Extrapolation of Mortality Trends

Demographers claim some expertise in predicting future mortality levels, and their method of choice is usually a mere extrapolation of past trends. Biologists and others sometimes criticize this approach because it seems to ignore underlying mechanisms. However, this critique is valid only insofar as such mechanisms are understood with sufficient precision to offer a legitimate alternative method of prediction. Although many components of human aging and mortality have been well described, our understanding of the complex interactions of social and biological factors that determine mortality levels is still imprecise. Furthermore, even if we understood these interactions and wanted to predict future mortality on the basis of a theoretical model, we would still need to anticipate trends in each of its components.

The extrapolative approach to prediction is particularly compelling in the case of human mortality:

First, mortality decline is driven by a widespread, perhaps universal, desire for a longer, healthier life.

Second, historical evidence demonstrates that mortality has been falling steadily, and life span has been increasing, for more than 100 years in economically advanced societies.

Third, these gains in longevity are the result of a complex array of changes (improved standards of living, public health, personal hygiene, and medical care), with different factors playing major or minor roles in different time periods.

Fourth, much of this decline can be attributed to the directed actions of individuals and institutions, whose conscious efforts to improve health and reduce mortality will continue in the future.

Even accepting this argument, there is still a question of what to extrapolate. Demographers tend to view death rates as the fundamental unit of analysis in the study of mortality patterns, because these rates are estimates of the underlying "force of mortality" or the risk of death at any moment in a person's lifetime. These risks change over age and time and vary across social groupings (by sex, race, education, income, etc.). Life expectancy and the expected maximum age at death (for a cohort of a given size) can be expressed as a mathematical function of death rates by age. Thus, the usual strategy is to extrapolate age-specific death rates into the future and then to use the results of such an extrapolation to compute forecasts of life expectancy or other parameters of interest.

Predictions of future life expectancy by such methods yield values that are not too different from what is observed today. For example, forecasts made in the late 1990s by the U.S. Social Security Administration put life expectancy in 2050 at 77.5 years for men and 82.9 years for women, compared to 72.6 and 79.0 years in 1995 (48). These forecasts were not true extrapolations, however, because they assumed a slowdown in age-specific rates of mortality decline in the future. Another study using similar data but based on a purely extrapolative technique yielded more optimistic results—a life expectancy at birth in 2050 of around 84 years for both sexes combined (49). Plausible forecasts for Japan yielded slightly higher values of life expectancy at birth: 81.3 years for men and 88.7 years for women in 2050, compared to 76.4 and 82.9 years in 1995 (50).

The life expectancy forecasts of Lee and Tuljapurkar (49) are reproduced here as Figure 6. These projections are based on a clever extrapolative technique pioneered by Lee and Carter (51), which has been very influential in the world of mortality forecasting over the last 15 years. The method yields a range of estimates for each calendar year during the forecast period (in this case, from 1997 to 2096). The inherent uncertainty of future trends is represented in the graph by plotting not only the median forecast, which may be considered the "best estimate," but also by showing two extreme forecasts. The

Year

FIGURE 6 Life expectancy at birth, United States, 1900-1996 (actual) and 1997-2096 (forecast). Source. From Ref. 49.

median forecast lies at the 50th percentile of the full distribution: half of the estimates lie below or above this value. Figure 6 also presents the fifth and 95th percentiles, thus showing relatively extreme trajectories of future life expectancy.

It is important to remember that these projections are mere extrapolations of the historical experience of one country during a particular time period (the United States from 1900 to 1996). The implicit assumption is that future trends will resemble past ones. This assumption is plausible given the fairly steady pace of mortality decline over the past century. Of course, extrapolation is not without its flaws. It could not, for example, have anticipated the rise of mortality in the former Soviet Union after 1990, the emergence of AIDS in certain populations during the 1980s, or the divergence of mortality trends between Eastern and Western Europe after 1960. However, such observations are less an indictment of extrapolation as a method of mortality forecasting than a demonstration that the greatest uncertainties affecting future mortality trends derive from social and political rather than technological factors.

An important issue for consideration in forecasting mortality is the time frame—both the time frame of the data that form the input to an extrapolation and the time horizon of the projection itself. Although short-term fluctuations have been common, long-term mortality trends in industrialized countries have been remarkably stable. When mortality decline slowed temporarily during the 1950s and 1960s (in the United States and other developed countries), predictions that the rise in human life expectancy had come to an end were commonplace. Similarly, the unusually rapid decline of mortality rates after 1968 fostered expectations of unprecedented gains in longevity that would continue for decades. With the benefit of hindsight, these were both overreactions to rather short-lived episodes in the history of mortality change.

Another common error results from an undue emphasis on trends in life expectancy. Although it continues to increase, the pace of change in life expectancy at birth has been slower in recent decades than during the first half of the twentieth century (Fig.1). As noted earlier, a slower rise in life expectancy at birth was inevitable once juvenile mortality was reduced to historically low levels. However, it does not follow from this observation that gains against mortality in the future will be slower than in the past. Although the increase in life expectancy has slowed down, the decline in death rates at older ages (where most deaths now occur) has quickened (35). An extrapolation of current trends in death rates suggests that life expectancy will continue to increase, though not as quickly as during the first half of the twentieth century. This slow but stable increase in average life span will be driven by the accelerating pace of mortality decline at older ages.

Better Mind Better Life

Better Mind Better Life

Get All The Support And Guidance You Need To Be A Success At A Better Life. This Book Is One Of The Most Valuable Resources In The World When It Comes To Better Living with Enhanced Mental Health.

Get My Free Ebook


Post a comment