A Comprehensive Look at the Empirical Performance of Equity Premium Prediction II

(With Amit Goyal and Athanasse Zafirov)

Our paper reexamines whether 29 variables from 26 papers published after Goyal and Welch (2008), as well as the original 17 variables, were useful in predicting the equity premium in-sample and out-of-sample. Our samples include the original periods in which these variables were identified, but ends later (in 2020). Most variables have already lost their empirical support, but a handful still perform reasonably well. Overall, the predictive performance remains disappointing
In the following description, you want to see a statistically significant in-sample (IS) statistic, two half (H1 and H2) coefficients that are not very different or sign changing, and an out-of-sample (OOS) R² that is statistically significant and positive. Variables without IS significance are not worth looking at. Beyond this, you want variables that did not just work in one year, say 1950, and then never again. (The full information about variables with better descriptions is in the paper).

/home/card-eqpredict.html Last modified: