Modeling Expected Stock Returns for Long and Short Horizons

Expected returns over long and short horizons are modeled using two approaches: an equilibrium asset pricing model and a vector autoregression (VAR). Empirical properties of returns that are consistent with the equilibrium model’s implications include (i) an annual "equity premium" of about six percent (ii) a U-shaped pattern of autocorrelations of returns with respect to investment horizon for the R-squared in projections of stock returns on predetermined financial variables. Parameters estimated in a monthly VAR for returns and these financial variables also imply autocorrelations, R-squared values, and conditional expected returns that are close to those computed with actual long-horizon returns. Simulations indicate that such a VAR is a reasonable approximation to the equilibrium model for representing the properties short- and long-horizon returns.