Cover Image
close this bookOvercoming Child Malnutrition in Developing Countries - Past Achievements and Future Choices. 2020 Vision for Food, Agriculture, and the Environment. Discussion Paper 30 (IFPRI, 2000, 73 p.)
View the document(introduction...)
View the documentForeword
View the documentAcknowledgments
View the document1. Exploring the Causes of Malnutrition
View the document2. Determinants of the Nutritional Status of Children
View the document3. Data and Methods
View the document4. New Evidence from Cross-Country Data, 1970-95
View the document5. How Has Child Malnutrition Been Reduced in the Past?: A Retrospective
View the document6. Projections of Child Malnutrition in the Year 2020
View the document7. Priorities for the Future
View the document8. Conclusions
View the documentAppendix: Cross-Country Studies: Methodological Issues and Past Findings
View the documentReferences
View the documentRecent Food, Agriculture and the Environment Discussion Papers

3. Data and Methods

Explanatory Variables

This study focuses on the underlying and basic determinants of child malnutrition. Explanatory variables representing all three of the underlying determinants described in the conceptual framework - food security, care for mothers and children, and the health environment and services - are considered. In addition, two variables representing the basic economic and political determinants of child malnutrition - national income and democracy - are considered. The choice of variables is guided by the conceptual framework (Figure 1), the experience of past studies, and data availability. A key determinant - poverty - is excluded from the analysis because of insufficient data.

Underlying Determinant Variables

Unfortunately, no cross-national data on food security from nationally representative household survey data are available. However, data do exist for one of its main determinants: national food availability. This variable is used as a proxy, even though it does not account for the important problem of food access, which is also essential for the achievement of food security (Sen 1981; El Obeid et al. 1999).

Similarly, there are no cross-national indicators of maternal and child care that cover the time span of the study. Women’s education and women’s status relative to men’s are chosen as proxies for this factor. The education level of women, who are the main caretakers of children, has several potentially positive effects on the quality of care. More educated women are better able to process information, acquire skills, and model positive caring behaviors than less-educated women. They make better use of health care facilities, interacting more effectively with health care providers and complying with treatment recommendations, and they are more likely to keep their living environment clean. More educated women tend to be more committed to child care, interacting more with the children in their care and stimulating them. Finally, education increases women’s ability to earn income, but this increases the opportunity cost of their time, which tends to mitigate against some important care-giving behaviors, for example, breast feeding (Engle, Menon, and Haddad 1999).

A lower status of women relative to men restricts women’s opportunities and freedoms, providing less interaction with others and less opportunity for independent behavior, which restricts the transmission of new knowledge and damages self-esteem and expression (Engle, Menon, and Haddad 1997). It is a particularly important determinant of two resources for care: mothers’ physical and mental health and their autonomy and control over resources in households. The physical condition of women is closely associated with the quality of caring practices, starting even before a child is born. A woman’s nutritional status during childhood, adolescence, and pregnancy has a strong influence on her child’s birth-weight and subsequent growth (Martorell et al. 1998; Ramakrishnan, Rivera, and Martorell 1998). A woman who is in poor physical and mental health provides lower quality care to her children after they are born, including the quality of breast feeding. In general, when the care a child’s mother receives suffers, the child’s care suffers as well (Ramalingaswami, Jonsson, and Rohde 1996; Engle, Menon, and Haddad 1999). While women are more likely to allocate resources at the margin to the interests of their children than are men, the lower their autonomy and control over resources relative to men’s, the less able they are to do so (Haddad, Hoddinott, and Alderman 1997; Smith and Chavas 1997). In short, low status relative to men restricts women’s capacity to act in their own and their children’s best interests. 3

3 This paper focuses on women’s status relative to men’s, rather than their absolute status, as much work indicates that it is the former that governs both women’s power and control over resources in household decisionmaking as well as the general value placed on women’s well-being at a societal level (Haddad, Hoddinott, and Alderman 1997; Smith 1998; Kishor and Neitzel 1996).

Women’s education and relative status also play a key role in household food security. In many countries women are highly involved in food production and acquisition. The household decisions made in these areas are influenced by women’s knowledge regarding the nutritional benefits of different foods and their ability to direct household resources toward food for home consumption (Quisumbing et al. 1995). Thus the effects on child malnutrition of women’s education and women’s status relative to men’s will partially reflect the influences of these variables on food security as well as on care of mothers and children.

Access to safe water is used as the indicator for health environment and services. Improvements in water quantity and quality have been shown to reduce the incidence of various illnesses, including diarrhea, ascariasis (round-worm), dracunculiasis (guinea worm), schistosomiasis, and trachoma (Hoddinott 1997). This variable was chosen as a proxy for health environment and services because it is the variable for which the most data are available, and because it is highly correlated with other measures of the quality of a country’s health environment and services: countries with high access to safe water are likely to have good health environments and health services overall.

Basic Determinant Variables

To broadly capture the availability of resources in countries, per capita national income is used as a measure, under the hypothesis that income plays a facilitating role in all of the underlying factors considered. A rise in national income may enhance the health environment and services as well as women’s education by increasing government budgets. It may boost national food availability by improving resources available for purchasing food on international markets, and for countries with large agricultural sectors, it reflects the contribution of food production to overall income generated by households. It may improve women’s relative status directly by freeing up resources for augmenting women’s lives as well as men’s. Finally, a plethora of recent studies have shown that the relationship between growth in national income and poverty is negative (see, for example, Ravallion and Chen 1997; Roemer and Gugerty 1997).

The political context within which child malnutrition is determined is accounted for by using democracy as an indicator. As for national income, it is hypothesized that democracy plays a facilitating role in all of the underlying factors considered. The more democratic a government, the larger the percentage of government revenue that may be spent on education, health services, and income redistribution. The more democratic a government is, the more likely it is to respond to the needs of all of its citizens, women’s as well as men’s, indirectly promoting women’s relative status. With respect to food security, the work of Dr and Sen (1994) and others clearly points to the importance of democracy in averting famine. Democratic governments are more likely to honor human rights - including the right to food and nutrition (Haddad and Oshaug 1998) - and to encourage community participation (Isham, Narayan, and Pritchett 1995), both of which may be important means to reducing child malnutrition.

The estimation technique used in this study allows us to consider explicitly only observed variables that change over time. However, one can implicitly control for unobserved time-invariant factors (see Chapter 4) that affect child malnutrition as well. Some important determinants of child malnutrition identified in the last section fall into the latter category, for example, climate and sociocultural environment.

How the Variables Are Measured

The analysis is based on data for 63 developing countries over the period 1970-96 (for a list of the countries included, see Table 2). The dependent variable is the prevalence of children under five who are underweight for their age. The availability of high-quality, nationally representative survey data for underweight children is the limiting factor for inclusion of countries. Data for the explanatory variables are matched for each country by the year in which the underweight data are available. For statistical reasons, only countries for which child malnutrition data are available for at least two points in time are included. The total number of country-year observations is 179. The average number of observations per country is 2.8. The average number of years between observations for a country is 6.9.

All South Asian countries are included in the sample, and more than half of the countries in Sub-Saharan Africa, East Asia, and Latin America and the Caribbean (LAC). The Near East and North Africa (NENA) region, for which only 5 of 14 countries are included, has the poorest coverage. Overall, the sample covers 57 percent of the developing countries and 88 percent of the 1995 population of the developing world. Although the data were not purposefully sampled in a random manner, the authors believe that they are adequately representative of the population of developing countries. 4

4 It is possible that countries with low rates of child malnutrition and high incomes are better able to conduct national surveys of malnutrition. If so, these countries would be overrepresented in the sample. However, it is also possible that national-level malnutrition surveys are carried out in low-income countries with high rates of malnutrition because institutions with external funding sources are more interested in studying them.

The measures employed for the explanatory variables, their definitions, and sample summary statistics are given in Table 3. Here, a brief description is given of each, along with their sources. Fuller descriptions are provided in the companion IFPRI research report (Smith and Haddad 2000).

As indicated earlier, the prevalence of underweight children under age five is used as a measure of child malnutrition. This measure represents a synthesis of height-for-age (long-term growth faltering or stunting) and weight-for-height (acute growth disturbances or wasting) data. 5 The largest share of the data, 75 percent, are from the WHO Global Database on Child Growth and Malnutrition (WHO 1997). These data have been subjected to strict quality control standards for inclusion in the database. The rest of the data are from ACC/SCN (1992, 1996) and World Bank (1997a).

5 While in the past national-level data on stunting and wasting were more rare than data on underweight, they are becoming increasingly available and more widely employed as indicators of child malnutrition (see ACC/SCN 1997 for the first review of trends in stunting, for example). Future cross-country panel data studies of the causes of child malnutrition will be able to use both of the indicators, which are likely to have different determinants (Victora 1992; Frongillo, de Onis, and Hanson 1997).

For per capita national food availability, data for the countries’ daily per capita dietary energy supplies are used. This measure is derived from food balance sheets compiled by the United Nations Food and Agriculture Organization (FAO) from country-level data on the production and trade of food commodities (FAO 1998).

Table 2 - Regional, country, and population coverage of the study

Region

Number of countries

Share of countries covered

Share of population covered a

Number of observations

Country (years in parentheses)


(percent)


South Asia

5

71

98

16

Bangladesh (82, 85, 89, 96), India-rural (77, 91), Nepal-rural (75, 95), Pakistan (77, 85, 90, 95), Sri Lanka (77, 80, 87, 93).

Sub-Saharan Africa

26

58

83

65

Benin (87, 96); Burkina Faso (87, 92); Cameroon (77, 91); Comoros (91, 95); Congo, Republic of (77, 87); Congo, Democratic Republic of (75, 86, 89, 94); Cd’Ivoire (86, 94); Ethiopia-rural (83, 92); Ghana (87, 93); Guinea (80, 95); Kenya-rural (82, 87); Lesotho (76, 81, 94); Madagascar (83, 92, 95); Malawi (81, 92, 95); Mauritania (81, 87, 90); Mauritius (85, 95); Niger (85, 92); Nigeria (90, 93); Rwanda (76, 92); Senegal (86, 92); Sierra Leone (74, 77, 90); Tanzania (87, 91, 96); Togo (76, 88); Uganda (77, 88, 95); Zambia (72, 85, 88, 92, 96); Zimbabwe (84, 88, 94).

East Asia

8

57

94

26

China (87, 92, 95), Indonesia (78, 87, 95), Laos (84, 94), Malaysia (83, 86, 90, 95), Myanmar (80, 83, 90, 95), Philippines (73, 82, 87, 93), Thailand (82, 87, 90), Viet Nam (80, 87, 94).

Near East and North Africa

5

31

37

14

Algeria (87, 92, 95), Egypt (78, 88, 92, 95), Jordan (75, 90), Morocco (87, 92), Tunisia (74, 88, 94).

Latin America and the Caribbean

19

68

85

58

Bolivia (81, 89, 93), Brazil (75, 89, 96), Chile (78, 82, 86, 95), Columbia (77, 86, 89, 95), Costa Rica (78, 82, 89, 94), Dominican Republic (86, 91), El Salvador (88, 93), Guatemala (77, 80, 87, 95), Guyana (71, 81, 93), Haiti (78, 90, 94), Honduras (82, 87, 93), Jamaica (78, 85, 89, 93), Mexico-rural (74, 79, 89), Nicaragua (80, 93), Panama (80, 92), Peru (75, 84, 91, 96), Trinidad and Tobago (76, 87), Uruguay (87, 92), Venezuela (81. 87, 90, 94).

Developing countries

63

57 percent of the developing countries

88 percent of the developing-country population

179


Source: Population data. United Nations 1998; regional grouping of developing countries, Smith and Haddad 2000.

a These percentages are calculated from countries’ 1995 populations.

Table 3 - Variable definitions and sample summary statistics


Definition

Mean

Standard deviation

Minimum

Maximum

Prevalence of child malnutrition

Percent of children under five less than -2 standard deviations below the median weight-for-age Z-score of the NCHS/WHO international reference

24.6

15

0.9
(Chile 1995)

71.3
(India 1977)

Access to safe water

Percent of population with access to safe water (percent)

56.2

23.7

6
(Ethiopia 1983)

100
(Mauritius 1985)

Female secondary school enrollment

Gross female secondary school enrollment rate (percent)

33.8

22.5

2.5
(Uganda 1977)

88
(Uruguay 1992)

Female-to-male life expectancy ratio

Ratio of female life expectancy at birth to male life expectancy at birth

1.062

0.03

0.97
(Nepal 1975)

1.15
(El Salvador 1988)

Per capita dietary energy supply

Daily per-capita dietary energy supply (kilocalories)

2,360

331

1,592
(Ethiopia 1992)

3,284
(Egypt 1995)

Per capita GDP

Per capita gross domestic product (in purchasing power parity-adjusted 1987 U.S. dollars)

2,306

1,779

306
(Ethiopia 1992)

8,612
(Chile 1995)

Democracy

Combined index of political rights and civil liberties (measured on a scale of 1 to 7 points, 1 = least democratic)

3.5

1.7

1 a

7
(Costa Rica 1978,1982,1989)

a The countries for which the democracy index number is 1 are: Algeria (1 995), Benin (1987), China (1992, 1995), Ethiopia (1983), Guinea (1980), Haiti (1994), Laos (1994), Mauritania (1995), Myanmar (1990, 1995), Uganda (1977), Viet Nam (1980, 1987, and 1994), and Zaire (1986).

For women’s education, female gross secondary school enrollment rates are used as a measure. The data are from the United Nations Educational, Scientific, and Cultural Organization’s UNESCOSTAT database (UNESCO 1998).

The measure used for women’s status relative to men’s is the ratio of female life expectancy at birth to male life expectancy at birth. 6 Life expectancy at birth is defined as the number of years a newborn infant would live if prevailing patterns of mortality at the time of his or her birth were to stay the same throughout his or her life. This measure was chosen because discrimination against females at all stages of life and inequity in investments in women relative to men are reflected in differences in life expectancy ratios. The source for life expectancy data is World Development Indicators (World Bank 1998).

6 There is no agreed-upon measure of “women’s status”. Most measures available in the literature are multiple-indicator indexes (UNDP 1997; Kishor and Neitzel 1996; Mohiuddin 1996; Ahooja-Patel 1993). The extension of human life is associated with an enhanced quality of life. Inequalities in life expectancy or mortality favoring males (beyond international biological norms) reflect discrimination against females and entrenched, long-term gender inequality (Sen 1998; Mohiuddin 1996). The gender life expectancy ratio was chosen as a measure of women’s relative status for two main reasons: (1) it is a good indicator of the cumulative investments in females relative to males throughout the human life cycle; (2) data are readily available for the countries and years included. The ratio of female-to-male infant mortality would be an even better index, not being influenced by public health risks such as cigarette smoking in adults, but such data were not available for sufficient countries and years.

To measure access to safe water, the percentage of a country’s populations with access to safe water is used, defined as the population share with reasonable access to an adequate amount of water that is either treated surface water or water that is untreated but uncontaminated (such as water from springs, sanitary wells, and protected boreholes) (World Bank 1997b). The data are from various years of UNICEF’s State of the World’s Children and reports from the World Health Organization (WHO 1996).

For per capita national income, real per capita gross domestic product (GDP), expressed in purchasing power parity-comparable 1987 U.S. dollars, is used. The data are from the World Bank’s World Development Indicators (World Bank 1998). 7

7 These data are only reported for 1980 to the present. To arrive at comparable purchasing power parity (PPP) GDP per capita figures for the 1970s data points, it was necessary to impute growth rates from the data series on GDP in constant local currency units and apply them to countries’ 1987 PPP GDPs.

For degree of democracy, an average of two seven-point country-level indexes from Freedom House (1997) are used. One is of political rights and one of civil liberties, 8 and each is given an equal weight. The combined index ranges from 1 to 7, with “1” corresponding to least democratic and “7” to most democratic.

8 Political rights enable people to participate freely in the political process, including choosing their leader freely from among competing groups and individuals. Civil liberties give people the freedom to act outside of the control of their government, including developing their own views, institutions, and personal autonomy (Ryan 1995).

Estimation Methods

The estimations of the effects on child malnutrition of the hypothesized determinants are based on multiple linear regression techniques. 9 Specifically, a panel-data econometric technique, country fixed-effects estimation, is used. The technique controls for country-specific factors that vary little over time - factors such as climate, characteristics of countries’ physical environments (soil type and topography, for example), and deeply embedded cultural and social mores. From a practical standpoint, it is carried out by including in the estimating equations a dummy variable for each of the 63 countries in the sample.

9 See Smith and Haddad (2000) for a detailed description of the methods employed.

In conformity with the conceptual framework, two child malnutrition regression models are separately estimated, an underlying determinants model and a basic determinants model. The underlying determinants model explores the effects of national food availability, women’s education and status, and access to safe water on the prevalence of malnutrition. The basic determinants model explores the effects of national income and democracy. As discussed in the appendix on methodological issues and past findings, when determinants lying at different levels of causality are included in the same regression equation, biased estimation results. Therefore, the models are kept separate to avoid this problem. It is then possible to explore the ways in which the basic determinants work through the underlying determinants to affect child malnutrition.

A number of tests have been performed to gauge the accuracy of the estimates and whether they differ across the developing-country regions (for details see Smith and Haddad 2000). One test determines whether any important variables have been left out of the analysis. Another tests for potential endogeneity of the explanatory variables, using an instrumental variables technique. Finally, a test for parameter stability determines whether there are significant differences across the developing regions. 10

10 The test for omitted variables bias is the Ramsey Regression Specification Error Test (RESET). A Hausman-Wu test is used to test for endogeneity of the explanatory variables. To identify appropriate instruments for undertaking this test, candidate instrument sets are subjected to relevance and overidentification tests. The parameter stability tests are Chow F-tests (see Smith and Haddad 2000).