台灣景氣指標長期趨勢估計法之研析

A Study on the Method of Estimating the Long-Term Trend of Taiwan's Business Indicators

台灣係採用「成長循環」（growth cycles）概念衡量景氣波動，因此在編製景氣指標時，一個非常關鍵的問題在於：如何穩定估計各構成項目長期趨勢並加以剔除。2007年7月，本會參考OECD作法修訂景氣指標，當時以PAT/MCD法剔除構成項目長期趨勢並平滑化；然實務上，每新增一個月份資料，構成項目的長期趨勢都會修正，連帶使構成項目循環波動成分發生改變。為評估何種剔除長期趨勢的方法較適合我國採用，本研究挑選「實質貨幣總計數M1B」、「股價指數」、「核發建照面積」、「工業生產指數」、「批發零售及餐飲業營業額指數」、「非農業部門就業人數」、「失業率（倒數）」、「製造業單位產出勞動成本指數」、「金融業隔夜拆款利率」等9項景氣指標構成項目模擬測試，比較PAT/MCD與Double HP-filter兩種剔除長期趨勢方法，結果顯示Double HP-filter法估計之循環波動穩定性較高。此外，本研究亦發現，採用Double HP-filter法編製而得之景氣指標，可大幅降低不規則項（irregular）干擾，有利轉折點的判定。

Taiwan uses the concept of 'growth cycles' to measure economic fluctuations. Hence, in compiling economic indicators, a crucial question is how to estimate the long-term trend of each component, and then remove it. In July 2007, drawing reference from OECD practice, the Council for Economic Planning and Development (CEPD) used the combined PAT/MCD approach to remove the long-term trend and carry out smoothing. However, with this method, the addition of new data each month causes change in the long-term timing of the time series, changing the long-term trend values, and hence resulting in change to the cyclic fluctuations after removal of the long-term trends. To assess which method of removing the long-term trend is more suitable for Taiwan, this study selected nine economic indicators for simulation and comparison of the PAT/MCD and Double HP-filter methods. The results showed that the Double HP-filter method yielded higher stability of cyclic fluctuations. In addition, the study found that using the Double HP-filter approach could significantly reduce irregular interferences, aiding the identification of turning points.