Bayesian Methods and Markov Switching Models for the Analysis of U.S. Postwar Business Cycle Fluctuations
- Author(s): Li, Jie
- Advisor(s): Chauvet, Marcelle
- et al.
This dissertation consists of five chapters addressing analytically and empirically U.S. Postwar business cycle fluctuations. Markov Switching models and Bayesian estimation methods are used to investigate United States macroeconomic dynamics in the last 60 years. Chapter 1 introduces the structure of this dissertation. Chapter 2 proposes a dynamic stochastic general equilibrium (DSGE) model with Markov Switching and heteroskedastic shocks to examine the role of agents' beliefs separately from changes in monetary policy in explaining inflation fluctuations. Bayesian analysis is conducted with Markov Switching to support regime switches in the private sector, in the implementation of monetary policy and in the volatility of shocks in the U.S. Postwar economy, which are related to the "Great Inflation", the "Great Moderation" and the 2008 financial crisis. A counterfacutal analysis found that if agents maintained a weak response to macroeconomic dynamics over time, there would be lower inflation during the "Great Inflation". In addition, irrespectively to monetary policy regimes, supply shocks are the main driver of inflation fluctuations, while demand shocks are the main source of changes in the output gap. However, when agents maintain a higher risk aversion towards consumption with a higher slope in the Phillips curve, demand shocks also play a role in driving inflation, even though supply shocks are still the main driver of inflation. Chapter 3 emphasizes on the monetary policy with an investigation on the assumption that policymakers commit to a Taylor rule, using a time-varying inflation-unemployment dynamic model on U.S. economy. This chapter is based on the conjecture that potential policymakers' misperception may be originated from unobserved deviations of unemployment from its natural rate. Five processes are proposed for policymakers' belief under commitment to inflation and unemployment and compare them with a baseline autoregressive process without commitment. The models are estimated using Bayesian techniques. Empirical results are as follows: First, policymakers' belief is very persistent even when it commits to a Taylor-type policy rule. Second, the run-up of U.S. inflation around 1980 can be mostly attributed to policymakers' misperception while the peak surge of inflation in 1974 is possibly a result of non-policy shocks. Third, models with commitment dominate models without commitment, especially in periods of large oscillations in inflation. In particular, when policymakers are committed to respond to a Taylor-type policy rule, the average loss function is considerably reduced over time, thus effectively lessening potential misperceptions. Chapter 4 introduces a simple version of adaptive expectation to a dynamic stochastic general equilibrium (DSGE) model to evaluate goodness of fitness and forecasting performance on U.S. macroeconomic indicators. Analytical maximum likelihood estimation results represent a DSGE model with adaptive expectation outperforms a DSGE model with rational expectation. In addition to providing a better fit of inflation and output gap in the U.S. Postwar macro economy, a DSGE model with adaptive expectation also leads to redundant lagged inflation in fitting inflation dynamics. Chapter 5 concludes and proposes future extension.