Carbon Pricing, Credit Reallocation, and Real Effects

Whether a higher carbon price raises or lowers bank credit to exposed firms is theoretically ambiguous, and the paper treats it as an empirical question. Carbon pricing is meant to promote green investment and accelerate the low-carbon transition, but investing in low-carbon technologies “typically requires significant external financing,” which a higher carbon price may discourage by worsening firms’ exposure to transition risk. So, in the authors’ framing: “higher carbon prices may spur green investment and raise credit demand; on the other hand, banks may view these firms as riskier and therefore restrict credit supply. Depending on which force dominates, credit may rise, fall, or remain unchanged.” The paper then asks whether whatever credit adjustment occurs is associated with real effects on investment, employment, and carbon emissions.

The November 2017 EU reform package that tightened the ETS permit cap and created the Market Stability Reserve, which raised permit prices from 5 to 15 euros per tonne within a quarter and to 20 euros by mid-2018. The authors emphasise two problems that make carbon-price effects hard to estimate: carbon price changes are typically not exogenous but depend on current and future economic and environmental conditions, and higher carbon prices can move both credit demand and credit supply by raising firms’ costs and affecting profitability. The 2017 reform is attractive on the first count because it delivers “a sharp and unexpected price increase within a stable institutional setting” – permit prices had been persistently low since the system’s 2005 launch.

By its ex-ante shortage of free permits – verified emissions minus freely allocated quota – scaled by total assets, with firms split at the median of the 2016 value. The variation comes from how the EU ETS allocates permits: sectors differ in their entitlement to free permits (energy suppliers receive none; the largest share goes to sectors at risk of carbon leakage), and within a sector more carbon-intensive firms receive relatively fewer free permits. Emissions and quota data for all ETS firms come from the European Union Transaction Log. The scaling by total assets follows De Jonghe, Mulier and Schepens (2020).

Large enough that the authors describe it as capturing sharp differences in incentives, not marginal ones. Within the sample of 354 Italian ETS firms, the average treated (above-median) firm “would face an increase in emission allowance costs from about 7% to more than 20% of the wage bill if it maintained its end-2016 emission intensity.” The average below-median firm “faces no meaningful adjustment, as it holds a marginal surplus of free permits.”

Difference-in-differences over 2016-2019 on quarterly loan-level records from the Italian Credit Registry, with banktime and sectortime fixed effects as the baseline. Banktime fixed effects proxy for idiosyncratic bank credit supply shocks, a first step toward isolating demand-driven credit changes (Amiti and Weinstein 2018). Because that may not fully separate demand from supply, the authors also use more granular fixed effects exploiting firm location, controlling for geographical heterogeneity in bank credit supply and comparing firms in a given sector facing similar local demand conditions. They keep the banktime and sector*time specification as the baseline to preserve statistical power, justifying this on the grounds that ETS firms “tend to be large, geographically dispersed, and active across multiple sectors,” so less likely to be tied to one bank’s geographical allocation priorities or to highly localized demand.

Highly exposed ETS firms experience a credit expansion of about 10% relative to less exposed ETS firms, and it comes through term loans. The authors read the term-loan composition – rather than credit lines or other short-term funding – as suggesting the borrowing finances new investment rather than liquidity needs. Firm-level survey evidence, run through the same difference-in-differences framework, points the same way: treated ETS firms are 29% more likely ex-post to report a willingness to increase external financing relative to the control group, and they “not only express this intention but also demand and obtain additional funds”; on questions targeting bank credit demand specifically, highly exposed firms report higher demand and are 23% more likely to cite fixed capital investment as the main reason, rather than working capital, debt restructuring, or changes in financing conditions.

On average, no – but they are more reluctant with firms that were already financially constrained. The carbon price hike is not associated with higher credit risk: highly exposed firms are no more likely ex-post to default, nor do they face tighter collateral requirements or shorter maturities, which the authors read as banks not perceiving a significant increase in borrower risk. However, the credit expansion is significantly weaker for treated firms with high short-term leverage, a more dispersed pool of lenders, or stringent collateral requirements – respectively proxies for liquidity risk and lender discipline (Diamond 1991), weak credit relationships (Petersen and Rajan 1994), and financing constraints (Kiyotaki and Moore 1997).

Treated ETS firms raise total bank credit by 21% and total liabilities by about 9% relative to the control group, and use the funds to expand productive inputs: long-term assets rise 8% and the wage bill 6%. The authors describe this as greater investment in physical and human capital.

The extra investment is not accompanied by a rise in carbon emissions, and if anything emission intensity falls – but the overall effect is not statistically significant. Treated firms show an ex-post relative decline in emission intensity, measured as emissions over revenues; the authors state plainly that “although the effect is not statistically significant, its magnitude is economically meaningful,” and align it with the hypothesis that firms adjust production in response to the carbon price shock by improving environmental efficiency. The load-bearing result on emissions is conditional, not unconditional (see Q10).

Entirely from treated firms that undertake green investment: within that group emission intensity falls about 28%. Green investment is identified from balance-sheet footnotes, classifying as green those firm-year observations that explicitly mention climate-related topics in the investment section (Accetturo et al. 2024); in the 2016-2019 ETS sample this measure varies little within firms, with firm fixed effects explaining about 75% of it, so the authors treat it as also proxying a firm’s broader commitment to greening its production technology. Among treated ETS firms, only those engaging in green investment see a significant expansion in bank credit, investment and the wage bill – “credit growth and the accumulation of physical and human capital thus accrue exclusively to firms committed to cleaner technologies.” It is within this group that the statistically and economically significant fall in emissions over revenues of about 28% emerges, and the authors note it “reflects primarily a sharper decline in the absolute level of emissions, rather than faster revenue growth.” A back-of-the-envelope calculation puts the resulting reduction in the burden of permit costs over revenues, relative to a scenario with unchanged emission intensity, at nearly 40%. The result holds whether firms are classified on ex-ante or ex-post green investment.

Banks more exposed to shocked ETS firms cut credit to non-ETS firms in brown sectors by about 1%, which is too small to change those firms’ total debt or real outcomes. The spillover analysis uses the full sample of 253,725 non-financial, non-ETS firms and compares credit provision by banks with heterogeneous ex-ante exposure to treated ETS firms, with firm*time fixed effects absorbing time-varying firm-level shocks including credit demand (Khwaja and Mian 2008; Amiti and Weinstein 2018). On average, banks’ ex-ante exposure does not significantly affect ex-post credit supply. The effect appears once sectors are split: defining brown sectors as the upper half of the NACE 2-digit greenhouse gas emissions distribution – sectors accounting for over 95% of corporate carbon emissions and where ETS firms are concentrated – banks with above-median exposure cut credit to non-ETS firms in brown sectors by about 1% relative to lower-exposure banks and to firms in green sectors, consistent with preserving sectoral diversification. “Nonetheless, this modest reduction does not translate into meaningful changes in total debt or other firm-level real outcomes for non-ETS firms.”

Pre-trend tests, a balance test, finer location-based fixed effects, exclusion of energy suppliers, an influence analysis, and an alternative control group. Specifications with time-varying coefficients verify credit adjustments are not driven by pre-existing trends. A balance test on normalized differences (Imbens and Wooldridge 2009; Imbens and Rubin 2015) shows treatment is not systematically related to firm characteristics: treatment and control groups differ significantly only in total emissions and emission shortages, not in total assets, indicating the treatment variable captures heterogeneous emissions and permit costs rather than size. Augmenting the baseline with banktimelocation and industrytimelocation fixed effects – separately, jointly, or interacted – leaves the findings unchanged. Results survive excluding energy suppliers, which may pass carbon costs into energy prices (Fabra and Reguant 2014), and an influence analysis rules out outlier-driven results. A specific concern is addressed head-on: the relative credit expansion might reflect low-exposure firms selling excess free permits at higher prices and reducing their own financing needs, but highly exposed ETS firms still expand credit relative to non-ETS firms, which are largely unaffected by the shock.

It runs opposite to the California ETS evidence, and the authors attribute the difference to the timing and nature of the shock rather than to method. Ivanov, Kruttli and Watugala (2024) use loan-level data on the introduction of California’s ETS and find an increase in credit risk and lower credit supply for exposed firms, particularly through shorter maturities. This paper finds the EU ETS carbon price shock does not significantly raise credit risk – consistent with Aiello and Angelico (2023) for Italian firms – which is what allows banks to accommodate higher demand. The explanation offered is that California’s enforcement “created substantial uncertainty about firm profitability,” whereas this shock came more than a decade after the EU ETS launch, when firms were better prepared; micro-level studies find higher EU ETS prices do not harm profitability (Martin, De Preux and Wagner 2014; Colmer et al. 2025). The two papers agree on one thing: banks discriminate across firms, favouring those with looser financing constraints. Relative to Antoniou et al. (forthcoming), who show that the early ETS surplus of free permits reduced loan spreads for permit-abundant firms, this paper’s contribution is complementary – it exploits a reform that curtailed free permits and raised their price at the same time.

That carbon pricing and bank credit complement each other in reducing firms’ carbon emission intensity – a link they say no prior study documents. The transition-risk literature reports mixed evidence on whether transition risk changes bank credit supply, and crucially finds that even sizable induced changes in loan portfolios “rarely translate into measurable improvements in firms’ environmental performance” (Kacperczyk and Peydro 2024; Sastry, Verner and Marques Ibanez 2024). The paper’s stated contribution is to show that bank credit can support reductions in carbon intensity when firms face proper incentives – here, a higher carbon price – to lower their environmental footprint. Relative to Accetturo et al. (2024) and Martin et al. (2024), which document that credit supply shocks affect how much green investment firms do, the focus here is on firms’ credit demand response to an environmental regulation shock rather than on supply-side effects.

That the EU ETS covers only a relatively small set of large firms, generally less financially constrained than typical small and medium-sized enterprises, so extending carbon pricing to the whole corporate sector need not reproduce these results. They spell out two reasons for caution. First, banks might become more reluctant to finance green investment for exposed and financially constrained SMEs, as suggested by their own finding that highly constrained exposed ETS firms get a smaller credit increase. Second, credit-supply spillovers could then generate real effects, because banks’ loan-portfolio exposure to the shock would be much larger and the associated credit cuts harder for affected firms to offset through alternative funding. Their closing line is that “more research is needed to assess the impact of large-scale environmental policies on credit allocation.”