Revisiting the Meese and Rogoff Puzzle: From Micro-Level Stock Market Evidence

Abstract

This paper revisits the Meese and Rogoff puzzle using micro-level financial market evidence. We employ a large panel of micro-level stock returns to forecast exchange rates and find that they significantly improve the predictive accuracy over the benchmark random walk. This advantage is particularly evident in signaling currency appreciation and depreciation, in the medium- to long-term forecasts, and during the periods of high exchange rate volatility. Compared to aggregate stock returns and classical structural models, micro-level stock returns can capture more detailed information on the stock market and achieve superior forecasting performance. Our findings offer a new perspective on addressing the Meese and Rogoff puzzle, emphasizing the importance of incorporating micro-level financial market information into exchange rate prediction.

Keywords: Meese and Rogoff Puzzle, Exchange Rate Forecasting, Micro-Level Information, Financial Market, Machine Learning

JEL Classification: C53, E17, F37