運用門檻回歸模型探討台灣股市指數走勢及交易訊息－信用交易指標應用

台灣股市交易中散戶投資者扮演著不可忽略的角色，因此探討散戶交易行為一直是可以深入討論的議題。為此，本研究以信用工具作為散戶交易行為之代理觀察變數，並且運用門檻回歸模型，探討不同信用工具交易餘額下台灣市場指數走勢關係。本研究期間為2005年1月3日至2013年8月9日，共計2116個交易日，以每日融資、融券及資券比與每日市場加權指數作為研究樣本。由門檻回歸實證結果發現在不同信用交易餘額下市場指數會對應出不同市場指數趨勢關係。研究結果顯示出較高信用交易工具餘額下產生了不同市場訊息。當融券餘額或融資餘額過高時，市場往往會產生反轉的訊息。最後，資券比指標結果也驗證相同結論，同時在較高資券比率水準下，市場指數波動程度跟著增加，隱含著市場融資相對過高情況下，市場風險產生了擴大現象。

關鍵字

個人投資者（散戶）；交易訊息；信用交易；門檻回歸模型

並列摘要

In Taiwan stock market domestic individual investors dominated daily trade volume. Due to the critical role, the research of information of domestic individual investor trading is always an interesting topic in the investment field. For this distinctive and interesting phenomenon and the gaps in previous study, this study uses three prevalent indexes as observed variables, outstanding of margin purchase, outstanding of short sale, and margin-short ratio. Based on the threshold autoregression model, furthermore, this study aims to discuss the probable trends of Taiwan stock market under different threshold values of the observed indexes. This study collected daily data of margin, short, margin-short ratio from January 3rd, 2005 to August 9th, 2013. Based on these 2116 daily data, the empirical results of the threshold regression model indicate that corresponding different estimating threshold values of observed variables, Taiwan market index could exhibit different time trend patterns. Hence, these results would infer that the credit trading indexes could convey the different contents of information depending on the values, especially the higher values, of the observed variables.