Green Credit and Corporate Environmental Violations: Evidence from Regulatory Arbitrage in China

Shuangquan Yang^*

• School of Management and Economics, Beijing Institute of Technology, Beijing, China

Objective: Test whether identity-based green credit improves or worsens firm-level environmental compliance in China and identify the governing mechanism. Study design and methods: We assemble a firm–year panel for 2,376 listed firms (2013–2023) by matching 18,432 environmental penalties from government websites to green-credit status extracted via text mining from 157,000 loan disclosures and reports. Identification combines (i) firm fixed-effects and event studies, (ii) an instrumental-variables design exploiting the staggered 2012 Green Credit Guidelines interacted with pre-policy eligibility, and (iii) a quasi-experiment using the January 2018 Environmental Protection Tax reform that standardized enforcement. Exogeneity is probed with lead-IV, never-eligible/SOE subsample placebos, a 500-draw permutation-IV, fake reform years, and balance/pre-trend checks. Results: Green-credit recipients are 19.7% more likely to be penalized. At the city level, a one–standard-deviation rise in green-credit penetration is associated with a 23% lower enforcement probability for comparable violations, consistent with regulatory forbearance. Yang Green Credit and Environmental Violations The 2018 reform attenuates ∼73% of the pre-reform differential, indicating that cross-agency discretion enables the effect. Heterogeneity is concentrated in private firms, weak-governance cities, and low-disclosure firms; effects are negligible where enforcement is stronger. IV first-stage strength and placebo/permutation tests support the exclusion restriction. Conclusions: Identity-based green credit can backfire under discretionary enforcement. Standardized, coordinated enforcement and performance-based contracts that tie financial benefits to verifiable environmental outcomes are essential. Back-of-the-envelope losses are ≈RMB 168 bn (0.138% of GDP), with sensitivity bounds of RMB 139–197 bn (0.110–0.156%).