Replication for: The Politics of Foreign Direct Investment into Developing Countries

Our paper focuses on methods to estimate causal effects in time series cross sectional data when the variables of interest exhibit a strong upward trend over time. We use Büthe and Milner (2008)'s analysis of the political effects of preferential trade agreements (PTAs) on foreign direct investment (FDI) flows as an example of the problems that trending variables create for time-series cross-section (TSCS) analysis. We argue unless the functional form of the trend in each country is known and correctly specified, detrending the variables of interest can bias the results and leaves out important information. We employ two alternative methods, matching and a fuzzy regression discontinuity design, to exploit the information that the trend offers in order to estimate the causal effect of PTAs on FDI flows. Our alternative analyses provides no support to Büthe and Milner (2008)'s conclusion that signing PTAs helps developing countries to attract FDI.本文聚焦于当核心关注变量随时间呈现显著上升趋势时，时间序列截面数据中的因果效应估计方法。我们以布特与米尔纳（2008）针对优惠贸易协定（Preferential Trade Agreements, PTAs）对外国直接投资（Foreign Direct Investment, FDI）流入的政治效应所开展的分析为例，阐释趋势性变量给时间序列截面分析（Time-Series Cross-Section, TSCS）带来的诸多问题。本文认为，若无法知晓并正确设定每个国家的趋势函数形式，对核心关注变量进行去趋势处理不仅会导致估计结果产生偏误，还会遗漏关键信息。我们采用匹配法与模糊回归断点设计这两种替代性方法，充分利用趋势所蕴含的信息，以估计PTAs对FDI流入的因果效应。本文的替代性分析并不支持布特与米尔纳（2008）提出的“签署PTAs有助于发展中国家吸引FDI”这一结论。