Equity premium prediction: The role of economic and statistical constraints

In this paper, we show that the equity premium is predictable out-of-sample when we use a predictive regression that conditions on a large set of economic fundamentals, subject to: (1) economic constraints on the sign of coefficients and return forecasts, and (2) statistical constraints imposed by shrinkage estimation. Equity premium predictability delivers a certainty equivalent return of about 2.7% per year over the benchmark for a mean-variance investor. Our predictive framework outperforms a large group of competing models that also condition on economic fundamentals, as well as models that condition on technical indicators.