Introduction Background and Significance
2 Background and Significance
Mortality rates among the oldest-old population have declined in most of Europe and in developed countries in Asia and the Pacific since the middle of the 20th century. However, significant differences persist among these populations in oldest-old mortality rates [3; 7; 16]. Using high-quality data for 28 developed countries, Kannisto [7] estimates mortality rates between ages 80 and 99 and other age-specific morality rates, the rates of change in mortality between the late 1950s and the late 1980s, and estimates the dates at which sustained morality declines at these ages began. While the mortality rates in all of these populations are declining, the rate of decline and the point at which sustained decline began differ considerably. Kannisto's estimates of data of sustained decline in mortality for women between ages 80 and 99 in Western Europe vary from the early 1950s for France to the early 1970s for Austria and West Germany. Similar variation is found for men, but the sustained decline generally began later.

Oldest-old mortality in the United States has also declined significantly [9; 26], with the most significant declines coming during the 1970s [3]. These declines placed mortality among the oldest old in the U.S. lower than elsewhere. Manton and Vaupel [14] compare female mortality between ages 80 and 99 in the U.S., England and Wales, France, Sweden and Japan. Female mortality rates in the U.S. were lower from 1950 until the mid-1960s. However, in the mid-1960s, female mortality rates in the U.S. dropped notably below the rates in the other countries, a pattern that held until the mid-1990s.

Analyses of age-at-death data for whites at oldest ages from various data sources has shown greater consistency than is the case for blacks [25], including some Medicare data [9]. For the black data, questions persist about the accuracy and consistency of the age distribution of mortality at oldest ages. The question of the accuracy of age-at-death data for blacks has been raised primarily in the literature on the mortality crossover. Mortality is higher for blacks relative to whites at every age up to the oldest old, when mortality falls below whites. Explanations for this pattern focus on the selection processes at earlier ages that affect population heterogeneity, leaving a healthier population at oldest ages [12; 13; 18]. Kestenbaum [10] reports evidence that the cross-over is real before age 90 using the Medicare EDB. Others argue that this crossover is an artifact of poor data that comes from systematic over-reporting of age at oldest ages [1; 2].

Elo and Preston [4] review evidence of age reporting problems for blacks in various data sources, evidence that repeatedly shows significant inconsistency in age reporting. They attribute the inconsistency to the lack of birth registration in large parts of the South during the period when the older blacks we examine were born, reducing accuracy. Others suggest that lower levels of literacy among older blacks reduce their accurate knowledge of their age. However, these factors would not necessarily result in systematic overstatement of age.

Age overstatement may be associated with the perceived higher social status that comes with extreme age. Myers [17] documents a case of a supposed centenarian who had greatly overstated his age in Pennsylvania, and there are other well-known cases of populations of supposed centenarians that prove to be false [6], and this may be the case with blacks. Indeed, Hendricks and Hendricks [5] maintain that systematic age overstatement among blacks is a result of the "greater prestige" of the very old in black families.

Age overstatement may occur when there is a direct benefit to being old, for example at the beginning of Social Security. The original process of attaining a Social Security card was not particularly stringent, only requiring documentary proof-of-age since the end of 1974, while entitlement to Medicare, established in 1965, required age verification [10]. Indeed, Elo and Preston [4] report concerns over the validity of the age data for both Medicare and Social Security for those who did not have birth certificates. (Note 1)

Preston et al. [21] raised the issue of systematic understatement of age by blacks, especially women, citing a small study by Peterson [19] on active-church women who suggests that understatement is due to vanity on the part of these older women. Preston et al. [21] linked death certificate data with Census records from 1900, 1910 and 1920. They conclude that the result of systematic understatement of age at death data for Blacks (and especially for women) results in an excess of reported deaths above age 95. They correct for the resulting shift of the age distributions that results, reporting no evidence of a crossover between ages 85 and 95, but uncertainty after that age. They note that their corrected mortality rates are higher than Kestenbaum’s [9] estimates using Social Security data.

Introduction Background and Significance

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