臺灣民眾通膨預期調查資料之分析

AN EMPIRICAL ANALYSIS OF CONSUMER INFLATION EXPECTATIONS IN TAIWAN

臺灣通膨預期之相關實證研究多根據國外專家調查之通膨預期資料，據作者所知，過去的文獻尚未有以民眾調查資料分析臺灣通膨預期的研究。我國公開民眾通膨預期調查的機構包括中央大學台灣經濟發展研究中心與國泰金控，皆為每月公布一次；央行經濟研究處資金流量科每季針對投信進行之產業景氣意向調查，亦包括通膨預期。國泰金控與中央大學的調查結果係民眾對未來物價走勢的看法，請受訪民眾回答其心目中未來物價水準可能上升、下降或持平。本文參照文獻上的作法，量化國泰金控與中央大學的調查資料，計算民眾心目中通膨預期的數值，並進行實證分析。過去量化通膨預期的文獻多採用全樣本的資料計算各月份的通膨預期，然而，政策制訂者必須以每月現有的資料計算通膨預期，本文有別於傳統文獻作法，每月更新通膨預期係採用樣本期初至該月份的所有的資料，以遞迴法計算各月之通膨預期值。本文實證結果之主要兩點結論如下：第一，若以國泰金控與中央大學調查資料量化所得之通膨預期，預測未來六個月的通膨，其預測績效優於傳統之ARMA模型，其中國泰金控超越的程度達到統計上之顯著水準，且量化後之通膨預期，其預測組合可進一步提升預測績效；第二，主計總處每季發布下一、二季通膨之預測時，若以國泰金控之通膨預期資料轉換成下一季之通膨預測，其預測精準度略優於主計總處；若以國泰金控、中央大學之通膨預期資料轉換成下二季之通膨預測，其預測精準度優於主計總處。

We present an analysis of inflation expectations in Taiwan derived from the analysis of data from surveys of households. To our knowledge, this is the first paper to analyze Taiwan's consumer inflation expectations. The surveys are conducted by the Research Center for Taiwan Economic Development of National Central University (NCU) and by Cathay Financial Holdings (Cathay). We modify the seminal methods of Carlson and Parkin (1975) and Batchelor and Orr (1988) and quantify their qualitative survey data. We depart from the assumption of a constant just noticeable difference (j.n.d.), under which the determination of δ requires full sample information. Instead we calculate the time-varying value of δ based on all data available up to the time of interest. As a consequence, our quantitative inflation expectations constructed in this paper are ex ante, as opposed to the ex post inflation expectations in previous works. Two conclusions emerge. The first is that the quantified household inflation expectations derived from the NCU and from the Cathay survey data in both cases outperform two benchmark models-the Autoregressive Moving Average (ARMA) model and the random-walk mode – in forecasting the six-month-ahead inflation. In particular, the forecasting accuracy of the quantified data derived from the Cathay survey performs statistically significantly better than the benchmark models. Our second conclusion is that our inflation forecasts for one or two quarters ahead outperform the DGBAS counterparts published in their press releases for GDP preliminary estimates.