The project decomposes historical market betas into one part that is predictive of future market betas, and another part that is not. If financial market prices are rational, it is only the first that should explain future rates of return. If markets are irrational, the latter part can matter, too. Thinking of what is behind this decomposition is the main idea of our initial project that has now been reduced to one year. That project of exploratory research is to help develop a fuller, theory-based research agenda on critical pricing factors, and their correspondence to rational market theory, in finance.
For example, it is plausible that extreme 1-day spikes are results of some historical accident and not systematic. Such daily observations are outliers in market-beta regressions. If the overall stock market rate of return was positive (negative) on this day, the calculated market beta of this stock would be high (low). This would be misleading. Fortunately, this means that spikes are almost like a natural experiment. To exploit this, we estimate one market beta based on all of a firm's historical returns, and one based on its winsorized returns. "Winsorized" data are those data that are left after outlier detection has been performed by specified statistical criteria and the flagged observations have been removed.
We first show that this decomposition works: The winsorized market beta predicts future market betas better than the unwinsorized market beta, and the difference between them is not predictive of future market betas. All hedging-based theories posit that only ex-ante market betas should matter. There should be no historical path-dependence. In contrast, behavioral theories, for the most part (as far as they can be pinned down) posit that investor biases respond primarily to their (historical) experiences.
Our technique of stripping out outliers in calculating crucial financial pricing statistics is more general in that it can apply to more than just market beta. We can test whether other factors that dominate the empirical pricing literature -- price momentum, HML, SMB, volume, volatility -- reflect rational pricing or not.
