IQMR 2026

Qualitative Causal Inference: From Fundamentals to Process Tracing Applications (Modules 2, 6)

Monday, June 15; Tuesday, June 16

Eggers Hall, Room 010

David Waldner (University of Virginia)

This module introduces students to Qualitative Causal Inference (QCI), a hybrid model of inference that has important implications for how we think about and implement qualitative methods to explain particular outcomes. Part I outlines the fundamentals of QCI, contrasting it to two alternative models of inference: identification-based causal inference, associated with the ‘credibility revolution’ in quantitative social science, and the detective model of causal inference which forms the logical foundation for process tracing. Part I of the module covers the fundamentals of QCI and Part II covers applications. The foundational material introduces the three main forms of reasoning currently used by social scientists, identifies their domain restrictions for valid inference, and discusses why the detective model (process tracing) is vulnerable to false positives. The module introduces qualitative causal inference as an identification strategy for determining when an association (for our purposes, a sequence of events, or concatenation) can be given a causal interpretation. Valid qualitative causal inference requires checking whether we can substitute a bias-free causal graph for a bias-saturated causal graph. That substitution requires satisfying three criteria: causal continuity along the front-door path and causal discontinuity along the back-door and side-door paths. Over its three sessions, the module also introduces three methodological innovations that are needed to implement the qualitative identification strategy: hypothetical interventions, invariant causal mechanisms, and event-history maps. By the end of the module, students should understand the material needed to justify the claim that an observed sequence of events is causal in nature. Part II of the course then pivots to applications. We learn how to create causal graphs and check their qualities in the first session, while the second session discusses how to evaluate causal graphs and how to implement remediation measures if systematic error is detected. The final session will discuss causal graphs that students generate to depict the causal claims of their research projects. Between the two modules, David Waldner will hold “office hours” on Monday evening to provide feedback as students experiment with constructing causal graphs.

Participants may enter the module sequence after it has begun, but their doing so is discouraged.

Qualitative Causal Inference: Fundamentals and Applications I (M2, June 15)

In Part I (Module 2) David Waldner outlines the fundamentals of QCI, contrasting it to two alternative models of inference: identification-based causal inference, which is associated with the so-called ‘credibility revolution’ in quantitative social science, and the detective model of causal inference, which typically takes the form, in qualitative research, of process tracing. In these sessions students will learn how to specify causal graphs in order to achieve unit-level causal inference, with attention to the mitigation of three potential sources of bias:

(1) spuriousness and endogeneity via the construction of a ‘front-door path’ that indicates causal continuity between X and Y; (2) confounding variable bias by checking for the presence of a “back-door path” that produces a non-causal association between X and Y; (3) measurement error by checking for the presence of a ‘side-door path’ that complicates taking the pre-treatment value of the outcome variable as a proxy measure of the counterfactual (and inherently unobservable) value of the outcome under control.

8:45am - 10:15am – Models of Inference

Required readings:

1:30pm - 3:00pm – Foundations: Causal Graphs, Hypothetical Interventions, and Invariant Causal Mechanisms

Required readings:

3:30pm - 5:00pm – Foundations: Event-History Maps

Required readings:

Qualitative Causal Inference: Fundamentals and Applications II (M6, June 16)

In Part II (Module 6) we learn how to apply qualitative causal inference to our research projects. Session 1 begins with the most basic question students would have: given a research project, how do we construct causal graphs and event-history maps and how do we test hypotheses? This session helps students answer that question in two ways: guidelines and recommended best practices along with multiple examples in which verbal arguments in published research are translated into causal graphs. Session 2 pivots from the construction of causal graphs and event-history maps to their evaluation: what are the criteria for determining whether we can claim a valid, unit-level causal inference? And what do we do if we cannot make that claim? Session 2 concludes with discussion of mediation strategies. Given that the world may not cooperate with our creative theorizing and rigorous methodological practices, we may conclude that our causal claims contain systematic errors or bias: What now? First, the session argues that detecting bias is the reward we receive for our hard work, as it is far easier to be ignorant of bias than to identify it. Second, the session presents three strategies of remediation: (1) signing the bias, (2) estimating relative magnitudes, and (3) “discretizing” outcomes. The session concludes with a discussion of what we can claim if these remediation measures are not sufficient to allay concerns about systematic errors: we can still make valuable contributions in the form of causal narratives. Session 3 is entirely devoted to open discussion of causal graphs presented by students. Here we will review best practices and give examples of how to construct causal graphs and event-history maps, while deriving a suite of hypothesis tests.

8:45am - 10:15am – Best Practices and Illustrative Examples for Constructing Causal Graphs and Event-History Maps

Required readings:

1:30pm - 3:00pm – Evaluation and Remediation

Required readings:

3:30pm - 5:00pm – Workshop

Students will submit their draft causal graphs during the lunch session and we will discuss them during this session. There are no required readings.