在經建會所編制的領先指標中同時包含了金融面及實質面指標，金融面變數與實質面變數的動態行為雖然大部分時期相同，但是有可能即使在相同的景氣狀態下兩者依然可能會有不同走勢的可能性。因此若我們不區分金融面指標及實質面指標可能有不同走勢的可能性，而將所有的變數混合在一起以傳統的單因子模型進行估計並預測景氣轉折點，有可能得到不理想的結果。本文的目的之一就是驗證上述推論的正確性，且將實證結果與臺灣經濟發展的歷程作一對比，我們更嘗試修正實證模型，希望能夠改善上述單因子模型之缺失。實證結果顯示除了印證若把所有的金融及實質指標置於馬可夫轉換單因子模型中進行推估，的確會產生景氣狀態預測失誤的情況。金融變數在認定臺灣景氣轉折點的日期與經建會所公告的日期較為一致；而實質變數所認定出的景氣轉折點日期與經建會所公告的日期則較不相同。根據樣本外的預測結果，本文中所嘗試的修正模型能夠改善臺灣景氣轉折點的預測。

The compostic leading indicator, compiled by the Council for Economic Planning and Department, consists of two sets of indicators: the real and financial leading indicators. This paper argues that the use of all leading indicators simultaneously may mix two different sets of information and thus provide a less accurate prediction of a future recession. We divide Taiwan's six leading indicators into two different sectors, real and financial sectors, and show that the two sectors may reveal different information. We use the Markov switching factor model to extract the common factor for each sector and find that the predicted recessions based on the two sectors are different in three periods. Finally, using financial variables seems to outperform the real variables in predicting a future recession.