Joint work with Aslan Bakirov and Francesco Del Prato. Under review.
Abstract. How much wage dispersion is visible in characteristics we observe? Using Portuguese matched employer-employee records linked to firm financial data, we sort workers and firms into observable cells and decompose log wages across worker-firm cells. Worker cells account for 35.0% of the variance, firm cells 6.7%, sorting 8.8%, and a worker-firm interaction 6.7%; the remaining 42.8% lies within cells. A second exercise splits the firm component into what firms pay similar workers (pay policy) and whom they employ (workforce composition). Pay policy dominates, accounting for about two-thirds of the firm component’s variance, and the two margins reinforce: higher-paying firms employ higher-wage workforces. Holding observable cells fixed across Portugal’s recovery, the fall in log-wage variance is explained by changing cell wage schedules rather than by workforce reallocation or re-sorting. The two-sided structure usually credited to latent worker and firm effects is thus visible in observed characteristics on the worker side, much less so on the firm side.
Abstract. We study structural estimation on networks in the empirically common case of a single large observed graph. We propose an adversarial estimator that minimizes statistical distance between observed and simulated node-specific distributions of local network neighborhoods. The paper provides two theoretical results: population identification via a divergence characterization of the estimation objective, and consistency under growing-graph asymptotics with cross-observation dependence. A key contribution is computational. We provide a reproducible estimation workflow that integrates fixed-point simulation, efficient focal-neighborhood data construction, and alternating minimax training with stabilization tools suitable for large-scale runs. The workflow is model-agnostic in a broad class of network structural models and is straightforward to implement with modern software. In benchmark simulations, the procedure scales to large graphs and recovers structural parameters with high precision.