TCS Daily


The Long and Short of Lifespan

By Arnold Kling - January 18, 2005 12:00 AM

"Unlike the cohort life table, the period life table does not represent the mortality experience of an actual birth cohort. Rather, the period life table presents what would happen to a hypothetical (or synthetic) cohort if it experienced throughout its entire life the mortality conditions of a particular period in time. Thus, for example, a period life table for 2002 assumes a hypothetical cohort subject throughout its lifetime to the age-specific death rates prevailing for the actual population in 2002." --Centers for Disease Control, United States Life Tables, 2002

If reading that description of lifespan data makes your eyes glaze over, I don't blame you. However, some important policy debates depend on lifespan estimates, and I do not think that people understand the tenuous, unreliable nature of those estimates. The purpose of this essay is to draw attention to this issue.

What's at Stake

Lifespan estimates come up in at least two policy arenas: comparing health care quality across countries; and forecasting the financial condition of Social Security. If, as I shall argue, we are under-estimating longevity, then we are under-estimating the quality of American health care and over-estimating the financial condition of Social Security.

One way to compare health care quality across countries would be to look at people with identical medical conditions and compare outcomes. Using that method, the United States often comes out ahead. For example, in a recent issue (not yet on line) of Cato's Letter, John Goodman writes,

"Among women who are diagnosed with breast cancer, only one fifth die in the United States, compared to one third in France and Germany, and almost half in the United Kingdom and New Zealand. Among men who are diagnosed with prostate cancer, fewer than one fifth die in the United States, compared to one fourth in Canada, almost half in France, and more than half in the United Kingdom."

These examples of better health outcomes do not seem to show up in the aggregate statistics on longevity. That is, when we compare overall longevity statistics in different countries, the results show little, if any, superiority in U.S. health care to that in other affluent countries. Given that those other countries spend less per capita on health care, this raises questions about the "bang for the buck" in U.S. health care spending.

The other issue affected by longevity is the viability of Social Security. The current debate over cutting benefits has nothing to do with benefits in the near term. What is at issue is the level of promises that we make to people in the future (the Democratic National Committee attack that I cited here included complaints about cuts for people who retire in 2075, meaning people who have not yet been born).

It would make sense to limit our promises to those that can be kept under the existing tax structure. However, the promises that we can make depend crucially on longevity. If, as I shall argue, our baseline projections for longevity under-estimate the lifespans of future retirees, then to maintain our current level of promises would be irresponsible, because with greater longevity the ratio of retirees to workers in the future will surge. It would be more prudent to gradually scale back promises, by raising the retirement age or shifting from wage indexing to price indexing -- see the Social Security Policy Primer in my book,.

A Simple Lifespan Table

Suppose we were to examine the mortality history of everyone in the United States born in 1910, following them up through the year 2000. The result would be a table with 90 columns of data, showing the percent of the 1910 cohort who died in each subsequent year. That is too much information to put into this sort of an essay.

To save space, let us pretend that people live in 20-year increments. They may die at birth, at age 20, at age 40, at age 60, at age 80, or at age 100. This simplification will allow me to present a life table that is less precise, but which still offers a reasonable portrayal of the main issue under discussion -- the calculation of life expectancy.

Based crudely on the information in the CDC table, the percentage of people born in 1910 who die at each point would be as follows:

1910

1930

1950

1970

1990

2010

15 %

7 %

16 %

30 %

30 %

2 %

That is, 15 percent of those born in 1910 died at birth (or closer to birth than to age 20), 7 percent died close to age 20, and so on. For the 2 percent that survived well past age 80, we assume that death occurs at age 100. The average is simply the number of years before death times the percent of people who lived to that age. For the cohort born in 1910, the longevity is (0 times .15) + (20 times .7) + (40 times .16) + (60 times .30) + (80 times .30) + (100 time .02) = 51.8 years.

How do we calculate the longevity of a person born in, say, 1950? The analogy would be to look at how many people born that year died in 1950, 1970, 1990, 2010, 2030 and 2050. The problem is that we only have actual data for the first three of those time periods. The reality is that we do not know the lifespan of people born in 1950, much less the lifespans of people born much more recently, about whom the promises for future Social Security benefits are such non-negotiable issues for the DNC and the AARP.

Whose Lifespan is it, Anyway?

This raises the question of how to calculate national average longevity. On the one hand, we do not have enough data to complete the lifespan picture for people born after 1910. On the other hand, we do not want to simply assume that lifespan is 51.8 years, just because that is our last good data point. We know that the lifespan has increased for later cohorts, because already their infant mortality is lower. What do we do to estimate today's lifespan?

We have to make the best use we can of the data that we have. As of 2000, our data on cohorts born since 1910 might look like this:

 

Percent Died at Age

Birth Year

birth

20

40

60

80

100

1910

15 %

7 %

16 %

30 %

30 %

2 %*

1930

6 %

7 %

14 %

31 %

?

?

1950

3 %

3 %

7 %

?

?

?

1970

2 %

2 %

?

?

?

?

1990

1 %

?

?

?

?

?

*assumption

To calculate a national longevity statistic from this table, we assume that someone has a 1 percent death rate at age zero, as was true for the 1990 population, a roughly 2 percent death rate at age 20, which is the information that we have for the 1970 population, a death rate at age 40 that is taken from the 1950 population, and so forth. This synthetic longevity calculation, which actually applies to no cohort in particular, gives an example in our data of 68.9 years.

What does the number 68.9 mean? It means that, using our 20-year death intervals, if you were born in 1990 and experienced the infant mortality of the 1990 birth cohort, the 20-year mortality of the 1970 birth cohort, and so on, you would live 68.9 years. The number applies to no one in the real world.

Application to Health Care Comparisons

The national longevity number, as calculated above, automatically excludes many improvements in health care. For example, suppose that because of the use of statin drugs to reduce cholesterol, people born in 1950 are going to have considerably lower death rates from heart attacks around age 60 than people born in 1930. The longevity calculation does not know this. All it knows is the death rate at age 60 for people born in 1930, and it applies that death rate to people born in 1950.

Thus, by its very definition, "average" longevity rules out measuring much of the increase in lifespan that recent advances in medical care have produced. For miracle drugs and other life-saving treatments, average longevity calculations will always be behind the curve in assessing their effects.

Comparing international health outcomes based on "average" longevity is a meaningless exercise. In such calculations, cohorts born before 1950 have a large influence on the estimated death rates over age 50 for people born today. In effect what you are comparing across countries is a combination of their current rates of infant mortality and the state of care of the elderly shortly after World War II. In reality, such longevity comparisons say very little about the quality of health care for today's adults.

Social Security Forecasts

Calculations about the future of Social Security are based on projections that the majority of 50-year-old Baby Boomers will not live past 85. This would be a safe prediction, provided that the Boomers experience mortality over the next 35 years that can be projected based on the mortality of people born in the 1920s and 1930s. Instead, it could turn out that the majority of those aged 50 will live past the age of 90, with even greater increases in lifespan for subsequent cohorts.

As the uncertainty about lifespan increases, the risk increases that Social Security's defined-benefit system will be unsustainable. Switching to a defined-contribution structure would greatly reduce the risk of a meltdown. Partial privatization would be a move in the direction of greater safety.

Calculating the lifespan may seem like an arcane, technical exercise. But I think that the more you understand it, the better informed you will be on the issues of health care and Social Security. Our health care advances may be doing more to improve longevity than the statistics show. In Social Security, the status quo is riskier and private accounts are less risky than most people realize.

 

Categories:
|

TCS Daily Archives