China’s carbon-trading pilots accelerated listed firms’ digital transformation, especially among non-state, small and eastern firms, with gains driven by increased R&D and better resource allocation and strengthened where governments actively supported the market.

Main Finding

The carbon emissions trading system (CETS) significantly promotes enterprises’ digital transformation (EDT) in China. This effect is robust to multiple checks and is transmitted mainly through increased R&D investment and improved resource-allocation efficiency. Government participation (market incentives and administrative supervision) strengthens the positive effect. The impact is heterogeneous: stronger for non-SOEs, small firms, firms in eastern regions, and firms in low-carbon or high‑tech industries.

Key Points

• Sample and scope: Panel of Chinese listed firms, 2007–2022 (28,840 firm-year observations from 2,024 firms). CETS treated as a quasi‑natural experiment using pilot-region timing (pilot launches in Shanghai, Beijing, Guangdong, Tianjin in 2013; Hubei and Chongqing in 2014; national market launched 2021).
• Identification: Multi‑period difference‑in‑differences (DID) with firm and year fixed effects; event‑study (parallel trend) tests.
• Robustness: Results hold under Goodman–Bacon decomposition (checks for heterogeneous timing bias), propensity‑score‑matched DID (PSM‑DID), synthetic DID (SDID), placebo tests, alternative dependent variables, and controls for other policy interventions.
• Mechanisms: Empirical mediation tests indicate CETS increases firms’ R&D spending and improves resource allocation efficiency, which in turn fosters EDT.
• Moderator: Higher government participation (through subsidies, tax incentives, stricter supervision) amplifies the CETS → EDT effect.
• Heterogeneity: Larger policy effect observed among non‑state‑owned enterprises, smaller firms, firms located in eastern China, and firms in low‑carbon or high‑tech sectors.
• Theoretical framing: Results are consistent with Porter‑type mechanisms where market‑based environmental regulation spurs innovation and productivity improvements that facilitate digitalization.

Data & Methods

• Dependent variable (EDT): Textual analysis of firm annual reports to construct a digital transformation index (frequency of digital-related keywords, expanded from literature and policy documents).
• Treatment: Indicator for whether a firm is located in a CETS pilot region in/after the launch year (multi‑period treatment timing).
• Econometric specification: EDT_it = β0 + β1 DID_it + β2 Controls_it + η_i + γ_t + ε_it (firm and year fixed effects). Event‑study specification used to test parallel trends and dynamic effects.
• Controls: Typical firm‑level covariates (size, leverage, profitability, industry and region dummies, etc.) and checks removing other overlapping policies.
• Robustness and validity checks: Goodman–Bacon decomposition, PSM‑DID, SDID, placebo interventions, alternative EDT measures.
• Mechanism tests: Mediation analyses for R&D investment and measures of resource allocation efficiency; moderation analysis for government participation (market incentives + regulatory supervision proxies).

Implications for AI Economics

• Environmental policy can be a powerful accelerator of firm-level digitalization and thus AI adoption. Market-based instruments like CETS create revenue/pressure incentives that increase firms’ willingness to invest in digital/AI technologies for monitoring, optimization, and trading.
• Policy complementarities matter: carbon markets plus active government support (subsidies, institutional capacity, supervision) produce larger digital/AI adoption effects than carbon markets alone. Designing carbon policies with explicit digital/AI support (training, subsidies for digital upgrades, data infrastructure) can increase effectiveness.
• Heterogeneous diffusion: smaller firms, non-SOEs, and less-advanced regions may need targeted support to capture AI benefits from environmental policy. Without targeted interventions, carbon markets may widen digital/AI adoption gaps.
• Market creation: CETS fosters demand for AI-enabled carbon management products (emissions monitoring, forecasting, trading algorithms, smart energy systems). This suggests growth opportunities for AI startups and platform providers specializing in emissions and sustainability tools.
• Modeling AI diffusion: Incorporate environmental regulation and carbon‑pricing signals as demand-side drivers in models of AI adoption, R&D allocation, and labor‑skill transitions. The empirical link through R&D and reallocation highlights channels to represent in structural and macro models.
• Labor & distributional effects: Faster digital/AI adoption driven by CETS implies shifting skill demands (more data, AI, and controls expertise), with potential displacement in routine tasks—policies should consider retraining and inclusive measures.
• Research directions: measure AI-specific investments (vs. broad “digital” keyword indices), long‑run productivity and welfare effects of CETS‑induced AI adoption, cross‑country comparisons where carbon market design differs, interactions with other climate policies (taxes, subsidies), and potential unintended consequences (market concentration, digital divide).

Suggestions for researchers and policymakers: - Researchers: Replicate using AI‑specific investment measures; examine post‑2021 national ETS effects; estimate firm‑level outcomes such as AI adoption rates, employment composition, and productivity changes. - Policymakers: Pair carbon trading with explicit digital/AI support (grants, tax credits, training programs) targeted to firms/regions less able to self‑finance digital transformation to maximize climate and digitalization co‑benefits.