Are the New-Generation Treasury Return-Predictive Factors Economically Significant? A Cross-Currency and Out-of-Sample Investigation

Abstract

Recent studies claim that a new generation of return-predicting factors can predict excess returns in the US Treasury market far better than the slope-related factors. The new-generation factors are, however, often difficult to interpret and far less parsimonious, and therefore doubts have been raised about their robustness, and about whether their high predicting power may be an artifact of data mining and overfitting. By predicting excess returns in one currency using return-predicting factors estimated from a different currency, we present strong evidence suggesting that these new factors are statistically informative, economically meaningful and, at least in their `restricted form, surprisingly robust. A new out-of-sample technique for same-currency returns that we introduce reinforces these conclusions. Our findings also point to a commonality of fundamental financial mechanism(s) at the origin of the predictability observed in all the Treasury markets examined.