本研究利用Bai and Perron(1998)提出內生結構性轉折點之程序，來尋找台灣股市在1980年至2004年之間，是否存在結構性轉折點並且進行相關之探討。實證結果如下: 1.在未考慮結構性轉折點之前，台灣股市的週報酬有明顯的變異數異質的現象存 在，代表當期的變異數會受前期變異數以及前期誤差所影響。 2.依循Bai and Perron(1998)結構性轉折點之程序，所找出的結構轉折點為1988 年12月的第4週。主要原因為全球性股災。 3.加入結構轉折點的平均數方程式，台灣股價週報酬仍然有顯著的GARCH現 象。經過進ㄧ步的檢測發現，常數項並無顯著性差異，但是，前期自我相關向 呈現顯著性的差異。同時考慮常數項以及前期自我相關項並無發現顯著性的差 異。 4.將結構轉折點加入平均數方程式以及變異數方程式，台灣股價週報酬仍然有顯 著的GARCH現象。 5.將結構轉折點加入平均數方程式以及變異數方程式所尋找出的結構轉折點為 1987年3月，原因亦為全球性股災。 In this paper we review the factors that lead to change on volatility of stock market and use alternative methodologies of endogenous breakpoint detection in order to analyze whether the volatility of Taiwan stock market has changed significantly over the period 1980-2004. We follow the framework of Bai and Perron(1998) and use their sequential procedure and estimated critical value. The main finding of this research are summarized as follows : 1. In the sample GARCH(1,1) without break. We can find the GARCH effect is signifiicant. The variance is modeled as deterministic of past variances and error term. 2. We use the framework of Bai and Perron(1998). We can find the sup LR statistic to the last week of December 1988 is maximum and the statistic above the critical. So the last week of December 1988 is structural break. The main reason is the global crash if stock market. 3. We consider the structural break in the mean equation. We still find the GARCH effect is significent. The constant term are not significantly different. But the lag return are significantly different. 4. We consider the structural break in the variance equation. We still can find the GARCH effect is significant. 5. When we consider the structure break in the mean equation and variance equation, we find the sup LR statistic to March 1987 is maximum.
