高頻率股市報酬率之日內循環及波動性之研究

Abstract

本文採用一分鐘高頻率股市報酬率資料為研究標的，企圖尋找，台灣加權股市每分鐘之報酬率，在加入日內循環因子後，是否對報酬率波動性（volatility）有更好的估計性，其次本研究亦探討台灣股市每分鐘報酬率是否有非對稱性效果，亦即報酬率殘差的正負項對波動性有不一樣的影響，最後本研究將資料分成三區段，探討開盤後三十分鐘，開盤後三十分鐘至收盤前三十分鐘，收盤前三十分鐘，探討這三區段對報酬率波動的影響，其結果如下： （一） 每分鐘報酬率所做的AR(1,2,3)-GARCH(1,1)模式，我們發現，每分鐘資料本期報酬率與前三期有顯著關係，在波動性方面ARCH（或GARCH）效應很顯著，在衡量非對稱模型上EGARCH(1,1)及SGARCH(1,1)上，都明顯未發現有非對稱性效果。 （二） 尋找日內循環中，我們發現每分鐘的報酬率資料，很難有日內循環的發生，或是其效果與原始資料所做的模式估計無差異。 （三） 所做的三區段的波動性循環中發現，除開盤後三十分鐘不顯著外，其餘兩時段均顯著且為正向關係，尤其是收盤前三十分鐘更為顯著。

Recently, the research on the volatility of the high frequency returns has been very popular and shown a strong impact on intraday periodicity. In Taiwan, most of the relevant researches did not take intraday periodic patterns for consideration. This study adopts deterministic filter model combining trigonometric functions with polynomial terms and attempts to find the intraday periodicity. The first purpose is to discuss whether the results of the estimation of the filter data incorporating periodic factor are better than raw data. The second is to use asymmetric model to find out whether there is asymmetric effect in the high-frequency stock returns in Taiwan. The Third is to separate high frequency data into three parts to understand whether the return volatility is significantly different in each part. The conclusions are as follows: First, we found that the result of estimation through filter data is not better than the result of estimation through raw data. Secondly, there is no asymmetric effect either in filter data or raw data. Third, the volatility presented in each of the three parts is significantly different