Can a dynamic correlation factor improve the pricing of industry portfolios?

We investigate whether a common trait shared by the major price patterns can be used to characterize the average returns of industry portfolios. We use a factor exploiting the premium induced by stocks with low and unstable correlations with the market and include it in the standard asset pricing models. The factor reduces the magnitude of alphas in these models and universally improves the description of the industry returns as measured by the GRS statistics. The finding is robust across different industry divisions and portfolio weightings.