The Contribution of Forecast Uncertainty to Lifespan Uncertainty

Ricarda Duerst , University of Rostock
Jonas Schöley, Max Planck Institute for Demographic Research

Period lifespan variability is sometimes interpreted as an indicator of lifespan uncertainty for individuals. Is this an admissible interpretation for living members of a cohort, given that future death rates are uncertain? To answer this, we express individual lifespan uncertainty as a function of variation in ages at death in a cohort lifetable and uncertainty about the future death rates. We then quantify the importance of each component to the total individual lifespan uncertainty of a birth cohort. A benign contribution of forecast uncertainty would facilitate the interpretation of lifespan variability as individual lifespan uncertainty. Using HMD data for seven European countries, we decompose the total variance in remaining lifespan at birth, age 40, and age 80 of an individual cohort member into cohort lifetable variance and forecast variance for cohorts born from 1871 through 2019. We do so under Lee-Carter cohort forecast variance and under an empirical forecast variance measure which allows for mortality shocks and trend changes. Even when accounting for the possibility of future mortality shocks and trend changes, the forecast uncertainty component tends to contribute less than 10% to the total lifespan uncertainty of an individual. Conditioning on the absence of future mortality shocks and continuation of linear trends, the forecast uncertainty contributes around 1% to total lifespan uncertainty. We give credibility to the interpretation of cohort lifetable lifespan variance as individual lifespan uncertainty by showing that the lifespan uncertainty of a cohort is mainly determined by expected lifetable variance with smaller contributions from forecast variance.

See paper

 Presented in Session 121. Methodological Innovations in Mortality Studies