Published online by Cambridge University Press: 29 December 2014
Africa’s arbitrary country borders have been seized upon as sources of“natural experiments”: having randomly assigned people to different countrytreatments, differences in outcomes on either side of the border can then beattributed to the institutions, demographics, or policies put in place ineach country. While methodologically attractive, the use of African bordersas sources of natural experiments presents several potential pitfalls. Wedescribe these pitfalls—some common to all studies that employjurisdictional boundaries, some unique to African borders—and offerguidelines for overcoming them. We conclude that African cross-borderstudies can provide research advantages similar to well-executed comparativecase studies, but that they frequently offer weaker inferential leveragethan is claimed.
John F. McCauley, Assistant Professor of Government & Politics,Department of Government & Politics, University of Maryland,College Park, 3140 Tydings Hall, College Park MD 20742, USA (mccauley@umd.edu). Daniel N. Posner,Professor of International Development, Department of PoliticalScience, University of California, Los Angeles 4289 Bunche Hall, LosAngeles CA 90095, USA (dposner@polisci.ucla.edu). The authors thank the UCLAGlobalization Research Center Africa for financial support; IbrahimaOuattara, Aimé Bado, Ollo Edmond Da, and Warhanti Da for theirexcellent research assistance, and Denis Cogneau, Elise Huillery, andthe DIAL research team for sharing data. Useful comments were receivedfrom Daniel de Kadt, Chad Hazlett, members of the Working Group inAfrican Political Economy, participants at the 2012 PrincetonExperimental Research Workshop, and two anonymous reviewers.