This paper examines empirically stochastic volatility of the KOSPI200 index. This paper adopts the Bayesian multi-state sampling technique for the estimation of the latent volatility process. In particular, this paper uses the simulation smoother proposed by De Jong and Shephard (1995) to draw the latent variables from its conditional distribution. The empirical results using two data sets, the KOSPI200 and KOSPI indices for 1/3/90-9/4/95, suggest that stochastic volatility model (SVM) is a reasonable description of the Korean stock prices and that the predicted volatility from the SVM may be useful in evaluating the stochastic volatility option pricing models like Hull and White (1987) in addition, although the level of volatility appears to have decreased since early 1990`s, the volatility gap between the KOSPI200 and KOSPI seems to have increased.