在全球化之投資趨勢中，藉由了解國際股市之關聯性，可作為政府制定跨國投資政策之參考，及國內投資人分散投資組合風險之依據。本文利用Longin and Solnik(2001)所提出之極端值理論，來研究臺灣與美國、日本、香港、新加坡及南韓等五個國家股市，在月報酬及週報酬下的極端值相關性。實結果顯示在月資料下，報酬為負時之相關性皆比報酬為正時來的大；並且報酬為負時，卻是逐漸下降。若以週資料為樣本時，實證結果發現不論報酬為正或負，其股價指數之相關性皆為下降之趨勢。最後，假設資料間為多元常態分配下，實證結果發，以週報酬或是月報酬率資料來看，不管股市報酬為正或為負，隨著門權值之增加，股市間之相關性反而愈小。

In today’s globalized investing environment, the government sets overseas-investing policy, and cosmetic investors decide how to diversify their portfolios, by examining the correlation among international equity market. This paper uses the extreme value theory proposed by Longin and Solnik (2001) to investigate the similarities between the international and Taiwan stock markets. Five national stock markets (those of Taiwan, the United States, Japan, Hong Kong, Singapore and South Korea) are used to test their extreme-returns relationships. The empirical results show that, in monthly data, the correlation of extreme negative returns among the five countries is greater than the correlation of extreme positive returns. When extreme returns are negative, the stock price index correlation rises when the absolute value of the correlation threshold increases; when extreme returns are positive, the stock price index correlation drops when the threshold value decreases. On the contrary, in weekly data, the empirical results indicate that the correlation of extreme returns among the five countries decreases regardless of whether the returns are negative or positive. Finally, if we assume that the sample data are in a multivariate normal distribution, no matter whether the data is weekly or monthly, or the returns are negative or positive, when the absolute value o the correlation threshold increases, the correlations tend to decrease.