| In this dissertation, we study some Bayesian econometric models and their applications on the term structure of risk-free interest rates. In Chapter 1, we propose a new Markov Chain Monte Carlo (MCMC) algorithm for the Markov switching models with ARMA-GARCH error. Due to the error structure, the conditional variance depends on all the past history of the state variable and it makes the maximum likelihood estimation hard to implement. We modify the single move procedure suggested by Carlin, Polson and Stoffer (1992) to generate the state variable. Through the simulations, we show that our MCMC algorithm works in such complicated models.;In Chapter 2, we examine the expectations theory of the term structure of risk-free interest rates using the Bayesian inference. The US data do not conform to the expectation theory with the constant term premium. We modify the Autoregressive Conditional Heteroskedasticity in Mean (ARCH-M) model which is introduced by Engle, Lilien, and Robins (1987) and estimate the time-varying term premium by our MCMC algorithm. We find that the time-varying term premium plays an important role in non-credible monetary regimes. We also find that the premium tends to decrease in economic expansions and increase in recessions and financial distress.;In Chapter 3, we estimate the dynamics of 3-, 6-, 12-month, 2-, 5- and 10-year risk-free interest rates. We set up a dynamic factor model with one factor ignoring the no-arbitrage conditions. We suggest a MCMC algorithm and estimate the model with it. We find that the patterns of the factors on the different maturities are different from period to period. |
