Mortality Crossovers: Reality of Bad Data?

Citation: Coale, Ansley J., Kisker (1986) Mortality Crossovers: Reality of Bad Data?. Population Studies (RSS)
Internet Archive Scholar (search for fulltext): Mortality Crossovers: Reality of Bad Data?
Tagged: uw-madison (RSS), wisconsin (RSS), sociology (RSS), demography (RSS), prelim (RSS), qual (RSS), WisconsinDemographyPrelimAugust2009 (RSS)

Summary

In populations for which data on deaths and on the living are little distorted by misreported ages, there is a close positive relation between mortality at younger and older ages. For example, there are a number of studies that reveal a positive association between mortality reduction early in life and a continued advantage in the mortality of a cohort over its lifetime, including Kermack et al. (1934) and Preston and van de Walle (1978). When misreporting of age is evident, low mortality at advanced ages accompanies high mortality at lower ages. This relation, which produces the so-called mortality crossover, has been attributed to heterogeneity in which the frailer members of the population at the younger ages when death rates are high, leaving only the very robust, who they enjoy low death rates when very old. The shift from relatively high to relatively low rates occurs above age 55, yet the recorded proportion surviving in the same cohort from age 55 to age 80 is higher in the population that should be eliminating the frail over this range as compared to the life expectancy above age 80/85. Positive rather than negative association between early and late mortality in the same cohorts is the rule, when death rates are reliably calculated.

Low reported mortality rates at advanced ages are often produced by age overstatement. Overstatement as age progresses is almost universal. Evidence of age overstatement among the elderly comes from the clearly overstated fraction of the population at extreme ages, the unbelievable life table estimates that come from subsequent census counts, and prevalence of age-heaping. Even in the censuses of France, where age heaping is negligible, the number aged 100 or over is seriously exaggerated. Overstatement of age among the elderly in the population censuses causes understatement of death rates. When the death rates reported at advanced ages are far below those in Japan, Sweden, Norway, and the Netherlands, the population in question almost certainly contains more persons reported as old than are genuinely present, rather than exceptionally robust survivors of a hazardous early life