African-American mortality at older ages: Results of a matching study
Preston and colleagues follow up Preston & Elo (1994) who find that death certificates for blacks (especially women) in the 1960 census underreport age. The current study matches 1980 and 1985 death certificates to 1900, 1910, and 1920 Censuses to reexamine the black mortality crossover in which it appears that blacks have higher death rates than whites at younger ages and lower death rates at older ages. The authors focus on the age at death entered on the death certificate (rather than the age implied by dates on records) b/c that is what NCHS uses for national mortality estimates. Age at death below age 90 was understated for men and women, while age at death was over-stated for men and women over 90. Discrepancies were concentrated around one year for early censuses, but were more concentrated around 2+ years when soc sec records are used for comparison. After adjustment w/age-specific growth rates, the only crossover still present was the 90+ age group. This may or may not be an error. The death certificates are ultimately shown to be the cause of the crossover error. The same pattern may be true for whites, but whites were not analyzed. Even if the pattern of misreporting is the same for whites, the age-mortality structure may not produce a crossover even if data corrections are made.