Time-varying Relationship of Price Indexes in Taiwan

物價指數衡量經濟中不同個體或部門問的交易條件，為總體經濟計量模型中重要的部份。例如，消費者物價指數(CPI)衡量生活成本，躉售物價指數(WPI)則為生產成本指數。經由複雜的經濟活動與統計編製原則，這些物價指數間存在高度的相關性，而在一般總體經濟模型中，往往假設這種關係是穩定的。本文旨在探討臺灣的物價指數間是否真正存在穩定的關係。我們應用指數加權遞迥最小平方法與參數拔靴法等計量方法對臺灣的資料作實證分析。實證結果發現在調整匯率的影響後，出口物價指數的函數相對穩定，但CPI與WPI的方程式則為不穩定。本研究提醒我們在建立臺灣總體經濟模型時，應特別注意物價間不穩定的關係，可攷慮依時變動模型或非線性模型。

關鍵字

依時變動；臺灣物價指數；指數權數遞迴最小平方法；參數拔靴法；穩定法

並列摘要

Price indexes are important variables in macroeconomic models. They are composed to measure various costs in an economy. For instance, a Consumer Price Index (CPI) is constructed to measure the cost of living and a Wholesale Price Index (WPI) summarizes the prices of transactions between enterprises and is intended as a measure of production cost. The price indexes of an economy are highly correlated via economic theory and index composition. The relationship is often treated as stable over time in macroeconomic modeling. In this paper we consider four monthly price indexes of Taiwan, namely the CPI, WPI, Export Price Index (PEX), and Import Price Index (PM). Our goal is to investigate the stability of the relationship between these indexes. We employ various methods, including exponentially weighted recursive least squares and parametric bootstrap methods, to show that, for Taiwan is Economy, the relationship is time-varying for the CPI and WPI. For the export price index (PEX), we found that the relationship is relatively stable after adjusting for the impact of exchange rates between the U.S. Dollar and Japanese Yen versus New Taiwan Dollar. Macroeconomic modelers should be cautioned about the time-varying relationship of price variables, and nonlinear models deserve further investigation.

並列關鍵字

Time-varying ； Taiwan price indexes ； Exponential weighting into the past ； Parameter bootstrapping ； Stability

參考文獻

Tiao, G. C.,G. E. P. Box(1981).Modeling multiple time series with applications.Journal of the American Statistical Association.76(376),802-816.

Google Scholar

Tiao, G. C.,R. S. Tsay(1983).Multiple Time SeriesModeling and Extended Sample Cross Correlations.Journal of Business and Economic Statistics.1(1),43-56.