Disagreement Over the Long-Run: The Effects of Higher Order Expectations on the Equity Premium

Abstract

I present a novel asset pricing model where agents have private signals about the long-run conditional mean growth rates of their consumption and dividend endowments. Agents use their private information in combination with publicly known observables to endogenously form model consistent expectations. In the absence of arbitrage, agents are left forecasting the beliefs of other agents, more commonly known to the literature as higher order expectations. The model generates a countercyclical price of risk for news about long-run growth. Agents' precision about the true mean growth rate improves during bad times, rendering prices more sensitive to new information. However, the price of risk quickly becomes small and insubstantial if the scale noise grows large, clouding agents' perception of the true conditional mean. Therefore the success of asset pricing models with long-run risks hinges on agents' heterogeneous beliefs about future growth sharing a tight distribution around the truth. The model also makes a new contribution to the consumption based literature in that it can match features of time-varying disagreement in real consumption growth forecasts from the Survey of Professional Forecasters. Heteroskedasticity in agents' endowments drives the cross-sectional dispersion of beliefs about future consumption growth.