Heterogeneous Agent Calibration via Sparse Bayesian Structural Estimation: A Constrained Optimization Framework for Macroeconomic Equilibrium Decomposition

Abstract

This paper develops a constrained sparse Bayesian structural estimation framework for calibrating heterogeneous agent models (HAMs) under aggregate demand uncertainty and endogenous credit market frictions. Existing methodological approaches frequently suffer from parameter identification failure and computational intractability when applied to high-dimensional income distribution panels. We address these limitations by integrating L1-regularized posterior inference with Krusell–Smith algorithm extensions, enabling tractable equilibrium decomposition across asymmetric household portfolios. Employing longitudinal microdata from fourteen OECD economies (2005–2023), our framework achieves a 34% reduction in simulated method-of-moments estimation error relative to conventional DSGE benchmarks. The results demonstrate that precautionary savings heterogeneity accounts for approximately 41% of unexplained cross-country output gap variance, providing actionable implications for macroprudential policy calibration and distributional fiscal consolidation strategies.