5 Cohort Patterns
|Age-specific mortality rates are plotted for the birth
cohorts 1895-1899, 1900-1904, and 1905-1909. The mortality rates begin at different ages,
since data collection did not begin until 1976, with the oldest age being that attained in
1993. Comparisons among all 14 regions are made for white males and white females.
Fourteen regions are too many to plot in a single figure, so three sets of regions are
plotted with the region with the lowest mortality rates (Minnesota, North Dakota and South
Dakota) used as the standard in all figures.
Figures 1-3 present the cohort mortality patterns for white males. The mortality rates in all of these figures appear to be consistent, with no signs of a crossover. For the oldest cohort--those born between 1895 and 1899--there is instability in the rates from the mid-90s on, especially for regions in the west. For the middle cohort--1900-1904--the mortality rates again appear to be consistent. Again, the mortality rates at older ages become unstable in the west, especially for Region 12 (Arizona, New Mexico and Nevada), and, to a lesser degree, for Regions 13 (California) and 14 (Oregon, Washington and Alaska). Recall that these regions are where SSNs were obtained, and these regions had relatively low populations in the 1930s. Mortality rates for the youngest cohort are also consistent, with less instability. There are notable regional differences, most notably the lower mortality rates in Region 8 relative to the Southern regions. Mortality rates in Region 8 are also lower than in the northeastern regions.
Figures 4-6 present the cohort mortality patterns for white females. As was the case with white males, the mortality rates in all of these figures appear to be consistent, with no signs of a crossover. There is notably less instability at the older ages among the females relative to the males, and the regional differences in mortality rates do not appear to be as large. Some variability is seen at the oldest ages in some of the western regions. In addition, for the middle cohort (1900-1904) there is a notable increase in mortality rates between ages 73 and 74 for Region 2 (NY, NJ and PA) for reasons that are not clear.
The cohort morality curves for white men and women are consistent for the three cohorts examined. Two patterns require comment. First, there is notable variation in the mortality rates at older ages in the western regions, again, probably a function of population size when Social Security cards were issued. These variations are greater for the men than for the women. Second, there are regional differences, especially for men. The South and the Northeastern regions appear to have higher mortality than Region 8 and other western regions. We estimated no formal test of difference between the mortality curves. The primary finding of this examination of mortality curves is the overall consistency and lack of crossover.
Mortality curves for black men and women are presented in Figures 7 and 8, with one figure for each cohort. We plot the mortality rates for the eight regions with substantial black populations.
There is considerably less consistency in the mortality rates for black males in all three cohorts. The inconsistency increases at oldest ages, but is apparent throughout. Mortality rates for black men who obtained their Social Security cards in California vary considerably in all three cohorts, but variations are notable for other regions as well, especially from the mid-80s and older.
Inconsistencies in cohort mortality rates are not as great for black women. The pattern for all three cohorts shows some variability among regions, with mortality rates for those obtaining their Social Security cards in California showing the greatest variability. The variability between regions may not be significant for the youngest cohort of black women, at least until the mid-80s and older.
In sum, the cohort mortality rates for blacks are inconsistent, especially for men. Regional morality rates rise and fall, even at relatively young ages. Above age 80, the mortality rates for men are highly inconsistent. This appears to be less of a problem for black women, especially for the youngest cohort. However, the black female mortality rates are not as consistent as their white peers.
Figure 9 compares national cohort morality rates for white men and women and black men and women for the three birth cohorts. The black mortality crossover occurs in every cohort, and in every race-sex-cohort group, the crossover occurs at ages 86-87. Whether the crossover is due to selection processes or to age misreporting, the consistency across cohorts in the age at which the crossover occurs is remarkable.
We also compare the mortality rates of the white populations of the U.S. with the mortality rates in Puerto Rico. Preston and Rosenwaike  report evidence of age misreporting in Puerto Rico. If this is the case, one indication would be a crossover in mortality rates similar to the pattern found for the black-white crossover. Another indication would be substantially lower mortality rates. As is shown in Figure 10, no clear crossover is found. Puerto Rican female mortality rates are very close to those of white American women. However, the mortality rates of Puerto Rican men are substantially lower than rates for white American men from the same three cohorts. This is consistent with Shai and Rosenwaikes  estimates of mortality rates at ages 65-71 for the years 1979-1981. Mortality rates at these ages for men born in Puerto Rico were substantially lower than the same age group of white men in the U.S. Mortality rates of Puerto Rican women born on the island and white women from the U.S. were essentially the same. Rosenwaike  comments on the lower old age mortality rates of all Hispanics relative to white Americans, reviewing possible measurement reasons for the differences as well as the "healthy migrant" hypothesis. He came to no firm conclusion for the differences, but suggests that they are indeed real.
|Evaluation of U.S. Mortality Patterns at Old
Using the Medicare Enrollment Data Base.
Allan M. Parnell and Cynthia R. Owens
© 1999 - 2000 Max-Planck-Gesellschaft ISSN 1435-9871