Marriage Selection and Mortality Patterns: Inferences and Fallacies

Citation: Goldman (1993) Marriage Selection and Mortality Patterns: Inferences and Fallacies. Demography (RSS)
Internet Archive Scholar (search for fulltext): Marriage Selection and Mortality Patterns: Inferences and Fallacies
Tagged: uw-madison (RSS), wisconsin (RSS), sociology (RSS), demography (RSS), prelim (RSS), qual (RSS), WisconsinDemographyPrelimAugust2009 (RSS), marriage (RSS), mortality (RSS)

Summary

Notes: Researchers have long wondered whether marital-status differences in mortality arise largely from marriage selection mechanisms or from causal processes typically known as marriage protection. Unfortunately, many investigators have relied on aggregated patterns of mortality differentials to make inferences about the relative importance of selection and causal processes. This paper treats 2 of these approaches. This first approach is based on age patterns of mortality differentials; the second, on the direction and strength of the relationship between the magnitude of the mortality differential and the relative size of the single population. About the 1st approach: The great majority of analysts focusing on marital status differentials in mortality have relied on the measure of relative risk known as the relative mortality ratio (RMR) the ratio of the death rate of a specified unmarried group (e.g., singles) to the death rate of the married group. In most populations, age schedules of relative mortality among singles are characterized by rising values from the twenties to mid-30s or 40s, followed by declining ratios through the oldest age groups. It has been argued that this pattern supports the idea of marriage selection in that if marriage operates its selection of the healthier lives, we should expect increasing levels of excess mortality of singles between ages 25 and 40 a growth which should then level off and stop altogether when nuptiality falls to very low levels. In other words, once marriages cease, the fraction of frail persons in both the single and ever-married populations diminishes progressively. About the 2nd approach: The idea behind this approach is that populations in which the great majority of persons marry should be characterized by greater selectivity effects among those who remain single than populations in which substantial proportions never marry. Supposing that the selective efficiency of marriage acts with the same strength in different populations, it follows that the lower the proportion remaining single, the higher must be the frequency among the unmarried of the less healthy and the impaired, and the higher should be the excess mortality of the singles. In other words, populations in which relatively few individuals remain single are typically those in which single persons experience high excess mortality. In this paper, a simple mathematical simulation model is used to demonstrate that many inferences derived from observed patterns are simply not justified. Marriage selection mechanisms, operating in the absence of any marriage protection factors, can produce unanticipated patterns in the RMR with respect to both age and the relative size of the single group. Relating to the 1st approach, mathematical analysis reveals that the age at which the hypothesized convergence of mortality rates begins depends on the relative rates of change of the 2 death rates and is not predictable: convergence need not begin at the end of the marriage span or even within the human life span. Relating to the 2nd approach, Goldman finds in her simulation that the relationship between the magnitude of the excess mortality among single and the relative size of this group depends critically on how the marriage rates for the healthy and the frail subgroups vary across populations. These findings highlight the importance of prospective data for assessing the relative importance of selection and causal factors in accounting for the excess of mortality of the unmarried.