* African Borders as Sources of Natural Experiments Daniel N. Posner Department of Political Science University of California, Los Angeles Los Angeles, CA 90095-1472
[email protected] This version: 12 February 2006 * Paper prepared for presentation in the Colloquium Series of the Yale Program in Agrarian Studies, 17 February 2006. I thank John McCauley for sharing his data on religious identification along the Côte d’Ivoire-Burkina Faso border and Anoop Sarbahi for his help assembling the GPCP rainfall data. The arbitrary nature of Africa’s borders is well known.1 Less well recognized is the opportunity that these arbitrary borders provide for social science research. This paper describes some of the shortcomings of traditional, observational studies and shows how natural experiments can provide for more reliable inferences about causality. The paper then draws on examples from three different projects to illustrate how the artificial nature of Africa’s borders can be used as a source of natural experiments. The Quest for Causal Inferences in Social Science Research In the social sciences, researchers typically make causal inferences through observational studies. These are analyses where the variables that the researcher is studying acquire their values through the unfolding of real-world events that are outside the researcher’s control (Collier et al 2004). The problem with observational studies is that they make it difficult or impossible to rule out the possibility that other, unmeasured variables may be the cause of the observed relationship between the variables of interest. Suppose, for example, we are interested in the linkage between smoking and cancer.