Abstract

We study the implications of contagious beliefs for business cycles. We introduce a real business cycle model in which heterogeneous firms dynamically switch between models of the world. Models spread based on both their consistency with data, as in models of learning, and their prevalence, as in models of social contagion. The spread of models generates endogenously persistent belief-driven fluctuations. If contagion is sufficiently strong, models can “go viral” and induce hysteresis in the model’s unique equilibrium. To take this framework to the data, we adopt a “micro-to-macro” approach in which we measure firms’ models using the sentiment of their language in regulatory filings. We find that firms accelerate hiring when they are optimistic, even though this sentiment does not positively predict future firm fundamentals or performance. Moreover, sentiment spreads contagiously at the aggregate and industry levels. Mapping our microeconomic estimates into the model, we find that contagious beliefs account for one fifth of business cycle fluctuations in aggregate output.