Decomposing world health inequality

https://doi.org/10.1016/S0167-6296(02)00123-6Get rights and content

Abstract

This study explores global inequality in health status and decomposes it into within- and between-country inequality. We rely on standardized height as our health indicator since it avoids the measurement pitfalls of more traditional measures of health such as morbidity, mortality, and life expectancy. It also avoids measurement problems associated with using monetary variables such as income or expenditure across time or place to compare welfare. Our calculation of world height inequality indicates that, in contrast with similar research on income inequality, within-country variation is the source of most inequality, rather than the differences between countries.

Introduction

This study documents global inequalities in health in a novel way. Rather than considering the health of people across the income distribution or some other measure of social stratification, we examine inequality in health status in and of itself.1 We are motivated by the fact that health is an important indicator of well-being. Just as measuring the dispersion of income is of interest, so too are statements about inequality in health status. Although the nature of health inequality differs in many important respects from the widely understood measures of income inequality, the measure of health inequality that we propose has some useful and convenient properties that make it particularly attractive and interesting. In addition, while our focus is on measurement and decomposition of global health inequality, our approach has potentially important applications to the broader objective of defining health inequality among individuals. We thereby seek to promote the consideration of health inequality as a worthy topic for research, independent of its correlation with other aspects of well-being, and of similar importance to income inequality.

Most of the literature on health inequality explores how health differs across various socio-economic dimensions. The positive correlation or “gradient” between health and socio-economic status has led researchers to focus on income-related inequalities in health status and access, or on the importance of relative income or social position as a determinant of health (e.g. Wagstaff et al., 1991, Contoyannis and Forster, 1999, Preston and Taubman, 1994, van Doorslaer et al., 1997, Mackenbach and Kunst, 1997, Navarro, 1998, Hummer et al., 1998, Gwatkin, 2001, Brockerhoff and Hewett, 2000).2 Much of this literature is thus grounded in a welfare framework where the fundamental measure of well-being is income, and differences in health outcomes are analyzed across the income (or expenditure) distribution. There are also papers that study how health varies across other socio-economic dimensions such as ethnicity, geographical location, gender, and assets (Minujin and Delamonica, 2002, Gwatkin et al., 1999).3 The problems of poor health status and health inequality are seen as a result of income inequality, or an underlying social process that contributes to inequality among socio-economic groups in the population. Murray et al. (1999) point out that the magnitude of health inequality measured in this way is conditioned by the critical choice of what variable is used to disaggregate the population into social groups.

Empirically, studies often find a positive correlation between health and many indicators of socio-economic status or various measures of social stratification. Nevertheless, the correlation between health and these other social indicators, including income and expenditures, is sufficiently weak that in any given sample, income and the measures of social stratification usually predict only a small portion of the variation in health status. This applies to a variety of health measures, including the standardized heights of children that we employ in this paper. Many other factors are important in explaining health status. A wide variety of social and economic circumstances and behaviors matter, including: the psychological state of the primary care-giver, weaning and other feeding practices, the social norms and behaviors that govern sexual transmission of diseases, and the natural occurrence of trace minerals and vitamins available in soils and foods. Community factors often matter, too. The availability and quality of the health care system and related public health measures such as water and sanitation, vaccination coverage, etc. have all been shown to be of equal or greater importance than income in determining child health.

This point is illustrated in the literature that looks at the determinants of health status, both at national and individual level. At the national level, a wide variety of cross-country studies show that while national incomes play a role in determining health status, there are other important factors such as health expenditures, social service infrastructure, education, and environmental infrastructure.4 Efforts to examine the determinants of health status at the individual level also show that, in some cases, income is not an important characteristic in explaining child malnutrition. For example, Skoufias (1998) models child health during the economic transition in Romania and finds that income does not matter in urban areas. Sahn and Alderman (1997) report a similar finding in Maputo, Mozambique for children less than 3 years of age, as do Thomas et al. (1996) for urban Côte d’Ivoire.5 In these and other cases,6 there are both a large number of other community infrastructure variables (e.g. sanitation and water, drug availability, and health clinic quality), prices, and household level covariates (e.g. education of various members, household composition) that affect child health outcomes. Even if these factors are positively correlated with incomes, or some other socio-economic dimension, they have independent and important effects on health. By focusing solely on income-related inequality in health, one therefore runs the risk of ignoring a large fraction of health inequalities, namely those that are uncorrelated with income. By implication, reducing income inequality will not necessarily be an effective way of dealing with inequality in health, since the gradient fails to capture the fact that a wide range of unobservable factors beyond income influence health. This is especially the case because income distributions have a long right tail.

That health inequality is intrinsically important, regardless of the correlation with income, is well illustrated by considering two populations, A and B, with equal levels of average health and equal levels of health inequality. However, assume that in population A, there is a strong correlation between health and income, and in B, just the opposite. We would certainly not want to adopt the view that health inequality in population A is a more serious public policy problem, owing to the stronger correlation with income (or some other measure of social stratification). To the extent that we can identify a cardinal measure of health inequality, which we do in this paper, comparisons of distributions of health are meaningful, regardless of whether health is correlated with welfare measured along other dimensions (Deaton, 2001).

Thus, in this paper we focus on inequalities in child health status, not the correlations between health status and other socio-economic indicators or the “gradient” as it is commonly termed. Perhaps the simplest way to distinguish what we do in this paper from the traditional approach is that our “univariate” approach7 orders individual well-being by health status or health condition, not income levels, and describes the inequality in health status across this health ordering.

A number of World Health Organization (WHO) policy statements and papers strongly articulate the need to reduce the differences in health status between countries and between socio-economic groups within countries (e.g. WHO, 1985, WHO, 1986, Whitehead, 2000). These calls beg the fundamental question as to why health inequality is of interest, as does our decision to focus on health inequality in and of itself. A number of considerations motivate this interest. First, Sen, 1979, Sen, 1985, Sen, 1987 argues that the notion of poverty is inadequately captured by income or expenditure. Poverty is the deprivation of basic capabilities, or the failure of certain basic functionings, not just low levels of income. Low incomes are only instrumentally significant, while deprivation of capabilities, such as poor health, are intrinsically important. Health, literacy, and so forth, are more direct measures of capability deprivation, or poverty, than income or expenditures. And to the extent that measures of health are appropriate arguments in the social welfare function, there are legitimate reasons to be concerned about health inequality directly.

Second, health inequality is more likely to reflect some notion of absolute deprivation in the population than does income inequality. This results from the fact that in examining incomes, it is possible, and in fact often the case, that an increase in inequality is caused by a lengthening of the right-hand tail of the distribution. In addition, observed increases in income inequality can be offset by increases in mean incomes, implying greater social welfare despite worsening inequality. In health, unlike incomes, there is a natural limit to improvements so that the distribution of health has no long right-hand tail. Worsening health inequality is therefore more likely to be caused by greater dispersion in the left-hand tail, i.e. by unhealthy people becoming more so. So it is more likely that health inequality is bad, at least in the context of social welfare functions that put more weight on the welfare of the poor.

A third argument for focusing on health inequality derives from the assertion of Wilkinson, 1996, Wilkinson, 1997 that lack of social cohesion and other disparities in socio-economic circumstances, including health status (as well as other considerations such as security), are health risks. Such risk goes beyond the correlation between low levels of income and low level of health. Wilkinson’s case rests on what he refers to as the “neuroendocrine” pathways through which psychosocial risk factors link health to the inequality in socio-economic circumstances. That is, psychosocial effects of relative deprivation, as might be measured by inequality of health status, in and of itself, are risk factors for poor health.

Finally, there is a practical reason for focusing on inequality in health, rather than incomes. It is often difficult to compare incomes and socio-economic correlates of health across time or place. Constructing income and expenditure measures is complex, especially in poor countries where analysts face numerous challenges. These include imputing the costs of home production and housing, valuing the rental value of durables, dealing with the problems that arise when own-account enterprises and self-employment make up large shares of consumption, the related difficulties of calculated profits from self-employment, and so forth. Furthermore, even minor deviations in survey design, such as the recall period or the number of commodities in the consumption module, can influence the outcomes of these surveys (e.g. Pradhan, 2000, Scott and Amenuvegbe, 1990). There is also the problem of converting nominal incomes into comparable units across time and place. For inter-temporal welfare comparisons, this requires appropriate price deflators, but these can be inaccurate at high inflation rates (Escobal and Castillo, 1994). Spatially, markets are poorly integrated in poor countries, prices vary dramatically, and reliable regional price data are scarce. Cross-country welfare comparisons require accurate purchasing power parity indices. High-quality indices of both types are scarce for poor countries. Finally, there is the difficulty of defining comparable socio-economic groups, which in fact, may have different compositions in different countries. For example, making urban versus rural, or male- versus female-headed household comparisons across countries may not be very meaningful if the definition of these categories and their determinants differ.

In contrast, many measures of health status, and particularly anthropometrics, are easily comparable across time and location. They are not plagued by the problems of survey design, comparability of nominal units, or socio-economic definitions that affect other welfare measures. So, we are better able to make inter-temporal and inter-country comparisons when focusing on an objective health outcome indicator, rather than trying to define comparable income levels of socio-economic groups, the latter of which also tend to change in composition over time.

While our primary concern in this paper is methodological, we do uncover an especially interesting policy-relevant finding on the extent to which global health inequality is a result of intra- versus inter-country variation in health status. Specifically, we describe global health inequality of young children, and decompose it into the contribution of within- and between-country inequality. As a first attempt to deal with health inequality in this way, the objective of this paper is modest insofar as none of the work is explanatory, merely descriptive. Nonetheless, the results give a sense of how serious health inequality is in the world, and the decomposition indicates the relative importance of intra- and inter-country inequalities.

The remainder of this paper is organized as follows. First, we discuss in greater detail how to construct and standardize a decomposable health inequality measure using children’s height. This is followed by a discussion of the data that we employ. We next present the results of our decomposition analysis. This is followed by some concluding remarks, including how our results compare to similar exercises that examine global income inequality.

Section snippets

Choice of health indicator

Health can be measured and characterized over various dimensions.8 Clearly, no single indicator, or even narrow group of indicators, captures health in all of these. In this paper we rely on a single, widely used indicator of health, the height of pre-school age children.9 We do so

Data

The data used in our analysis are primarily from the Demographic and Health Surveys (DHS), a project that has collected representative data on child health, fertility, contraceptive use, and related demographic data throughout the developing countries for the past 15 years.17 The data are all collected in a comparable fashion, following a set of detailed

Results

In Table 2 we initially present the country-specific levels of stunting, using the −2 height-for-age z-score as the cut-off point, in keeping with standard procedure. In addition, we show a ranking of countries from the least to the most healthy child populations. The table also includes transformed average height in centimeters, as well as the standard deviation, using girls of 24 months as the reference group. We can see that there is a large divergence in the levels of child health across

Discussion

In this paper we have developed an innovative approach to measuring, and decomposing world health inequality. We are motivated by the underlying proposition that quantifying health inequality, like income inequality, is an important area of research. Health is a well-recognized measure of well-being that arguably belongs in a social welfare function. To borrow Sen’s terminology, poor health is an intrinsically important measure of capability deprivation. Thus, reducing inequality in health

References (80)

  • S. Anand et al.

    Human development in poor countries: on the role of private incomes and public services

    Journal of Economic Perspectives

    (1993)
  • Appleton, S., Song, L., 1999. Income and Human Development at the Household Level: Evidence from Six Countries. Mimeo,...
  • Baten, J., 2002. Did partial globalization increase inequality? Did inequality stimulate globalization backlash? The...
  • Beaton, G.H., et al., 1990. Appropriate uses of anthropometric indices in children: a report based on an ACC/SCN...
  • Behrman, J.R., Deolalikar, A.B., 1988. Health and nutrition. In: Chenery, H., Srinivasan, T.N. (Eds.), Handbook of...
  • J. Bound

    Self-reported objective measure of health in retirement models

    Journal of Human Resources

    (1991)
  • M. Brockerhoff et al.

    Inequality of child mortality among ethnic groups in sub-Saharan Africa

    Bulletin of the World Health Organization

    (2000)
  • P. Bustos

    Growth in indigenous and nonindigenous Chilean schoolchildren from three poverty strata

    American Journal of Public Health

    (2001)
  • H.R. Carr

    Time trends in inequalities in health

    Journal of Biosocial Science

    (1988)
  • Deaton, A., 1999. Inequalities in income and inequalities in health. NBER Working Paper...
  • Deaton, A., 2001. Relative deprivation, inequality and mortality. NBER Working Paper...
  • Deolalikar, A., 1992. Nutrition and health. In: Roy, S., James, W.E. (Eds.), Foundations of India’s Political Economy:...
  • M. de Onis et al.

    Is malnutrition declining? An analysis of changes in levels of child malnutrition since 1980

    Bulletin of World Health Organization

    (2000)
  • Dow, W., et al., 1997. Health care prices, health and labor outcomes: experimental evidence. Labor and Population...
  • Escobal, J., Castillo, M.C., 1994. Sesgos en la medición de la inflactión en contextos inflacionarios: el caso Peruano....
  • G. Firebaugh

    The trend in between-nation income inequality

    Annual Review of Sociology

    (2000)
  • E.E. Gakidou et al.

    Defining and measuring health inequality: an approach based on the distribution of health expectancy

    Bulletin of the World Health Organization

    (2000)
  • Government of Australia, 1997. National Nutrition Survey Australia 1995. Australian Bureau of Statistics and...
  • Graitcher, P.L., Gentry, E.M., 1981. Measuring children: one reference for all. Lancet ii,...
  • Gwatkin, D.R., 2001. Poverty and inequalities in health within developing countries: filling the information gap. In:...
  • Gwatkin, D.R., Rustein, S., Johnson, K., Pande, R., Wagstaff, A., 1999. Socio-Economic Differences in Health,...
  • Habicht, J.-P., et al., 1974. Height and weight standards for preschool children. How relevant are ethnic differences...
  • Haddad, L., Hoddinott, J., Alderman, H. (Eds.), 1997. Intrahousehold Resource Allocation in Developing Countries:...
  • J. Haughton

    Falling fertility in Vietnam

    Population Studies

    (1997)
  • Hill, A., Mamdani, M., 1989. Operational Guidelines for Measuring Health Through Household Surveys. Mimeo, Centre for...
  • R.A. Hummer et al.

    Sociodemographic differentials in adult mortality: a review of analytic approaches

    Population and Development Review

    (1998)
  • A. Kroeger

    Response errors and other problems of health interview surveys in developing countries

    World Health Statistics Quarterly

    (1985)
  • Martorell, R., Habicht, J.-P., 1986. Growth in early childhood in developing countries. In: Falkner, F., Tanner, J.M....
  • R. Martorell

    Acute morbidity and physical growth in rural Guatemalan children

    American Journal of Diseases in Childhood

    (1975)
  • Mata, L., 1978. The Children of Santa Maria Cauque: A Prospective Field Study of Health and Growth. MIT Press,...
  • Cited by (133)

    • Child health inequality in Sub-Saharan Africa

      2022, Economics and Human Biology
      Citation Excerpt :

      We gather information from comparable household surveys carried out under the Demographic and Health Surveys (DHS), covering 33 SSA countries in the 2009‐2016 period. Starting from the height-for-age z-score (HAZ), our measure of health is the standardized height of children below five years old (relative to the WHO reference standards) (Pradhan et al., 2003), adjusted by the age-gender distribution of children in each country.2 Thus, changes in height inequality along the age distribution are not caused by differences in the composition of child gender and/or age structure across country-years.

    • Inequality and Life Expectancy in Africa and Asia, 1820‐2000

      2022, Journal of Economic Behavior and Organization
    View all citing articles on Scopus

    The research for this paper was conducted while Menno Pradhan was a Visiting Fellow at Cornell University.

    View full text