Cohort Patterns Discussion
6 Period Mortality Rates
We now turn to estimates of period age-specific mortality rates. The figures presented follow the same procedure as the cohort figures: regional comparisons are made with the white results split into three sets. Rather than estimate the mortality experience of three cohorts, we estimate the mortality rates for the periods 1980-84, 1985-1989, and 1990-1993. The only limiting factor is the youngest age at which we can estimate mortality corresponds to the youngest members of the cohort.

The white male period mortality rates for the three periods are shown in Figures 11-13. In all of the figures, the mortality curves are consistent and stable until the mid-90s. After age 95, there are fluctuations in the mortality rates of all of the regions, doubtless due to the small number of survivors in each region. As in the cohort estimates, Region 12 (AZ, NM, NV) has more variation, due probably to the small population in the 1930s. There are modest regional differences, with western regions having slightly lower mortality rates.

Figures 14-16 show the age specific mortality rates for white women by region for the three periods. As was the case with the men, the mortality rates are consistent. However, the instability in the mortality rates does not appear until the late 90s. As with the cohort mortality rates, Region 12 shows variation at earlier age, and a less stable curve. However, the overall pattern is clearly consistent, especially below age 95, and regional differences are small.

These period age specific-mortality rates support the conclusion in our examination of the cohort age-specific mortality rates that the Medicare EDB data produce consistent rates for white men and women. Variations occur primarily when there are relatively small numbers of individuals in each region and age. Figure 17 plots the national age-specific mortality rates for men and women, respectively. Mortality in the most recent period is slightly below the rates of the earlier periods. The mortality rates increase steadily until between ages 100 and 105, and drop at older ages. These national figures reduce the regional variation at oldest ages.

We turn now to the period mortality rates for older blacks. We plot the mortality rates for same three periods (1980-84, 1985-89 and 1990-93) that we used for the white period rates. As with the cohort rates for blacks, we confine our examination to the eight regions with substantial black population.

Period age-specific mortality rates for black men are plotted in Figure 18. No regular or smooth curves appear for any region for any of the three periods. Year-to-year fluctuations are readily apparent. These fluctuations begin during the 70s in some regions (e.g. Region 13--CA), and become pronounced in all regions by the mid-80s. This is the general pattern in all three periods.

Such year-to-year fluctuations are not evident in the period age-specific mortality rates for black women, shown in Figure 19. For the 1990-93 estimates, fluctuations begin in the 80s for Regions 13 and 6 (WV, KY, TN), but they are not pronounced. In general, the mortality curves are consistent, with little regional variation well into the 90s for the 1980s, with some more variations for the 1990-93 rates.

As with the whites, we plotted the national age-specific mortality rates for the three periods. Figure 20 shows the national period mortality for black men and women. The increase number of cases smoothes the curves noticeably, but the year-to-year variations for the men are still apparent, beginning in the late 80s. By age 95, the rates for men fluctuate significantly. As expected, the mortality rates for the black women are more stable, with some variations in the 90s, but no significant variations until after age 100.

Figure 21 compares the national period age-specific mortality rates of white and black men and women. As was the case with the cohort comparison, there is a crossover of mortality. Both male and female black mortality rates drop below the white rates at age 86, and they do this in each period. This was the same age for the crossover in the cohort rates. Further, in the mid-90s, mortality rates of black men drop below the mortality rates of white women. Recall that Kestenbaum [9] argued that the age data for blacks in selected years of the Medicare EDB appeared to be accurate, and that he thought that the mortality crossover was real, not an artifact of age misreporting. The mortality patterns of these older black women are consistent in both the cohort and period comparison. This is certainly not a conclusive test of the crossover, but the pattern appears to be credible. However, the regional and national inconsistencies of both cohort and period mortality rates of black men, and the decline below the rates for white women raise questions about these data.

The final figure compares the national period rates of men and women with the period mortality rates of Puerto Rican men and women. We compare the male and female period age-specific mortality rates of Puerto Rico and the white population in the U.S. Figure 22 covers the three periods, 1980-84, 1985-89 and 1990-93, respectively. As was the case with the cohort rate comparisons, the mortality rate for Puerto Rican men is lower than the rate for white men on the mainland. This pattern occurs in all three period, with the mortality rate of Puerto Rican men approaching and appearing to drop below the mortality rates of white women in the 90s. However, this may be due to the instability in the rate estimates. There is clearly considerable year-to-year fluctuation in the rates, especially after the mid-80s. As noted earlier, this pattern has been found in other data. One complicating pattern in interpreting this difference is the regular and easy migration between Puerto Rico and the mainland. The EDB regional identifier is based on the location where the individual received a Social Security card, not where they lived.

The relationship of mortality rates between the Puerto Rican women and the white women is also similar to the pattern found in the cohort estimates. For 1980-84, the mortality rate of Puerto Rican women appears to fall below white women, before beginning to fluctuate in the 90s. For the other two periods, the rates are very similar.

Cohort Patterns Discussion

logo70.gif (2450 bytes)

Evaluation of U.S. Mortality Patterns at Old Ages
Using the Medicare Enrollment Data Base.

Allan M. Parnell and Cynthia R. Owens
1999 - 2000 Max-Planck-Gesellschaft ISSN 1435-9871