A Bayesian Perspective on Theory-Blind Data Collection

Copestake, Goertz, and Haggard’s (CGH) “Veil of ignorance Process Tracing” (VPT)—which in essence entails placing a firewall between data collection and hypothesis testing1—is an interesting addition to a growing list of proposals made in recent years that aim to address potential sources of bias in qualitative social science. Many of these proposals (e.g., pre-registration, time-logging whether evidence was discovered before or after a hypothesis was devised) import prescriptions from large-N, frequentist, statistical research that, from a Bayesian perspective, are not applicable to qualitative research. Bayesian reasoning provides its own safeguards against the problems of confirmation bias and ad hoc hypothesizing, without imposing procedural constraints that would interfere with the inherently iterative, dynamic, and interactive nature of case-study research—where we go back and forth between hypothesizing, data collection, and analysis (Fairfield and Charman 2019)