Boosting the Predictability of the Market Risk Premium

Simulations in canonical models place these correlations much closer to -1 than 0. Avoiding the slack bounds against 0, we construct tight bounds based on correlations between the pricing kernel and functions of returns. Empirically, these bounds work well in terms of out-of-sample R-squared (26% against the benchmark of a long 124-year sample of historical risk premiums). The new bound draws information from both risk-neutral and physical moments of returns.