A Bayesian perspective on commodity style integration

Abstract

Commodity style integration is appealing because by forming a unique long-short portfolio with exposure to K mildly correlated factors, a larger and more stable risk premium can be extracted than with any of the standalone styles. A key decision that a commodity style-integration investor faces at each rebalancing time is the relative weighting of the factors. We propose a Bayesian optimized style-integration (BOI) strategy with excellent out-of-sample performance. Focusing on the problem of a commodity investor that seeks exposure to the carry, hedging pressure, momentum, skewness, and basis-momentum factors, the evidence suggests that the BOI portfolio achieves better Sharpe ratios and certainty equivalent returns, among other performance metrics, than the $1/K$ style-weighted integrated portfolio, and a battery of sophisticated optimized integrations. The findings survive the consideration of longer estimation windows, various commodity score schemes, and alternative Bayesian priors.