Abstract

The economic gains from using equity premium forecasts are usually assessed by comparing a forecast-based strategy to a strategy based on the trailing historical mean. Whether these economic gains are statistically significant remains mostly untested. This paper shows that a buy-and-hold benchmark can be much harder to beat than the historical-mean benchmark and that the practice of not testing the statistical significance of economic gains can lead to questionable conclusions. The findings rest on an examination of many hypothetical sample periods and the replication of two widely cited papers (Rapach et al., 2010, 2016).