Case Studies and Statistics in Causal Analysis: The Role of Bayesian Narratives

Abstract

Case study method suffers from limited generalisation and lack of extensive comparative method both of which are prerequisites for the standard co-variation approach to causality. Indeed, in the standard model co-variation and comparative method are logical prerequisites for any causal explanation. Nevertheless, those that advocate case studies characteristically aspire to make causal inferences whilst promoting the virtues of detailed study and eliminating artificial comparison enforced by statistical samples. It is this latter aspect that recommends case methodology to many qualitatively orientated social scientists. It is proposed that the social sciences should find a “small N” (qualitative) conception of causal inference which is logically prior to any inter-unit comparison and generalisation and which complements “large N” (quantitative) statistical studies where this is not the case. The method advocated is called Bayesian Narratives which can depend upon subjective causal and counterfactual statements. Bayesian Narratives, in turn, require ethnographic data collection, in contrast to statistical sampling, in the pursuit of causal connections.