Comparative-Case Attribution Test¶
Comparative analysis — instantiates Agency / Structure Attribution Balance
Uses real cases where the actor varied under similar structures, or the structure varied under similar actors, as natural experiments that discriminate among competing attributions.
The Comparative-Case Attribution Test answers the replaceability question without imagining anything. Instead of a hypothetical swap, it finds actual cases: similar structures that hosted a different actor, or a similar kind of actor placed under a different structure, and reads the difference in outcomes as evidence. Its defining move is that history has already run the experiment several times, and the analyst's job is to line the runs up. Where the counterfactual probes reason about one manipulated world, this test reasons across several worlds that genuinely occurred — which makes its evidence about attribution harder to wave away, and grounds it in observed variation rather than assumption.
Example¶
An inventor is celebrated as the sole originator of a device. The test assembles the real comparison set. On one axis it finds independent laboratories that reached the same result within months of one another, each with a different lead figure — a case where the actor varied while the structural conditions (mature prior art, available instruments, an active patent market) were nearly shared. On the other axis it finds a similarly skilled inventor who, working under a weaker funding structure and a hostile patent regime, never brought a comparable device to market — a case where the structure varied while the actor-type was held roughly constant.
Reading the set discriminates among models. The multiple-independent-arrivals cases cut hard against the "unique individual" model: when several different people cross the line under the same conditions, the conditions are doing much of the work. The failed-under-weaker-structure case supports the structural model. The test does not crown one winner; it returns a set of surviving models with their relative support — here, "the invention was over-determined by the state of the field, with the focal inventor decisive mainly on timing and packaging."
How it works¶
- State the claim as competing models. Actor-unique, structure-driven, coalition-driven — so the comparison has something to adjudicate rather than merely illustrate.
- Select cases that vary one axis. Similar structure with a different actor (Mill's Method of Difference), or similar actor-type under a different structure, so the varying factor is the suspected cause.[1]
- Match on confounders, and record the misses. Get the cases as alike as possible on everything else, and write down every dimension on which they fail to match.
- Read which model the differences support. Let the pattern of outcomes across the real cases raise or lower each model's support, and keep more than one alive.
Tuning parameters¶
- Comparison design — most-similar cases (differ on the suspected cause, alike elsewhere) versus most-different cases (alike only on the suspected cause). The choice determines what a match can and cannot prove.
- Case count — few cases studied deeply versus many studied shallowly. Depth catches confounders; breadth resists over-reading a single quirky pair.
- Matching stringency — how close a comparison case must be before it counts. Loose matching finds cases easily but lets confounders in; tight matching is clean but often leaves you with almost none.
- Varying axis — whether the actor or the structure is the dimension allowed to differ across cases, which decides whether you are testing replaceability or conditionality.
When it helps, and when it misleads¶
Its strength is that it disciplines speculation with reality: the strongest available evidence on whether a person was replaceable is often that a comparable situation with a different person actually happened, and you can see how it turned out. It converts arguments about imagined worlds into arguments about the record.
Its failure mode is the small-N curse: no two real cases match on everything, so an unmatched confounder can masquerade as the cause — the "many variables, small number of cases" problem that makes comparison suggestive rather than conclusive. The classic misuse is cherry-picking the one comparison case that flatters a prior while ignoring the cases that cut against it. The guarding discipline is to pre-specify the comparison set and the matching criteria before looking at outcomes, to report the unmatched confounders honestly, and to keep the alternative-model set open rather than declaring the comparison decisive.
How it implements the components¶
attribution_uncertainty_and_alternative_model_set— its output is a set of competing models with their relative support and residual uncertainty, disciplined by how the real cases came out.opportunity_constraint_and_selection_map— comparing similar actors under different structures exposes exactly how the option set and selection differed across cases, which is the structural variable the comparison turns on.
It does not run hypothetical manipulations: swapping the actor is counterfactual_actor_substitution_test (the Counterfactual Actor-Substitution Probe) and perturbing the structure is structural_counterfactual_and_constraint_test (the Structural-Constraint Relaxation Probe); this test grounds the same replaceability question in cases that actually occurred rather than ones imagined.
Related¶
- Instantiates: Agency / Structure Attribution Balance — it supplies the empirical, cross-case discipline the counterfactual probes cannot.
- Consumes: Network and Institutional Position Map supplies the structural descriptions used to judge whether two cases are comparable.
- Sibling mechanisms: Counterfactual Actor-Substitution Probe · Structural-Constraint Relaxation Probe · Actor–Structure Evidence Matrix · Process Tracing Across Levels · Narrative Focal-Actor Audit · Network and Institutional Position Map · Paired Micro/Macro Timeline · Turning-Point Opportunity Analysis · Plural Causal Synthesis Review
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: The method compares already-existing cases selected to vary actor or structure, matches confounders, and updates support among competing attribution models, so its operative form is comparative causal analysis.
Nearest alternative: Experiment, Test & Rehearsal — The cases function as natural experiments, but the mechanism does not deliberately impose variation or expose a target; it analyzes variation that history already produced.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Political Science
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Comparative social science established controlled cross-case variation to discriminate actor and structural explanations.
Related originating lineages:
- History & Historiography — Comparative history supplies temporally grounded cases and observed variation rather than imagined counterfactuals.
- Philosophy — Mill's method of difference supplies the canonical inductive contrast behind the attribution test.
- Sociology & Anthropology — Comparative sociology contributes actor-versus-structure attribution across real institutional settings.
Review resolution: The complete mechanism is the comparative-politics design that treats matched real cases as natural experiments, but it synthesizes Mill's causal logic with comparative historical and sociological evidence. All three are formative rather than mere application domains, so they are retained as alternates.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Mill, J. S. A System of Logic, Ratiocinative and Inductive: Being a Connected View of the Principles of Evidence and the Methods of Scientific Investigation. 2 vols. John W. Parker (1843). Defines the Method of Difference as comparing cases alike in every circumstance but one and treating the differing circumstance as causal or necessary. registry ↩