## APPENDIX II: IMPULSE RESPONSE FUNCTIONS FOR ALTERNATIVE SVAR

## Source details

**Canonical URL:** [APPENDIX II: IMPULSE RESPONSE FUNCTIONS FOR ALTERNATIVE SVAR](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020232-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020232-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020232-print-pdf.pdf.json)

---

### I. Introduction
- “Transition risks” are the risks stemming from the transition to a “low-carbon economy” that emits fewer greenhouse gases (GHG).
- Transition risks can be driven by changes in policy, advances in technology, or shifts in market sentiment.
- Paper complements the IMF’s climate-related stress testing framework by introducing transition risk analysis focused on policy-driven transition channels.
- Two main questions:
  - How does a substantial increase in domestic carbon pricing impact banks’ credit exposures (loans) by affecting corporates’ operating costs and profitability under severe assumptions?
  - How does a drastic increase in global carbon taxes affect banks’ loan losses through a fall in the revenues of domestic oil producers?
- Exercise conducted in partial equilibrium and does not account for the use of revenues from higher carbon taxes or gains from expansion of clean energy sectors.

### II. Literature review — context and methodological precedents
- Mark Carney’s September 2015 Lloyd’s speech introduced taxonomy (physical vs transition risks) and stressed “stranded assets” and “unburnable reserves”.
- TCFD (2017) recommends scenario-based assessments for financial implications of climate risks.
- Representative empirical/modeling contributions:
  - Weyzig et al. (2014): banks’ losses equal to 0.4 percent of total assets in a rapid transition.
  - University of Cambridge Institute for Sustainability Leadership (2015): 5-year scenario with a global $100/ton CO2 carbon tax.
  - Battiston et al. (2016): network-based approach; second-round losses on average larger than first-round losses.
  - Dutch central bank framework (Vermeulen et al., 2018; 2019): four scenarios combining technological breakthroughs and policy stances; uses “transition vulnerability factors”.
  - NGFS (2020) Guide: practical guidance for climate-related financial risk assessment.
  - Bank of England and other authorities: insurer/insurer-investor stress tests with transition scenarios.

### III. Carbon pricing in Norway — current state
- Norway’s carbon pricing is advanced and complex; only Switzerland and Luxembourg estimated to have a smaller “carbon pricing gap” than Norway (OECD, 2018).
- Statistics Norway: average price paid per ton of CO2-equivalents in 2019 was NOK 419 (about USD 45).
- Global average carbon price reported as USD 2.
- IMF benchmark range to achieve Paris targets: USD 50-100.
- National emissions charges vary across sectors (example: domestic aviation up to NOK 750 per ton CO2-equivalents in 2019).
- Example policy change: January 2020 removal of carbon tax exemptions for the fishing sector.
- Conclusion: Norway is among leading jurisdictions in carbon pricing alignment with Paris goals, reducing but not eliminating rapid-transition risks.

### IV. Measuring the impact of higher domestic carbon prices: the “corporate cost of emissions” channel
- Method: firm-level balance sheet approach estimating whether additional carbon-costs impair firms’ ability to service debt and thereby affect banks’ debt-at-risk.
- Four hypothetical carbon-tax scenarios vary by:
  - Average level: $75 per ton CO2-equivalent and $150 per ton CO2-equivalent.
  - Differentiation: uniform carbon price vs parallel shift preserving current differentiation up to averages of $75 or $150.
- Key modeling assumptions:
  - ‘No-pass-through’: firms fully absorb increased costs (no change in output prices, quantities, or inputs); results represent an upper bound of earnings loss.
  - Firms are representative of their sectors, allowing sector-to-firm mapping.
- Emissions estimation:
  - Use Exiobase input-output tables to calculate CO2-equivalent GHG emissions per NOK of output for 163 sectors in 2016 — scope 1 emissions only.
  - Firm-level data: 2017 Orbis with NACE 2 sectoral breakdown (conversion to 85 sectors).
- Financial health metric:
  - Interest Coverage Ratio (ICR) = EBIT / Interest Expense.
  - Indicator: share of companies per sector for which ICR drops from between 1 and 2 to below 1.
- Sectoral vulnerabilities (from Figure 2):
  - A = Agriculture, forestry and fishing
  - E = Water supply, sewerage, waste management and remediation activities
  - H = Transportation and storage
  - D = Electricity, gas, steam and air conditioning supply (small impact due to high renewable share)
- Partial pass-through sensitivity (firms with ICR between 1 and 2 before the increase):
  - No-pass-through: 7.3 percent see ICR fall below 1.
  - 50 percent pass-through: 4.8 percent fall below ICR 1.
  - 70 percent pass-through: 3.4 percent fall below ICR 1.
- Aggregation to bank exposures (sector-level supervisory data) — Banks’ corporate debt at risk from higher carbon prices:
  - Uniform carbon price to $75 average:
    - drop below ICR 2: all banks 2.3%
    - drop below ICR 1: all banks 2.2%
    - most exposed bank: drop below ICR 2 = 9.0%; drop below ICR 1 = 9.1%
  - Uniform carbon price to $150 average:
    - drop below ICR 2: all banks 4.0%
    - drop below ICR 1: all banks 4.0%
    - most exposed bank: drop below ICR 2 = 15.9%; drop below ICR 1 = 15.8%
  - Parallel increase of price to $75 average:
    - drop below ICR 2: all banks 2.2%
    - drop below ICR 1: all banks 2.2%
    - most exposed bank: drop below ICR 2 = 8.1%; drop below ICR 1 = 8.1%
  - Parallel increase of price to $150 average:
    - drop below ICR 2: all banks 4.3%
    - drop below ICR 1: all banks 4.3%
    - most exposed bank: drop below ICR 2 = 16.0%; drop below ICR 1 = 16.1%
- Key interpretations:
  - Uniform carbon price hike to $150: about 4 percent of all corporate bank exposures see ICR drop below threshold.
  - Most affected bank can see around 16 percent of corporate exposures with ICR dropping below threshold.
  - Concentration in highly affected sectors increases bank-level exposure at risk.
  - Results similar under parallel vs uniform increases and for ICR thresholds of 2 vs 1.
- Caveat: assumption that a bank’s borrowers are representative of their sector; deviations can affect bank-level results.

### V. The impact of higher global carbon prices through the external demand channel
- Motivation: Norway’s reliance on oil and gas implies global carbon-price-driven demand changes materially affect Norwegian economy and financial stability.
- Norwegian oil sector size (2019):
  - around 14 percent of GDP
  - 19 percent of total investments
  - 21 percent of state revenues
  - 36 percent of total exports
- Modeling approach:
  - Model global supply and demand for oil to compute new equilibrium producer prices and quantities after exogenous global carbon prices.
  - Translate changes in oil sector revenues into impacts on Norwegian banks’ loan losses via a SVAR linking oil sector revenues to the broader economy and financial sector.
- Global market setup and parameters:
  - Rystad Energy data for global oil supply in 2018 based on breakeven cost curves.
  - Global oil demand price elasticity assumed -0.24.
  - Minimum CO2 content assumed 0.43 ton/barrel for tax per barrel calculations.
  - Example baseline oil price: $60 per barrel.
- Comparative statics results:
  - Global carbon price of $75 per ton CO2-equivalent:
    - tax per barrel = $31.4
    - equilibrium quantity reduced roughly 7 percent
    - consumer price increases about 36 percent
    - producer price falls about 16 percent
    - global producers face a 26.5 percent drop in revenues
    - estimated reduction in Norwegian oil revenues: 26.5 percent
  - Global carbon price of $150 per ton CO2-equivalent:
    - equilibrium quantities fall by over 13.5 percent
    - consumer price increases by more than 80 percent
    - producer price decreases by 23.4 percent
    - global oil producers face a reduction in revenues by more than 38 percent
    - estimated reduction in Norwegian oil revenues: more than 38 percent
- Contextual comparison:
  - During 2014-16 oil price downturn, Norwegian oil revenues fell by about 43 percent.
  - Carbon-price-driven decline may be seen as more permanent, potentially generating larger consumption and investment responses than a cyclical shock.
- Next step:
  - Introduce oil sector revenues into a structural vector-autoregression (SVAR) to assess interactions and estimate loan-loss impacts.

### A. Data and Methodology (SVAR)
- Time series and sample:
  - Quarterly data from Q1 2010 to Q2 2019.
  - Sources: Statistics Norway and the IMF World Economic Outlook database.
- Variables included in the baseline SVAR:
  - Oil sector revenues, real GDP, bank loan losses, the Norwegian Regional Network Survey, and the NOK/USD exchange rate.
  - Four lags of each variable; identification via a Cholesky decomposition with ordering: Oil sector revenues, Regional Network Survey (export-oriented oil sector category), GDP, bank loan losses, NOK/USD exchange rate.
- Transformations:
  - Oil revenues and GDP: first differences to induce stationarity.
  - Loan loss rates, survey results, and the exchange rate: included in levels.
- Role of the Regional Network Survey:
  - Summarizes views of executives from over 300 Norwegian enterprises and organizations; included to capture expectations for the oil sector and improve SVAR dynamics.
- Rationale and limitations:
  - Using the oil price directly is rejected because a carbon tax drives a wedge between consumer and producer prices.
  - Carbon tax data itself preferred, but consistent long historical series are not available.
  - Time-series variation dominated by the 2014-16 period of low oil prices.

### B. Results (SVAR: oil revenue shock → loan losses and related dynamics)
- Impulse responses:
  - Calculated to a 1 standard-deviation shock in oil revenues; horizontal axis in quarters.
  - Loan losses of banks and mortgage corporations show a significant reaction to shocks in oil revenues.
  - Model symmetric: a drop in oil revenues leads to an increase in loan losses.
  - Maximum reaction occurs after 6 quarters and has an average size of 0.2 percentage points.
  - GDP shocks have no significant impact on loan losses; exchange rate shocks reaction is insignificant; oil survey shocks have a small positive impact.
- Quantified impacts under carbon price scenarios:
  - Drop in oil revenues following a $75 carbon price:
    - Implied increase in loan losses of around 0.6 percentage points (0.5 percentage points assuming an initial oil price of $40/barrel).
  - Carbon prices rising to $150:
    - Implied increase in loan losses of 0.9 percentage points (0.8 percentage points assuming an initial oil price of $40/barrel).
- Benchmarks:
  - Average loss rate of the whole banking system in 2019: 0.35 percent.
  - Loss rate in Q4 2016: 0.88 percent.
- Robustness checks:
  - Alternative SVAR including investment and consumption (removing GDP) yields similar results; maximum point-estimate reaction of 0.26 percentage points.
  - Re-ordering endogenous variables does not change qualitative or quantitative results for loss rates.

### VI. CLIMATE POLICY-ENFORCED OUTPUT REDUCTION: THE PORTFOLIO CHANNEL
- Concept:
  - Climate policy could force reductions in scope 3 emissions by reducing oil-sector output.
  - Carbon Tracker Initiative estimate: Equinor must reduce overall output by about 45 percent before 2040 to be consistent with Paris targets.
- Carbon Tracker Initiative-based projected production/production-reduction requirements (rounded to nearest 5 percent):
  - Equinor: 45 percent
  - ExxonMobil: 55 percent
  - Shell: 10 percent
  - Chevron: 35 percent
  - BP: 25 percent
  - Total: 35 percent
  - Eni: 40 percent
  - ConocoPhillips: 85 percent
  - Petrobras: 65 percent
- Three-step analytical approach:
  1. Calculate impact on earnings for Norwegian oil majors given required output reductions (accounting for fixed and variable costs; assume no behavioral change).
  2. Use a dividend discount model to estimate change in firms’ share prices, assuming current market pricing reflects fair value.
  3. Estimate impact on Norwegian households’ and other institutions’ balance sheets assuming portfolios representative of the Oslo Børs All-share Index.
- Scenario definitions:
  - Scenario 1: Norwegian oil majors’ output drops by 45 percent due to unspecified policy mix that does not entail a change in oil prices.
  - Scenario 2: Output reduction partly achieved by increasing global carbon price to US$75.
  - Scenario 3: Output reduction partly achieved by increasing global carbon price to US$150.
- Equinor illustrative results and extrapolation:
  - A 45 percent production reduction could reduce net operating income by up to 80 percent while likely leaving Equinor profitable.
  - Carbon price of US$75–150 could further decrease output by approximately 8–12 percent.
  - Distributing production reduction linearly between 2020 and 2040 and adjusting income/dividends implies estimated change in share price of up to -50 percent.
- Spillover to broader equity market:
  - Other Norwegian equities correlate on average one-to-one with the oil sector over horizons up to one year.
  - Correlation of daily returns of the Oslo Børs Benchmark Index with the energy sector index over a 1-year horizon averaged 90 percent and never below 84 percent in the past 10 years.
- Portfolio impact estimates (selected figures):
  - Households (2018 holdings):
    - Directly held about 22 percent of their financial assets in domestic equity; another 3 percent indirectly via pension claims, life insurance policies, and investment fund shares.
    - Modeled reduction in oil production would lower the value of households’ financial assets by about 11 to 12 percent.
  - Insurers and pension funds:
    - Hold 10.6 percent of financial assets as domestic equity; potential drop in portfolio value around 5 percent.
  - Non-money-market investment funds:
    - Hold 11.7 percent of portfolios in domestic equity; potential drop between 5 and 6 percent.
  - Banks and mortgage corporations:
    - Direct holdings of domestic equity in 2018: 2.6 percent of financial assets.
- Table 3: Portfolio effects summary
  - Output reduction only:
    - Change in output of oil companies: -45%
    - Change in share price of oil companies: -43%
    - Impact on assets of Norwegian households: -11%
    - Impact on assets of Norwegian insurers and pension funds: -5%
    - Impact on assets of Norwegian non-money-market investment funds: -5%
  - Output reduction and $75 carbon price:
    - Change in output: -53%
    - Change in share price: -47%
    - Impact on households: -12%
    - Impact on insurers/pension funds: -5%
    - Impact on non-money-market investment funds: -5%
  - Output reduction and $150 carbon price:
    - Change in output: -57%
    - Change in share price: -49%
    - Impact on households: -12%
    - Impact on insurers/pension funds: -5%
    - Impact on non-money-market investment funds: -6%
- Caveats:
  - Analysis is static and preliminary; assumes sudden shock with no room for adaptation by agents.
  - Results conditional on assumptions about portfolio representativeness, dividend discount parameters, and no behavioral adjustments by firms.

### VII. CONCLUSIONS — key findings and implications
- Main questions revisited:
  - How does an increase in domestic carbon pricing impact banks’ credit exposures (loans)?
  - How does an increase in global carbon taxes affect banks’ loan losses via a fall in domestic oil producers’ revenues?
  - How would reductions in domestic oil production affect share prices and net wealth of domestic shareholders?
- Three main results:
  1. Domestic carbon price increase could cause inability to service debt for higher-emission firms when profits are low relative to interest expenses.
     - At a carbon price of $75 per ton CO2 equivalent, most at-risk sectors: agriculture, forestry and fishing; water supply, sewerage, waste management and remediation activities; and transportation and storage.
     - Banks’ debt at risk from an increase in carbon prices is small on average but can be significant when lending is concentrated in high-risk sectors.
  2. Global carbon price increases could lead to a fall in oil sector revenues causing a significant increase in banks’ loan losses.
     - Carbon price of US$75 estimated to increase loan loss rates by about 0.6 percentage points.
     - Carbon price of US$150 estimated to increase loan loss rates by 0.9 percentage points.
     - Compare to loss rate of 0.88 percent in Q4 2016.
  3. Climate policy curbing oil-sector total output could reduce valuations of oil producers, implying portfolio effects for Norwegian households and asset managers (estimated household asset value declines about 11 to 12 percent under scenarios considered).
- Broader methodological and policy observations:
  - Stress testing climate-related transition risks presents challenges across risk exposures, scenarios, models mapping shocks to impacts, and outcomes—particularly for multi-decade horizons.
  - Sharp increases in carbon prices are powerful and parsimonious for decarbonization scenarios, though politically contentious; technology and sentiment shocks are relevant but harder to model.
  - Transmission channels need more granular characterization; micro adjustments can propagate through value chains, suggesting usefulness of general equilibrium frameworks.
  - Combining partial-equilibrium micro approaches with richer macro/CGEs is a promising direction to better capture transmission channels and cross-border spillovers.
- Research implications:
  - Further investigation required to capture indirect exposures (e.g., real estate, consumers), second-round financial amplification, cross-border spillovers, and to develop methodologies suited to long horizons.

*Source: Excerpt from the chapter “APPENDIX II: IMPULSE RESPONSE FUNCTIONS FOR ALTERNATIVE SVAR” in the supplied IMF working paper content.*

### REFERENCES ____________________________________________________________________ 333

### REFERENCES

### References section
- Begins on page 333.

### APPENDIX I: GREENHOUSE GASES
- Titled "APPENDIX I: GREENHOUSE GASES".
- Begins on page 377.

*Source: wpiea2020232-print-pdf - REFERENCES ____________________________________________________________________ 333*

### APPENDIX II: IMPULSE RESPONSE FUNCTIONS FOR ALTERNATIVE SVAR __________ 388

### APPENDIX II: IMPULSE RESPONSE FUNCTIONS FOR ALTERNATIVE SVAR __________ 388

### I. Introduction
- “Transition risks” are the risks stemming from the transition to a “low-carbon economy” that emits fewer greenhouse gases (GHG).
- Transition risks can be driven by changes in policy, advances in technology, or shifts in market sentiment.
- The paper complements the IMF’s climate-related stress testing framework by introducing transition risk analysis focused on policy-driven transition channels.
- Two main questions addressed:
  - How does a substantial increase in domestic carbon pricing impact banks’ credit exposures (loans) by affecting corporates’ operating costs and profitability under severe assumptions?
  - How does a drastic increase in global carbon taxes affect banks’ loan losses through a fall in the revenues of domestic oil producers?
- Exercise is conducted in partial equilibrium and does not account for the use of revenues from higher carbon taxes or gains from expansion of clean energy sectors.

### II. Literature review — context and methodological precedents
- Mark Carney’s September 2015 Lloyd’s speech introduced taxonomy (physical vs transition risks), stressed “stranded assets” and “unburnable reserves”, and highlighted stress testing for climate-related exposures.
- TCFD (2017) recommends scenario-based assessments for financial implications of climate risks.
- Key empirical and modeling contributions summarized:
  - Weyzig et al. (2014): sensitivity analysis estimating banks’ losses under transition scenarios; finds banks’ losses equal to 0.4 percent of total assets in a rapid transition.
  - University of Cambridge Institute for Sustainability Leadership (2015): 5-year scenario analysis with a global $100/ton CO2 carbon tax in the ‘Two Degrees’ scenario; portfolio impacts vary by strategy.
  - Battiston et al. (2016): network-based approach capturing second-round effects; second-round losses on average larger than first-round losses.
  - Dutch central bank framework (Vermeulen et al., 2018; 2019): four scenarios combining presence/absence of technological breakthroughs and active/passive policy; uses multi-country macroeconometric model and “transition vulnerability factors”.
  - NGFS (2020) Guide on Scenario Analysis provides practical guidance for climate-related financial risk assessment.
  - Bank of England and other authorities conducting insurer/insurer-investor stress tests with transition scenarios.

### III. Carbon pricing in Norway — current state
- Norway’s carbon pricing is advanced, complex and evolving; only Switzerland and Luxembourg estimated to have a smaller “carbon pricing gap” than Norway (OECD, 2018).
- Statistics Norway: average price paid per ton of CO2-equivalents in 2019 was NOK 419 (about USD 45).
- Global average carbon price reported as USD 2.
- IMF benchmark range considered necessary to achieve Paris targets: USD 50-100.
- Carbon pricing varies substantially across industries: some industries covered only by national emissions charges, others by both national charges and EU ETS. National emissions charges vary across sectors (e.g., domestic aviation up to NOK 750 per ton CO2-equivalents in 2019).
- Carbon pricing evolves via annual reviews; example: January 2020 removal of carbon tax exemptions for the fishing sector.
- Conclusion: Norway is among leading jurisdictions in carbon pricing alignment with Paris goals, reducing but not eliminating rapid-transition risks.

### IV. Measuring the impact of higher domestic carbon prices: the “corporate cost of emissions” channel
- Approach: firm-level balance sheet method estimating whether additional carbon-costs impair firms’ ability to service debt and thereby affect banks’ debt-at-risk.
- Four hypothetical carbon-tax scenarios vary by:
  - Average level: $75 per ton CO2-equivalent (mid-point of High-Level Commission on Carbon Prices range) and $150 per ton CO2-equivalent (supporting more rapid transition).
  - Differentiation across sectors: uniform carbon price vs parallel shift preserving current differentiation up to averages of $75 or $150.
- Key modeling assumptions:
  - ‘No-pass-through’: firms fully absorb increased costs (no change in output prices, quantities, or inputs); results represent an upper bound of earnings loss.
  - Firms are representative of their sectors, allowing sector-to-firm mapping.
- Emissions estimation:
  - Use Exiobase input-output tables to calculate CO2-equivalent GHG emissions per NOK of output for 163 sectors in 2016 — scope 1 emissions only.
  - Firm-level data: 2017 Orbis with NACE 2 sectoral breakdown (conversion from Exiobase 163 sectors to NACE 2 with 85 sectors).
  - Current sector tax rates estimated by crossing sector tax payments with historical sector emissions.
- Financial health metric:
  - Interest Coverage Ratio (ICR) = EBIT / Interest Expense (EBIT = earnings before interest and taxes).
  - Indicator: share of companies per sector for which ICR drops from between 1 and 2 to below 1 (unable to cover interest expenses).
- Sectors most affected in Figure 2:
  - A = Agriculture, forestry and fishing
  - E = Water supply, sewerage, waste management and remediation activities
  - H = Transportation and storage
  - Rationale: combination of currently low carbon prices in some sectors and high carbon-intensity.
  - Small impact on D = Electricity, gas, steam and air conditioning supply due to high renewable share in Norwegian electricity.
- Partial pass-through sensitivity:
  - Of firms with ICR between 1 and 2 before the increase, 7.3 percent see ICR fall below 1 under no-pass-through.
  - Allowing 50 percent pass-through: 4.8 percent fall below ICR 1.
  - Allowing 70 percent pass-through: 3.4 percent fall below ICR 1.
- Aggregation to bank exposures (sector-level supervisory data):
  - Banks’ exposure at risk computed as share of total sectoral exposure for which ICRs drop below thresholds.
  - Table: Banks’ corporate debt at risk from higher carbon prices
    - Uniform carbon price to $75 average:
      - drop below ICR 2: all banks 2.3%
      - drop below ICR 1: all banks 2.2%
      - most exposed bank: drop below ICR 2 = 9.0%; drop below ICR 1 = 9.1%
    - Uniform carbon price to $150 average:
      - drop below ICR 2: all banks 4.0%
      - drop below ICR 1: all banks 4.0%
      - most exposed bank: drop below ICR 2 = 15.9%; drop below ICR 1 = 15.8%
    - Parallel increase of price to $75 average:
      - drop below ICR 2: all banks 2.2%
      - drop below ICR 1: all banks 2.2%
      - most exposed bank: drop below ICR 2 = 8.1%; drop below ICR 1 = 8.1%
    - Parallel increase of price to $150 average:
      - drop below ICR 2: all banks 4.3%
      - drop below ICR 1: all banks 4.3%
      - most exposed bank: drop below ICR 2 = 16.0%; drop below ICR 1 = 16.1%
- Key interpretations:
  - Following a uniform carbon price hike to $150, about 4 percent of all corporate bank exposures see ICR drop below threshold.
  - The most affected bank can see around 16 percent of corporate exposures with ICR dropping below the threshold.
  - Banks concentrated in highly affected sectors see higher exposures at risk versus diversified banks.
  - Results are little changed under parallel vs uniform increases and similar for ICR thresholds of 2 vs 1.
- Caveat: assumption that a bank’s borrowers are representative of their sector; deviations (e.g., banks lending to lower-emission firms) can affect bank-level results.

### V. The impact of higher global carbon prices through the external demand channel
- Motivation: Norway’s heavy reliance on oil and gas implies global carbon-price-driven demand changes can materially affect Norwegian economy and financial stability.
- Norwegian oil sector size (2019):
  - around 14 percent of GDP
  - 19 percent of total investments
  - 21 percent of state revenues
  - 36 percent of total exports
- Modeling approach:
  - Model global supply and demand for oil to compute new equilibrium producer prices and quantities after exogenous global carbon prices.
  - Translate changes in oil sector revenues into impacts on Norwegian banks’ loan losses via a SVAR linking oil sector revenues to the broader economy and financial sector.
- Global market setup and parameters:
  - Rystad Energy data for global oil supply in 2018 based on breakeven cost curves (global and Norwegian producers).
  - Global oil demand price elasticity assumed -0.24 (median from literature, Caldara et al. 2016).
  - Minimum CO2 content assumed 0.43 ton/barrel for tax per barrel calculations.
  - Example baseline oil price: $60 per barrel.
- Comparative statics results:
  - Global carbon price of $75 per ton CO2-equivalent:
    - tax per barrel = $31.4
    - equilibrium quantity reduced roughly 7 percent
    - consumer price increases about 36 percent
    - producer price falls about 16 percent
    - global producers face a 26.5 percent drop in revenues
    - estimated reduction in Norwegian oil revenues: 26.5 percent
  - Global carbon price of $150 per ton CO2-equivalent:
    - equilibrium quantities fall by over 13.5 percent
    - consumer price increases by more than 80 percent
    - producer price decreases by 23.4 percent
    - global oil producers face a reduction in revenues by more than 38 percent
    - estimated reduction in Norwegian oil revenues: more than 38 percent
- Contextual comparison:
  - During 2014-16 oil price downturn, Norwegian oil revenues fell by about 43 percent.
  - A carbon-price-driven decline may be seen as more permanent by agents, potentially generating larger consumption and investment responses than a cyclical oil price shock.
- Next step described:
  - Introduce oil sector revenues into a structural vector-autoregression (SVAR) to assess interactions between the oil sector, banks, and mortgage corporations and to estimate loan-loss impacts from a drop in oil sector revenues.
- Modeling assumptions reiterated:
  - Instantaneous changes in carbon and oil prices (comparative statics).
  - For Norway, consumer behavior assumed unchanged due to already-high domestic carbon taxes and high share of hydropower and electric vehicles.
  - Focus on producer-side adjustments for Norwegian producers.

*Italic: Source: Excerpt from the chapter “APPENDIX II: IMPULSE RESPONSE FUNCTIONS FOR ALTERNATIVE SVAR” in the supplied IMF working paper content.*

### introduction of a carbon tax. It would be preferable, of course, to use carbon tax data itself,

### wpiea2020232-print-pdf - introduction of a carbon tax. It would be preferable, of course, to use carbon tax data itself,

### A. Data and Methodology
- Time series and sample:
  - Quarterly data from Q1 2010 to Q2 2019.
  - Sources: Statistics Norway and the IMF World Economic Outlook database.
- Variables included in the baseline SVAR:
  - Oil sector revenues, real GDP, bank loan losses, the Norwegian Regional Network Survey, and the NOK/USD exchange rate.
  - Four lags of each variable; identification via a Cholesky decomposition with ordering: Oil sector revenues, Regional Network Survey (export-oriented oil sector category), GDP, bank loan losses, NOK/USD exchange rate.
- Transformations:
  - Oil revenues and GDP: first differences to induce stationarity.
  - Loan loss rates, survey results, and the exchange rate: included in levels.
- Role of the Regional Network Survey:
  - Summarizes views of executives from over 300 Norwegian enterprises and organizations on recent economic developments and the outlook ahead; included to capture expectations for the oil sector and improve SVAR dynamics.
- Rationale and limitations:
  - Using the oil price directly is rejected because a carbon tax drives a wedge between consumer and producer prices.
  - Carbon tax data itself preferred, but consistent long historical series are not available, motivating proxies.
  - The time-series variation is dominated by the 2014-16 period of low oil prices.

### B. Results (SVAR: oil revenue shock → loan losses and related dynamics)
- Impulse responses:
  - Calculated to a 1 standard-deviation shock in oil revenues; horizontal axis in quarters.
  - Loan losses of banks and mortgage corporations show a significant reaction to shocks in oil revenues.
  - By construction, model is symmetric: a drop in oil revenues leads to an increase in loan losses.
  - The maximum reaction occurs after 6 quarters and has an average size of 0.2 percentage points.
  - GDP shocks have no significant impact on loan losses; exchange rate shocks reaction is insignificant; oil survey shocks have a small positive impact.
- Quantified impacts under carbon price scenarios:
  - For a drop in oil revenues following a $75 carbon price:
    - Implied increase in loan losses of around 0.6 percentage points (0.5 percentage points assuming an initial oil price of $40/barrel).
  - For carbon prices rising to $150:
    - Implied increase in loan losses of 0.9 percentage points (0.8 percentage points assuming an initial oil price of $40/barrel).
- Benchmarks:
  - Average loss rate of the whole banking system in 2019: 0.35 percent.
  - Loss rate in Q4 2016: 0.88 percent (height of problems caused by the drop in the oil price between 2014 and 2016).
- Robustness checks:
  - Alternative SVAR specification including investment and consumption (removing GDP) yields similar results; maximum point-estimate reaction of 0.26 percentage points in that specification.
  - Re-ordering endogenous variables does not change qualitative or quantitative results for loss rates.

### VI. CLIMATE POLICY-ENFORCED OUTPUT REDUCTION: THE PORTFOLIO CHANNEL
- Concept and motivation:
  - Climate policy could force reductions in scope 3 emissions by reducing oil-sector output (via higher carbon taxes and supply-side measures).
  - Carbon Tracker Initiative estimate: Equinor must reduce overall output by about 45 percent before 2040 to be consistent with Paris targets.
- Table summary (Carbon Tracker Initiative-based projected production/production-reduction requirements):
  - Examples of required production reductions to 2040 (rounded to nearest 5 percent): Equinor: 45 percent; ExxonMobil: 55 percent; Shell: 10 percent; Chevron: 35 percent; BP: 25 percent; Total: 35 percent; Eni: 40 percent; ConocoPhillips: 85 percent; Petrobras: 65 percent.
- Three-step analytical approach:
  1. Calculate impact on earnings for Norwegian oil majors given the required output reductions (accounting for fixed and variable costs; assume no behavioral change).
  2. Use a dividend discount model to estimate change in firms’ share prices, assuming current market pricing reflects fair value.
  3. Estimate impact on Norwegian households’ and other institutions’ balance sheets assuming portfolios representative of the Oslo Børs All-share Index.
- Scenario definitions:
  - Scenario 1: Norwegian oil majors’ output drops by 45 percent due to unspecified policy mix that does not entail a change in oil prices.
  - Scenario 2: Output reduction partly achieved by increasing global carbon price to US$75.
  - Scenario 3: Output reduction partly achieved by increasing global carbon price to US$150.
- Equinor-specific illustrative results and extrapolation:
  - A 45 percent production reduction would likely leave Equinor profitable but could reduce net operating income by up to 80 percent.
  - Imposition of a carbon price of US$75–150 could further decrease output by approximately 8–12 percent.
  - Distributing production reduction linearly between 2020 and 2040 and adjusting income/dividends accordingly implies estimated change in share price of up to -50 percent (uncertainty from cost of capital and post-reduction growth assumptions).
- Spillover to broader equity market:
  - Other Norwegian equities correlate on average one-to-one with the oil sector over horizons up to one year.
  - In the past 10 years, correlation of daily returns of the Oslo Børs Benchmark Index with the energy sector index over a 1-year horizon averaged 90 percent and never below 84 percent.
- Portfolio impact estimates:
  - Households (2018 holdings):
    - On average held about 22 percent of their financial assets directly in domestic equity, plus another 3 percent indirectly via pension claims, life insurance policies, and investment fund shares.
    - If equity holdings are representative of the Oslo Børs All-share Index, the modeled reduction in oil production would lower the value of households’ financial assets by about 11 to 12 percent.
  - Insurers and pension funds:
    - Hold 10.6 percent of financial assets as domestic equity; potential drop in portfolio value around 5 percent.
  - Non-money-market investment funds:
    - Hold 11.7 percent of portfolios in domestic equity; potential drop between 5 and 6 percent.
  - Banks and mortgage corporations:
    - Direct holdings of domestic equity in 2018: 2.6 percent of financial assets; indirect effects via wealth declines of borrowers could be non-negligible.
- Table 3: Portfolio effects summary (scenarios and impacts)
  - Output reduction only:
    - Change in output of oil companies: -45%
    - Change in share price of oil companies: -43%
    - Impact on assets of Norwegian households: -11%
    - Impact on assets of Norwegian insurers and pension funds: -5%
    - Impact on assets of Norwegian non-money-market investment funds: -5%
  - Output reduction and $75 carbon price:
    - Change in output: -53%
    - Change in share price: -47%
    - Impact on households: -12%
    - Impact on insurers/pension funds: -5%
    - Impact on non-money-market investment funds: -5%
  - Output reduction and $150 carbon price:
    - Change in output: -57%
    - Change in share price: -49%
    - Impact on households: -12%
    - Impact on insurers/pension funds: -5%
    - Impact on non-money-market investment funds: -6%
- Caveats:
  - Analysis is static and preliminary; assumes sudden shock with no room for adaptation by agents.
  - Results conditional on assumptions about portfolio representativeness, dividend discount parameters, and no behavioral adjustments by firms.
  - Further work needed to integrate side-effects that could weaken or strengthen results.

### VII. CONCLUSIONS — key findings and implications
- Main questions addressed:
  - How does an increase in domestic carbon pricing impact banks’ credit exposures (loans)?
  - How does an increase in global carbon taxes affect banks’ loan losses via a fall in domestic oil producers’ revenues?
  - How would reductions in domestic oil production affect share prices and net wealth of domestic shareholders?
- Three main results:
  1. Domestic carbon price increase could cause inability to service debt for firms with higher emissions, especially when profits are low compared to interest expenses.
     - At a carbon price of $75 per ton CO2 equivalent, the sectors most at risk: agriculture, forestry and fishing; water supply, sewerage, waste management and remediation activities; and transportation and storage.
     - Banks’ debt at risk from an increase in carbon prices is small on average but can be significant when lending is concentrated in high-risk sectors.
  2. Global carbon price increases could lead to a fall in oil sector revenues causing a significant increase in banks’ loan losses.
     - Carbon price of US$75 estimated to increase loan loss rates by about 0.6 percentage points.
     - Carbon price of US$150 estimated to increase loan loss rates by 0.9 percentage points.
     - Compare to loss rate of 0.88 percent in Q4 2016.
  3. Climate policy curbing oil-sector total output could reduce valuations of oil producers, implying portfolio effects for Norwegian households and asset managers (estimated household asset value declines about 11 to 12 percent under scenarios considered).
- Broader methodological and policy observations:
  - Stress testing climate-related transition risks presents challenges across the four typical elements of a stress test: risk exposures, scenarios, models mapping shocks to impacts, and outcomes—particularly when horizons extend to decades.
  - Scenario choice: sharp increases in carbon prices are powerful and parsimonious for decarbonization scenarios, though politically contentious; other policy instruments and shocks (technology, sentiment) are relevant but harder to model.
  - Transmission channels need more granular characterization; micro adjustments at firm level can propagate through value chains and final demand, suggesting usefulness of general equilibrium frameworks.
  - Combining partial-equilibrium micro approaches with richer macro/CGEs is a promising direction to better capture transmission channels and cross-border spillovers.
- Research implications:
  - Further investigation required to capture indirect exposures (e.g., real estate, consumers), second-round financial amplification, cross-border spillovers, and to develop methodologies suited to long horizons.

*Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020232-print-pdf.pdf*

### REFERENCES

### REFERENCES

### Key citations and themes
- Major topics covered by the references:
  - Climate risk assessment of financial institutions and sovereign bond portfolios (e.g., Battiston et al.; EIOPA – Financial Stability Report, December 2019).
  - Climate-related stress testing and systemic risk analysis (e.g., Battiston et al., 2016; Battiston & Martinez-Jaramillo, 2018; Borio, Drehmann & Tsatsaronis, 2014).
  - Carbon bubble, stranded assets, and energy transition risks (Carbon Tracker Initiative, 2011; 2013; 2019; Weyzig et al., 2014).
  - Guides and reports for central banks and supervisors on climate scenario analysis (NGFS, 2020).
  - Fiscal and policy analysis on climate mitigation (IMF, 2019 Policy Paper; IMF, 2019 Fiscal Monitor).
  - Empirical and modelling work on CO2 price paths, climate-economy models, and declining CO2 price paths (Kent, Litterman & Wagner, 2019; Nordhaus, 1992; 1994; Nikas et al., 2019).
  - Financial-sector vulnerability analyses and stress testing (IMF, 2020 Financial Sector Assessment Program - Norway; Vermeulen et al., 2018; 2019).
- Representative referenced works (authors and years preserved as in source):
  - 2° Investing Initiative (2018)
  - Battiston S., Jakubik P., Monasterolo I., Riahi K., van Ruijven B. (2019)
  - Battiston, S., Mandel, A., Monasterolo, I., Schuetze, F., & Visentin, G. (2016)
  - Bolton, P. and Kacperczyk, M. (2020)
  - Borio, C., Drehmann, M., & Tsatsaronis, K. (2014)
  - CTI (2011); CTI (2013); CTI (2019)
  - Dunz, N., Naqvi, A. and Monasterolo, I. (2018)
  - ECB (2019)
  - Georgieva K. (2020)
  - High-Level Commission on Carbon Prices (2017)
  - IMF (2019); IMF (2019); IMF (2020)
  - Kent, D.D., Litterman, R.B. and Wagner, G. (2019)
  - McGlade, C., Ekins, P. (2015)
  - Monasterolo, I. and de Angelis, L. (2020)
  - NGFS (2020)
  - Nordhaus, W. D. (1992); Nordhaus, W. D. (1994)
  - NOU (2018)
  - Sims, C. A. (1992)
  - Sinn, H.-W. (2012)
  - TCFD (2017)
  - University of Cambridge Institute for Sustainability Leadership (2015)
  - Vermeulen, R., Schets, E., Lohuis, M., Kolbl, B., Jansen, D.-J. and Heeringa, W. (2018); (2019)
  - Weyzig F., Kuepper B., van Gelder J.W., van Tilburg R. (2014)
  - Wood, R., Stadler, K., Bulavskaya, T., et al. (2015)

### Appendix I: Greenhouse Gases — Definitions and metrics
- Definition:
  - Greenhouse gases absorb and emit radiation at specific wavelengths within the spectrum of terrestrial radiation emitted by the Earth’s surface, atmosphere, and clouds; this property causes the greenhouse effect.
  - Primary greenhouse gases listed: Water vapour (H2O), carbon dioxide (CO2), nitrous oxide (N2O), methane (CH4) and ozone (O3).
  - Human-made greenhouse gases also noted: halocarbons and other chlorine- and bromine containing substances (dealt with under the Montreal Protocol).
  - Kyoto Protocol gases additionally listed: sulphur hexafluoride (SF6), hydrofluorocarbons (HFCs) and perfluorocarbons (PFCs).
- Differences across GHGs:
  - Persistency and global warming potential (GWP) vary across gases.
  - Methane: described as more potent than CO2 in terms of global warming but with far shorter persistence—“around ten years”.
  - Carbon dioxide: average atmospheric residence time of “around 100 years”.
- CO2 equivalent (CO2-eq):
  - Definition preserved: “A carbon dioxide equivalent or CO2 equivalent, abbreviated as CO2-eq, is a metric measure used to compare the emissions from various greenhouse gases on the basis of their global-warming potential (GWP), by converting amounts of other gases to the equivalent amount of carbon dioxide with the same global warming potential.”
  - Conversion basis: generally based on a “100-year horizon (sometimes a 20-year one)” and CO2 is taken as unit of measure with GWP of 1 by definition.
  - Examples with exact figures:
    - Methane has a GWP of “96 over 20 years”.
    - Methane has a GWP of “32 over a 100-year horizon”.
- Notes and glossary sources:
  - IPCC, AR5 Climate Change 2013: The Physical Science Basis, Glossary (link preserved in source).
  - GWP explanation and 100-year GWP of CO2 = 1 (European Commission link preserved in source).

### Appendix II: Impulse Response Functions for Alternative SVAR — Key findings and model details
- Empirical takeaways (verbatim findings from figure captions/notes):
  - “Shocks to oil revenues have a significant and economically meaningful impact on loan losses.”
  - “Shocks to investment have a small positive impact on loan losses...”
  - “...while consumption shocks have a small negative impact”
  - “The reaction to FX shocks is insignificant.”
  - “A shock to the oil survey increases loan losses somewhat.”
- SVAR estimation details (preserved exactly as in source note):
  - “Impulse response functions are calculated with a structural vector-autoregression (SVAR) using quarterly data from 2010 to 2019.”
  - “Variables are transformed to induce stationarity and we include 4 lags.”
  - Identification:
    - “Identification is achieved through a Cholesky decomposition with the following ordering: Oil sector revenues, Regional Network Survey (export-oriented oil sector category), investment, consumption, bank loan losses, NOK/USD exchange rate.”

*Content unit: wpiea2020232-print-pdf - REFERENCES*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020232-print-pdf.pdf_