台灣經濟成長率之混合頻率預測-MIDAS迴歸應用

NCUIR > School of Management at National Central University > Graduate Institute of Economics > Electronic Thesis & Dissertation > Item 987654321/68654

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/68654

Title:	台灣經濟成長率之混合頻率預測-MIDAS迴歸應用
Authors:	涂育嘉;Tu,Yu-Jia
Contributors:	經濟學系
Keywords:	經濟成長率預測;混合頻率資料模型;因子模型;Forecasting economic growth rate;MIDAS regression model;Factor model
Date:	2015-07-14
Issue Date:	2015-09-23 13:57:56 (UTC+8)
Publisher:	國立中央大學
Abstract:	本文以Ghysels et al. (2004) 所提出的MIDAS 迴歸模型(Mixed Data Sampling)對台灣經濟成長率進行混合頻率預測, 主要分為兩個部分: 第一, 使用領先與同時指標構成項目, 進行單變量迴歸、多變量迴歸及組合預測。第二, 使用能處理大量資料的因子模型, 並將資料區分成實質面與金融面變數。透過樣本外預測結果發現, MIDAS迴歸模型有較佳的預測績效, 且顯示此方法具有一定的準確性與即時性, 也發現金融相關資料能提供好的預測能力。而台灣尚未運用MIDAS 迴歸模型在總體的經濟成長率預測, 本文算是一個新的嘗試, 提供往後對總體經濟預測方面, 若存在資料頻率不同時,有其他的處理方式。;This paper aims to construct a mixed-frequency model based on MIDAS regression, which was proposed by Ghysels et al. (2014), to forecast Taiwan economic growth rate. There are two sections in this paper. The first one is the use of leading indicator and coincident indicator in Univariate regression, Multivariate regression and Forecast combination. Another part is that we adopt Factor model to analyze the respective factors between various macroeconomic variables. In this estimator, we distinguish variables into financial variables and non-financial variables. The results of pseudo out-of-sample show that the MIDAS regression models perform better and have the certain degree of accuracy and timeliness. Furthermore, we find financial variables can provide a good prediction power. Although the MIDAS regression models are not widely used in prediction of Taiwan economic growth rate, we try to make a new attempt on it and it can also be a good alternative in macroeconomic forecast with different frequencies data.
Appears in Collections:	[Graduate Institute of Economics] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	729	View/Open

社群 sharing

Loading...