Skip to main content
9th World Conference on Information Systems and Technologies

Full Program »

Forecasting The Retirement Age: A Bayesian Model Ensemble Approach

In recent decades, most countries have responded to continuous longevity improvements and population ageing with pension reforms. Increasing early and normal retirement ages in an automatic or mechanical way as life expec-tancy at old age progresses has been one of the most common policy re-sponses of public and private pension schemes. This paper provides compa-rable cross-country forecasts of the retirement age for public pension schemes for selected countries that introduced automatic indexation of pen-sion ages to life expectancy pursuing alternative retirement age policies and goals. We use a Bayesian Model Ensemble of heterogeneous parametric models, principal component methods, and smoothing approaches involving both the selection of the model confidence set and the determination of op-timal weights based on model’s forecasting accuracy. Model-averaged Bayesian credible prediction intervals are derived accounting for both sto-chastic process, model, and parameter risks. Our results show that statutory retirement ages are forecasted to increase substantially in the next decades, particularly in countries that have opted to target a constant period in retire-ment. The use of cohort and not period life expectancy measures in the pen-sion age indexation formula would raise retirement ages even further. These results have important micro and macroeconomic implications for the design of pension schemes and individual lifecycle planning.

Jorge Bravo
Universidade Nova de Lisboa - NOVA IMS & Université Paris-Dauphine PSL & MagIC & CEFAGE-UE

Mercedes Ayuso
University of Barcelona, Department of Econometrics, Statistics and Applied Economy, Riskcenter-UB


Powered by OpenConf®
Copyright ©2002-2020 Zakon Group LLC