Comovement and Return Predictability
15 Pages Posted: 22 Aug 2007
In this paper, we investigate whether comovement can generate predictability in returns. Barberis and Shleifer (2003) argue that style investing causes comovement of stocks within a style to be larger than that warranted by fundamentals. This naturally suggests that comovement may measure "investor sentiment", a construct that both academics and practitioners would like to measure with precision. The former are interested in measuring sentiment for scientific reasons and the latter for mercenary ones. Unfortunately, instruments to measure sentiment are either sparse or of poor quality (Baker and Wurgler (2007)). But since our interest is not in aggregate sentiment but comovement with respect to a particular style, we can use an extraordinarily simple and yet precise measure: the R2 of a regression of the stock on the returns of style it belongs to. We measure comovement with respect to size and book-to-market based portfolios because of their ubiquitous role in both retail and institutional portfolio management. We find that between 1970 and 2006, a long-short portfolio strategy based on low and high comovement stocks generates gross returns of 0.9% per month for up to 3 months and 0.7% per month up to 12 months. The excess returns (alpha) after controlling for size, book-to-market and momentum effects are up to 0.86% per month. Although comovement is correlated with trading volume, our results withstand the incorporation of volume in to the trading strategy (Lee and Swaminathan (2000)). As we complete our analysis over the next few months, we expect to (a) use an alternative measure of comovement, (b) examine measures of investment style beyond the traditional size and value-growth grid, (c) examine time variation in excess returns, and (d) determine if changes in comovement can explain the reversal in returns over long-horizons (DeBondt and Thaler (1985)).
Suggested Citation: Suggested Citation