Declining improvement in life expectancy over the generations

The question of life expectancy gains is a controversial one for scientists. In a study published in PNAS (Proceedings of the National Academy of Sciences of the United States of America), Carlo Giovanni Camarda (INED), José Andrade (MPIDR) and Héctor Pifarré i Arolas (University of Wisconsin) estimated life expectancy for the generations born between 1939 to 2000 using several models. Regardless of method used, the authors’ conclusion is the same: life expectancy progress has slowed considerably.

Many studies on advances in life expectancy focus on a cross-cutting indicator estimated for a given year, not for a given generation or cohort. While calculations of a generation’s or cohort’s life expectancy are much more precise and accurate indicators, they can only be done after the cohort in question has died out, meaning that over the generations, estimates are based on increasingly less precise and accurate projections. For example, in 2025, the 1940 birth cohort has already lived its first 85 years, and we need only establish projections for the 20 remaining years (assuming a maximum age of 105), whereas the 2000 birth cohort has only lived 25 years and there are still 80 years of projections to do. 

The life expectancy of today’s oldest generation was observed to rise sharply in the first half of the 20th century, from 62 years for people born in 1900 to 80 for those born in 1938. Are the following generations going to experience a similar gain? Is their life expectancy going to exceed the symbolic threshold of 100? A recent study in PNAS* suggests that none of these generations will attain life expectancy of 100 years. Even if adult survival increased twice as fast as the authors of this article have predicted, cohort life expectancy** would not reiterate the rapid rises observed for cohorts born early in the 20th century. 

The authors’ study draws on data from 23 high-income (or low-mortality) countries. Through its analysis of cohort-based data and the priority given to data-based statistical methods rather than theoretical reasoning contributes, the study is a solid contribution to debate on longevity stagnation. If current generations followed the same trajectory as that observed in the early 20th century, a person born in 1980 could live to be a hundred. The authors checked whether the pace of life expectancy gains was slowing for living cohorts. For this they used data from the Human Mortality Database (HMD), while six statistical mortality forecasting methods, including the United Nations approach (World Population Prospects) [1], were used to predict future developments. 

“All the methods show that life expectancy for people born between 1939 and 2000 is increasing more slowly than in the past. Depending on method used, the slowdown varies between 37% and 52%.” The authors predict that people born in 1980 will not, on average, live to be a hundred, and that none of the groups studied will reach that threshold. The deceleration is explained primarily by the fact that the major earlier gains in longevity were due to spectacular improvements in survival early in life. Today, mortality at young ages is already very low—meaning that there is little additional gain to be made on that score. And the pace of improvement among older people is not enough to offset the fall in the speed of life expectancy improvement. 

From 1900 to 1938, life expectancy rose around five-and-a-half months with each new generation. For 1939 to 2000 birth cohorts, the gain is no more than two-and-a-half to three-and-a-half months, depending on method used. 

According to the authors, even if adult and older-generation survival were to improve twice as fast as predicted, life expectancy gains would remain below those observed in the early 20th century. 

Mortality forecasts are never certainties. Expected future developments can be thrown off course by pandemics, revolutionary new medical treatments, and major societal changes. Moreover, the above estimates characterize populations, not individuals, and should be thought of as scientifically informed projections. 

Changes in life expectancy affect social cohesion and people’s life plans. They oblige states to adapt their health and retirement systems and social policies. They influence personal decisions on savings, retirement, long-term planning. Given that life expectancy gains are slowing, both governments and individuals will have to adjust their expectations. 

[1] Two major approaches have enabled scientists to construct cohort mortality profiles: methods based on periods, such as Lee-Carter, Smooth Constrained Mortality, Compositional Data Analysis, United Nations World Population Prospects (2024), and methods based on cohorts: Linear Lee-Carter and Cohort Segmented Transformation Age-at-Death Distributions

* J. Andrade, C. G. Camarda, and H. Pifarré i Arolas, “Cohort mortality forecasts indicate signs of deceleration in life expectancy gains”, Proceedings of the National Academy of Sciences of the United States of America (35) e2519179122, at https://doi.org/10.1073/pnas.2519179122 (2025)

** Life expectancy by cohort tracks an entire generation (for example, all persons born in 1900) and measures how many years people in this cohort actually lived, on average, once all cohort individuals have died (or on the basis of forecasts if the cohort has not died out). It takes account of medical advances, pandemics, wars, etc. that individuals in a given generation experience. 

Observed and forecasted life expectancies by cohort. The dotted vertical black line marks the 1938 birth cohort; the red curve represents optimal gains (highest life expectancy found for each cohort); the blue curve, the median; the grey line, values specific to each country; the pink curve, linear extrapolation from optimal gains between 1900 and 1938. © MPIDR