Uncovering attention heterogeneity

Heterogeneity in individuals’ attention to news shapes how aggregate shocks propagate through the economy. We propose a novel method to uncover the full distribution of attention in surveys of consumers and professional forecasters, by developing a finite mixture model to cluster forecasts according to their level of attention. Our results reveal substantial heterogeneity in attention, which, when ignored, leads to an overestimation of average attention. Leveraging our clustering, we construct decade-long panel datasets on attention and provide new insights into its drivers. We document an asymmetric effect of aggregate shocks on the distribution of attention and assess how attention varies with socioeconomic characteristics. Finally, we identify systematic deviations from Bayesian updating and related theories designed to account for behavioral biases. In particular, we find that individuals’ attention is less sensitive to changes in uncertainty and signal precision than predicted by these theories. Overall, this paper opens new avenues by providing the first panel datasets on attention and illustrating how they can inform theories of imperfect information through within-individual variations.