brycewang-stanford Theory Building (cps-theory-building)
CPS rewards arguments that travel. A good CPS theory is a mechanism with scope conditions that generates observable implications a comparativist can test in more than one setting. Theory can be verbal-causal, formal/game-theoretic, or comparative-historical — but it must do more than label; it must predict what should differ across cases and when the mechanism turns off.
When to trigger
- Turning a comparative puzzle into a stated argument with implications
- A reviewer said the paper "lacks a theory," "is descriptive," or "the mechanism is a black box"
- Your finding has no portable claim — nothing another comparativist could import
- You need to specify scope conditions and the contrast with the rival from
cps-literature-positioning
What a CPS argument must deliver
- A named mechanism. State the causal process in one sentence: actor → incentive/constraint → behavior → outcome. Give it a portable name so it can be cited and carried.
- Observable implications. List 2–4 predictions, including at least one that distinguishes your mechanism from the rival (not just "outcome happens").
- Scope conditions. Say where the mechanism should hold and where it should fail. A theory that predicts everything predicts nothing; scope conditions make it comparative and falsifiable.
- Comparative statics / variation. State how the outcome should change as the key cause varies
across cases or over time — this is what the design (
cps-research-design) will exploit. - Micro-foundations where the claim is macro. If the argument is about regimes or systems, name the actor-level behavior that must hold for the macro pattern to obtain.
Modes of theorizing in CPS
| Mode | Strength | Watch out for |
|---|---|---|
| Verbal-causal | Accessible, flexible | vague mechanism, unfalsifiable scope |
| Formal / game-theoretic | Precise comparative statics | model-empirics gap; assumptions doing all the work |
| Comparative-historical | Process and sequence, path dependence | selection on the outcome; ad hoc periodization |
| Typological | Organizes variation | categories without a mechanism behind them |
The portability test (CPS-specific)
Write one sentence: "In a different region/regime — say ___ — the same mechanism predicts ___." If you cannot, the argument is a local description, not a comparative theory. Then write the distinguishing implication: "If the rival account were right instead of mine, we would see ___; my mechanism predicts ___ instead."
Checklist
- Mechanism stated in one sentence, with a portable name
- 2–4 observable implications, at least one distinguishing the rival
- Scope conditions: where it holds and where it fails
- Comparative statics stated (how outcome moves with the cause)
- Micro-foundations supplied if the claim is macro-level
- Portability sentence written for a second setting
Anti-patterns
- A label masquerading as a mechanism ("institutions matter")
- No scope conditions — the theory "explains" every case, so it is unfalsifiable
- Formal model whose comparative statics are never tested, or a loose analogy to the empirics
- Comparative-historical narrative that selects on the outcome
- A finding with no portable claim — nothing another comparativist could carry to a new case
Output format
【Mechanism】one sentence: actor → incentive → behavior → outcome (+ portable name)
【Observable implications】incl. the one distinguishing the rival
【Scope conditions】holds where / fails where
【Comparative statics】how outcome moves with the cause
【Portability】a named second setting where it should also hold
【Next】cps-research-design
Supplementary resources
../../resources/exemplars/library.md— verified CPS papers with portable mechanisms../../resources/worked-examples/01-introduction.md— argument stated early in CPS house style