Socio-Economic Status and Health

The significance of this topic hardly needs arguing. Health is clearly one of the most important dimensions of personal and family well-being. Causal pathways from health to all measures of SES-income, wealth, and education- are very important and potentially run in both directions, while third factors may affect both. The study of the SES-health nexus requires extensive, complex and longitudinal datasets. To understand the effect of policy on outcomes, studies that take an international comparative perspective are of great importance.

Contributions by PRC Staff

Recent research of PRC staff has used a variety of data sources to explore a number of issues, including:

Disparities in Adult and Child Health

Loughran, Datar, and Kilburn (2004) show that birth weight, net of family fixed effects, is a significant predictor of childhood test scores and that parental investment can significantly mitigate the effects of low birth weight. As another dimension of adult health, Lakdawalla has studied adult obesity, contributing both theoretical insights and empirical analysis to understand the prevalence and consequences of obesity (Lakdawalla and Philipson, 2002b). In related work, Finch has contributed to our understanding of health disparities among children. In particular, he has examined the effects of race and birth weight on respiratory disease and infant mortality (Finch, Boardman and Hummer, 2001; Finch, Frank and Hummer, 2000), with a broader interest in understanding how differences in birth outcomes contribute to SES differentials later in life (Finch 2002a; Finch 2002b). Beckett has advanced research on health disparities among adults, providing evidence that converging health inequalities between groups defined by SES cannot be explained by mortality selection (Beckett, 2000).

In a paper that was selected as the “Best Research Article of the Year” by the National Institute for Health Care Management, Goldman and Smith (2002) show that differential adherence by education levels to complex treatment regimes can help explain socio-economic differentials in health status and that such differentials could be overcome with training, monitoring, and other interventions. In a recent paper, Goldman and Smith (forthcoming) show that the diffusion of medical drug technology for the treatment of hypertension played little role in exacerbating the SES gradient in health.

Smith has addressed the two-way interactions between SES and health in several papers. Smith (1999) set out the issuses involved and suggested analytical methods of estimating these effects that have been largely followed in the subsequent literature. Smith (2005) estimates the impact of new unexpected health events on susequeunt SES meaures and finds relatively small impacts on out-of-pocket medical expenses and much larger effects on labor supply, income, and wealth. In another paper, Smith (2004) shows that childhood health events have large future impacts on several adult SES measures, including schooling, labor supply, and family income. These effects persist and are even stronger when estimated within siblings.

In several papers Hurd with co-authors has addressed two issues. The first is to use longitudinal data to learn about the causal flows between SES and health. To test for causality from SES to health, the method controls for baseline health and finds whether the onset of a new health condition is related to SES. To test for causality from health to SES, the method controls for SES and finds whether health causes a transition in wealth. Adams et al (2003) find only weak links from SES to health and mixed evidence of a link from health to wealth change in the population aged 70 or over. Van Soest and Michaud extend the analysis of Adams et al. (2003) using techniques for dynamic panel data models controlling for unobserved heterogeneity, applied to the HRS cohort.

In a related paper Hurd, McFadden and Merrill (2001) find that subjective survival conditional on SES and health has predictive power for in the population aged 70 or over. Hurd and Kapteyn (2003) compared the relationship between SES and health in the U.S. and the Netherlands with the objective of finding how the relationship is affected by public policy. In the Netherlands the social safety net weakens the relationship by providing insurance against income decline in the case of a wealth shock and universal access to health care services. Kapteyn, Smith and van Soest (2004) analyze cross country differences in work disability, using vignettes to correct for differences in response scales between countries and socio-economic groups. They exploit the RAND Internet survey, an ongoing RO1 project with Kapteyn as PI, giving the opportunity to experiment with innovative ways of asking questions, improving identification of response scale and genuine work disability differences.

Back to Top

Population and the Environment

Sastry has documented significant effects on mortality in Malaysia from air pollution associated with an extremely widespread series of forest fires in Indonesia in 1997 that were exacerbated by the El Niño weather phenomenon (Sastry, 2002). His estimates indicate that the effects of a high air pollution day associated with the smoke haze increased total all-cause mortality by 22 percent. The presence of such mortality effects serve as a sentinel indicator for other correlated, but unmeasured, morbid outcomes.

Back to Top

Health and Educational Outcomes in Childhood

Datar and Sturm have studied the effect of childhood obesity on children’s academic and social-behavioral outcomes in the early school years, using national data on kindergartners (Datar & Sturm 2004 a & b). Using the same data, they have also examined whether school physical education programs are effective in combating obesity among young children (Datar & Sturm 2004c). They find that increased physical education instruction in the early school years reduces body mass among girls significantly.

Back to Top

Health of Immigrants

Smith and his collaborators have used data from the New Immigrant Survey to examine prospective changes in the health of new legal immigrants. This research demonstrated the large empircal importance for health selection in migrant flows to the United States. The findings indicate that immigrant health may actually improve over time rather than deteriorate, as suggested by prior research. More generally, this research combines traditional demographic methods such as cohort analysis with data on economic outcomes of immigrants.

Back to Top


Maestas has studied the effect of the Medicare program on health care utilization and health, finding a large increase in utilization as individuals cross the eligibility threshold at age 65, small gains in health, and a narrowing of disparities in access to care (Card, Dobkin and Maestas, 2004). Disparities in utilization of hospital care do not narrow after 65, and ongoing work aims to understand why blacks in particular do not share in the gains in hospital care experienced by whites. In related work, Maestas is examining the effect of medical expenditure risk on household financial decisions, finding that higher risk causes older individuals to substitute toward safer financial investments (Maestas and Goldman, 2004). The work suggests that Medicare reforms, such as the new prescription drug benefit enacted by the Medicare Modernization Act of 2003, may have spillover effects on financial markets.

Back to Top

Future Directions and Scientific Objectives

Through these funded and pending projects and future projects still in the planning stages or further on the horizon, our future research in this area has two primary scientific objectives:

  • To deepen our understanding of the measurement and meaning of health status, considering various dimensions of physical and mental health for children and adults
  • To extend prior investigations of the interrelationship between health and socio-economic status to more fully understand the feedback effects throughout the life course

Examples of these new directions include:

Child and Adult Health

PRC staff will use a number of databases to examine issues related to child and adult health. In terms of child health, Loughran is using longitudinal data from the PSID and NLSY to address different dimensions of the long-run consequences of early child health. More generally, he will examine the role of parental decisions and other determinants of child health, survival, and physical and cognitive development, and outcomes in adulthood such as education, health status, earnings and occupational status. Sastry will explore aspects of child and adult health and their determinants based on data from the L.A.FANS, while Bird is using data from a nationally representative database of individuals in treatment for HIV to examine differences across men and women in health care utilization and the relationship to risk taking behavior. Maestas and Buchinsky will study issues in the measurement of health status over time, with attention to how individuals assess their current health status and whether self-assessments shift with changes in the economy, changes in medical technology, and reference groups.

Sastry is actively pursuing research on SES and health in two different settings. First, with funding from NICHD he is studying socio-economic inequalities in child health and survival in Brazil. In on-going research, Sastry is examining trends in socio-economic inequalities in under-five mortality in Brazil over several decades during which the mortality transition unfolded. In a recent paper, Sastry (2004) found that mother's education emerged as the key factor underlying socio-economic inequalities in under-five mortality even as levels of education for women increased and inequality in schooling fell. In current work he is extending his look at other measures of socio-economic status and is examining the pathways through which socio-economic status affects children's health and survival. Second, Sastry is collecting new biomarker data on stress and health for adults and children in Los Angeles as part of Wave 2 of L.A.FANS, including cortisol (a stress hormone), blood pressure, C-reactive protein (a marker of acute inflammation), Epstein-Barr virus antibodies (a marker of immune function), cholesterol, diabetes, and pulmonary function. With funding from NIEHS and OBSSR, he and colleagues plan to investigate the effects of socio-economic status--as well as social context and family environment--on these physiological markers of stress and health across the lifecourse.

Other Relationships between Health and Socio-Economic Status

Other future work in the Center will continue to investigate the interrelationships between various dimensions of health and socio-economic status. For example, Van Soest will analyze the interaction mechanisms between neonatal mortality and birth spacing in India, allowing for unobserved heterogeneity in mortality ("frailty") as well as birth spacing ("fecundity"). Hurd and Kapteyn plan to extend their work on cross-national analyses of the relation between SES and health in two different ways. On the one hand, they plan to exploit the new and rich datasets that are becoming available in Europe (including ELSA and SHARE) to improve the robustness and accuracy of their empirical results. On the other hand, they are developing formal models of the relation between SES and health that generate precise predictions of patterns we should observe in different countries with different institutions. As part of the new program project “Economic and Health Determinants of Retirement,” Kapteyn, Hurd, and co-authors will study the role of health and workplace conditions in retirement models, exploiting data from the HRS, SHARE, and ELSA.

Goldman and Lakdawalla will advance their research examining persistent health disparities across education groups. Future work by Bird on the social determinants of health status, including gender roles, will also contribute to this area. Karney and his colleagues are planning to extend their work on stress and marriage to examine the mechanisms through which socio-economic status affects the health of family relationships. Finally, Goldman, Hurd, and Lakdawalla are building a demographic model that incorporates health status, disease, and disability to forecast future trends in health care expenditures. This modeling capability, which bridges economic and demographic models, has already been applied to the demand for nursing home care (Lakdawalla et al., 2003) and will have broader applicability to consider the interrelationships between health and other demographic and economic outcomes. Recently Goldman was a awarded a Roybal center for health policy simulation in which this activity will be further expanded.