## wpiea2025062-print-pdf

## Source details

**Canonical URL:** [wpiea2025062-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025062-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025062-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025062-print-pdf.pdf.json)

---

### Carbon stress model (firm module)
- Four-step simulation of a hypothetical CO2 price increase on firm creditworthiness:
  - Determine Scope 1 and Scope 2 emissions for each company.
  - Compute additional costs from a higher CO2 price.
    - Raw stress scenario: costs come on top of already existing expenses.
    - Enhanced stress scenario: already paid expenses under the EU ETS (increase to 100 EUR) are recognized, Scope 2 costs are added, and firms are allowed to pass on carbon costs to consumers.
  - Run financial projection in line with EU ETS and IFRS accounting rules.
    - Emission cost incurred via obligation to purchase permits for every tCO2 emitted in fiscal year t; allowances bought and paid via cash and deposits in period t and certificates submitted in year t+1.
    - Carbon costs are assumed to be paid via the “cash and deposits” balance sheet item; if insufficient, companies borrow (increase indebtedness).
  - Use stressed positions as basis for stressed credit risk ratings assuming an otherwise static balance sheet (no additional investments; profits and losses affected only by additional CO2 costs).
- Five financial ratios stressed: EBIT, self-financing ability, net indebtedness ratio, return on cash flow, EBITDA - ROI.
- One ratio not stressed: capital interest burden (total interest payments divided by total debt), assuming average interest rate of a firm is not impacted by the carbon price shock.

### Bank module (translating firm PD impacts to banks)
- Firm-level PD impacts translated into bank-specific capitalization-ratio changes by stressing banks’ credit exposure to non-financial firms.
- Effect computed for:
  - the 776 firms with firm-level data available, and
  - all other firms recorded in AnaCredit using extrapolated stressed PDs.
- Where firm-level emissions are unavailable, country-sector average emission intensities are used; a Monte-Carlo simulation accounts for within-sector heterogeneity by combining observed dispersion with Eurostat country-sector averages.
- Key bank assumptions:
  - Banks follow the internal rating-based (IRB) approach and IFRS.
  - Unstressed Common Equity Tier 1 ratio: CET1r = CET1 / RWA.
  - Stressed CET1r:
    - CET1r* = (CET1 − Σf∈firms ∆provisionsf) / (RWA + Σf∈firms ∆RWAf).
  - Households, government, financial-sector credit exposure and market risk exposure left unstressed.
- LGD linked to PD increases per Frye and Jacobs (2012), raising expected losses and risk weights and thus reducing CET1r.

### Metrics (firm- and bank-level)
- Firm-level:
  - Stressed PD factor F*PD = PD* / PD (relative increase in PD). Example interpretation: a one notch downgrade ≈ PD factor of 1.5.
  - Stressed PD difference ∆PD* = PD* − PD (absolute scale); when multiplied by exposure, measures change in expected loss.
- Bank-level:
  - ∆CET1r* = CET1r* − CET1r (change in Common Equity Tier 1 ratio).
- Bank-level results present averages of Monte-Carlo simulation distributions unless indicated otherwise.

### Data sources and coverage
- Firm-level:
  - Consolidated financial statements from European Records of IFRS Consolidated Accounts (2023); fiscal year 2021 consolidated financial statements for 881 groups in participating countries.
  - Emission data from ERICA, ISS ESG, and Carbon 4 Finance (Scope 1 and Scope 2).
  - Main firm-level sample: 776 firms for fiscal year 2021.
  - For missing firm-level emissions, Scope 1 emission intensities at country-sector level from Eurostat (2024) scaled to derive Scope 2 emissions.
  - EU ETS coverage: 1.36 billion tons of CO2 equivalent covered in 2021 (European Environmental Agency (2023) cited).
- Bank-level:
  - Bank capitalization and portfolio composition from EBA (2021) transparency exercise for the largest 112 European banks; sample restricted to 81 banks where at least 5% of RWA stem from corporate RWA and AnaCredit data available.
  - AnaCredit exposures scaled to match consolidated exposures reported to EBA.

### Stylized facts and sample statistics
- Firm-size classification (revenue thresholds preserved):
  - Small if revenue < EUR 250 Mio., Medium if EUR 250 Mio. ≤ revenue < EUR 1,500 Mio., Large if revenue > EUR 1,500 Mio.
- Emission data by source and firm size:
  - Small: Corp. publication 47, Modeled 205, Total 252
  - Medium: Corp. publication 169, Modeled 79, Total 248
  - Large: Corp. publication 257, Modeled 19, Total 276
  - Total: Corp. publication 473, Modeled 303, Total 776
- Sector classification: 469 firms in climate policy relevant sectors (CPRS); 307 firms in non-CPRS.
- Baseline creditworthiness: Of 776 entities, 469 reach investment grade on the S&P-equivalent rating scale.
- Emission intensity (median total intensity, tCO2/Mio EUR of revenue):
  - Kendall’s tau between Scope 1 and Scope 1+2 intensities = 0.75.
  - Fossil fuels 231
  - Utilities 434
  - Energy-intensive manufacturing 34
  - Transportation 39
  - Buildings 25
  - Non-CPRS 17
- Global Scope 1 emissions and EU ETS coverage by sector (mtCO2; totals and shares preserved exactly):
  - Fossil-fuel: Scope 1 emissions 86, EU ETS share 67%, Free share 40%
  - Utility: Scope 1 emissions 346, EU ETS share 60%, Free share 2%
  - Energy-intensive: Scope 1 emissions 246, EU ETS share 37%, Free share 30%
  - Buildings: Scope 1 emissions 5, EU ETS share 5%, Free share 3%
  - Transportation: Scope 1 emissions 70, EU ETS share 20%, Free share 12%
  - Non-CPRS: Scope 1 emissions 16, EU ETS share 10%, Free share 7%
  - Total: Scope 1 emissions 770, EU ETS share 48%, Free share 16%
  - Interpretation: 770 MtCO2 direct emissions globally in 2021; 373 MtCO2 (48%) covered through the EU ETS.
- Bank-level sample statistics:
  - 81 banks report EUR 7,016 bn in total risk-weighted assets.
  - CET1r varies between 8.4% and 38%; system-level CET1r is 14.9%.
  - Corporate credit risk exposure accounts for 43.7% of total RWA on average.
  - Share of exposure of the 776 IFRS groups in banks’ portfolios is on average 7%.

### Results — Impact on firm probabilities of default (raw stress scenario)
- Sample: 776 firms.
- Sectoral stressed PD factor F*PD = PD*/PD:
  - Fossil-fuel sector:
    - Median PD increase: 18%
    - Third-quartile PD increase: 66%
  - Utility sector:
    - Median PD increase: 19%
    - Third-quartile PD increase: 113%
  - All other sectors:
    - No widespread shift in median PD factor, but significant impact for individual companies.
- Distributional notes (Table 5, selected):
  - Total (N = 776): min 1.00; p10 1.00; p25 1.00; p50 1.01; p75 1.03; p90 1.23; max 57.89
  - Fossil-fuel (N = 161): min 1.00; p50 1.18; max 6.78
  - Utility (N = 45): max 30.23
  - Energy-intensive subsector example ("Rubber, plastic and glass, ceramics", N = 22): p75 1.88; p90 45.94; max 57.89
- Absolute stressed PD difference ∆PD* (Table 6, selected):
  - Total (N = 776): min 0.000; p10 0.000; p25 0.000; p50 0.000; p75 0.000; p90 0.020; max 17.55

### Heterogeneity within the energy-intensive sector
- 236 firms in the energy-intensive sector in the sample.
- Subsector impacts:
  - “Metallic manufacturing and metal production”, “wood and paper industry”, and “rubber, plastic and glass production, ceramics” record strong increases comparable to fossil fuel and utility sectors.
- Aggregate debt-weighted sample PD increases from 0.41% to 0.60%.
- PD increases measured on an absolute scale are below 1 percentage point for almost all firms and below 0.1 pp for most firms except utilities.
- “Metal production and manufacturing of metal products” shows a widespread shift among upper-half firms; baseline PDs in that subsector are among the highest.

### Impact on firm PDs and rating migrations (raw stress scenario)
- Under the raw stress scenario:
  - 88% of firms would not be downgraded.
  - Downgrades typically affect firms with high absolute emissions: firms downgraded from investment grade to non-investment grade produce 43% of total emissions.
  - Utility sector: more than half of the 45 firms would be downgraded by at least one notch.
  - Fossil fuel sector: 8 of 16 companies associated with a downgrade.
  - 17% of companies in energy-intensive industries would be downgraded.
  - Stress simulation indicates drop below investment grade for only 6% of sample companies (from an initial 60% with investment grade).
  - Measured in terms of debt, 4% of total volume migrates to non-investment grade.
- Table 7 (selected downgrades per CPRS sector):
  - Total (N = 776): >= one notch 12%; >= three notches 3%; to non-investment grade 6%
  - Fossil-fuel (N = 165): >= one notch 50%; >= three notches 6%; to non-investment grade 15%
  - Utility (N = 45): >= one notch 53%; >= three notches 20%; to non-investment grade 35%

### Enhanced stress scenario (accounting for incurred carbon costs and pass-through)
- Aggregate outcomes:
  - 91% of firms see their rating unchanged or upgraded.
  - Aggregate debt-weighted PD increases from 0.41% to 0.48% (12 bp lower stressed PD compared to raw stress).
- Model outputs for energy producers:
  - Simplified model yields a revenue boost of 18.4% across the board for energy producers; total additional carbon costs sum to 3.6% of total revenue.
  - Renewable energy suppliers do not see an increase in marginal costs while prices go up.
- Counter-intuitive observations:
  - Carbon pricing can be financially beneficial for carbon-efficient firms.
  - Direct relationship between emission intensities and transition risk observed in raw scenario vanishes cross-sectorally in enhanced scenario.
- Enhanced scenario summary (Table 8):
  - Total (N = 776): Downgrades >= one notch 9%; to non-investment grade 3%; Upgrades >= one notch 4%
  - Utility (N = 45): Upgrades >= one notch 69%

### Impact of using firm-level data vs. country-sector averages
- Estimation error when using country-sector-level averages instead of firm-level emissions:
  - Average absolute deviation for the 776 firms is 0.19 pp.
  - Deviations range from 0.03 pp in buildings to 1 pp for utilities.
- Electricity sector flagged as particularly inadequate for sector-average approximations.

### Impact on bank capitalization ratios (CET1r)
- Sample: 81 bank groups (570 banks); credit information for 704 IFRS-groups; dataset includes approximately 3.3 million firm-bank relationships for other companies.
- Raw stress scenario:
  - Aggregate CET1r decreases by 72 basis points (bps).
  - Comparable historical shock: CET1r decreased from 15.6% in Q4 2019 to 15.0% in Q2 2020 during COVID-19.
  - 19 banks experience a CET1r drop by 100 bps or more.
  - Only one bank drops below a CET1r threshold of 10% due to the stress.
  - Two sample banks already start with a CET1r below 10%.
- Enhanced stress scenario:
  - Aggregate CET1r decreases by 19 bps.
  - Five banks see their CET1 ratio increase under this scenario.
  - No bank falls below a capitalization ratio of 10% due to the stress.

### Drivers of the CET1r impact
- Higher PDs → higher risk weights and higher credit risk provisions.
- IFRS9 staging effect:
  - Exposures with a PD increase by a factor of 2 or more are reclassified into Stage 2 and require lifetime provision.
  - Aggregate staging effect:
    - Without staging, mean aggregate CET1r decline in raw stress would be 51 bps (instead of 72 bps).
    - Without staging, mean aggregate CET1r decline in enhanced stress would be 17 bps (instead of 19 bps).
- Base PDs and cyclical point of the economy amplify stress impact via provisions and risk weights.

### Quantifying intra-sector heterogeneity effects on CET1r
- Using simulated firm-level emission data (vs. Eurostat sectoral averages) leads to a higher ∆CET1r* for almost all banks in both scenarios.
- Absolute deviation more pronounced in raw scenario: four banks have a mean difference over 30 bps.
- Relative deviation: on average 21% higher ∆CET1r* impact in both raw and enhanced scenarios.

### Bank-level drivers: size, portfolio intensity, and exposure composition
- Bank size:
  - No significant association between bank size and mean CET1r* impact; at best a small trend of larger banks being marginally harder hit in raw scenario.
- Portfolio emission intensity:
  - Raw scenario: strong negative relationship between portfolio emission intensity and mean CET1r impact.
  - Enhanced scenario: link between emission intensity and deterioration in credit quality vanishes; emission intensity no longer a reliable short-term proxy for carbon transition risk.
  - Banks where energy producers account for over 10% of exposure are highlighted in analysis comparisons.

### Conclusions’ robustness and sensitivity
- Carbon price sensitivity:
  - Worsening of creditworthiness roughly scales proportionally with carbon price increase; PDs increase exponentially.
  - Example: If carbon price increases by EUR 200 instead of EUR 100, stress in terms of log(F*PD) will be 100% higher on average.
- Pass-through modeling:
  - log(F*PD) decreases proportionally as pass-through rate increases.
  - Simple pass-through scheme: non-energy firms modeled with firm-by-firm revenue compensation of 50% of additional cost.
  - Pass-through may lead to net benefits depending on market structure; charging higher prices impacts demand.
- Financing additional carbon costs:
  - Results robust to financing choice (running down cash reserves vs borrowing): deviation between approaches for F*PD corresponds to factor 0.99 and 1.01 for the 5th and 95th percentile.
- Method and scope:
  - Sample of 776 European non-financial firms; over 2.5 million non-financial firms with loans at 81 largest European banks subjected to same stress scenarios.
  - Macroeconomic transmission channels of carbon pricing are not accounted for; analysis framed as near-term sensitivity.

### Impact summaries (conservative vs more realistic scenarios)
- Conservative scenario (costs increase by EUR 100 per tCO2; no pass-through):
  - 88% of firms not downgraded by even one notch.
  - Debt-weighted PD increases from 0.41% to 0.60%.
  - Aggregate CET1r projected to decline by 72 bps, from 14.9% to 14.2%.
- More realistic scenario (costs increase from EUR 3 to EUR 100 per tCO2; pass-through feasible):
  - Debt-weighted PD increases from 0.41% to 0.48%.
  - Aggregate CET1r impact is 17 bps (described as vanishingly small).
  - Some cleaner firms improve creditworthiness due to windfall profits.
- Heterogeneity effects:
  - Accounting for intra-sector heterogeneity leads to a 21% higher aggregate CET1r impact in both scenarios.
- Macro effects on output and unemployment are uncertain and not included.

### Three policy recommendations
- First:
  - When discussing negative fallout of carbon pricing, "see credit risk only where it’s due." Significantly higher carbon prices do not, in the short run, endanger financial stability via the corporate credit risk channel (long-term effects beyond scope).
- Second:
  - Banks and regulators should account for differential impacts between and within sectors.
  - Banks should assess forward-looking transition risk at the firm level (not rely on industry averages) and embed this in risk management, pricing, limits, and origination standards.
  - Supervisory authorities should investigate concentration risk where a bank’s portfolio is concentrated in the most affected sectors and react appropriately.
- Third:
  - Avoid treating emission intensities as the sole indicator for transition risk and do not overemphasize the role of the for-profit financial sector in accelerating the transition.
  - Pass-through of carbon costs depends on market structures and is not proportional to firms’ emission intensities.

*Italic: Content extracted from IMF Working Paper "Credit Risk Where It’s Due: Carbon Pricing and Firm Defaults" (selected sections).*

### introduction in 2011, the model has consistently shown excellent performance regarding discrim-

### IMF WORKING PAPERS — Credit Risk Where It’s Due: Carbon Pricing and Firm Defaults

### Carbon stress model (firm module)
- The model simulates the impact of a hypothetical CO2 price increase on firm creditworthiness by following four steps:
  - Determine Scope 1 and Scope 2 emissions for each company.
  - Compute additional costs stemming from a higher CO2 price.
    - In the raw stress scenario, these costs come on top of already existing expenses.
    - In the enhanced stress scenario, already paid expenses under the EU ETS (increase to 100 EUR) are recognized, Scope 2 costs are added, and firms are allowed to pass on carbon costs to consumers.
  - Run the financial projection in line with EU ETS and IFRS accounting rules.
    - Emission cost is incurred via the obligation to purchase permits for every tCO2 emitted during the production process in fiscal year t; allowances for expected emissions are bought and paid via cash and deposits within the same period (t) and certificates are kept in stock and submitted in the year after (t+1).
    - Carbon costs are assumed to be paid via the “cash and deposits” balance sheet item; if insufficient, companies borrow (increase indebtedness).
  - Use stressed positions as the basis for stressed credit risk ratings assuming an otherwise static balance sheet (no additional investments; profits and losses affected only by additional CO2 costs).
- Five financial ratios are stressed when modeling an increase in carbon price: EBIT, self-financing ability, net indebtedness ratio, return on cash flow, EBITDA - ROI. One ratio is not stressed: capital interest burden (total interest payments divided by total debt), assuming the average interest rate of a firm is not impacted by the carbon price shock.

### Bank module (translating firm PD impacts to banks)
- Firm-level PD impacts are translated into bank-specific capitalization-ratio changes by “stressing” banks’ credit exposure to non-financial firms.
- The effect is computed for:
  - the 776 firms with firm-level data available, and
  - all other firms recorded in AnaCredit (Eurosystem credit register) using extrapolated stressed PDs based on the observed link between emission intensity and PDs.
- Country-sector average emission intensities are used when firm-level emissions are not available; a Monte-Carlo simulation accounts for within-sector heterogeneity by combining observed dispersion in firm-level emission intensities with Eurostat country-sector averages to derive distributions for bank-level metrics.
- Assumptions for bank treatment:
  - Banks follow the internal rating-based (IRB) approach and IFRS to determine capital requirements and expected loss provisions.
  - The unstressed Common Equity Tier 1 ratio is CET1r = CET1 / RWA.
  - Using stressed PDs and stressed LGD, additional provisions and increased risk-weighted assets are obtained for each firm, resulting in a stressed CET1 ratio at bank level:
    - CET1r* = (CET1 − Σf∈firms ∆provisionsf) / (RWA + Σf∈firms ∆RWAf).
  - Households, government and financial-sector credit exposure as well as market risk exposure are left unstressed.
- Firms’ LGD is expected to increase in line with the PD according to Frye and Jacobs (2012); higher PD and LGD lead to higher expected losses (IFRS9 provisioning, including lifetime provisions) and increased risk weights (Basel IRB), decreasing numerator (available capital) and increasing denominator (RWA) of CET1r.

### Metrics (firm- and bank-level)
- Firm-level metrics:
  - Stressed PD factor F*PD = PD* / PD, measuring relative increase in PD.
    - Credit rating scales typically show an exponential link between rating grades and PDs; F*PD can be interpreted similarly to a rating migration (a one notch downgrade approximately corresponds with a PD factor of 1.5).
  - Stressed PD difference ∆PD* = PD* − PD, measured on an absolute scale; when multiplied by exposure, it measures change in expected loss.
- Bank-level metric:
  - ∆CET1r* = CET1r* − CET1r measures change in the Common Equity Tier 1 ratio after versus before the stress.
- Unless indicated otherwise, bank-level results present the averages of the Monte-Carlo simulation distributions.

### Data sources and coverage
- Firm-level:
  - Consolidated financial statements from European Records of IFRS Consolidated Accounts (2023); for fiscal year 2021, consolidated financial statements for 881 groups in participating countries (Austria, Belgium, France, Italy, Germany, Greece, Portugal, and Spain).
  - Emission data from ERICA, ISS ESG, and Carbon 4 Finance (Scope 1 and Scope 2). Reported information preferred to modeled; when multiple sources exist, the source closest to peer-group median is used.
  - Main firm-level sample used in the analysis: 776 firms for fiscal year 2021 (consolidated financial statement and carbon emissions available).
  - For firms without firm-level emissions, Scope 1 emission intensities at country-sector level (NACE level 1 and, for high-emitting subsectors, level 2) from Eurostat (2024) are used and scaled to derive Scope 2 emissions using observed firm-level factors and Eurostat electricity-sector data.
  - EU ETS emissions and costs: installations in EU ETS database matched to ORBIS to determine firm-controlled installations and free allowance allocations; for firms without financial statements, sector-level share of paid EU ETS emissions is used as proxy in enhanced stress scenario.
  - EU ETS coverage: 1.36 billion tons of CO2 equivalent were covered by the EU ETS in 2021 (European Environmental Agency (2023) cited).
- Bank-level:
  - Bank capitalization and portfolio composition from European Banking Authority (2021) transparency exercise (FINREP and COREP) for the largest 112 European banks; values used: CET1, total and corporate RWA, portfolio composition.
  - Granular borrower-lender exposure data from AnaCredit (Eurosystem-wide credit register); missing PDs imputed using country-sector averages. AnaCredit exposures scaled up to match consolidated exposures reported to EBA. Portfolio restricted to banks where at least 5% of RWA stem from corporate RWA and AnaCredit data are available, resulting in a sample of 81 banks.

### Stylized facts and sample statistics
- Firm-level sample structure (emission data by source and firm size; revenue thresholds preserved):
  - Small if revenue < EUR 250 Mio., Medium if EUR 250 Mio. ≤ revenue < EUR 1,500 Mio., Large if revenue > EUR 1,500 Mio.
  - Emission data by source and firm size:
    - Small: Corp. publication 47, Modeled 205, Total 252
    - Medium: Corp. publication 169, Modeled 79, Total 248
    - Large: Corp. publication 257, Modeled 19, Total 276
    - Total: Corp. publication 473, Modeled 303, Total 776
- Sector classification: 469 firms belong to climate policy relevant sectors (CPRS) and 307 firms to non-CPRS.
- Baseline creditworthiness: Of 776 entities, 469 reach investment grade on the S&P-equivalent rating scale (baseline predictions from rating model).
- Emission intensity (Scope 1 and Scope 1+2 normalized by revenue):
  - Kendall’s tau between Scope 1 and Scope 1+2 intensities = 0.75.
  - Median total intensity (tCO2/Mio EUR of revenue): fossil fuels 231, utilities 434.
  - Energy-intensive manufacturing median intensity 34 tCO2/Mio EUR (noting heterogeneity with some high observations).
  - Transportation median intensity 39 tCO2/Mio EUR.
  - Buildings median intensity 25 tCO2/Mio EUR.
  - Non-CPRS median intensity 17 tCO2/Mio EUR.
- Global direct (Scope 1) emissions and EU ETS coverage by sector (in mtCO2) — totals and shares preserved exactly:
  - Fossil-fuel: Scope 1 emissions 86, EU ETS share 67%, Free share 40%
  - Utility: Scope 1 emissions 346, EU ETS share 60%, Free share 2%
  - Energy-intensive: Scope 1 emissions 246, EU ETS share 37%, Free share 30%
  - Buildings: Scope 1 emissions 5, EU ETS share 5%, Free share 3%
  - Transportation: Scope 1 emissions 70, EU ETS share 20%, Free share 12%
  - Non-CPRS: Scope 1 emissions 16, EU ETS share 10%, Free share 7%
  - Total: Scope 1 emissions 770, EU ETS share 48%, Free share 16%
  - Interpretation: 770 MtCO2 emitted through direct emissions globally in 2021, 373 MtCO2 (48%) covered through the EU ETS; roughly one third of companies’ global emissions are already priced under the EU ETS.
- Bank-level sample statistics:
  - 81 banks report a total of EUR 7,016 bn in total risk-weighted assets.
  - CET1r varies between 8.4% and 38% across institutions; system-level CET1r is 14.9%.
  - Corporate credit risk exposure accounts for 43.7% of total RWA on average.
  - Share of exposure of the 776 IFRS groups in banks’ portfolios is on average 7%.

### Results — Impact on firm probabilities of default (raw stress scenario)
- Sample: 776 firms.
- Sectoral impacts (stressed PD factor F*PD = PD*/PD):
  - Fossil-fuel sector:
    - Median PD increase: 18%
    - Third-quartile PD increase: 66%
  - Utility sector:
    - Median PD increase: 19%
    - Third-quartile PD increase: 113%
  - All other sectors:
    - No widespread shift in median PD factor, but significant impact observed for individual companies.
- Presentation: Results shown as boxplots of the stressed PD factor by sector; distributions capture 25th, 50th, and 75th percentiles and whiskers as min/max or 1.5× interquartile range.

_Italic: Content extracted from IMF Working Paper "Credit Risk Where It’s Due: Carbon Pricing and Firm Defaults" (selected sections: introduction, carbon stress model, bank module, metrics, data, stylized facts, and initial results)._

### Appendix for a summary table.  Note:  the four outliers in the “Rubber, plastic and glass products,

### Appendix for a summary table.  Note:  the four outliers in the “Rubber, plastic and glass products,

### Heterogeneity within the energy-intensive sector
- A more granular subsector classification highlights the heterogeneity within the energy-intensive sector (Figure 5, Panel B).
- 236 firms in our sample belong to the energy-intensive sector.
- The sectors “metallic manufacturing and metal production”, “wood and paper industry” as well as “rubber, plastic and glass production, ceramics” record a strong increase due to the stress, comparable to the fossil fuel and utility sector.
- On aggregate, the debt-weighted sample PD increases from 0.41% to 0.60%.
- Measured on an absolute scale (Figure 6), PD increases are below 1 percentage point (pp) for almost all firms and below 0.1 pp for most firms in all but the utility sector.
- The “metal production and manufacturing of metal products” subsector shows a more widespread shift for the upper half of the 17 firms; baseline PDs in that sector are among the highest.
- Summary tables are provided in the Appendix (Table 5 and Table 6).

### Impact on firm PDs and rating migrations (raw stress scenario)
- The increase in PDs translates into mild rating migrations.
- Under the raw stress scenario, 88% of firms would not be downgraded.
- Downgrades typically affect firms with high absolute emissions: firms downgraded from investment grade to non-investment grade produce 43% of total emissions.
- In the utility sector, where emission intensities are highest, more than half of the 45 firms would be downgraded by at least one notch.
- In the fossil fuel sector, 8 of 16 companies would be associated with a downgrade.
- 17% of companies in the energy-intensive industries would be downgraded.
- The stress simulation indicates a drop in credit quality below investment grade for only 6% of the sample companies (from an initial 60% with investment grade).
- Measured in terms of debt, 4% of total volume migrates to a non-investment grade.
- Detailed statistics for downgrades are in Table 7 in the appendix.

### Enhanced stress scenario (accounting for incurred carbon costs and pass-through)
- Accounting for existing carbon costs and increased revenue because of cost pass-through reduces stress significantly: 91% of firms see their rating unchanged or upgraded.
- In the enhanced stress scenario, the aggregate debt-weighted PD increases from 0.41% to 0.48%, which marks a 12 bp lower stressed PD compared to the raw stress scenario.
- The simplified model yields a revenue boost of 18.4% across the board for energy producers, whereas the total additional carbon costs sum up to 3.6% of total revenue.
- Renewable energy suppliers do not see an increase in marginal costs while prices go up, driven by more polluting technologies (as evidenced by the European energy crisis in 2022).
- Modeling caveats: marginal cost pass-through and effective prices in the electricity market depend on demand and supply structure; firms may operate different technologies and contracts yielding further complexity.
- Counter-intuitive results highlighted:
  - Carbon pricing can be financially beneficial for carbon-efficient firms, creating transition opportunities.
  - The direct relationship between emission intensities and transition risk observed in the raw scenario vanishes on a cross-sectoral level in the enhanced scenario because utilities, on average higher emission intensity, are less exposed to carbon price induced credit risk.
- Table 8 in the Appendix shows detailed statistics on up- and downgrades.

### Impact of using firm-level data vs. country-sector averages
- Most countries lack standardized firm-level emission disclosures; standardized PDs are available only for the largest firms.
- To quantify estimation error from using country-sector-level averages, PD impact of the raw stress scenario is compared for firms where both firm-level and sector-average data are available.
- The scatter plot (Figure 8) compares PD impact using firm-level emissions data (x-axis) and sector average emissions (y-axis).
- The discrepancies are significant:
  - The average absolute deviation for the 776 firms is 0.19 pp.
  - Deviations range from 0.03 pp in the sector buildings to 1 pp for utilities.
- Estimation errors are particularly high in sectors with a wide range of emission intensities; electricity sector highlighted as inadequate for sector-average approximations.

### Impact on bank capitalization ratios (CET1r)
- The analysis translates PD increases into changes of the capitalization ratios of euro area banks for both raw and enhanced stress scenarios according to equation (5).
- The sample consists of 81 bank groups (consisting of 570 banks) under the ECB’s direct supervision with required disclosures; for the sample of 776 IFRS-groups, credit information for 704 IFRS-groups is included, encompassing 4,370 firm-bank relationships; dataset also includes approximately 3.3 million firm-bank relationships for other companies.

Raw stress scenario:
- The aggregate CET1r of the banking sector decreases by 72 basis points (bps).
- This shock is manageable and comparable to the impact of COVID-19 where CET1-ratio decreased from 15.6% in Q4 2019 to 15.0% in Q2 of 2020.
- Some banks are more affected: 19 banks experience a CET1r drop by 100 bps or more.
- Only one bank drops below a CET1r threshold of 10% due to the stress.
- Two sample banks already start with a CET1r below 10%.

Enhanced stress scenario:
- Under the enhanced stress scenario, the aggregate CET1r of the banking sector decreases by 19 bps.
- Five banks see their CET1 ratio increase under this scenario.
- Not a single bank falls below a capitalization ratio of 10% due to the stress.

### Drivers of the CET1r impact
- Higher PDs lead to both higher risk weights and higher credit risk provisions.
- Credit risk provisions are affected by:
  - Higher PDs (and LGDs) leading to higher expected losses.
  - IFRS9 staging: exposures with a significant increase in credit risk (e.g., significant PD increases) and impaired assets require lifetime provisions rather than one-year provisions.
- In the analysis, exposures that experience a PD increase of a factor of 2 or more are reclassified into Stage 2 and require lifetime provision.
- Aggregate staging effect:
  - Without staging, mean aggregate CET1r in the raw stress scenario would decrease by 51 bps (instead of 72 bps).
  - Without staging, mean aggregate CET1r in the enhanced stress scenario would decrease by 17 bps (instead of 19 bps).
- Base PDs drive the stress on firms and banks: higher base PDs (e.g., due to a sluggish economy) yield stronger impacts in terms of ∆PD*.
- The cyclical point of the economy determines stress impact via higher provisions and risk weights.

### Quantifying intra-sector heterogeneity effects on CET1r
- Accounting for intra-sector heterogeneity by using simulated firm-level emission data leads to a higher impact ∆CET1r* compared to the use of Eurostat sectoral emission data for almost all banks in both raw and enhanced scenarios.
- The deviation is caused by firm-level data deviating from Eurostat sectoral averages and the nonlinear impact of CO2 price shocks on PDs, risk weights, provisions, and risk-weighted assets.
- In absolute terms the deviation is more pronounced in the raw scenario with four banks having a mean difference over 30 bps.
- The relative deviation is nearly identical for raw and enhanced scenarios with an on average 21% higher ∆CET1r* impact in both scenarios.

### Bank-level drivers: size, portfolio intensity, and exposure composition
Bank size:
- Bank size does not appear significantly associated with the mean CET1r* impact from a carbon price increase.
- Setting aside the most affected banks, there is at best a small trend with bigger banks being marginally harder hit in the raw stress scenario; no such trend in the enhanced scenario.

Portfolio emission intensity:
- Banks with higher portfolio emission intensities tend to be hit harder by rising carbon prices.
- In the raw stress scenario there is a strong negative relationship between a bank’s portfolio emission intensity and its mean CET1r impact.
- Emission intensities are a very good proxy of carbon transition risk at both firm and bank portfolio levels under the raw scenario.
- In the enhanced stress scenario, market mechanics lead to windfall profits for carbon-efficient producers; the clear link between emission intensity and deterioration in credit quality vanishes at firm-level.
- Consequently, emission intensities are no longer a good proxy for short-term carbon transition risk in the enhanced scenario.
- Banks that consider emission intensity as a proxy and manage it via interest rates or lending limits may not always benefit cleaner firms compared to more polluting ones from other sectors.
- For comparability, Scope 1 emission intensity is used on the x-axis in Figure 13 for both scenarios; banks where energy producers account for over 10% of the exposure are highlighted.

*International Monetary Fund — Appendix for a summary table (from IMF Working Paper "Credit Risk Where It’s Due: Carbon Pricing and Firm Defaults")*

### conclusions’ robustness.

### conclusions’ robustness.

### Carbon price sensitivity
- Results are sensitive to the assumed carbon price increase.
- When assuming a higher increase in global carbon prices:
  - Worsening of creditworthiness measured on the rating grades roughly scale proportionally with the carbon price increase.
  - PDs increase exponentially.
- Example (from Figure 14 in the source):
  - If the carbon price increases by EUR 200 instead of EUR 100, the stress in terms of log(F*PD) will be 100% higher on average.
  - Moving from a global carbon price of EUR 3 per tCO2 (Parry, 2021) to EUR 200 is noted as highly unlikely and would strain underlying assumptions (no cost pass-through, static balance sheet structure, lack of macroeconomic impact).

### Pass-through modeling
- Impact of the stress on firms’ creditworthiness scales roughly with the pass-through rate.
- Model mechanics:
  - log(F*PD) decreases proportionally as the pass-through rate increases (see Figure 15 in the source).
  - The simple pass-through scheme applied for non-energy firms models pass-through as a firm-by-firm revenue compensation of 50% of the firms’ additional cost.
- Caveats:
  - This simple scheme may not reflect market realities.
  - Depending on market structure, pass-through might lead to net benefits for firms in industries other than energy production.
  - Charging higher prices will also impact demand for firms’ products.

### Financing additional carbon costs
- Results are robust to the financing employed by companies to pay the carbon price (running down cash reserves versus borrowing from banks).
- Quantitative comparison:
  - Deviation between both approaches for F*PD corresponds to a negligible factor of 0.99 and 1.01 for the 5th and 95th percentile.

### Discussion — scope, sample, and approach
- Sample and scope:
  - Sample of 776 European non-financial firms.
  - Over 2.5 million non-financial firms with loans at the 81 largest European banks are subjected to the same stress scenarios.
- Method:
  - Used a well-established probability of default (PD) model and the European credit register Anacredit.
  - Two scenarios analyzed.
  - Macroeconomic transmission channels of carbon pricing are not accounted for.

### Impact on firms (findings)
- Conservative scenario (costs of global direct and energy-related emissions increase by EUR 100 per tCO2; firms unable to pass on costs):
  - 88% of firms are not downgraded, even by one notch.
  - Debt-weighted PD of the sample of 776 firms increases from 0.41% to 0.60%.
- More realistic scenario (costs increase from EUR 3 to EUR 100 per tCO2; firms can pass on some carbon costs):
  - Debt-weighted PD of the sample increases from 0.41% to 0.48%.
  - Some cleaner firms benefit and see improved creditworthiness due to windfall profits from rising market prices.
- Conclusion:
  - Existence of both transition risks and transition opportunities from carbon pricing.
  - Aggregate firm-level impact is limited in the near-term analysis presented.

### Impact on banks (findings)
- Conservative scenario (no pass-through by firms):
  - Aggregate Common Equity Tier 1 capital ratio (CET1r) projected to decline by 72 basis points, from 14.9% to 14.2%.
  - This reduction is considered non-negligible but not expected to pose a systemic threat to banking system stability.
- More realistic scenario (cost pass-through feasible):
  - Impact on CET1r is 17 basis points (described as vanishingly small).
- Heterogeneity effects:
  - Accounting for intra-sector heterogeneity of emission intensity leads on aggregate to a 21% higher impact on the CET1r under both scenarios.
- Note:
  - Aggregate macroeconomic effects on output and unemployment are uncertain and not included; results are framed as a near-term sensitivity analysis.

### Three policy recommendations
- First policy implication:
  - When discussing negative fallout of carbon pricing, see credit risk only where it’s due. Significantly higher carbon prices do not, in the short run, endanger financial stability via the corporate credit risk channel (long-term effects beyond scope).
- Second policy recommendation:
  - Banks and regulators should account for differential impacts between and within sectors.
  - Banks should assess forward-looking transition risk at the firm-level (not rely on industry averages) and embed this in risk management, pricing, limits, and origination standards.
  - Supervisory authorities should investigate concentration risk where a bank’s portfolio is concentrated in the most affected sectors and react appropriately.
- Third policy recommendation:
  - Avoid treating emission intensities as the sole indicator for transition risk and do not overemphasize the role of the for-profit financial sector in accelerating the transition.
  - Passing-through carbon costs depends on market structures and is not proportional to firms’ emission intensities; consequently, transition risk will not necessarily scale with emission intensities and implementing a carbon price may not by itself shift financing away from high-emission ventures.

### Administrative notes
- Declaration of Competing Interest:
  - The authors report no potential conflict of interest.
- Funding:
  - This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
- Acknowledgements:
  - Thanks to Gerhard Winkler, Sabine Wukovits and participants in the OeNB research seminar, the workshop on “Climate change risk and credit assessment” hosted by Banca d’Italia (October 2021), and the MCM and ICD seminars at the International Monetary Fund for inputs and comments.

*Source: wpiea2025062-print-pdf - conclusions’ robustness.*

### References

### wpiea2025062-print-pdf - References

### Major referenced literature and data sources
- Environmental taxes and economic growth: Abdullah, S., Morley, B., 2014. Energy Economics 42, 27–33.
- Studies on carbon pricing, pass-through, windfall profits, and EU ETS: Alexeeva-Talebi (2011); Hobbie et al. (2019); Sijm et al. (2006); Cludius et al. (2020); Fabra & Reguant (2014); Keppler & Cruciani (2010).
- Climate stress tests and financial-sector analyses: Battiston et al. (2017); Alogoskoufis et al. (2021) ECB economy-wide climate stress test; Emambakhsh et al. (2023) The Road to Paris; Guth et al. (2021) OeNB climate risk stress test; Vermeulen et al. (2018) DNB Occasional Studies.
- Firm-level and bank-level credit risk under climate transition: Capasso et al. (2020); Carbone et al. (2021); Dumrose & Höck (2023); Nguyen et al. (2023); Ramos-García et al. (2023); Zhang & Zhao (2022).
- Methodology and policy discussion on carbon pricing: Hepburn et al. (2020); Parry (2021); Parry, Black, Roaf (2021); Metcalf & Stock (2023); Venmans et al. (2020).
- Data and institutional sources cited: Anacredit (European Central Bank, 2016); EBA transparency exercise (European Banking Authority, 2021); Eurostat; EU ETS data viewer (European Environmental Agency, 2023); World Bank State and Trends of Carbon Pricing (2023).

### Key methodological and data references used in the study
- Basel IRB formula and risk-weighting reference: Basel Comittee on Banking Supervision, 2023. CRE31.5 - Risk-weighted assets for corporate, sovereign and bank exposures.
- Credit assessment and rating systems: ICAS model and ECB Occasional Paper Series on in-house credit assessment systems (Auria et al., 2021); Leitner & Mayer (2015).
- IFRS 9 provisioning conventions and LGD–PD linkage: Frye & Jacobs (2012).
- Emissions and carbon-price reference values: EUR 60 per tCO2 corresponds to the average price of EUA allowances in 2021 (used in equation (8)).

### Appendix A.1 — Detailed firm-level statistics (selected exact values)
- Sample totals and CPRS classification (Table 4):
  - Total N = 776; Investment-grade share 60%
  - Fossil-fuel: N = 168; Investment-grade 1%
  - Utility: N = 45; Investment-grade 76%
  - Energy-intensive: N = 236; Investment-grade 67%
  - Buildings: N = 80; Investment-grade 53%
  - Transportation: N = 92; Investment-grade 59%
  - Other: Scientific R&D N = 1; Investment-grade 10%
  - Other: Other N = 296; Investment-grade 56%
- Distribution of the stressed PD factor F*PD in the raw stress scenario (Table 5) — selected statistics:
  - Total (N = 776): min 1.00; p10 1.00; p25 1.00; p50 1.01; p75 1.03; p90 1.23; max 57.89
  - Fossil-fuel (N = 161): min 1.00; p50 1.18; max 6.78
  - Utility (N = 45): max 30.23
  - Energy-intensive subsector examples: Rubber, plastic and glass, ceramics (N = 22): p75 1.88; p90 45.94; max 57.89
- Distribution of the stressed PD difference ∆PD* (in percentage points) in the raw stress scenario (Table 6) — selected statistics:
  - Total (N = 776): min 0.000; p10 0.000; p50 0.000; p75 0.000; p90 0.020; max 17.55
  - Fossil-fuel (N = 160): p75 0.030; p90 0.140; max 3.84
  - Rubber, plastic and glass, ceramics (N = 22): p90 11.27; max 17.55
- Downgrades per CPRS sector in the raw stress scenario (Table 7):
  - Total (N = 776): >= one notch 12%; >= three notches 3%; to non-investment grade 6%
  - Fossil-fuel (N = 165): >= one notch 50%; >= three notches 6%; to non-investment grade 15% (computed as share of firms starting with an investment-grade rating)
  - Utility (N = 45): >= one notch 53%; >= three notches 20%; to non-investment grade 35%
- Enhanced stress scenario downgrades/upgrades (Table 8):
  - Total (N = 776): Downgrades >= one notch 9%; to non-investment grade 3%; Upgrades >= one notch 4%
  - Utility (N = 45): Upgrades >= one notch 69%

### Appendix A.2 — Detailed bank-level statistics (Table 9)
- CET1 ratio (%) — mean 16.79; min 8.40; p25 13.47; median 15.66; p75 18.24; max 37.97
- Portfolio intensity — mean 134.11; min 3.60; p25 72.09; median 116.39; p75 162.38; max 554.29
- Share corporate RWA (%) — mean 43.70; min 6.00; p25 36.00; median 43.00; p75 52.00; max 83.00
- Share IFRS-groups (%) — mean 6.13; min 0.00; p25 1.68; median 3.93; p75 8.42; max 32.15

### Appendix A.3 — Impact methodology (exact equations and parameter values)
- Raw stress scenario (A.3.1):
  - Firms’ additional carbon costs computed as: ∆costs = 100×global Scope 1 emissions. (equation (6))
- Enhanced stress scenario (A.3.2):
  - Cost pass-through to revenue (equation (7)):
    - For most carbon intensive energy firm: ∆revenue = 90%×∆carbon costs
    - For other energy firms: ∆revenue = F_marginal × revenue (where F_marginal is defined as ∆revenue/revenue for the most carbon intensive NACE 35 firm)
    - For all other firms: ∆revenue = 50%×∆carbon costs
  - Firms’ additional carbon costs (equation (8)):
    - ∆carbon costs = 100×(Scope 1 emissions + 90%×Scope 2 emissions) − 60×(EU ETS emissions - EU ETS free allowances)
    - Note: EUR 60 per tCO2 corresponds to the average price of EUA allowances in 2021.
  - Overall costs considering pass-through (equation (9)):
    - ∆costs = ∆carbon costs − ∆revenue
- Bank-level aggregation and stressed CET1 computation (A.3.3):
  - Banks’ RWA split into RWA_i,f (exposures to the 776 firms), RWA_i,s (exposures to firms in other sectors), and RWA_other_i (RWAs not affected).
  - CET1 ratio formulation:
    - CET1r_i = CET1_i / (RWA_other_i + Σ_f RWA_i,f + Σ_s RWA_i,s)
    - Stressed CET1 ratio:
      - CET1r*_i = (CET1_i − Σ_f ∆p_i,f − Σ_s ∆p_i,s) / (RWA_other_i + Σ_f RWA*_i,f + Σ_s RWA*_i,s)
    - Where ∆p_i,f/s is the change in credit risk provisioning and RWA* are stressed RWA after the shock.
  - RWA_i,f and RWA*_i,f computed with Basel IRB risk weights using exposure-level PD and LGD inputs; LGD* inferred via Frye and Jacobs (2012) linking long-run and point-in-time PD and LGD parameters.
  - Provisioning follows IFRS 9: unimpaired exposures (Stage 1) provisioned over one year; Stage 2 and 3 over lifetime; an initial Stage 1 exposure moves to Stage 2 if its PD more than doubles because of the carbon pricing shock.
  - Use of Anacredit for exposure (GCA) and scaling procedures with FINREP and COREP to align sectoral exposure shares and prudential exposure values.
- Monte-Carlo simulation for firm-level emission intensity (A.3.4):
  - For the 2.5 mio. borrowers with no firm-level emissions, draw 1000 times from sectoral deviation factors to simulate firm-level emission intensity; an upper emission intensity threshold is applied to limit the stressed PD factor to 50.
  - Country-sector emission intensities used from Eurostat on NACE 1-digit level; if unavailable, EU-sector averages used; high-emitting subsectors of sections C and H use NACE 2-digit; steel and cement simulated on NACE 4-digit based on firm-level data set.
- Extrapolation of PD impacts (A.3.5):
  - Functional form: PD* = f(PD, EI, EUETS) (equation (10))
  - Expectation of exponential relationship between emission intensity and stressed PD factor: F*PD = PD*/PD.
  - Regression models estimated on the 776-sample for a carbon price shock of EUR 100:
    - Table 10 (Scenario coefficients; exact reported values):
      - Raw scenario dependent variable log(F*PD): intensity coefficient 0.0007*** (0.00002); log(PD) 1.0009***; non-paid intensity 0.0004***; paid intensity 0.0003***; constant 0.0090**; Observations 756; R2 0.6672.
      - Enhanced non-energy scenario dependent variable log(PD*): intensity coefficient 0.0007*** (0.00002); log(PD) 1.0104*** (0.0824); non-paid intensity 0.0004** (0.0002); paid intensity 0.0001 (0.0001); constant 0.0096; Observations 729; R2 0.9983.
      - Enhanced energy scenario dependent variable log(PD*): intensity coefficient 0.0007*** (0.00002); constant −0.5341 (0.4801); Observations 36; R2 0.8277.
    - Note markers: * p<0.1; ** p<0.05; *** p<0.01.
  - Robustness (Table 11) shows control variables (leverage, liquidity, profitability, size) add no additional statistically significant impact beyond intensity:
    - Example coefficients: leverage −0.0106 (not significant), liquidity −0.0229, profitability 0.0426.
    - Observations 756; R2 0.6686.
  - Summary statistics of controls (Table 12):
    - Leverage mean 0.38; min −2.61; p25 0.26; median 0.38; p75 0.51; max 0.97
    - Liquidity mean 0.15; min 0.00; p25 0.06; median 0.12; p75 0.19; max 0.92
    - Profitability mean 0.05; min −1.54; p25 0.02; median 0.05; p75 0.08; max 0.40
    - log(Total assets) mean 14.01; min 8.88; p25 12.50; median 13.86; p75 15.39; max 20.09
    - log(Revenue) mean 13.41; min 5.91; p25 11.98; median 13.39; p75 14.87; max 19.34

_Italic: Content derived from the PDF file "wpiea2025062-print-pdf - References"._

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025062-print-pdf.pdf_