Predictability and Underreaction in Industry-Level Returns: Evidence from Commodity Markets
47 Pages Posted: 16 Feb 2012 Last revised: 12 Sep 2013
Date Written: September 11, 2013
This paper finds significant evidence that commodity price changes can predict industry-level returns for horizons between one trading day and up to six trading weeks (30 days). We find that for the 1985-2010 period, 40 out of 49 U.S. industries can be predicted by at least one commodity. Our findings are consistent with Hong and Stein’s (1999) “underreaction hypothesis”. Unlike other papers in the literature, we pinpoint the length of underreaction by using daily data. We provide a comprehensive examination of the return linkages among 25 commodities and 49 industries. This enables us to present a more detailed investigation of underreaction and investor inattention hypotheses. Our results indicate modest out-of-sample forecast ability. Finally, we implement data-mining robust methods to assess the statistical significance of industry returns reactions to commodity price changes; these results confirm there is evidence that commodity data can predict equity returns more than four trading weeks ahead.
Keywords: Asset pricing, Commodity markets, Equity markets, Industry-level returns, Information and market efficiency, Predictability, Underreaction
JEL Classification: G12, G14, G15, E44
Suggested Citation: Suggested Citation