Do crude oil futures still fuel portfolio performance?

Abstract

Motivated by significant changes in the dynamics of crude oil markets, we revisit the question of whether crude oil futures are a valuable addition to typical stock and bond portfolios. To answer this question, we use a mean-variance framework with established heuristic and classic weighting schemes and, in addition, derive optimal asset weights based on forward-looking estimates of expected returns (via univariate forecast combinations and multivariate LASSO regressions) and covariances (via DCC-GARCH and GO-GARCH models). We find that crude oil futures tend to be of limited benefit for investors. In recent data, their optimal weights are very low and they typically do not enhance out-of-sample portfolio performance. These findings hold in a variety of robustness checks with respect to different datasets, estimation windows, regression specifications, GARCH models, and transaction costs.

Unprofitability of food market investments

“…Because comparable active investing is also possible in bond, nonfood commodity, and currency markets, another natural extension of our work would be to leave the stock market narrative (e.g., by using the data set of Ilmanen et al, 2021). Third, there is a growing literature popularizing forward‐looking estimation of portfolio optimization inputs (see Vinzelberg & Auer, 2021). Because food futures are more predictable than other assets (see Mehlitz & Auer, 2021), they may turn out to be more beneficial in a forward‐looking framework than in a backward‐looking one.…”

A comparison of minimum variance and maximum Sharpe ratio portfolios for mainstream investors

“…Second, in the single-index model, we use an alternative market index composed of 60% stocks, 25% bonds and 15% commodities: a heuristic allocation which can often be found in the literature (see Auer and Schuhmacher, 2013;Vinzelberg and Auer, 2021). While we…”

“…Finally, the maximum drawdown (MDD) quantifies the largest cumulative loss in the out-of-sample period, that is, the worst buy-sell decision the investor could have made (see Schuhmacher and Eling, 2011). Besides descriptively looking at risk, we follow Vinzelberg and Auer (2021) by statistically comparing the Sharpe ratios of MinVar and MaxSR portfolios via the Ledoit and Wolf (2008) bootstrap test [17]. In contrast to earlier proposals, this test can take into account non-normality of returns and, in addition, performs well in small samples.…”