政治學的諸多主題,例如選舉競爭、政黨結盟、以及國際衝突等,皆涉及到行為者之間的策略互動。本文引介近來在美國政治學界新興的統計模型:「逆推統計法」(statistical backward induction),此模型的特色在於使用改良後的統計方法,驗證由賽局模型所推導出的策略互動,進而符合「理論模型的實證意涵」(empirical implications of theoretical models, EITM)的研究典範。本文從「隨機效用模型」(random utility model)出發,探討一般常見的「樣本選擇模型」(sample selection model)與「巢狀勝算對數模型」(nested logit model)為何無法有效處理具有策略互動性質的資料,並逐步解釋「逆推統計法」如何結合賽局理論與統計模型來分析這類資料。本文最後採用「逆推統計法」,一方面分析影響美國貿易代表處如何擬定「特別三○一名單」,以及被列名國家如何回應的因素,一方面說明如何將「逆推統計法」應用於外交政策的實證研究。
While conventional statistical methods usually assume that the error term in the models are independent and identically distributed (i.i.d.), this assumption is usually violated when observations are interdependent due to the strategic interactions among players. The violation of the i.i.d assumption results in the inefficient estimation of standard errors that can further invalidate the hypothesis testing. This paper discusses the method of statistical backward induction (SBI) developed by Curtis S. Signorino and his coauthors that can be used to analyze different kinds of strategic interactions in politics, such as electoral competitions, party coalitions, and international conflicts. After demonstrating how to derive the SBI estimator, this paper applies SBI to analyze how the U.S. government uses the Special 301 Report to coerce its trade partners into protecting the intellectual property rights (IPR) of American products. It shows that one country's trade surplus with the U.S. is a key determinant for the U.S. to nominate this trade partner in the Special 301 Report. Meanwhile, it is the dependence on the U.S. market that affects the nominated country's decision to ignore or comply with the U.S. threat of trade retaliation implied by the Special 301 Report.