Stock Return Predictability in a Production Economy

Abstract

This paper demonstrates that the standard real business cycle model, augmented by capital adjustment costs and variable capital utilization, yields countercyclical consumption growth volatility and equity risk premia predictability. In the model, as in the data, the investment to capital ratio and the fraction of investment in output are procyclical due to consumption smoothing. These two results imply that as productivity declines due to adverse productivity shocks, output falls at increasing rates (since in high productivity states the growth of the economy's stock of capital offsets the effect on output of productivity declines) and investment must be reduced by increasing rates if consumption were to fall at a constant rate. Convex capital adjustment costs and time to build (note that even in the standard RBC model capital takes one period to build) imply a desire for investment smoothing and a trade-off between investment smoothing and consumption smoothing. Since high productivity states are more likely in the data (as well as in the model) than low productivity states, much of the economy's consumption smoothing (i.e. investment fluctuations) occurs when productivity fluctuates between two relatively high states. The desire for investment smoothing implies that in the less frequent cases of transition into lower productivity states, more of the decline in output is absorbed by consumption. Consequently, in the model consumption growth is countercyclical (which is consistent with the empirical findings in Kandel and Stambaugh (1990) and Bekaert and Liu (2004)) as is the equity risk premium. Our model incorporates variable capital utilization rates which give rise to a countercyclical output gap. Model simulations show that the output gap serves as a measure for recessions and is positively correlated with future excess stock returns. In the data we find that the output gap is a strong predictor of future excess returns over both short and long horizons, and does a better job, both in-sample and out-of-sample, than other variables known to predict excess returns.