“Currency Carry Trade Return Predictability and Asset Pricing Implications” by Gurdip Bakshi
the University of Maryland
This paper studies the time-series return predictability of currency carry trades, constructed by selecting currencies to be bought or sold against the U.S. dollar, based on forward discounts. Changes in a commodity index, currency volatility and, to a lesser extent, a measure of liquidity predict in-sample the returns of dynamically re-balanced carry trades, as evidenced by individual and joint p-values in monthly predictive regressions at horizons up to six months. Predictability is further supported through out-of-sample metrics and combination forecasts, and a predictability-based decision rule produces sizeable improvements in the Sharpe ratios and skewness profile of carry trade returns. Our evidence also indicates that predictability can be traced to the long legs of the carry trades and their currency components. We test the theoretical restrictions that a factor model, with average currency returns and the mimicking portfolio for the innovations in currency volatility as risk factors, imposes on the coefficients in predictive regressions.