了解影響匯率報酬的因素不論在交易、避險或是投機方面，皆是非常備受關注的議題。本文研究目的是為探討同時運用市場超額報酬、利差交易和動能策略是否更能解釋外匯市場的超額報酬，即表示在實證分析上應用三因子模型有更佳的解釋力。 而本文中使用OLS迴歸及 Fama-Macbeth 兩步驟橫斷面迴歸分析，結果發現皆顯示相較於 Lustig, Roussanov, and Verdelhan (2011) 論文裡的兩因子模型，加入動能策略因子形成的三因子模型在未包含交易成本及考慮交易成本的情況下皆應該是較適切的模型。

本研究欲探討國家商品貿易特性在匯率報酬評價中扮演的角色，決定匯率報酬的因素非常多，包含利率、市場波動、國際貿易及國家政治等非常廣泛的因素，而國家商品貿易特性也會是影響匯率報酬評價的可能因素之一。本研究以「進口比率」(Import Ratios) 衡量國家的商品貿易特性，也以該數值建構投資組合。研究結果發現，去除商品貿易特性特殊之國家後，進口比例(Import Ratio)越高之投資組合，其遠期外匯貼水也偏高，且外匯超額報酬也隨之遞增。 在Ready, Roussanov, and Ward(2013)論文中認為，國家的商品貿易特性是造成不同國家利率高低差異的原因，所以該作者認為國家商品貿易特性極有可能是利差交易背後的原因。然而，本研究的Fama-Macbeth 兩步驟橫斷面迴歸實證結果發現，國家的商品貿易特性確實是造成國家利率差異的因素之一，但利差交易背後的風險背後的因素，雖然包含國家商品貿易因素，但仍包含其他因素，且商品貿易因子(IMX)無法取代利差交易因子(HML)在外匯超額報酬評價模型中的角色。 此外，本研究亦嘗試在Lustig所提出之市場因子(RX)和利差交易因子(HML)的兩因子模型中，再額外加入商品貿易因子(IMX)，構成匯率評價的三因子模型，但研究結果發現不論是在遠期外匯貼水投資組合或商品貿易投資組合中，三因子模型都沒有優於兩因子模型。 / There are many factors in determinant of exchange rate returns, such as interest rates, market volatility, international trade and politics. The purpose of this research is considering commodity trade in the pricing model of excess return of currency market. This research use “Import Ratios” to measure the characteristic of different countries’ commodity trade. We use import ratios to construct “Import Ratio Sort Portfolio”. After removing the countries which commodity trade characteristics are special, we could see when import ratios is higher, the forward discount and exchange rate return are also higher in import ratio sort portfolio. Ready, Roussanov, and Ward(2013) thought the commodity trade is the reason that cause interest rate differences between countries. In this research, the result of Fama-Macbeth two-step regression show that commodity trade is one of the reasons that cause interest rate differences. It means that there are other risks behind carry trade. In the pricing model of excess return of currency market, HML factor can’t be replaced by IMX factor. We also try to construct three-factor model, which consider excess return, carry trade, and commodity trade simultaneously. But the result shows that three-factor model can not have better explanatory power than Lustig, Roussanov, and Verdelhan(2011)’s two-factor model.