Sustainable Asset Pricing with Heterogeneous Agents: A Computational Approach

Environmentally responsible assets trade at a persistent premium over identical conventional securities. How strong must investor preferences for sustainability be to explain this? I address this with a two-agent model where an ESG-conscious and a neutral investor trade green equity, brown equity, and a risk-free bond across twelve exogenous shocks. Neural networks approximate equilibrium prices and portfolios by satisfying optimality conditions across the state space. Calibrated to German twin bond data, the model matches the observed green premium (−2.53 vs. data −2.44 basis points) and the equity premium (4.86% vs. data 4.81%) with a small ESG preference intensity (ϕ∗ = 1.0 × 10−4). Green equities earn lower risk-adjusted returns than brown (Sharpe ratios 0.200 vs. 0.207), consistent with the green premium reflecting investor preferences rather than risk compensation, though this partly reflects the model’s symmetric dividend structure. The cross-sectional wealth distribution—not the ESG regime—drives green premium variation.