A large psychology literature argues that, due to selective memory recall, decision-makers’ forecasts of the future are overly influenced by the perceived news. We adopt the diagnostic expectations (DE) paradigm [Bordalo et al. (2018), Journal of Finance, 73, 199–227] to capture this feature of belief formation, develop a method to incorporate DE in business cycle models, and study the implications for aggregate dynamics. First, we address (1) the theoretical challenges associated with modelling the feedback between optimal actions and agents’ DE beliefs and (2) the time-inconsistencies that arise under distant memory (i.e. when news is perceived with respect to a more distant past than just the immediate one). Second, we show that under distant memory the interaction between actions and DE beliefs naturally generates repeated boom–bust cycles in response to a single initial shock. We also propose a portable solution method to study DE in dynamic stochastic general equilibrium models and use it to estimate a quantitative DE New Keynesian model. Both endogenous states and distant memory play a critical role in successfully replicating the boom–bust cycle observed in response to a monetary policy shock.

We use a high-frequency identification approach to document that individual politicians affect asset prices. We exploit the regular flow of viewpoints contained in Congress members’ tweets. Supportive (critical) tweets increase (decrease) the stock prices of the targeted firm and the corresponding industry in minutes around the tweet. The bulk of the stock price effects is concentrated in the tweets revealing news about future legislative action. The effects are amplified around committee meeting days, especially when the tweet originates from committee members and influential politicians. Overall, we show that Congress members’ social media accounts are an important source of political news.