Title
On the Long Run Volatility of Stocks: Time-Varying Predictive Systems
Speaker
Rob McCulloch, The University of Chicago
Abstract
A widely held belief is that investing in stocks is less risky if you have a long investment horizon. Thus, we are often often advised to put more stocks in our pension portfolio when we are young and less when we are old. In "Predictive Systems: Living with Imperfect Predictors", Pastor and Stambaugh (2008) develop a framework for estimating expected returns. In "Are Stocks Really Less Volatile in the Long Run" (2009), they use this framework to assess the conventional wisdom that stocks are less volatile over long horizons than short horizons. They show that this conclusion is only reached by ignoring important parameter uncertainty. They also argue that a key component of prior information concerns the correlation between unanticipated expected return and the unpredictable return.
The predictive system framework consists of a vector auto regression in the stock return, the latent expected return for the next period, and a set of variables thought to be able to predict returns. We examine the sensitivity of the results to prior and model specification and find that we can find different results for "reasonable" priors and models.
