Analysis of Call Rate Behaviour Using Regime-switching Diffusion Process Model (in Korean)

This paper estimates a regime-switching diffusion process model for Korean call rate. Parameters of the diffusion model with nonlinear general drift and volatility function vary depending on the economic states and are estimated using maximum likelihood method. In order to obtain the maximum likelihood estimates, we use the algorithm of Hamilton (1989) and calculate transition probability density function(TPDF) using Choi (2015a)'s method from the approximate log-TPDF obtained by At-Sahalia (2008)'s irreducible method. We find evidences that there exist two different regime and that the volatility term of the model explains the movements of the call rate better than the drift term. Also, the model shows that the transition probability varies depending on the level of interest rate. As interest rate increases, the probability of remaining at the same regime "L" next period decreases while the probability for the regime "H" increases. The estimated smoothed probability of regime "H," which is close to 1 and 0 during the period with high and low volatility respectively, is a clear evidence of the existence of two different regimes. Lastly, comparison with existing models demonstrates that the generalized model in this paper is preferred in explaining the call rate behaviour.