Macroeconomic Effects of Quantitative Easing Using Mid-sized Bayesian Vector Autoregressions

Abstract

The paper estimates macroeconomic effects and decomposes transmission channels of quantitative easing in the United States using 15-variable Bayesian vector autoregressive model with stochastic search variable selection prior, distinguishing between Treasury bond purchases, mortgage-backed securities purchases and Operation Twist. A positive quantitative easing shock has a strong, negative impact on unemployment and no impact on prices, with Treasury purchases and Operation Twist found to be more effective than purchases of mortgage-backed securities. Opposite to the assumptions usually made in the literature, quantitative easing transmits to the real economy mostly via the stock market instead of long-term rates. Among numerous extensions to the baseline model, spillbacks are found to account for 40% of the impact of Treasury purchases on unemployment and commercial paper purchases have similar effects on the economy as purchases of Treasury bonds and mortgage-backed securities. However, baseline estimates are not found to be very robust, and thus substantial uncertainty regarding the macroeconomic effects of QE persists.