多國性股價報酬率的統計特性及星期效果研究－自相關條件異質性模型之應用

The Statistical Properties and Weekday Effects of the Multinational Stock Returns-An Application of the ARCH Model

黃柏農(Bwo-Nung Huang)

《中國財務學刊》 2卷2期 (1995/01) Pp. 43-76

https://doi.org/10.6545/JoFS.1995.2(2).3

摘要

本文以修正簡單迴歸的星期效果模型當中所具有的序列相關及自相關條件異質等問題之後，以日股價報酬率在跨越1980年至1992年的研究區間，及針對德國、香港、英國、臺灣、美國及日本股市進行週天效應之研究，研究結果發現：倫敦及臺北股市存在顯著的負週末效應（week end effects），此外，紐約及東京股市的週二及週三亦具有相當顯著的正效果。而東京的週六及倫敦的週五之報酬率亦相當顯著。一般而言，修正序列相關與異質現象後之t統計量會變得較顯著，此外如果股市具有較明顯的異質現象時，修正前後的實証結果會有很大的差異。

並列摘要

In this paper, we correct the problems of serial correlation and autoregressive conditional heteroscedasticity in the simple regression weekday effect model, then use daily stock ret urns spanning from 1980 to 1992 to exp lore the weekday effect of the Germany, Hong Kong, British, Taiwan, U.S. and Japan stock markets. As a result, we find that t here exists significant negative weekend effects in the London and Taipei stock markets. Yet it is significantly positive in the Tuesdays and Wednesdays of the New York and Tokyo market s. In addition, the stock returns of the Saturdays and Fridays in Tokyo and London markets respectively are also quite significant. Generally speaking, the t statistic after correcting the serial correlation and heteroscedasticity would become more significant. However, a great difference would appear between the empirical results before and after modification if stock markets are patently heteroscedastic.