A machine learning approach to forecasting carry trade returns

ABSTRACT

Carry trade refers to a risky arbitrage in interest rate differentials between two currencies. Persistent excess carry trade returns pose a challenge to foreign exchange market efficiency. Using a data set of 10 currencies between 1990 and 2017, we find (i) a machine learning model, long short-term memory (LSTM) networks, forecast carry trade returns better than linear and threshold models and other machine learning models; and (ii) excess carry trade returns deteriorate after the 2007–2008 global financial crisis in all model forecasts, indicating that the uncovered interest rate parity may still hold in the long run.

KEYWORDS:

• Carry trade
• uncovered interest rate parity
• machine learning
• long short-term memory networks

JEL CLASSIFICATION:

• C45
• F21
• F31
• F37

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ See Burnside, Eichenbaum, and Rebelo (2008).

² See Davis, Liu, and Sheng (2020) on machine learning applications for time series problems and Fischer and Krauss (2018) on the application of LSTM networks in predicting stock returns.

³ Colombo, Forte, and Rossignoli (2019) employs support vector machines to predict carry trade direction.

⁴ We also confirm that $q_t-\bar{q}$ is stationary by the panel cointegration tests.

⁵ A rule of thumb for splitting between training and test sets is 5:1. Results are robust with other splitting.

⁶ Results are consistent under an alternative 80% training versus 20% validation rule.

⁷ The null versus alternative hypotheses are that the model (random walk, VAR, TECM, or RNN) has the same forecast accuracy versus that it has less forecast accuracy with LSTM networks.

⁸ The Sharpe ratio is measured as the difference between the portforlio return and the risk-free one-month bond rate, divided by the standard deviation of the portfolio. The gain/loss ratio is measured as the ratio of the probability of gains over that of losses.

Additional information

Funding

Xiao Wang acknowledges financial support from the National Natural Science Foundation of China (grant number: 72003181).