裁至目前為止經建會都尚未公佈台灣第十次完整的景氣循環高峰及谷底日期,而且在2003年年初,台灣景氣似乎又有下跌的趨勢而形成所謂的「雙重衰退」。根據經建會所編製之「綜合領先指標」資料,我們發現台灣綜合領先指標領先台灣景氣狀況的「最適」期間應該為3個月,符合綜合領光指標應該領先景氣狀態3至6個月的預期。Probit模型認定出的高峰及谷底轉折點日期與經建會所定義的「台灣景氣循環基準日期」誤差較大,而且會產生誤認的情況;相反的,根據Probit馬可夫轉換模型的當期機率所認定出的高峰及谷底轉折點日期與經建會所定義的「台灣景氣循環基準日期」幾乎完全相同。根據本文的實證結果,我們認為台灣第十次的景氣循環的確已經結束,而且台灣並沒有出現所謂的「雙重衰退」,印證了Chen (2003)所得到的預測結果。
Using Taiwan data, we employ the Probit-Markov Switching (Probit-MS) model to identify business cycle turning points. The merit of the Probit-MS model is that it incorporates the dependent structures of the leading indicator, which is not taken consideration in the traditional Probit model. It is unambiguous that the best forecast horizon for Taiwan's leading indicator in predicting the business condition is three months. The turning points identified by the Probit-Markov switching model are greatly enhanced when compared with those from the Probit model, and beyond this, the model-identified business cycle dates are highly consistent with the officially identified turning points. Our empirical results provide further evidence in favor of Chen's (2003) predictions.