What problem does it solve? Structural estimation can silently produce confident, wrong answers: an optimizer converges on a parameter the data never identified, or a misspecified model fits in-sample and fabricates counterfactual results for mergers, taxes, or welfare calculations. This Skill enforces a disciplined workflow — model card, per-parameter identification, Monte Carlo recovery, gradient verification, and equilibrium re-solving — so structural estimates are validated before any counterfactual is trusted. ## Core Features & Use Cases - Model card gate: Documents primitives, per-parameter identification, the target counterfactual, and the estimation plan before any estimation machinery is built. - Monte Carlo recovery: Proves the estimator recovers known parameters from distant starting values across the parameter space, with objective-surface profiling and Hessian eigenvalue checks for non-identification. - Analytical gradient discipline: Derives group-by-group gradients (or uses AD), verified against finite differences, avoiding nested-loop tolerance bias. - Counterfactual simulation: Re-solves equilibria for merger pricing, dynamic discrete choice, entry games, auctions, consideration sets, and search models, bounded by the weakest assumption. - Use Case: An economist estimating a random-coefficients (BLP) demand model uses this Skill to write the model card, verify recovery with simulated data, estimate with pyblp, and simulate post-merger prices with the Bertrand equilibrium re-solved. ## Quick Start Use the structural-estimation skill to estimate a BLP demand model on my market-level data and simulate the price effects of the proposed merger.
