以總體、廠商及高頻資料所進行之經濟預測

Abstract

在本文中，我們評估總體資料、廠商資料和高頻資料能否幫助預測美國工業生產指數和通貨膨脹，並藉由動態因子模型和因子混頻抽樣迴歸模型(MIDAS)進行實證研究。研究結果顯示，除了廣泛使用於預測工業生產指數和通貨膨脹的總體資料外，廠商以及高頻資料可能也包含有助於長期預測的訊息。

In this thesis, we assess the performance of a large-dimensional set of macro-level, firm-level and daily predictors in forecasting the industrial production and inflation of the U.S. We base this empirical study on the dynamic factor model and the factor mixed data sampling regression (MIDAS). The empirical study shows that the firm-level and high-frequency predictors may contain useful information in addition to the widely used macro-level predictors in the long-term forecast of the industrial production and inflation.