REGIME-SWITCHING PRODUCTIVITY GROWTH AND BAYESIAN LEARNING IN REAL BUSINESS CYCLES

Abstract

Growth in total factor productivity (TFP) in the USA has slowed down significantly since the mid-2000s, reminiscent of the productivity slowdown of the 1970s. This paper investigates the implications of a productivity slowdown on macroeconomic variables using a standard real business cycle (RBC) model, extended with regime-switching in trend productivity growth and Bayesian learning regarding the growth regime. I estimate the Markov-switching parameters using US data and maximum-likelihood methods, and compute the model solution using global projection methods. Simulations reveal that, while adding a regime-switching component to the standard RBC setup increases the volatility in the system, further incorporating incomplete information and learning significantly dampens this effect. The dampening is mainly due to the responses of investment and labor in response to a switch in the trend component of TFP growth, which are weaker in the incomplete information case as agents mistakenly place some probability that the observed decline in TFP growth is due to the transient component and not due to a regime switch. The model offers an objective way to infer slowdowns in trend productivity, and suggests that macroeconomic aggregates in the USA are currently close to their potential levels given observed productivity, while counterfactual simulations indicate that the cost of the productivity slowdown to US welfare has been significant.