- 期刊
以時間序列頻譜分析方法探討整合與分隔市場理論對台灣股市適用性之研究
An Empirical Study of Cointegration Theory in Taiwan Stock Market-Spectral Analysis Approach
摘要
本文係嘗試應用一種屬於頻率範疇(Frequency Domain)的方法－頻譜分析(Spectral Analysis)來探討台灣股市與美、日、英、港股市間之關聯。 由於國際股市之關聯常係呈短期性或週期性的，而非長期性的平均走勢，因此用傳統的時間範疇(Time Domain)分析法－迴歸分析等，常不易精確的衡量出短期週期循環的關聯。另外，傳統迴歸分析雖有分析容易、資料需求不多等優點，但在衡量出不同序列間之領先或落後週期或不同序列間關係變動的動態性時，常滲有主觀判定的成份，至許多現象不易展現。 本文為截取二種方法之優點，除以頻譜分析探討台、美、日、英港股市關聯外，亦輔以自身相關迴歸及對數迴歸法，俾使所欲探討的現象能夠完全展現。文中係同時以股價指數及報酬率實證模型來做分析，並發現台灣股市與美、日、英、港股市皆有1~3個月左右的短期性週期循環及1~2週左右的領先落後期數，此點用傳統多元迴歸分析方法較不易衡量出來。
並列摘要
This paper attempts to apply the method of Frequency Domain attributed to Spectral Analysis to discuss the relationship of stock market among Taiwan, America, Japan, England and Hong Kong. The relationship among the international stock markets always exhibit short-term cyclical phenomenon, therefore the traditional regression analysis should not be fit-table to interpretate these relationships well. Based on this unsideration, this paper tries to combine the merits of Spectral Analysis and Regression Analysis to testify the exact relationships of stock market among the above countries. In our empirical results, we find that Taiwan, America, Japan, England, Hong kong stock market have about one-three months periodical cycles, and also have about one-two weeks leads or lags.