本研究是針對臺灣房價指數來作預測，並以季資料為分析頻率；研究期間為2001年第1季至2014年第4季共56個樣本值。採ADF單根檢定，移動性Chow檢定，TAR及AR-GARCH模型等計量方法，來探討本文之研究目的。經實證結果發現：臺灣房價指數為一階整合之非定態數列，且存在一時間結構轉變點（為2009年第1季）。在結構轉變點以前，落遲1期房價對本期房價的影響有0.517723；而在此點之後，落遲1期房價對本期房價的影響調升為0.657938。再就動態與靜態預測評估下，AR（1）-GARCH（1,1）是比TAR（1）模型有較佳之預測力。且此兩模型均顯示：未來三年（2015至2017年），臺灣房價有下修趨勢，然跌幅會逐漸減緩。

This study focuses on Taiwan housing price index to make forecast, and it uses quarter frequency data to analysis, the study period is from the first quarter of 2001 to the fourth quarter of 2014, there are 56 samples in the period. Using ADF unit, moving Chow test, TAR and AR-GARCH models for the purpose of this research. The empirical results show that: Taiwan housing price index is a non-steady-state first-order integration of the time series number, and there is a structural change point (for the first quarter of 2009). The lag one period of housing price has impact on the current price before the structure change point, and the value is 0.517723. The effect on the current housing price after the structure change point is raising to 0.657938. Viewing on dynamic and static forecasts, AR (1) -GARCH (1,1) model has better prediting power than TAR(1) model. Also the two models show: the next three years (from 2015 to 2017), Taiwan housing price has a down trend, but the decline ratio will gradually slow down.