Estimation and Identification Issues in Monetary Policy Rules

Thesis (Ph.D.)--University of Washington, 2015The dissertation explores the links between macroeconomic phenomena and monetary policy and to develop new econometric methods. In the first chapter, “Monetary Policy Rules and Macroeconomic Stability Revisited: Limited Information Approach under Identifying Restrictions, provides a new approach to limited information estimation consistent with the forward-looking monetary policy rule. Recently, the weak identification in the conventional estimation method has drawn attention to the estimation of a forward-looking monetary policy rule. This paper identified a particular range for the value of the concentration parameter, for which the generalized method of moments (GMM) suffers from the weak identification problem, while the proposed method does not. This implies that GMM estimation generates spurious weak identification in the estimation of a forward-looking monetary policy rule. The proposed approach allows us to provide stronger messages to the estimation of a forward-looking monetary policy rule. The estimation results confirm a change of monetary policy in the U.S. In the 1960-1979 sample, the policy was inactive and it did not react sufficiently to the expected deviation of inflation from its target. In contrast, under the 1979-1997 sample monetary policy actively responses to the inflation with a high degree of interest smoothing. The second chapter of the dissertation is the extension of the first chapter, “Estimation of a Time-varying Forward-looking Monetary Policy Rule: Limited Information Approach. In this chapter, I estimate a time-varying forward-looking monetary policy rule by considering a time-varying structural vector auto-regression (VAR) model for the monetary transition mechanism. Assuming that the time variation comes from the coefficients and the variance covariance matrix, I illustrate this via modeling multivariate stochastic volatility. In a foundational paper, Primiceri (2005) estimated a time-varying structural VAR with stochastic volatility after assuming monetary policy shocks to be independent of any other innovations without forward-looking variables. Because agents are assumed to be rational, monetary policy changes can be incorporated into future forecasts. Kim and Nelson (2006) used a single equation to investigate the estimation of a forward-looking monetary policy rule in relation to the forward-looking behavior of agents. To account for the endogeneity, they suggested a two-step estimation technique based on the control function approach. However, as Chon and Kim (2014) argued, the error term in instrumenting equations for forward-looking variables follows moving-average (MA) dynamics, resulting in additional information loss. Consequently, this paper illustrates that one can recover this MA structure after considering the reduced-form of the time-varying VAR; the procedure suggested in this paper resolves the possible weak identification issues. The third chapter of the dissertation is “Stock Market Reaction to Monetary Policy Changes: Identification through Heteroskedasticity with Markov-switching.” This paper investigates the estimation issues surrounding the response of asset prices to monetary policy changes. Because of the simultaneous relationship between stock prices and policy decisions, and because both react to numerous other variables, estimation of the impact of stock price to monetary policy action is difficult. In this paper, I use the heteroskedastic structure of monetary policy shocks to identify stock market reactions to monetary policy changes following Rigobon and Sack (2004). Especially, in order to consider all possible sources which affect shifts in monetary policy shocks, such as the alteration of expectations about the future path of the monetary policy and a change in the timing of policy moves, I incorporate the Markov-switching framework to detect different state endogenously. The procedure proposed in this paper can reduce the potential bias caused by mis-specified timings in the shifts of monetary policy shocks and produce more precise estimate of the monetary policy actions on the stock market. Since the stock market is forward-looking, I focus on the surprised part of the policy actions within the conventional event-study framework. The empirical finding tells us that the heteroskedasticity on event day may well be a consequence of the asymmetric effects on the different types of policy actions: expansionary policy vs. contractionary policy. Also, we found that the unanticipated 25-basis point increase would decrease 1.91 percent in the S&P 500 returns