Learning, Observability and Time-varying Macroeconomic Volatility

Abstract

Our paper provides a theoretical explanation for the time-varying macroeconomic volatility by introducing the unobservability of regime switching and learning. With the unobservability of regime switching, agents must endogenously form their expectations using best-performed forecasting models. We find that if the regime switching is observable to agents, agents do not shift their expectation frequently and so will not generate a larger macroeconomic volatility. However, with the unobservability of regime switching where no agents can know which regime is dominant, allowing endogenous expectation formation would give rise to larger macro fluctuations (first-layer amplification mechanism), which is made through agents frequently shifting their expectations. Furthermore, we consider the policy implication under the zero lower bound. Our simulations show that in the unobservable regime switching, the economy is more likely to fall into a deflationary trap. To avoid the deflation risk, the policy maker should set a higher expected inflation based threshold. If the expected inflation is under the threshold, an aggressive policy rate will be implemented; otherwise, the normal Taylor-rule monetary policy will be used. Furthermore, to reduce the deflation risk, the strategy for the policy maker is to raise the threshold, and this will generate larger macroeconomic fluctuations (second-layer amplification mechanism) due to more frequent policy strategy switching. We argue that sometimes only with unobservability, the policy maker faces a dilemma between avoiding deflation risk and maintaining macroeconomic stability, and huge macroeconomic fluctuations do not necessarily result from bad luck or bad policy but from the two-layer amplification mechanism caused by the unobservability.