物價指數是經濟預測的重要指標,及早預知消費者物價指數走勢將有助於對價格震盪提出因應對策,惟目前有關消費者物價指數的研究大多內含於總體經濟模型,少有單就消費者物價之走勢進行深入研究,本研究遂蒐集1991年1月至2003年12月計156筆臺灣地區消費者物價指數及其七大基本分類指數進行分析,運用時間數列分析方法,採單變量ARIMA模式、轉換函數模式及多變量VARMA模式建構總指數及七大基本分類指數預測模式,並以樣本內預測來檢測模式之合適性及樣本外預測評估各預測模式預測效力。實證結果發現,以各七大基本分類指數配適各種時間數列模式,再依實務編算方法重新求算而得之總指數預測效果較佳,其中,食物類、居住類、交通類及教養娛樂類指數採多變量VARMA模式:衣著類、雜項類指數用單變量ARIMA模式擬合;醫藥保健類指數使用介入模式最爲理想。
Consumer price index is an important indicator for economic forecast. Exploring the trend of the consumer price index would provide much help for policy making to deal with the vibration of the price. Therefore, the purpose of this article is to establish a better model for exploring the changing consumer price indices iii Taiwan. In this research 156 monthly data, from January 1991 to December 2003, including the general consumer price indices and its seven basic group indices are collected for estimating the models. The univariate ARIMA model, the transfer function model, and the multivariate VARMA model arc used to analyze the time series data. The in-sample data arc used to evaluate the suitability of those forecast models, and the out-sample data are used to measure the forecast errors. Study results show that the suitable model for the general consumer price index in Taiwan would he the model integrated by the seven basic group consumer price indices with the weighted average, where the food, housing, transportation & communication and education & entertainment categories are fitted by the VARMA model, the clothing and miscellaneous indices are fitted by the ARIMA model, and the medicines & medical care indices are fitted by the intervention model, respectively.