Advanced Process Tracing Methods

This hands-on course aims to provide participants with the methodological tools to refine their use of process-tracing methods in their own substantive research, while also enabling them to embed process-tracing case studies in mixed-methods research design. The goal of process-tracing is to make strong within-case inferences about causal mechanisms. The course requires active participation.

The promise of process-tracing as a methodological tool is that it enables the researcher to study more-or-less directly the causal mechanism(s) linking an independent variable (or set of variables) and an outcome, allowing us to open up the ‘black box’ of causality itself. A classic example from medical science is scholarship on the association between smoking and cancer. While a strong empirical correlation had been well established for many years, it was only recently that medical scientists using techniques analogous to process-tracing have provided strong proof that a biological mechanism actually exists that causally links smoking and cancer (Bunge 1997). Within political science methodology, process-tracing is arguably the only method that allows us to study causal mechanisms, allowing us to understand how an X (or set of X’s) produces Y instead of simply studying correlations and associations, and therefore is an ‘…invaluable method that should be included in every researcher’s repertoire.’ (George and Bennett 2005:224).

The course starts by differentiating process-tracing from other methods; including both large-n quantitative methods, but also other small-n methods such as analytical narratives, comparative case studies etc. Here we define process-tracing by the interest in studying causal mechanisms in single case studies in ways that enable within-case causal inferences to be made. We discuss the three variants of Process Tracing: theory-testing, theory-building and explaining outcome process-tracing.

The course thereafter will discuss in-depth the ontological and epistemological underpinnings of process-tracing, including a thorough introduction to the Bayesian logic of inference and how it contrasts with the frequentist logic of inference and the comparativist logic of elimination. We discuss the types of inferences that can be made using process-tracing designs.

The next session discusses what causal mechanisms are, and how they differ from other types of causal theorization. The focus will be upon enabling students to translate causal theories (X->Y) into causal mechanisms that can be studied empirically using process-tracing.

The core of the course will be hands-on workshops dealing with topics such as conceptualizing and operationalizing mechanisms, along with tools for evaluating evidence. The session on conceptualization (Day 2, session 2) introduces the conceptualization process, and how theories of causal mechanisms should be formulated in order to enable the analyst to capture the transmission of causal forces through the mechanism. We will also discuss the challenges of conceptualization when we are dealing with more complex mechanisms (long-term, incremental mechanisms, multiple interacting mechanisms, and the challenge of studying equilibrium situations). The afternoon project work will deal with conceptualizing the theories in your own research as causal mechanisms disaggregated into parts composed of entities engaging in activities.

Day 3 focuses upon the operationalization phase, where empirical predictions are developed for what evidence we should find if the parts of a hypothesized causal mechanism are present. The focus will be upon test strength and how it affects our ability to update our confidence in the validity of a theory. The second session will use Moravcsik’s 1999 article as a practical example of test strength and how it can be improved. The afternoon project work deals with developing and strengthening tests in your own research.

Day 4 focuses on the challenges of gathering and working with empirical material in process-tracing research. Empirical material needs to be evaluated before it can be admitted as evidence upon which we can make causal inferences. These two sessions deal with the evaluation process, where raw empirical observations are assessed for their content, accuracy and probability, enabling us to use it to update the degree of confidence we have in the hypothesized causal mechanism actually being present using the Bayesian logic of inference. The session will involve a lecture on source criticism, and then in-class discussions of a set of empirical materials from the Cuban Missile Crisis. The afternoon project work will involve the detailed evaluation of your own empirical material in your research.

The first session on day 5 deals with mixed-methods, focusing on how case selection strategies can in certain situations enable the inferences made in the single case using process-tracing to shed light on causal relationships in the broader population of the phenomenon being studied. We illustrate when and how the theory-building and testing variants of process-tracing can be nested into a broader research program; something that is possible as they share a common nomothetic ambition with other political science research methods, where systematic factors are exclusively focused on. On the other hand, explaining outcome process-tracing studies cannot be embedded in a nested analysis due to the inclusion of non-systematic, case-specific factors in the analysis, giving them a more idiographic flavour that has a family resemble to historical research.

In the concluding session we will debate the strengths and limitations of process-tracing, discussing when it can and cannot be used, along with practical challenges that you face in your own research.

This course requires that you are familiar with the recent literature on case study methods (post 2010), and assumes familiarity with basic concepts related to Process-tracing. In particular, you should have basic knowledge about causal mechanisms and empirical tests and how they are used in case studies.