This dissertation is a collection of three research essays that study the predictability of aggregate returns on alternative asset classes (commodity and currency futures), together with more traditional asset classes (stocks and bonds), and exploit such predictability via the construction of strategic and tactical portfolio allocation strategies.
The first essay examines the diversification benefits of investing in aggregate commodity and currency futures. We consider a risk-averse investor with mean-variance preferences who exploits the possibility of predictable time variation in asset return means, variances, and covariances. To this end, return predictability is modelled jointly in a first order vector autoregression framework with time-varying variances and covariances following a dynamic conditional correlation specification. We implement unconditional and conditional efficient portfolio strategies designed to exploit predictability, together with more traditional and/or ad hoc ones yet hitherto relatively unexplored in this context (including the equally weighted, fixed weight, volatility timing, and reward-to-risk timing strategies). We find that, for all portfolio strategies, commodities and currencies do not improve the investment opportunity set of the investor with an existing portfolio of stocks, bonds and T-bills, and an investment horizon of one-month. Our findings, which reverse the conclusions of previous studies that focus on static portfolio strategies, are robust across several performance metrics.
The second essay provides a comprehensive study of the statistical and economic significance of aggregate commodity return predictability using a variety of commodity, financial, corporate and treasury bond markets, and macroeconomic variables. Using forecast combination methods designed to deal with model uncertainty and parameter instability, we find that aggregate commodity returns are predictable both in and out-of-sample and outperforms the historical average benchmark in terms of standard statistical evaluation metrics, including the Campbell and Thompson (2008) out-of-sample R^2 and Clark and West (2007) MSFE-adjusted statistic. We also show in a dynamic asset allocation setting that commodity return predictability is economically significant as measured by average realized utility gains and Sharpe ratios for an investor with mean-variance preferences and relative risk aversion of three. This is the case whether trading commodity futures is considered as a stand-alone investment strategy or as a diversification strategy. The results are also robust to transaction costs of 20 basis points per dollar of trading. Further analysis to shed more light on the economic drivers of predictability shows that the ability of commodities to generate sizeable statistical and economic significance has links to the state of the business cycle. Specifically, we find that, consistent with the extensive literature on equity and bond return predictability, and more recently in the commodity markets literature, commodity return predictability and portfolio performance is stronger during recessions relative to expansions. Fernandez-Perez, Fuertes, and Miffre (2017), for example, show that commodity portfolios that capture backwardation and contango signals, such as the term structure and hedging pressure, can play a leading role as an indicator of future economic activity.
The third essay studies the cross-predictability of returns using predictor variables specific to the commodity, stock and bond markets, and market integration thereof. If markets are integrated, then the same predictor variables should forecast the returns of the various asset classes. We present evidence supporting the hypothesis that the commodity market is only partially integrated with the stock and bond markets. Extensive in-sample and out-of-sample tests of predictability show that the information content of commodity-specific predictors have statistically significant predictive power for stock and bond excess returns at both short- and long-horizons. The results continue to hold even after controlling for a comprehensive pool of traditional stock and bond predictors. Portfolio analysis using mean-variance spanning tests confirms the robustness of the in-sample and out-of-sample predictive regression tests. Specifically, we also find that the returns on a commodity predictability-based trading strategy constructed using stock and bond predictors does not improve the investment opportunity set of a mean-variance investor with an existing portfolio composed of commodities and a commodity predictability-trading strategy constructed using commodity predictors. As a commodity, stock or bond stand-alone investment strategy, the information content of commodity-specific predictors also generate substantial utility gains for a mean-variance investor with a relative risk aversion as high as five who imposes realistic portfolio weights constraints. Our results have two implications. First, at the margin, different traders seem to price the strategies that exploit the predictability of each asset class. Second, there appears to be different marginal risk arbitrageurs for each asset class in-line with the literature on limits to speculation (see, for instance, Garleanu and Pedersen (2011); Adrian and Shin (2010); Jylha and Suominen (2009)), and capital mobility (see, for instance, Duffie and Strulovici (2012); Duffie (2010)). Traders in the stock and bond markets seem to value information from the commodity market but not vice versa. The reported evidence could also be regarded as providing a possible explanation for why diversification benefits disappear out-of-sample (see Cotter, Eyiah-Donkor, and Poti (2017), Ahmed and Tsvetanov (2016), and Daskalaki and Skiadopoulos (2011)).