Why life expectancy improvements slowed: cohort effects and acceleration–deceleration dynamics of period life expectancy at older ages

This study examines acceleration–deceleration patterns in partial period life expectancy at ages 65–94 (LE) across nine European populations in 1980–2019 using data from the Human Mortality Database. Applying detrended Age–Period–Cohort models and a frailty framework under standard assumptions, we disentangle long-run linear mortality change, non-linear cohort patterns, and changes in cohort frailty composition. We find that the diminishing returns to proportional mortality decline, expected to result in the LE slowdown, operate only on the relative scale. In absolute terms, the long-run linear trend implies rising LE gains. Non-linear cohort effects were a significant factor in the acceleration of LE improvements from the 1990s to the late 2000s, particularly among men; removing them substantially weakens or eliminates this acceleration. The pronounced sex difference is consistent with the well-documented cohort pattern of smoking-related mortality. The post-2010 deceleration, by contrast, persists once cohort effects are removed, reflecting changes in period mortality conditions rather than cohort replacement. Finally, under a standard frailty model, rising average frailty among survivors masks part of the underlying mortality improvement, so that observed LE gains understate the improvement by one-twelfth. This attenuation widens when improvements slow and narrows when they accelerate, so selection amplifies rather than smooths the cohort-patterned deviations. Overall, acceleration and deceleration in LE gains arise from distinct mechanisms, underscoring the importance of accounting for cohort effects on period mortality when interpreting recent stagnation.