Abstract
In this article, we revisit recent evidence indicating that the choice of performance measure appears to be irrelevant for the ranking of investment alternatives in the commodity market. Extending the previous literature in several important ways, we provide the following insights into the rankings produced by the 13 most popular performance measures for 24 commodities. First, ranking differences are somewhat larger in the spot market than in the futures market. Second, when we use daily instead of monthly data, performance measures that model reward based on average returns still produce similar performance rankings. However, when data of higher frequency is used for performance measures modeling reward based on higher partial moments, performance rankings differ crucially from those produced by measures focusing on average returns. Finally, the degree of ranking (dis)similarity appears to vary over time. Empirically, then, the choice of performance measure can matter. Nevertheless, our findings do not invalidate recent theoretical results on ranking similarity, because population rankings may not be identical with sample rankings, which are subject to estimation error.