Sociodemographic Differentials in Adult Mortality: A Review of Analytic Approaches

Citation: Hummer, Rogers, Eberstein Sociodemographic Differentials in Adult Mortality: A Review of Analytic Approaches.
Internet Archive Scholar (search for fulltext): Sociodemographic Differentials in Adult Mortality: A Review of Analytic Approaches
Tagged: uw-madison (RSS), wisconsin (RSS), sociology (RSS), demography (RSS), prelim (RSS), qual (RSS), WisconsinDemographyPrelimAugust2009 (RSS)

Summary

Hummer et al. indicate that most research on sociodemographic differentials "has been primarily descriptive and focused on issues of accurate reporting, data quality, and measurement error rather than on conceptual development and understanding, and tests of theoretical principles" (p. 555). Relatively few studies have closely examined the pathways by which sociodemographic differentials are translated into differences in mortality. Fewer still have taken a longitudinal or multi-level approach. But, this is changing as demographers are becoming more sophisticated both methodologically and theoretically. Hummer et al. proceed to delineate the contributions of previous approaches, point to promising developments, and make suggestions for future studies. Most previous studies on socioeconomic differentials are strictly demographic in nature. "The central goal of this approach is the accurate documentation of mortality patterns within and between demographic groups, usually by age, sex, and race/ethnicity, but sometimes including such factors as educational level, occupation, and marital status. The most frequently used data for describing adult mortality patterns are death certificates combined with U.S. census data. . . Much of what is known about mortality differentials-some of which is non-intuitive-stems from these careful descriptions" (p. 557). For instance, two phenomena-the epidemiological paradox (lower adult mortality among Hispanics compared to non-Hispanic whites) and the racial mortality crossover (mortality rates among blacks higher than whites at ages up to about 85, and then lower thereafter)-are the result of descriptive research. But, the primary limitation of such research is that it is not explanatory. As a result, speculation regarding the causal mechanisms behind observed differentials is frequently employed. In the 1960s, it became much more common to incorporate socioeconomic variables into studies of mortality. "Thus, demographic and socioeconomic variables came to be seen as inter-correlated and interacting determinants of mortality . . ." (p. 559). For instance, Kitagawa and Hauser's (1973) seminal work demonstrated that income and education have strong, inverse associations with mortality. Since this study, SES variables have come to dominate the work on mortality in the U.S. Importantly, more recent thinking on this subject focuses on wealth (i.e., "stocks") rather than income (i.e., "flows") because of the less transitory nature of the former. Additionally, variables on marital status, other forms of social support and stress have proven important in studies of mortality-particularly marital status. This approach has proven more fruitful than the strictly demographic approach, but it too has limitations, such as the lack of truly explanatory models (i.e., no proximate determinants), and the omission of early life influences. To remedy these problems, Hummer et al. suggest that (1) proximate determinants of mortality (e.g., exercise) be incorporated into studies which hope to uncover causal mechanisms and make useful policy suggestions, (2) time be incorporated into studies to better understand childhood influences, the duration of events (e.g., time between marriages), and so on. Either longitudinal data or carefully crafted retrospective designs may be used to uncover the importance of lifecourse events, (3) studies refine outcome measures (e.g, use only specific causes or multiple underlying causes of death) to better understand the causal mechanisms, and (4) macro-level variables (e.g., mean income in a census tract) be included to estimate the effects of forces operating beyond the individual.