Using Bootstrap Uncertainty to Make Portfolio Choices More Robust
Summary
This article explains how bootstrap resampling can represent uncertainty in estimates used for portfolio decisions. Instead of relying on one calculated mean or on returns generated from an assumed distribution, it repeatedly samples observed returns with replacement and recalculates the quantity of interest. The resulting range of estimates shows how sensitive a decision may be to variations in the available history. The author notes that ordinary daily resampling breaks serial dependence, so strategies whose results depend on autocorrelation may require resampling in longer blocks.
The method is applied to portfolio allocation and leverage choices for bonds and equities, including Kelly-style growth optimisation and mean-variance analysis. Rather than selecting the single best point estimate, the article examines how performance changes across bootstrapped samples and parameter combinations, then recommends choosing a robust point within a broad high-performing region. The example’s proposed allocations depend on assumed forward returns, volatility, and correlation, and its leverage analysis relies on assumptions that may fail in markets. The article presents a decision framework, not a general-purpose allocation prescription.
Key ideas
- Bootstrapping produces a distribution of estimates by repeatedly resampling historical observations with replacement.
- Daily resampling can erase autocorrelation, so dependent strategies may require resampling blocks of time.
- Portfolio choices should account for uncertainty in estimated returns and other model inputs.
- A broad region of acceptable allocation and leverage choices can be more robust than a single estimated optimum.
- The example’s numerical outcomes depend on its return, volatility, and correlation assumptions.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.