Statistical analysis of democratization: a constructive critique

Abstract

ABSTRACTWhile statistical analysis of the origins and stability of democracy has received a great deal of scholarly effort, and made major contributions to comparative politics, valid causal inference in this tradition remains exceptionally difficult and, perhaps, elusive. Three important challenges create difficulty for scholars who seek to assess the state of knowledge in this domain. First, most existing work on the origins of democracy starts (often without any explicit discussion) from the supposition that the key causes of democracy are largely randomly assigned. Second, much existing literature is far from specific about which of many methodological changes produces a new finding — making it hard for scholars to know how to interpret any resulting changes in the causal inference. Finally standards for credible statistical research have evolved and improved over time, requiring scholars to periodically decrement the credibility they attach to existing findings. This essay closes with ideas about how research in this domain might be improved by closer attention to detail and replication, by greater use of alternative strategies for causal inference including forward path analysis, and by a broader use of contemporary statistical tools for optimal multivariate description.