房地產景氣特性之再確認──多變量馬可夫轉換之應用

Reconfirmation of the Turning Points of Real Estate Cycle Indicators in the Taiwan Real Estate Market-An Application of Markov-Switching Vector Auto-Regression Models

探討房地產景氣的變化一向是房地產領域關注的重要議題，精確而有效的模型不僅有助於瞭解房地產與總體經濟變數波動特性的差異，更有利於相關房地產政策制訂時的依據。在研究景氣變化的計量模型中，馬可夫轉換模型是較常被利用來進行此一研究的工具，然現階段的實證結果卻無法完全令人滿意，主要原因是單變數的馬可夫轉換模型無法完全反應房地產景氣變化的特性所致。本文則嘗試採用多變量的馬可夫轉換模型，直接對構成房地產指數的多個內容變數同時進行分析，藉由多變量的架構連結變數間的相互關係來確認景氣轉折與持續期間。由本文實證結果我們將發現，不論是領先指標或基準循環指標下掌握轉折的能力有顯著改善外，更重要的是，對於房地產景氣的擴張期與收縮期的特性捕捉，多變量馬可夫模型更能顯示其優越之處，此一實證結果將可有效提供決策當局制訂房地產政策時的參考。

How to precisely confirm the features of real estate cycles has always been an important issue in the real estate research field. An unerring description of a business cycle outline is not just a helpful tool for the authority that tries to formulate policies in response to future economic change, but it also provides a useful message for people who are currently facing buying/renting decisions. Unfortunately, we are still lacking a good theoretical model that can support us when making such decisions. While a one-variable Markov switching model performs a good prediction in terms of capturing the turning point in a business cycle, it still fails to estimate the cycle's duration, a problem that has not been overcome. This research attempts to employ a multi-structure Markov switching model to revise the duration problem. Based on the confirmation purpose, we compare several settings of Markov switching structures, and adopt new measurements to evaluate each kind of model. Besides, we simultaneously test the performance of leading indicators and reference series. The empirical results show that, when the Markov-switching vector auto-regression model is used, the ability to estimate the cycle's duration really outperforms that of the single series Markov switching model.