Board ties transmit ESG practices: firms with shared directors adopt peers' ESG outcomes, boosting their own ESG scores; the spillover operates through performance-based learning and green innovation and is strongest when directors have overseas experience or when firms prioritize environmental issues.

Main Finding

Firms’ ESG performance spreads through director interlocks: a focal firm’s ESG score is positively associated with the ESG performance of firms with which it shares directors. This network spillover is robust to multiple identification strategies and is partly mediated by performance-based learning and green innovation.

Key Points

• Positive inter-firm spillover: ESG outcomes of connected firms predict improvements in focal firms’ ESG performance, consistent with inter-organizational mimicry.
• Robust identification: results hold under instrumental-variable estimation, Oster’s omitted-variable-bias test, and several alternative model specifications.
• Mediation channels:
  • Performance-based learning: firms learn from peers’ ESG outcomes and adapt practices.
  • Green innovation: interlocked firms stimulate adoption of environmentally-oriented innovation in focal firms.
• Heterogeneous effects (stronger spillovers for):
  • Firms with directors who have overseas experience.
  • Firms with greater environmental attention (higher salience of environmental issues).
  • Firms in low-pollution industries or facing less stringent environmental regulation.
  • State-owned enterprises.
• Policy-relevant insight: director networks are an important mechanism for the diffusion of ESG practices across firms.

Data & Methods

• Data (general description): firm-level ESG performance measures and a director interlock network constructed from shared board membership between firms; additional firm controls and measures of green innovation and environmental attention used for mediation and heterogeneity tests.
• Empirical strategy:
  • Main regressions relate focal-firm ESG to a weighted average of connected firms’ ESG, controlling for firm and time variation plus covariates.
  • Robustness checks include alternative specifications and sample splits.
  • Causal inference approaches:
    • Instrumental variable (IV) approach to address endogeneity of network peers’ ESG.
    • Oster’s method to assess sensitivity to omitted variable bias.
  • Mediation analysis to test roles of performance-based learning and green innovation (e.g., tests using innovation outcomes or intermediate adoption measures).
• Tests of heterogeneity across director experience, industry pollution/regulatory stringency, firm attention to environmental issues, and ownership type.

Implications for AI Economics

• Network diffusion models: the paper reinforces that board-level ties are effective channels for technology and practice diffusion. Similar models can be applied to study diffusion of AI adoption, algorithms for ESG reporting, or AI-driven sustainability practices across firms.
• Role of governance in AI uptake: director characteristics (e.g., overseas experience) matter for cross-firm transfer—important when modeling how governance networks influence adoption of advanced AI tools or governance standards.
• Measurement and identification lessons: using inter-firm networks and IV/Oster approaches provides a template for causally studying how peer AI investments or AI-related outcomes propagate across firms.
• AI-enabled green innovation: findings suggest that peer effects catalyze green innovation; AI can both accelerate and be a channel for such innovation (e.g., AI for emissions monitoring), so research should consider network spillovers when estimating AI’s environmental impacts.
• Policy design: regulators aiming to accelerate responsible AI or ESG-compliant AI deployment can leverage influential nodes in director networks (or target firms with salient environmental attention) to generate broader diffusion effects.
• Data opportunities: firm-board networks combined with AI-derived ESG metrics (e.g., from textual disclosures) enable richer studies of diffusion dynamics and causal impacts of AI-related governance interventions.