The Carry Trade and UIP when Markets are Incomplete

We propose a new model to explain the failure of UIP and the profitability of the carry trade and to link these two phenomena to the Balassa-Samuelson effect and the Backus-Smith puzzle. The key features of our model are market incompleteness and partial risk sharing through tradable goods. In the model, carry trade profits are due to two independent channels. First, a purely nominal channel, which works even in complete markets, makes the carry trade risky due to (endogenously) counter-cyclical inflation. Second, a real channel, which, due to imperfect risk sharing, makes the carry trade risky exactly when risk sharing is needed most. The model is consistent with several empirical facts. In particular: (i) real and nominal currency appreciations are positively related to local output growth, (ii) carry trade profits are positively related to output growth and negatively to inflation in the target (high interest rate) country, (iii) ex-ante, target countries are smaller and have higher expected inflation volatility, but there do not appear to be systematic differences between high and low interest rate countries in loadings on world output growth, in expected output growth, or in output volatility. Leading existing models give opposite predictions.