Module 4 · Lesson 4.5
Climate Stress Testing
Design, Execution, and Supervisory Practice
33Slides
8Charts
3Levels
1Case Study
5References
01
Overview
Where This Lesson Fits
This is the fifth of six lessons in Module 4: Green Finance Risk Management. Building on the carbon metrics and CVaR tools from Lesson 4.4, this lesson teaches how to stress-test an entire institution against climate scenarios — from design to capital impact assessment.
Pedagogical arc: “You have measured the risk. Now stress-test the institution's resilience.” Carbon metrics tell you HOW MUCH exposure you carry; stress tests tell you WHETHER your institution can survive it.
Learning Outcomes
- Distinguish climate stress tests from traditional stress tests on four dimensions
- Design a climate stress test using the five-element framework
- Calculate climate-adjusted PD, LGD, and expected losses under scenario stress
- Compute $\Delta CAR$ to assess institutional capital adequacy under climate scenarios
- Evaluate supervisory stress test frameworks across ECB, BoE, and ASEAN jurisdictions
Prerequisites
Lesson 4.4 (Carbon Metrics and Climate Value-at-Risk) required. Lessons 4.1 (Climate Risk Taxonomy), 4.2 (TCFD/TNFD Frameworks), and 4.3 (Scenario Analysis) recommended.
02
Slide Deck
Lecture Slides (33 Slides)
Download Slide Deck (PDF)03
Foundation Level
What Is a Climate Stress Test?
A climate stress test assesses a financial institution's resilience to climate-related shocks under specified scenarios. Unlike traditional stress tests that replay historical crises, climate stress tests project unprecedented future states using climate science and economic models.
Four Key Differences from Traditional Stress Tests
| Dimension | Traditional | Climate |
|---|---|---|
| Horizon | 3–5 years | 30+ years |
| Uncertainty | Parametric (known distributions) | Deep (no historical precedent) |
| Data | Historical loss data | Novel climate projections |
| Scope | Institution-level | Economy-wide systemic |
ASEAN ContextThe Bank of Thailand (BOT) launched its climate stress testing program in 2023, joining a growing list of ASEAN regulators catching up with the ECB and BoE who started several years earlier.
Five Design Elements
The Five-Element Framework
- Scenario Selection: Choose climate scenarios (NGFS Net Zero 2050, Delayed Transition, Hot House World) that define the macro trajectory of temperature, carbon prices, and policy actions over the projection horizon.
- Transmission Channels: Map how climate shocks transmit to financial variables — credit risk (PD/LGD changes), market risk (equity/bond repricing), and operational risk (supply chain disruption, compliance costs).
- Portfolio Mapping: Classify the entire balance sheet by sector, geography, and climate sensitivity. Assign each exposure to the relevant transmission channel and scenario pathway.
- Loss Estimation: Calculate expected losses per exposure using climate-adjusted PD, LGD, and EAD. Aggregate across the portfolio to derive total stressed provisions.
- Capital Impact: Compute post-stress capital adequacy ratio ($\Delta CAR$) and compare against regulatory minimums. Determine whether capital buffers are sufficient.
ASEAN ExampleThe BOT uses NGFS scenarios (step 1) calibrated with Thailand's 2011 flood data to model physical risk transmission channels (step 2), with particular focus on the agricultural sector.
Supervisory Landscape
Major Supervisory Exercises
| Regulator | Year | Scope | Key Finding |
|---|---|---|---|
| ECB | 2022 | 104 banks, 3 scenarios, 30-year horizon | EUR 70B aggregate losses |
| BoE CBES | 2021 | 19 banks/insurers, bottom-up + dynamic BS | GBP 225B cumulative losses |
| HKMA | 2021 | 20 banks, physical risk focus | Significant flood/typhoon exposure |
| MAS | 2022 | 8 D-SIBs, transition + physical | Material sector concentration risk |
| BOT | 2023 | Thai banking system, NGFS-based | Agriculture dominates physical losses |
ASEAN MomentumBy 2025, all major ASEAN regulators have either completed or launched pilot climate stress testing programs. Singapore (MAS) and Thailand (BOT) lead the region, with Vietnam (SBV), the Philippines (BSP), and Indonesia (OJK) actively developing frameworks.
Top-Down vs. Bottom-Up Approaches
Comparison
| Dimension | Top-Down | Bottom-Up |
|---|---|---|
| Granularity | Aggregate sector/portfolio level | Individual borrower level |
| Models | Uniform regulator-supplied | Bank's own internal models |
| Speed | Fast, standardized | Slow, resource-intensive |
| Accuracy | Lower (misses borrower heterogeneity) | Higher (captures idiosyncratic risk) |
| Examples | ECB 2022, HKMA 2021 | BoE CBES 2021, MAS 2022 |
Best practice: Combine both approaches. Use top-down for system-wide comparability and bottom-up for institution-specific accuracy.
ASEAN PracticeThe BOT uses a top-down approach for speed and system-wide coverage, as most Thai banks lack the internal models needed for bottom-up analysis. MAS uses a combined approach, leveraging Singapore's D-SIBs' more advanced risk infrastructure.
Reading a Stress Test Output
Interpreting Results: Pre-Stress vs. Post-Stress CAR
The central output of any climate stress test is the change in Capital Adequacy Ratio (CAR) from pre-stress to post-stress levels.
Capital Adequacy Ratio
$$CAR = \frac{\text{Tier 1 Capital}}{\text{RWA}}$$What Happens When CAR Breaches 10.5%?
- Capital action: The bank must raise additional Tier 1 capital (equity issuance, retained earnings)
- Risk reduction: Divest high-risk exposures, reduce lending to carbon-intensive sectors
- Deleveraging: Shrink the balance sheet by not renewing maturing loans, selling assets
The regulatory minimum of 10.5% comprises the 8% Basel III minimum plus the 2.5% capital conservation buffer. Systemically important banks face additional surcharges.
ASEAN BenchmarksTypical Thai bank pre-stress CAR: ~15–17%. Under NGFS Hot House World: ~11–13%. The 2–4 percentage point drop consumes most of the buffer above the 10.5% regulatory floor, leaving limited headroom for additional shocks.
04
Intermediate Level
Transmission Channel Modeling
Credit Risk Channel
PD and LGD adjustments under climate scenarios. The core formula:
Climate-Adjusted PD
$$PD_{\text{climate}} = PD_{\text{base}} \times (1 + \text{stress\_multiplier})$$The stress multiplier captures the incremental default probability driven by physical damage, transition costs, or policy shocks specific to each sector and scenario.
Market Risk Channel
- Equity repricing: Coal equities −40% under Net Zero, oil & gas −25%, renewables +15%
- Bond spreads: Carbon-intensive issuers +50–200 bps spread widening
- Real estate: Flood-zone properties −10–30% in collateral value
Operational Risk Channel
- Supply chain: Disruption from extreme weather events affecting borrower operations
- Compliance: Increasing regulatory burden and reporting requirements
- Litigation: Climate-related lawsuits up +300% since 2015 (Grantham Research Institute)
Worked Example (Thai Agriculture Sector)
Thai agriculture sector: $PD_{\text{base}} = 5\%$, stress multiplier $= 0.8$ (from NGFS Hot House physical risk calibration).
$$PD_{\text{climate}} = 5\% \times (1 + 0.8) = 5\% \times 1.8 = 9\%$$
ASEAN ContextThe stress multiplier for Thai agriculture is among the highest in the ASEAN region due to flood vulnerability, monsoon dependence, and low crop insurance penetration (~7%).
Expected Loss Under Climate Stress
Expected Loss Formula
$$EL = PD \times LGD \times EAD$$Five-Sector Worked Example (Thai Mid-Size Bank)
| Sector | EAD (THB B) | PDclimate | LGD | EL (THB B) |
|---|---|---|---|---|
| Agriculture | 90 | 9.0% | 65% | 5.27 |
| Manufacturing | 120 | 6.5% | 50% | 3.90 |
| Real Estate | 100 | 4.5% | 55% | 2.48 |
| Power | 60 | 8.0% | 45% | 2.16 |
| Services | 130 | 3.5% | 40% | 1.82 |
Total Stressed Provisions
$$\text{Provisions} = \sum_i EL_i = 5.27 + 3.90 + 2.48 + 2.16 + 1.82 = \text{THB } 15.63\text{B}$$ASEAN ContextFor a representative Thai mid-size bank, total stressed provisions of THB 15.63B (~$450M) under the NGFS Hot House scenario would require drawing on capital buffers. Agriculture and manufacturing together drive ~59% of total climate-stressed losses.
Capital Adequacy Impact
Delta CAR
$$\Delta CAR = \frac{\text{Regulatory Capital} - \text{Climate Losses}}{\text{RWA}}$$Worked Example
- Pre-stress CAR: 14.2% (Regulatory Capital THB 56.8B / RWA THB 400B)
- Climate losses: THB 18.5B (from Expected Loss aggregation)
- Post-stress capital: THB 56.8B − THB 18.5B = THB 38.3B
- Post-stress CAR: THB 38.3B / THB 400B = 9.6%
The result: post-stress CAR of 9.6% breaches the 10.5% regulatory minimum (Basel III minimum 8% + 2.5% conservation buffer). This triggers mandatory capital action.
Capital Headroom Analysis
Pre-stress headroom: 14.2% − 10.5% = 3.7 percentage points. Climate losses consume 4.6 percentage points, exceeding the buffer by 0.9pp.
ASEAN ContextThai banks typically have 2–4 percentage point headroom above the 10.5% minimum. Under Hot House World, this headroom may be fully consumed. Banks with narrower buffers (common among smaller Thai commercial banks) face the greatest capital adequacy pressure.
Data Proxies and Gap-Filling
Four-Tier Data Hierarchy
| Tier | Data Source | Availability |
|---|---|---|
| Tier 1 | Borrower-level climate exposure data | <10% of ASEAN portfolios |
| Tier 2 | Sector-geography proxy data | ~15% coverage |
| Tier 3 | Sector average stress parameters | ~35% coverage |
| Tier 4 | Expert judgment and qualitative assessment | Remainder (~40%) |
ASEAN-Specific Challenges
- SME dominance: 80%+ of borrowers are SMEs with no climate data reporting capability
- Flood maps unlinked: National flood risk maps exist but are not linked to banking exposure databases
- Different proxy methods: Each regulator uses different gap-filling approaches, making cross-country comparison difficult
Improvement Pathway
Mandatory climate reporting → linked geospatial-financial databases → borrower-level exposure data. Timeline: 3–5 years for meaningful coverage improvement in leading ASEAN jurisdictions.
Worked Example: Thai Mid-Size Bank
Bank Profile
THB 500B total assets, 6 sectors, NGFS Delayed Transition scenario.
Results by Sector
| Sector | EAD (THB B) | ΔPD | ΔLGD | ELstressed (THB B) |
|---|---|---|---|---|
| Agriculture | 80 | +4.0pp | +10pp | 6.00 |
| Manufacturing | 110 | +3.5pp | +5pp | 4.95 |
| Real Estate | 90 | +2.0pp | +8pp | 2.70 |
| Power | 50 | +5.0pp | +5pp | 3.25 |
| Transport | 40 | +3.0pp | +5pp | 1.60 |
| Services | 130 | +1.5pp | +3pp | 2.47 |
Capital Impact Summary
- Total stressed losses: THB 20.97B (~THB 22B rounded)
- Pre-stress CAR: 15.1%
- Post-stress CAR: 12.3%
- Headroom above 10.5%: 1.8 percentage points
Key FindingAgriculture and manufacturing together drive ~52% of total stressed losses (THB 10.95B out of THB 20.97B). Despite having only 38% of total EAD, these two sectors face the highest stress multipliers due to physical vulnerability and transition exposure respectively.
05
PhD Extension
Bottom-Up Stress Testing Methodology
Logistic PD Model with Climate Covariates
$$\text{logit}(PD_i) = \alpha + \beta_1 X_i^{\text{fin}} + \beta_2 X_i^{\text{climate}} + \varepsilon_i$$Climate Covariates
- Flood frequency: Number of major flood events per decade in borrower's operating region
- Drought index: Standardized Precipitation-Evapotranspiration Index (SPEI) for agricultural borrowers
- Carbon price: Shadow carbon price under NGFS scenarios ($0–250/tCO₂)
- Energy transition index: Composite measure of renewable energy adoption and fossil fuel phase-out pace
Estimated Coefficients
| Covariate | $\beta$ | t-stat | Marginal Effect |
|---|---|---|---|
| Flood frequency | 0.45 | 3.8 | +2.1pp PD per unit |
| Drought index | 0.38 | 3.1 | +1.7pp PD per unit |
| Carbon price | 0.32 | 2.9 | +1.4pp PD per $50/t |
| Energy transition | 0.28 | 2.4 | +1.2pp PD per unit |
Research QuestionCan machine learning models (gradient boosting, neural networks) improve PD prediction under climate scenarios compared to logistic regression? How do non-linear interactions between climate covariates affect out-of-sample predictive accuracy?
Dynamic Balance Sheet Assumptions
Static vs. Dynamic Balance Sheet
| Assumption | Static | Dynamic |
|---|---|---|
| Portfolio composition | Frozen at test date | Evolves with management decisions |
| Management response | None (passive) | Rebalancing, divestment, capital raising |
| Loss estimate | Higher (worst-case) | Lower (15–30% reduction) |
| Comparability | High (uniform) | Low (depends on assumed behavior) |
| Used by | ECB 2022 (for comparability) | BoE CBES 2021 (for realism) |
Impact on Results
Dynamic balance sheet assumptions typically reduce estimated losses by 15–30% compared to static assumptions, because management is assumed to divest high-risk assets, raise capital, and restructure portfolios over the 30-year horizon. However, this introduces subjectivity about future management decisions.
Research QuestionHow do management adaptation assumptions affect cross-bank comparability? Can a standardized set of adaptation rules (e.g., “divest X% of coal exposure by 2035”) preserve comparability while incorporating dynamic realism?
Second-Round Effects
The Amplification Cycle
- Climate Shock: Physical or transition event causes direct losses to exposed borrowers
- Bank Losses: Credit losses reduce bank capital, impairing lending capacity
- Credit Contraction: Banks tighten lending, reduce credit supply to the real economy
- GDP Decline: Reduced credit availability causes further economic contraction, which feeds back to stage 1
Three Amplification Channels
- Interbank contagion: Bank A's losses reduce its ability to meet interbank obligations, transmitting stress to Bank B
- Fire-sale externalities: Distressed asset sales depress market prices, forcing mark-to-market losses at other institutions
- Real economy feedback: Credit contraction reduces investment, employment, and output, increasing default rates across the economy
Magnitude: Second-round effects amplify first-round losses by +20–50% (Battiston et al. 2017). Most current supervisory stress tests do not capture these effects, meaning they systematically underestimate total system losses.
Research QuestionCan agent-based models capture non-linear amplification dynamics in ASEAN banking networks? How do interconnection patterns between ASEAN banks (cross-border lending, shared exposures) affect systemic amplification?
Comparing Supervisory Approaches
Detailed Cross-Jurisdictional Comparison
| Dimension | ECB | BoE | BOT | MAS |
|---|---|---|---|---|
| Scope | 104 banks | 19 banks/insurers | Thai banking system | 8 D-SIBs |
| Methodology | Top-down, static BS | Bottom-up, dynamic BS | Top-down, NGFS-based | Combined approach |
| Scenarios | 3 NGFS (30yr) | 3 custom (30yr) | NGFS + 2011 flood | 2 NGFS + local |
| Key finding | EUR 70B losses | GBP 225B losses | Agri dominates | Sector concentration |
| Limitation | No 2nd-round | Complex, costly | Limited data | Small sample |
Key Insight
The ECB exercise has the largest scale (104 banks), the BoE the most innovative methodology (dynamic balance sheet, bottom-up), and the BOT the best ASEAN-specific calibration (2011 flood anchor). No single approach is clearly superior — each reflects its jurisdiction's priorities and constraints.
Research QuestionWhat would a unified ASEAN climate stress testing framework look like? How could it balance cross-country comparability with country-specific calibration (e.g., Thailand's flood risk vs. Philippines' typhoon risk vs. Vietnam's sea-level rise)?
Research Frontier: Acharya et al. (2023)
Four Arguments Why Current Stress Tests Underestimate Risk
- Tail risks underestimated: Current tests use scenario means, not worst-case tails. Climate damage distributions are fat-tailed, meaning extreme outcomes are far more likely than normal distributions suggest.
- Non-linearities ignored: Climate impacts are non-linear (e.g., crop yield drops sharply above 2°C warming, not gradually). Linear stress multipliers miss these threshold effects.
- Network effects omitted: First-round tests ignore interbank contagion, fire sales, and credit contraction feedback loops that amplify losses by 20–50%.
- No historical precedent: Climate change is unprecedented — there is no historical crisis that can serve as an adequate calibration anchor for 30-year projections.
Proposed Improvements
- Agent-based models: Simulate individual bank behavior and interaction to capture emergent systemic risk
- Extreme Value Theory (EVT): Use fat-tailed distributions instead of Gaussian assumptions for loss modeling
- Multi-model ensembles: Run stress tests across multiple climate models and average results to reduce model uncertainty
ASEAN Application
Network effects may be larger in interconnected ASEAN banking systems where cross-border exposures (Thai banks in Vietnam, Singaporean banks across ASEAN) create contagion channels that single-jurisdiction tests miss entirely.
Research QuestionHow can climate stress tests incorporate tipping points (permafrost melt, ice sheet collapse, Amazon dieback) and cascading failures? What are the implications of irreversibility for capital adequacy planning?
06
Quantitative Lab
Lab: Simplified Climate Stress Test on an ASEAN Bank Portfolio
Task:Execute a simplified climate stress test on a 1,000-loan ASEAN bank portfolio
Method:Python (pandas, numpy, matplotlib)
Output:4-panel dashboard: sector EL breakdown, pre/post CAR comparison, PD shift waterfall, scenario comparison
Step-by-Step Instructions
- Load portfolio: Generate or load a 1,000-loan portfolio with sector, EAD, PD_base, LGD_base, and physical_exposed flag per loan
- Select NGFS scenario: Choose one of Net Zero 2050, Delayed Transition, or Hot House World; load corresponding sector stress multipliers
- Apply PD/LGD shocks: Calculate climate-adjusted PD and LGD per loan using scenario-specific multipliers and physical exposure flags
- Calculate expected loss: Compute $EL_i = PD_i^{\text{climate}} \times LGD_i^{\text{climate}} \times EAD_i$ per loan and aggregate
- Compute post-stress CAR: Subtract total provisions from CET1 capital, divide by RWA
- Generate 4-panel dashboard: Sector EL bar chart, pre/post CAR gauge, PD shift waterfall, Net Zero vs Hot House comparison
Python Pseudocode (Steps 3–5)
# Step 3: Apply climate-adjusted PD and LGD
df['pd_climate'] = df['pd_base'] * (1 + df['stress_multiplier'])
df['lgd_climate'] = np.where(df['physical_exposed'],
df['lgd_base'] + df['damage_fraction'], df['lgd_base'])
# Step 4: Calculate stressed expected loss
df['el_climate'] = df['pd_climate'] * df['lgd_climate'] * df['ead']
total_provisions = df['el_climate'].sum()
# Step 5: Compute post-stress CAR
post_stress_cet1 = cet1_capital - total_provisions
car_post = post_stress_cet1 / rwa
df['pd_climate'] = df['pd_base'] * (1 + df['stress_multiplier'])
df['lgd_climate'] = np.where(df['physical_exposed'],
df['lgd_base'] + df['damage_fraction'], df['lgd_base'])
# Step 4: Calculate stressed expected loss
df['el_climate'] = df['pd_climate'] * df['lgd_climate'] * df['ead']
total_provisions = df['el_climate'].sum()
# Step 5: Compute post-stress CAR
post_stress_cet1 = cet1_capital - total_provisions
car_post = post_stress_cet1 / rwa
Post-Stress Capital Adequacy
$$CAR_{\text{post}} = \frac{CET1 - \sum_i EL_i^{\text{climate}}}{\text{RWA}}$$Data Sources
07
Case Study: Thailand — BOT Climate Stress Testing Framework
Context: The 2011 Thai Floods and Their Legacy
Thailand's 2011 floods caused $46.5 billion in damages, inundated 800+ factories in the industrial belt north of Bangkok, and drove banking sector NPLs from 2.9% to 3.8%. This event served as the anchor calibration for the Bank of Thailand's climate stress testing program launched in 2023.
BOT Framework Design
- Scenarios: NGFS Net Zero 2050, Delayed Transition, and Hot House World
- Horizon: 2050 (27-year projection from 2023 baseline)
- Sector granularity: Agriculture, manufacturing, real estate, power, transport, services
- Physical risk anchor: 2011 flood recalibrated using climate projections (50-year return period compressed to 20-year under Hot House)
Calibration: From 2011 to Hot House
The BOT scales the 2011 flood impact forward: if a 50-year flood becomes a 20-year flood under Hot House, the annualized loss probability increases by 2.5×. PD scaling follows: $PD_{\text{HotHouse}} = PD_{2011} \times 2.5$ for physically exposed sectors.
Results: Thai Banking System
| Scenario | Total Losses | Top Sector | Pre-CAR | Post-CAR |
|---|---|---|---|---|
| Hot House World | THB 28B | Agriculture (THB 15B) | 16.2% | 12.8% |
| Net Zero 2050 | THB 19B | Power (THB 7B) | 16.2% | 14.1% |
Under Hot House World, losses are concentrated in agriculture (THB 15B), manufacturing (THB 7B), real estate (THB 4B), and other sectors (THB 2B). Total assets: THB 600B.
Key Numbers
Hot House: Agriculture physical losses THB 15B (~$430M) dominate total losses
Net Zero: Power transition losses THB 7B (~$200M) lead, as coal phase-out strands assets
The physical risk scenario (Hot House) produces 47% higher losses than the transition scenario (Net Zero)
Net Zero: Power transition losses THB 7B (~$200M) lead, as coal phase-out strands assets
The physical risk scenario (Hot House) produces 47% higher losses than the transition scenario (Net Zero)
Framework Evaluation
| Strengths | Gaps |
|---|---|
| Uses NGFS scenarios (global comparability) | No second-round effects modeled |
| 2011 flood provides empirical anchor | Limited borrower-level data (~15% Tier 1–2) |
| Agriculture sector granularity | Static balance sheet assumption |
| Regular update cycle planned | No interbank contagion channel |
| Integration with BOT supervisory review | Cross-border exposures not captured |
Discussion Questions
- Why does the Hot House World scenario produce higher losses than Net Zero for Thai banks, despite Net Zero imposing carbon pricing?
- How would incorporating second-round effects (credit contraction, fire sales) change the post-stress CAR estimates?
- What are the limitations of using 2011 flood data as a calibration anchor for 30-year climate projections?
- How should the BOT adapt its framework to capture cross-border climate risk transmission (e.g., Thai banks' exposures to Vietnamese and Cambodian agriculture)?
08
Key Concepts
Glossary
- Bottom-up stress test
- Approach where individual banks use their own internal models to estimate climate-stressed losses at the borrower level
- CAR (Capital Adequacy Ratio)
- Tier 1 Capital divided by Risk-Weighted Assets; regulatory minimum is 10.5% (Basel III 8% + 2.5% conservation buffer)
- CBES
- Climate Biennial Exploratory Scenario — the Bank of England's climate stress testing exercise for banks and insurers
- CET1 (Common Equity Tier 1)
- Highest quality regulatory capital, consisting of common shares and retained earnings; absorbs losses first
- Climate stress test
- Assessment of a financial institution's resilience to climate-related shocks under specified scenarios over multi-decade horizons
- $\Delta CAR$
- Change in Capital Adequacy Ratio from pre-stress to post-stress: $(Regulatory Capital - Climate Losses) / RWA$
- Dynamic balance sheet
- Stress test assumption where portfolio composition evolves with management decisions (divestment, rebalancing, capital raising)
- EAD (Exposure at Default)
- Total amount of exposure to a borrower at the time of default; used in Expected Loss calculation
- Expected Loss (EL)
- $EL = PD \times LGD \times EAD$ — the expected credit loss per borrower under given scenario assumptions
- Fire-sale externality
- Market price depression caused by distressed asset sales at one institution, forcing mark-to-market losses at others
- ICAAP
- Internal Capital Adequacy Assessment Process — a bank's own assessment of capital needs, including climate risk provisions
- LGD (Loss Given Default)
- Percentage of exposure lost when a borrower defaults; climate-adjusted LGD incorporates physical damage to collateral
- NGFS scenarios
- Network for Greening the Financial System reference scenarios: Net Zero 2050, Below 2°C, Delayed Transition, NDCs, Hot House World
- $PD_{\text{climate}}$
- Climate-adjusted Probability of Default: $PD_{\text{base}} \times (1 + \text{stress\_multiplier})$, capturing incremental default risk from climate scenarios
- Physical risk channel
- Transmission of acute (floods, storms) and chronic (sea-level rise, temperature) climate impacts to financial losses
- RWA (Risk-Weighted Assets)
- Total assets weighted by credit risk; denominator in the Capital Adequacy Ratio
- Second-round effects
- Amplification of first-round climate losses through interbank contagion, fire sales, and credit contraction feedback (+20–50%)
- Static balance sheet
- Stress test assumption where portfolio composition is frozen at the test date with no management response
- Stress multiplier
- Factor applied to baseline PD to capture climate-driven incremental default risk; varies by sector, geography, and scenario
- Top-down stress test
- Approach where the regulator applies uniform models and assumptions across all banks for system-wide comparability
- Transmission channels
- Pathways through which climate shocks translate to financial losses: credit risk, market risk, and operational risk
09
References
Key References
- 2022 ECB. 2022 Climate Risk Stress Test. European Central Bank.
- 2022 BoE. Results of the 2021 Climate Biennial Exploratory Scenario (CBES). Bank of England.
- 2024 UNEP FI. Good Practice Guide to Climate Stress Testing. United Nations Environment Programme Finance Initiative.
- 2023 Acharya, V., Berner, R., Engle, R., Jung, H., Stroebel, J., Zheng, X. & Zhao, Y. “Climate Stress Testing.” Annual Review of Financial Economics, 15, 291–326.
- 2024 BOT. Climate Risk Stress Testing Framework. Bank of Thailand.
Supplementary References
- 2017 Battiston, S., Mandel, A., Monasterolo, I., et al. “A Climate Stress-Test of the Financial System.” Nature Climate Change, 7(4), 283–288.
- 2024 NGFS. Climate Scenarios for Central Banks and Supervisors (4th vintage). Network for Greening the Financial System.
- 2022 MAS. Guidelines on Environmental Risk Management. Monetary Authority of Singapore.
10