| 摘要: | 本研究利用因子模型結合傳統基本面模型(包含購買力評價模型、泰勒法則模型、貨幣模型以及未拋補利率評價模型), 利用樣本外預測之方式與拔靴法檢定, 發現基本面加入因子之模型能降低傳統基本面模型對匯率預測之RMSPE。相較於其他基本面加入因子之模型, 購買力評價結合因子模型之表現最佳。另外, 本研究檢視5 個範圍期間(h = 1, 4, 6, 8, 12), 並利用各種不同基本面加入因子模型, 在去除短期範圍期間(h = 1) 之下, 13個目標國家中, 加拿大、丹麥、歐盟、挪威、紐西蘭、韓國、新加坡以及英國, 皆曾出現過全數顯著之表現。最後, 若僅將範圍期間專注在中長期(h = 4, 6), 除了貨幣加入因子模型之外, 多數國家的基本面加入因子模型對匯率之預測表現優於隨機漫步模型。
;In this study, we combined factor models with traditional fundamental models (including purchasing power parity model, Taylor rule model, monetary model, uncovered interest parity model), using out of sample forecast and bootstrap test to find that RMSPE of exchange rate forecast for factors combined with fundamental models lower than traditional fundamental models. In particular, factors combined with purchasing power parity model outperform other factors combined with fundamental models. Moreover, this study examined five horizons (h = 1, 4, 6, 8, 12). Out of short horizon (h = 1), there are Canada, Denmark, Euro, Norway, New Zealand, South Korea, Singapore,and United Kingdom ever have significant forecast from our object countries. Finally, if we focused on middle horizon (h = 4, 6), out of factors combined with monetary model, there are more countries have evidence to support that forecast of exchange rate for factors combined with fundamental models outperform random walk model. |
