| Summary: | 碩士 === 國立中央大學 === 經濟學研究所 === 99 === This study uses dynamic factor analysis proposed by Stock and Watson (2002a) to forecast inflation in Japan. This method summarizes large amounts of economic information by a few estimated factors, and uses them as predictors to construct the forecasts. The objective is to investigate whether CPI subcomponents play an important role in estimating factors to obtain better predictors to improve forecasting performance. The dataset composed of 46 macroeconomic series and 103 CPI subcomponents from 1986 to 2009.
The results indicate that the subcomponent-included factor model outperforms the benchmark autoregressive model at the one, two, and four quarter ahead horizons and the factor model, which does not include the disaggregated CPI components, resulting in a deterioration of forecasting performance at the one, two, and four quarter ahead horizons. The results support the argument that using information from CPI subcomponents contributes to substantial improvements in accurate inflation forecasting.
|
|---|
