Return Predictability and Optimal Portfolio Choice: Evidence from Commodity and Global Futures Markets

Abstract

Time variation in expected returns is understood to be a common feature across aggregate asset classes as diverse as equities, currencies and bonds. Less is known on aggregate commodity return predictability. Findings of this thesis provide evidence of time variation of commodity returns and assesses its economic value within portfolio allocation strategies. Further findings demonstrate the effect aggregate hedge fund capital has on the profitability of carry and momentum strategies across global futures markets. Besides the financial implications, the findings of this thesis also affect policies with respect to international macroeconomics, poverty alleviation, energy and climate policy, and commodity market regulation.

First, I study variation of expected commodity return portfolios. I run regressions of 1 to 9 months portfolio holding returns on lagged average futures discounts (AFD). I find that the AFD predicts commodity portfolio returns with 9 months $R^2$ values of 10 percent. Most predictable variation is a result of spot premia variation, while term premia are only significantly time-varying on short term horizons. Variation in the AFD is procyclically related to macroeconomic conditions. The procyclical relation leads to strong return predictability: the AFD and US industrial production growth rates forecast up to 16 percent of the commodity portfolio holding return variation at the 9 months horizon.

Second, I study the statistical and economic value of macroeconomic, financial and commodity market specific variables in predicting commodity returns. I estimate the models within a data-rich Bayesian model averaging (BMA) framework. I find that commodity portfolio returns and volatility are predictable across commodity sectors. Posterior model probabilities reveal that most of the predictable variation in commodity returns is due to macroeconomic variables of industrial production growth and the variation of the aggregate commodity basis. Portfolio volatility is related to lagged dividend yield, default spread, and inflation growth. I further find that an investor will pay a high performance fee to switch from a dynamic portfolio strategy based on a simple autoregressive benchmark models to a BMA model. In contrast a conditional volatility strategy does not generate significant economic gains.

Third, in collaboration with James B. Grant, we provide evidence that hedge funds capital is a key determinant for the profitability of carry and momentum strategies in futures markets across asset classes. We parameterize carry and momentum portfolios from the perspective of a utility maximizing risk averse investor. We find that the returns to optimal carry and momentum strategies yield high Sharpe ratios (above 1.2), which are not a compensation for traditional risk exposure or time-varying risk due to macroeconomic cycles or funding liquidity, however they are related to pro-cyclical hedge fund capital flows. Larger capital flows lead to higher carry and momentum returns, implying that expected returns decrease with the total amount of assets under management by hedge funds. We argue that these findings are consistent with the notion of limits to arbitrage.