Financial crisis and slow recovery with Bayesian learning agents

In a simple continuous-time model where the learning process affects the willingness to hold liquidity, we provide an intuitive explanation of business cycle asymmetry and post-crisis slow recovery. When observing a liquidity shock, individuals rationally increase their subjective probability of re-encountering it. It leads to an upward jump in liquidity preference and a discrete fall in consumption. Conversely, as a period without shocks continues, they gradually decrease the subjective probability, reduce liquidity preference, and increase consumption. The recovery process is particularly slow after many shocks are observed within a short period because people do not easily change their pessimistic view.