若我們可以認定預測房地產景氣循環的高峰和谷底形成的所在位置，則這個訊息就可以提供給決策者和投資大眾一個相當重要的參考指標。本研究選用包含了投資面、生產面、交易面與使用面房地產活動的台灣房地產景氣綜合領先指標，運用雙變量馬可夫轉換自我迴歸模型MS-ARX進行房地產景氣循環轉折點之認定，最適模型為截距項、變異數皆受到不可觀察變數所影響且遞延（lag）期數為8季的L（1）–MSIH（2）–ARX（8）模型。該模式確認了房地產景氣綜合領先指標領先基準循環一季。整體而言，綜合領先指標模型對於房地產研究中心所發布的四次谷底與三次高峰皆有相當接近的認定。 Projecting the peak and trough of a real estate cycle could help decision makers and investors. In this study, we adopt the Composite Leading Index of Taiwan’s real estate cycle which includes the investment, product, transaction and use phase of the real estate life cycle. We use bivariate Markov-switching autoregressive model to identify the turning points of real estate cycle. The model with the best fit is L(1)- MSIH(2)-ARX(8), of which, the lags is 8 and both the intercepts and variance are regime-dependent. The predicted dates of the turning points from the L(1)- MSIH(2)-ARX(8) model are so consistent with the peaks and troughs of the reference cycle announced by the Taiwan Real Estate Research Center that the Leading Index can present its forecast function. However, we use L(1)- MSIH(2)-ARX(8) model to identify that the Leading Index leads the reference cycle by one season. Generally speaking, L(1)-MSIH(2)-ARX(8) produced rather accurate results in terms of identifying the turning points of real estate cycles in Taiwan.
臺灣土地研究, 11(2), 155-177 Journal of Taiwan land research