臺灣景氣指標季節調整方法之研析

A Study of the Method of Seasonal Adjustment of Taiwan Business Indicators

美國商務部普查局（Census Bureau）已在2013年1月正式發布新季節調整程式（X-13 ARIMA-SEATS），本文嘗試應用X-13 ARIMA-SEATS進行研究，並以同時指標及對策信號中的「機械及電機設備進口值」為例，結合該數列日資料及X-13ARIMA-SEATS程式，重新建構季節調整模型。透過上述模型建構過程及結果，期能對目前景氣指標季節調整工作提出具體建議。研究首先利用「機械及電機設備進口值」日資料自行推估每月交易日效果，其次應用X-13ARIMA-SEATS軟體建構regARIMA模型並進行季節調整。其中，除了處理華人地區特有的春節效應，更採用軟體新增的SEATS季節調整方法，藉以比較X-11與SEATS兩者季節調整結果差異。本文主要實證結果如下：一、觀察歷年平均週一至週日之機械及電機設備進口值，與直接採用X-13ARIMA-SEATS所推估之交易日效果並不相符，而根據該數列日資料所推估之交易日因子與實際平均統計結果則較為相近。二、本數列的春節前、中、後因子解釋變數均為顯著，顯示農曆春節對國內企業進口機械電機設備的活動確實存在影響。三、透過診斷檢定分析，SEATS季節調整方式的診斷結果普遍較X-11理想。此外，不論是水準值或年增率，透過本研究所建構的模型，季節調整後數列均較原始數列平滑，季節性波動已大致去除。

The United States Census Bureau released X-13ARIMA-SEATS seasonal adjustment program in January 2013. This paper attempts to apply the program. In the process of this empirical study, first, daily data of imports of machinery and electrical equipment are used to compute monthly trading day factors. Second, the X-13ARIMA-SEATS program is employed to construct the regARIMA model. In the model, regressors are used to estimate the Chinese New Year holiday effect, and then X-11 and SEATS seasonal adjustments methods are used, respectively. Finally, the two adjustment methods are compared by a common set of diagnostics. The main empirical results are as follows: 1. The trading day factors computed from daily imports of machinery and electrical equipment data are consistent with the historical average on Monday to Sunday. 2. The three regressors of before, during and after Chinese New Year holiday are all significant, reflecting that import of machinery and electrical equipment were actually affected by the holiday. 3. The diagnostic results from SEATS seasonal adjustment are generally better than X-11 adjustment. In addition, through the model constructed for this study, for both original value and annual growth rate of import of machinery and electrical equipment, the seasonally adjusted series are smoother than unadjusted series.