Forecasting Cohort Mortality: Lee–Carter Methods and CCP-Splines

Accurate mortality forecasts are central to policy, insurance, and demographic research. Yet most existing approaches rely on age–period models, limiting their ability to capture the real experiences of birth cohorts. We address this gap by developing and evaluating new models for cohort mortality forecasting. Specifically, we extend the Lee–Carter framework with two cohort-specific variants and introduce Cohort Constrained P-splines (CCP-splines), a flexible method that embeds demographic constraints into a smoothing framework. Out-of-sample validation demonstrates that CCP-splines outperform not only the cohort LC variants but also the conventional diagonal Lexis approach, which derives cohort patterns from classic Lee–Carter age-period forecasts. Applications to eight populations confirm the advantages of CCP-splines, showing improved fit, reduced bias, and more reliable uncertainty estimates. By combining statistical rigour with demographic knowledge, this study provides a practical and flexible framework that positions CCP-splines as a new benchmark for cohort mortality forecasting.