reading 4 - positivist process tracing: testing, building and revising theories Flashcards
process-tracing methods
what is process-tracing?
3 components
= research method for tracing causal mechanisms using detailed, within-case empirical analysis of how a causal mechanism operated in real-world cases
-> shift focus from causes and outcomes to hypothesized causal mechanisms in between (mechanisms are not causes, but processes triggered by causes)
three core components:
- theorization about causal mechanisms linking causes and outcomes
- analysis of observable empirical manifestations of theorized mechanisms
- complementary use of comparative methods to select cases and to enable generalizations of findings from single-case studies to other causally similar cases
process = causal mechanism that links a cause(s) with an outcome
!! IS NOT a descriptive narrative of what happened
process-tracing methods -> better overall theories + greater scrutiny causal logic + knowledge of how causal process operates in real-world cases + better understanding of how the mechanism works -> stronger within-case causal inferences
process-tracing methods
what are we tracing?
how causal mechanisms operate in real-world cases
causal mechanism = focus on the hypothesized causal process in between cause and effect
- mechanisms are causal processes triggered by causes
for the rest = disagreement about the nature of mechanisms
- minimalist understanding = causal mechanism not unpacked in much detail = appropriate in early research process (many plausible mechanisms) + explore if similar mechanisms operate in other cases
- systems understanding = elements mechanisms are unpacked theoretically and studied empirically in the form of traces left by the activities associated with each part of the process
holistic view mechanisms: effect of mechanism is more than the sum of their parts
process-tracing methods
what are traces?
traces = mechanistic evidence = observational evidence left by the activities associated with mechanisms or their parts
- traces as empirical fingerprints that might have been left by operation of a mechanism
- there can be an issue of access: maybe we can’t access the fingerprints
Bayesian reasoning = logical underpinning -> points to direction of asking the right questions about what empirical fingerprints we should expect the causal mechanism to leave
observation = raw empirical material from sources that have not been evaluated for content and accuracy
mechanistic evidence = after evaluation
process-tracing methods
case selection and generalization (1.4)
process-tracing to explain a particular historical outcome -> case selection principles irrelevant (case not a case of a narrower theoretical phenomenon)
often: process-tracing to understand how a given mechanisms works within a bounded population of causally similar cases
diff types of process-tracing -> diff case selection guidelines, usually: typical cases, if we understand typical cases we can study deviant cases where the mechanism should have been operated but broke down (= theoretical-revision process-tracing)
selecting cases requires prior cross-case knowledge of population of cases (e.g. use Qualitative Comparative Analysis)
contextual sensitivity mechanistic explanations -> not enough to assume mechanistic homogeneity based on cross-case comparison
- process-tracing seems recipe for piecemeal, noncumulative research, but: without taking mechanistic heterogeneity seriously lifts the level of abstraction about mechanisms -> generalizations, theorized mechanisms tell little/nothing about how processes work in the real world
process tracing also in natural sciences
!embracing complexity -> claims more contextually specific BUT still cumulative with collaborative research over a longer period
process-tracing methods
causal homogeneity
can refer to 2 levels:
- causes: equifinality = same outcome is produced by different causes in different cases
- mechanistic heterogeneity: different operative mechanisms may link together the same causes and outcomes in diff contexts
= challenge to generalizing mechanistic explanations
process-tracing methods
4 variations of process-tracing
theoretical focus -> goal to test/build/revise theories of causal mechanisms that can in principle be present in multiple cases
case focus -> goal to develop a case-specific mechanistic explanation that can account for the big and important things going on in the case
- theory-testing process-tracing
- purpose = Is hypothesized causal mechanism present and does it function as theorized?
- focus = theory - theory-building process-tracing
- purpose = What is the causal mechanism between the cause and outcome?
- focus = theory - theoretical revision process-tracing
- purpose = Why did the mechanism break down in the case?
- focus = theory - explaining-outcome process tracing
- purpose = What mechanistic explanation accounts for the historical outcome?
- focus = case
process-tracing methods
theory-testing process-tracing
- conceptualizing a plausible hypothetical causal mechanism (can be minimalist or more unpacked/system) based on existing theorization and empirical research
- operationalization: propositions about potential empirical fingerprints that might have been left in a given case by the activities associated with a mechanism and its parts
- collection and assessing available empirical record to see if the mechanism was present
- no evidence found -> new round of theory building using the insights gained
process-tracing methods
theory-building and theoretical-revision process-tracing
theory building = developing theories of mechanisms
*existing theory used to guide what to look for (or they use hunches)
- empirics first
- how did we get there?
- when we know there might be a relationship, but don’t know what mechanism links them
- defining and operationalizing key theoretical concepts
- investigate empirical material, using it as clues that can shed light on an underlying causal mechanism (making a descriptive narrative helps)
- inferring tha tthhe found observable empriical material is actual evidence that reflects empirical fingerprints
- tentative hunches about potential mechanisms -> evaluation (uniqueness in relation to the tentative hypothesized mechanism or its part)
theoretical revision = focus on understanding why mechanisms did not work as expected in deviant cases + use this info to better understand the conditions required for the mechanism to work
- investigate where in the process the mechanism broke down -> investigate diff with typical
- uncover unknown omitted conditions that must be present for the mechanism to funtion properly
process-tracing methods
explaining-outcome process-tracing
aims to trace causal mechanisms to produce a comprehensive explanation of a particular historical outcome
uses abductive analysis as a way of building explanations: juxtapostition between empirical material and theories
2 diff starting points:
- theory-first = same steps as theory-testing (existing cause + associated mechanisms are tested to see if they account for the outcome) -> usually not sufficient cause -> testing or building path for second stage of research
- empirics first = emprical evidence to build new mechanism that can account for the elements of the elements of the outcome that were unaccounted for using the first mechanism = theory-building steps
process-tracing methods
what are we tracing (2.1)
Disagreement about what we’re tracing when engaging in process-tracing
Most agree: process-tracing is tracing causal mechanisms
But what are causal mechanisms?
Two positions about nature of causal mechanisms that are compatible with the goal of tracing processes empirically within cases
Most relevant early in a project = minimalist understanding: mechanisms relatively black-boxed as “sketches” in which the parts and the causal logics linking them are not specified
= form of plausibility probe to see whether there is any evidence of a link
Tracing a full-fledged mechanism makes sense when we have a strong hunch that a particular mechanism might be operative + want to test if it works as we theorized
Tracing events is not the same as tracing a causal mechanism
process-tracing methods
the nature of causal mechanisms - 4 positions on what we are tracing:
Not compatible with within-case research design =
- Descriptive narratives : series of events in between occurrence of cause and outcome = !!! Is not the same as explaining a process that links together causes and consequences - Intervening variables = not compatible with goal of tracing the process using in-depth within-case analysis
Compatible with within-case research design =
- Minimalist - Maximalist (systems)
process tracing methods
Understandings process-tracing that are not compatible with the goal of tracing mechanisms within cases
Process tracing = research that traces an empirical process in the form of a sequence of events between the occurrence of a cause and an outcome
BUT: causal mechanisms are more than a series of events (more than a plausible descriptive narrative about what happened), it also looks at the causal logic between events
Abell: analytical narratives concept goes a bit further than just tracing events = narrative should include action linkages between events build on subjective counterfactuals
-> problems:
1. Counterfactual understanding of causation -> attention away from how process worked within a case, focus on diff that variations in action linkages make for the outcome across cases (comparing actual vs hypothetical scenario) 2. It reduces the scope of RQ: only those that can be assessed by asking actors whether things could have been different
Process tracing / mechanisms as intervening or mediating variable between C and O (variables that stand between C and O)
Problem with this =
Theoretical level: counterfactual understanding of causation -> mechanisms/mediators need to be assessed by measuring the impact of the mechanism’s absence on values of the outcome across cases with everything else held constant -> theoretical black-boxing of the causal links in our theory = lose focus on the process itself
!investigating mechanisms as intervening can still be useful BUT not compatible with goal of learning how processes actually work within cases (bc intervening-variable approach -> need to explore difference that variation makes across cases)
process tracing mechanisms
minimalist understandings of mechanisms
(also: diagnostic evidence or causal process observations)
Mechanisms as linking causes and outcomes
BUT how they link is not unpacked in any detail = high level of abstraction
Theorized mechanistic explanation is: superficial (parts of the process and causal logics linking them not specified at all) OR incomplete (causal logics that link parts of the process are not specified)
- Superficial = cause -> one-liner description of process (blackboxed) -> outcome - Incomplete = cause -> entity -> entity -> outcome Arrows are greyboxed: the causal mechanism is not specified
Focus on finding within-case mechanistic evidence of a link (finding footprints left by a causal mechanism)
-> not enough info to answer the “how does it work” question
E.g. Tannenwald article on impact norms (nuclear taboo) on US decision-making to not use atomic weapons, plausible links: personal moral convictions decisionmakers OR domestic opinion OR world opinion
When to use?
- When there are many diff plausible mechanisms -> explore if there is any within-case mechanistic evidence before dev. More step by step analysis
- as follow-up: see if a mechanism found in several cases operates more broadly within a bounded population (-> focus on critical parts of the mechanism)
process tracing methods
unpacked mechanisms - mechanisms as systems
= core elements of the causal mechanism are unpacked theoretically + studied empirically in the form of traces that the activities associated with parts of the process leave within cases
Mechanism = system of interlocking parts that transmit powers or forces between a cause to an outcome
Lower level of abstraction than minimalistic understandings of mechanisms: trying to capture how actual causal processes play out within cases
-> strong inferences/evidence on how causal processes actually worked in real-world cases
Mechanistic explanations as systems -> understood in a holistic way: effects of the mechanism more than the sum of its parts
- Parts have no independent existence in relation to producing an outcome - Explanation can't be reduced to counterfactual dependencies bc changing one stepp can destroy or transform the system
Parts of the mechanism = entities (actors, organizations, structures) that engage in activities (manifestations of causal powers of entities in each part = transmit causal forces or powers through a mechanism)
Analytical value =
1. More logical scrutiny to causal claims: have to explain how everything is connected -> better causal theories + knowledge about conditions that must be present for the mechanism to work (good for real world policy situations) 2. Empirical analysis studies the workings of each part -> evidence for each part leads to stronger causal inference + revision of theory
-> deeper explanatory knowledge of causal relationships -> stronger claims about causation possible
!trade-off = Higher internal validity of our findings about process BUT limited ability to generalize outside of the studied cases
process tracing methods
(counterfactual understanding of causation)
= claim a cause produced an outcome because its absence would result in the absence of the outcome
