最近幾年應用Hamilton(1989)提出的馬可夫轉換模型來探討臺灣景氣循環特性的文獻，所認定出的影氣循環轉折點往往與經建會公佈的日期不一致，尤其是在1990年代之後。本文嘗試解決此一問題。我們發現，即使利用多變量動態馬可夫轉換單因子模型，仍無法準確認定出1990年代後期臺灣的景氣循環轉折點，其主原因是：模型中所假設的各個重要總體經濟變數，如國內生產毛額、消費、出品及進口，之共同因子發生變化，這肇因於臺灣經濟結構改變。而本文的實證結果並進一步發現，應用單變量馬可夫轉換單因子模型，以實質固定資本形成毛額的資料進行估計，能夠準確地捕捉臺灣在1990年代後期的景氣循環轉折點。

　　　　　This paper investigates the business cycle characteristics in Taiwan. We want to resolve the identification puzzle of Taiwan's business chronologies, and adopt two alternative approach to accomplish this job. First, we revise Hamilton's(1989) Markov-switching model to search another better one. Second, since previous empirical studies have witnessed the failure of GDP in capturing Taiwan's business cycles , we replace GDP with other important macroeconomic series to capture Taiwan's business cycle fluctuations. We show that the univariate Markov-switching factor model in dating business chronologies will be more consistent with the CEPD-defined business chronologies than that of the multivariate case. The key point is the common factor of various marcoeconomic series no longer hold for the post-1990s is Taiwan. However, the univariate factor model with regime-switching in terms of real gross domestic fixed capital formation still preserve the ability to track Taiwan's business fluctuations even in the late-1990s.