Economic linkages inferred from news stories and the predictability of stock returns

We show that news stories contain information about economic linkages between firms and that information diffuses slowly across linked stocks. Specifically, we identify linked stocks from co-mentions in news stories and find that these stocks subsequently cross-predict each other's returns. Content analysis of shared news stories uncovers previously undocumented types of firm linkages, whose effects are distinct from those analyzed in prior literature. Furthermore, our results indicate that information can flow from smaller to larger stocks and across industries. Both limited attention and slow processing of complex information contribute to slow information diffusion.