本文採用Lo and MacKinlay(1988、1989)之傳統變異比率檢定法,及Wright(2000)的無母數變異比率檢定法--秩與符號,探討美國股票、外匯、原油與黃金四個主要市場的市場效率性。由於過去的文獻在探討市場效率時,普遍未嚴謹考慮資產價格是否存在序列相關及異質變異數之現象,而造成檢定效力不彰、結果不一致的問題。 本研究改善過去之不足,除採用Lo and MacKinlay(1988、1989)所提之傳統變異比率檢定,並以Wright(2000)所提出的無母數變異比率檢定--秩與符號,考量資產價格存在異質變異數下進行檢定,並根據各個市場所發生的重大事件、制度改變、政府管制等改變,進行次樣本分析,以避免資料在全樣本期間下可能會受到時間移轉的影響,而抵消掉不具效率性的因素及效果,使其提升評估市場效率之檢定力。實證結果顯示,在制度改變或政府管制等改變之後的次樣本期間,外匯市場--英鎊對美元之即期匯率及黃金市場--紐約商品交易所之黃金現貨價格,具有效率性。而原油現貨及S&P500股票市場,在制度改變或政府管制等改變之後的次樣本期間,在不同觀察值間隔(K)下,均不具效率性。 In this study, we apply both traditional variance ratio test proposed by Lo and MacKinlay(1988、1989)and nonparametric variance ratio test proposed by Wright(2000), to investigate the efficiency of American stock, foreign exchange, oil, and gold markets. Reviews of the past papers show that the test results, due to the ignorance of the serial correlation and heteroscedasticity, are mostly inconsistence in testing market efficiency. We improve the results by using both variance ratio test proposed by Lo and MacKinlay(1988、1989)and non-parametric variance ratio test proposed by Wright(2000), under the consideration of heteroscedasticity. Dividing the whole sample period into two subperiods according to important event, rule change and government regulation also vastly improve the test result. The test experiments in this paper show that both the exchange rate market for Pound to US Dollar and the gold market are efficient in the subperiods after rule change and after government regulation. However, both oil market and S&P500 stock market are inefficient under various observation durations (K) in the subperiods after rule change and after government regulation.