We show that technical indicators deliver stable economic value in predicting the U.S. equity premium over the out-of-sample period from 1966 to 2014. Results tentatively improve over time and beat alternatives over a large continuum of sub-periods. By contrast, economic indicators work well...