This paper tries to use Multivariate Time Series Variance Com-ponent Analysis (MTV) and Principal Component Regression (PCR) to analyze the structure of Taiwan stock prices and other macroeco-nomic variables from Jan. 1986 to Dec. 1990. It is found that finan-cial variables are most important factors to be related to Taiwan stock prices, and fol1owed by production and trade variables, foreign stock prices and Taiwan stock price expectation. We also found that the performance of prediction of MTV is better than that of PCR.