Uncovering interfirm links through textual topic similarity: A comomentum analysis in financial markets

Abstract

Using an unsupervised topic modelling methodology, we construct a cross-firm similarity measure based on the various topics extracted from Management Discussion and Analysis texts. Our findings indicate that the returns of firms with similar textual topics predict the focal firms’ future stock returns. A long-short portfolio constructed on this basis yields an annualised alpha of 17.03%. Further analyses show that the return predictability is stronger for stocks subject to limited investor attention and limits to arbitrage. Additionally, our textual linkage measure can also predict future earnings surprises. Overall, mispricing due to sluggish information incorporation acts as a potential explanation for return predictability.