aeja-theory-model — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited aeja-theory-model (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
AEJ: Applied is empirical-first. Theory earns its place only when it interprets the estimate, sharpens the estimand, or unlocks a magnitude the design cannot deliver alone — never as the headline. Pick the lightest tool that does the job and keep the empirical estimate the star.
| Theory's job | Right amount of model | Where it goes |
|---|---|---|
| Name the mechanism | a few equations / a conceptual framework | short section before results |
| Map a reduced-form coefficient to a structural parameter | a sufficient-statistic / envelope argument | inline derivation + appendix |
| Deliver a welfare or counterfactual number | a calibrated or partially-structural model | a dedicated section, clearly bounded |
| Discipline heterogeneity / sign predictions | a simple model generating testable comparative statics | framework section, tested in results |
Where possible, express the welfare/policy object as a function of estimable elasticities (a Harberger/Chetty-style sufficient statistic) rather than estimating a full structural model. This keeps the credibility in the reduced-form design while delivering an economic magnitude. State the assumptions under which the sufficient statistic is valid and what it omits.
If the question genuinely requires out-of-sample counterfactuals or unobservable primitives, a small structural model is acceptable — but tie each parameter to a data feature, validate against an untargeted moment, and never let the model's assumptions silently replace the identification the design provided.
A clean RD shows a tuition subsidy raises enrollment by 4.2pp (s.e. 1.1). The number is credible but the policy question is the welfare gain. Instead of building a full college-choice model, the paper uses a sufficient-statistic argument: the marginal value of public funds depends on the enrollment elasticity (estimated) and the fiscal externality of an extra graduate (calibrated from administrative tax data). This yields an MVPF of ~1.3 (illustrative) with a stated range, while the credibility still rests on the RD — the AEJ: Applied ideal.
you then test, or a channel-distinguishing test in the data — not more notation.
sufficient statistic of estimable elasticities; state the assumptions that make it valid.
against an untargeted moment; keep the credibility anchored in the reduced-form design.
【Theory's job】mechanism / reduced-to-structural mapping / welfare / comparative statics
【Tool chosen】framework / sufficient statistic / small structural model
【Key relation】estimand = f(estimable elasticities / parameters): ___
【Validity assumptions + what it omits】[...]
【Magnitude delivered】[number + uncertainty + scope], or "none — interpretation only"
【Next step】aeja-robustness~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.