台股漲跌停時距之持續性研究

Abstract

由於財務上常以固定的時間間隔做為資料分析的基礎，而這種交易資料的分析可能會造成市場上部分的訊息流入沒有被充分衡量，為解決該項缺失，本研究由此切入，利用Engle 和 Russell（1998）的自我相關條件時距（autoregressive conditional duration；ACD）模型，估計在台股漲跌停限制下，探討台灣股票市場漲跌停時距之間是否存在自我相關，以及過去的漲跌停時距與期望漲跌停時距和未來的期望漲跌停時距之間是否分別存在顯著關係，並對未來發生漲跌停的交易時距進行合理的預測，甚至更進一步分別探討漲停時距、跌停時距是否存在自我相關；本文亦承襲 Kim 和 Limpaphayom（2000）所使用影響股票容易達漲跌停的五個變數並進而探討這些變數對ACD模型中的係數以及平均漲跌停時距是否具有顯著的解釋能力，實證結果發現，本研究對ACD模型中的α、β係數的特性補捉尚不完善，但對平均漲跌停時距較具有顯著的解釋能力。

Financial transaction data inherently arrive in irregular time intervals, while standard econometric techniques are based on fixed time interval analysis. The choice of a fixed interval for financial data analysis is very perilous as it may leave the investigator with many uninformative points, or disguise the periods of most interest. In order to avoid these problems, we use data from stocks listed in Taiwan Stock Exchange and apply the Autoregressive Conditional Duration（ACD）model that seem well suited for limit-hit duration. Kim and Limpaphayom（2000）who empirically investigate the determinants of limits-hit frequency and find that stock volatile stocks, actively traded socks, and small market capitalization stocks hit price more often than other stocks. We apply the parameters to examine whether these variables have significant explanatory power on coefficient of ACD model and mean limit-hit duration.