Investing in Commodity Indices: Risk Factors and Trading Strategies

This paper researches risk factors and trading strategies using 27 Dow Jones-UBS commodity indices. A stepwise regression procedure with several criteria is established that leads to the final time-series risk model choice. It consists of sector, liquidity, materials sector equity, momentum and non-commercial hedging pressure factors. The model has an average out-of-sample R2 of 0.57. In the strategy part of the research, two types of framework for every strategy are considered. In one the strategy goes long/short in the sectors, in the other it goes long/short in individual commodi- ties within each sector. The Henriksson-Merton non-parametric test is used to evaluate strategies. The strategy based on the roll yield deviation from its' 5-year average is the best across sectors with the IR of 0.54 and the 12-month momentum with open interest growth as confirmation is chosen within sectors with the IR of 0.28. The strate- gies have no structural breaks in the sample period and are robust to the choice of parameters. As the two strategies have low correlation because of different construction framework, the combination of them can reduce risk.