distance to health and hospital facilities, etc. In addition to the objectives of the survey, other considerations that went into the final selection of the survey design were as follows: Resources. All work in connection with the survey, right from designing and pretesting questionnaires to finally collecting and sorting completed schedules, had to be finished in a maximum time period of 6 months, between February and July 1989, and on a fixed budget. Experience. In the absence of any previous survey experience, the researcher wanted to guard against taking on more than she could handle. Rapport with project staff and management. Having worked as Project Co-ordinator earlier and being a member of the Indian Administrative Service, this already existed in the present situation. Thus, field level cooperation was assured. Data availability. The project maintained individual records at the village level even for earlier time periods. Hence it was possible to do more than a single cross- section survey and obtain weight records for more than one time period. But in order to do so, it was necessary to go to each village level center for data collection. Given these considerations, two alternative designs were examined. Survey Design 1: Draw the sample from a wider area of TINP’s operation, covering its very early phases and its newest phases, but concentrate only on children currently enrolled. Since the project was established in different years in five regional phases, this procedure would generate a sample consisting of: 61 * Currently enrolled children from Phase I where the project has been in operation from October 1980. * Currently enrolled children from Phase II where the project has been in operation from March 1982. * Currently enrolled children from Phase in where the project was introduced between February 1983 to May 1983. * Currently enrolled children from the later Phases IV and V where project operation started in September 1984 and January 1985, respectively. While all sample children would be selected from those enrolled at the time of the survey, the time dimension would be indirectly introduced in as much as the children from different project phases would represent areas where the project has been in operation for a different number of years. A major advantage would be the ease of picking up data from current records and contacting fa m ilies that are currently participating. The drawbacks of this approach are (i) the difficulty in separating variability due to region and that due to the duration of project operation since phases differ with respect to both, time and region; and (ii) the wider the area of coverage the more thinly the sample would be spread, making the task of monitoring and quality control that much more difficult. Of course, this second problem could be overcome by selecting only the first two or three phases for sample selection. Survey Design 2: Consider only the three earliest phases of project implementation. This would allow for the maximum possible time for program impact and also permit the sample to capture some variability due to duration of program implementation. Then select the sample as follows: * Draw a sub sample, approximately 1/3 of the sample, from currently enrolled children in each of the phases I, II and III. * Draw a second sub sample from older siblings of those currently enrolled and who had participated in the program at an earlier date. This would generate another sub sample from the same area, but relating to an earlier time period. 62 * Draw the final part of the sample from the ’base line’, i.e., the time when the program was just put in place in each of the three phases, but from the same areas. This would provide the earliest data on weights and would have to serve as a proxy for the pre-program situation since comparable monthly weight records are not maintained in areas where the program did not exist. Thus, from each of the selected project areas we would obtain three subsamples for three different time periods, including data on siblings. This would be better suited to make comparisons over time, to separately examine regional differences, and to capture variability due to duration of implementation, while allowing some control for genetic and household factors. In other words, this sample design involves the drawing of successive samples from the same population (area). It does not, however, allow for the strict experimental design type control group, to primarily study effects of an intervention program, possible under much more controlled situations of the following kind: Group Time 1 Program Time 2 Treatment Sample A Yes Sample A Control Sample B No Sample B This was impossible in the field situation with human subjects because of two reasons: (i) The same sample (A and B) would have aged between the two time periods and would not have in time 2 weight profiles comparable with those of time 1. So, at the very least, two different samples would have to be drawn for times 1 and 2 in each of the treatment and control groups as follows: Group Time 1 Program Time 2 Treatment Sample A1 Yes Sample A2 Control Sample B1 No Sample B2 63 (ii) For ethical and political reasons weights cannot ordinarily be taken systematically over time when no program benefits are going to accrue to the children. Hence it would not be possible to generate samples B1 and B2. Given these considerations, the final survey design chosen is tabulated in table 4.1. Table 4.1 THE SURVEY DESIGN Time 1 (current) Time 2 (older sibs) Time 3 (base lin e ) Phase I (e a rly program) Sample A1 Sample A2 Sample A3 Phase I I (middle program) Sample B 1 Sample B2 Sample B3 Phase I I I (la te program) Sample C 1 Sample C2 Sample C3 4.2 Sample Selection The specifications of research goals, together with an evaluation of resources available (time, personnel and funds) were the two considerations that went into determining not only the survey design, but also the sample size. Ultimately there was an inevitable arbitrariness in the choice of the exact sample size. It was decided to select a total sample of approximately 1000 preschool children. The actual sample turned out to be a little higher - 1227. A stratified random sample was drawn in multiple stages in order to fulfill three types of conditions as completely as possible: (i) It had to be representative of the situation prevalent in the rural areas of Tamilnadu and also provide sufficient variation in circumstances (eg., economic, regional, with respect to time) in order to make it 64 possible to relate child weights to many of the exogenous factors with which they might be expected to vary, using econometric analysis, (ii) It had to be selected cost efficiently, yielding maximum possible information for given resources, (iii) It could not be allowed to be so scattered as to make supervision, quality control and uniformity in interpretation by the interviewers difficult to ensure. Different sample selection procedures were adopted at the various stages to capture as much of the population variability as possible, while economizing on resource use. Stage 1: Selection of Districts. The project had been set up in a phased manner covering the rural areas of six districts in the state of Tamilnadu. Phase I was the earliest, having started in October 1980 and Phase V was the latest with a start date of January 1985. In order to allow for maximum possible time for project impact it was decided to draw the sample from the earliest three phases. This meant that the districts selected would have to be out of Madurai ... Phases I and II Ramanad ... Phase III Pudukottai ... Phase HI. Pudukottai was dropped since the project management was in the middle of making several changes in program implementation and reorganizing the administrative hierarchy there. Hence, the districts chosen were two of the earliest project districts, covering the three earliest phases of program implementation as indicated below (source: the Project Coordination Office, Madras): 65 D fs tric t Selected Phase S tart Date Duration of Program(*) Years Months Madura i I Oct 1980 8 6 I I Mar 1982 7 1 Ramnad I I I Feb 1983 6 2 <*) Upto A pril 1989. Of the two districts selected, Madurai can be considered economically more developed as compared with Ramnad. The normal rainfall in Madurai (important for agricultural operations) is higher than the state average, while that for Ramnad is a little below the state average. The percentage of area sown more than once to total cropped area, showing the extent of double cropping, is significantly higher in Madurai as compared to Ramnad, indicating higher overall productivity of land and higher levels of economic activity, including employment generation. Figures for the normal annual rainfall (in mm.) and the actual percentage of area sown more than once to total cropped area in the two districts for 1983-84 and 1984-85, are shown in tables 4.2 and 43. Table 4 .2 % AREA S O W N M O R E THAN O N C E TO TOTAL CR O PPED AREA FO R SELECTED DISTRICTS, 1983 - 84 & 1984 - 85 D is tr ic t 1983 - 84 1984 - 85 Madurai 10.85 8.38 Ramnad 1.63 2.17 Source: The periodic Season and Crop Reports published by Department of S ta tis tic s , Govt of Tamilnadu. 66 Table 4 .3 N O R M A L ANNUAL RAINFALL (MM) D is tric t R ain fall Madurai 163.9 Ramnad 132.9 State average 138.3 Source: The periodic Season and Crop Reports, Department of S ta tis tic s , Govt of Tamilnadu. Thus, two of the earliest project districts were selected, one economically relatively better than the other. Stage 2: Selection of Blocks. Blocks are administrative divisions under a district made on the basis of a population criterion and geographical contiguity for development purposes. In the selected districts the number of blocks, average total population per block and average population 6-36 months old are summarized in table 4.4. Table 4 .4 M EA N TOTAL & 6-36 M O NTHS O LD POPULATION PER B LO C K FO R SELECTED DISTRICTS D is tr ic t No. of Mean Total Mean 6-36 % 6-36 Month Blocks Pop/Block Month Pop/Block Pop/Block Madurai 33 80,560 4,074 5.06 Ramnad 34 66,600 3,226 4.84 Source: Calculated on the basis of figures as on A p ril 1988, O ffic e of the Project Co-ordinator, TINP, Madras. Two blocks were selected from each district. In order to avoid selecting both ’good’ or both ’bad’ blocks, it was decided to select blocks purposively on the basis of 67 some index of overall economic development of the area. Keeping in mind that the concern was with an exclusively rural population, the ratio1 Gross area irrigated Gross cropped area or GAI/GCA was considered a fair overall index. The extent of area irrigated is crucial for at least 3 reasons: (i) Assured irrigation allows farmers to choose relatively high value crops which bring in greater incomes, (ii) It also results in higher productivity, other things being the same, (iii) Irrigation allows double cropping of the same piece of land, (iv) Because of (i) to (iii) above, not only do landed households benefit, but so do others due to better employment opportunities in agricultural and related occupations. The ratio in net terms would ignore double cropping because of which the gross ratio was preferred. This ratio was examined for each of the 33 blocks of Madurai and 34 blocks of Ramnad. After removing extreme cases the following 4 blocks were chosen, one each with a low and high GAI/GCA ratio from each of Madurai and Ramnad districts: D is tric t Phase Blocks (GAI/GCA)* High or Low Selected Madurai I Kottampatti 39.30 Low I I Chellampatti 68.05 High Ramnad I I I K ariap atti 20.00 Low I I I Sivagangai 79.20 High Source ( * ) : Season and Crop Reports, Department of S ta tis tic s , Government of Tamilnadu. 1 This ratio measures total area irrigated as a proportion of total area cropped, including double cropping. This is better than the corresponding net ratio which ignores double cropping, since it takes into account only the actual acreage on ground. 68 Kottampatti block had to be selected anyway on another ground - it was the only block in Phase I of the project, when the program had just been introduced in Tamilnadu. Population profiles of the selected blocks are given in table 4.5, including percentage of population in the 6-36 months age group. Table 4 .5 POPULATION PROFILES OF THE SELECTED BLO CKS APRIL 1988 Block Total ( Population 6 - 3 6 Months Population % Population 6 -3 6 Months Kottampatti 95,251 5,087 5.34 CheIlampatti 81,210 3,881 4.78 K ariapatti 51,425 2,925 5.69 Sivagangai 68,071 3,596 5.28 Source: O ffice of the Project Co-ordinator, TINP, Madras. Stage 3: Selection of Villages / Community Nutrition Centers. Community Nutrition Centers (CNCs) operate at the village level2. The stage three sampling unit was the CNC which automatically implied village selection as well. It was not possible to arrive at the number of CNCs in the sample without reference to the ultimate sample size of approximately 1000 children. The following questions had to be answered in the process of CNC selection: 2 A strict one-to-one correspondence between villages and CNCs did not exist when large villages had more than one CNC or, occasionally, a village and its hamlet together had only one CNC. 69 * How many children are enrolled at a particular time in a typical CNC in the selected blocks? * What proportion of them should be selected? * Should a given sample size be drawn from a larger number of CNCs (implying a smaller number from each CNC) or a smaller number of CNCs (implying a larger number from each). This choice would depend upon whether intra or inter-CNC variability was greater. Taking into account all of the above points it was decided to gather and code data from 1% of the 229 CNCs in operation in the sample blocks. This resulted in covering 16 CNCs and 16 villages. No figures about the relative magnitudes of inter as against intra CNC variability were available. However, discussions with the project staff seemed to indicate that intra CNC variability was considerable. Sampling adequate numbers from each CNC would capture help capture intra-CNC variability. Covering 16 villages randomly selected was expected to capture inter-CNC variability as well. In each block CNCs were independently selected using circular systematic sampling with a random start. The selection resulted in the following numbers in the sample: Sample Block ( 1 ) Total C N C No. ( 2 ) 7% of Column 2 (3) No. of CNCs in Sample (A) Kottampatti 6 8 A. 8 5 Chellampatti 6 6 A. 6 A K ariap atti A T 2.9 3 Sivagangai 5A 3.8 2 Total 229 16.0 16 70 Stage 4: Selection of Children. This was the final stage. Two lists were drawn up for each CNC on the basis of current and old records available in the centers. These were: List 1 consisting of children currently enrolled, together with the number of months in the program, and names of older siblings, if any had been enrolled in the program earlier; List 2 consisting of children enrolled at the time the program was just introduced in the village. Out of list 1 two categories of children were selected: (a) All children currently enrolled provided they had been enrolled for at least 12 months. Those just joining were eliminated since they would yield very few monthly weight records. This generated 423 currently enrolled children3. (b) All cases of older siblings of the currently enrolled child selected in (a) above. This generated 373 older siblings. Out of list 2 base line children were selected using simple random sampling, independently from each village CNC. The sampling fraction was either 1/3 or 1/2, depending upon the total number of children in the complete list for each CNC. In larger villages where CNC enrollment was high, the sampling fraction was 1/3, whereas in smaller villages it was 1/2. This resulted in the selection of 431 base line children4. Thus the final sample was composed as indicated in table 4.6. 3 423 children remained after deleting 2 cases of migration for which socioeconomic and demographic particulars could not be gathered, and 1 case where inconsistencies could not be rectified even by going back to the original questionnaires. 4 431 children remained after dropping 7 cases of families who no longer lived in the village. 71 Table 4.6 COMPOSITION OF THE TINP SAMPLE DISTRICT P H A SE B LO C K NO. O F CNCs C U R R EN T CHILDREN O LDER SIBS B A SE CHILDREN TOTAL % O F S A M P L E Madura i I Kottampatti 5 161 154 137 452 36.8 II Chellampatti 4 100 71 105 276 22.5 Ramnad I I I K ariapatti 3 54 50 90 194 15.8 I I I Sivagangai 4 108 98 99 305 24.9 Total 16 423 373 431 1,227 100.0 Percent 34.5 30.4 35.1 100.0 Thus, the sample consists of data drawn from rural households in 16 randomly selected villages from 4 development blocks from the two earliest program districts in the state of Tamilnadu. An attempt was made to capture the as much variability in the data as possible. One of the districts (Madurai) is considered economically more developed than the other (Ramnad). The 4 blocks were selected on the basis of an indicator of overall economic development of the block - GAI/GCA. Two of the blocks had a high GAI/GCA ratio and two had a low one, with one of each kind from each district. Thus, between region variability due to varying levels of overall economic development is expected to be captured by the data. Within block variability is expected to be captured by randomly selecting more than one village in each block - between 3 to 5 villages per block are selected. Between village variability is captured by covering 16 randomly selected villages. Within village variability is captured by randomly selecting 1/3 to 1/2 of the children enrolled at the base line in each village, and all children enrolled for not less than 12 months from those currently enrolled. This, together with weight data of all participating older siblings of those currently enrolled, are expected to capture variations over time. The 72 data are expected to be representative of the situation in rural Madurai and Ramnad districts, and, perhaps, to a lesser extent of rural Tamilnadu in general. Nothing can be said about the data’ s representativeness of special situations without further study: for example, hill districts of Tamilnadu or elsewhere where agro-climatic factors and disease patterns may be different. Moreover, since the data are exclusively rural, they cannot be taken to represent urban populations or even populations partly urban since urban areas typically tend to have better facilities like hospitals, protected water supply, communication facilities, etc., available even to the relatively economically worse off households. These are likely to be especially relevant to health status when the issue of concern is weights of preschoolers. Weight records are available only for program participants, which could lead to a selectivity bias. This is the usual problem with data generated by clinics when information is available only for participants who make up a small percentage of the target population, and who differ from non-participants in important ways, say, by socioeconomic status or disease patterns. However, since participation rates are observed to be high5 in TINP areas, the problem is not serious. Martorell’s (1986) evaluation of program data specifically referring to the selectivity bias problem is as follows: " The data indicate that registration by the CNC is very high and that most of the children registered are weighed every month. Unlike other projects, the concern with bias in the case of TINP is much less. Project data may be used to evaluate impact without the usual strong lamentations about coverage and bias." To the extent there are program impacts, the current weight data cannot be considered representative of a pre-program situation. The base data, however, are 5 An average of over 90% of the under 3 year old children in the community were reported to be participating, except in the early program months when participation rates ranged between 70% to 80% in each phase (Office of the Project Coordinator, TINP; Chidambaram, 1989). 73 better in representing a pre-program situation since they contain weight observations starting from the first month of program introduction, before the program can be considered well established and when it was relatively new to the households in each of its phases. Keeping the above in mind, the data can be considered representative of a pre-program and program situation in rural Tamilnadu. To a lesser extent they can be considered representative of rural southern India and other parts of rural India. Findings based on TINP data can be of relevance to other LDCs, especially in South Asia, provided overall indicators, apart from economic, are also similar. In particular, if indicators of levels of social development like literacy rates, especially female literacy rates, infant mortality rates, health care and hospital facilities, water supply, etc., are broadly similar. 4.3 The Data The TINP sample contains data on 1227 preschool children enrolled in TINP, from 854 households. The children have been drawn- from 3 categories: Children Households (a ) Those cu rren tly enrolled between March-April 1989 . . . 423 423 (b ) Older siblings of those currently enrolled . . . 373 (c ) Children enrolled when the project was ju st introduced excluding sibs of those currently enrolled . . . 431 431 Total 1,227 854 74 Of the 373 siblings, 107 were also in the base, in addition to the 431 mentioned in (c) above. On each of the sample children data were collected on a number of individual level, household level, and communal level variables, during the period March to April 1989. The data include, first of all, monthly child weights, measured in kgs., over the duration of each child’s participation in the program, and also the nutritional grade in which the child was placed each month, depending upon observed weight deficits as compared with ’ norms’. In addition, at an individual level, the data contain information on the date of birth6, sex, birth order for each child, and some health related information like the number of times dewormed, administration of vitamin A, treatment for diarrhoea, etc. At a household level information is available on the socioeconomic attributes of the household (religion, caste, parental education, occupation, annual income and expenditure, savings, indebtedness), kind of dwelling unit (type of roof, walls, number of rooms, whether electrified), assets (type of land operated if any, farm animals owned, vehicle owned), and other household particulars (number of children, number of adults, whether joint family). At a communal level data is available on variables like the distance from the village to the nearest health facility, distance to the nearest general hospital, whether the village is connected by a pucca road, kind of water supply and sanitation. 4.4 Data Sources The data that make up the TINP sample are based on three sources: 6 In the present research obtaining accurate dates of birth was crucial since, whether, and the extent to which a child manifests weight deficits cannot be established without reference to its age. 75 (a) On going and past program records maintained in each CNC at the village level were used to obtain weight and grade observations over time for each sample child. This was also the source for other individual level variables like the date of birth and sex of each child, information on supplementary feeding, vitamin A administration, deworming, referral for medical care, treatment for diarrhoea, etc. The program records were maintained by trained Community Nutrition Workers (CNWs) in each CNC. The CNW is usually a mother (to promote motivation and carry greater conviction among others regarding child care) with at least 8 years of schooling, who is given 2 months of preservice training. She is not a government servant with fixed horns of work and generally lives in the local area itself. Her work, including record maintenance is supervised by a Nutrition Supervisor, with one supervisor for 10 CNWs. Because of regular supervision, and uniformity of training, recorded weight data is not expected to suffer from any serious measurement error problems. Accuracy of weight records was not systematically and independently checked during the field work (except for a few cases randomly checked and found correct when weighing was actually in progress), and available information was recorded onto the questionnaires. Martorell’ s (1986) evaluation notes about TINP indicate that reliance can indeed be placed upon age and weight records, except for the second decimal point: "I was astonished to see record after record of regular monthly weighings. The CNWs ... know their communities well and incorporate children into the system shortly after birth. Birth dates are recorded precisely and age estimation is not a problem." He goes on to observe, "The scale adopted for weighing is easy to use and can be read with ease even when children are upset and crying. All the weighings I observed were done correctly but most values were rounded to 76 the nearest tenth of a kilogram ( xxjc) and only occasionally were fives recorded as second decimals (xx.x5)." (b) Survey responses addressing households of the sample children were used to obtain socioeconomic and other household level information, and also community level information. Draft questionnaires were first prepared, translated into Tamil, the local language, with the help of bilingual translators7, and pretested in a non-sample village. After carrying out modifications based upon feedback from the pretesting, the questionnaires were finalized. Survey responses (coded) were recorded directly onto the questionnaire forms. There were a few open ended question as well. Mothers of the sample children were the respondents, except in cases where the mother had died or was not available in the village (19 cases). In such situations another adult relative responsible for the care of the child was the respondent. The respondents were remarkably forthcoming in their responses to most questions. The exceptions were responses to questions about income, expenditure and, sometimes, indebtedness. As often found in other similar surveys, here there was frequently hesitation in responding due to either reluctance to give information, or incomplete information available with the respondent, or both. Answers to these question were elicited at the end, so as to minimize interference with the rest of the responses. The unreliability of self reported household income and expenditure makes them less useful for analysis. Information about kind of dwelling unit - type of wall, type of roof, number of rooms, and whether electrified - was verified by the interviewers actually looking at each dwelling unit. 7 Actually, the translation was done in two steps. First, the questionnaires were translated into Tamil by one translator. Then, the Tamil version was retranslated into English by another, and the two versions compared, to ensure that loss of detail or nuance was at a minimum. Only after reconciliation of the two versions were the questionnaires used. 77 Other household particulars like number of adults, number of children, whether joint family, religion and caste, educational level of each member of the household, etc., were obtained on the basis of survey responses. Community information about type of water, distance to health and hospital facility, whether the village was connected by a pucca road, etc., was cross checked with the local CNW. (c) It was considered important to obtain information on mother’s heights8 in order to be able to capture genetic effects and effects of other family background characteristics (eg., non-genetic factors like diet). So even though the program maintained no records, actual measurements of mother’s heights were taken and recorded in the questionnaires in all cases where the mother was available. Unlike child measurements which vary over time as the individual grows, adult heights do not change. So only one time measurements were necessary. Among the 423 currently enrolled children mother’s heights were taken in 417 cases (of the remaining 6, the mother was not available in the village in 4 cases, and had died in 2 cases). Among the 431 base line children mother’ s heights were taken in 418 cases (of the remaining 13, the mother was not available in 8 cases, and had died in 5 cases). Ideally, both, mother’s and father’s heights would be of relevance. But it is much harder to obtain father’s heights since fathers are usually not available during the day. There was another consideration as well. The interviewers and program staff (who knew the community well, and helped in the survey) felt that fathers would be less cooperative in allowing themselves to be measured, whereas the mothers were accustomed O This suggestion came from Lee Lillard. 78 to the idea of body measurements since they either took or sent their children regularly to the CNCs. Hence, only mother’s heights9 were obtained. Thus, the sample data contain information on a variety of individual, household and community variables gathered from three sources - ongoing and past program records, survey responses and actual measurements of mother’s heights. While care was taken to verify the accuracy of information wherever possible, or evaluate reliability based upon verification done by others (as in the case of child weights) some errors are inevitable. Perhaps, least reliable is the information on household income and expenditure. 9 In taking height measurements, the guidelines recommended for field workers were followed. While measuring heights sounds simple, care has to be exercised about a number of points. The interviewers were trained with the help of a public health nurse in charge of supervising health related inputs in the sample area. The instructions followed were: In each village a single suitable wall should be selected, perpendicular to the ground, with a flat, even floor. Measurements should be marked off in cm. All subjects should be measured against the same scale. Each subject should be asked to stand against this scale, erect, with no footwear, with the center of the back touching the scale, feet parallel, and heels, buttocks and back of the head touching the wall, and arms hanging by the side. A smooth ruler should be held on top of the head, at a right angle to the scale, firmly pressing the head in the center. Height is then read off from the lower edge of the ruler to the nearest 0.5 cm. Each reading is taken twice to ensure correctness of measurement. Chapter V 79 CHILD WEIGHTS: LEVELS AND TRENDS In this chapter two questions are addressed. First, what are the levels of weights for different ages observed among preschoolers in the sample from TINP areas, and are there changes in the levels over time? In section 5.1 this is explored using some individual growth curves. Section 5.2 examines the same question based on separate curves of mean weight by age for earlier and later participants using all weight observations from the sample. Second, how do mean weights and trends from the sample compare with those observed in other Indian data sets? To answer this, in section 5.3, published data on mean weight and trends for preschoolers of different ages for the entire state of Tamilnadu are examined. So also are age wise averages observed among children from better off Indian households. Section 5.4 summarizes the findings. 5.1 Some Individual Growth Curves: TINP Sample Preliminary explorations on levels and trend based on individual growth curves are carried out in this section. The curves in figure 5.1 show the weight growth of two sisters observed over different time periods. Individual monthly weights have been plotted on a weight chart used to monitor weight growth in TINP program areas. The heavy lines separate observed weights into 5 grades, the grades indicating increasing degrees of weight deficit as compared with the norm adopted: normal and grades I, I I , ni and IV. Comparing weight curves for siblings helps control for genetic factors, household economic and social status. The children belong to a landless, scheduled caste household. Since the siblings are both female, it also controls for sex. Apart from individual variability, the sisters mainly differ with respect to date of birth - there is a gap of 27 months in their ages. Of the two growth curves the upper curve relates to a currently enrolled female (weights taken during the 28 months from January 1987 to April 1989) while the lower curve to her older sister (weights taken for 31 months between October 1984 to April 1987). The growth curve of the later bom, currently enrolled child is higher for all corresponding ages as compared with her older sister’s, with the gap between the two being the narrowest at lower ages. Weights of the later bom are in the ’normal’ grade in the early months, falling well into grade I after the 8th month. With supplementary feeding growth picks up once again, the level continues to be in grade I. For the earlier bom sister, weights start out being in grade II, move towards grade III, are briefly in grade III and return to grade II with supplementary feeding and medical attention. As the children age there is a flattening of the curves. That the child from a later cohort has a higher curve than her older sibling indicates that time, including duration of program operation, might have a positive effect on weight outcomes. Possible negative birth order effects of being later born, if any, have been compensated for in this case. Figure 5.2 shows the growth curves of 2 siblings, a male and a female, with 2 m overlapping period of 12 months in being weighed. The male is 19 months older than his sister. Once again, being siblings, genetic and socioeconomic factors are largely controlled for. The children belong to a landed household. One would expect the effect of time to benefit the later born sister. However, in spite of belonging to an earlier birth cohort, the male child has a higher growth curve (March 1985 to May 1987) as compared 81 with that of his later born sister (June 1986 to December 1988). Weight observations for the male are available only from his tenth month of age. He is either in the ’normal’ grade, or in the upper part of grade I, very close to the ’ normal’. For his later born sister, weight observations are in the ’ normal’ range to begin with, slip into grade I with a flattening of curve, and for the most part are in the lower part of grade I, approaching the ’ normal’ grade after age 33 months. That the sister’s curve is well below that of her earlier born brother at all ages indicates the possible positive effects of being male. 5.2 Mean Weights Based on the TINP Sample Levels and trends are examined in this section based on mean weights for different ages using all weight data from the TINP sample. The TINP sample contains information on monthly weights of children between the ages of 6 - 36 months, separately for the currently enrolled children (1987 - 1989), their older siblings (1983 - 1988), and base line children when the program had just been introduced in each of the 16 sample villages (1980 - 1984). Though the children are different, exactly the same villages are covered for the current, sibling and base line data. The base line data contain information on weights when the program was just introduced in each of the 16 sample villages in the 4 sample blocks. So program effects are expected to be minimum. Baseline children belong to the early birth cohorts in the sample. The current data can be taken to represent a situation where the program is well established and known among the village population. Currently enrolled children are from later birth cohorts. If there is an improvement in weights over time, either due to beneficial program effects, or otherwise, we would expect means for the currently enrolled to be higher than 82 base line means. The sibling data contain weight information on older siblings of the currently enrolled, when they were in the preschool age group and participated in the program. Their birth cohorts lie largely in between the currently enrolled and the base line children. There is some overlap, though, between the older siblings and the younger children among the currently enrolled, and between the younger siblings and the older children from the base line. Each child is weighed every month while it is 6 - 36 months of age for the duration of its participation in the program. The average duration of program participation was 21 months. In calculating mean weights for each monthly age, fluctuations in weights due to temporary infectious episodes tend to be evened out. Also, since each of the current, sibling and base data sets cover more than one calendar year, fluctuations due to a single bad or good monsoon or agricultural year also tend to get evened out. (This will be a possible problem when we examine trends in mean weights by calendar year). Figures 5.3 to 5.6 graph the mean weights (in kgs.) for monthly ages observed in the sample1 . Figure 5.3 graphs the mean weights by age for all children with separate curves for the currently enrolled, siblings and base line children. Figures 5.4 to 5.6 present the mean weights separately for males and females. The following 2 points are observed: (1) Mean weights of the currently enrolled children for each monthly age are higher than those of their siblings, which, in turn, are higher than the mean weights of the base line children (figure 5.3). Between the three groups of children the gap between the base line children and the currently enrolled is the highest indicating that duration of program operation might be influencing weights. Testing for the significance of the 1 See appendix, tables 1 to 3 for means, SDs and number of observations for each age. 83 differences in mean weights it was found that both the current and sibling means were significantly higher than the means for base line children for most of the ages. The differences between the currently enrolled and their siblings were not statistically significant in most cases. However, that mean weights of the currently enrolled are higher than sibling mean weights for each age is significant for another reason. In view of the expected negative birth order effects, mean weights among older siblings (the earlier bom) can be expected to be higher than the mean weights of the currently enrolled (the later born). Yet mean weights for the currently enrolled are higher for each age. Thus, controlling for age, a definite upward trend in mean weights is observed over time in the sample from TINP areas. (2) When male and female means are compared, male means are higher for each monthly age. This is true for the currently enrolled, their older siblings and base line children (figures 5.4 to 5.6). Male-female differences by age were found to be significant in all cases. 5.3 Comparison With Weights Based on Other Indian Data How mean weights and trends observed in the sample from TINP areas compare with those observed in other Indian data sets is examined in this section. First, mean weights based on data for the entire state of Tamilnadu for the period 1974 - 1982, compiled by the National Nutrition Monitoring Bureau (NNMB), Hyderabad, are observed and trends examined. Mean or median child weights observed among well off households from two other parts of India are also reported in order to get an idea about high weight values that can be expected among preschoolers. 84 The NNMB, set up as an apex nutrition monitoring body in India, has been collecting and compiling extensive weight data (among other things) on a representative sampling basis for different states in India. Mean weights among preschoolers for Tamilnadu are available for most years from 1974 to 1982. The NNMB data are much wider in coverage as compared with TINP and include urban children also, even though they are less detailed. The period of coverage starts well before TINP was introduced in 1980. The NNMB mean weights can be taken to represent pre-program and non-program mean weights - i.e., an indication of levels and trends in the general preschool child population in Tamilnadu. In the last 3 years of NNMB data, 1980 - 1982, the integrated health and nutrition program, TINP, had been introduced in 2 districts of Tamilnadu. However, for two out of the last three years, 1980 and 1981, there is no overlap between the TINP districts and the NNMB sample districts. In 1982, one out of the four NNMB sample districts was also a TINP district (Madurai), but the other three NNMB sample districts were outside TINP areas. So in 1982 alone there is a possibility of a small overlap2. NNMB data are compiled by year of age and not by monthly age. For comparability with the sample data from TINP areas we restrict ourselves to an examination of NNMB data for children who are less than 36 months old. Mean weights for the 3 yearly age groups of 0, 1 and 2 years for males and females from Tamilnadu over 2 The possibility could not be absolutely ruled out because information from the NNMB about actual villages and urban areas covered was not easily available. My discussions with the doctors at NNMB (K. Vijayraghavan and J.G. Sastry) seem to suggest that the possibility of overlap was negligible. The NNMB sample was intended to represent the entire state of T am ilnadu. 85 the period 1974 - 1982 are graphed in figures 5.7 and 5.83. The following points are observed: (1) The figures show no visible trend with fluctuations between years. For this 8 year period a time trend fitted separately for males and females in each age group was either negative and insignificant or positive and insignificant. (2) Female mean weights are below male means for ages 1 and 2 years but not for age 0 (under 12 months). As against a lack of trend in Tamilnadu as a whole, we observed increasing mean weights by age in the sample TINP blocks with means for the currently enrolled children being higher than those for their siblings, which, in turn, were higher than base line means. However, the data were compiled differently from the NNMB. In order to permit more direct comparisons between the means from the sample blocks and NNMB means, the weight data from the sample blocks were regrouped by yearly ages and separated by calendar year. The distinction between currently enrolled, siblings and base line children was no longer maintained. Three points may be taken into consideration while comparing NNMB and sample means. One, the sample data are exclusively rural while the NNMB data are intended to represent the entire state, including urban areas. To the extent inclusion of relatively better off urban children results in higher means, we might expect lower mean levels in the sample as compared with NNMB data. Two, in respect of trend, mean weights are now distinguished by calendar year and data for more than one calendar year are no longer pooled together as was the case in section 5.1. So seasonal fluctuations due to good and bad years are likely to be visible (Chambers et al., 1979; Chambers, 1982). 3 See appendix table 4 for means, SDs and number of observations. 86 Three, though the NNMB data relate to an earlier period (1974 to 1982) as compared with the sample (1980 to 1989), there are 3 years of overlap from 1980 to 1982. Direct level comparisons can be made for this period. For this period the sample however covers only children from the base line where the lowest means were observed. Figure 5.9 graphs the mean weights during the overlap period for each of the 3 ages, separately for males and females, using the NNMB and sample data from TINP areas in turn. Based on the sample data figures 5.10 to 5.15 contain graphs of mean weights by calendar year for the period 1980 - 1989 separately for all children, males and females4. The following observations are made: (1) During the three year overlap period differences between the NNMB and sample mean weights are significant in 9 out of 18 comparison points. Of these 9 points NNMB means are higher than sample means at 5 points with the reverse being the case at the remaining 4 points. Overall, neither is clearly uniformly above the other (figure 5.9). The overlap period is not long enough to draw conclusions about fluctuations. (2) When means from the sample are examined for the entire period, 1980 - 1989, separately for the 3 yearly ages, a clear upward trend is observed for ages 0 and 1 in respect of all children (figures 5.10 and 5.11), and also in respect of males and females (figures 5.13 and 5.14). Time trends fitted for each of these ages were all positive and significant. (3) For children 24 months of age and above (i.e., age = 2 years) the time trend fitted continued to be positive and significant, but was less strong as compared with the trends observed among younger children (figures 5.12 and 5.15). 4 See appendix tables 4 to 6 for the data. 87 (4) Female mean values are significantly lower than male for corresponding age groups (figures 5.13 to 5.15). Average weights among children from well off Indian households from the cities of Hyderabad and Delhi are reported in tables 5.1 and 5.2. These are substantially higher than the means for corresponding ages in the sample in respect of, both, males and females. In fact, they are closer to the US reference weights. This is consistent with the findings from other developing countries that differences in body sizes between well-nourished children across ethnic groups are minor (Martorell and Habicht, 1986; Martorell, Mendoza and Castillo, 1988). 5.4 Summary Preliminary explorations based on individual growth curves for siblings from the TINP sample suggest that weights of children from later cohorts could be higher than weights of the earlier bom, and weights of male children higher than those of females. Next, based on the sample, mean weights by age calculated separately for the currently enrolled, their older siblings and base line children are examined. We find that mean weights have increased significantly over the 1980 - 1989 period. Means for later participants are higher for each age as compared with the means for earlier participants for all children and separately for males and females. Mean weights for female children are significantly lower than those for males for each age, and this continues to hold good over time. When the sample mean weights by yearly age groups are compared with mean weights observed in the NNMB data for the state of Tamilnadu as a whole during the three year overlap period, neither set of mean weights are found uniformly significantly 88 above the other. NNMB means are higher in respect of some ages for some years with the reverse being the case in respect of others. In both cases female means are significantly lower than male means. As against a lack of trend for Tamilnadu as a whole based on NNMB data for the period 1974 to 1982, a significant upward trend is observed in the sample areas during 1980 to 1989. Sample mean weights are substantially lower than the mean or median weights observed among well off Indian children. sr«m IC S fiu iriD jr/riS ftj W e ig h t i n KGs 5«* I9f & r @ B , D f t c © w V 0) fi) f t H ft) H- (D t r 0 “ n r n 0 3 • • 3 II O ^ 0 C D f t 3 H H _ V O V O 00 00 a s - ■sj i i > > t J t J i-f H- H- H H H H V O V O 00 00 v j V O M IH IH IIII Q W O 5 ! ► 3 a o a w < M c o o a • - 3 S O C O H t o M J O C O a J O O 3 i -9 H 3 U C O a I t1 M a H - v Q C t D C O $ 6 T W H . IG co /rd B jririS A ) Weight in KGs. 90 F i g u r e 5 . 2 GROWTH CURVES OF TW O SIBLIN G S FROM TINP SAMPLE Year 1 Male, earlier born: Mar 1985 - May 1987 Female, later born: Jun 1986 - Dec 1988 @ y ? fe 6 lD ^ u S ls3 r 6T65)i @ j!5 1 u C c U © MEAN WEIGHT (Kgs Figure 5.3 MEAN WEIGHTS: CURRENT, SIBS & BASE CHILDREN TINP Sample Current children, 1987-1989 Siblings, 1983-1988 Base line children, 1980-1984 LEGEND AGE IN MONTHS Figure 5~4 MEAN WEIGHTS: CURRENTLY ENROLLED MALES & FEMALES TINP Sample, 1987-1989 10 O' Q Males Females LEGEND: 14 16 18 20 22 24 26 28 30 32 34 36 10 AGE IN MONTHS MEAN WEIGHT (Kgs. Figure 5.5 MEAN WEIGHTS: MALE & FEMALE SIBLINGS TINP Sample, 1983-1988 LEGEND: 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 AGE IN MONTHS ^ Figure 5.6 MEAN WEIGHTS: BASE LINE MALES & FEMALES TINP Sample, 1980-1984 0 9 8 7 LEGEND Males Females 6 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 AGE IN MONTHS M E A N WEIGHT Figure 5.7 FEMALE MEAN WEIGHTS, 1974 - 1982 NNMB Data For 0 - 2 Year Olds, Tamilnadu o --- 1974 197S 1976 1977 1978 1980 Y E A R S + 1+ Year# 4 2+ Year* M E A N WEIGHT MALE MEAN WEIGHTS, 1974 - 1982 NNMB Data For 0 - 2 Year Olds, Tamilnadu 10 - 1974 1975 1977 1978 1980 Y G K R S □ 0+ Yeor + 1 + Years O 2+ Years MEAN WEIGHT (Kgs. Figure 5.9 MALE & FEMALE MEAN WEIGHTS, 1980 - 1982 TINP Sample & NNMB Data MALE FEMALE LEGEND Sample NNMB 1980 1982 1980 1982 YEARS M E A N WEIGHT < K g O Figure 5.10 MEAN CHILD WEIGHTS, 1980 - 1989 TINP Data For Age < 12 Months 7.4 - 7.2 - 68 - 66 - 6 4 - 6 2 - as - 6 4 - 1981 1982 1984 1987 M EA N WEIGHT <K q b .> Figure 5.11 MEAN CHILD WEIGHTS, 1980 - 1989 TINP Data For 12 <= Age < 24 Months 8 3 as 8 7 86 aa a4 8 3 8 2 a i 8 7.3 7.8 7.7 7.6 7.S 7.4 7.3 7.2 7.1 7 1380 1381 1382 1383 1384 13SS 1385 1387 1388 Y E A R S M E A N W EIGHT ( K 9 O Figure 5.12 10.4 1Q3 1Q2 1Q1 10 9.9 9.8 9.7 9.6 9.3 9.4 9.3 9.2 9.1 9 8 9 88 8 7 86 8 3 1980 1981 1982 1983 1984 1983 1986 1987 1988 19a» YEARS M S MEAN CHILD WEIGHTS, 1980 - 1989 TINP Data For 24 <= Age <=36 Months M E A N WEIGHT (K G S.} Figure 5.13 MALE & FEMALE MEAN WEIGHTS, 1980 - 1989 TINP Data For Age < 12 Months 7.4 - 7.2 - 68 - 66 - 6 4 - 6 2 - 5 8 - 66 - 6 4 - 1981 1987 1988 Y E A R S □ R*mdes + ■ Mdes M E A N Vr'EIGHT <KGS.) Figure 5.14 MALE & FEMALE MEAN WEIGHTS, 1980 - 1989 TINP Data For 12 <*= Age <24 Months a s - aa - a7 - ae - aa - a4 - a3 &2 - ai 8 7.9 - 7.8 7.7 -i 7.6 - 7.5 7.4 - 7.3 - 7.2 - 7.1 - 7 1989 8 Figure 5.15 1Q4 10 3 1Q2 1Q1 10 9 9 9.8 9 7 9 6 9 8 9 4 9 3 9 2 91 9 8 9 88 8 7 86 88 1980 1981 1982 1983 1984 19S5 1986 1987 1988 1989 Y»RS □ F e m d e s + ■ M des 8 MALE & FEMALE MEAN WEIGHTS, 1980 - 1989 TINP Data For 24 <= Age <=36 Months Table 5.1 50TH PERCENTILE VALUES OF WEIGHTS OF WELL-TO-DO HYDERABAD BOYS AND GIRLS Age Boys Girls (Months) Weight Weight (Kgs.) (Kgs.) 12 9.3 8.7 15 9.9 9.2 18 10.5 9.8 21 11.2 10.1 24 11.6 10.5 27 12.0 10.9 30 12.5 11.3 33 13.0 11.8 36 13.5 12.3 Source: Rao, H., et al (1976). Table 5.2 50TH CENTILE VALUES OF WEIGHTS OF MALE AND FEMALE INDIAN CHILDREN OF A HIGH SOCIOECONOMIC GROUP Age Males* Females (Months) Weight Weight (Kgs.) (Kgs.) Birth 2.97 2.96 6 7.44 6.86 12 9.45 8.76 18 10.49 9.84 24 11.56 10.92 30 12.38 11.79 36 13.52 12.96 * Means are provided as 50th centiles are not available. Source: Shanti Ghosh (1986). 104 Chapter VI 105 METHODOLOGY Having observed the levels and trends in preschooler weights, and the relatively high percentages of children manifesting different degrees of weight deficits compared with ’ norms’, the next question is to examine the causes of these child weight outcomes. But before so doing, this chapter first contains a discussion on the choice of variables in modelling weight outcomes in section 6.1; then, after a description and interpretation of the variables used in the analysis in section 6.2, the econometric procedure employed and its rationale are outlined in section 6.3. 6.1 Choice of Variables. Modelling of health and nutritional outcomes using anthropometric indicators, though of relatively recent origin among economists, is not new in medical, public health and nutrition literature. A number of existing studies have examined the biological, social and economic factors and mechanisms underlying variations in anthropometric outcomes. There is general agreement that the environment in which a child lives (the micro household and the more macro community) influences growth through its effects, ultimately, on diets and the incidence and severity of infections (Gordon et al., 1963; Jelliffe and Jelliffe, 1978; Scrimshaw and Underwood, 1980; Black et al., 1982; Chen and Scrimshaw, 1983; Martorell and Yarbrough, 1983; Martorell and Habicht, 1986). Variables affecting diet and disease loads have been thought of as either 106 proximate determinants (eg., food consumption, breast feeding, timing of introduction of solid foods, diarrheal episodes, etc.) or underlying socioeconomic or cultural factors (eg., economic status of the household, environmental sanitation, dietary and health practices, etc.). Economists have viewed anthropometric outcomes as a health-nutrition production function (Rosenzweig and Schultz, 1983; Rosenzweig and Wolpin, 1986; Barrera, 1987; Behrman and Wolfe, 1987, Strauss, 1990; Akin and Guilkey, 1990; etc.). A methodological issue concerns the choice of variables to be included as regressors. Estimation of an equation that includes as regressors not only variables that are exogenous but also those that come under the purview of household choice causes problems of simultaneity resulting in estimates that are biased and inconsistent. Consider, for example, the variable ’utilization of health facility1 . If availability of such a facility is not in question, its utilization can be thought of as a household choice variable. Then, to the extent children who fall ill more frequently tend to be taken to the health facility more often, utilization of health facility will have downward biased and even negative coefficients. Similarly, if healthier children are better able to breast feed as compared with the less healthy, the breast feeding variable will have coefficients that are upward biased. No doubt these variables are in fact associated with health and nutritional outcomes and do belong in a well specified health-nutrition production function. But using OLS rather than a simultaneous equations approach will cause biased estimates. On the other hand, it needs to be recognized that the possibility of choice does not automatically rule out treating a variable as exogenous. Consider breast feeding - mothers can choose its duration. However, if due to cultural factors most mothers breast feed as long as possible, it is not really a choice variable. In fact, there may not be much variation in the duration of breast feeding among mothers and across children. A similar 107 argument is applicable in the case of variables like family size or government sponsored mandatory immunization programs. A clear distinction between whether a variable is necessarily exogenous or endogenous may not always be possible for all circumstances and the issue needs to be examined separately in each case. Variables that could come under the purview of household choice have been used as regressors in several studies. Levinson (1974) uses household calorie and protein availability as covariates in regressions of standardized child weights and heights using data from rural India. Battad (1978) uses a food consumption variable, using a rural Phillipine sample. Heller and Drake (1979), while treating illness as endogenous, take food consumption, age at weaning, birth interval etc., as exogenous. Wolfe and Behrman (1982) include average calorie intake, length of breast feeding and use of refrigeration among regressors. Ryan, Bidinger, Rao and Pushpamma (1984) use energy and protein consumption and mother’s labor market participation among explanatory variables. Martorell, Leslie and Moock (1984) include months of breast feeding and food types eaten among their regressors to explain heights and weights on the basis of Nepalese data. The possibility of bias in respect of such specifications has been pointed out in Rosenzweig and \ Schultz (1982), Barrera (1987), Strauss (1990). However, in as much as the possibility of choice need not result in endogeniety in each case, the criticism is not automatically applicable. It would have been useful if the above studies had contained a discussion on I the issue of exogeniety or endogeniety of the chosen regressors and the likelihood of simultaneity and bias. This issue has been explicitly recognized in more recent studies. Two of the early ones are those by Chernichovsky and Coate (1983) where reduced form equations are estimated when analyzing child health of US children, and by Chernichovsky, A 108 y" Kielmann, Kielmann and Rienke (1983) where the correlates of preschool child growth in ! I Pimjab are studied. Other studies which are careful recognizing the problem of \ simultaneity resulting in biased estimates are Schultz and Wolpin (1984), Behrman and Deolalikar (1985), Behrman and Wolfe (1987 & 1987a), Horton (1986), Barrera (1987), Thomas, Strauss and Henriques (1987) and Strauss (1990). Though the Tamilnadu data set has a number of variables that should in fact belong in a well specified health or nutrition production function (eg., utilization of health facilities, length of breast feeding, age of introduction of solid food, immunization, number of children etc.), it is proposed here to choose only those variables that can be largely considered as given to the household in a particular time period. Because of the a program a health facility is available in all the sample blocks, Utilization is treated as endogenous and not included as a regressor. Instead, distance to the health facility is included as a regressor. Again the program, through its communication component, continuously promotes breast feeding, the timely introduction of solid food, immunization against polio and DPT, spacing of children through the use of modem contraceptives. Thus parents are made conscious and could be taking deliberate decisions on these issues. Such variables are therefore excluded as regressors. No doubt, a fully simultaneous system / 0.. / u of equations would be the ideal approach, but estimating only a small part of the system is/ I the more modest aim here for two reasons: (a) While the patterns of human growth are now well documented, the underlying biological causes and processes are still imperfectly understood. As available today, therefore, theory is less than perfect, (b) Data availability is a second constraint. In particular, the data set does not have complete information on many variables known to influence preschooler weights - birth weights are available for a 109 very small proportion of the children, there is no information about the mother’s prenatal history, etc. Hence the more modest approach is proposed here. 6.2. Variable Description. Dependent variables. Monthly child weight and standardized weight-for-age scores are the dependent variables. The standardized scores are calculated by dividing each weight observation by the age-sex specific median weight from the US NCHS standards, allowing high degree of reliability, is responsive to the independent variables of interest, and is commonly used in the biomedical and public health literature for its growth and health significance. Evidence from many developing countries suggests that weight for well not (Martorell, Habicht, Yarbrough, Malina and Klien, 1974; Martorell and Habicht, 1986). Since age-sex specific standards are used, the coefficients most affected when comparing actual weight with standardized weight will be those of age and sex. For example, even if observed female weights are lower than male weights for a particular age, since NCHS median female weights are also lower than median male weights, observed female weights in the data will be divided by smaller numbers in the standardization, thus reducing the male-female gap in standard weights as compared with actual weights. Individual variables. Child age is used in two ways. First, monthly age of the child and square of age are used to capture the effects on weight due to age, the squared term allowing for a curved relationship. In another set of equations dummies for age groups are direct comparisons between children of different ages and sex. Weight is measured with a nourished children is close to the NCHS median while that for undernourished children created to further explore the relationship between weight and age. Weight is expected to vary with age (increasing as the child grows older) the way age-sex standardized weight is not. Standardized weight can fall rapidly when there is evidence of growth faltering during certain stages of human growth, for example, during weaning. Evidence from developing countries has shown a continual worsening of nutritional status as age increases owing to environmental influences and the cumulative effects of inadequate nutrient intakes. Birth order is expected to have an adverse impact on weight, higher birth orders being associated with lower weight and standard weight. Empirical evidence for inequalities in resource allocation by birth order for resources such as schooling and educational attainments among siblings exists (eg., Lindert, 1978; Birdsall, 1979; King and Lillard, 1983; Behrman and Taubman, 1986). This could be due to greater strain on household resources of time and purchasing power. In the case of weight a higher proportion of children in the household could have adverse effects due to the increased probability of infectious episodes. In addition, a biological factor pointed out in Horton (1988) is maternal depletion. Later bom children are bom to older mothers and tend to be of lower birth weight. She has also referred to culturaLand socioeconomic factors which can cause a tendency among parents to favor earlier birth orders. For example, if the oldest son is important in funeral rites or if parents plan in their old age to depend upon their oldest children who become economically independent the earliest. In the present sample birth order can be considered exogenous, unlike birth interval or the number of children. The program’s messages include child spacing and a small family size. To the extent households respond, birth interval and family size become endogenous. Hence the preference for birth order as a regressor. To capture possible effects of sex, a dummy for females is used. Gender preference has been observed in certain societies. Especially in South Asia a bias against females has been reported for IMR, childhood morbidity and schooling attainments. Sex differences, if any, were expected to be captured by examining age-sex standardized weight by sex. Household variables. Mother’s height is used as a proxy for her genetic and health endowments. As pointed out in Strauss (1990), it is also likely to pick up other family background characteristics not captured by the education variable. For instance, father’s genetic and health endowments (to the extent there is a tendency to look for similar characteristics in mate selection), cumulative historical socio-economic status, potential attainment levels, are some examples. The likelihood of picking up other family background characteristics not captured by the education variable is particularly important for the rural Tamilnadu sample where as much as 66% of the mothers have no education, 24% have only upto 5 years of education, and just 10% have more than 5 years. Mother’s education is used as a class variable based on years of completed formal schooling. The variable is allowed to take on 3 levels, one each for those with no schooling, for those with some schooling of upto 5 years, and for those with more than 5 years of schooling. Two dummies were constructed to capture the differential effects of the 3 levels under mother’ s education, illiterate mothers with no schooling being ( the omitted category. The cutoff point chosen is 5 years since that is when a child passes from primary school to middle school. Better educated mothers may be more efficient at producing child health and nutritional outcomes, may bring in additional resources either on their own or through selective mating with more educated men, and improve resource allocation within the household as a result of better knowledge and access to information (Welch, 1970; Michael, 1973; Rosenzweig and Schultz, 1985; Barrera, 1987). Land operated by the household is used to capture long term household economic status. If one of the reasons for low weight is inadequate command over resources, then households with better economic status should be under less strain and should have children with higher standard weight. Land is treated as a class variable and 3 levels of land operated were considered - the landless households, households operating land but with uncertain irrigatjon-facihties, jm d households operating land with assured irrigation. Dummies for the two landed classes were created, the landless being the omitted category. While information on annual income from various sources and expenditure was available, the probability of measurement error problems was considered high. Moreover, weights of older siblings when they were in the corresponding 6-36 month age group were also to be analyzed by household economic status. Land operated by the household was likely to change much more slowly over time than income or expenditure. An index of type of dwelling unit was also used instead of land to capture long term household economic status. This combined the following 5 house characteristics: * rented or owned house (scores 0 or 1 respectively); * roof type - thatched, tiled, asbestos or RCC (scores 0, 1, 2, 3, respectively); * wall type - nil, mud, stone or brick (scores 0, 1 or 2 respectively); * number of rooms; * whether electrified (score 1 if yes, 0 otherwise). The total score obtained by each household was normalized to vary between 0 and 1 from the lowest to the highest and was called HOUSINDX. Most conclusions remained substantially unchanged using HOUSINDX instead of land and the results are not reported here. , , , _ 1 1 3 . Belonging to one of the scheduled castes (SC) can in the Indian context be taken as evidence of belonging to a particularly socially deprived category (Betielle, 1969 & 1974). A dummy for a household not belonging to the SC category was created to examine the effect of caste. Household structure is related to intrahousehold distribution and is intended to capture the effect of being the child of the household head. Whether being the child of the household head makes a difference is examined by using a dummy for father being the household head1. Household structure could make a difference if differential command over resources within a household results in specially advantageous positions for certain children (those of the household head) as against others. Strauss (1990) finds being the child of a senior and junior wife of the household head has a positive impact on weight. Community variables. Development block was used to capture the effects of the immediate macro economy in which the household resides. The data were drawn from 4 development regions or blocks, two of which were considered economically more developed than the others (as measured by GAI/GGA, chapter V). The 4 blocks, Kottampatti, Chellampatti, Vadipatti and Sivagangai, were assigned levels 1, 2, 3 and 4, respectively. Dummies were created for the first three levels with 4 being the omitted category. A development block is a geographically contiguous region of approximately 75,000 population used as an administrative unit for carrying out government development programs like water supply 1A positive effect of being the child of the household head was expected. So either parent being the household head was not considered as, generally, female headed households tend to be the outcome of widowhood or desertion by a spouse, with consequent economic and social problems likely to adversely influence child weights. improvements, road construction, other rural works, family planning, etc. Block can be taken to represent the overall level of economic development and the availability of various services in each region. Distance to the Primary Health Center (PHC) measured in kilometers was obtained for each of the 16 villages. So was distance to the nearest general hospital. The PHC tended to be used for relatively smaller ailments and immunizations while the general hospital was used for ailments considered more serious and deliveries. Distance to the nutrition center was available but not used since there was one located in each village, and generally within walking distance. In a separate specification for data on currently enrolled children alone, kind of water supply (protected or otherwise) and whether the village was connected by a pucca road were also included as regressors. Water supply, with its effects on childhood intestinal infections and the consequent synergistic relationship between nutrient consumption, absorption and disease, has significance for weight. Four types of water sources were observed in the data. Of these, water from open tanks and wells was considered less hygienic as compared with water from deep bore wells and hand pumps. The former sources could be used by cattle and other animals and also by humans for washing purposes in ways that could contaminate the water. A dummy for hygienic water was created in cases where the water source recorded was a bore well or hand pump, the other two sources forming the omitted category. The existence of a pucca road connecting the village is expected to capture the overall availability of other services (including the frequency with which local officials are likely to come and inspect villages for monitoring existing development programs and inclusion in new development activities), the extent of integration into the rest of the economy, the relative ease with which children can be taken to a hospital, and, perhaps, the overall level of development itself. These two variables were not included in weight equations for base and sibling data because a lot of new water connections and roads had been built under local self sufficiency schemes in recent years, and current information on the type of water supply and the existence of a pucca road was not useful in analyzing base line weight. 6.3 Econometric Procedure. 6.3.1 Parameter Estimation. Weight is thought of as being determined by individual level variables, household specific variables and communal variables. These explanatory variables are divided into two categories: class variables which include only categorical dummies for the groups of interest to the present analysis, and other covariates important in explaining variations in weights. Conceptually, the reduced form functional relationship estimated is weight = f {covariates; class variables; interactions} ... (1) Initially only covariates and dass variables will be used in the analysis, with interaction terms being introduced as a subsequent step. In what follows is described the econometric procedure adopted to investigate the variations in the 2 dependant variables, monthly weight of preschool children and standardized weight-for-age scores. The purpose is to measure: (a) the variation in the dependent variables due to exogenous, non household choice regressors or covariates which are individual level variables like age of the child and birth order; household level variables which include mother’s height, an index of type of house to capture household economic status; and communal variables like distance to the nearest primary health care facility and general hospital; (b) the differential impacts of 6 categorical or class variables which \ include 4 household level indicators, land operated representing long term economic V V X status, maternal education. SC. and household structure: an individual level variable, the sex of the child; a variable to represent the impact of the immediate macro economy relevant for the households captured by the development block in which the household is located; (c) possible interactions between some of the class variables themselves like land operated and sex (do the sexes manifest differential nutritional status across households with differential access to land?); or between (a) and (b) above like birth order and maternal education. That is if, for example, birth order is a significant regressor, then are there differences in its response coefficient across different levels of maternal education (does maternal education partly compensate for the negative birth order effects)? Consider the equation where, to begin with, there are no interactions: K ^ ij = B 0j + lE.Bk ^ k ij + e ij (2 ) I where, i = the individual observation i; j = 1... G refers to the levels or values of the class variable (eg., block = 1, 2, 3 or 4); k = 1 ... K are the K regressors other than the class variables; Yy = the i th observation on the dependant variable for the j th level of the class variable (eg., the first observation for the second block would be Y1 2 ); Xyj = is an observation on the k th explanatory variable for the i th observation and j th level of the class variable (eg., if mother’s height is the k th explanatory variable Xkl2 would be the 1st observation on mother’s height in block 2); Bk = slope coefficients that are initially assumed to be constant over the class variable levels (j = 1... g) and where k = 1...K, K being the number of regressors; B0: = the intercept terms, different for each level of the class variable but constant over individual observations; e;j = the random error. The error term e^ is assumed to have zero mean and constant variance and to be independently identically distributed over individuals and the class variable (eg., block). Since B0 j are assumed to be fixed parameters, the dummy variable or covariance analysis is used to make inferences about group effects. If intercept differences (Bq j) are found significant a natural extension would be to investigate and test for the possible differences in slope coefficients also. The least squares dummy variable technique is used to obtain estimates of B0g and Bk where B0g represents intercept terms, one for each level of the class variable or group being considered (eg. one for each block), and Bk represents the slope coefficient of the k th regressor assumed to be the same for all levels of a group. It is convenient to write (2) above as where Dgj are dummy variables which take the values 0 or 1. Specifically, = 1 if g=j - 0 otherwise. 118 6.3.2 Hypothesis Testing. Testing for intercept differences. The hypothesis being tested is regarding the existence of statistically significant differences between the different levels of the 6 class variables under consideration, land, maternal education, sex, SC, household structure and block, taken one at a time. More precisely, we test for each class variable under consideration, whether its levels differ by intercept while slopes are assumed to be constant for all levels of the class variable. The null hypothesis is HO : B0 j = ... = B0g = ... = Bog H I : B0g are not all equal, under the assumption of a common set of slope coefficients, Bj ... Bk. Testing for heterogeneity of slopes. The above test was conditional on the assumption of a common set of slope coefficients. This assumption requires to be tested before it can be concluded that level differences between levels of a class variable are insignificant. Moreover, whether or not slopes differ can be of interest in itself, involving the testing for slope differences of the effect of a class variable on the regression between the independent regressors and the dependant variable. For example, if there is an overall relationship between birth order and weight, i.e., Bk is not zero for k = birth order, then do these slope coefficients differ by levels of mother’s education with children of less educated mothers showing significantly different response coefficients as compared with those of more educated mothers? HO : Bkl = Bj^ = ... = BkG k = 1..JK, j = 1...G. H I : Bk j are not all equal. 119 This can be handled by introducing interaction terms between class variables and covariates [eg., (birthorder)*(matemal education)]. Alternately, equation (2) can be modified as: K K Y y = B 0 + B pj + ^ B k X y j + ] > B kj X ^ j + e y k=i *»• K = (Bo + Bq j) +^(Bk + Bj^) XW j + ey ... (4) k-cl where B0 and Bk are average intercept and regression coefficients for regressors Xky; B0 j and Bk j represent group effects of level j of the class variable, with j = 1, 2, ..., G as before. The null hypothesis would then be: HO : Bk j = 0, j = 1, 2, ..., G H I : Bk j are not all zero. Possible intercept differences ( B qj’s being different) are not relevant to either version of the hypothesis. Testing for interactions between class variables. If the variance in the dependant variable can be explained by more than one class variable then it is of interest to examine interactions. For example, if the dependant variable weight for age varies with the levels of any one of the class variables, say, land operated, then are these differences related systematically to the levels of another class variable, say, sex. In other words, if differences in land held explain weight variations then are the effects of land different between males and females? Do the different dass variables substitute or complement each other in their effect on the dependent variable? For this the model described by (3) is modified to include more than one class variable. The reduced form functional relationship of interest continues to be (1) above. The specification with interactions between 2 class variables can be thought of as follows: Y ijk = m + V j + tj + (vtyj + B Xjj + eijk ... (5) where, yjjk = k th weight observation from the (ij)th level combination of the classes v and t (eg., i th land and jth sex level or group); m = mean intercept; V ; = effect of the ith level of the class variable v, say, land; tj = effect of the jth level of the class variable t, say sex; (vt)jj = the interaction effect for the i th level of the class variable v and the j th level of the class variable t; B = the regression effect of the covariates, vector or scaler; Xjjk = observed values of the exogenous variables or covariates; eijk = error variation; An extension of this model to include more than two class variables with multiple combinations is conceptually straightforward (interpretation becomes more difficult, however). The class combinations of interest which appear significant have been presented as part of the results below. Dummies are created separately for each class variable and combinations of class variables to allow for interactions. For example, if weight is observed from 3 levels of land operated (no land, land with uncertain irrigation and land with assured irrigation), separately for the 2 sexes, dummies for each land category, sex and land-sex combination are created as follows: 1 21 1 for the mean (intercept) 3 for the levels under land 2 for the levels under sex 6 for the interactions between land and sex. Thus, dummies are created separately for each class variable and combinations of class variables to allow for interactions. Of course, in the process of estimation one level of each class variable is omitted to overcome the problem of linear dependance. The intercept thus represents the mean of the omitted levels. The other parameter estimates represent differences from the omitted levels. Whether or not group differences and their interactions are significant is tested. We now go on to the regression results in the following chapters. C hapter VII 122 REGRESSION RESULTS: EFFECTS OF GROUPS AND OTHER COVARIATES ON CHILD WEIGHT The main issue in this chapter is the examination of the role of individual, household and community level variables1 in explaining weight outcomes. These variables in combination largely make up the environment in which children grow. Some of the variables of interest are categorical in nature, for example, type of land operated, sex, whether belonging to a deprived community (SC), type of water, etc. Maternal education level has also been treated as a class variable. Group effects of belonging to one or other levels of these class variables are examined using categorical dummies in section 7.1. The other covariates like age of the child, birth order, mother’s height, and the two distance variables can be more easily treated as continuous, and their effects are explored in section 7.2. The effects of age are examined both ways. 7.1 Group Effects. This section examines how weights and standardized weight scores of preschoolers vary between the different levels of certain groups or classes. Reduced form equations for weight and standard weight were estimated separately for (a) currently enrolled children, who were 6-36 months of age at the time of the survey (b) older siblings of those currently enrolled when they were 6-36 months of age; (c) children from 1 Described in chapter VI, section 6.2. 123 the base line when they were 6-36 months of age. Several alternate model specifications of the equations for weight and standard weight were tried. The specifications in tables 7.1 to 7.5 allow us to examine group or class effects, i.e., intercept differences between the different levels of the class variables - land, mother’s education, sex, SC, household structure, development block, type of water and whether the village is connected by a pucca road. In addition, what happens to weight over time can be examined. Weight records for base line children extend over the period 1980 to 1984; those for siblings from 1983 to 1988; and for the currently enrolled the period covered is 1987 to 1989. Weight variations over time are studied by making two kinds of comparisons: one, comparing base line regressions with current, and two, comparing the regressions of the currently enrolled with those based on sibling data (63 % of the households in the current data set had an older child who had participated earlier). The first kind of comparison allows us to look at weight variations in an early program situation, starting with the first month of program operation in each block versus current weight variations, i.e., after 6 to 8 1/2 years of program operation. The second kind of comparison allows us to compare weight variations of the currently enrolled with those of their older siblings who participated in the program at an earlier time period when they were of corresponding age group. Effects of land operated by households. Long term household economic status influences total resources available to the household, including food. It also affects the capacity to cater for periodic income fluctuations due to rainfall and other seasonal factors, by influencing household capacity to store food stocks in good times and to borrow in bad times, thus, smoothing consumption flows over time. To the extent greater resource 124 availability at the household level results in greater entitlements for preschoolers as well, a positive effect on weight is expected. A clear positive association between long term household economic status as measured by the type of land operated and weight outcomes is observed for weight and standard weight. As compared with children from landless households (the omitted category), children from both the landed categories show positive and significant coefficients indicating higher intercepts for base, sibling and current data. Tables 7.1 to 1 3 contain coefficient estimates and the corresponding t-values. Further, for sibling and current data the positive effects of land operated are the strongest for the best land category - land with assured irrigation. Thus, there are intercept differences between the weight and standard weight of children from households of varying economic status, the differences between the landed and landless categories being the most significant. A comparison of least square mean values of standard weight based on the model in table 7.4 is carried out in table 7.6. For base line data we see that the mean standard weights for the two landed categories, while not significantly different from each other (p = 0.27 for a two tailed t-test2 testing for equality of means carried out in columns 6 - 8), are both significantly higher than the landless category. The means are 0.712, 0.723 and 0.719 for children of the landless, those operating land with uncertain irrigation and those operating land with certain irrigation. A similar pattern is observed for current data, though the levels of the mean values are higher than the base levels. Starting with children from landless households, the means are 0.720, 0.742 and 0.746 for the three land categories. For sibling data mean standard weights for the three land categories, 2 Since higher weight outcomes for better landed categories are expected, a one tailed t- test would also suffice. This would result in lower p-values, increasing the significance of the differences. 125 starting from the landless, are 0.704, 0.724 and 0.730. Columns 6 to 8 of table 7.6 contain the p-values at which the null hypothesis of equality of least square means across the 3 levels of land operated is tested. The p-values indicate significant differences in standard weight outcomes when a comparison between any of the landed categories versus the landless is carried out. For base and current data the two landed categories are much less different from each other, but in both cases are significantly higher than the landless. Similar conclusions result from a comparison of least square mean weights based on tables 7.1 and 7.3, and of standard weights based on table 1 2 and 7.5, but are not reported. Thus, child health and nutritional outcomes measured by weight and standard weight do vary with household economic status captured by land operated. Some land is better than none in all cases, and, at least for sibling and current data, land with assured irrigation is the best of all for weight outcomes. We also observe that over time the levels of standard weights have risen for each land category. Current standard weights are higher than those of their older siblings when the siblings were of the corresponding ages. Current standard weights are also higher than the base line levels. In some of the earlier studies household economic status has not always been found to make such a significant difference. In the Horton (1988) study with Bicol data total assets were found positively related to height-for-age but the effect on weight- for-height was insignificant. In Barrera (1987) household income (excluding mother’s) was insignificant in explaining height-for-age for children less than 10 years of age, and especially the preschool age group where maternal characteristics appear more important (education and height). In Strauss (1990), however, land per adult was found positive and significant for child weight outcomes though not for child height. With Indian data from 126 West Bengal Sen and Sengupta (1983) do find that children from landed families exhibit a lower ’undernourishment index’ calculated on the basis of weight-for-age as compared with those from the landless. Effects of maternal education. Mother’s education can influence the kind of care the child receives. Better educated mothers may increase household resources through selective mating with wealthier men or by themselves. They are likely not only to have more information, but also to seek out more information about health care, making them more efficient at producing child health outcomes. Especially under circumstances where environmental sanitation and protected water supply cannot be taken for granted, maternal education may be important in reducing disease loads that children have to bear. Better educated mothers are more likely to boil drinking water at least during diarrheal episodes, be more careful about immunization, etc. In the context of the program, maternal education could promote the adoption of practices suggested by the program.3 Results in tables 7.1 to 7.5 show the positive and significant effects of maternal education on weight. As compared with children of mothers with no schooling (called ’ illiterate mothers’), those of mothers with upto 5 years schooling and more, have positive and highly significant intercepts in 5 out of 6 cases. As explained in the section on variables, two dummies were constructed for mother’s years of schooling to capture the effects of maternal education: one for those with some but only upto 5 years of schooling and another for those with more than 5 years of schooling illiterate mothers with no schooling being the omitted category. 3 Of course, the program itself has an educational component, which may act as a substitute for formal schooling. 127 For base line data, compared with children of illiterate mothers with no schooling, those of mothers with more than 5 years schooling have significantly higher intercepts. But children of mothers with schooling of upto 5 years are not significantly different, with the coefficient being negative and insignificant, especially for standardized weights. Thus, in the early months of the program a little schooling is not different from no schooling. More than 5 years of schooling is required for significant improvements in standard weight. As the program is better established, children of mothers with even some education begin to show significant improvements over the children of illiterate mothers. For sibling data both education levels show significantly higher intercepts as compared with the omitted category. The same is true for current data with the coefficient for the highest education level being the strongest and most significant. Mother’s education level does seem to make a significant difference to child weight outcomes. Moreover, comparing the coefficient magnitudes for current data with the base, over time the importance of education seems to have increased. For both education levels the coefficients for current data are larger than those for the base. Comparing the coefficient magnitudes for current as against sibling data, stronger magnitudes are observed for more than 5 years schooling. Wolfe and Behrman in their 1987 study have argued that effects of maternal education derived from reduced form equations may be overstated unless family background variables are kept constant. To do this they use information on schooling of mother’s sibling and her parents and find that the effect of maternal schooling gets eliminated. However, Horton (1986 and 1988) using data from the Philippines and estimating reduced form equations for height and weight-for-height finds dear effects of parent’s education. So do Barrera (1987) using the Bicol (Philippines) data, and Thomas, Strauss and Henriques (1987) using evidence from Brazil. Strauss (1990) for rural Cote 128 d’ Ivoire finds a stronger effect of education on weight-for-height than for height. In their work mother’s height is controlled for and this helps to capture unobserved genetic and non-genetic (e.g., dietary) family background characteristics. Mother’s height is included in the present specification also. An examination of the least square mean standardized weight-for-age scores (table 7.7) by levels of mother’s education is instructive. Column 3 contains the means for each level of mother’s education. Columns 6 - 8 test the null hypothesis of equality of pairwise means across the three levels. For the base line data level 1, more than 5 years of schooling, has the highest standard weight (0.732 as compared with 0.710 and 0.713 for the two lower levels). The other two levels are not significantly different from each other indicating that more than 5 years of schooling are required before weight differences become significant. For sibling data even some education seems to make a difference. Standard weights for children of mothers with education levels 1 and 2 are significantly higher than those for level 3 (0.726 and 0.724 as compared with 0.707 for children of illiterate mothers). For current data once again means are highest for level 1, next for level 2 and the lowest for level 3 (0.753, 0.736 and 0.720 for levels 1, 2 and 3 respectively), each very significantly different from the other. This could be became literate mothers may be more likely to derive benefits from program operation by utilizing more efficiently the information they learn from the program. Thus, maternal education as measured by years of formal schooling does make a difference for child weight-for-age outcomes, higher levels being associated with higher means. Initially, for base line data, some education (upto 5 years) does not make a significant difference, but more than 5 years schooling does. For sibling and current data, 129 however, even some education is associated with higher mean standard weights. For current data, in addition, more than 5 years schooling is associated with the highest means. The effects of time can be observed by comparing current mean standard weights for each level of maternal schooling with corresponding mean weights for base and sibling data. Means based on current data are higher than those for sibling and base data for each level of maternal education indicating improvements in standard weights for all groups. The differences between each group are highly significant for current data, but not for base and sibling data. For the base some education is not significantly different from no education, with only the highest maternal education category making a significant difference. For sibling data the two education categories are both associated with higher mean standard weights, but are not significantly different from each other. Education levels seem to make more of a difference over time probably due to the greater ability of more educated mothers taking advantage of program benefits. Effects of sex. Gender preference has been observed in certain societies. Especially in south Asia, a bias against females has been reported for IMR and in the overall sex ratio of the population resulting from differential mortality in the younger age groups (Visaria, 1971; Visaria and Visaria, 1981; Bardhan, 1984; Mazumdar, 1991). In the case of anthropometric outcomes the prevalence of lower values for female children has not been well established. For data from West Bengal a sex bias against females has been observed (Sen, 1984; Sen and Sengupta, 1983). But the significance of some of these differences has been questioned (Kakwani, 1986). Outside of Asia, however, it appears that differences between the sexes are small and often insignificant; see, for example, results from Africa (Strauss, 1988; Svedberg, 1988) and Latin America (Schofield, 1979; Thomas, et al., 1988). Attempts to measure differences by sex in the intrahousehold distribution of 130 nutrients suggest boys tend to be favored (Behrman and Deolalikar, 1988, for India; Senaur et al., 1988, for the Philippines). But there is little evidence of difference between the sexes in the allocation of expenditures in the Cote d’ Ivoire (Deaton, 1987) and the US (Gronau, 1985 & 1988). For the TINP data sex was found to be a significant regressor in determining child weight outcomes. The regressions for weight in tables 7.1 and 7.3 with a dummy for females have highly significant and negative coefficients showing lower weights for female children as compared with males. By itself this cannot, however, be taken as an indicator of sex bias since the NCHS standards also have lower median weights for females as compared with males for each monthly weight. To examine the question of sex bias we look at regressions where the age-sex standardized weight scores are used as the dependent variable (tables 7.2, and 7.4). Sex bias for anthropometric outcomes can only be identified relative to another population (in this case the NCHS median weights) since standardizations based on the sample preclude tests for bias4. Negative and significant coefficients are observed for sex (females) for base and sibling data. This shows that female preschoolers are further away from their age-sex specific NCHS median than male preschoolers are from theirs. However, for current data the sign is reversed without any decline in the statistical significance of the coefficient. Comparing base with current the absolute t-value in fact goes up from 2.94 to 7.10 in table 7.2 and from 3.05 to 7.04 in table 7.4. Similarly, comparing siblings with current the absolute t-value goes up from 4.72 to 7.10 in table 7.2 and from 4.68 to 7.04 in 4 If there is a systematic sex bias in the reference population but not in the sample being studied, then standardizations could impart a spurious bias (Thomas, 1989). 131 table 7.4. This indicates a reduction in absolute difference between male and female children in the current data5. A comparison of least square means for weights and standard weights is carried out in table 7.8. Female least square mean weights are significantly lower than male in all three situations - base, sibling and current. The lower mean weights for base and sibling data are due to their lower mean ages. Age-sex independent comparisons of least square mean standard weight show the significantly lower standard weights for base and sibling data and a rejection of the null hypothesis of equality of means by sex. For base and sibling data respectively, p = 0.0023 and 0.0001 for a 2-tailed t-test, strongly rejecting the null hypothesis of equality of means by sex. The means for standard weights are 0.715 for females and 0.721 for males in the base, 0.715 and 0.724 for females and males respectively in the case of siblings. Thus for females weight is significantly less than their NCHS standard as compared with males. There is a significant reduction in this male-female difference over time with female standard weight being higher than male. The values in the current data for females and males respectively were 0.739 and 0.727 respectively. Perhaps this reduction in male-female differences over time can be partly explained due to the negative selection for supplementary feeding by the program. That is, while all children are weighed every month and receive health care inputs (eg., dewarming, vitamin A administration, treatment for diarrhoea, immunization etc.) a daily supplementary food is given only to the worst off cases. These worst off cases consist of children identified as in grades III and IV of malnutrition or in whose cases growth 5 Whether this reversal in sign indicates a reversal of sex differences is not automatically evident. Even though the coefficient for being female is positive and significant for current data on standard weight, there is the question of the appropriateness of NCHS standards. 132 faltering is observed on the basis of inadequate weight gain over successive months (less than 300 grams over 3 months for children upto 12 months of age, and over 4 months for children older than 12 months) regardless of the grade. To the extent female children tend to belong to the worst off categories they would receive the extra input of supplementary feeding more than proportionately, partly offsetting sex differences. This is in fact observed in the data: in the sample of all the children identified by the program as belonging to grades HI and IV, over 65% are females. It will be seen in chapter V m that this conclusion about the existence of a sex differences gets modified when interactions between sex and land and between sex and caste (whether SC) are allowed and households are separated by the different categories of land operated or by whether or not they belong to the SC group. Tentative explanations for the possible causes are also given there. Once again, over time comparisons show improvements in standard weights. Higher standard weights are observed for both male and female children in the case of current data as compared with base and sibling data. Effects of caste. Data showed the prevalence of 13 major caste groups of which 3 were scheduled castes (SC). Since the most important difference was expected between the scheduled caste households versus the others, a dummy for not belonging to one of the scheduled castes was created. In the Indian context, and especially in rural areas, belonging to an SC category can be considered as evidence of lower social status as compared with other households. Historically they have been considered as untouchable and as being outside the four tier ’ vama’ system. Even today in rural areas they reside a little away from the main dwelling area in a village. Economically also, they tend to be deprived, largely engaged as agricultural laborers or other less skilled operations. 133 The dummy for not belonging to the SC group had positive coefficients for base, sibling and current data (tables 7.1 to 7.4). Thus, not belonging to the SC category is beneficial, resulting in higher weight and standard weight outcomes. The t- values for standard weights are 1.92, 4.79 and 3.06 for base, sibling and current data respectively in table 7.2 and 1.87, 4.78 and 2.96 respectively in table 7.4. Comparing coefficient magnitudes over time we observe that the effect of not belonging to the SC group becomes stronger among siblings and then is reduced for current data. The gap between children from the SC and non-SC households for current data is a little less than the gap for base data. A comparison of least square mean standard weights in table 7.9 shows that not belonging to the SC category is always associated with higher standard weights as compared with children from SC households. Over time there is an increase in standard weights for both the SC and non-SC categories for current data as compared with the base and with sibling data. Thus, both categories have improved over time, but the gap is still significant, though it is less for current data as compared with siblings. Again, negative program selection for supplementary feeding could partly explain this, and partly it could be a trend effect. Effects of household structure. The intention here was to capture the effects of at least one of the child’s parents being the household head. The presumption was that if the head is some relative other than a parent, like an uncle or a grandparent, this could have a negative impact upon child weight. If parents have the greatest interest in their child’ s growth as compared with other relatives, and to the extent decision making by the household head and access to household resources affects child weight, being the child of the household head can be advantageous. 134 A household dummy for whether father is the head of household was created. A dummy for either parent (mother or father) being the head was not created since the intention was to look for a positive effect and households with mothers as heads are more likely to face negative influences of husband’s death or desertion, with consequent reduction in economic and social status. The coefficients of father not being the head turned out to be positive and significant in all the three regressions for the base, sibling and current data. This was contrary to what had been initially expected. It turns out that nc> t being the child of the household head is advantageous with respect to weight outcomes. While initially puzzling, this result is consistent with an alternate explanation. A closer scrutiny of the data showed that for the present sample not being the child of the household head implies that the child is likely to come from an extended family. Joint families are more likely to be landed and less likely to belong to SC households. Pooling of resources in such households is much more prevalent and feasible as compared with nuclear families where a parent is almost invariably the household head. This is likely to be particularly beneficial in rural areas where there can be considerable variability in production, availability of unskilled and other work, and earnings between good and bad seasons. Such a pooling of resources could improve household economic status and lessen the variability in incomes over time, offsetting the possible advantages of being the child of the household head. The result obtained indicates that belonging to an extended family is beneficial with respect to weight outcomes. This conclusion appears to be the opposite of the finding in Strauss (1990) for data from Cote d’ Ivoire where being the child of a senior or junior wife was 135 associated with higher weight outcomes as compared with not being the child of the household head. The advantage of belonging to an extended family is reduced over time as is evident from the smaller magnitudes of the coefficients and smaller t-values for current data as compared with the sibling and base data. A comparison of least square mean standard weights in table 7.10 shows that for both categories there is an improvement over time with current means being greater than those for siblings and the base, and the gap between the two categories being less among the currently enrolled children as compared with the gap observed for base or sibling data. This could be due partly to a trend effect over time and partly due to the operation of the program. Effects of development block. The intention was to examine whether the immediate macro economy in which the household resides, i.e., development block, is systematically related to weight outcomes. Blocks Kottampatti, Chellampatti (from Madurai district), Kariapatti and Sivagangai (from Ramnad district) are assigned levels 1, 2, 3, and 4 respectively. Dummies are created for the first three levels, Sivagangai being the omitted category. Blocks 1 and 3 were considered economically less developed than 2 and 4 (see chapter IV for details). If block level economic development were to be systematically related to weight outcomes, then one would expect lower standard weights for block levels 1 and 3 as compared with the omitted block, Sivagangai (SIV), assigned level 4, and not very significantly different standard weights for block level 2 as compared with level 4. The results obtained (tables 7.1 to 7.4) were not consistent with such an expectation. For the base data the three blocks with levels 1, 2 and 3 showed significantly higher intercepts than the omitted category as seen from the positive and significant \ ) I 136 coefficients. Blocks 1 and 2 were almost identical, the coefficients in the standard weight specifications being 0.1086 and 0.1089 in table 7.2 and 0.1092 and 0.1091 in table 7.4. For sibling data blocks 1 and 2 have significantly higher intercepts while block 3 has a significantly lower intercept. This again does not match levels of economic development. For current data, again, significantly higher intercepts are observed for blocks 1 and 2. For block 3 the intercept, while higher than that of the omitted block, is not significantly so. As observed for sibling data the two Madurai district blocks show higher standard weights as compared with the two blocks from Ramnad district. Least square means for each block for base, current and sibling data are compared in table 7.11 and they indicate a similar pattern. Except for the base data where block 3 has the highest means, overall blocks 1 and 2 tend to show higher standard weights than blocks 3 and 4. Thus the block level index of economic development does not systematically explain child weight outcomes as does the indicator of household economic status and as do other household and individual variables. It is possible that blocks may differ not only by economic but also by social characteristics, which could influence weight outcomes. In order to check this the block-wise scheduled caste composition of the children was examined. Overall, the percentage of children coming from scheduled caste households varied between 17% in block 4 to 28% in block 2. The scheduled caste composition did not systematically explain weight variations across blocks. Perhaps a more detailed analysis of percentages belonging to other castes as well would be instructive, but was not attempted. Three points seem to emerge in making comparisons across blocks: (a) District rather than a block level index of economic development separates standard weights better, especially for sibling and current data, the two blocks in _ ---------------------------------------i3 7 Madurai district showing higher weights as compared with the blocks in Ramnad district. This is also consistent with duration of program operation: blocks belonging to the early phases I and II doing better than phase ID. (b) Differences between the blocks are less for current data as compared with sibling and base data indicating that variations across blocks decrease over time. (c) Overall levels of mean standard weight tend to be higher for the current data as compared with base and sibling data. The exceptions are Kariapatti block for base data and Kottampatti block for sibling data. Thus our initial hypothesis of blocks differing in weight by an overall index of development is not supported by the data. It is of interest to see that a higher level of regional aggregation, i.e., the district level, seems to separate weight outcomes better. Also, the duration of program operation appears to be a relevant explanatory variable. Phases I and II (the early program phases) of Madurai district have higher intercepts than the later phase III blocks of Ramnad district. Effects of duration of program operation. The effects of duration were explored further by using the difference between the year in which a child was weighed and the year of introduction of the program in each village as one of the regressors. The regression results showed that longer duration was positively and significantly associated with weight and standard weight. This was true for all children indicating that program operation had a positive influence on weight outcomes. Over time the magnitudes of these effects and their significance became smaller. The base line coefficient estimates for weight and standard weight were 0.264 (t = 12.52) and 0.020 (t = 11.93) respectively, whereas those for the currently enrolled children were 0.035 (t = 3.15) and 0.003 (t = 2.90), respectively. This could indicate that although 138 duration continued to be positive and significant, after 6 to 8 1/2 years of program operation in the various villages, differences in duration (which were around two years) mattered less. Perhaps, after over a decade of program operation in all areas, small Variations in duration may not make a significant difference to weight outcomes. Effects of water source. Whether the water source used for drinking is protected or not can influence the incidence and severity of intestinal infections. Data from , i 1 Latin America show that diarrheal diseases are by the major cause of death from infections in children under five (Puffer and Serrano, 1973). Water is an important source by which intestinal infections can be introduced. WTiile infections and physical growth are ! not significantly related in DCs, data from LDCs indicate a very significant relationship j between common childhood ailments, especially diarrhea, and growth (Martorell and J ! Yarbrough, 1983). Possible causes are the lighter disease load experienced by children from DCs making it hard to detect effects on physical growth, and the ample nutritional resources available before, during and after illness, enabling children to quickly make up any loss in weights. The mechanisms by which infections affect growth (like loss of appetite, nutrient losses, malabsorption by the body, etc.) have already been discussed in chapter n. I I Hence, better quality water can be expected to be positively associated | with higher weights and standard weights. As explained in chapter VI, a dummy for relatively hygienic water was created (when the water source was a deep bore well or hand pump), with other less hygienic sources being the omitted category (open tanks and wells). ' l i I I 139 Separate regressions for current data alone were run with a water source dummy, and a dummy for non-existence of a pucca road connecting the village6. In the specifications with water and road dummies, the variable ’development block’ was omitted since water supply and rural road works are taken up blockwise. The regression for standard weight is reported in table 73; results from the weight regression were largely the same and are not reported. As compared with the omitted category, the dummy for hygienic water has a positive and very significant coefficient (t = 5.15) indicating the importance of water quality for weights. The least square mean standard weight for better water is 0.737 as compared with 0.726 for poorer water, the null hypothesis of equality of means being rejected at a p = 0.0001 (table 8.12). Thus, type of water is significant in explaining child weight outcomes. Effects of road. Whether or not the village is connected by a pucca road indicates not only the ease with which it might integrate with the rest of the economy and its level of development, but also the overall level of services available in the area. In some ways it might better capture the level of development than the dummy for block. The village not being connected by a pucca road is found to be negatively associated with weight and standard weight, though regressions for standard weight alone have been reported (table 73). The regression coefficient is negative and very significant (t = -7.75). Differences in the least square mean standard weight between the two groups are found significant at p = 0.0001 in table 7.12 6 Current water supply and road conditions were not applicable to base and sibling data as a number of new water connections and roads had recently been established. 140 12. Effects of Other Covariates. In this section weight variations associated with the age of the child, birth order, mother’s height and two distance variables are explored. Unlike group or class effects of categorical variables examined above, the explanatory variables of concern in this section are treated as continuous. Tables 7.1 to 7.4 contain the regression results. Effects of age. Effects of age are different for weight and standard weight. Consider the dependent variable weight first. From table 7.1 it is seen that the coefficients for age are positive and very significant whereas those for square of age are ' negative and significant indicating a concavity. Weight increases with age but at a decreasing rate. The effects for base, sibling and current data are similar. For standardized weight age and age squared continue to be very significant but with signs reversed (table ! 7.2). This suggests that there is a lag in weight growth relative to the standard which gets j j wider with age, and this happens at an increasing rate during the 6-36 months period of ; a child’s life. For standard weight regressions the magnitudes of the age coefficients are j i higher for current data as compared with, both, base and sibling data. To investigate further the effects of aging on child growth, child ages are i divided into 6 age group levels: Age in Months Age Group Level AGE < = 9 1 9 < AGE < = 12 2 12 < AGE < = 18 3 18 < AGE < = 24 4 24 < AGE < = 30 5 30 < AGE 6 Dummies are created for levels 1 to 5, age greater than 30 months being the omitted category. The regressions for weight and standard weight are reestimated using age group dummies instead of the age and the square of age terms. The results are available in tables 73 and 7.4. Direct comparisons across age groups are possible with standardized weight-for-age scores. Hence we examine the coefficients of age group dummies in the standard weight specification (table 7.4) and also look at the corresponding least square means of each age group for base, sibling and current data (tables 7.13 a & b). A U-shaped growth pattern is observed for base, sibling and current data. Standardized weight scores bottomed out at level 3 of the AGE dummy, i.e., lowest standard weights are observed for children in the age group 12 to 18 months. The decline in growth relative to the NCHS median starts right away at 6 months of age, standard weights being highest at level 1 of the age dummy (between 6 to 9 months of age), with the sharpest decline thereafter between levels 1 and 2. For base data level 1 is not different from the omitted category of level 6. The decline continues between levels 2 and 3, but the slope is flatter, reaching its lowest at level 3 (between the ages of 12 to 18 months). After that weight picks up relative to the standard, increasing continuously until the child is 36 months old and the weight observations stop. The U-shaped growth pattern is typical of many developing countries, and has in the literature been related to growth faltering associated with the introduction of solid foods and the termination of breast feeding. The difference here is that the upward trend starts earlier (between 12 to 18 months of age) than reported in other studies where the bottoming out appears to occur around 2 years7. This could be due to the support available from the program. 7 For heights the bottoming out has been reported to occur later, around 4 years (Barerra, 1987; Svedberg 1987; Strauss, 1990). 142 The general growth pattern is similar for base, sibling and current data with one significant difference. Current levels appear significantly higher than those for base and sibling data for each age group. A graph plotting mean standard weights from tables 7.13 a and b against age groups for base, sibling and current data shows this clearly (figure 7.1). Over time there has been an improvement in standard weights. This is particularly significant when we compare current and sibling data. Current data contains weights of later bom children as compared with sibling data. Household and community characteristics for siblings are the same as those for current children from the same households. Individual characteristics (age, sex, birth order) are different. But age has been controlled for and birth order is expected to have an adverse effect upon current standard weights. To the extent birth order has a negative impact upon standard weights, a stronger negative influence is expected to be operating upon current data as compared with sibling data. Yet current standard weights are higher for each age group. This could be because of the operation of the program. Birth order. Negative and highly significant coefficients for birth order were obtained for base, sibling and current data. Higher birth order children tend to have lower weights and standard weights. Similar results have been obtained in Horton (1988). As pointed out earlier in the section on variables, empirical evidence for inequalities in resource allocation by birth order for resources such as schooling and educational attainments among siblings exists (eg., Lindert, 1978; Birdsall, 1979; King and Lillard, 1983; Behrman and Taubman, 1986). This could be due to greater strain on household resources of time and purchasing power. In the case of nutritional status a higher proportion of children in the household could adversely affect nutritional status due to the increased probability of infectious episodes. In addition, a biological factor pointed 143 out in Horton (1988) is maternal depletion. Later born children are bom to older mothers and tend to be of lower birth weights. She has also referred to cultural and socioeconomic factors which can cause a tendency among parents to favor earlier birth orders. For example, if the oldest son is important in funeral rites or if parents plan in their old age to depend upon their oldest children who become economically independent the earliest. To the extent resources available to a household are a limiting factor causing the negative relationship, better household economic status should counter the effect of birth order. Similarly, apart from economic strain, to the extent, in addition, maternal depletion or a greater proportion of children in the household leading to greater probabilities of infections cause the negative relationship, maternal education should counter the effects of birth order since better educated mothers are likely to be relatively more aware of the problems, more receptive in adopting new ideas and practices made available by the program and more efficient at producing child health and nutrition outcomes. Interaction effects between birth order and land operated by the household and between birth order and mother’s education level are examined in chapter VQI. Mother’s height. This was another significant predictor with positive coefficients for weights and standard weights, statistically highly significant for base, sibling and current data. Mother’s height is used as a prosy for her genetic and health endowments. As pointed out earlier it is also likely to pick up other family background characteristics not captured by the education variable. The distance variables. Distances to the nearest primary health care facility and general hospital have negative coefficients (not all significant) for weights and standard weights. This indicates that distance probably contributes to underutilization, with adverse effects upon weights. The coefficients are significant in 4 out of 6 cases. 144 13 Conclusion. The individual, household and communal level variables selected as regressors for child weights above can be thought of as the basic or underlying determinants (as against proximate factors, many of which interact with health, nutritional status and growth). The variables used as regressors were chosen such that the direction of causation would run from them to weights, and not the other way round. Not all effects turned out to be what was initially expected (eg., the effect of father being the household head and the significant reduction of sex differences for current data). Over time comparisons were possible because of the availability of base line and sibling data, in addition to current. Increases in weight and standard weight over the period 1980 to 1989 are notable. The existence of the program could be contributing to improvements over time. This is also supported by the significant positive influence of program duration on weight outcomes. i j STANDARD WEIGHT SC O R E S r Figure 7.1 LEAST SQUARED MEAN STANDARD WEIGHTS Base, Sibling & Current Data 0.76 0.75 - 0.74 0.73 0.72 - 0.71 0.7 - < A G E <= 12 12 < A G E <= 18 18 < A G E <= 24 24 < A G E <= 30 3 0 < A G E □ B A SE D A T A A G E GROUPS. M O N T H S 4- SIB D A T A C U R R E N T D A T A T a b le 7 . 1 146 REDUCED FORM REGRESSIONS FOR WEIGHT: BASE, SIBLING & CURRENT DATA Regressors Base Sibling Current 1. Intercept 2.4111 (7.38) -0.7232 (-2.57) 3.1856 (13.75) 2 . Child age (months) 0.1808 (17.78) 0.1881 (27.76) 0.1756 (29.63) 3 . Child age squared (months) -0.00061 (-2.99) -0.00074 (-4.82) -0.00059 1 (-3.89) 4 . Birth order -0.0861 (-10.31) -0.1229 (-14.29) -0.0928 (-14.98) 5. Sex (Male = 0) -0.5630 (-20.76) -0.5173 (-23.17) -0.3575 (-19.46) 6. Mother's height (cms) 0.01756 (8.69) 0.0359 (20.03) 0.0132 (8 .92) 7 . Mother's education (Illiterate = 0) i * Upto 5 years schooling -0.0428 (-1.33) 0.1844 (6.85) 0.1673 (7.24) * More than 5 years schooling 0.2177 (4.78) 0.1764 (4.69) 0.3406 (10.47) 8. Land operated (No land = 0) * Land with uncertain irrigation 0.1282 (4.10) 0.2254 (9.02) 0.2402 (11.56) * Land with certain irrigation 0.1034 (2.21) 0.2880 (7.48) 0.2702 (8.10) 9 . Whether scheduled caste (Yes = 0) 0.0634 (1.71) 0.1175 (4.50) 0.0719 (3.30) 10. Whether father is HH head (Yes = 0) 0.3154 (8.83) 0.1773 (5.89) 0.0935 (3.95) cont... Table 7.1 cont... 147 . Distance to primary -0.0014 -0.0055 -0.0120 health center (kms.) (-0.58) (-2.79) (-7.38) . Distance to general -0.0324 -0.0006 -0.0076 hospital (kms.) (-13.51) (-0.56) (-4.38) . Development block (Siv = 0) * Kot 0.2307 0.2747 0.0875 (5.07) (8.13) (3.09) * Chel 0.2352 0.1087 0.2344 (5.37) (3.09) (8.15) * Kari 1.1013 -0.3087 0.1195 (15.25) (-4.88) (2.37) N 6671 7002 9156 R-squared 0.5785 0.7117 0.6671 F-statistic 570.71 1077.64 1144.53 Mean squared error 1.1119 0.8241 0.7462 148 T a b le 7 . 2 REDUCED FORM REGRESSIONS FOR STANDARDIZED WEIGHT: BASE, SIBLING & CURRENT DATA Regressors Base Siblings Current 1. Intercept 0.5018 (19.2) 0.2909 (11.81) 0.6135 (28.63) 2. Child age (months) -0.0014 (-1.72) -0.0049 (-8.31) -0.0093 (-16.90) 3. Child age squared (months) 6.00668E—05 (3.57) 0.00013 (9.70) 0.00023 (16.25) 4. Birth order -0.0069 (-10.42) -0.1239 (-16.47) -0.0086 (-15.08) 5. Sex (Male = 0) -0.0064 (-2.94) -0.0092 (-4.72) 0.0121 (7.10) 6. Mother's height (cms) 0.0014 (8.67) 0.0029 (18.73) 0.0013 (9.26) 7. Mother's education (Illiterate = 0) * Upto 5 years schooling -0.0028 (-1.10) 0.0172 (7.31) 0.0154 (7.21) * More than 5 years schooling 0.0191 (5.24) 0.0196 (5.95) 0.0326 (10.82) 8. Land operated (No land = 0) * Land with uncertain irrigation 0.0106 (4.24) 0.0200 (9.15) 0.0217 (11.30) * Land with certain irrigation 0.0068 (1.81) 0.0265 (7.85) 0.0248 (8.05) 9. Whether scheduled caste (Yes = 0) 0.0057 (1.92) 0.0109 (4.79) 0.0062 (3.06) 10. Whether father is HH 0.0249 head (Yes = 0) (8.75) 0.0158 (5.98) 0.0077 (3.50) cont... 149 T a b l e - 7 . 2 c o n t . . . 11. Distance to primary -3.71900E-05 health center (kms.) (-0.19) 12. Distance to general -0.0026 hospital (kms.) (-13.34) 13. Development block (Siv = 0) -0.00061 (-3.57) -4.894 74E-05 (-0.42) - 0.0011 (-7.37) •0.00072 (-4.50) * Hot * Chel * Kari 0.0186 (5.13) 0.0189 (5.42) 0.0872 (15.12 0.0231 (7.83) 0.0086 (2.78) -0.0302 (-5.45) 0.0062 (2.36) 0.0160 ( 6 . 01 ) 0.0051 (1 . 10 ) N 6658 R-squared 0.1169 F-statistic 54.94 Mean squared error 0.0071 7002 0.1651 86. 32 0.0063 9156 0.1470 98.45 0.0064 Table 7.3 150 REDUCED FORM REGRESSIONS FOR WEIGHT WITH DUMMIES FOR AGE GROUPS: BASE, SIBS AND CURRENT Regressors Base Siblings Current 1. 2 . Intercept Child age (Age > 30 months = 0) 7.8002 (24.59) 4.7675 (16.75) 8.3489 (35.33) k 6 months<age<=9 months -4.0423 (-42.98) -4.1274 (-95.26) -3.9396 (-89.96) k 9 months<age<=12 months -3.4446 (-46.46) -3.4780 (-77.01) -3.3094 (-73.38) *12 months<age<=18 months -2.7339 (-61.96) -2 .7260 (-76.92) -2.5825 (-62.94) *18 months<age<=24 months -1.8581 (-49.72) -1.8046 (-51.76) -1.7256 (-41.44) *24 months<age<=3 0 months -0.9429 (-27.81) -0.8940 (-26.01) -0.9056 (-20.68) 3. Birth order -0.0886 (-10.35) -0.1219 (-13.72) -0.0936 (-14.60) 4 . Sex (Male = 0) -0.5735 (-20.61) -0.5190 (-22.49) -0.3584 (-18.86) 5. Mother's height (eras) 0.0178 (8.59) 0.0359 (19.36) 0.0134 (8.76) 6. Mother's education (Illiterate = 0) * Upto 5 years schooling -0.0504 (-1.43) 0.1816 (6.52) 0.1643 (6.88) * More than 5 years schooling 0.1944 (4.16) 0.1773 (4.56) 0.3401 (10.10) 7. Land operated (No land = 0) * Land with uncertain irrigation 0.1427 (4.45) 0.2265 (8.76) 0.2397 (11.15) k Land with certain irrigation 0.1133 (2.36) 0. 2876 (7.22) 0.2759 (7.99) cont... i 151 T a b le 7 . 3 c o n t . . . 8 . Whether scheduled caste (Yes = 0) 0.0697 (1.83) 0.1161 (4.30) 0.0699 (3.10) 9. Whether father is HH head (Yes = 0) 0.3110 (8.49) 0.1725 (5.54) 0.0911 (3.72) 1 0 . Distance to primary health center (kms.) -0.00041 (-0.17) -0.0052 (-2.57) -0.0118 (-7.08) 11. Distance to general hospital (kms.) -0.0329 (-13.37) -0.00055 (-0.54) -0.0076 (-4.25) 12 . Development block (Siv = 0) * Kot 0.2668 (5.72) 0.2725 (7.80) 0.0935 (3.19) * Chel 0.2380 (5.30) 0.1032 (2.84) 0.2430 (8.17) * Kari 1.1150 (15.40) -0.3021 (-4.62) 0.1306 (2.50) N R-squared F-statistic Mean squared error 6671 0.5564 439.14 1.1705 7002 0.6920 825.57 0.8808 9156 0.6440 869.81 0.7982 T a b le 7 . 4 152 REDUCED FORM REGRESSIONS FOR STANDARDIZED WEIGHT WITH AGE GROUP DUMMIES: BASE, SIB S AND CURRENT Regressors Base Sibling Current 1. 2. Intercept Child age (Age > 3 0 months = 0) 0.5189 (21.06) 0.2699 (11.23) 0.5481 (25.92) * 6 months<age<=9 months -0.0039 (-0.54) 0.0011 (0.29) 0.0148 (3.78) * 9 months<age<=12 months -0.0313 (-5.44) -0.0295 (-7.74) -0.0202 (-5.01) *12 months<age<=18 months -0.0356 (-10.39) -0.1326 (-10.88) -0.0262 (-7.15) *18 months<age<=24 months -0.0269 (-9.25) -0.0218 (-7.39) -0.0203 (-5.44) *24 months<age<=30 months -0.0132 (-5.02) -0.0110 (-3.80) -0.0168 (-4.29) 3. Birth order -0.0070 (-10.47) -0.0124 (-16.46) -0.0087 (-15.13) 4 . Sex (Male = 0) -0.0066 (-3.05) -0.0091 (-4.68) 0.0120 (7.04) 5. Mother's height (cms) 0.0014 (8.89) 0.0029 (18.85) 0.0013 (9.26) 6. Mother's education (Illiterate = 0) * Upto 5 years schooling -0.0030 (-1.10) 0.0171 (7.29) 0.0155 (7.24) * More than 5 years schooling 0.0187 (5.15) 0.0194 (5.92) 0.0324 (10.76) 7 . Land operated (No land = 0) * Land with uncertain irrigation 0.0105 (4.24) 0.0203 (9.29) 0.0218 (11.35) * Land with certain irrigation 0.0067 (1.79) 0.0266 (7.92) 0.0252 (8.15) cont... 153 Table 7.4 cont... . Whether scheduled 0.0055 0.0109 0.0060 caste {Yes = 0) (1.87) (4.78) (2.96) . Whether father is HH 0.0251 0.0157 0.0078 head (Yes = 0) (8.82) (5.98) (3.54) . Distance to primary —1.84976E--05 -0.00062 -0.0011 health center (kms.) (-0.10) (-3.60) (-7.37) . Distance to general -0.0026 -4.87 673E-05 0.00070 hospital (kms.) (-13.53) (-0.44) (-4.38) . Development block (Siv = 0) * Kot 0.0192 0.0230 0.0060 (5.30) (7.80) (2.30) * Chel 0.0191 0.0085 0.0160 (5.47) (2.77) (6.02) * Kari 0.0880 -0.0299 0.0049 (15.29) (-5.43) (1.05) N 6658 7002 9156 R-squared 0.1210 0.1700 0.1465 F-statistic 48. 07 75 . 30 82.52 Mean squared error 0.0071 0.0063 0.0064 . T a b le 7 . 5 154 REDUCED FORM REGRESSIONS FOR WEIGHT AND STANDARDIZED WEIGHT WITH DUMMIES FOR WATER & PUCCA ROAD (Current data) Regressors Weight Standardized Weight 1. Intercept 2. Child age (Age > 30 months = 0) * 6 months<age<=9 months 8 . 286 (35.24) -3.9339 (-89.93) 0.5429 (25.83) -0.0155 (3.96) * 9 months<age<=12 months -3.2994 (-72.23) -0.0193 (-4.79) *12 months<age<=18 months -2.5735 (-62.79) -0.0254 (-6.94) *18 months<age<=24 months -1.7157 (-41.24) -0.0194 (-5.23) *24 months<age<=30 months -0.8984 (-20.54) -0.0162 (-4.15) 3. Birth order -0.0966 (15.52) -0.0090 (-16.12) 4. Sex (Male = 0) -0.3558 (-18.71) 0.0122 (7.24) 5. Mother's height (cms.) 0.0143 (9.32) 0.0013 (9.65) 6. Mother's education (Illiterate = 0) * Upto 5 years schooling 0.1454 (6.13) 0.0130 (6.11) * More than 5 years schooling 0.2973 (8.69) 0.0280 (9.13) 7. Land operated (No land = 0) * Land with uncertain irrigation 0.2843 (12.93) 0.0262 (13.34) * Land with certain irrigation 0.3585 0.0320 (10.39) (10.39) cont... 155 8 . 9. 10. 11. 12 . 13 . T a b le 7 . 5 c o n t . . . Whether scheduled 0.0671 0.0060 caste (Yes = 0) (2.96) (2.98) Whether father is HH 0.0589 0.0053 head (Yes = 0) (2.41) (2.41) Distance to primary -0.0137 -0.0012 health center (kms.) (-10.10) (-9.65) Distance to general -0.0036 -0.0005 hospital (kms.) (-3.38) (-5.11) Hygeinic water 0.1246 0.0111 (Unhygeinic = 0) (5.16) (5.15) Whether pucca road -0.1954 -0.0156 (Yes = 0) (-8.67) (-7.75) N R-squared F-statistic Mean squared error 9156 9156 0.6449 0.1504 921.95 89.89 0.7960 0.0064 Table 7.6 LEAST S Q U A R E M E A N S T A N D A R D W EIG H T B Y CATEG O RIES O F L A N D O PER A TED : Land Level LS Mean Standard Prob > |t | Prob > | (i or j) Standard Error H O : LS mean=0 H O : LS Mean(i)=Li Weights LS Mean 1 2 <1> (2 ) (3) (4) (5) ( 6 ) (7) B A S E D A TA Land with certain 1 0.7189 0.0038 0.0 0.2708 irrigation Land with uncertain 2 0.7228 0.0027 0.0 0,2708 irrig atio n No land 3 0.7123 0.0027 0.0 0.0739 0.0001 SIBLING D A TA Land with certain 1 0.7302 0.0031 0.0 0.0472 irrig atio n Land with uncertain 2 0.7238 0.0021 0.0 0.0472 irrig atio n No land 3 0.7035 0.0021 0.0 0.0001 0.0001 C U R R E N T D A TA Land with certain 1 0.7455 0.0029 0.0 0.2652 irrig atio n Land with uncertain 2 0.7422 0.0018 0.0 0.2652 irrig atio n N o land 3 0.7204 0.0018 0.0 0.0001 0.0001 t| i Mean(j) 3 (8) 0.0739 0.0001 0.0001 0.0001 0.0001 0.0001 T a b le 7 .7 L E A S T S Q U A R E M E A N S T A N O A R D W E I G H T B Y L E V E L S O F M A T E R N A L E D U C A T I O N : Maternal Level LS Mean Standard Prob > |t | Prob > | Education ( i or j) Standard Error HO : IS mean=0 H O : LS Mean(i)=LS Weights LS Mean 1 2 (2) (3) (4) (5) ( 6 ) (7) B A S E D A TA More than 5 years 1 0.7315 0.0039 0.0 0.0001 schooling Upto 5 years 2 0.7097 0.0030 0.0 0.0001 schooling No schooling 3 0.7128 0.0025 0.0 0.0001 0.2319 SIBLING D A TA More than 5 years 1 0.7264 0.0031 0.0 0.5182 schooling Upto 5 years 2 0.7241 0.0023 0.0 0.5182 schooling No schooling 3 0.7069 0.0018 0.0 0.0001 0.0001 C U R R E N T D A TA More than 5 years 1 0.7525 0.0029 0.0 0.0001 schooling Upto 5 years 2 0.7355 0.0020 0.0 0.0001 schooling N o schooling 3 0.7201 0.0017 0.0 0.0001 0.0001 t| Mean(j) 3 (8) 0.0001 0.2319 0.0001 0.0001 0.0001 0.0001 157 T a b le 7 . 8 158 L E A S T S Q U A R E M E A N W E I G H T & S T A N D A R D W E I G H T B Y S E X Sex Level Weight & Standard Prob > | t | Prob > | t | HO: std weight Error HO: LS Mean=0 LS Mean<1)=LS Mean(2) LS Means LS Mean <1> (2 ) <3> (4 ) (5 ) ( 6 ) BA SE DATA WEIGHT LEAST S Q U A R E M EA N S Female 1 8.8110 0.0279 0 .0 0 .0 Male 2 9.3739 0.0318 0 .0 SIBLING DATA Female 1 8.2074 0.0230 0 .0 0 .0 Male 2 8.7247 0.0228 0 .0 CU RR EN T DATA Female 1 7.8197 0.0187 0 .0 0 .0 Male 2 8.1771 0.0191 0 .0 STANDARD WEIGHT LEAST SQ U A R E M E A N S B A SE DATA Female 1 0.7147 0.0025 0 .0 0.0023 Male 2 0.7213 0.0029 0 .0 SIBLING DATA Female 1 0.7149 0 .0 0 2 0 0 .0 0 .0 0 0 1 Male 2 0.7242 0 .0 0 2 0 0 .0 CURRENT DATA Female 1 0.7389 0.0017 0 .0 0 .0 0 0 1 Male 2 0.7268 0.0018 0 .0 T a b le 7 . 9 L E A S T S Q U A R E M E A N S T A N D A R D W E IG H T B Y W H E T H E R S C SC Level Standard Standard Prob > |t| Prob > |t| HO: Weight Error HO: LS Mean=0 LS Mean(1)=LS Mean(2) LS Mean LS Mean (1> (2 ) (3 ) (4 ) (5 ) ( 6 ) BASE DATA Not S C 1 0.7208 0 .0 0 2 2 0 .0 0.0609 S C 2 0.7152 0.0034 0 .0 SIBLING DATA Not S C 1 0.7246 0.0016 0 .0 0 .0 0 0 1 S C 2 0.7137 0.0024 0 .0 CURRENT DATA Not SC 1 0.7390 0.0014 0 .0 0.0031 S C 2 0.7330 0 .0 0 2 2 0 .0 Table 7.10 LEAST S Q U A R E M E A N STANDARD WEIGHT BY W H ETH ER FATHER IS H H HEAD: Whether Level Standard Standard Prob > | t | Prob > j t | HO: father is Weight Error HO: LS Mean=0 LS Mean(1)-LS Mean(2) H H head LS Mean LS Mean (1) ( 2 ) <3) (4 ) (5 ) ( 6 ) BASE Father not head 1 0.7305 0.0033 0 .0 0 .0 0 0 1 Father head 2 0.7054 0.0023 0 .0 SIBLINGS Father not head 1 0.7270 0.0026 0 .0 0 .0 0 0 1 Father head 2 0.7113 0.0018 0 .0 CURRENT Father not head 1 0.7399 0 .0 0 2 1 0 .0 0.0004 Father head 2 0.7321 0.0017 0 .0 T a b le 7.11 L E A S T S Q U A R E M E A N S T A N D A R D W E I G H T B Y D E V E L O P M E N T B L O C K Development Block Level LS Mean Standard Prob > |t | Prob > |t| ( i or j) Standard Error H O : LS mean=0 H O : LS Meanti)=LS Meantj) Weights LS Mean 1 2 3 4 (1) (2) t3) 14) t5) t 6 ) t7) (8 ) t9) B A S E D A TA Kottampatti 1 0.7056 0.0031 0 .0 0.9658 0.0001 0.0001 Chellampatti 2 0.7055 0.0029 0 .0 0.9658 • 0.0001 0.0001 Kariapatti 3 0.7744 0.0050 0 .0 0.0001 0.0001 • 0.0001 Sivagangai 4 0.6864 0.0036 0 .0 0.0001 0.0001 0.0001 • SIBLING D A TA Kottampatti 1 0.7718 0 .0 0 2 2 0 .0 0.0001 0.0001 0.0001 Chellampatti 2 0.7273 0.0026 0 .0 0.0001 • 0.0001 0.0057 Kariapatti 3 0 .6 8 8 8 0.0047 0 .0 0.0001 0.0001 • 0.0001 Sivagangai 4 0.7188 0.0028 0 .0 0.0001 0.0057 0.0001 ■ C U R R E N T D A TA Kottampatti 1 0.7353 0.0021 0 .0 0.0001 0.8122 0.0217 Chellampatti 2 0.7453 0 .0022 0 .0 0.0001 • 0.0233 0.0001 Kariapatti 3 0.7342 0.0041 0 .0 0.8122 0.0233 • 0.2931 Sivagangai 4 0.7293 0.0025 0 .0 0.0217 0.0001 0.2931 T a b le 7.12 LEAST SQUARE MEAN STANDARD WEIGHT BY WATER QUALITY & PUCCA ROAD (Current data) (1) Level Standard Weight LS Mean (2) (3) Standard Error LS Mean (4) Prob > |t| HO: LS Mean=0 (5) Prob > |t| HO: LS Mean(l)=LS Mean(2) ( 6 ) WATER QUALITY Hygeinic Not hygeinic 1 2 0.7370 0.7259 0.0016 0.0023 0.0 0.0 0.0001 WHETHER VILLAGE IS CONNECTED BY A PUCCA ROAD Connected 1 0.7393 Not connected 2 0.7237 0.0021 0.0017 0.0 0.0 0.0001 Table 7.13a L E A S T S Q U A R E M E A N S T A N D A R D W E I G H T B Y A G E G R O U P S : SIBS A N D C U R R E N T Age Groups Level LS Mean Standard Prob > | t | Prob > | t | (Age in months) (i or j) Standard Error HO: LS mean=0 HO: LS Mean(i)=LS Mean(j) Weights LS Mean 1 2 3 4 5 6 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) SIBLINGS Age <= 9 1 0.7359 0.0034 0.0 . 0.0001 0.0001 0.0001 0.0010 0.7708 9 < Age <= 12 2 0.7053 0.0035 0.0 0.0001 . 0.4341 0.0439 0.0001 0.0001 12 < Age <= 18 3 0.7022 0.0026 0.0 0.0001 0.4341 . 0.0004 0.0001 0.0001 18 < Age <= 24 4 0.7130 0.0026 0.0 0.0001 0.0439 0.0004 . 0.0003 0.0001 24 < Age <= 30 5 0.7238 0.0025 0.0 0.0010 0.0001 0.0001 0.0003 0.0001 30 < Age <= 36 6 0.7348 0.0025 0.0 0.7708 0.0001 0.0001 0.0001 0.0001 CURRENT Age <= 9 1 0.7623 0.0025 0.0 0.0001 0.0001 0.0001 0.0001 0.0002 9 < Age <= 12 2 0.7273 0.0027 0.0 0.0001 0.0334 0.9779 0.2842 0.0001 12 < Age <= 18 3 0.7212 0.0021 0.0 0.0001 0.0334 0.0123 0.0004 0.0001 18 < Age <= 24 4 0.7272 0.0022 0.0 0.0001 0.9779 0.0123 0.2082 0.0001 24 < Age <= 30 5 0.7307 0.0025 0.0 0.0001 0.2842 0.0004 0.2082 0.0001 X 30 < Age_<£ 36_______6 0.7475 0.0035 _____ 0.0_ 0.0002 0.0001 0.0001 0.0001 0.0001 Table 7.13b LEAST S Q U A R E M E A N S TA N D A R D W EIG HT B Y A G E G R O U PS : B A S E A N D C U R R E N T Age Groups Level LS Mean Standard Prob > |t | Prob > |t| (Age in months) ( i or j ) Standard Error H O : LS mean=0 HO : LS Mean(i)=LS Mean(j) Weights LS Mean 1 2 3 4 5 6 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) B A S E Age <= 9 1 0.7326 0.0073 0.0 . 0.0022 0.0001 0.0021 0.2055 0.5917 9 < Age <= 12 2 0.7051 0.0057 0.0 0.0022 . 0.4888 0.4533 0.0019 0.0001 12 < Age <= 18 3 0.7009 0.0034 0.0 0.0001 0.4888 . 0.0193 0.0001 0.0001 18 < Age <= 24 4 0.7096 0.0030 0.0 0.0021 0.4533 0.0193 . 0.0001 0.0001 24 < Age <= 30 5 0.7232 0.0027 0.0 0.2055 0.0019 0.0001 0.0001 . 0.0001 30 < Age <= 36 6 0.7365 0.0026 0.0 0.5917 0.0001 0.0001 0.0001 0.0001 C U R R E N T Age <= 9 1 0.7623 0.0025 0.0 . 0.0001 0.0001 0.0001 0.0001 0.0002 9 < Age <= 12 2 0.7273 0.0027 0.0 0.0001 . 0.0334 0.9779 0.2842 0.0001 12 < Age <= 18 3 0.7212 0.0021 0.0 0.0001 0.0334 . 0.0123 0.0004 0.0001 18 < Age <= 24 4 0.7272 0.0022 0.0 0.0001 0.9779 0.0123 . 0.2082 0.0001 24 < Age <= 30 5 0.7307 0.0025 0.0 0.0001 0.2842 0.0004 0.2082 . 0.0001 30 < Age < = 36_______ 6 0.7475 0.0035________0.0 0.0002 0.0001 0.0001 0.0001 0.0001 Os C hapter V III 164 REGRESSION RESULTS: SEX AND BIRTH ORDER INTERACTIONS In this chapter the effects of two individual level regressors - sex and birth order - are explored further and an attempt made to understand the reasons for the negative effects. Two specific issues are addressed. First, does the observed male-female gap differ significantly across households of varying economic and social status? This question is explored in section 8.1 by examining the effects of interactions between the class variables sex and land operated, and between sex and whether the child comes from a scheduled caste household. Second, does maternal education and better household economic status partly compensate for the observed negative effects of birth order? If so, then the effects of birth order on weight outcomes would differ significantly across children whose mothers have different levels of education, and across children belonging to households of different economic status. To examine this the covariate birth order is allowed to interact, in turn, with the class variables maternal education and land operated, and the sign and significance of the interaction term coefficients are studied in section 8.2. Section 8 3 is the concluding section. 8.1. Sex Interactions In chapter VII the independent effects of a sex dummy showed that being female had a negative effect upon standard weight for base and sibling data, with a 165 positive sign for current data indicating a reduction in absolute male-female differences over time. This conclusion gets modified when household economic status and caste are controlled for. Regressions for weight and standard weight were reestimated allowing for two kinds of interactions with the female sex dummy: (a) interaction between sex and land operated by the household, and (b) interaction between sex and whether the household belonged to the SC category. The motivation for so doing was that, perhaps, the extent of male-female differences in weight outcomes depended upon a household’s economic status and social status. Better off households could be expected to have a lower male-female gap as compared with the worse off. As before, land operated can be taken as evidence of relatively higher socioeconomic status as compared with the landless, and belonging to one of the scheduled castes (SC) can be taken as evidence of lower social status as compared with non-SC households1. In what follows are discussed the results of, first, interactions between sex and land operated, and second, interactions between sex and whether or not the child comes from an SC household. Since the concern here is with the sex of the child, only regressions for the age-sex standardized weights are tabulated and discussed. As mentioned before, this allows direct comparisons between male and female weights relative to their respective NCHS standards. Interactions between land and sex. We examine here the question of whether sex differences vary across households operating different levels of land. 1 It is not possible to uniquely separate households on the basis of sodal or economic status and there must inevitably be overlaps. Inequalities are often ’cumulative.’ when households high on the economic scale also tend to be high on the social and power scales as well. Yet, land operated has been fundamentally considered as an indicator of command over resources which may, in turn, lead to higher social status. Belonging to an SC household has been primarily considered an indicator of lower social status, which may also be associated with a lower command over resources (Beteille, 1974). 166 Household economic status as measured by land operated has already been found to be important in explaining weight variations. Regressions in table 8.1 for base, sibling and current data contain interactions between a female sex dummy and dummies for the two type of land operated (landless = 0). It appears that the sex bias against female children is most in evidence in the better off landed households. Consider the coefficients and t-values of the interaction terms. All interaction terms for the landed categories have negative signs. This indicates that as compared with children from landless households, those from landed households exhibit a greater sex-bias against females. The t-values are significant in 4 out of 6 cases. The 2 cases where the coefficients are not significant are both for households not operating the best land - land with uncertain irrigation. Children from the best off households, Le., those operating land with assured irrigation, show negative and significant coefficients as compared with children from landless households. This indicates that the male-female gap is in evidence primarily in the best off households even though the mean standard weight without separating data by sex is higher in these households. Comparisons of the least square mean standard weights based on this model with interactions between land and sex are available in tables 8.2 to 8.4, separately for base, sibling and current data. Pairwise comparisons of mean standard weight for the 6 land-sex level combinations (3 levels of land and 2 of sex give us 6 sex-land combinations) are carried out and the null hypothesis of equality of pairwise means is tested in columns 7 -12 of each table. For the base line, within each level of land operated male and female means differ significantly (table 8.2). Females have higher standard scores for the two lower land categories whereas males do better than females in the highest land category - 167 land with assured irrigation. For female children moving to a higher level of land operated results in higher mean weight and standard weights only when the move is from the landless to the landed category. A further increase in the land category does not increase, and in fact decreases, standard weight (compare levels 1, 3 and 5). On the other hand, for male children higher land levels are always associated with significantly higher standard weight scores (compare levels 2, 4 and 6). Male standard weight for the lowest land category is not significantly different from female standard weight for the highest land i category. There are no sex differences among the lower land categories as there are j among children from households operating the best lands. I I For sibling data lower standardized weight scores are observed for female j I children as compared with males in both categories of landed households, but not among j the landless (table 83). Higher land operated categories are associated with higher S standard weight for males and females, but the increase is much sharper for male children ; 1 moving from the landless to the landed category. i Among the currently enrolled, female children have significantly higher | I mean standard weight scores as compared with males not only among the landless, but also among the landed households with uncertain irrigation (table 8.4). Among the highest ' level of land operated males have higher standard weight scores. Once again differences by sex seem to be most in evidence among the better landed households with female standard weight scores being higher than male among other households. Over time the male-female gap diminishes. For data on the currently ! I enrolled the male-female gap among the better landed households is less than that i observed among their older siblings and that for base line data. The associated p-value j I testing for equality of means using a 2-tailed t-test for current data is much higher than for I 168 sibling and base data. The p-values are 0.1427 for current data, 0.0056 for sibling and 0.0027 for base line data. Thus, the data show that while mean standard weights are higher among children from better landed households, female means are not always higher when we move from the landless to the better land categories, the way male means are. This result was difficult to interpret. We found household economic status to be strongly positively associated with weight. It would not have been unreasonable to presume that the differences by sex often observed for south Asian countries in connection with infant and child mortality and even schooling, would be reduced with improvements in household economic status. The results above are not consistent with such a presumption. Improved economic status does not reduce sex differences. This points towards looking for other explanations. Two suggestions in an attempt to explain this result are given below. The explanations are tentative and cannot, without further exploration, be considered definitive. One possible explanation for such a result could be that among households that have a lower economic status, and especially the landless (a very high percentage of scheduled caste households come under this category), every adult member tends to engage in some economic activity. Among households that have a higher economic status, especially the privileged land owners (those with assured irrigation in the present situation - who are likely to have wells with pumpsets, or good lands located closer to the head reaches of surface irrigation systems), -female members tend not to go' out and work for a wage. This could be for reasons of ’prestige’ and related social reasons. The resulting lesser economic value of female members could be one possible reason. Another is the greater prevalence of the practice of dowries among the better landed households making the female child a very expensive proposition for her natal household. Both these 169 explanations or pressures would also be relevant, though perhaps to a lesser degree, for socially upwardly mobile households who try to ’sanskritize’ (Beteille, 1974), i.e, adopt certain practices of households higher on the social scale, including their taboos against work for women. These pressures can be expected not to operate among the landless. The considerable reduction in male-female differences over time could be due to the negative selection for supplementary feeding by the program. While a child’ s sex is not one of the criteria for selection, in so far as females tend show relatively greater growth faltering and have lower weights, they are likely to be selected more often for an additional input of supplementary feeding. Interactions between SC and sex. Interactions between household economic status and a female dummy showed no differences by sex among the landless households and the strongest sex differences among those with the best land. Possible explanations included factors like social status or prestige resulting in tendencies for female members not working for a wage among the landed and higher pressures for customs like dowry payments. These factors could result in a relatively lower economic value of female members among better landed households. To the extent social status is a factor influencing the prevalence of differences by sex, it is possible that separating children from SC versus non-SC households might give further insight into the pattern of these differences. In the Indian context where caste considerations are relevant in determining social status, especially in rural areas, the SC households are a specially deprived category. These households reside separately from the main dwelling area in a village, in an area called the ’cheri’ or ’colony’. To examine the possible differential male-female gap in standard weight among SC versus non-SC households, we look at the coefficient of the interaction term 170 between the female sex dummy and a dummy for whether a child comes from a non-SC household. Negative signs for this interaction term are obtained in each case - for base line, sibling and current data. This indicates stronger sex differences among the non-SC households as compared with SC households, thus indicating a greater prevalence of the male-female gap among households of higher social status. Comparisons of pairwise least square mean standard weights by whether SC are carried out in table 8.5. Lower scores for females are observed among non-SC households as compared with males, but higher scores for females are observed among the relatively socially deprived SC households. This tendency seems to have declined over time, being less for current data as compared with base and sibling data, similar to what is observed for interactions between sex and land. Thus, the widest male-female gap can be placed largely in the landed upper caste households, possibly because of the lower economic and social value of women. On the other hand, the regression results also suggest that higher child weights are associated with children from landed households. Hence, the relative neglect of female children is less likely to be fatal as compared with poorer households (Bardhan, 1988). It was seen in chapter V that average weights of male and female preschoolers from well-off urban families were closer to the US NCHS median weights than to TEMP levels. In fact, for the under 3 years age group among the well-off, average male weights ranged between 90.00% to 9538% of the male NCHS standards for * corresponding ages, and average female weights ranged between 91.11% to 9538% of their corresponding standards. In the light of the above result showing wider male-female differences among the better off classes in TINP areas one could reasonably ask why females from urban well-off families did no worse than males. An explanation for this 171 apparent contradiction lies in the fact that better off households from rural areas are not comparable with well-off urban households with a history of much better access to hospitals, health care facilities, information and communication. Moreover, the urban well- off are relatively modernized, perhaps with more than one generation of education available to them. This could reduce the probability of the relative neglect of females resulting in lower weight outcomes. The explanation does not imply that pressures for dowries or the preference for male children are non existent. But, even if operative, such preferences are less likely to result in poorer growth of females. 8.2 Birth Order Interactions. Standard weight scores were earlier found to vary negatively with birth order, as evident from negative and significant coefficients of the birth order variable in the standard weight regressions. Higher levels of mother’s education and better household economic status were found to be associated with higher weights and standard weights - differences of intercept. To the extent resources available to a household to buy food are a limiting factor causing the negative relationship, better household economic status could partly counter the effect of birth order. Similarly, apart from economic strain, to the extent maternal depletion or a greater proportion of children in the household leading to greater probabilities of infections cause the negative relationship, maternal education could counter the effects of birth order since better educated mothers are likely to be more aware of the problems, more receptive in adopting new ideas and practices made available by the program and more efficient at producing child health outcomes. Interaction effects between birth order and maternal education level, and between birth order and land operated by the household are examined here. The regressions for standard weight are reestimated allowing for interactions between (a) birth order and maternal education levels, and (b) birth order and land operated by the household (table 8.6). The interaction terms allow an examination of whether the negative relationship between birth order and weights is different for the 3 levels of mother’s education or for the 3 levels of land operated. Thus we look for significant slope differences (as distinct from intercept differences examined earlier, captured by dummies). This is discussed in turn below. Interactions between birth order and mother’s education. Does mother’s education partly compensate for the negative effects of high birth orders? To the extent it does, the magnitudes of the negative coefficients would be smaller for children of more educated mothers as compared with those of less educated. Looking for slope differences, we examine the coefficients of the interaction terms in table 8.6. Significant differences in the response coefficients for children of educated mothers as compared with those of illiterate are observed. All interaction terms are significant except one - that in respect of current data for mothers with upto 5 years of schooling. As compared with children of illiterate mothers, the negative birth order slopes indeed differ for educated mothers. Overall, the coefficient magnitudes of the interaction terms are higher for children of mothers with some schooling as compared with children of illiterate mothers. Further, in the case of base and sibling data, the coefficients are even higher for mothers with more than 5 years of schooling. This shows that slopes or the response coefficients capturing birth order effects for children of the highest maternal education category are furthest from the corresponding slopes of the omitted category, 173 with the slopes for the intermediate education category being in between. The picture is less dear cut for current children where the highest maternal education level does not result in a higher interaction coeffident. The compensatory effects of maternal education appear less clear cut for current data, but are evident for base and sibling data. So far we have looked at interaction coefficients relative to the omitted category of children of mothers with no schooling. Actual regression coeffidents are also estimated, separately for each level of maternal education, for base, sibling and current data (table 8.7). Looking vertically down the columns for base and sibling data shows smaller absolute values of the coefficients indicating lower negative effects of birth order for children of mothers with higher education levels. In fact birth order does not have a significant impact upon children of mothers with the highest level of education. For this maternal education level the birth order coeffident is positive and insignificant in the case of base line children (coeffident = 0.0020, t = 0.99) and negative and insignificant in the case of siblings (coeffident = -0.0030, t = -0.90). High maternal education does alter the negative relationship between birth order and weights - the effects of maternal education are compensatory. This pattern is less perfect for current data which includes the weight records of later bom children to households represented in the sibling data. Here the negative response coefficient in the case of some schooling is less strong as compared with no schooling showing some compensatory effects. But the coeffident for more than 5 years of schooling shows stronger birth order effects. Three possible explanations for the less perfect compensatory effects of maternal education for current data can be thought of: (a) Data for the currently enrolled children consist of weight records on higher birth order children on an average, as compared with sibling data. The mean birth orders for sibling 174 and current data are 2.2 and 2.8, respectively. Other things being equal, in a horizontal comparison of birth order coefficients between current and sibling data one should expect higher absolute coefficient values for current children in each education category because of the expected stronger effects of higher birth orders2, (b) For current data, now that the program has had 6 to 8 years of operation in the community, the advantage of formal maternal schooling could have diminished. The program itself is an alternate educational j experience for mothers, providing information on child health and nutrition with its growth | * charts for each child, monthly weighings of children, selective feeding of only the worst off j j cases, home visits by the project staff and several other communication activities. Of ! course, it is still possible that better educated mothers derive greater benefits from the program which shows up in terms of intercept differences, (c) A high percentage of mothers of the currently enrolled have already had an older child participate in the ! program (63%). This gives them an additional advantage, independent of formal schooling, 1 I as compared with households with a child participating for the first time. j Thus, the TINP data show that maternal education does partly compensate for negative birth order effects. This conclusion is strongest for the base line I when possible program effects are minimum. It continues to hold good for siblings. The j advantage of higher formal maternal schooling gets reduced for the currently enrolled children after 6 to 8 years of program operation. Of course intercept differences due to maternal education are independently relevant. Interactions between birth order and land operated. Does higher household economic status compensate for the negative effects of birth order? We have ’ ' I I 2 But except in one case this does not happen. For current data other things cannot really j be considered equal. The effect of the program may have softened the negative birth order effects. 175 already seen the positive effects of household economic status as measured by land operated on weight outcomes in terms of higher intercepts for weight and standard weight in the case of children from landed households as compared with the landless. The question being examined here is whether the negative birth order-weight relationship differs significantly between children of landed versus landless households. In other words, are slope differences significant? And if so, does having land compensate for the negative effects of high birth orders? If it is inadequacy of resources that is important in causing the negative relationship, then one can expect a compensatory effect to operate for the economically better off households. An examination of the coefficients and t-values of the interaction terms between land operated and birth order in the regressions in table 8.6 allows us to test for slope differences from the landless category, separately for base, sibling and current data. We observe the following: (a) For base line data the intermediate level of land is not significantly different from the landless category, whereas the highest level of land with certain irrigation is. (b) For siblings the highest level of land operated is not significantly different from the landless category, whereas the intermediate category is. (c) For current data both the landed categories are significantly different from the landless, though not from each other. However, the coefficient magnitudes and signs indicate a more strongly negative relationship for the landed categories. It appears that the differential effects of birth order across land operated categories are not so dear cut. Estimating the actual coeffidents of birth order, separately for each land category (table 8.7) we see how their magnitudes and signs differ across land categories by making vertical comparisons. Land with certain irrigation compensates for negative birth order effects among base line children as seen by the positive and insignificant birth order 176 coefficient for this category (coefficient = 0.0020, t = 0.92), whereas the lower land categories have significant negative birth order coefficients. For sibling data initially the coefficient magnitude goes up from -0.0031 to -0.0135 as we move from the landless to the intermediate land category, and then falls to 0.0053 for the best land category. Some compensation seems to operate for the best land category as compared with the intermediate, but the smaller negative coeffident for the best land category is not statistically significantly different from the landless. Among the currently enrolled the negative birth order coeffidents are much stronger for the two landed categories. It appears that the compensatory effects of land are not dear cut like those of maternal education. 83 Conclusion The focus of this chapter was on exploring further the effects of sex and birth order - two individual level variables - in terms of their interactions with other selected regressors. The differential effects of sex across households of varying sodoeconomic status examined above indicate that lower weight outcomes for female children relative to males are observed largely among landed and non scheduled caste households. The relatively lower economic and sodal status of women among rural upper classes may be a possible explanation. Further anthropological and sodological evidence for this would be of interest. However, since child weights are higher among landed households, the relative neglect of female children may be less likely to result in mortality as compared with children from worse off households. Over time the reduction of the gap between the sexes could be due to program effects. The time period is too short for there 177 to be discernible social changes in favor of females. While being later bom is associated with lower weight outcomes, maternal education is seen to have significant compensatory effects. These effects are much more clear cut than the compensatory effects of land. This indicates that the negative effects of being female or of higher birth orders are less due to resource constraints, and more due to other factors. In the case of sex explanations include factors like prestige and the custom of dowry payments contributing to a lower economic value of a female child. In the case of birth order explanations include factors that can be countered by education resulting in better health care, like higher probability of infections, maternal depletion and lower birth weight. T a b le 8 . 1 178 REDUCED FORM REGRESSIONS FOR STANDARDIZED WEIGHT WITH SEX INTERACTIONS: BASE, S IB S AND CURRENT Regressors Base Siblings Current 1 . Intercept 0.4777 (18.19) 0.2686 (10.83) 0.5979 (27.60) 2. Child age (months) -0.0013 (-1.56) -0.0050 (-8.45) -0.0093 (-16.94) 3. Child age squared (months) 5.78607E-05 (3.45) 0.00013 (9.83) 0.00023 (16.30) 4 . Birth order -0.0069 (-10.46) -0.0119 (-15.77) -0.0088 (-15.34) 5. Sex (Male = 0) 0.0266 (5.10) 0.0103 (2.36) 0.0219 (6.09) 6. Mother's height (cms. ) 0.0014 (8.87) 0.0030 (19.06) 0.0013 (9.73) 7. Mother's education (Illiterate = 0) * Upto 5 years schooling -0.0031 (-1.03) 0.0145 (6.10) 0.0149 (6.95) * More than 5 years schooling 0.0203 (5.61) 0.0185 (5.59) 0.0330 (10.92) 8. Land operated (No land = 0) * Land with uncertain irrigation 0.0101 (2.81) 0.0363 (11.60) 0.0217 (7.74) * Land with certain irrigation 0.0227 (4.05) 0.0398 (8.50) 0.0361 (8.58) 9. Whether scheduled caste (Yes = 0) 0.0272 (5.92) 0.0119 (3.86) 0.0111 (3.68) 0. Whether father is HH head (Not = 1) 0.0295 (10.11) 0.0172 (6.54) 0.0071 (3.22) cont... 179 T a b le 8 . 1 c o n t . . . Distance to primary health center (kms.) Distance to general hospital (kms.) Development block (Siv = 0) Kot Chel Kari INTERACTIONS BETWEEN (Male =0; No land = Land with uncertain irrigation, female Land with certain irrigation, female —6.71424E—05 (-0.35) -0.0026 (-13.52) 0.0154 (4.21) 0.0179 (5.11) 0.0881 (15.17) SEX & LAND. 0) -0.00039 (-0.08) -0.0287 (-3.91) -0.00055 (-3.16) - 0.0011 (-7.31) INTERACTIONS BETWEEN SEX & SC. (Male = 0; SC - 0) —4.89464E—05 -0.00071 (-0.65) (-4.41) 0.0248 (8.38) 0.0094 (3.04) -0.0256 (-4.56) * Female, non-SC -0.0365 (-6.24) -0.0312 (-7.23) -0.0261 (-4.04) -0.00014 (-0.03) 0.0064 (2.44) 0.0163 ( 6 . 11) 0.0050 (1.07) -0.00044 ( -0. 12) -0.0250 (-4.25) -0.0094 (-2.40) N 6658 R-squared 0.1258 F-statistic 50.26 Mean squared error 0.0070 7002 0.1716 76.14 0.0062 9156 0.1498 84.70 0.0063 Table 8 .2 Land ( 1) Land with certain irrig atio n Land with uncertain irrig atio n N o land L E A S T S Q U A R E M E A N S T A N D A R D W E I G H T B Y S E X F R O M H H s O P E R A T IN G D IF F E R E N T C A T E G O R I E S O F L A N D (BASE DATA) Sex Level LS Mean Standard Prob > |t| ( i or j ) Standard Error H O : LS mean=0 Weights LS Mean <2) (3) (4 ) (5) ( 6 ) Female 1 0.7153 0.0044 0 .0 Male 2 0.7356 0.0055 0 .0 Female 3 0.7309 0.0029 0 .0 Male 4 0.7229 0.0030 0 .0 Female 5 0.7212 0.0028 0 .0 Male 6 0.7128 0.0032 0 .0 Prob > |t | H O : LS Mean(i) = LS Meantj) 1 2 3 4 5 6 (7) ( 8 ) (9) ( 10) ( 11) ( 12) . 0.0027 0.0007 0.1390 0.2245 0.6388 0.0027 * 0.4230 0.0171 0.0130 0.0001 0.0007 0.4230 • 0.0316 0.0033 0.0001 0.1390 0.0171 0.0316 • 0.6444 0.0049 0.2245 0.0130 0.0033 0.6444 • 0.0233 0.6388 0.0001 0.0001 0.0049 0.0233 Table 8.3 L E A S T S Q U A R E M E A N S T A N D A R D W E I G H T B Y S E X F R O M H H s O P E R A T IN G D IF F E R E N T C A T E G O R IE S O F L A N D (SIBLING DATA) Lend Sex Level LS Mean Standard Prob > |t | Prob > |t | H O : LS Mean(1) = LS Mean(j) ( i or i) Standard Error H O : LS mean=0 1 2 ( 1) ( 2 ) (3) Weights (4) LS Mear (5) Land with certain irrig atio n Female 1 0.7227 0.0042 Male 2 0.7385 0.0041 Land with uncertain Female Irrig atio n 3 0.7140 0.0026 Male 4 0.7350 0.0025 No land Female 5 0.7090 0.0027 Mate 6 0.6987 0.0027 (6 ) (7) ( 8 ) (9) ( 10) ( 11) 0 2 ) 0 .0 . 0.0056 0.0486 0.0083 0.0031 0.0001 0 .0 0.0056 • 0.0001 0.4271 0.0001 0.0001 0 .0 0.0486 0.0001 • 0.0001 0.0934 0.0001 0 .0 0.0083 0.4271 0.0001 ■ 0.0001 0.0001 0 .0 0.0031 0.0001 0.0934 0.0001 • 0.0021 0 .0 0.0001 0.0001 0.0001 0.0001 0.0021 Table 8.4 L E A S T S Q U A R E M E A N S T A N D A R D W E I G H T S B Y S E X F R O M H H s O P E R A T I N G D IF F E R E N T C A T E G O R IE S O F L A N D (C U R R EN T DATA) Land Sex Level LS Mean Standard Prob > |t | Prob > |t | H O : LS Mean(i) = LS Mean(j) <i or j ) Standard Error H O : LS mean=0 1 2 3 4 5 6 ( 1) ( 2 ) (3) Weights (4) LS M ean (5) Land with certain irrig atio n Female 1 0.7368 0.0040 Male 2 0.7447 0.0038 Land with uncertain Female irrig atio n 3 0.7469 0.0021 Male 4 0.7302 0.0024 No land Female 5 0.7257 0 .0 0 2 2 Male 6 0.7085 0.0023 ( 6 ) (7) (8 ) (9) ( 10) ( 11) ( 12) 0 .0 . 0.1427 0.0168 0.1413 0.0107 0.0001 0 .0 0.1427 • 0.5845 0.0003 0.0001 0.0001 0 .0 0.0168 0.5845 • 0.0001 0.0001 0.0001 0 .0 0.1413 0.0003 0.0001 • 0 .1100 0.0001 0 .0 0.0107 0.0001 0.0001 0 .1 1 0 0 • 0.0001 0 .0 0.0001 0.0001 0.0001 0.0001 0.0001 Table 8.5 L E A S T S Q U A R E M E A M S T A N D A R D W E I G H T S B Y S E X F R O M S C V E R S U S N O N -S C H O U S E H O L D S S C Sex Level LS Mean Standard Prob > |t | Prob > |t| H O : LS Mean(l) = LS Mean(j) ( i or j ) Standard Error H O : LS mean=0 1 2 3 4 Weights LS Mean (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) B A S E D A TA Yes Female 1 0.7271 0.0038 0.0 0.0020 0.0131 0.0105 Yes Male 2 0.7102 0.0046 0.0 0.0020 0.1134 0.0001 N o Female 3 0.7178 0.0021 0.0 0.0131 0.1134 0.0001 No Male 4 0.7373 0.0026 0.0 0.0105 0.0001 0.0001 • SIBLING D A TA Yes Female 1 0.7093 0.0034 0.0 0.0353 0.0004 0.0001 Yes Male 2 0.7181 0.0030 0.0 0.0353 0.3566 0.0001 N o Female 3 0.7211 0.0020 0.0 0.0004 0.3566 0.0005 N o Male 4 0.7300 0.0021 0.0 0.0001 0.0001 0.0005 • C U R R E N T D A TA Yes Female 1 0.7356 0.0027 0.0 0.0003 0.5237 0.4321 Yes Male 2 0.7222 0.0030 0.0 0.0003 0.0001 0.0002 N o Female 3 0.7373 0.0019 0.0 0.5237 0.0001 0.0826 No Male 4 0.7333 0.0018 0.0 0.4321 0.0002 0.0826 T a b le 8 . 6 184 REDUCED FORM REGRESSIONS FOR STANDARDIZED WEIGHT WITH BIRTH ORDER INTERACTIONS: BASE, S IB S AND CURRENT Regressors Base Siblings Current 1. Intercept 0.5056 (19.29) 0.2898 11.72 0.5931 (27.72) 2. Child age (months) -0.0014 (-1.65) -0.0049 (-8.35) -0.0092 (-16.90) 3. Child age squared (months) 5.88970E—05 (3.51) 0.00013 (9.76) 0.00022 (16.25) 4. Birth order -0.0092 (-7.04) -0.0074 (-5.40) -0.00013 (0.14) 5. Sex (Male = 0) -0.0074 (-3.43) -0.0072 (-3.61) 0.0134 (7.89) 6. Mother's height (cms.) 0.0014 (8.83) 0.0029 (18.47) 0.0013 (9.29) 7. Mother's education (Illiterate = 0) * Upto 5 years schooling -0.0128 (-2.60) 0.0074 (1.58) 0.0116 (2.77) * More than 5 years schooling -0.0029 (-0.45) 0.0027 (0.35) 0.0472 (6.99) 8. Land operated (No land = 0) * Land with uncertain irrigation 0.0117 (2.59) 0.0428 (10.35) 0.0595 (15.86) * Land with certain irrigation -0.0138 (-1.75) 0.0287 (3.89) 0.0635 (10.33) 9. Whether scheduled caste (Yes = 0) 0.0050 (1.70) 0.0090 (3.91) 0.0050 (2.49) 0. Whether father is HH head (Not = 1) 0.0254 (8.89) 0.0166 (6.29) 0.0051 (2.31) cont. T a b le 8 . 6 c o n t . . 185 11. Distance to primary health center (kms.) 12. Distance to general hospital (kms.) 13. Development block (Siv = 0) * Kot * Chel * Kari 14. INTERACTIONS BETWEEN (Illiterate = 0) * Upto 5 years schooling * More than 5 years schooling 15. INTERACTIONS BETWEEN (No land = 0) * Land with uncertain irrigation * Land with certain irrigation — 3 . 19468E— 05 ( - 0 . 1 6 ) -0.0026 (-13.67) 0.0171 (4.63) 0.0192 (5.48) 0.0895 (15.46) -0.00070 (-4.09) - 0 . 0 0 1 0 ( - 6 . 8 9 ) —4.89513E-05 -0.00087 (-0.74) 0.0193 (6.43) 0.0083 (2.64) -0.0297 (-5.35) (-5.44) 0.0044 (1.69) 0.0149 (5.61) 0.0066 (1.42) BIRTH ORDER & MATERNAL EDUCATION. 0.0036 (2.15) 0.0089 (4.33) 0.0045 (2.46) 0.0086 (2.49) 0.0017 (1.33) -0.0059 (-2.59) BIRTH ORDER & LAND OPERATED. -0.00025 (-0.17) 0.0070 (2.92) -0.0104 (-6.37) -0.0023 (-0.83) -0.0138 ( -11.88) -0.0137 (-7.44) N 6658 R-squared 0.1211 F-statistic 45.70 Mean squared error 0.0071 7002 0.1718 72.42 0.0063 9156 0.1614 87.92 0.0063 186 T a b le 8 . 7 REGRESSIONS COEFFICIENTS COMPUTED FOR BIRTH ORDER INTERACTIONS Regressors Base Siblings Current 1. INTERACTIONS BETWEEN BIRTH ORDER & MATERNAL EDUCATION * Illiterate -0.0069 (-7.09) -0.0117 (-10.29) -0.0090 (-11.28) * Upto 5 years schooling -0.0034 (-2.15) -0.0072 (-4.60) -0.0074 (-6.54) * More than 5 years schooling 0.0020 (0.99) -0.0030 (-0.90) -0.0149 (-6.80) 2. INTERACTIONS BETWEEN BIRTH ORDER & LAND OPERATED * No land -0.0050 (-3.83) -0.0031 (-1.78) -0.0013 (-1.11) * Land with uncertain irrigation -0.0053 (-5.33) -0.0135 (-9.41) -0.0151 (-14.79) * Land with certain irrigation 0.0020 (0.92) -0.0053 (-2.18) -0.0150 (-9.08) C hapter IX CONCLUSION In this chapter, the main points of the study are summarized in section 9.1, some policy implications are discussed in section 9.2, and finally in section 93, limitations of the study are pointed out, together with suggestions for further research. 9.1 Summary The central problem of this study was to examine and explain the patterns of weight variations among preschoolers. The evidence for this study comes from rural Tamilnadu where a health and nutrition program has been in operation since late 1980. Weights and standardized weight scores are used as the dependent variables. A number of individual, household and communal variables expected to influence weight outcomes are used as regressors. The questions investigated and results obtained are summarized below. First, what are the levels of weight values for different ages observed in the TINP data, are there any changes in the means over time, and how do the mean weights of rural TINP preschoolers compare with weights observed in other Indian data sets? We find that weights in the TINP areas are comparatively low, but trend upward over time. Female weights tend to be significantly below male weights for corresponding ages. Preschooler weights among well off Indian children are much closer to the NCHS levels than to what is observed in TINP areas, which is about 70% below the US NCHS median. The upward trend in TINP areas from 1980 - 1989 contrasts with a lack of trend in Tamilnadu as a whole (1974 - 1982). Although there may be other explanations, the result is consistent with favorable program effects. Second, how do weights and age-sex standardized weight scores vary with a number of individual (age, sex, birth order), household (land operated, maternal education, maternal height, caste, household structure) and community level variables (development block, distances to the nearest health facility and hospital, quality of water supply, pucca road connections)? This is examined using multiple regression analysis. In respect of individual level variables we find that while weight increases with age (at a diminishing rate), the situation is different for standard weight. Standard weight shows a U-shaped pattern - weight falls relative to the NCHS standards right from 6 months of age, with standard weight reaching its lowest level between the ages of 12 to 18 months, and picking up thereafter until the child is 36 months old. This growth faltering relative to standards could be related to the introduction of solid foods, likely to be less than adequate, the accompanying infections introduced through water and food, and the withdrawal of breast feeding. Being a female child is associated with lower weight. Analysis of standard weight indicates that females are further away from their age-sex specific NCHS median than males are from theirs. However, the absolute difference between males and females becomes less over time. In Indian data gender bias against females has been reported in the case of infant mortality rates and in the masculine sex ratio in the population. Differences by sex in anthropometric outcomes have not been well established. Outside of Asia the differences have been observed to be small and often insignificant. The present study shows the prevalence of significant male-female differences in weight outcomes, with 189 the difference becoming smaller in current data. This reduction in weight differences could be attributed to the negative selection for supplementary feeding by the program. Since females tend to be in grades IQ and IV more often than males, they would tend to be selected more often. Birth order is found to have negative and highly significant effects on weight and standard weight. This could be due to economic strain resulting from a larger number of children, maternal depletion resulting in lower birth weights for the later bom, increased probability of infections due to a higher proportion of children in the household, or even cultural preferences for earlier birth orders. Empirical evidence for inequalities among siblings by birth order in respect of schooling attainments exists. For anthropometric outcomes Horton (1988) finds negative birth order effects using Phillippine data. The results of this study are similar. The effects of five household level variables: household economic status, maternal education, maternal height, caste and household structure were examined. A dear positive assodation is observed between household economic status as measured by land operated and weight outcomes. For base, sibling and current data children from landless households tend to be the worst off in terms of weight and standard weight. In the literature household economic status has not always been found to make such a significant difference. Particularly for the preschool ages, maternal characteristics like height and education have turned out to be more important. But in another study based on Indian data (Sen and Sengupta, 1983) and in Strauss (1990) using data from Cote d’ Ivoire, household economic status has been found to have significant positive effects. Maternal education has positive and significant effects on weight and standard weight. In the early program months more than 5 years of maternal schooling is 190 required for significant increases in standard weight (base line data). Over time, as the program is better established, children of mothers with even some schooling (less than 5 years) show significant improvements over children of illiterate mothers, with the strongest effects for mothers with over 5 years schooling. Better educated mothers are more likely to seek out information about health care and nutrition and act on it. This is specially important in a situation where environmental sanitation and protected water supply cannot be taken for granted. In the present situation maternal education could promote the adoption of practices suggested by the program. It has been argued in Wolfe and Behrman (1987) that controlling other family background variables, the effect of maternal schooling gets eliminated. Information on mother’s sibling and her parents is used to capture family background. In this study, however, maternal schooling has been found to be of considerable significance. Here maternal height has been used to capture family background characteristics, in addition to genetic factors. Mother’s height is found to be a positive and significant predictor for weight and standard weight. It is used as a prosy for her genetic and health endowments and also other family background characteristics not captured by the education variable. This is in keeping with other studies. Belonging to one of the scheduled castes is taken as an indicator of low social status. Not belonging to a scheduled caste is found to have positive and significant effects on weight outcomes. This could also be capturing another dimension of household economic status. Whether children of household heads tend to do better than others was examined using a dummy for father being the head of household. It turned out that not being the child of the household head was associated with higher weight and standard 191 weight. A closer scrutiny showed that not being the child of the household head was indicative of belonging to an extended household in the present data set. In such households either a grandparent of the child, the father’s older brother, or an older male relative would be considered the household head. The result indicates that as compared with children from nuclear households, children from joint families tend to have higher weight outcomes. Pooling of resources prevalent in such joint families is likely to be particularly beneficial by reducing seasonal fluctuations in income streams in rural areas with beneficial effects on weight. In the literature there is not much evidence on the effects of household structure on child health outcomes. Strauss (1990) has examined this issue and he finds that being the child of the household head is beneficial. The present result contrasts with such a conclusion indicating the strong positive effects of belonging to a joint family. At the community level variations in weight outcomes with region as captured by development block, distance to the nearest health facility and hospital, quality of water supply and the existence of pucca road connections were studied. The results show that a block level index of economic development does not systematically explain child weight outcomes. This contrasts with the result in respect of household level economic status. While other explanations cannot be ruled out, it is possible that blocks differ in their social characteristics - for example, caste structure. Further, the duration of program operation in the region is found to vary positively with weight and standard weight. It is of interest to note that at a higher level of regional aggregation, taking district instead of block as the regional unit, level of economic development is systematically related to weight outcomes. I i 192 Distances to the nearest health facility and general hospital were found to be negatively (not always significantly) associated with weight outcomes. Distance contributes to underutilization with possible adverse effects upon weight and standard Weight. Water quality was found to have positive and very significant effects upon weight. Unhygienic water is a recognized as a source of intestinal infections. Unlike in the DCs, physical growth and infections have been found to be closely related in data from LDCs. To the extent unhygienic water contributes to infections it can be expected to result in lower weight outcomes. For villages connected by a pucca road higher weight outcomes were found as against villages not so connected. The existence of a pucca road promotes integration with the rest of the economy and availability of services in the area resulting in better overall development. Third, does the male-female gap differ significantly across households of varying economic status and social status? To answer this interaction effects between the variables sex and land operated, and between sex and whether the child belongs to a scheduled caste household were allowed and regression equations reestimated for standard weight. Economically and socially better off households could be expected to have lower male-female differences if resource crunch was the primary cause. Contrary to such an expectation, we find that as compared with children of landless households, the male- female gap is higher in the landed households. Thus, while mean standard weights of male children are higher in better landed categories, female means are not necessarily higher, indicating that improved economic status does not necessarily reduce sex differences. Since social status was expected to influence the male-female weight gap, differences by 193 scheduled caste versus other households were examined. We find a significantly higher male-female gap among the socially better off non-scheduled caste households. Two possible explanations that could contribute to a relatively lower economic value of a female child and to looking upon the birth of a daughter as a burden among the better off upper classes are suggested. First, unlike in the landless or scheduled caste households where all adults tend to be economically active, among households of relatively higher economic and social status female members tend not to work for a wage due to prestige related reasons. Second, the practice of dowries causing parents of females to bear large financial outflows at the time of marriage is much more prevalent among the upper classes. Both reasons contribute to the relatively lower economic value of a female child among these classes, making her bringing up an expensive proposition for the natal household. Such factors could influence the quality of child rearing of females relative to males, resulting in lower female weight outcomes than necessitated by household economic status. Fourth, does maternal education and household economic status counter the observed negative effects of higher birth order? To answer this regressions for standard weight are reestimated allowing for interactions between birth order and maternal education level, and between birth order and land operated by the household. Thus, we look for differences in the response coefficients (slope differences) of birth order by levels of maternal education and by household economic status. The results show that maternal education does compensate for negative birth order effects. This conclusion is strongest for the base line data when program effects are minimum and continues to hold good for sibling data. The compensatory effect of maternal schooling gets reduced for current data after 6 to 8 years of program operation. Perhaps, over time the program itself 194 is an alternate source of education for mothers with reference to health care and nutrition. Of course, intercept differences due to maternal education are independently relevant. Unlike maternal education, the compensatory effects of land operated are not dear cut. Household economic status, while resulting in higher intercepts for weight outcomes, does not seem to systematically counter the negative effects of higher birth order. This result suggests that inadequacy of resources is not a major factor resulting in the negative relationship between weight and birth order. Factors that can be influenced by maternal education through the adoption of better health care practices are likely to be more important - factors like maternal depletion, greater probability of infections due to a higher proportion of children in the household, and even preference for earlier birth orders. 9.2 Polity Implications The historical experience of countries shows that long established differences in body size between dasses have either disappeared or are fast disappearing - probably for the first time in documented history. This narrowing of differences resulted not from a convergence of means between the better off and the worse off groups to some middle level, but rather from the increase in the means of the lower groups (Tanner, 1978; Bielicki, 1986). In the present case better childhood environments at the household and community level (higher economic and social status, higher levels of maternal education, better water quality, nearness to health and hospital facilities, etc.) are assodated with higher weight outcomes, indicating that improvements in body size are positively assodated with economic and sodal development. But the lack of an upward trend in all-India data 195 and also in data from Tamilnadu as a whole (Ganguly, 1979; NNMB, Hyderabad data), i.e., in non-program areas, indicate that narrowing of differences will perhaps cover a time period considered long by modem standards. It is here that special health and nutrition interventions together with improvements in environmental factors like water quality have a role to play in accelerating the process. To the extent governments are concerned not only with agricultural, industrial and trade policies, but also with broader development objectives encompassing living standards, social modernization, investment in human capital, the empirical findings have important policy implications. (i) Direct interventions in the health sector in the form of health and nutrition programs are beneficial in many ways and should be promoted. They can accelerate the process of narrowing of differences in body sizes between classes. Moreover, to the extent such interventions select the relatively worst off individuals with respect to their physical condition they have the added advantage of narrowing down systematic differences even within a household (for eg., between sexes and birth orders). The effect of narrowing differences even within households can be particularly important where household behavior results in relatively greater deprivations for some of its members, like females or the later bom. In situations where literacy levels are low, especially among females, health and nutrition programs can themselves serve as an alternate source of education with positive effects upon health outcomes. (ii) In making investments in human capital in the health sector when resources are limited, it is economical to direct interventions at the early phases of human growth and development, when growth is most rapid and the consequences of sustained deprivations largely irreversible, i.e., until early childhood. The more popular feeding programs in schools, while relatively easier to organize (since children assemble regularly 196 in school anyway), and much more visible, intervene when it is too late for an important reason - considerable irreversible growth retardation is already likely to have occurred. (iii) Public investments in water quality are recommended. We observed that community improvements in quality of drinking water make a significant difference to child weight through decreasing the incidence and severity of intestinal infections. The effects of water quality on mortality are already well established. Such investments have the additional advantage that they require little cooperation by the household in terms of time or effort required for utilization. Further, they are neutral across sexes and birth orders. 9.3 Limitations and Further Research This study uses information on weight-for-age of preschoolers. Since no information on child heights was available analysis of weight-for-height was not possible. Weight-for-height has the advantage of combining information on two types of anthropometric measurements. It is often preferred for being nearly age independent and is especially useful in communities where child age is not accurately known. Comparing the results based upon weight-for-age with height-for-age and weight-for-height would have been of considerable interest. However, the measurement of height among very young children, especially those under two years of age who cannot quite stand erect, tends to be error prone except under hospital or laboratory conditions. Supine length instead of erect height is usually the preferred measure for this age group1. Due to 1 But if some children are measured in the supine position (the youngest ones) and others in an erect position (the older ones), there is the problem of combining length and height data in view of differences between the two types of measurement. operational difficulties in obtaining accurate and regular height/length measurements under rural field conditions for a very large number of children these measurements were discontinued very early by the program management. Hence the study had to be restricted to the analysis of weight-for-age. Chapter II contained a more detailed discussion on this. All information in the TTNP sample relates to children participating in the program. Usually this creates problems of selectivity bias. But because of high participation rates, the problem is not serious in the present case. However, a related limitation is the difficulty of having a good control group of non-participants to evaluate program effects more precisely. Birth order is used as a regressor in preference to family size or the number of children, since, unlike birth order, the latter are choice variables at the household level and cannot be properly considered as exogenous. Significant negative effects of birth order on child weight are observed. However, it can be argued that birth order is also endogenous since higher birth order children occur in households which have decided to have a larger number of children. Regressions dropping birth order were also run but the effects of other explanatory variables were substantially unchanged. An important factor not considered in the above study is the effect of seasonality. Individuals, specially children, are not at equal risk in all seasons. Crop failures and wet seasons have been identified as two sources of exceptional stress (Chambers, 1982; Lipton, 1983). Proportions of children with inadequate weight for age and those presenting at health centers and hospitals with severe wasting or stunting vary seasonally, tending to rise in the later wet seasons. Similarly, greater proportions of such cases are observed during periods of unexpected stress induced by crop failures. Using rainfall data and/or monthly or quarterly prices of primary foodgrains to examine seasonality could 198 provide additional insight. Inadequacy of resources and time prevented explorations of seasonality. This could be the subject of future research. The UNP data being exclusively rural, rural-urban comparisons are precluded. Especially for child weights, the better infrastructure facilities in urban areas available even to the urban poor (eg., protected water, electricity, better hospital facilities and com m unication) would have made possible inferences about the relative importance of urb anisation and economic status versus community level variables. Rural-urban comparisons were, regretfully, outside the scope of the present study but such an analysis would be of interest in a future study. Also of considerable interest for future research would be the analysis of male-female differences in other cultural contexts. For example, the analysis of sex differences by economic status in the Matlab data would give further insight into the generalizability of the results of this study. All through the study the assumption has been that higher weights are better since the complication of obesity and the emphasis on weight control does not exist in rural Tamilnadu, and, more generally, among the poor in many LDCs. Yet, there is no reason to think that weights (so also heights and other signs of growth and acceleration in maturation) are good in themselves. It is only because higher weight outcomes and other signs of growth are associated with improvements in health and the overall biological condition of children and, therefore, in future, of adults, that anthropometric outcomes are valued. Data on older children from Belgium (Renson et al., 1983), on conscripts from Poland (Brajczewski, 1985), children from Czechoslovakia (Parizkova and Berdychova, 1977) and from Canada (Jequier, 1979) show that while larger body sizes were generally associated with better socioeconomic backgrounds or urbanization, they were also 199 associated with inferior performance in certain functional traits like absolute grip strength, grip strength for weight, stroke volume, pulse rate after exercise, etc. These findings may serve as a reminder that whatever makes children from better environments grow more steadily and faster does not necessarily make them more fit physiologically or even more likely to achieve fuller expressions of intellectual or cultural potential. However, in many societies such associations do still obtain and, in spite of important exceptions, differences in childhood growth are associated with differential morbidity, mortality, and physiological functioning capacity (Tanner, 1982; Bielicki, 1986). APPENDIX 201 Table 1 M E A N W E IG H T S A N D S D B Y A G E , P O O L E D (3) T IN P S A M P L E Current Data Si bling Data Base Data A G E IN N M E A N S D N M E A N S D N M E A N S O M O N TH S (KGS.) (KGS.) (KGS.) 6 317 5.89 0.68 139 5.83 0.79 20 5.56 1.03 7 358 6.04 0.71 174 5.98 0.77 30 5.97 0.84 8 380 6.26 0.79 184 6.18 0.79 46 6.10 0.92 9 386 6.45 0.83 189 6.40 0.85 50 6.21 0.92 1 0 387 6.62 0.84 197 6.58 0.87 67 6.37 0.91 1 1 397 6.80 0.89 202 6.75 0.87 78 6.48 0.94 1 2 404 7.00 0.91 209 6.95 0.95 102 6.82 1.02 13 407 7.17 0.93 212 7.12 0.95 110 6.90 1.02 14 410 7.31 0.91 214 7.26 0.96 128 7.03 0.98 15 409 7.47 0.91 220 7.38 0.97 140 7.22 0.95 16 412 7.60 0.94 227 7.56 0.96 150 7.33, 0.92 17 414 7.74 0.97 223 7.68 0.95 154 7.45 0.96 18 413 7.89 0.96 224 7.82 0.97 186 7.62 0.91 19 407 8.07 0.98 228 7.99 0.96 187 7.78 0.94 20 394 8.21 1.10 231 8.12 1.02 2 2 1 7.92 0.94 2 1 369 8.33 1.10 228 8.26 1.02 226 8.07 1.00 22 337 8.50 1.04 235 8.42 0.99 231 8.27 1.03 23 324 8.67 1.01 238 8.56 1.00 236 8.37 1.03 24 303 8.76 1.02 239 8.68 1.00 278 8.61 1.11 25 287 8.86 1.02 245 8.83 1.00 284 8.73 1.13 26 265 9.03 1.00 245 9.02 1.06 310 8.88 1.14 27 255 9.21 1.03 244 9.19 1.05 309 9.04 1.15 28 237 9.35 1.04 244 9.32 1.04 318 9.19 1.16 29 212 9.49 1.08 248 9.45 1.09 315 9.37 1.17 30 188 9.65 1.01 249 9.62 1.05 341 9.49 1.22 31 162 9.80 1.09 245 9.76 1.02 346 9.67 1.20 32 136 9.96 1.09 249 9.91 1.05 375 9.82 1.17 33 113 10.13 1.14 244 10.08 1.06 368 9.98 1.19 34 86 10.47 1.09 244 10.26 1.05 374 10.15 1.21 35 66 10.63 1.10 245 10.42 1.03 368 10.31 1.19 36 37 10.68 1.28 246 10.63 1.03 378 10.51 1.22 Total 9272 6961 6726 a: Both sexes. 202 Table 2 FEMALE MEAN WEIGHTS AND SD BY AGE, TINP DATA Current Data Sibling Data Base Data AGE IN | MONTHS j N MEAN (KGS.) SD N MEAN (KGS.) SD N MEAN (KGS.) SD 6 | 167 5.82 0.67 79 5.78 0.72 12 5.11 0.89 7 j 188 5.93 0.73 95 5.89 0.78 19 5.64 0.70 8 j 196 6.11 0.77 100 6.08 0.78 30 5.78 0.84 9 j 199 6.27 0.84 99 6.27 0.84 32 5.90 0.81 10 j 202 6.42 0.85 107 6.42 0.88 40 6.17 0.80 n j 209 6.61 0.87 108 6.59 0.90 46 6.23 0.80 12 I 216 6.82 0.90 113 6.74 0.99 59 6.59 0.84 13 | 216 6.99 0.95 111 6.86 0.96 64 6.67 0.88 14 j 221 7.15 0.95 114 7.03 1.00 71 6.89 0.88 15 | 216 7.32 0.92 117 7.12 1.02 78 7.07 0.87 16 j 219 7.43 0.90 123 7.35 1.02 86 7.18 0.87 17 j 221 7.58 0.91 119 7.45 0.96 90 7.30 0.94 18 j 222 7.71 0.95 116 7.61 0.97 105 7.46 0.82 19 | 218 7.88 0.93 120 7.81 0.98 105 7.66 0.89 20 j 211 8.03 0,95 121 7.95 1.03 124 7.70 0.90 21 i 199 8.16 1.00 118 8.02 1.04 126 7.82 0.92 22 j 179 8.28 1.02 122 8.25 0.99 129 8.06 0.93 23 j 171 8.46 0.99 124 8.40 0.99 133 8.17 0.95 24 j 165 8.57 0.97 123 8.53 1.00 154 8.42 1.03 25 | 158 8.67 0.96 128 8.69 1.00 157 8.55 1.04 26 j 143 8.86 0.94 127 8.90 1.08 169 8.75 1.06 27 j 137 9.07 0.99 125 9.05 1.09 169 8.91 1.09 28 | 128 9.22 0.99 124 9.19 1.08 176 9.04 1.11 29 j 115 9.36 1.04 127 9.33 1.11 172 9.19 1.18 30 j 99 9.56 1.00 128 9.51 1.10 179 9.34 1.22 31 i 83 9.67 1.06 126 9.65 1.04 180 9.47 1.18 32 | 70 9.89 0.91 128 9.79 1.08 195 9.65 1.14 33 j 54 10.04 0.94 124 9.96 1.06 193 9.81 1.14 34 | 40 10.34 0.98 125 10.13 1.06 195 9.94 1.17 35 | 32 10.56 1.04 128 10.30 1.05 192 10.14 1.20 36 16 10.27 1.32 128 10.50 1.04 195 10.37 1.98 Total | 4910 3647 3675 203 Table 3 M A L E M E A N W E IG H T S A N D S D B Y A G E , T IN P D A T A Current Data Sibling Data Base Data A G E IN M O N TH S N M E A N (ICGS.) S D N M E A N (KGS.) S D N M E A N (KGS.) S D 6 150 5.96 0.68 60 5.91 0.87 8 6.25 0.85 7 170 6.16 0.67 79 6.08 0.76 1 1 6.52 0.79 8 184 6.42 0.78 84 6.30 0.79 16 6.70 0.76 9 187 6.63 0.79 90 6.55 0.83 18 6.76 0.86 1 0 185 6.83 0.77 90 6.77 0.82 27 6.66 1.01 1 1 188 7.03 0.77 94 6.94 0.80 32 6.83 1.02 1 2 188 7.22 0.82 96 7.19 0.83 43 7.13 1.16 13 191 7.38 0.83 1 0 1 7.40 0.84 46 7.22 1.13 14 189 7.51 0.86 100 7.52 0.84 57 7.21 1.07 15 193 7.63 0.87 103 7.67 0.82 62 7.40 1.03 16 193 7.80 0.88 104 7.80 0.82 64 7.52 0.96 17 193 7.93 0.93 104 7.94 0.86 64 7.66 0.94 18 191 8.10 0.95 108 8.04 0.91 81 7.84 0.97 19 189 8.29 0.94 108 8.19 0.91 82 8.01 0.95 20 183 8.41 0.97 110 8.30 0.97 97 8.18 1.05 2 1 170 8.52 1.18 110 8.47 0.96 100 8.37 1.02 22 158 8.74 1.02 113 8.61 0.96 102 8.53 1.09 23 153 8.90 0.99 114 8.72 0.99 103 8.61 1.07 24 138 8.98 1.03 116 8.85 0.97 124 8.84 1.16 25 129 9.09 1.05 117 8.98 0.98 127 8.94 1.19 26 122 9.22 1.05 118 9.15 1.02 141 9.03 1.21 27 118 9.36 1.06 119 9.34 0.99 140 9.17 1.21 28 109 9.51 1.09 120 9.45 0.98 142 9.36 1.21 29 97 9.64 1.10 1 2 1 9.57 1.05 143 9.58 1.13 30 89 9.74 1.01 1 2 1 9.73 0.99 162 9.67 1.20 31 79 9.94 1.10 119 9.88 0.99 166 9.89 1.18 32 66 10.04 1.26 1 2 1 10.04 1.02 180 10.01 1.17 33 59 10.22 1.29 1 20 10.20 1.05 175 10.16 1.21 34 46 10.59 1.17 119 10.40 1.02 179 10.38 1.21 35 34 10.71 1.17 117 10.55 1.00 176 10.47 1.17 36 2 1 11.00 1.18 118 10.77 1.03 183 10.66 1.23 Total | 4362 3314 3051 Table 4 FEMALE AND MALE MEAN WEIGHTS BY AGE FOR YEARS BETWEEN 1974 & 1982: NNMB DATA, TAMILNADU Females Males Year Age (years) No. of obs. Mean (kgs.) Weight sd No. of obs. Mean (kgs.) Weight sd 1979 0 51 6.75 0.56 46 6.39 0.86 1980 0 57 6.20 1.27 53 6.10 1.36 1981 0 36 5.60 1.40 35 6.50 1.49 1982 0 60 6.00 1.30 70 6.50 1.50 1974 1 27 7.80 N.A. 34 7.80 N.A. 1976 1 42 7.90 0.97 41 8.60 1.28 1978 1 49 7.70 1.15 67 8.40 1.35 1979 1 32 9.02 0.70 36 9.07 3.41 1980 1 48 8.00 1.30 58 8.30 1.45 1981 1 43 7.40 1.34 38 8.20 1.24 1982 1 50 7.70 1.35 55 7.70 1.36 1974 2 37 9.70 N.A. 27 9.50 N.A. 1976 2 35 9.10 1.39 34 9.90 1.78 1978 2 69 9.30 1.47 68 9.60 1.55 1979 2 53 9.03 1.50 40 9.49 0.63 1980 2 64 9.40 1.26 52 9.80 1.57 1981 2 33 8.20 1.79 52 9.70 1.61 1982 2 60 8.90 1.55 69 9.90 1.63 Source: NNMB, National Institute of Nutrition, Hyderabad. N.A. stands for 'not available'. 205 T a b le 5 MEAN WEIGHTS BY YEARLY AGE GROUPS: POOLED TINP DATA, 1980 - 1989. Age Group (Months) Age (Year) Year Total N Mean Weigh (Kgs.) < 12 0 1980 46 5.93 < 12 0 1981 35 6.36 < 12 0 1982 142 6.12 < 12 0 1983 265 6.20 < 12 0 1984 293 6.29 < 12 0 1985 381 6.36 < 12 0 1986 355 6.14 < 12 0 1987 1399 6.37 < 12 0 1988 681 6. 43 < 12 0 1989 9 7.30 12<=Age<24 1 1980 179 7. 50 12<=Age<24 1 1981 417 7.99 12<=Age<24 1 1982 439 7.52 12<=Age<24 1 1983 1133 7.55 12<=Age<24 1 1984 595 7.77 12<=Age<24 1 1985 834 7.87 12<=Age<24 1 1986 860 7.94 12<=Age<24 1 1987 1539 7 .57 12<=Age<24 1 1988 2942 7.82 12<=Age<24 1 1989 525 8.41 24<=Age<=36 2 1980 195 9.25 24<=Age<=36 2 1981 789 9. 69 24<=Age<=3 6 2 1982 795 9. 81 24<=Age<=36 2 1983 1500 9.41 24<=Age<=3 6 2 1984 1447 9. 51 24<=Age<=3 6 2 1985 793 9. 71 24<=Age<=36 2 1986 925 9. 67 24<=Age<=36 2 1987 1017 9.81 24<=Age<=36 2 1988 1585 9. 34 24<=Age<=36 2 1989 1094 9.65 T o t a l : 2 3 2 0 9 Tabte 6 206 FEM A LE & M A LE M E A N WEIGHTS B Y YEARLY A G E GROUPS: TINP DATA, 1980 - 1989. Females Males Age Group Age Year (Months) (Year) Total Mean Weight N (Kgs.) TOTAL M E A N WEIGHT N (Kgs.) < 12 0 1980 32 5.49 14 6.92 < 12 0 1981 22 5.93 13 7.09 < 12 0 1982 78 6.02 64 6.24 < 12 0 1983 138 5.94 127 6.49 < 12 0 1984 138 6.08 155 6.47 < 12 0 1985 207 6.25 174 6.49 < 12 0 1986 202 6.03 153 6.28 < 12 0 1987 763 6.21 636 6.56 < 12 0 1988 348 6.27 333 6.60 < 12 0 1989 5 7.29 4 7.31 12<=Age<24 1 1980 91 7.46 88 7.54 12<=Age<24 1 1981 237 7.61 180 8.42 12<=Age<24 1 1982 224 7.31 215 7.76 12<=Age<24 1 1983 657 7.36 476 7.82 12<=Age<24 1 1984 315 7.50 280 8.08 12<=Age<24 1 1985 391 7.61 443 8.10 12<=Age<24 1 1986 463 7.63 397 8.30 12<=Age<24 1 1987 835 7.42 704 7.75 12<=Age<24 1 1988 1612 7.63 1330 8.06 12<=Age<24 1 1989 266 8.25 259 8.58 24<=Age<=36 2 1980 99 9.25 96 9.26 24<=Age<=36 2 1981 385 9.57 404 9.81 24<=Age<=36 2 1982 423 9.54 372 10.01 24<=Age<=36 2 1983 738 9.28 762 9.53 24<=Age<=36 2 1984 827 9.28 620 9.83 24<=Age<=36 2 1985 400 9.53 393 9.90 24<=Age<=36 2 1986 445 9.41 480 9.92 24<=Age<=36 2 1987 521 9.52 496 10.10 24<=Age<=36 2 1988 861 9.23 724 9.48 24<=Age<=36 2 1989 591 9.45 503 9.90 T otal: 12314 10895 207 REFERENCES Adelman, Irma and Cynthia Taft Morris, 1973. Economic growth and social equity in developing countries. Stanford, California: Stanford University Press. S Aguillon et al., 1982. "The relationship of family characteristics to the nutritional status of preschool children.1 ' Food and Nutrition Bulletin Vol. 4, no. 4. Ahluwalia, M., 1976. "Inequality, poverty and development." Journal of Development Economics 3:307-42. , 1978. India: occasional papers. World Bank Staff Working Paper No. 279, Nov. Ahluwalia, M. et al., 1979. Growth and poverty in developing countries. World Bank Staff Working Paper no.309 (revised), Washington,D.C. s Akin, John, S., and David Guilkey, 1990. "The production of infant health: input demand and health status differences related to gender of the infant." Mimeo, University of North Carolina. ^ Anderson, Mary Ann, 1979. "Comparison of anthropometric measures of nutritional status in preschool children in five developing countries.” American Journal of Clinical Nutrition vol. 32, 2339-45. Bardhan, P.K., 1970. "On the minimum level of living and the rural poor.” Indian Economic Review 5, no. 1, April. , 1971. "On the minimum level of living and the rural poor: a further note." Indian Economic Review 6, no. 1, April. , 1974. "On life and death questions." Economic and Political Weekly Aug., vol. 9. , 1984. Land, labor and rural poverty. Columbia University Press, New York. , 1988. "Sex disparities in child survival in rural India." in Srinivasan, T.N. and Bardhan, P.K., eds. Rural poverty in South Asia. Columbia University Press. Barrel, RA.E. and M.G.M. Rowland, 1979. "Infant foods as a potential source of diarrhoeal illness in rural West Africa." Trans. Rov. Soc. Trop. Med Hvg. 73:85- 90. v' Barrera, Albino, 1987. "The role of maternal schooling and its interaction with public health programs in child health production." Mimeo, Yale University. ✓Batliwala, Srilatha, 1981. "Rural energy scarcity and nutrition: a new perspective." Economic and Political Weekly. 17(9):329-33. 208 y Battad, Josephine, 1978. "Determinants of nutritional status of preschoolers." Philippine Economic Journal vol. 17, nos. 1 & 2, 154-167. Beaton, G.H. & H. Ghassemi, 1979. Supplementary feeding programs for voung children in developing countries. UNICEF/UN, New York. y ------ 1 1982. "Supplementary feeding programs for young children in developing countries." Amer. J. Clin. Nutr. 35:864. y Behrman, J. and A. Deolalikar, 1985. "How do food and product prices affect nutrient intakes, health and labor force behavior for different family members in rural India?" Processed. Department of Economics, University of Pennsylvania. y ------ 1 1988. "Health and Nutrition." in Chenery, H. and T.N. Srinivasan eds., Handbook of Development Economics. Amsterdam, North Holland. ^ Behrman, J. and P. Taubman, 1986. "Birth order, schooling and earnings." Journal of Labor Economics 4:121-45. ^Behrman, J. and B. Wolfe, 1987. "How does mother’s schooling affect family health, nutrition, medical care usage and household sanitation?" Journal of Econometrics 36:185-204. v' Beisel, W.R., 1977. "Magnitude of host nutritional responses to infection." Amer. J. Clin.Nutr. 30, pp 1236 - 1247. Berg, A., 1973. The nutrition factor: its role in national development. Washington, DC: Brookings Institution. •s 1 1981. Malnourished people: a policy view. WB Poverty and Basic Needs Series, The World Bank, Washington DC. ^Berg, A. & JAustin, 1984. "Nutrition policies and programs: a decade of redirection." Food Policy 9, 4. Bergman, P. and M. Goray, 1984. "The timing of adolescent growth spurts of ten body dimensions in boys and girls of the Wroclaw longitudinal twin study." J. Hum. Evol. 13:339-437. Beteille, A., ed., 1969. Social Inequality. Penguin, Harmondsworth. , 1974. Studies in Agrarian Social Structure. Oxford University Press, Delhi. Bhan, M.K., and Shanti Ghosh, 1986. Successful growth monitoring: some lessons from India. UNICEF, New Delhi. /Bielicki, T., 1986. "Physical growth as a measure of the economic well-being of populations: the twentieth century." in Falkner, F. and J.M. Tanner eds., Human 209 growth: a comprehensive treatise, vol. 3, 2nd edition, Plenum Press, New York and London. Birdsall, N., 1979. "Siblings and schooling in urban Colombia." unpublished Ph.D. diss., Yale University. v/Black, R.E., K.H. Brown, S. Becker, A. Alim and M.H. Merson, 1982. "Contamination of weaning foods and transmission of enteropathogenic escherichia coli diarrhoea in children in rural Bangladesh." Trans. Rov. Soc. Trop. Med. Hvg. 72:259-264. & '■'''Blau, D., 1984. "A model of child nutrition, fertility and women’s time allocation: the case of Nicaragua." in T.P. Schultz and K. Wolpin eds., Research in population economics. Greenwich, Conn.:JAI Press. Brajczewski, C., 1985. "Spoleczne roznice w budowie fizycznej i wydolnosci mlodych mezczyzn." unpublished manuscript. s/ Brown, K.H., R.E. Black, E. Robert and S. Becker, 1982. "Seasonal changes in nutritional status and the prevalence of malnutrition in a longitudinal study of young children in rural Bangladesh." Amer. J. Clin. Nutr. August, 36:303-313. Brundtland, G.H., K. Liestol, and L. Walloe, 1980. "Height, weight and menarcheal age of Oslo schoolchildren during the last 60 years." Ann. Hum. Biol. 7:307-322. v Buschang, P.H. and R. Malina, 1983. "Growth in height and weight of mild to moderately undernourished Zapotec school children." Human Biol. 55:587-597. Caldwell, J.C., 1979. "Education as a factor in mortality decline: an examination of Nigerian data.” Population Studies 33:395-413. Carloni, Alice Stewart, 1981. "Sex disparities in the distribution of food within rural households." Food and Nutrition 7:3-12. ^/Cassidy, Claire Monod, 1980. "Benign neglect and toddler malnutrition." in Greene, L.S. and F.E. Johnston, eds. Casterline, J., E. Cooksey and A.F. Ismail, 1987. "Household income and child survival in Egypt." Processed, Population Studies Center, Brown University. Chambers, R. et al.,1979. "Seasonal dimensions to rural poverty: analysis and practical implications." Journal of Tropical Medicine and Hygiene, vol. 82, no. 8, August, 156-171. v Chambers, R., 1982. "Health, agriculture and rural poverty: why seasons matter." Journal of Development Studies vol. 18, no. 2, Jan. '''Chandra, R.K., 1980. Immunology of nutritional disorders. London: Edward Arnold. 2 1 0 Chaudhury, Maitreyi, 1984. "Malnutrition of rural children and the sex bias." Economic and Political Weekly May, 25th. vChen, L.C., A.K.M. Chowdhury and S.L. Huffman, 1980. "Anthropometric assessment of energy protein malnutrition and subsequent risk of mortality among pre-school aged children." Amer. J. Clin. Nutr. 33 pp 1836-45. Chen, L., E. Huq, and S.D’ Souza, 1981. "Sex bias in the family allocation of food and health care in rural Bangladesh." Population and Development Review 7(l):55-70. Chen, L. et al, 1982a. "Anthropometric assessment of energy protein malnutrition and subsequent risk of mortality among pre-school aged children." in Sukhatme, P.V., ed. Newer concepts in nutrition and their implications for policy. , 1982b. "Where have the women gone? Insights from Bangladesh on low sex-ratio of India’s population." Economic and Political Weekly March 6th. v'Chen, L. and N.S. Scrimshaw, eds., 1983. Diarrhea and malnutrition: interactions, mechanisms and interventions. Plenum Press, New York. /Chemichovsky, Dov, 1979. "The economic theory of the household and impact measurement of nutrition and related health programs." In Klien, R. et al., eds., Evaluating the impact of nutrition and health programs. New York: Plenum Publishing. Chemichovsky, D., and D. Coate, 1983. "An economic analysis of the diet, growth and health of children in the United States." in Salkever, D., I. Sirageldin and A. Sorkin, eds., Research in human capital and development, vol. 3, Greenwich, CT:JAI Press. Chemichovsky, D., N.S. Kielman, A. Kielman and W. Rienke, 1983. "Correlates of preschool child growth in rural Punjab." in Child and maternal health services in rural India: the Narangwal experiment, eds. A. Kielman and Associates, Baltimore: Johns Hopkins University Press. Chidambaram, G., 1989. Tamilnadu Integrated Nutrition Project - terminal evaluation. Government of Tamilnadu, Madras. Coreil, Jeannine, 1983. "Allocation of family resources for health care in rural Haiti." Social Science Medicine 17(11). Dandekar, K., 1975. "Why has the proportion of women in India’s population been declining?" Economic and Political Weekly Oct. 18th. '/Dandekar, V.M., 1982. "On measurement of undemutrition" Economic and Political Weekly Feb. 6th. S Dasgupta, P and D. Ray, 1986. "Inequality as a determinant of malnutrition and j 211 unemployment: theory." The Economic Journal No. 384, vol. 96, December. y , 1990. "Adapting to undernourishment: the clinical evidence and its implications." Forthcoming in Poverty and hunger: the poorest billion, eds. Dreze, J., and A. Sen, Oxford:OUP. •/ Dasgupta, Partha, 1991. "Nutrition, nonconvexities and redistributive policies." The Economic Journal. Vol. 101, no. 404, January. s/Dasgupta, Rajaram, 1984. "Nutritional situation in India." Economic and Political Weekly August 25th. Das, V. and Nicholas, R., 1981. "’ Welfare, and ’ well-being’ in South Asian societies." Mimeographed. New York: ACLS-SSRC Joint Committee on South Asia. Deaton, A., 1982. "Inequality and needs: some experimental results for Sri Lanka." Population and Development Review 8 (Supplement):35-52. , 1987. "The allocation of goods within the household: adults, children and gender." Working paper, Princeton University. Delgado, D.L. et al., 1986. "Effects of health and nutrition interventions on infant and child mortality in rural Guatemala." in United Nations, Determinants of mortality change and differentials in developing countries. New York. den Hartog, A.P., 1973. "Unequal distribution of food within the household." FAO Newsletter (Oct-Dec) 10:4. , 1982. "Unequal distribution of food within the household (a somewhat neglected aspect of food behavior)." Nutrition Newsletter 10(4):8-17. Dixon, Suzanne D., Robert A. LeVine and Berry T. Brazelton, 1982. "Malnutrition: a closer look at the problem in an East African village." Develop. Med. Child Neurol. 24:670-685. \/ Dowler, E A , Payne, P.R., Seo, Y.O., Thomson, A.M., and Wheeler, EA., 1982. "Nutrition, status indicators: interpretation and policy making role.” Food Policy 7, pp 99-112. Dreze, J., and A. Sen, 1989. "Hunger and public action.: Oxford: Clarendon Press. Dugdale, A., and Payne, P., 1977. "Patterns of lean and fat deposition in adults." Nature 266, No. 5600. Dyson, T. and Moore, 1983. "On kinship structure, female autonomy, and demographic behavior in India." Population and Development Review 9, pp 35-60. 212 Easterlin, R A , 1974. "Does economic growth improve the human lot?" in PA. David and M.W. Reder, eds., Nations and households in economic growth. Academic Press, New York. Evans, M.E., 1944. "Illness history and physical growth. I. A comparative study of the rate of growth of pre-school children of five health classes." American Journal of Diseases of Children 68, 390 - 394. Eveleth, P.G. and J.M. Tanner, 1976. Worldwide variations in human growth. Cambridge University Press, Cambridge. Evenson, Robert, Barry Popkin and Elizabeth King-Quizon, 1980. "Nutrition, work and demographic behavior in rural Philippine households: a synopsis of several Laguna household studies." Ch.l in Binswanger, H., et al. eds., Rural household studies in Asia. Singapore: Singapore UnivJPress. Falkner, F., and J.M. Tanner eds., 1986. Human growth: a comprehensive treatise, vol.3, 2nd edition, New York: Plenum Press. FAO/WHO Expert Committee, 1973. Energy and protein requirements. FAO Nutrition Meeting Report Series No. 52; WHO Technical Report Series No. 522. FAO, 1983. World food survey. Fields, Gary, S., 1980. Poverty, inequality and development." Cambridge University Press. Field, John, O., R.I. Miller, and W.D. Drake, 1981. Kottar: Malnutrition, intervention, and development in a south Indian district. USAID, Washington DC. Fischbein, S., 1977. "Intra-pair similarity in physical growth of monozygotic and of dizygotic twins during puberty." Ann. Hum. Biol. 4:417-430. Florencio, C A , 1980. "Household behavior from the nutritionist’s perspective." Ch.2 in Binswanger et al., eds. Rural household studies in Asia. Singapore University Press. ^^Floud, R., A. Gregory and K. Watcher, Height, health and history: nutritional status in the United Kingdom. 1750-1980. Draft, unpublished. Floud, R. and K. Watcher, 1982. "Poverty and physical stature: evidence on the standard of living of London boys, 1770-1870." Social Science History 6. \ / Fogel, R., 1986a. "Nutrition and the decline in mortality since 1700: some preliminary findings." in Engerman, S. and R. Gallman, eds., Lone term factors in American economic growth. Chicago: NBER Studies in Income and Wealth, no. 51. , 1986b. "Physical growth as a measure of the economic well being of populations: the 18th and 19th centuries." in Falkner and Tanner eds. 213 , 1987. "Biomedical approaches to the estimation and interpretation of secular trends in equity, morbidity, mortality and labour productivity in Europe, 1750-1980." Mimeo, Center for Population Economics, University of Chicago. Fogel, R., S.L. Engerman and J. Trussell, 1982. "Exploring the use of data on height: the analysis of long term trends in nutrition, labour, welfare and labour productivity." Social Science History 6. Freund, Rudolf, J., and Ramon C. Littel, 1981. "SAS for linear models: a guide to the ANOVA and GLM procedures." SAS Institute Inc., Cary, NC, USA. v' Ganguli, P., 1979. "Progressive decline in statture in India: a study of sixty population groups." in WA. Stini, ed., Physiological and morphological adaptation and evolution. World Anthropological Series, Monton, The Hague, The Netherlands. Garn, S. and C. Rohman, 1966. "The interaction of nutrition and genetics in the timing of \ growth and development." Pediatric Clinics of North America 13:353-379. j v' Ghosh, Shanti, 1974. "A longitudinal study of length, weight and head circumference from j birth to two years among children of a high socioeconomic urban community in [ Delhi." Indian Pediatrics 11, 395. j , 1986. "A longitudinal study of the outcome of a birth cohort." Project no. 01-658-2, Delhi. Gopalan, C., 1979. "The child in India" 13th Jawaharlal Nehru Memorial Lecture, Jawaharlal Nehru Memorial Fund, New Delhi. j v/- ------- , 1983. "Measurement of undemutrition: biological considerations.” Economic and | Political Weekly 19th April. j I v / --------, 1984."The Tamil Nadu Nutrition Projects.", Bulletin of the Nutrition Foundation of ! India 5. j I I Gopalan, C. and KLVijayraghavan, 1971. Nutrition Atlas of India. NIN, ICMR, Hyderabad. | Gopaldas, Tara and S. Seshadri, eds., 1987. Nutrition monitoring and assessment. Oxford ■ University Press, Bombay, Calcutta and Madras. i Gordon, J.E., I.D. Chitkara and J.B. Wyon, 1963. "Weanling diarrhea." Am. J. Med. Sri. 245:129/345-161/377. I I Government of India, 1983. Census of India. 1981 Paper 2 of 1983. j ✓ Government of Sri Lanka, 1976. Sri Lanka nutrition status survey. Sep. 1975 - March 1976. ] Ministry of Health. ! Government of Tamilnadu, Season and crop reports for various years, Department of I i I I ___________________________________________________________________________________________________________________________________________________________________________ J 214 Statistics, Madras. Greene, L.S. and Johnston, F.E., eds., 1980. Social and biological predictors of nutritional status, physical growth and neurological development. Academic Press, New York. Grewal, Tina, Tara Gopaldas and VJ.Garde, 1973. "Etiology of malnutrition in rural preschool children (Madhya Pradesh)." Journal of Tropical Pediatrics and Environmental Child Health 19(3):265-70. Gronau, R., 1985. "The allocation of goods within the household and estimation of adult equivalence scales: how to separate the men from the boys." Manuscript, Hebrew University. , 1988. "Consumption technology and the intrafamily distribution of resources: adult equivalence scales re-examined." Journal of Political Economy 96.6:1183-1205. Guhan, S., 1981. A primer on poverty: India and Tamilnadu. Madras Institute of Development Studies, Madras. Y’ Gwatkin, D.R., 1979. "Food policy, nutrition planning and survival: the cases of Kerala and Sri Lanka." Food Policy. November. Haaga, J. and J. Mason, 1985. "Intra-family food distribution and nutritional surveillance in developing countries." Cornell Nutritional Surveillance Program, Working Paper # j 44, Ithaca, NY. Habicht, J.E., R. Martorell, C. Yarborough, R. Malina and R. Klien, 1974. "Height and weight standards for preschool children: how relevant are ethnic differences in j growth potential?" Lancet 1:611-614. j I - Hanumantha Rao, D., K. Satyanarayana and G. Gowrinath Sastry, 1976. "Growth patterns * of well-to-do Hyderabad preschool children." Indian J. Med. Res. 64:629. • I ^ Hardy, M.C., 1938. "Frequent illness in childhood, physical growth and final size. American j Journal of Physical Anthropology. 23, 241- 246. j l Harriss, Barbara, 1983. "Implementing food distribution policies: a case study in south India.", Food Policy. 8. , 1985. "Meals and noon meals in Tamil Nadu.” , Mimeo, Nutrition Policy Unit, London School of Hygiene and Tropical Medicine, London. Heller, P.S. and W.D. Drake, 1976. "Malnutrition, child morbidity and the family decision process." Ann Arbor, Michigan: Center for Research on Economic Development, Discussion Paper # 58. 1 1979. "Malnutrition, child morbidity and the family decision process." Journal of j Development Economics 6:203-35 June. i 215 Horowitz, Grace, 1980. Intra-familv distribution of food and other resources. Report to the Nutrition Economics Group, OICD, USDA, Washington DC., July. Horton, Susan, 1986. "Child nutrition and family size in the Phillippines." Journal of Development Economics 23:161-176. > 1988. "Birth order and child nutritional status: evidence from the Phillippines." Economic Development and Cultural Change 36:341-354. ILO, 1976. Employment, growth and basic needs: a one world problem. Geneva. ILO, 1984. World labor report. Geneva. Jallade, J.-P., 1974. Public expenditures on education and income distribution in Columbia. Baltimore: Johns Hopkins University Press, WB Staff Working Paper no. 18. Janes, M.D., 1970. 'The effects of social class on physical growth of Nigerian Yoruba children." Bull. Int. Epidemiol. Assoc. 20:127-136. , 1974. "Physical growth of Nigerian Yoruba children." Trop. Geogr. Med. 26:389-98. Jellife, D.B., 1966. The assessment of nutritional status of the community. Geneva: WHO. Jellife, D.B. and E.F.P. Jellife, 1978. "The volume and composition of human milk in poorly nourished communities: a review." Am. .T . Clin. Nutr. 31:492-515. </ ------- » eds., 1979. Human nutrition: a comprehensive treatise. Plenum Press, New York. Jequier, J.C. et al., 1979. "Exteme und interne Fehlerquelle einer Langschnittuntersuchung" in Die Motorische Entwicklung im Kindes-und Jugendalter. K. Willimczik and M. Grosser eds., pp 151-160, Hoffman, Schomdorf. Johansson, Ryan, S., 1988. T he computer paradigm and the role of cultural information in social systems." in Historical Methods Fall, Vol 21, no. 4. Johansson, Ryan, S., 1990. "Theory, policy, and policy in the political economy of fertility." Session 6N, Sunday 8:45-10:45, Social Science History Association, Minneapolis, October 21. Kakwani, N.C., 1980. Income inequality and poverty. Oxford University Press, New York. , 1986. "Is sex bias significant?" Wider Working Paper 9, Helsinki. Katona-Apte, Judit, 1983. T he significance of intra-household food distribution in food programs." Food and Nutrition Bulletin 5(4):22-29. Kennedy, Eileen, 1983. "Determinants of family and preschooler food consumption." Food 216 and Nutrition Bulletin 5(4):22-29. Y ' Kielman, A A., Uberoi, I.S., Chandra, R.K., and Mehra, V.L., 1976. "The effect of nutritional status on immune capacity and immune responses in pre-school children in a rural community in India.” Bulletin of the WHO, 54, 477 - 483. King, E.M. and L. Lillard, 1983. "A censored discrete choice model of school attainment levels and sibling correlation in schooling." Paper presented at Econometric Society meetings, San Francisco, mimeo. Kumar, S., 1987. "The nutrition situation and its food policy links." in J. Mellor, C. </ Delgado and M. Blackie, eds., Accelerating food production in sub-Saharan Africa. Baltimore, John Hopkins Press. Kynch, J. and A. Sen, 1983. "Indian women: well-being and survival." Cambridge Journal of Economics 7. Laughlin, E.H., and W.G. Mullin, 1955. "Certain aspects of deficiency diseases of the N / tropics and treatment of some related infections." Annals of the New York Academy of Sciences, 1955, 63, 276. Lechtig, A., H. Delgado, R. Martorell, D. Richardson, C. Yarbrough and R.E. Klien, 1978. "Effect of maternal nutrition on infant mortality." in W.H. Mosley ed., Nutrition ^ and human reproduction. New York: Plenum Press. Levinson, FJ., 1974. Morinda: an economic analysis of malnutrition among voung children y in rural India. Cambridge, MA: Cornell/MIT International Nutrition Policy Series. Lindert, P.H., 1978. "Sibling position and achievement." Journal of Human Resources 12:198-219. y Lindgren, G., 1976. "Height, weight and menarche in Swedish urban school children in relation to socioeconomic and regional factors." Ann. Hum. Biol. 3:501-528. Lipton, Michael, 1968. Assessing economic performance. Staples, London. , 1983. Poverty, undernutrition and hunger. WB Staff Working Paper, no. 597, Washington DC. j Lozy, M., et al., 1980. Childhood malnutrition in rural Egypt: results of the Ministry of Health’s ’weighing exercise’. MIT-Cairo University Health Care Delivery Systems Project, Monograph # 4, June. Marchione, TJ., and Prior, F., 1980. "The dynamics of malnutrition in Jamaica.” in Greene, L.S. and Johnston, L.S., eds., Social and biological predictors of nutritional status. Academic Press, New York. Marchione, TJ., 1981. "Child nutrition and dietary diversity within the family: a view from 217 the Caribbean." Food and Nutrition Bulletin Vol. 3, no. 2. /Martorell, Reynaldo, 1980. "Interelationships between diet, infectious diseases, and v nutritional status." in Greene and Johnston eds. pp 81-106. , 1986. "Evaluation of the Tamilnadu Integrated Nutrition Project." Unpublished mimeo. Martorell, R., H. Delgado, V. Valverde and R. Klien, 1981. "Maternal stature, fertility and infant mortality." Human Biology 53:303-312. /Martorell, R. and J.P. Habicht, 1986. "Growth in early chilhood in ^ developing countries." in Falkner, F. and J. Tanner eds., Human growth: a comprehensive treatise, vol. 3, 2nd edition, Plenum Press: New York and London. Martorell, R., J.P. Habicht, C. Yarbrough, R. Malina and R. Klien, 1974. "Height and weight standards for preschool children: how relevant are ethnic differences in growth potential." Lancet l(April):611-15. Martorell, R. and TJ. Ho, 1985. "Nutritional status, morbidity and mortality." Pop. Dev. v ' Rev. 10(suppl. 7):49-68. Martorell, R., J. Leslie, and P.R. Mook, 1984. "Charecteristics and determinants of child nutritional status in Nepal.” Amer. J. Clin. Nutr. 39: 74-86. Martorell, R., F. Mendoza and R. Castillo, 1988. "Poverty and stature in children." in \ f Waterlow, J.C. ed., Linear growth retardation in less developed countries. Nestle Nutrition Workshop Series, vol. 14, Vevey/Raven Press, New York. Martorell, R., and C. Yarbrough, 1983. "The energy cost of diarrheal diseases and other v ' common illnesses in children." in Chen, L.C. and N.S. Scrimshaw, eds., Diarrhea and malnutrition: interactions, mechanisms and interventions. ppl25-141, Plenum Press, New York. Mazumdar, Vina, 1991. "Declining sex ratio." in Frontline. May 11-24, vol. 8, no. 10. Meredith, H.V., and V.B. Knott, 1962. "Illness history and physical growth. HI. Comparative anatomic status and rate of change for school children in different long term health categories." American Journal of Diseases of Children 103:146- 151. Messer, Ellen, 1983. "The household focus in nutritional Anthropology: an overview." Food v"" and Nutrition Bulletin 5(4). Michael, R.T., 1973. "Education in non-market production." Journal of Political Economy pp 306-327. 218 Miller, B,, 1981. The endangered sex. Cornell University Press, Ithaca. Mitra, Ashok, 1985. "The nutrition situation in India." in Biswas, Margaret and Per Pinstrup-Andersen, Nutrition and development. Oxford University Press. National Center for Health Statistics, 1976. Growth Charts. US Department of Health and Human Services, Public Health Service, Health Resources Administration. Norgan, N.G., Ferro-Luzzi, A. and Durnin, J., 1974. "The energy and nutrient intake of 204 New Guinea Adults." Phil. Trans. Rov. Soc.. 268b:309-48. Parizkova, J. and J. Berdychora, 1977. "The impact of ecological factors on somatic and motor development of preschool children." in Eiben, O., ed., Growth and development and physique, pp 235-242, Akademiai Kiado, Budapest. Payne, Philip, 1982. Contributions to UNICEF/FAO/NPU/ICAR Workshop on Nutrition in Agriculture. Pacey, Arnold and Philip Payne, eds., 1985. Agricultural development and nutrition. FAO and UNICEF, Westview Press Inc., Colorado. Payne, Philip, and Peter Cutler, 1984. "Measuring malnutrition: technical problems and ideological perspectives." Economic and Political Weekly. August 25th. . /P e lto , Gretel, 1981. "Methodological issues in nutritional Anthropology." Food and Nutrition Bulletin 3(4):21-28. Pinstrup-Anderson, Per, 1983. "Estimating the nutritional impact of food policies: a note ^ on the analytical approach." Food and Nutrition Bulletin 5(4):16-21. Pitt, Mark M. and Rosenzweig, Mark R., 1984. "Health and nutrient consumption across and within farm households." University of Minnesota, May. Piwoz, Ellen Gail and Fernando E. Viteri, 1984. "Studying health and nutrition behavior by v/r examining household decision-making, intra-household resource distribution and the role of women in these processes." Food and Nutrition Bulletin 7(4). Pollit, E., and Leibel, 1980. "Biological and social correlates of failure to thrive." in Greene, L.S., and Johnston, F.E., eds. Prokopec, M. and V. Lipkova, 1978. "Biological response to social change: acceleration of growth in Czechoslovakia." in Auxologv: human growth in health and disorder. L. Gedda and V. Parisi eds., pp 208-225, Academic Press, London. Puffer, R.R., and C.V. Serrano, 1973. "Patterns of mortality in childhood: report of the inter-American investigation of mortality in childhood." Scientific Publ. no. 262, Pan American Health Organization, Washington DC. 219 Rao, Pralhad, 1980. "Determinants of nutritional status in India." Ch. 6 in Binswanger et \ / al., eds. Rural household studies in Asia. Singapore University Press. Reddy, V., V. Jagadeesan, N. Raghuramulu, C. Bhaskaram, and S.G. Srikantia, 1976. "Functional significance of growth retardation in malnutrition.” Amer. J. Clin. Nutr. 29:3-7. Renson, R., G. Beimen, M. Ostyn, J. Simons, D. Van Gerven and R. Wellens, 1983. "Sociogeographic variation of physical fitness of 12-19 year old Belgian boys." Bull. Soc. Rov. Belee. Anthropol. Prehist. 94:177-192. Reutlinger,S. and M. Selowsky, 1976. Malnutrition and poverty. Baltimore: John Hopkins University Press. Rona, R.J., A.V. Swan and D.G. Altman, 1978. "Social factors and height of primary school children in England and Scotland." J. Epidemiol. Commun. Health 32:147- 154. /Rosenberg, I.H., N.W. Solomon and R.E. Scneider, 1977. "Malabsorption associated with ^ diarrhea and intestinal infections." Amer. J. Clin. Nutr. 30:1248-1253. Rosenzweig, M.R. & T.P.Schultz, 1982a. "Market opportunities, genetic endowments, and the intra-family distribution of resources: child survival in rural India.", American Economic Review. June, vol. 72. , 1982b. "Child mortality and fertility in Colombia: individual and community effects." Health Policy and Education 2:305-348. , 1983. "Estimating a household production function: heterogeneity, demand for health inputs and the effect of birth weight." Journal of Political Economy 91:723- 746. v"-------, 1985. "Health and nutrient consumption across and within farm households." Review of Economics and Statistics 67, May:212-23. Rosenzweig, M.R. and K. Wolpin, 1986. "Evaluating the effects of optimally distributed public programs." American Economic Review 76:470-482. Rowland, M.G.M., RA.E. Barrell and R.G. Whitehead, 1978. "Bacterial contamination in traditional Gambian weaning foods." Lancet 1:136-138. ^Rowland, M.G.M., TJ. Cole and R.G. Whitehead, 1977. "A quantitative study into the role ^ of infection in determining nutritional status in Gambian village children." British Journal of Nutrition 37:441-450. Rowland, M.G.M., 1984. "The diarrhoea-malnutrition complex." Diarrhoea Dialogue no. 6, Aug., p 4; also editorial, p 1. 220 Ryan, J.G., P.D. Bidinger, N. Prahlada Rao and P. Pushpamma, 1984. "The determinants ^ of individual diets and nutritional status in six villages of Southern India." ICRISAT Research Bulletin no. 7, Andhra Pradesh, India. Satyanarayana, K., A. Nadamuni Naidu, M.C. Swaminathan and B.S. Narsinga Rao, 1980. "Effects of nutritional deprivation in early childhood on later growth: a community study without intervention." Am. J. Clin. Nutr. 34:1636-7. Schofield, S., 1974. "Seasonal factors affecting nutrition in different age groups and ^ especially preschool children." Journal of Development Studies vol. 11, no. 1: 22- 40. ^ ------- 1 1975. Village nutrition studies: an annotated bibliography. Institute of Development | Studies, University of Sussex. j ^ ------- ; 1979. Development and the problems of village nutrition. London: Croom Helm. J I Schultz, T J., 1980. "Assessing the causes and consequences of the incidence of diseases i * and death: the interpretation of relations among health status, household j economics and health environment." in Proceedings of the meeting on j socioeconomic determinants and consequences of mortality. Mexico City, 19-25 t June 1979, New York and Geneva: United Nations and WHO. I j , 1984. "Studying the impact of household economic and community variables on | child mortality." in Mosley, H. and L. Chen eds., "Child survival: strategies for j research", Population and Development Review, supplement to vol 10. Schultz, T.P. and K. Wolpin, eds., 1984. Research in Population Economics. Greenwich, ' Conn: JAI Press. ^ Scrimshaw, N.S., 1977. "Effect of infection on nutrient requirements." American Journal of Clinical Nutrition 30:1536-44. Scrimshaw, N.S. and BA. Underwood, 1980. "Timely and appropriate complementary feeding of the breast fed infant: an overview." Food and Nutrition 2:19-22. Seckler, D., 1982. "Small but healthy: a basic hypothesis in the theory, measurement and v policy of malnutrition." in P.V. Sukhatme, ed. Newer concepts in nutrition and their implications for policy. s/ Sen, A.K., 1982. Choice, welfare and measurement. Blackwell, Oxford and MIT Press, Cambridge, MA. , 1984. Resources, values and development. Harvard University Press. , 1985a. Commodities and capabilities. North-Holland Elsevier Science Publishers, B.V. 221 f 1985b. Women, technology and sexual divisions. UNCTAD and INSTRAW. ------; 1988. "Family and food: sex bias in poverty." in T.N. Srinivasan and P. Bardhan eds., Rural poverty in South Asia. Columbia University Press. Sen, A.K. and S.Sengupta, 1983. "Malnutrition of rural children and the sex bias." Economic and Political Weekly Annual no., vol. XVIII nos. 19-21. Senaur, B., M. Garcia and E. Jacinto, 1988. "Determinants of intrahousehold allocation of food in the rural Philippines." American Journal of Agricultural Economics 70.1:170-180. Senaur, Benjamin and Marito Garcia, 1991. "Determinants of the nutrition and health y / status of preschool children: an analysis with longitudinal data." Economic Development and Cuktural Change, vol. 39, no. 2, January. Singh, JJ.B. Wyon and J.E.Gordon, 1962. "Medical care in fatal illness of a rural Punjab population." Indian Journal of Medical Research, November. y Srinivasan, T.N., 1979. "Malnutrition: some measurement and policy issues.” Mimeo, Washington DC: World Bank. y 1 1981. "Malnutrition: some measurement and policy issues." Journal of Development Economics 8.1:3-20. Srinivasan, T.N. & P. Bardhan eds., 1974. Poverty and income distribution in India. Calcutta: Statistical Publishing Society. , eds., 1988. Rural poverty in South Asia.. Columbia University Press. y 1 1977. "Development, poverty and basic human needs: some issues." Food Research Institute Studies, 16. Stini, W A , 1969. "Nutritional stress and growth: sex differences in adaptive response." American Journal of Physical Anthropology 31:417-426. Stini, WA. et al., 1980. "Lean tissue growth and ap'd disease susceptibility in bottle fed versus breast fed infants." in Greene and Johnston eds. Strauss, J., 1990. "Households, commodities and preschool children’s nutrition outcomes: evidence from Cote d’ Ivoire." Economic Development and Cultural Change. 231- 261. Sukhatme, P.V., ed. 1982. Newer concepts in nutrition and their implications for policy. y Maharashtra Association for the Cultivation of Science, Pune 411004. ; 1988. "Energy intake and nutrition: on the autoregulatory homeostatic nature of the energy requirement." in T.N. Srinivasan and P. Bardhan eds. 222 Svedberg, P., 1987. "Undernutrition in sub-Saharan Africa: a critical assessment of the V evidence." World Institute for Development Economics Research, Helsinki, manuscript. , 1988. "Undernutrition in sub-Saharan Africa: is there a sex bias?" Wider Working Paper, Helsinki. Swaminathan, M.C., 1987. "Non-invasive parameters for field monitoring and evaluation: their sensitivity, specificity and cost." in Nutrition monitoring and assessment. Tara Gopaldas and S. Seshadri eds., Oxford University Press, New Delhi. y Tanner, J.M., 1976. "Growth as a monitor of nutritional status." Proc. Nutr. Soc. 35:315- 322. , 1978. Foetus into man: physical growth from conception to maturity. London: Open Books Publishing. , 1981. A history of the study of human growth. Cambridge University Press, Cambridge. , 1982. "The potential of auxological data for measuring economic and social well being." Social Science History 6:571-81. Thoday, J.M., 1965. "Genetidsm and environmentalism." in Meade and Parkes Thomas, D., 1989. "Intra household resource allocation: an inferential approach." Mimeo, Yale University. Thomas, D., J. Strauss and M. Henriques, 1987. "Child survival, nutritional status and y household characteristics: evidence from Brazil." Economic Growth Center discussion paper 542, Yale University. 1 1988. "Child survival, height for age and household charecteristics in Brazil." Revised mimeo, forthcoming, Journal of Development Economics. Trussell, J. and C. Hammerslough, 1983. "A hazards model analysis of the covariates of infant and child mortality in Sri Lanka." Demography 20:1-26. UNICEF, 1981, A sample survey report on the health and nutrition status of children covered bv the mother and child care urogram under the UNICEF assisted food 'S rehabilitation program in West Bengal. (First Assessment, 1979). Calcutta: UNICEF. US Department of Health, Education and Welfare, 1977. NCHS growth curves for v ' children, birth - 18 years. National Center for Health Statistics, Washington DC. S Vaidyanathan, A., 1984. "Food consumption and the size of people: some Indian evidence." 223 Working Paper 186, Center for Development Studies, Trivandrum. Van Praag, 1978. "The perception of welfare inequality." European Economic Review 10. s/ Van Praag, B.M.S. et al., 1982. "A comparison between the food ratio poverty line and the Leyden poverty line." Review of Economics and Statistics 64. Vijayraghavan, K., D. Hanumantha Rao, G.N.V. Brahmam and K.V. Rameshwar Sarma, (undated). "Assessment of nutritional status." Mimeo, National Institute of Nutrition, Field Division, ICMR, Hyderabad. Visaria, P., 1971. The sex ratio of the population of India. Census of India, 1961, Monograph no. 10, New Delhi. Visaria, P and Leela Visaria, 1981. "Population scene after 1981 census." Economic and Political Weekly 17, Spl. No. Waterlow, J., R. Buzina, W. Keller, J. Lane, M. Nichaman and J. Tanner, 1977. The presentation and use of height and weight data for comparing the nutritional V status of groups of children under the age of ten years. Bulletin of the World Health Organization, 55:489-98. Welch, F., 1970. "Education in production." Journal of Political Economy 78.1:35-39. White, Benjamin, N.F., 1980. "Rural household studies in anthropological perspective." in Binswanger et al. eds. Rural household studies in Asia. Singapore University Press. Wolfe, B. and J. Behrman, 1982. "Determinants of child mortality, health and nutrition in a developing country." Journal of Development Economics 11:163-193. , 1987. "Women’s schooling and children’s health: are the effects robust with adult sibling control for the woman’s childhood background?" Journal of Health Economics 6:239-254. World Bank, 1984. World development report. Oxford University Press. Yarbrough, C., J.P. Habicht, R.M. Malina, A. Lechtic and R.E. Klien, 1975. "Length and weight in rural Guatemalan Latino children: birth to seven years of age." American Journal of Anthropology 42:439-447.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

An economic analysis of cooperative group behavior: Theory and evidence

PDF

Irrigation institutions and technical change: A case study from South India

PDF

Life-cycle demographic decisions and the economic well-being of single parents in the United States, 1964-1989

PDF

The influence of industrial pressure groups within the People's Republic of China

PDF

Patronage, dependence and debt in peasant agriculture

PDF

Countertrade: A theoretical and empirical investigation

PDF

The economics of dowry

PDF

Changing role of the household as an income equalizer: An empirical study of Kuznets hypothesis on the case of Taiwan

PDF

Knowledge and use of the contraceptive effect of breastfeeding: Evidence from Malaysia and Guatemala

PDF

The size distriubition of Turkish manufacturing establishments across sectors and over time and its responsiveness to economic policy changes

PDF

Intensive growth vs. extensive growth and their historical implications

PDF

Adaptation, assets, and aspiration. Three essays on the economics of subjective well-being

PDF

Three essays on the determinants of firm structure and contractual choice

PDF

The role of structure, interdependence, and interventionary policy in the diffusion of innovations

PDF

A case study of underutilization of canal irrigation in Bihar, India

PDF

Adaptive economizing in disequilibrium: Essays on economic dynamics

PDF

An econometric evaluation of IMF program effectiveness

PDF

Comparative analysis of the transition to deliberate fertility control and its links to modernization in seven Asian nations

PDF

The political economy of the state and development strategies: A case study of Turkey, 1930-1980

PDF

Examining the portfolio balance theory of foreign exchange rates: A time series analysis

Asset Metadata

Creator Rajivan, Anuradha Khati (author)

Core Title Weight variations among preschoolers: An analysis of evidence from rural Tamilnadu

Degree Doctor of Philosophy

Degree Program Economics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Economics, General,health sciences, human development,health sciences, public health,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Easterlin, Richard A. (committee chair), Heer, David M. (committee member), Kuran, Timur (committee member), Nugent, Jeffrey B. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c20-277023

Unique identifier UC11260194

Identifier DP23382.pdf (filename),usctheses-c20-277023 (legacy record id)

Legacy Identifier DP23382.pdf

Dmrecord 277023

Document Type Dissertation

Rights Rajivan, Anuradha Khati

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA