What does tempo-adjusted period life expectancy measure? Taking a cohort perspective, I show that under conditions of constant linear mortality shifts the tempo-adjusted period indicator translates exactly to the cohort born e * 0 (t) years earlier. I discuss the implications of cohort translation for the interpretation and application of tempo-adjusted period life expectancy. 1 Office of Population Research, Princeton University, E-mail: josh@princeton.edu http://www.demographic-research.org 71Goldstein: Found in translation?
IntroductionLife expectancy at birth is at root a cohort concept. It tells us how long, on average, the members of a cohort survive. Actual life expectancy can only be known fully for cohorts born long ago. To summarize recent mortality conditions and period-to-period variation, the hypothetical concept of period life expectancy is conventionally used. But even period life expectancy refers conceptually to a cohort -the hypothetical one that lives according to the rates observed in a single period. When mortality conditions are improving, period life expectancy is less than that of the cohort born in the period. This is because the hypothetical cohort following the period life table is deprived of future mortality improvement.I recite this basic property of period life expectancy because the "tempo adjusted" method of measuring period life expectancy -as developed by Bongaarts & Feeney (2002) -arrives at exactly the opposite conclusion. According to Bongaarts and Feeney, period life expectancy overstates longevity when mortality conditions are improving. They conclude: "Our main finding is that the conventional calculation of period life expectancy at birth gives a misleading indication of how long we live. We are not living as long as we thought we were." (p. 25).To be fair, Bongaarts and Feeney, except at a few points, are not talking about cohorts. Instead, they intend e * 0 as a period measure that tries to improve upon period life expectancy. What such an improved period indicator actually measures is the subject of much debate as is clear from many of the papers in this volume. The approach taken here is to recast tempo-adjustment in cohort terms. Doing this enables us to resolve the counter-intuitive direction of tempo-adjustment by showing which cohort B&F are referring to when they say "we."The approach is similar to that of Goldstein & Wachter (2004), which showed -using a different model of temporal mortality change -the correspondence between period life expectancy e 0 and the life expectancy of particular cohort. Here, I look at which cohort has the life expectancy equal to current tempo-adjusted life expectancy e * 0 . I find that under linearly shifting mortality, defined below, tempo-adjusted life expectancy for year t translates to the cohort dying in year t: this is the cohort born e * 0 years earlier.
2An additional assumption is needed for this simple cohort translation of e * 0 to hold exactly. B&F's tempo-adjustment assumes that deaths are postponed uniformly...