Use when the quantitative model is the bottleneck for an American Economic Journal: Macroeconomics (AEJ: Macro) manuscript — DSGE, New Keynesian, heterogeneous-agent (HANK / Aiyagari-Bewley), or structural estimation — and calibration, parameter identification, solution accuracy, or counterfactual validity need discipline. For empirical shock identification see aejmac-identification.
SkillScore82Maintained
Automatic
◆Research pack
Last verified 8/15/2026
What it changes
Passed 18 of 22 test cases with the skill, and 6 of 22 without it.
Without skill27%
With skill82%
Checks it passed that the model alone missed
The Next step section states aejmac-robustness.
The Next step section specifies aejmac-robustness.
The response uses the bracketed header 【Numerics】.
and 16 more in the full report
14 fixed4 passed either way2 still failing2 broke
Pass-rate lift
+55 pp
Tokens
+16%
Agent turns
0%
Response time
-26%
Model gemini-3.6-flashMeasured 8/15/2026v118 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 14 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 14
Prompt
I need to evaluate the quantitative rigor of a calibrated general equilibrium model with incomplete markets analyzing tax reform welfare effects. Provide a formal critique formatted under the following key sections: Model type, Headline quantity, Parameter discipline, Identification (structural), Numerics, Counterfactual validity, and Next step.
What was checked, without and with the skill
Checked
Without skill
With skill
The response formats output headers using bracketed markers like 【Model type】.
Missed
Passed
The Parameter discipline block addresses untargeted moment validation.
Missed
Passed
The Next step section specifies aejmac-robustness.
Missed
Passed
case-03 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md72 linessingle file
Quantitative Theory & Model Discipline (aejmac-theory-model)
When to trigger
Parameters are calibrated or estimated but it is unclear what disciplines each one
A DSGE/HANK model is solved but the solution method / accuracy is unstated
A counterfactual or welfare number is reported with no validity argument (Lucas critique)
Untargeted moments are never shown, so the model's fit is asserted not demonstrated
You are unsure the model clears AEJ: Macro's quantitative-discipline bar
The AEJ: Macro model bar
AEJ: Macro welcomes quantitative-theoretical macro, but the standard is discipline, not decoration: a calibration or structural estimate must be tied to data, the solution must be accurate enough for the claim, and the counterfactual must be defensible. The model exists to deliver a broad-interest macro quantity (a multiplier, a welfare cost, a share of inequality, a propagation magnitude), not to display machinery.
Discipline paths
Path A: Calibration discipline
Source every parameter. Externally calibrated (cited micro/macro estimates) vs. internally calibrated (matched to targeted moments) — label each and give the target.
Targeted moments table. Show data vs. model on the moments you matched.
Untargeted-moment validation. Show the model matches moments it was not asked to match — this is the credibility payoff for calibration.
Sensitivity. Report how the headline quantity moves with the key parameters (and which moment moves which parameter).
Path B: Structural estimation discipline
Name what identifies each parameter — the data feature / moment, not "the likelihood." Report a sensitivity / informativeness measure (e.g., a sensitivity matrix) so readers see which data move which parameter.
Estimator stated (MLE / GMM / SMM / indirect inference / Bayesian) with priors (if Bayesian), starting values, tolerances, and multi-start evidence of a global enough optimum.
Monte Carlo recovery: simulated data return the true parameters.
Path C: Solution accuracy & numerics
State the solution method (perturbation order, projection, value-function iteration, sequence-space Jacobian for HANK) and why it suffices for the nonlinearity/size of shock studied.
For occasionally-binding constraints (ZLB, borrowing limits) or large shocks, justify global vs. local methods.
Report accuracy diagnostics (Euler-equation errors, grid/refinement checks) where the claim depends on accuracy.
Set and report seeds for any simulation.
Path D: Counterfactual & welfare validity
Argue the estimated/calibrated parameters are policy-invariant enough for the counterfactual (Lucas critique); show they are not functions of the policy you change.
State the welfare metric (consumption-equivalent, etc.) and carry uncertainty into the counterfactual quantity.
For HANK: be explicit about the distributional channel and the role of the MPC distribution / liquidity.