Objective. To analyze the effects of socioeconomic, regional, and ethnic conditions on chronic malnutrition in four Andean countries of South America: Bolivia, Colombia, Ecuador, and Peru.
|Key words||Equity, malnutrition, nutritional status, poverty, socioeconomic factors.|
This article analyzes empirical evidence from recent household surveys on chronic malnutrition and its relation to social, regional, and ethnic disparities in four Andean countries of South America: Bolivia, Colombia, Ecuador, and Peru.
There are two immediate causes of chronic child malnutrition (extremely low height-for-age): insufficient access to nutrients and high disease exposure. However, socioeconomic, regional, and ethnic factors frequently play a significant role in nutritional outcomes.
While malnutrition has generally been declining in recent decades, it is still a severe problem, affecting at least 27% of children younger than 5 years of age in the developing world and 21% of the children under 5 in the four Andean countries of our study (1-4). Poverty and social inequality are pervasive in the Andean nations, and in recent years the prospects for social development there have been affected by economic crisis, social conflict, and political instability (5, 6).
Child malnutrition has generally declined in Latin American countries over the last several decades as socioeconomic conditions have improved and fertility levels have declined. Nevertheless, such progress has been slow, poverty and social inequality are still widespread, and malnutrition prevalence rates remain high in many countries, including several of the Andean countries in our study.
Although the association between poverty and malnutrition is generally well-known, more empirical and analytical work is still needed on the relationships between social inequality and malnutrition within Latin American countries. In fact, concerns about health and equity are becoming more prominent, and specific policies to promote nutrition among the poor have been called for.
INFORMATION SOURCES AND STUDY METHODOLOGY
Our study is based on household surveys that were done in the mid- and late-1990s and that included anthropometric measures on children younger than 5 years. Demographic and Health Surveys (DHS) were the main sources of data for Colombia (a survey done in 1995), Peru (1996), and Bolivia (1997), with a Living Standard Measurement Survey (LSMS) providing the information for Ecuador (1998). All those surveys had national coverage, representative sample sizes, and detailed questionnaires on socioeconomic conditions, access to health services, and maternal and health outcomes (7-10).
Conventionally, children's anthropometric measures are transformed to a normalized scale of z scores of height-for-age, weight-for-age, and height-for-weight. A child is affected by chronic malnutrition (stunting) when his or her z score of height-for-age is lower than -2, or two standard deviations, below the international normalized median. Using the same criterion, global malnutrition is defined by low weight-for-age scores, and acute malnutrition corresponds to low height-for-weight.
Both global malnutrition and acute malnutrition may reflect occasional weight loss as a result of recent disease episodes. Stunting, however, is more representative of insufficient growth due to persistent dietary deficiencies and/or susceptibility to illness. This paper takes chronic malnutrition and its associated height-for-age z score as indicators of nutritional status.
DHS surveys have common questionnaires, allowing the development of multivariate indices for different countries. For our study of the four Andean nations we established an integrated socioeconomic status (SES) index that included three dimensions: education, housing, and employment.
Each dimension, in turn, derives from a particular set of indicators, as shown in Table 1. We applied categorical principal components analysis to develop multivariate indices of education, housing, and employment from the indicators, and then we estimated a global SES index from those three dimensional indices. The procedure permits the integration of both numerical and categorical variables, and it is a generalized extension of the classical method of principal components, which was originally restricted to just numerical variables.
Most of the questions in DHS surveys and other household surveys, as well as an important part of the social indicators (such as the household source of drinking water, or the occupational group), are categorical variables, either nominal or ordinal. Numerical indicators alone, such as years of schooling or the time spent to get drinking water, are usually insufficient for summarizing the questionnaire's most relevant topics.
As a statistical procedure, the categorical principal components analysis (CATPCA) simultaneously provides optimal quantification of categorical variables and reduces the dimensionality of the data, by summarizing in a reduced number of factors most of the information provided by the original indicators (11-14). CATPCA handles nominal, ordinal, and numeric indicators. An index (based on the first principal component) may be interpreted as the linear combination of constituent indicators, which captures the maximum possible amount of information provided by the indicators. The index optimizes the explained proportion of the total original indicator variance. While this paper emphasizes our findings, a detailed explanation of the methodology applied in this research is also available.3
In the particular case of Ecuador, where the LSMS is the only household survey with child anthropometric measures from the 1990s, no multivariate indices were estimated. For comparative analysis, aggregate per capita consumption is used as a proxy for the SES index. (Aggregate per capita consumption is the total household consumption divided by the number of household members. A regional price index was used to compensate for local differences in the cost of living.) Aggregate consumption is regarded as a more reliable and stable indicator of SES than household income, which has short-term fluctuations and is frequently underreported.
Bolivia, Ecuador, and Peru have significant indigenous populations, with their own cultures and languages. Although the DHS and LSMS surveys do not include a direct question concerning race or ethnicity, they each have a question about the language spoken at home. If the answer included an indigenous language, the household was identified as indigenous. Although this procedure may underestimate the size of indigenous groups, it is useful for identifying ethnic differences.
We prepared descriptive data on stunting prevalence by place of residence (large cities vs. small cities, etc.), geographic region (highland region vs. other areas of a country), ethnicity, and SES. We also calculated an estimate of the socioeconomic gradient in stunting as well as concentration indices (concentration indices are widely used indicators of social inequality in health and are based on a generalization of the Gini coefficient) (1). In addition, we conducted a specific regional analysis for the three Andean countries with severe geographical disparities, where we prepared separate regressions for the highland area of each country and for the other regions in the same country. Both linear regression models and smoothed regression curves were estimated in each case.
The regional, ethnic, and socioeconomic distribution of stunting in the four Andean countries
The Andean countries have persistently suffered from pronounced social, regional, and ethnic inequality as well as widespread poverty, with economic and social indicators falling below the averages for Latin America as a whole. Further, in the Andean countries over the last two decades, the economic achievements from structural adjustment programs and export promotion efforts have been weak, social conflict has been growing, and various forms of political violence have emerged.
The Andean countries, however, are internally diverse and heterogeneous. In broad terms, Colombia has historically achieved significant and more stable economic growth rates, coupled with efficient industrialization and diversification. On the other hand, Bolivia, Ecuador, and Peru remain less diversified economically, with lower per capita incomes. Those three countries also have deeper ethnic and regional disparities, which have shaped their societies from the colonial period. Table 2 presents basic economic and social indicators for the four Andean countries of this study.
Table 3 presents stunting prevalence for each of the four countries, broken down by place of residence, ethnicity, region, and SES. In Bolivia, Ecuador, and Peru, the historic development of the Inca Empire and other indigenous civilizations took place mostly in the Andean highlands (the altiplano high plateau, in the case of Bolivia). During the colonial period (from the 1530s to the 1820s), an important fraction of both the indigenous population and the white elite remained concentrated in the highland/altiplano areas, with the exception of the city of Lima, Peru, which is located on the Pacific coast. Most of the migration to the lowlands of the coast or other tropical lands took place during the republican period, particularly from the late nineteenth century to the present, as an outcome of export expansion, urbanization, and economic diversification. As a result, most of the deprived indigenous population still live in the highlands, where traditional social structures prevail. Consequently, in general for a particular country, a defined geographic difference exists between this highland region and the other, nonhighland areas of the rest of the country. Colombia, however, has a different regional configuration, as the Andean mountains there split into two divergent chains, and its indigenous population is smaller and culturally less defined.
In Bolivia, Ecuador, and Peru the national prevalence of chronic malnutrition is high, around 26% to 27%, while Colombia has a notably lower rate of 15% (Table 3). Among the eight countries of Latin America and the Caribbean with DHS III surveys (Bolivia, Brazil, Colombia, the Dominican Republic, Guatemala, Haiti, Nicaragua, and Peru), only two of them have worse conditions than do Bolivia, Ecuador, and Peru: Guatemala (46% chronic malnutrition in 1998) and Haiti (32% in 1994) (16, 17).
Urban-rural differences are extremely high in Peru, with a rural chronic malnutrition prevalence about three times as high as that of Lima. Both Bolivia and Ecuador also have large urban-rural differences. Colombia, in contrast, has a relatively low pattern of disparity, with a 1.5 ratio between the rates in the rural areas and in the large cities.
The ethnic divide is striking in Bolivia, Ecuador, and Peru (Table 3). In all three cases, prevalence among indigenous peoples is more than twice that of the rest of society. Similarly, the highland region of each of the three countries has a prevalence rate that is substantially above the rate in the lowland areas of the respective country.
Regional disparities in Colombia (not presented in Table 3) are smaller and not statistically significant at the 5% level. The DHS survey of Colombia defines five main regions, and none of them departs more than 3 percentage points from the national average of 14.9%. The Pacific region of Colombia has the country's highest stunting rate, 17.3%, while the central region has the lowest, 13.2%. Even at the smaller, departmental level, Colombia has only one outlier. That is the southwestern department of Nariño, which borders the highland area of neighboring Ecuador and has a prevalence of 35% (Nariño's DHS sample had 158 valid cases, a relatively small number).
From a socioeconomic perspective, the malnutrition patterns are less clearly defined. Peru has the strongest association between stunting and social inequality, due to both its concentration index (-0.31) and its extreme ratio (9.5 to 1) between stunting in the highest and lowest deciles (Table 3). (A high concentration index value indicates a strong association between social inequality and stunting.) Bolivia, Colombia, and Ecuador have lower levels of socioeconomic inequality in stunting. Later in this article we will analyze why the socioeconomic patterns in the four countries are less clearly defined than are the other results. As these descriptive data suggest, specific regional or ethnic differences may be blurred in national socioeconomic gradients.
The socioeconomic gradient in child malnutrition
There is growing evidence linking health and socioeconomic status, both in industrialized and developing countries (18). Given that, it is worth exploring the size and shape of the gradient between SES and stunting in the four Andean countries of our study and also in the various regions within those countries.
As shown in Figure 1, all four countries follow a clear linear trend. That figure presents the calculated smoothed locally weighted regression curve (LOWESS regression), based on average stunting estimates for each percentile of SES for the countries. Bolivia, Ecuador, and Peru each has a high gradient, with a linear slope ranging from -0.35 in Ecuador to -0.49 in Peru. Colombia has the lowest gradient, with a slope of -0.22. ("High" and "low" refer to absolute values, without regard to the negative sign of the slopes.)
Hypothetically, at least two factors could explain the difference between Colombia and the three other countries. One possibility is per capita income. Colombia's per capita income is the highest of the four nations. Child malnutrition is often the result of poverty and of underconsumption and inadequate access to protein and micronutrients. Higher per capita incomes are linked with reduced poverty and better access to food and basic goods, even in the highly inegalitarian situation found in these four Andean countries.
The argument concerning income, however, fails to explain all the results of our study. For example, malnutrition prevalence is very similar in Bolivia and Peru, in spite of the fact that the per capita income is twice as high in Peru.
There is a second possible explanation for the differences between Colombia and the three other Andean countries of our study: divergent long-term development patterns. Rosemary Thorp has compared the process of social transformation, institution-building, and economic growth in Peru and Colombia, from the mid-nineteen century to the present (19). According to Thorp, the two countries followed very different paths and achieved very different results. In Colombia, the different regions of the country were more similar in their economic growth. This was due to the way in which the coffee economy and other export sectors developed, as well as patterns in land ownership and foreign investment in the country. Colombia developed a stronger institutional capacity in the public sector, leading to a more stable and decentralized process of economic growth. Social expenditure has been stronger and more efficient in Colombia, resulting in higher education levels and lower rates of infant mortality than in Peru. Peru was affected by an extreme regional concentration of wealth in Lima and lower efficiency in public institutions. These patterns resulted from specific traits of the historical development of exports of guano, sugar, copper, and oil, and they led to weaker social and geographical diffusion of economic growth. Nevertheless, there was an overall weakness of social development in both countries, according to Thorp (19). Viewed from a broader perspective, her argument suggests the importance of human and institutional factors such as education, transparency, decentralization, and good governance.
The highland regions versus other areas
The socioeconomic effect, however, explains only a part (albeit a significant one) of the differences in child malnutrition found in the four Andean countries of our study. To analyze regional effects more closely, Figure 2 explores regional SES-stunting functions, differentiating between the highland area and the other regions of the country for Bolivia, Ecuador, and Peru. Regional inequalities in Colombia are small and thus are not plotted.
Figure 2 shows clearly that there is a higher stunting prevalence in the highland area than in the remaining regions of each of the countries. For a given national socioeconomic level, the predicted stunting rate in the highland region of a country is about 10% higher than the rates in the other parts of the same country. The slopes, however, are similar for all the curves.
Table 4 provides a detailed analysis of linear regression models for Bolivia, Ecuador, and Peru. In all three countries there is a statistically significant difference in the intercept between the highland region and the other regions of the country, at the 5% significance level. Slope differences, on the other hand, are rather small.
Regional differences in stunting prevalence exist in Ecuador, Peru, and Bolivia, in addition to the socioeconomic gradient. In all three countries, highland prevalence rates are about 10% higher than those of the other regions.
A recent country study of Ecuador provides an explanatory hypothesis for these regional differences within each country. An analysis of apparent food consumption reported in the LSMS survey showed that poor rural households in the highland area have a diet mostly based on carbohydrates from potatoes and flour, with insufficient intakes of protein and micronutrients (20). In contrast, in Ecuador's coastal and Amazon basin regions, fish and other protein sources are important components of the diet even in poor households.
A multiple regression model presented in the same study (20) identifies four social determinants of stunting. Two of those factors are socioeconomic: aggregate per capita consumption and mother's educational level. In addition, there are ethnic and regional effects, with risks being higher for indigenous households and for highland households.
An international SES index
National percentiles of SES are specific to each country and do not permit international comparisons of living conditions. For example, Colombian children have better social indicators than do their Bolivian counterparts in the same percentiles.
In order to compare data across countries, we developed an internationally comparable SES index from DHS III surveys in eight Latin American countries: Bolivia, Brazil, Colombia, the Dominican Republic, Guatemala, Haiti, Nicaragua, and Peru. Our index makes it possible to analyze the effect of absolute changes in living conditions on stunting. The resulting LOWESS curves for three of the Andean countries are plotted in Figure 3.
Curvilinear effects are visible in the figure, with different slopes for different SES levels. The curves follow a logistic-function path, with the slopes generally being steeper in the curves' first half than they are in their second half. This pattern suggests steeper relationships between SES and stunting at low to medium levels of economic development. However, the very poorest households have slightly lower slopes, so the initial reductions in stunting, for given socioeconomic improvements, will be comparatively small. In general, the reductions in stunting coming from SES improvements depend on socioeconomic level, with the benefits of reduced stunting being strongest among poor to medium-status households and then flattening out among better-off households.
The gradient in Colombia remains flatter than for the two other countries. Even comparing similar living standards, malnutrition rates in Colombia are lower than those in Bolivia and Peru. Both the intercept and the gradient seem different. This finding supports the hypothesis that, in addition to differences in absolute living conditions, Colombia's nutritional profile is better than that of its counterparts. Among the factors that might explain the differences are a consistently higher long-term investment in education and health, a more balanced regional economic configuration, and more efficient public policies.
The stunting rates in the four Andean countries of our study rank among the worst in Latin America and the Caribbean, being surpassed only by Guatemala and Haiti, in the group of countries with available information from DHS surveys. Lower SES was associated with stunting in all four of the Andean countries. Bolivia, Ecuador, and Peru have steeper gradients than Colombia does. The difference is only partially explained by economic factors, and the influence of consistently more efficient public sector performance as well as higher social expenditures in Colombia is suggested.
The explanatory factors for national, regional, and ethnic differences in child malnutrition need to explored further. Of particular importance would be a closer examination of the social and ethnic determinants of stunting as well as their contributions to achieving nutritional improvements.
In the four countries analyzed, and particularly in Bolivia, Ecuador, and Peru, malnutrition remains widespread. It contributes to the intergenerational perpetuation of poverty and inequality and limits prospects for human development.
The strong socioeconomic gradient in stunting confirms the need to reduce inequality and to focus resources on the poor, instead of working only to reduce the national averages without regard to the social distribution of the improvements. Moreover, the concavity of the SES-stunting function (Figure 3) suggests there will be higher returns for social investments made among those who belong to the poor and middle social strata.
The steep SES gradient confirms the strategic importance of education, housing, and employment as elements of policies aimed at improving nutrition. Rather than operating in isolation, health programs should be integral components of comprehensive social policies. The Colombian experience suggests the long-term importance of institutional factors as well as the returns that can come from sustained social and health investment.
The effects of specific regional or ethnic factors in Bolivia, Ecuador, and Peru might be reduced by better targeting resources to the deprived rural highlands regions. Another remedy might be to implement participatory programs integrated into the fabric of indigenous cultures. Further, enriching the diet of persons in highland regions could help in the development of those areas.
Acknowledgments. The authors would like to acknowledge the academic contribution and valuable comments of Dr. Ichiro Kawachi to this paper as well as the institutional research support from the Harvard Center for Society and Health and from the Pan American Health Organization.
1. Wagstaff A, Watabane N. Socioeconomic inequalities in child malnutrition in the developing world. Washington, D.C.: World Bank; 2000.
2. Smith I, Haddad I. Overcoming child malnutrition in developing countries: past achievements and future choices. Washington, D.C.: International Food Policy Research Institute; 2000.
3. Semba R, Bloem M, eds. Nutrition and health in developing countries. Totowa, New Jersey, United States of America: Humana Press; 2001.
4. de Onis M, Frongillo E, Blössner M. Is malnutrition declining? An analysis of changes in levels of child malnutrition since 1980. Bull World Health Organ 2000;78(10):1222-1233.
5. Londoño L. Poverty, inequality and human capital development in Latin America, 1950-2025. Washington, D.C.: World Bank; 1996.
6. United Nations Economic Commission for Latin America and the Caribbean. The equity gap: a second assessment. Santiago de Chile: ECLAC; 2000.
7. Macro International Inc. Demographic and Health Survey (DHS), Colombia, 1995 [electronic database]. Available from: http://www.measuredhs.com. Accessed 29 February 2000.
8. Macro International Inc. Demographic and Health Survey (DHS), Peru, 1996 [electronic database]. Available from: http://www.measuredhs.com. Accessed 29 February 2000.
9. Macro International Inc. Demographic and Health Survey (DHS), Bolivia, 1997 [electronic database]. Available from: http://www.measuredhs.com. Accessed 29 February 2000.
10. World Bank, Instituto Nacional de Estadísticas y Censos. Living Standard Measurement Survey, Ecuador, 1998 [survey database]. Washington, D.C.: World Bank; 1999.
11. Van de Geer J. Multivariate analysis of categorical data: theory. London: Sage; 1993.
12. Van de Geer J. Multivariate analysis of categorical data: applications. London: Sage; 1993.
13. Meulman J, Heiser W. SPSS Categories 10.0. Chicago, Illinois, United States of America: SPSS; 1999.
14. Meulman J. Optimal scaling methods for multivariate categorical data analysis. Leiden: Leiden University, 2000.
15. United Nations Development Program. Human development report. New York: Oxford University Press; 2001.
16. Macro International Inc. Demographic and Health Survey (DHS), Guatemala, 1998 [electronic database]. Available from: http://www.measuredhs.com. Accessed 29 February 2000.
17. Macro International Inc. Demographic and Health Survey (DHS), Haiti, 1994 [electronic database]. Available from: http://www.measuredhs.com. Accessed 29 February 2000.
18. Kawachi I, Berkman L. Social epidemiology. New York: Oxford University Press; 2000.
19. Thorp R. Economic management and economic development in Peru and Colombia. London: Macmillan; 1991.
20. Larrea C, Freire W, Lutter C. Equidad desde el principio: situación nutricional de los niños ecuatorianos. Washington D.C.: Organización Panamericana de la Salud; 2001.
Desigualdad social y malnutrición infantil en cuatro países andinos
Objetivo. Analizar los efectos de las condiciones socioeconómicas, regionales y étnicas en la desnutrición crónica en Bolivia, Colombia, Ecuador y Perú.
1 Facultad Latinoamericana de Ciencias Sociales ¾ Sede Ecuador, Quito, Ecuador, and Harvard Center for Society and Health, Boston, Massachusetts, United States of America. Send correspondence to: Carlos Larrea; e-mail: firstname.lastname@example.org
2 Food and Nutrition Program, Pan American Health Organization, Washington, D.C., United States of America.
3 Larrea C. Inequidad social, salud reproductiva y nutrición en ocho países de América Latina: análisis comparativo de las Encuestas DHS III [unpublished paper]. Washington D.C.: Pan American Health Organization; 2000.