Prevailing theories suggest that the quality of information about the macroeconomy is an important determinant of the equity premium. However, the empirical relationship between information quality and the equity premium has been obscured by the difficulty of measuring time-varying information quality, which is dynamically intertwined with uncertainty. By solving a nonlinear filtering problem, I extract the time series of information quality from data on professional forecasters, while differentiating information quality from uncertainty, volatility, and heterogeneity. Surprisingly, I find that the equity premium increases with information quality in contrast to standard asset pricing theory. In the U.S. stock market, a one-standard-deviation increase in information quality predicts a 3% increase in quarterly excess market returns with R2 up to 7.4% (and out-of-sample R2 up to 6.1%). Furthermore, information quality outperforms all the predictors studied in Goyal and Welch (2008) both in-sample and out-of-sample, while improving the return predictability by the dividend-price ratio. This strong evidence of return predictability is very robust to various specifications. Also, information quality is negatively correlated with the real risk-free rate, as opposed to the predictions of standard asset pricing theory. I propose a stylized consumption-based model of ambiguity aversion to future signals. This new model can simultaneously account for all these empirical findings and the equity premium puzzle, unlike the existing preferences: CRRA, Epstein-Zin, and ambiguity aversion to the mean growth rate, introduced by Ju and Miao (2012). Three components of the model-implied stochastic discount factor --- information quality, uncertainty, and aggregate signal --- explain up to 83% of the cross-sectional variation in the average returns on size, book-to-market, and momentum-sorted portfolios.
Read full abstract