Reasearch

We study how attention to economic information varies across individuals and over time. Using a new finite mixture approach to classify forecasts by attention level, we document substantial heterogeneity which, when ignored, overstates average attention. Applying our method to consumer and forecaster surveys, we construct decade-long panel datasets showing that attention relates systematically to socioeconomic characteristics--with highest attention levels among low-income and minority groups--and responds non-uniformly to aggregate shocks. Our approach offers new empirical discipline for expectation formation theories, highlighting, for instance, that attention adjusts less to uncertainty and signal precision than predicted by standard models.

 Many consumers do not respond to unanticipated income windfalls, a feature I attribute to adjustment friction in expectations. In the model, consumers incur a fixed cost to adjust cash-on-hand expectations and choose adjustment dates based on noisy signals of expectation error—the difference between actual and expected cash-on-hand. This results in an inattention region where consumers ignore news and do not respond to cash-on-hand changes if the signal remains within the region boundaries. Evidence from expectation and expenditure data supports this departure from Bayesian updating. I investigate implications for the marginal propensity to consume, both qualitatively and quantitatively.

Using data from U.S. public companies over recent decades, we document a paradox: while prospective market leaders have gained market share at an increasing speed, overall turnover in leadership has slowed. Although the dynamism of market shares suggests more contestable markets, the persistence of leadership indicates the opposite. We address this puzzle with a model of endogenous growth, where improved consumer access to market data accelerates market share dynamics and increases firms’ incentives to innovate. Greater competition enhances R&D productivity but also amplifies the misallocation of knowledge production. Consequently, innovation occurs through larger but less frequent spikes. Our model aligns with the data, predicting higher market concentration, increased markups, reduced turnover, and slower growth, even alongside more dynamic market shares and higher R&D spending. Ultimately, the paper shifts attention from insufficient competition in the goods market to inefficient competition in the R&D sector as a potential driver of secular fall in aggregate productivity.

Rounding in survey data obscures the identification of key parameters. We show that this issue is both conceptually and empirically important for estimating two measures of attention: the fraction of individuals who adjust their forecasts and the weight they place on news when doing so. We propose a simple fix-treating forecast revisions as interval data-which restores point identification. Our results reveal that many zero revisions arise from rounding rather than non-adjustment, and that even when individuals adjust, they remain largely inattentive.