The Trimmed Bootstrap: A new empirical simulation technique for pension finance researchers

Abstract

Empirical simulation in pension finance literature has gained prominence in recent years. However, to date, the merit in applying such methodologies to pension finance research has not been robustly examined. My research examines this issue and identifies that contemporary non-parametric simulation methods used in the literature are frequently highly inaccurate. To circumvent this issue, I derive a new methodology entitled the trimmed bootstrap. This significantly more accurate technique can aid pension finance researchers in deriving sound advice for plan members. Further, through an optimisation test, this research finds evidence that the optimal asset allocation technique for plan members is dynamic over time. More specifically, this research identifies that the optimal technique de-risks towards retirement, akin to a target-date style asset allocation. However, beyond retirement date, and once withdrawals commence in the portfolio, the optimal technique returns risk to the asset allocation. This creates a ‘V’ shaped asset allocation glide path. Finally, the trimmed bootstrap methodology is robustly examined in a variety of international markets. Through a series of out-of-sample tests and a novel forward-looking measure, I show evidence that the trimmed bootstrap should become the default empirical simulation technique for pension finance researchers. The trimmed bootstrap provides researchers with a powerful tool to better predict future market states to optimise retirement outcomes: this is the main contribution of this research to the pension finance literature.