本研究由設定檢驗的觀點,評估行政院主計處的一季前經濟成長率預測表現, 並提供可能的預測更新模型。實證顯示前述預測的誤差項具有顯著的一階自我相關結構,並與部分總體擴散指標之落後項相關。亦即, 此預測誤差具有可再預測性, 故主計處的預測表現未臻預測均方差極小化之最適結果。依據檢定所得出的訊息, 我們設計出數個預測更新模型。其中最簡單的模型, 僅需主計處上一期的預測誤差, 即可對其當期預測值進行即時的更新。預測更新模型的樣本內、外表現,皆優於主計處原有的一季前預測,亦明顯優於未納入主計處預測的擴散指標模型。

In this empirical study, we evaluate the one-quarter-ahead economic-growthrate forecasting performance of Taiwan’s Directorate General of Budget,Accounting and Statistics (DGBAS) and provide certain forecast updating models to improve the forecasting performance. Specifically, the DGBAS forecasting error sequence has significant autocorrelation of order one and is correlated with the macroeconomic diffusion indexes. This motivates us to propose a set of models for updating the DGBAS forecast. Among these models, the simplest one needs only the current DGBAS forecast and the previous DGBAS forecasting error, and the best model extends this simple model by including certain properly selected diffusion indexes as explanatory variables. The updated models considerably outperform the DGBAS forecast and the pure diffusion index models (that do not account for the DGBAS forecast) in both the in-sample and out-of-sample comparisons.