The Effect of Risk Changes on Carry Trade Returns and Speculative Behavior

Abstract

This paper analyzes the time-series reaction of carry trade returns to changes in various risk factors. Using non-linear methods, I find that implied currency volatility is an informative time-series predictor. Increases (declines) in the implied currency volatility (or, generally, perceptions of future risk) result in lower (higher) subsequent carry trade returns. The reaction to extreme changes in these risk perceptions is even more pronounced. Using futures positioning data, it is shown that speculators tend to sell carry trade positions upon a perception of increased risk and vice versa. A piecewise linear threshold model is proposed to predict short-term carry trade returns; it outperforms a variety of benchmarks (including the random walk) on almost all metrics. Robustness tests suggest that this performance does not depend on certain model settings or the sample period (i.e. data mining); instead, a rollingly updated model would lead to even better results.