Research > Research Themes > Human Resources and Endowments > Population heterogeneity: advances in understanding

Resources and endowments can be both observed and unobserved; Penn’s PSC has been at the forefront in the development and application of methods for taking into account unobserved resources and endowments in the measurement of the effects of other resources and events on life chances, including health.

The most well-known class of such models are so-called fixed effects models, in which case multiple observations on primary units—couples, for example, or individuals over time—are used to adjust for unmeasured factors presumed constant within units. Many are differencing models, e.g., (Allison) that hospitalization of a spouse can causally increase the chance that the other spouse dies, but (in a study with P. England) that there is no causal relation between occupational sex segregation and pay; and (Todd and Behrman) that, in Bolivia, investment in early childhood development appears causally, positively related to subsequent health and education—an effect not visible in the crosssection with standard controls. Several papers in Demography (Watkins, Kohler, Behrman) use similar methods to establish that the effects of social networks on reproductive health behavior exist beyond selfselection into networks. This interdisciplinary collaboration, sociology/demography/economics, is mirrored in work by Behrman with Schnittker and Kohler, which uses twin data to investigate the effects of schooling attainment on health in the US, Denmark, and China; schooling appears to proxy for endowments (family background, genetics). Behrman, with P. Taubman, pioneered the use of twins to adjust for unmeasured endowments; other recent applications show (with M. Rosenzweig, in the American Economic Review) that the strong positive association between the schooling of a mother and that of her child are reversed with twin-based adjustment for unobserved endowments and (in Review of Economics & Statistics) that, in something that is not observable in the crosssection, that nutritional investments in utero leading to higher birth weight have a large positive effect on later life outcomes such as educational attainment. The deleterious effects of higher observed birth weight are not caused by nutritional supplements, but reflect genetic and other unobserved endowments.

Other PSC contributions to knowledge of population heterogeneity include Ewbank’s seminal work integrating cohort and clinical data on gene incidence with demographic models for the effects of genetic heterogeneity on mortality; ~15% of variances among European populations in cardiovascular mortality are attributable to the differential prevalence of the APOE gene associated with heart disease and Alzheimer’s. Schurr, through his lead role in the Genographic Project, is at the forefront of the measurement of genetic diversity across self-defined ethnic populations, and their reconciliation with genetic populations, in the sense of homogeneous haplogroups. Rosenbaum, in his canonical text, Observational Studies, in the Journal of the American Statistical Association, and elsewhere, has developed sensitivity tests for measuring hidden bias, which are increasingly being employed in both the health care and social sciences literatures. Ríos-Rull has been a pioneer in the introduction of heterogeneity into agent-based models for overlapping generations, the key methodology for dynamic equilibrium models with increasing applications to understanding macro-demographic processes.

Although attention to population heterogeneity has been very important for the field, an important stream of thought and research at the PSC provides some balance against total preoccupation with a hard-to-see world. The BRCA gene for breast cancer is devastating from a clinical perspective, but does not explain crosspopulation differences in mortality (Ewbank). Preston and Elo reveal, in Demography, that a substantial fraction of the black-white mortality crossover at older ages is attributable to differential misreporting of age, and thus that the elegant theories explaining this crossover may be explaining facts not in evidence. Pauly has recently considered the role of heterogeneity on optimal coinsurance rates in health insurance. There are advantages to having more individual information, but a policy in which coinsurance is lower for services with higher (marginal) benefits relative to costs is superior to providing more information on differences between patients. Polsky shows that lower premiums of managed care plans can not be explained by favorable risk selection of enrollees. Participation in high school athletic programs leads to higher adult wages, but this premium is not primarily due to selection on predetermined characteristics valued in the labor market (Postlewaite in the Journal of Economic Inequality). Smith argues (in Population and Development Review) that the obsession with micro-level unobservables detracts from an understanding of population-level causal interventions in terms of their effects on the relations among individuals that constitute a group or a market, a society or an economy.