Chapter 6. How Has Child Malnutrition Been Reduced in the Past? A Retrospective

The approximate reductions in the prevalence of child malnutrition that took place between 1970 and 1995 for the developing countries as a group and for individual regions are summarized in Table 1. How were these reductions brought about? In this chapter, based on the regression analysis of the last, the contributions of improvements in health environments, women's education, women's status relative to men's, national food availability (underlying-determinant variables), national incomes, and democracy (basic-determinant variables) are explored.

To answer the question, the FE parameter estimates of Table 7 (columns 3 and 4) are used to formulate a predicting equation for the change in child malnutrition prevalence over 1970-95. Using the underlying-determinants model as an example, the base predicting equation (from equation 9) is

(19)

where DES1, DES2, and DES3 are the segments of the DES linear spline and is the FE term. This latter term is wiped out using a first-difference transformation as follows:

(20)

Each variable's absolute and percentage contributions to the total change are then calculated.³⁹ The absolute and percentage contributions, respectively, of SAFEW for example, are

(21)

³⁹ This procedure is similar to that undertaken in the estimation of "population attributable risk" (common in epidemiology) in which information on the risk of contracting a disease (such as lung cancer) from exposure to risk factors (such as smoking) is combined with information on the prevalence of the risk factor to determine the number of cases of the disease that are associated with the risk factor (Kahn and Sempos 1989).

For information on how much each variable has actually changed over the period for the sample, the data set is expanded to include all available data for the variables at six points in time: 1970, 1975, 1980, 1985, 1990, and 1995. Variable means for each of these years as well as their total change over the study period are given in Table 15. Appendix Tables 25-29 contain the information by region. The estimated contribution of the proxy measure for each variable is employed to estimate its contribution. For example, the contribution of increases in per capita DES is used to estimate the contribution of national food availability.

The upper panel of Table 16 reports the results for the underlying-determinant variables over 1970-95. The total predicted contribution for the sample is a reduction of 15.9 percentage points (row 4). The number is quite close to the 15.5 percentage-point reduction estimated for all developing countries in Table 1.⁴⁰

⁴⁰ Two sources of differences in the "actual" and predicted numbers are (1) any differences between the study sample and the developing countries as a whole in terms of the etiology of child nutrition; and (2) a restricted data set, in terms of years covered, which is employed in the regression equations due to the need to match data on the explanatory variables with the particular years for which underweight data are available in the former. That the two numbers are close indicates that the sample is representative of the developing countries as a whole. It also gives supporting evidence that the model provides a good predicting equation for underweight in the historical period covered. The predicted child malnutrition values for individual regions can also be compared with the numbers reported in Table 1. They are quite close for Sub-Saharan Africa, East Asia, and Latin America and the Caribbean. However, the predicted value is 28 percent higher than the actual for South Asia; it is 100 percent lower than the actual for the Near East and North Africa region.

Figure 12 summarizes the estimated percent contribution of each underlying determinant variable to the reduction in the prevalence of child malnutrition. The change in each over the period on an equivalent (100 percent) scale is also presented to aid in interpretation (Figure 13). Improvements in women's education have contributed by far the most, 43 percent. This contribution is the combined result of both the strong effect of the determinant and a fairly large increase in it over the period, as shown in Figure 13. The contribution of improvements in health environments has also been substantial: almost 20 percent. Improvements in food availability have contributed around 25 percent of the reduction in child malnutrition, not only because the effect of this variable is strong but also because increases have been substantial, rising from 2, 092 kilocalories per capita in 1970 to 2, 559 in 1995. The smallest contribution (12 percent) came from improvements in women's status as gauged by the female-to-male life expectancy figures. While this factor has a strong impact on child malnutrition, it has improved very little compared to the other variables (Figure 13). Together, the contributions of women's education and relative status have contributed to more than half of the 1970-95 reduction in child malnutrition in the developing countries. It can thus be surmised that much of the reduction has come from improvements in maternal and child care, confirming its critical role in the etiology of child nutrition. Some of the effect may also be working through the pathway of improved household food security.

Table 15 - Underlying-determinant and basic-determinant variable means, 1970-95

Variable			1970 (1)	1975 (2)	1980 (3)	1985 (4)	1990 (5)	1995 (6)	Absolute change, 1970-95 (7)	Average annual change, 1985-95 (8)
Underlying-determinant variables
	Access to safe water (percent)		30.2	45.4	52.4	60.7	69.9	70.3	40.1	0.96
	Female secondary school enrollment (percent)		15.6	25.4	28.4	30.6	36.4	46.6	31.0	1.6
	Female-to-male life expectancy ratio		1.022	1.026	1.033	1.040	1.045	1.048	0.024	0.0008
		Female life expectancy (years)	55.2	58.5	60.4	63.5	65.3	66.0	10.79	0.25
		Male life expectancy (years)	54.0	56.9	58.5	60.8	62.4	63.0	9.04	0.22
	Per capita dietary energy supply (kilocalories)		2,092	2,089	2,226	2,380	2,472	2,559	467	17.9
Basic-determinant variables
Per capita GDP (US$ PPP)			1,011	1,163	1,361	1,378	1,673	2,121	1,111	44.8
Democracy (1 = least democratic)			2.85	2.99	3.75	3.31	3.24	2.71	-0.14	-0.06

Notes: These data are population-weighted. They are estimated using data for the countries in the data set only (Comoros was dropped from the sample due to the absence of population data). In some cases where data were not available for a sample country for a particular year, extrapolations were undertaken.

Table 16 - Estimated contributions of underlying- and basic-determinant variables to changes in the prevalence of child malnutrition, by region, 1970-95

Variable		All regions (1)	South Asia (2)	Sub-Saharan Africa (3)	East Asia (4)	Near East and North Africa (5)	Latin America and the Caribbean (6)
		(percentage-point change in underweight rate)
Underlying-determinant variables
	Health environment	-3.06	-4.56	-2.07	-2.74	-0.45	-1.80
	Women's education	-6.82	-4.61	-3.39	-9.27	-9.64	-6.98
	Women's status relative to men's	-1.84	-3.85	+ 1.27	-1.36	+ 0.28	-1.65
	National food availabilitya	-4.14	-3.44	-0.048	-6.11	-2.34	-0.77
	Total percentage-point change	-15.9	-16.5	-4.2	-19.5	-12.4	-11.2
Basic-determinant variablesb
	Per capita national incomea	-7.39	...	...	...	...	...
	Democracy	+ 0.18	...	...	...	...	...
	Total percentage-point change	-7.2	...	...	...	...	...

Notes: The estimates in this table are obtained by multiplying the coefficients on the proxy variables for each determinant (see Table 7, columns [3] and [4]) by the change in the proxy from 1970-95. The changes are obtained from Appendix Tables 25-29.

^a These estimates take into account the changing coefficient on the proxy variable (DES and GDP) as its level changes.

^b As discussed in Chapter 5, the regression coefficients of the basic-determinants model cannot be applied to the regions separately due to fundamental structural differences across the regions. Thus, contributions for the basic-determinant variables are not broken down by region.

Figure 12 - Estimated contributions of underlying-determinant variables to reductions in developing-country child malnutrition, 1970-95

Source: IFPRI Cross-Country Child Malnutrition Determinants Data Set, 1997/98.

Figure 14 traces out the contributions of the underlying-determinant variables to changes in child malnutrition prevalences for five-year intervals during the study period, starting with 1970-75 and ending with 1990-95. The change in the prevalence of malnutrition over the intervals, in percentage points, is marked on the vertical axis. The bar falls below zero when there has been an increase in a variable, which is associated with a reduction in child malnutrition. A bar value above zero indicates that there has been an increase in child malnutrition due to a reduction in the level of a variable.

There are several points to note from the figure.

· There has been a fairly steady decline in the prevalence of developing-country child malnutrition of about 3.2 percentage points every five years since the early 1970s. The largest reduction, 4 percentage points, came in the 1975-80 period. Since then, the reductions have been smaller.

· The contribution of improvements in health environments has declined over the 25-year period; in the early 1990s it made very little contribution.

· Women's education made its greatest contribution in the early 1970s and early 1990s. Its contribution dropped dramatically between 1970 and 1980; since then it has gradually increased. It contributed 84 percent of the total 2.7 percentage-point reduction in the underweight prevalence in the early 1990s.

· Corresponding to the world food crisis of the 1970s, food availability declined during 1970-75, leading to a slight increase in child malnutrition. As the Green Revolution picked up, the developing countries saw substantial increases in their food supplies. The contribution of food availability to declines in malnutrition were steady and substantial in the late 1970s and early 1980s. Despite continued increases in food availability in the late 1980s and early 1990s (see Table 15), their contribution to reductions in child malnutrition has leveled off due to a decline in strength of their impact as they have increased.

· As improvements in women's status have fluctuated over the 25-year period, so too has their contribution to malnutrition reductions. The greatest contribution was made in the early 1980s; since the late 1980s it has declined considerably.

Figure 13 - Increases in underlying-determinant explanatory variables, 1970-95, on an equivalent scale of 100 percent

Source: IFPRI Cross-Country Child Malnutrition Determinants Data Set, 1997/98.

Notes: This figure shows the increase in each underlying-determinant variable over 1970-95 on an equivalent scale that allows comparison across variables even though they are measured in different units. Each variable is transformed by equating the minimum of its developing-country range to 0, the maximum to 100, and the in-between numbers to their relative positions on this scale. The variable ranges are given in Table 11, column (3). The shaded area in the bar for each variable starts with its 1970 transformed value and ends with its 1995 transformed value.

Figure 14 - Changes in child malnutrition in developing countries: Estimated contributions of underlying-determinant variables, five-year periods, 1970-95

Source: IFPRI Cross-Country Child Malnutrition Determinants Data Set, 1997/98.

The bottom panel of Table 16 gives the contribution of the basic-determinant variables. Democracy has actually deteriorated slightly. Despite its potential positive contribution, this declining trend is associated with a slight increase in child malnutrition. Improvements in per capita national income, however, have been quite large. For the full sample of countries, per capita GDP rose from $1,011 in 1970 to $2,121 in 1995, more than doubling. This large increase, in combination with the strong influence of the variable, has facilitated an estimated 7.4 percentage-point reduction in child malnutrition. The influence of national incomes in reducing malnutrition throughout the developing world over the 25-year period since 1970 has thus been quite strong. Figure 15 breaks the two variables' contributions down into five-year periods. While democracy made positive contributions in the 1970s, since then its decline has put a drag on child nutrition improvements. Aside from the early 1980s, the contribution of national income has been steadily increasing since the 1970s. Its greatest contribution came most recently: in the 1990-95 period alone it contributed to a 3 percentage-point decline in the prevalence of malnutrition.

Figure 15 - Changes in child malnutrition in developing countries: Estimated contributions of basic-determinant variables, five-year periods, 1970-95

Source: IFPRI Cross-Country Child Malnutrition Determinants Data Set, 1997/98.

Since there are fundamental differences across the regions in the strength of impact of the basic-determinant variables, precise region-specific estimates of their contributions cannot be undertaken. It can be inferred, however, that the contribution of national income has been positive for all regions except Sub-Saharan Africa (where it declined). The contribution of democracy has been positive for Sub-Saharan Africa, NENA, and LAC, but negative in South Asia and East Asia.

The regions' experiences have also differed regarding the relative contributions of the underlying-determinant variables. The section on regional differences in Chapter 5 demonstrated that there are no significant differences in the functional relationships between the underlying-determinant variables and child malnutrition. Thus their contributions can be quantified using the full-sample coefficient estimates of Table 7, column (3).

The overall reduction in the prevalence of child malnutrition in South Asia for the 25-year period is estimated to be 16.5 percentage points. As illustrated in Figure 16, the greatest contributions to this reduction have come from increased education of women and improvements in health environments, at about 28 percent each. Improvements in women's relative status have accounted for about 25 percent of the reduction and improvements in food availability about 20 percent. Figure 16 shows that the factors' relative contributions have fluctuated substantially over the study period. In the early 1970s, reductions in child malnutrition from improvements in health environments, women's education, and women's relative status were completely undermined by reductions in food availability. As a result, no progress was made. By the late 1970s, food availability began to improve and it contributed substantially to reductions in child malnutrition in the 1980s and early 1990s. The 1980s saw a precipitous decline in the prevalence of malnutrition in the region (over 11 percentage points) due to improvements in all of the factors. However, in the 1990-95 period, the pace of improvement has been severely curtailed by slower growth in health environment improvements and food availability.⁴¹ While the relative status of women in South Asia continues to be the lowest of all developing-country regions (the 1995 female-to-male life expectancy ratio was 1.02), small improvements have made a steady contribution in the past 25 years.

⁴¹ The declining contribution of food availability is partially due to declining impact as food supplies increased.

Figure 16 - Changes in child malnutrition in South Asia: Estimated contributions of underlying-determinant variables, five-year periods, 1970-95

Source: IFPRI Cross-Country Child Malnutrition Determinants Data Set, 1997/98.

The total reduction in Sub-Saharan Africa's child malnutrition rate over the study period is estimated to be only 4.2 percentage points. This overall net reduction masks the negative effect that deteriorations in women's relative status have posed over the study period. Figure 17 gives the percent contributions of the remaining factors to reductions in the region's prevalence of child malnutrition. The largest share of the reduction was brought about by increases in women's education, followed by improvements in health environments. A very small contribution was made by improvements in food availability. Figure 17 gives a fuller picture of the role the underlying-determinant variables have played. Improvements in health environments have made their greatest contribution since 1985. Increased education of women made strong contributions in all periods except for the late 1980s, when it actually declined. Women's relative status has continually declined in the region since the 1970s, most precipitously after 1985. Its contribution has thus been to worsen child malnutrition in the region throughout the study period. Changes in food availability have played a large role overall. However, the role was not always positive. Substantial improvements in the late 1980s and early 1990s were outweighed by deteriorations, for the most part, during the 1970-85 period.

Figure 17 - Changes in child malnutrition in Sub-Saharan Africa: Estimated contributions of underlying-determinant variables, five-year periods, 1970-95

Source: IFPRI Cross-Country Child Malnutrition Determinants Data Set, 1997/98.

The prevalence of child malnutrition has declined the most in East Asia, about 20 percentage points. The greatest contribution to this decline came from increases in women's education, followed by improvements in food availability and health environments (Figure 18). The early 1970s witnessed a very large reduction in child malnutrition of more than 6 percentage points (Figure 18), most of which was due to increases in women's education. Progress since this period has not been as great, but it has continued steadily. The contributions of improvements in the health environment and women's relative status declined over the period and were minimal by the 1990s. Improvements in food availability have taken place at a relatively fast pace in East Asia, rising from 1, 998 kilocalories per capita in 1970 to 2, 720 kilocalories in 1995. Overall they have contributed to a reduction in the prevalence of child malnutrition of about 6 percentage points.

Box 1 Why Has Child Malnutrition Been Rising in Sub-Saharan Africa? Sub-Saharan Africa is the only region in which the prevalence of child malnutrition has been increasing. From 1985 to 1995, it increased from 29.9 percent to 31.1 percent (see the table below). Of the four underlying-determinant variables, only one - women's relative status as proxied by the female-to-male life expectancy ratio - was moving in the wrong direction during the period. Two others, national food availability and women's education - both of which remain at extremely low levels - were almost stagnant. In addition, national income for the region declined significantly: per capita GDP fell by $52. The decline in this important basic determinant of child malnutrition is responsible for slow progress in all of the underlying-determinant factors and a slight increase in poverty. Therefore, it seems likely that deterioration in women's relative status and per capita national income, along with stagnation in women's education and food availability, at least partially explains the deterioration in child malnutrition in the region. Other factors may be deterioration in the capacity and outreach of government services under the impact of debt and structural adjustment; the rising incidence of HIV/AIDS (Ramalingaswami, Jonsson, and Rohde 1996); and conflict (Messer, Cohen, and D'Costa 1998). The decline in the ratio of women's life expectancy to men's is puzzling. It may be because women in Sub-Saharan Africa are more vulnerable to HIV/AIDS than men are, which itself reflects women's lower status (Brown 1996; Howson et al. 1996). Trends in the determinants of child malnutrition in Sub-Saharan Africa, 1985-95
	1985	1995	Percentage change, 1985-95
Child malnutrition (percent underweight)	29.9	31.1	4.0
Access to safe water (percent)	33.5	48.8	45.7
Female secondary school enrollment (percent)	16.4	19.0	15.8
Female-to-male life expectancy ratio	1.066	1.054	-1.1
Per capita dietary energy supply (kilocalories)	2.035	2,136	5.0
Per capita national income (PPP US$)	830	778	-6.3
Democracy	2.01	2.44	21.4
Poverty (percent)a	38.5	39.1	1.6
Sources: Tables I and 26. Poverty data are from Ravallion and Chen 1996, Table 5. Notes: With the exception of the poverty rates, these data are population-weighted means over all countries in the data set in each region. The poverty measure employs an international poverty line of $1 per person per day at 1985 power parity. a Poverty figures are for 1983 and 1993.

Figure 18 - Changes in child malnutrition in East Asia: Estimated contributions of underlying-determinant variables, five-year periods, 1970-95

Source: IFPRI Cross-Country Child Malnutrition Determinants Data Set, 1997/98.

Almost all of the reduction in child malnutrition in the Near East and North Africa region has come from increases in women's education (Figure 19).⁴² The state of health environments has fluctuated widely - sometimes improving, sometimes deteriorating - the net result being a very small contribution of 3.6 percent. Women's relative status deteriorated in most periods, muting child nutrition improvements. Improvements in food availability have contributed 19 percent to the reduction in child malnutrition. Although food availability has improved in all periods, its marginal impact has declined. By 1995, per capita DES had reached 3, 172 kilocalories, a point past which food availability has no or minimal impact.

⁴² The results for this region should be treated with caution given that the majority of countries in the region are not represented in the sample and that the model overpredicts the total reduction in child malnutrition for it substantially (see footnote 40).

The Latin America and Caribbean region experienced an estimated 11 percentage-point reduction in child malnutrition during the period, most of which took place during 1970-80. Since then, reductions in child malnutrition have continued at a much slower pace (Figure 20). As for the other regions, the greatest contribution (62 percent) came from expansions in female education (Figure 20). The contribution of improvements in health environments has steadily declined. Strong improvements in women's relative status in the 1970s were followed by small improvements in the 1980s. By the early 1990s there was a slight decline, muting the overall reduction in malnutrition of the period. Food availabilities improved in the 1970s, but declined in the early 1980s. Their contribution has been minimal since the early 1980s.

Figure 19 - Changes in child malnutrition in Near East and North Africa: Estimated contributions of underlying-determinant variables, five-year periods, 1970-95

Figure 20 - Changes in child malnutrition in Latin America and the Caribbean: Estimated contributions of underlying-determinant variables, five-year periods, 1970-95

Source: IFPRI Cross-Country Child Malnutrition Determinants Data Set, 1997/98.