Foreign exchange risk premia: from traditional to state-space analyses

TL;DR

This paper explores foreign exchange risk premia using traditional regressions and advanced state-space models, revealing their time-evolving nature, stationarity, and volatility characteristics.

Contribution

It introduces a comprehensive analysis combining traditional and state-space methods to better understand the dynamics of foreign exchange risk premia.

Findings

Risk premia are time-evolving and significantly stationary.

State-space models effectively capture the time variability of risk premia.

Volatility in risk premia is relatively high but not dominant in prediction errors.

Abstract

This paper examines foreign exchange risk premia from simple univariate regressions to the state-space method. The adjusted traditional regressions properly figure out the existence and time-evolving property of the risk premia. Successively, the state-space estimations overall are quite rationally competent in examining the essence of time variability of the unobservable risk premia. To be more precise, the coefficients on the lagged estimated time-series are significant and the disturbance combined from the observation and transition equations in the state-space system, rational and premium errors, respectively, is statistically white noise. Such the two residuals are discovered to move oppositely with their covariance approaching zero suggested by the empirics. Besides, foreign exchange risk premia are projected and found significantly stationary at level and relatively volatile throughout time with some clustering. This volatility is however not quite dominant in the deviations of forward prediction errors.