台灣房地產景氣循環週期之研究：頻譜分析之應用

Abstract

本文研究方法迴異現有探討台灣房地產景氣循環的文獻，補足迄今分析台灣房地產景氣變化，嘗試應用頻譜分析(Spectral Analysis)，將台灣2001年至2016年間各樣本變數由時域(time domain)轉換成頻域(frequency domain)，並在頻域中分析出組合所有頻率的頻譜，找出各樣本變數中影響性較大的頻率後，再篩選符合之樣本變數，將這些影響性較大的各頻率進行樣本變數間的頻譜連動性(Spectral Coherence)分析，並以此分析結果來度量兩時間序列的連鎖效應(ripple effect)大小，最後加以整理並找出頻譜主頻率(Major Cycle)來作為本文之研究結果。 實證研究結果發現，台灣2002年至2016年間，週期以1~2年(第16頻)為主，顯示台灣房地產波動較為快速，與以往文獻所分析出的週期較不相同，另外過去所認定與房地產景氣循環較具有相關性的樣本變數，如房價所得比、房貸負擔率、租金指數等，其頻率與其他變數較不相同外，頻譜連動性亦不高，可能非為房地產景氣循環之構成要素之一。

This study investigates in the first time the real estate trade cycle in Taiwan with the Spectrum Analysis. We use the discrete Fourier Transform to transform every sample data from the time domain into the frequency domain, and the monthly sample data cover the period from 2002 to 2016 in Taiwan. All harmonic amplitude of each sample datum is separated individually. Ten most influenced frequencies among all sample data are found. We use the Spectral Coherence to estimate the ripple effect, and find the Major Cycle in Taiwan’s real estate market. Our major findings show that the dominant cycle of the Taiwan’s real estate market is around 1 to 2 years from 2002 to 2016. It means that Taiwan’s real estate has highly frequent Major Cycle. It’s different from those of previous literature. On the other hand, some of related sample data, such as House Price Income Ratio, Housing Loan Burden Rate, and Rent Index, etc., do not show the higher Spectral Coherence. It indicates that these variables do not seem to be composite factors of Taiwan’s real estate cycle.