Chinese listed firms that adopt AI pay their executives more, driven by productivity gains and easier financing; stronger corporate governance materially curbs this pay uplift.

Main Finding

Using A‑share listed company data (2007–2023) and a firm‑level AI application indicator constructed by textual analysis, the paper finds that corporate adoption of artificial intelligence is associated with significantly higher executive compensation. This main result is robust to instrumental‑variable strategies and multiple robustness checks. Mechanism tests indicate the effect operates (at least partially) through three channels—relief of financing constraints, improvements in total factor productivity (TFP), and increases in executives’ human capital—and stronger corporate governance mitigates the AI → compensation effect.

Key Points

• Primary result: AI adoption → higher executive pay in listed firms (statistically significant and robust).
• Robustness: authors address endogeneity using IV approaches and perform a suite of robustness tests (details in paper).
• Mechanisms (empirically supported):
  • Alleviating financing constraints — AI increases information transparency, lowers transaction costs, and can attract policy support/subsidies, easing external financing and enabling higher pay.
  • Boosting productivity — AI raises TFP via automation, labor‑structure upgrading, and complementing high‑skill labor; productivity gains translate into larger pay for executives.
  • Enhancing executive human capital — AI adoption raises the skill/content demands on managers, increasing their human‑capital returns and bargaining power.
• Moderation: Better internal corporate governance reduces the positive effect of AI on executive compensation, consistent with governance curbing managerial power and rent capture.
• Theoretical grounding draws on optimal contracting, compensating differentials, human‑capital theory, and managerial‑power frameworks.
• Policy context: Chinese government support for AI deployment (policy drives mentioned) is relevant to financing and subsidy channels.

Data & Methods

• Sample: A‑share listed companies, 2007–2023.
• AI measure: firm‑level indicator constructed via textual analysis of corporate disclosures (to capture AI application).
• Outcome: executive compensation (aggregate/top executives; precise compensation measure reported in paper).
• Empirical strategy:
  • Panel regressions linking AI indicator to executive pay.
  • Endogeneity checks using instrumental variables.
  • Mechanism tests linking AI to financing constraint measures, TFP, and proxies for executive human capital.
  • Moderation analysis testing interaction with corporate‑governance quality.
• Robustness: multiple tests (alternative specifications, samples, controls; IVs) reported to support causal interpretation.

Implications for AI Economics

• Redistribution of returns within firms: AI adoption appears to shift a portion of firm gains toward top management, with implications for income inequality and internal pay structures. Researchers should quantify how much of productivity/dividend gains accrue to executives versus workers or capital.
• Channels matter for policy: because AI raises pay via eased financing and productivity, interventions that affect transparency, financing access, or productivity diffusion will influence compensation outcomes. Policymakers aiming to limit excessive managerial rent should prioritize governance reforms and disclosure standards.
• Corporate governance as an equilibrating force: stronger governance frameworks can limit managerial rent extraction during technological change. This suggests complementarities between technology policy and corporate‑governance reform when managing distributional effects of AI.
• Measurement and methods: textual analysis of filings is a productive approach to measure firm AI engagement; however, future work should refine measures to distinguish intention, adoption intensity, and effective deployment. Instrumental strategies and natural experiments will remain important for causal claims.
• Research agenda:
  • Heterogeneity: examine industry, firm size, ownership (state vs. private), and country differences in the AI → pay link.
  • Worker outcomes: measure impacts on ordinary employee wages, employment composition, and internal pay gaps to assess broader distributional effects.
  • Dynamic effects: study long‑run returns to AI investments and whether executive compensation growth is sustained or reverts as technologies mature.
  • Complementarity with regulation: evaluate how disclosure, taxation, and compensation‑governance policies can shape distributional outcomes from AI adoption.

Limitations to keep in mind (from methods described): AI measurement via text may capture disclosures/intent rather than realized deployment; residual endogeneity may persist despite IVs; details on IV validity and exact governance/capital constraint measures should be consulted in the full paper.