Qualitative Research Methods PDF

Qualitative Research Methods Process Tracing and Causal Mechanisms 8 November 2024 Stigler Center Manuel Cabal 1 / 26 Agenda 1. Finish last session’s lecture (Cases and Case Studies) 2. Theory reading: Beach & Pedersen...

Qualitative Research Methods Process Tracing and Causal Mechanisms 8 November 2024 Stigler Center Manuel Cabal 1 / 26 Agenda 1. Finish last session’s lecture (Cases and Case Studies) 2. Theory reading: Beach & Pedersen (2019) 3. Application reading: Winward’s (2021) research on Indonesia Stigler Center Manuel Cabal 2 2 / 26  What is process tracing? Method for tracing causal mechanisms using detailed, within-case empirical analysis Three components 1. What do we trace? Causal mechanisms: The process in between Cause and Outcome Thus, first, we need to theorize the process through which a Cause produces an Outcome 2. What does it mean to “trace” causal mechanisms? Observe the “empirical fingerprints” or “traces” left by the operation of a theorized causal mechanism 3. Generalization ✓ Think comparatively: What types of cases are we likely to find similar causal processes? ✓ It may not be an objective (e.g., single, important event) ✓ It may be the other way around: I start with a comparative design and use process tracing to analyze the causal processes within each case. Stigler Center Manuel Cabal 3 3 / 26  (1) What are “causal mechanisms”? Process in between Cause and Effect ✓Analytically distinct from the cause itself: In the statement C → O, we focus on “→” ✓The Cause triggers something that leads to an Outcome of interest… That “something” are the mechanism and the Outcome What “mechanisms” are not: ✓Sheer sequence of events (descriptive narrative) with little “productive” or causal meaning We must unpack the why and how, not just the what happened. Stigler Center Manuel Cabal 4 4 / 26 Sequence of events: Resulting problem: “Just so” story: An underspecified (poor theory) and untestable (poor science) tale. Stigler Center Manuel Cabal 5 5 / 26  (1) What are “causal mechanisms”? Process in between Cause and Effect ✓Analytically distinct from the cause itself: In the statement C → O, we focus on “→” ✓The Cause triggers something that leads to an Outcome of interest… That “something” are the mechanism and the Outcome What “mechanisms” are not: ✓Sheer sequence of events (descriptive narrative) with little “productive” or causal meaning We must unpack the why and how, not just the what happened. ✓Intervening variables Stigler Center Manuel Cabal 6 6 / 26  Cause → [Mechanism as Intervening Variable] → Outcome ✓Same problem as before: Treating it as a “variable” in the same sense as in linear regression, then, it doesn’t have a temporal meaning. ✓In linear regression (and quant research traditions, more generally), the term “variable” has implicit a “counterfactuals” understanding of causality: Something is a Cause because, if it’s not there, then the Outcome will not happen. ✓If we treat Mechanisms as Intervening Variables, we risk saying the same but just adding another variable: if the Cause and the Intervening Variable, then the Outcome does not happen. In other worlds, the risk is that I am avoiding specifying the how exactly the causality is produced. I am, again, leaving a why/how inside a “box”. To be sure, I can talk about intervening variables in write a good case study. For example, a factor that might attenuate the causal effect. However, the authors invite me to also specify mechanisms. Stigler Center Manuel Cabal 7 7 / 26  Two versions of causal mechanisms Depends mostly on the state of research and other practical considerations… 1. Minimalist version o We focus more on whether what we are saying makes sense by searching for diagnostic evidence (the fingerprints of the hypothesized mechanism) o Thus, we don’t go into much detail about the process o Back and forth: the search may lead to refining the hypothesized mechanism Neither the parts nor the logics linking them together are specified The parts (entities) are specified but not the logics linking them together Stigler Center Manuel Cabal 8 8 / 26 2. Maximalist version: a full-fledged mechanism o Less abstract theorization and more details about the process o What are these details? Entities that engage in activities Each part is an entity that engages in activities Each activity is the manifestation of each part’s causal power o Holistic: the parts do not have independent causal power to explain C → O… only the whole Stigler Center Manuel Cabal 9 9 / 26 The Nuclear “taboo” Tannenbaum (1999) theorizes that norms against the nonuse of A-weapons (cause) (a nuclear “taboo”) contributed to U.S. decision-makers’ avoidance of using those weapons (outcome), but the mechanism remains firmly within a theoretical gray-box because no mechanism is detailed to link the cause to the outcome. The closest she gets to unwrapping causal mechanisms is in the conclusion, where she mentions three plausible links between norms and nonuse in the form of minimalist “one-liners”: ✓ Constraints imposed by individual decision-makers’ personal moral convictions ✓ Domestic opinion ✓ World opinion Stigler Center Manuel Cabal 10 /10 26  (2) Within-case analysis: The “traces” Detective work: Observational manifestations of the process’s parts The mechanism’s “fingerprints” withing the case(s) We need to immerse ourselves in the theory and ask: What would it look like if the hypothesized mechanism is correct? What would it look like if the hypothesized mechanism is wrong? Answer these questions for each part of the mechanism Stigler Center Manuel Cabal 11 /11 26  Application: Sierra Leone Brast, Benjamin. 2015. “The Regional Dimension of Statebuilding Interventions.” International Peacekeeping. Available here ✓International efforts of state building are complex (multi-causal) ✓Research problem: over-focus on national and local actors...Missing a key piece ✓“Peer” recognition (regional actors) matters a lot for statehood Stigler Center Manuel Cabal 12 /12 26  Example: Sierra Leone Brast, Benjamin. 2015. “The Regional Dimension of Statebuilding Interventions.” International Peacekeeping ✓International efforts of state building are complex (multi-causal) ✓Research problem: over-focus on national and local actors...Missing a key piece ✓“Peer” Unstructured recognition interviews Archives (official documents) (regional actors) matters a lot for statehood Statistical data ► Combinable with methods, like elite surveys or survey experiments. Stigler Center Manuel Cabal 13 /13 26  (3) Generalizing I can bring other cases to test… ✓Whether my mechanism works in similar cases as a probe of a more general phenomenon ✓The existence of the whole C → O chain within a structured comparison framework See Winward’s article on Indonesia: from Central Java to West and East Java I also reflect about the (bounded) population of cases to which my findings may apply. In other words, I ask myself about “context” Under what conditions should I see this mechanism at work? For what other kinds of cases may this mechanism be relevant? In asking about context, I keep in mind… ✓Causal heterogeneity (equifinality) ✓Mechanistic heterogeneity Stigler Center Manuel Cabal 14 /14 26 Some possibilities to situate my case study in a broader conversation. In other words, to generalize my findings… At the level of Causes At the level of Mechanisms Heterogeneity (Equifinality) Heterogeneity My case? C1 My case? M1 C2 O1 C1 M2 M3 O1 C3 Stigler Center Manuel Cabal 15 / 26 15  Added value Why do we go over this much work? What’s the payoff ? Our logic can be more easily scrutinized Social sciences are a collective endeavor Sheds lights on contextual conditions C → O may already have empirical support (credible causal connection) But under what conditions? Everywhere? All of the time? Context matters Falsifiability Clarifies the observations that falsify our proposed mechanism… and how to revise our theory is if that’s the case Stigler Center Manuel Cabal 16 /16 26  Common concern Will we be generating propositions with such narrow scope that they have little value for knowledge accummulation? The authors argue: On the contrary—the proposition will provide detail about how things work in the real world, because our proposition will be contextualized: Under what conditions C → O holds and when are we likely to see “→” Why exactly does it work? Ex: medicine trials IV: DV: At stake: Expecting variations across Medicine Heal populations and lifestyles, or even No Medicine Do not Heal secondary effects… Knowledge accumulation happens as contextualized knowledge “piles up”. I am resigning to the ambition of giving causal statements so general that can explain a lot but are difficult to apply in the real world. Stigler Center Manuel Cabal 17 /17 26 Diverse Objectives Winward’s article on Indonesia Evangelista’s chapter on the Cold War Stigler Center Manuel Cabal 18 /18 26 Application: How “state weakness” relates to state violence in Indonesia Winward (2021) Stigler Center Manuel Cabal 19 / 26 19  Theoretical mechanism Notice: This is not a “variable” More like a value of a variable You may be asking: Where does the theory come from? Induction (combined with exiting theories’ building blocks) Stigler Center Manuel Cabal 20 /20 26  Empirical implications Testing: Is the mechanism actually present in the selected (typical) case? And does it work as theorized? For this phase, define the expected empirical implications… and then go into the case with diverse data collection techniques to gather the “fingerprints” the mechanism in action in order to support its plausibility First, I need to show or characterize the Outcome and Cause… Outcome ✓Quant data (secondary sources: published articles) ✓ Reflects on the credibility of the secondary sources’ data Stigler Center Manuel Cabal 21 /21 26 Cause ✓ Used five indicators/measurements ✓ Primary and secondary sources, too Stigler Center Manuel Cabal 22 /22 26 Primary sources 50 semi-structured interviews through snowball sampling: former political prisoners, soldiers, and perpetrators, over 8 months of fieldwork Anonymous and in Indonesian language. Probed consistency. Cross-checked with archival sources. Army publications: Military histories of the communist party, anticommunist campaigns, and histories of the individual territorial divisions involved in the violence. U.S. diplomatic cables Stigler Center Manuel Cabal 23 /23 26 Empirical implications for each Part of the mechanism (listed with ✓) defined prior to the analysis, based on the collected data: Part 1: Security forces approach potential collaborators for information ✓ Recorded events: Security forces coordinating rallies with local elites ✓ Recorded events: Local elites sanctioning violence by security forces after the fact ✓ Interviews: Local elites talk about these approaches from security forces Part 2: Civilian local elites widen targeting criteria ✓ Quantitative data: Greater levels of violence when opponents of targeted group (i.e., communists) are electorally strong ✓ Interviews: How denunciations were organized so that they deviated from the targeting criteria of the security forces ✓ Recorded victims that do not conform that the targeting criteria Stigler Center Manuel Cabal 24 /24 26 Part 3: Security forces arrest massively, which pressures on the state’s carceral infrastructure, so that they decide to execute more prisoners ✓ Military records: Massive numbers of prisoners, along with difficulties screening and caring for them ✓ Interviews or archival sources: Security forces justified executions to alleviate costs of caring for prisoners Part 4: Given mass arrests, security forces use torture, which leads to false confessions and more executions (and more arrests) ✓ Military documents and interviews: Use of torture Stigler Center Manuel Cabal 25 /25 26  Interviews: If you are surprised that interviewees were so candid to the researcher, the documentary The Act of Killing may be illustrative. Executioners from the security forces became national heroes and remained unpunished. They “normalized” their behavior in a context of “emergency”. Stigler Center Manuel Cabal 26 /26 26  Generalizability Test beyond the “typical” case (Central Java) West Java: High intelligence capacity East Java: Low intelligence capacity What cases are likely beyond the scope of these findings? (Discussed in the literature review section) ✓ Violence perpetrated by state actors because we are assuming that the perpetrators prefer selective than indiscriminate violence. Non-state actors may prefer indiscriminate violence for other purposes (and because they don’t face international backlash) ✓ State violence with selectivity in a context of territorial control. Thus, does not apply to civil wars (e.g., 1930s Spain, today’s Sudan) Stigler Center Manuel Cabal 27 /27 26