The newsworthiness of an event is partly determined by how unusual it is and this paper investigates the business cycle implications of this fact. In particular, we analyze the consequences of information structures in which some types of signals are more likely to be observed after unusual events. Such signals may increase both uncertainty and disagreement among agents and when embedded in a simple business cycle model, can help us understand why we observe (i) occasional large changes in macro economic aggregate variables without a correspondingly large change in underlying fundamentals (ii) persistent periods of high macroeconomic volatility and (iii) a positive correlation between absolute changes in macro variables and the cross-sectional dispersion of expectations as measured by survey data. These results are consequences of optimal updating by agents when the availability of some signals is positively correlated with tail-events. The model is estimated by likelihood based methods using individual survey responses and a quarterly time series of total factor productivity along with standard aggregate time series. The estimated model suggests that there have been episodes in recent US history when the impact on output of innovations to productivity of a given magnitude were more than eight times as large compared to other times.
