Levels of AI Investment, Environmental Disclosure Performance, and Firms' Interest-Bearing Debt Financing Levels (Journal of innovation and development)

Summary

• The study analyzes a sample of Chinese A-share listed companies from 2010 to 2023 comprising 17,781 observations.p.6
• The sample firms exhibit an average interest-bearing debt ratio (Leverage) of 0.2010 and a median of 0.1865.p.7
• Main empirical models control for firm-level and year fixed effects, showing a baseline lnAI coefficient of 0.2010 on firm leverage.p.8
• Investigating environmental disclosures shows a negative correlation between lnAI and EG_Index with a coefficient of -0.5507.p.8
• Heterogeneity analysis indicates that the positive impact of AI on debt financing is significant for large enterprises with a coefficient of 0.3172, but not for small enterprises.p.9
• Robustness testing via Propensity Score Matching yields an lnAI coefficient of 0.2011 on the matched sample.p.10

Key insights

1. In the primary market during the first half of 2026, China's total AI investment exceeded 300 billion yuan, accounting for 48.6 percent of total primary market investment.p.1
2. Firm leverage is defined as the sum of short-term borrowings, long-term borrowings, and non-current liabilities due within one year, divided by total assets.p.6
3. AI investment is measured via the natural logarithm of 1 plus the AI investment amount, displaying a sample average value of 0.0120.p.7
4. Stepwise regression shows that higher environmental disclosure performance (EG_Index) reduces the interest-bearing debt ratio with a coefficient of -0.0040.p.8
5. Bootstrap indirect effect tests yield a 95% Confidence Interval containing 0, indicating that environmental disclosure does not present a statistically significant mediation effect.p.9
6. Large companies enjoy greater access to debt financing from AI investments due to stronger information systems and higher AI implementation capabilities.p.9
7. Using one-period lagged AI investment levels (L1_lnAI) yields a statistically significant positive coefficient of 0.2568 on debt ratios.p.10
8. Replacing the log AI variable with raw AI investment in 100 million yuan units (AI_raw) produces a baseline coefficient of 0.1727.p.11
9. A placebo test with 500 random permutations confirms the true coefficient of 0.2010 is not driven by random chance.p.12