Work

Bayesian Causal Inference in Process Tracing – The Importance of Being Probably Wrong

Year: 2013
Type: preprint
Abstract: In response to King, Keohane, and Verba’s (1994) frequentist plea for causal inference, the qualitative methods literature particularly emphasized causal inference can and should be Bayesian. The appealing feature of Bayesian causal inference is that one uses evidence to update our confidence in a causal statement. The present paper starts with reviewing the existing arguments on Bayesian case studies and process tracing. A comparison of the claims with Bayes theorem shows that they are incomplete. It is either impossible to engage in Bayesianism or Bayesian inference is only possible under specific conditions that are left implicit. A formal, but simple exposition of Bayes theorem and an empirical example show that Bayesian causal inference requires the specification of three parameters by invoking theory and empirical insights. The formal exposition is used to formulate guidelines on valid case selection and causal inference in Bayesian process tracing. Moreover, I derive minimum requirements that must be met for carrying out informal Bayesianism.
Language: English
Open Access status: green
OpenAlex ID: W2289015334
Cites:
Cited by: 2
Related to: 10