Working Paper: CEPR ID: DP15305
Authors: Ilias Filippou; David Rapach; Mark Taylor; Guofu Zhou
Abstract: We establish the out-of-sample predictability of monthly exchange rate changes viamachine learning techniques based on 70 predictors capturing country characteristics,global variables, and their interactions. To guard against overfi tting, we use the elasticnet to estimate a high-dimensional panel predictive regression and find that theresulting forecast consistently outperforms the naive no-change benchmark, which hasproven difficult to beat in the literature. The forecast also markedly improves theperformance of a carry trade portfolio, especially during and after the global financialcrisis. When we allow for more complex deep learning models, nonlinearities do notappear substantial in the data.
Keywords: exchange rate predictability; elastic net; carry trade; deep neural network
JEL Codes: C45; F31; F37; G11; G12; G15
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| machine learning predictors (C45) | monthly exchange rate changes (F31) |
| elastic net forecast (C45) | performance of carry trade portfolio (G15) |
| forecasts (G17) | better predictions of currency depreciation (F31) |
| global foreign exchange volatility (F31) | effectiveness of predictors (C52) |