Three essays on Bayesian analysis of Korean economy

This dissertation applies recently developed Markov Chain Monte Carlo algorithms for estimation of either atheoretic or theoretic model to capture dynamics in less well-behaved economy and explores estimated dynamics for various analyses. Chapter 1 discusses the outline of this dissertation.; Chapter 2 applies a Bayesian method for filtering trend growth rates. We consider a VAR model with drifting parameters and time-varying stochastic volatility to let the economy's law of motion evolve. We propose the industrial production to inventory ratio as a cyclical indicator. We found trend growth rate decreases over the sample period. There were still substantial uncertainties and difficulties in estimating the trend, even when we focused the trend output growth rate rather than the trend output level. Trend growth rate in Korea shows a transitional path to a steady state.; Chapter 3 applies a Bayesian method to explore inflation dynamics. We consider an unobserved component model with drifting parameters and time-varying stochastic volatility to extract core inflation from volatile inflation. We extend the Stock and Watson (2005) model by allowing time-varying stochastic volatility to be estimated by simulation proposed by Cogley and Sargent (2005) instead of calibration. We found that time-varying core inflation has been moderated dramatically in Korea. Estimated core inflation shows a transitional path where the economy has evolved to a steady state. Stochastic volatilities of inflation, either measurement innovation or state innovation, were reduced substantially. Uncertainties in retrospective estimation of core inflation near the terminal period were still substantially high and resulted in substantial uncertainties in forecasting core inflation. Time-varying stochastic core inflation can be well approximated by a one-side filter such as an exponential filter making it suitable for practical use since it can be computed easily and quickly.; Chapter 4 applies recently developed Bayesian methods to estimate a small-scale DSGE model for less well-behaved economy. We consider the both first and second order approximations in a DSGE model and compare the estimates of the structural parameters delivered by the corresponding Kalman filter and sequential Monte Carlo filter. Contrary to the claims of the literature, we found the second order approximation for less well-behaved economy and the corresponding sequential Monte Carlo filter methods do not deliver the better posterior in terms of the marginal data density. Quite different estimates are delivered by the different filtering methods and across the different sample periods even by the same filtering method. Unstable steady states estimated in this chapter are consistent with transitional paths separately estimated in chapter 2 and chapter 3, respectively. However, generalized impulse response in the second order approximated model can capture an important feature of response dynamics to exogenous shocks missed in the first order approximated model.