## ppea2021041

## Source details

**Canonical URL:** [ppea2021041](https://www.imf.org/-/media/files/publications/pp/2021/english/ppea2021041.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/pp/2021/english/ppea2021041.pdf.md)
- [Structured JSON version](/-/media/files/publications/pp/2021/english/ppea2021041.pdf.json)

---

### Scope and purpose, recent evolution, and priorities
- Reviews quantitative tools of financial stability assessments under the Financial Sector Assessment Program (FSAP).
- Concentrates on the FSAP macroprudential stress testing framework:
  - interaction among solvency, liquidity, and contagion risks in the banking sector;
  - assessment of nonbank financial institutions (NBFIs), interactions with banks, and market impact;
  - assessment of nonfinancial sectors and links to the financial sector;
  - macroprudential policy analysis.
- Recent evolution since the 2014 FSAP Review:
  - expanded models for nonfinancial sectors and NBFIs;
  - increased interconnectedness analysis across banks, NBFIs, and nonfinancial sectors;
  - incorporated two-way feedback effects between the financial sector and the real economy.
- Approaches developed and used include:
  - growth-at‑risk (GaR);
  - structural vector autoregression (SVAR) models;
  - dynamic stochastic general equilibrium (DSGE) models;
  - agent-based models (ABM).
- Staff-identified priority areas:
  - interaction between solvency, liquidity, and contagion risks;
  - risks in nonbank financial sectors (cross-sectoral interactions and market impacts);
  - risks in nonfinancial sectors (links to banks’ balance sheets);
  - broader interconnectedness analysis;
  - macro-financial interactions using GaR and DSGE and micro/structural integration of stress-test results;
  - macroprudential policy: use stress-test results with early warning indicators (e.g., debt-at-risk, debt-service-to-income ratios (DSTI)) and microdata-based calibration of borrower-based tools.

### Pandemic-era implications and demand
- COVID-19 increased demand for:
  - household and nonfinancial corporate (NFC) sector vulnerability analysis;
  - bank stress testing focused on nonfinancial sector spillovers.
- Corporate vulnerabilities rose as firms increased debt amid extreme earnings shocks; liquidity risks could turn into insolvencies if recovery is delayed.
- Interest in quantitative calibration of macroprudential policy measures (MPMs) increased; some authorities released macroprudential buffers.

### Standardization, automation, and data constraints
- Push to standardize and automate core risk analyses (e.g., public GaR tool).
- Shift from excel-based tools to program codes for efficiency and accuracy.
- Two types of data constraints:
  - Availability (data gaps): granular sectorized exposures across borders, cyberattacks, climate, fintech.
  - Access: increased sharing of confidential supervisory data with safeguards, but limited access to some datasets (e.g., G-SIBs data collected by the BIS).
- Transaction and settlement (activity-based) data analysis remains rare because of technical challenges; joint analyses with national authorities are used when direct access is limited.

### Emerging risks: climate change, cyber, fintech
- Climate change: three-stage FSAP approach
  1. Climate financial risk diagnostic to set scope and identify physical and transition risks.
  2. Design of climate scenarios.
  3. Design of macro-financial scenarios and integration into standard FSAP stress tests.
- Key Fund features in climate assessments:
  - focus on risks over the three- to five-year FSAP horizon;
  - close scrutiny of physical risks for many Fund members.
- Reverse stress testing approaches will be explored given high uncertainty.
- Cyber risk: growing systemic threat; FSAP examples include Euro Area, Singapore, and simulated collateral access shocks; data constraints remain large.
- Fintech: early-stage quantitative analysis; pilots (2019 Singapore and 2019 Korea) estimated potential non-interest income impacts and unit cost of intermediation effects.

### Regulatory and accounting reform implications
- IFRS 9 adoption changed loan-loss-provision estimation to expected credit losses (ECLs) with 12 months and lifetime horizons; interplay with prudential provisioning must be explicitly modeled.
- Basel III / 2017 reforms (e.g., floor on risk weights for IRB) necessitate updates to solvency stress test tools and reflect jurisdictional implementation timelines.

---

### Bank solvency stress test toolbox upgrades
- Focus on enhancing satellite models linking macro-financial scenarios to banks’ balance sheets and income statements.
- New risks incorporated: climate change and cyber-risk in client distress analysis; bank operational risks and counterparty risks; impact of fintech on banks' income.
- Accounting/provisioning updates:
  - move toward accounting (IFRS9) expected loss metrics (12 months and lifetime ECLs);
  - disentangle accounting and prudential layers in capital estimation.
- Model selection/data:
  - shift to Bayesian Model Averaging (BMA) to address model uncertainty;
  - greater use of granular data.
- Regulatory changes:
  - incorporate Basel III / 2017 reforms, including transitory arrangements for varying country implementation timelines.

Key satellite model features and practices:
- Purpose: translate macro and market factors into granular risk factors to project balances and capital.
- Most important models: loan default rates (credit risk), net interest income, capital requirements; also trading losses, fees and commission income, operational losses.
- Common explanatory variables: GDP growth, inflation, unemployment, income, output gap, policy rate, property prices, equity prices, yield curves, FX, credit spreads, commodities, volatilities.
- Model complexity: determined by data availability and portfolio materiality; can operate at loan, portfolio, bank, or country levels.
- Validation criteria: in-sample and out-of-sample performance (root mean squared error), sign/significance of coefficients; expert judgment against benchmark crisis episodes.

Credit risk modeling specifics:
- Credit costs typically projected as changes in expected losses = PD × LGD with both deteriorating in downside scenarios.
- PD measurement depends on jurisdictional supervisory data (IRB PDs, NPLs, transition matrices, or IFRS9 lifetime PDs).
- Methods include linear regressions with variance adjustments, quantile regressions, BMA, and non-traditional machine learning methods.
- LGDs are more often calibrated (e.g., via provisioning rates, historical coverage).

PPNR and fair-value impacts:
- Require multiple sub-models for assets/liabilities, contractual run-offs, pre-payments, new lending, defaulted assets, and other P&L items.
- Fair-value valuation often approximated using duration; full valuation needs non-linear effects and granular sensitivities when available.

Structural credit risk models:
- Use micro/granular data and behavioral/macro drivers to estimate losses when historical tail events are scarce; permit counterfactual policy experiments (e.g., Merton-type, CCA).

Simplified tools and surveillance/crisis-response tools:
- Global Bank Stress Test (GST): publicly available bank-level financial statements covering 33 jurisdictions; modified in Fall 2020 GFSR for COVID-19 measures.
- Universal Stress Test (UST): expand to more countries using country-aggregate banking sector data.
- Development of macro scenario stress testing tools for NFCs and households using firm-level and household survey data.

Linking solvency, liquidity, and contagion:
- Feedback loop: capital losses → funding problems → asset fire-sales → further capital losses; amplification near regulatory capital minima.
- Recent FSAPs included higher wholesale funding costs and interbank contagion within solvency stress tests (examples: 2018 Euro Area, 2018 France, 2017 Japan, 2018 Brazil).
- Model development: structural joint solvency–liquidity models and haircuts impact models using transaction-level data.
- Next frontier: agent-based models (ABMs) to capture market microstructure, redistribution of losses, and liquidity gaps.

---

### Nonbank Financial Institutions (NBFIs): practices and priorities
- NBFIs include mortgage/leasing companies, asset managers, insurers, pension funds; footprint grew rapidly after the GFC.
- Systemic transmission: NBFI shocks can affect banks and markets via deposits/wholesale funding, credit lines, exposures to NFCs and households, and market activities (repos, securities lending, credit derivatives, insurance).
- FSAP coverage:
  - First NBFI stress test in Norway, 2005 (insurance).
  - By 2019, seven out of eight FSAPs included NBFI stress tests.
- Insurance solvency stress tests:
  - Focus on market risk (stocks, real estate, corporate bonds), interest rate risk (low-for-long), and liability shocks (morbidity, mortality, longevity, catastrophe).
  - Capital effects derived from asset discounts and revaluation of policy liabilities.
  - Cross-sector interactions considered in conglomerates (e.g., 2018 Belgium).
- Asset manager liquidity stress tests:
  - Focus on liquidity mismatch and potential fire sales leading to market liquidity declines and contagion.
  - Methodologies: calibrate redemptions, estimate selling strategies (pro-rata/waterfall), estimate price impact using elasticities and dealer inventories or market turnover.
  - COVID-19 re-emphasized redemption/run risks and connections to bank credit lines.
- Pension fund stress tests:
  - Defined contribution pass risk to members; defined benefit can create contingent sponsor/government liabilities.
  - Methodology similar to insurers: asset price drops, low-yield environment impacts.
- Challenges and next steps:
  - Shift emphasis to NBFIs' contribution to systemic risk through contagion and interconnectedness, cross-sectoral interactions, and bank/market impacts.
  - Scenario design for insurers must consider differing liability duration and government bond yield treatment.
  - Data and standards gaps: no Basel-like international prudential standards for NBFIs; need FSIs methodology for insurance sector; data gaps for hedge funds and new NBFIs.
  - Policy implications: macroprudential measures may need extension to market-based financing and NBFIs; develop liquidity provision frameworks for crises while avoiding premature prudential measures without full systemic understanding.

---

### Nonfinancial sector vulnerabilities: NFCs, households, sovereigns
- Nonfinancial Corporations (NFCs):
  - NFC leverage rose in advanced economies and EMs since the GFC; pandemic sometimes increased leverage further.
  - Quantitative approaches emphasize interest coverage ratio (ICR) analysis (Chow, 2016); distress often proxied by ICR < two.
  - Debt-at-risk measures quantify corporate debt held by firms below ICR thresholds.
  - Three PD approaches: structural, empirical, hybrid (hybrid has best predictive performance); BuDA is a hybrid bottom-up default analysis.
  - NFC analysis complements bank stress tests and can be mapped into bank PDs via historical relationships (Tressel and Ding forthcoming).
- Households:
  - Vulnerabilities driven by debt (mortgages): leverage measures (DTI, debt-to-asset), repayment ability (DSTI, LTV).
  - Microdata (household surveys, credit registries) crucial to calibrate LTV/DSTI and identify pockets of vulnerability.
  - Examples: Italy, Luxembourg, Netherlands, Ireland, Brazil, France, Switzerland used microdata-based approaches.
  - Staff plan to expand microdata use and integrate household models into system-wide stress testing (framework documented in Gross and Población, 2017).
- Sovereign/public sector:
  - Treated mostly as market-risk valuation of sovereign securities; challenges include rarity of sovereign distress in historical data and heterogeneous treatment under accounting and Basel rules.
  - Pandemic-related fiscal expansions raise sovereign risks in coming years, especially in EMDEs; alternative techniques may be needed to incorporate sovereign default scenarios and credit risks from government exposures.

---

### Interconnectedness, systemic liquidity, and systemic liquidity stress testing
- Interconnectedness analysis expanded to exposure-based mapping (interbank, cross-border), price-based measures, and hybrid methods.
- Mapping tools:
  - Balance sheet analysis (BSA), supervisory who-to-whom data, security-level data, FMI settlement/clearing data, BIS consolidated/locational statistics.
- Modeling approaches:
  - Exposure-based: Espinosa-Vega and Sole (2010), CoMap (Covi, Gorpe, Kok).
  - Price-based: Diebold and Yilmaz (2014), CoVaR, SRISK, SyRIN.
  - Hybrid: CCA, CCA combined with VAR/GVAR.
- Best practice: combine exposure- and price-based approaches and interpret interconnectedness alongside stress tests and nonbank/market analysis.
- Policy recommendations:
  - Strengthen monitoring of interlinkages across institutions and real economy;
  - Close data constraints, especially entity-level cross-sectoral exposures;
  - Enhance analytical tools and cross-border cooperation in supervision and resolution.
- Systemic liquidity:
  - Systemic liquidity risk arises from simultaneous liquidity difficulties across institutions and markets; distinct from institution-level liquidity risk due to amplification via interconnectedness.
  - Data needs: granular activity-based FMI data and who-to-whom exposures; behavioral modeling under stress is challenging.
  - Incorporating central bank and government backstops changes assessments; reserve-currency jurisdictions have greater mitigation capacity than small open economies without reserve currency.
  - Development agenda: pilot exercises, combine institution-level liquidity tests with market transaction data, invest in big-data processing and agency collaboration.

---

### Agent-Based Models (ABMs) and dynamic balance-sheet approaches
- Limitations of static/quasi-static stress tests:
  - Static assumptions can mis-rank banks; dynamic balance sheets and macro-feedback effects can reverse vulnerability rankings.
  - Recommendation: incorporate dynamic balance sheets, macro-feedback, and endogenous scenarios.
- ABMs:
  - Bottom-up simulation of heterogeneous agents with heuristic behavior; can model solvency-liquidity interactions, market microstructure, and non-linear systemic dynamics.
  - Experimental prototypes show temporary shocks can become long-lived, eroding solvency and credit growth and that local bank regulatory tightening may be ineffective without system-wide policies.
  - Two prototypes cited: Valderrama (forthcoming) and Eurace 2.0 (Gross and coauthors, forthcoming).
- Macroprudential policy use:
  - FSAPs can use stress tests and indicators to recommend macroprudential tools (e.g., countercyclical capital buffers, DSTI/LTV calibration).
  - Examples of policy-informing FSAP work: 2018 France, 2018 Romania, 2017 Netherlands, 2018 Peru.
  - Future: link stress-test severity to economic buoyancy and use microdata-based tools to calibrate borrower-based measures.

---

### Efficiency improvements, standardization, and tool development
- Need to improve efficiency to expand FSAP coverage within resource constraints.
- Standardization efforts:
  - Provide guidance notes, code files on a refreshed webpage; develop a tool to efficiently estimate satellite models; migrate to program code-based tools.
- Ongoing projects:
  - MCM updating guidance notes and excel tools; building internal operational reference notes and stress-testing codes.
  - Growth-at-Risk (GaR): user-friendly package with Excel interface and underlying Python codes; published on GitHub.
  - Corporate sector stress test tools: Tressel and Ding forthcoming; BuDA access for macro scenario corporate stress tests.
- Constraints: tools must be flexible to country-specific supervisory formats, accounting standards, business models, and implementation of Basel rules.

---

### Climate risk scenario analysis: operational approach and mapping into financial stability
- Appendix I three-stage approach:
  - Stage 1: diagnostic and climate financial risk heatmap (C-RAM informed by G-RAM).
  - Stage 2: design climate scenarios (physical and transition risks); NGFS scenarios leveraged; include near-term and long-term horizons.
  - Stage 3: map scenarios into banking resiliency via macro approach (map stage 2 into macro-financial scenarios and apply standard stress testing) or micro approach (borrower-level assessments aggregated to bank-level).
- Trade-offs:
  - Macro approach: feasible short-term, uses standard macro-financial models and CGE (e.g., GTAP) for sectorization; can use NGFS and other carbon-price paths.
  - Micro approach: more accurate but requires granular exposures, firm/household balance sheets, and sectoral modeling; feasible in few FSAPs with requisite data.
- Examples of carbon-price assumptions (level of assumed carbon prices in scenario (2020 US$/ton CO2)) as presented:
  - NGFS: Orderly 0103132350; Disroderly 21017841; Hothouse 77711
  - Bank of France: Orderly transition 05075180; Delayed transition 000700; Sudden transition 00170900
  - Bank of Canada: 2°C (consistent) scenario 080190600; 2°C (delayed action) scenario (abrupt transition) 000800; Nationally determined contributions 02070190
  - ECB: Orderly 760114360; Disroderly 75468845; Hothouse 7141416
  - ESRB: Sudden transition USD 100 per tonne carbon price shock 100
  - Dutch National Bank: Sudden transition USD 100 per tonne carbon price shock 100
  - WEO: Adverse (average) 1/71927101
  - Note: "1/ Assumes the baseline is the NGFS hothouse scenario."
- Operational considerations:
  - Need close inter-departmental and external collaboration (World Bank, NGFS, vendors).
  - Data needs: novel, very granular data; leverage catastrophe models for physical risk and NGFS scenarios for transition risk.
  - Applicability extends beyond banks (insurers, mutual funds) with sector-specific scenario methodologies.
- Practice examples:
  - Bank of England: bottom-up, 30-year horizon (60 years for physical risk), static balance sheet, reporting every 5 years.
  - Bank of Canada: top-down, 80-year horizon, 18 regions and 33 sectors.
  - Other authorities vary by horizon, granularity, reporting frequency, and static vs dynamic balance-sheet assumptions.

---

### Data initiatives, access, and recommendations
- Improvements since the GFC: G-20 Data Gap Initiatives; FSB annual global NBFI reports; activity-based FMI data and big-data techniques emerging.
- FSAP access to confidential supervisory data:
  - Mandatory assessments: sharing increased from 75 percent in first assessments to 97 percent in latest.
  - Voluntary assessments: all jurisdictions shared at least some confidential supervisory data in past five years, breadth varies.
  - Some access only in secure physical locations, increasing costs.
- Remaining gaps and recommendations:
  - Gaps: entity-level cross-sectoral linkages, activity-based FMI data, G-SIB exposures, hedge funds, new regulatory-arbitrage NBFIs.
  - Suggested approaches: expand country data development via Statistics and IT Departments; encourage standard practice for supervisory data sharing; enable electronic remote access; develop code-based analysis for counterparties to run when FSAP access is restricted.

---

### Use of FSAP tools in Article IV surveillance
- Article IV teams have used vulnerability indicators, balance sheet analysis, and GaR.
- Planned expansion:
  - Broader menu of FSAP quantitative tools selectable by desk economists.
  - Simple stress testing tools: standardized GST and UST; system-wide FX liquidity stress testing tool using BSA data to link balance-of-payments shocks to bank liquidity.
  - NFC and household tools: Tressel and Ding (forthcoming) corporate tools; MCM household tools to calibrate borrower-based MPMs.
  - Macro-feedback models to inform calibration of macroprudential tools and buffer sizing.
- Stress-test results can inform reserve adequacy, borrower-based MPMs, cyclical capital measures, sovereign-bank nexus in public debt sustainability, and climate policy recommendations.

*International Monetary Fund.*

### INTRODUCTION _____________________________________________________________________ 4

### INTRODUCTION

### Scope and purpose
- Reviews quantitative tools of financial stability assessments under the Financial Sector Assessment Program (FSAP).
- Concentrates on the main elements of the FSAP’s macroprudential stress testing framework:
  - interaction among solvency, liquidity, and contagion risks in the banking sector;
  - assessment of the health of nonbank financial institutions (NBFIs), their interactions with banks and their impact on financial markets;
  - assessment of the health of nonfinancial sectors and their links to the financial sector;
  - macroprudential policy analysis.
- Also reviews recent improvements in microprudential bank solvency stress testing and new tools for emerging risks (climate change, fintech, cyber).
- Discusses data constraints, adoption of quantitative tools by Article IV teams, and potential efficiency improvements.

### Recent evolution of FSAP quantitative work
- Since the 2014 FSAP Review, quantitative tools:
  - expanded to include models for vulnerabilities in nonfinancial sectors and NBFIs;
  - increased analysis of interconnectedness between banks, NBFIs, and nonfinancial sectors;
  - incorporated two-way feedback effects between the financial sector and the real economy.
- Approaches developed and used include:
  - growth-at‑risk (GaR);
  - structural vector autoregression (SVAR) models;
  - dynamic stochastic general equilibrium (DSGE) models;
  - agent-based models (ABM).
- Some FSAPs have begun quantitative calibration of macroprudential tools.

### Priority areas going forward
- Staff-identified focal areas (driven by staff assessment and FSAP Review survey results):
  - Interaction between solvency, liquidity, and contagion risks: develop models incorporating complex interactions to better capture systemic risk.
  - Risks in nonbank financial sectors: focus on cross-sectoral interactions and market impacts, recognizing data limitations.
  - Risks in nonfinancial sectors: develop models analyzing links between nonfinancial sector health and banks’ balance sheets.
  - Interconnectedness analysis: expand scope to include cross-financial segments and cross-sectoral linkages.
  - Macro-financial interactions: use GaR and DSGE primarily to build macro scenarios; develop micro and structural approaches to integrate stress test results back into macro-financial developments.
  - Macroprudential policy: use macroprudential stress test results alongside early warning indicators (e.g., debt-at-risk, debt-service-to-income ratios (DSTI)) to inform buffer sizing and ex-ante policy assessment; consider microdata-based calibration of borrower-based tools.

### Pandemic-era implications and demand
- The COVID-19 shock increased demand for:
  - household and nonfinancial corporate (NFC) sector vulnerability analysis;
  - bank stress testing focused on nonfinancial sector spillovers.
- Corporate vulnerabilities rose as firms increased debt to cope with cash shortages amid extreme earnings shocks.
- Underlying liquidity risks could turn into insolvencies if recovery is delayed, potentially spilling over to the financial sector.
- Interest in quantitative calibration of macroprudential policy measures (MPMs) increased with some authorities releasing macroprudential buffers.

### Standardization, automation, and efficiency
- Enhancements will be complemented by increased standardization and automation for core risk analysis where feasible.
- Example: development and public dissemination of the GaR tool (IMF, 2017b) with ITD collaboration.
- Staff are working to standardize core risk analysis (especially bank stress tests and satellite models) across data environments, with guidance notes and code files to be provided on a refreshed IMF webpage.
- Shifting away from excel-based tools to program codes could increase efficiency and accuracy.

### Data constraints
- Two types of data constraints affect quantitative risk analysis:
  - Availability (data gaps): granular sectorized exposures across borders, cyberattacks, and data in emerging areas (climate, fintech) remain inadequately collected.
  - Access: while most national authorities now share confidential supervisory data with FSAP teams under safeguards, access to some datasets (e.g., G-SIBs data collected by the BIS) is limited.
- Transaction and settlement (activity-based) data analysis remains rare due to technical challenges handling large confidential datasets.
- FSAP teams sometimes conduct joint analyses with national authorities who have data access, using code and information-sharing platforms that do not require direct FSAP access to raw data.

### Emerging risks: climate change emphasis
- Operational approach prioritized for climate-related financial stability risks, reflecting:
  - very high uncertainty over timing and likelihood;
  - complex micro-level dependencies;
  - data availability limitations.
- Staff envisage a three-stage approach for climate risk in FSAPs:
  1. Climate financial risk diagnostic to set scope and identify relevant physical and transition risks.
  2. Design of climate scenarios.
  3. Design of macro-financial scenarios and integration into standard FSAP stress tests.
- Key Fund features in climate assessments:
  - focus on risks over the three- to five-year FSAP horizon;
  - close scrutiny of physical risks, which may be relatively more relevant for many Fund members.
- Given high uncertainty, reverse stress testing approaches will be explored as a complementary perspective.

### Regulatory and accounting reform implications
- Ongoing regulatory and accounting reforms require adjustments to bank stress test tools:
  - 2017: BCBS introduced additional Basel III requirements limiting application of IRB approaches for risk-weighted assets.
  - Adoption of international financial reporting standards 9 (IFRS 9) in many jurisdictions changed loan-loss-provision estimation to expected credit losses (ECLs), altering models and data structures for estimating bank credit risks.

### Scoping and methodology for FSAP quantitative work
- The scoping process determines approach and methodologies for each FSAP:
  - Identifies material risks and vulnerabilities based on the preliminary Risk Assessment Matrix (see background paper on scope SM/21/54).
  - Choice of quantitative approaches depends on identified risks, data availability and structure, and jurisdiction-specific features (supervisory framework, accounting rules).

### Macroprudential vs microprudential stress testing
- Definitions and distinctions:
  - Microprudential stress test: forward-looking supervisory tool assessing an individual bank’s balance sheet; outcome leads to bank-specific supervisory measures.
  - Macroprudential stress test: similar framework but focuses on systemic risk by incorporating amplification and contagion channels affecting the whole financial system (e.g., interactions between solvency, liquidity, and contagion risks); expanded to study NBFI risks, nonfinancial sector vulnerabilities, and interconnectedness. Work is underway to incorporate two-way feedback effects between the real economy and bank health through macro-financial channels and dynamic bank balance-sheet modeling.
- Objective of macroprudential stress tests: recommend macroprudential measures to mitigate systemic risks.

_International Monetary Fund._

### 11.      Staff continue to upgrade the bank solvency stress test toolbox currently used by

### 11.      Staff continue to upgrade the bank solvency stress test toolbox currently used by

### Upgrades overview
- Focus: enhance satellite models that translate macro-financial scenarios into banks’ balance sheets and income statements and incorporate new methods to address changes in regulatory and accounting rules and new sources of risks.
- New risks incorporated: climate change, and cyber-risk as part of bank clients’ distress analysis; bank operational risks and counterparty risks; impact of fintech on banks' income.
- Accounting and provisioning: move from approximation based on expected losses approach (often proxied by changes in probability of default × loss-given default) to use of accounting (IFRS9) expected loss metrics, based on accounting definitions of 12 months and lifetime expected credit losses; disentangling accounting and prudential layer and their interplay.
- Model selection and data: shift from econometric models with ad-hoc specifications to Bayesian Model Averaging (BMA) models; based on more granular data.
- Regulatory changes: incorporate revisions to risk weights in the standardized approach, removing the use of internal risk models for certain asset classes, and a minimum leverage ratio; reflect Basel III / 2017 reforms where adopted, including transitory arrangements for varying country implementation timelines.

### Satellite models: role and design
- Purpose: link macroeconomic and market factors to forecast key bank parameters and translate macro-financial scenarios into granular risk factors to project bank balance sheets and capital.
- Most important satellite models: those forecasting loan default rates (credit risk), net interest income, and capital requirements; additional models for trading losses, fees and commission income, and operational losses.
- Explanatory variables commonly used: GDP growth, inflation, unemployment, income, output gap, policy rate, property prices, equity prices, yield curves; financial market indicators (FX, equities, credit spreads, commodities, rates, FX volatilities, equities volatilities, rates volatilities).
- Model complexity: chosen according to data availability and portfolio materiality; can be built at loan, portfolio, bank, or country levels.

### Model comparison and validation
- Criteria used:
  - in-sample forecast performance measures;
  - out-of-sample forecast performance based on a truncated sample, measured by the root mean squared error (or similar measures) over the forecasting period;
  - the sign and significance of coefficient estimates.
- Expert judgment may be applied to assess how banks’ risk metrics behaved historically relative to macro environment against benchmark crisis episodes (e.g., the GFC or the European sovereign debt crisis).

### Credit risk (loan default) modeling
- Credit costs: projected as changes in expected losses on loans; usually calculated as changes of the probability of default (PDs) multiplied by the loss-given-default (LGD), as both PDs and LGDs deteriorate in downside scenarios.
- Probability of default:
  - Assessed differently across jurisdictions depending on supervisory data structure (e.g., PDs when IRB-regulated; NPL data or transition matrices when Basel standardized approach; lifetime PD when IFRS9 data available).
  - Independent variables include local, regional, and global risk factors grouped by exposures to material geographies.
  - Methods: traditional linear regression with variance adjustments, quantile regressions, Bayesian Model Averaging (BMA) to address model uncertainty; non-traditional approaches include machine learning, random forests, and neural networks.
- Loss given default:
  - More likely calibrated than estimated.
  - If PDs proxied by NPLs, required provisioning rates by regulation and historical provision coverage rates could proxy LGDs.
  - LGDs can be calibrated based on experience summarized academic studies on credit risk modeling.

### Pre-provision net revenue (PPNR) and fair-value impacts
- Objective: project components of PPNR, including trading losses, fees and commissions income, and others.
- Requirement: many separate sub-models for stock of assets and liabilities, contractual run-offs, pre-payments, new lending, defaulted assets, and broader P&L items beyond net interest income.
- Idiosyncratic risk: PPNR models need to incorporate and project bank-specific factors such as pricing behavior, business strategy, and solvency–funding interactions.
- Fair-value instruments:
  - Valuation changes often approximated using duration; full valuation should incorporate non-linear effects for large shocks.
  - Additional granular data needed: delta sensitivities by major index/counterparty, breakdown of long vs. short positions and cash vs. derivative positions, and correlations between cash and derivative curves to stress basis risk.
  - When granular data are not available, a modified duration approach can provide reasonable proxy estimates.

### Structural credit risk models
- Use cases: effective when historical data is short or absent or when portfolio structure recently shifted.
- Approach: combine borrower risk measures (e.g., leverage, default rates) with behavioral and macroeconomic risk drivers (e.g., income, profitability, interest rates) and rely on micro/granular data (e.g., credit registry, household, and corporate surveys).
- Advantages: better estimates of losses in absence of long-run historical tail-loss events; permit counterfactual policy experiments.
- Other structural approaches: Merton-type credit risk models using firms’ balance sheet structure and equity prices.

### Changes in regulatory and accounting rules
- IFRS 9:
  - Introduces accounting expected loss metrics (12 months and lifetime expected credit losses).
  - In some jurisdictions IFRS 9 could be tighter than prudential provision requirements; in others prudential requirements are more conservative.
  - IFRS 9 provisions are more responsive to cyclical factors and may be larger than prudential impairments during a recession.
  - Both provisioning layers and their interplay need explicit modeling to estimate capital positions under adverse scenarios and to model banks’ behavioral responses and macro-financial amplification mechanisms.
- Basel III / 2017 reforms:
  - Finalized Basel III framework to be integrated into solvency stress test tools in adopting jurisdictions.
  - BCBS reforms on risk-weighted assets introduced a floor in 2017 on risk weights for banks using the internal-rating-based approach.
  - Stress test tool updates reflect these reforms and include transitory arrangements for countries with different implementation timelines.

### Simplified tools for surveillance and crisis response
- Global Bank Stress Test (GST):
  - Macro scenario stress testing tool using publicly available bank-level financial statements covering 33 jurisdictions.
  - Modified in Fall 2020 GFSR to incorporate COVID-19 related mitigation measures (such as loan guarantees).
  - MCM prepared a methodology note and will share relevant codes so desk economists can update data and scenarios (and models if needed).
- Universal Stress Test (UST): expansion to include more countries using country-aggregate banking sector data.
- Development underway: macro scenario stress testing tools for NFCs and households using firm-level and household survey data.
- Demand: interest in household and corporate sector analysis and bank stress tests increased in the context of the COVID-19 crisis.

### Linking solvency, liquidity, and contagion risks
- Importance: interaction of solvency and liquidity risks can drive severity of financial crises; runs on liabilities can force asset liquidations, causing losses and potential insolvency.
- Feedback loops: capital losses → funding problems → asset fire-sales → further capital losses; amplification becomes disproportionately larger as capital approaches required minimum.
- Recent FSAP practices:
  - Inclusion of higher wholesale funding costs due to solvency deterioration in stress scenarios (e.g., 2018 Euro Area, 2018 France, 2017 Japan).
  - Incorporation of solvency and contagion interaction via interbank networks; 2018 Brazil FSAP integrated contagion analysis within solvency stress tests to capture additional capital impact from exposures to defaulting banks.
- Model development:
  - Structural models for joint solvency–liquidity stress testing (Cont, Kotlicki, and Valderrama, 2020; Gross, Leika, and Valderrama, forthcoming; Krznar and Matheson, 2017).
  - Models that gauge impact of haircuts on liquid assets using transaction-level data (Han and Leika, 2019).
- Next frontier: agent-based models to incorporate demand and supply conditions, market microstructure, redistribution of losses/gains, and liquidity gaps and surpluses among institutions (e.g., Valderrama, forthcoming).

### Risks in nonbank financial institutions (NBFIs)
- Definition: NBFIs are institutions engaged in shadow banking activity or financial intermediation outside the traditional banking system.
- Examples: mortgage/leasing companies, asset managers, insurers, pension funds.
- Trend: rapid growth of NBFIs after the GFC; growing footprint over the past quarter-century.
- Systemic considerations:
  - Solvency distress of some NBFIs (e.g., investment funds and insurers) is in principle contained by investor/policyholder contractual loss absorption, unlike bank depositors.
  - However, shocks to NBFIs can generate systemic impacts via interconnectedness with banks and markets: lending to banks (deposits or wholesale funding), borrowing from banks (credit lines), exposures to NFCs (bonds), households (mortgages), and market activities (repos, securities lending, credit derivatives, insurance) affecting asset prices.

*Source: 2021 FSAP REVIEW—BACKGROUND PAPER ON QUANTITATIVE ANALYSIS*

### 27.      Risks to NBFIs have received increased attention in recent FSAPs. The first FSAP to

### Risks to NBFIs have received increased attention in recent FSAPs

### Overview
- The first FSAP to include stress tests of NBFIs was Norway in 2005, which analyzed the insurance sector.
- Since then, more FSAP exercises have incorporated stress tests of NBFIs, culminating in seven out of eight FSAPs in 2019.
- FSAPs have conducted stress tests for insurers and investment funds and examined their impact on asset prices.
- Key NBFI characteristics relevant for systemic analysis:
  - (Life) insurers do not typically “fail” suddenly because liabilities are long-term and policyholders cannot cancel contracts prematurely without large haircuts.
  - Many large life insurers have “mutual” structures where policyholders are equity holders expected to absorb losses.
  - Investment funds are highly substitutable; the default of an asset manager is unlikely to affect industry-wide capacity to provide services.

### Insurance Solvency Stress Tests
- Solvency stress tests of insurers are the most common NBFI stress test in FSAPs, applied more often to life insurers because they build up large asset holdings.
- Major risk exposures and test approaches:
  - Insurers—especially life insurers—are most exposed to market risk from securities; FSAPs analyze scenarios including falling prices of stocks, real estate, and corporate bonds.
  - Capital effects are derived by applying discounts to asset values and revaluing policy liabilities at new interest rates; risk-based capital requirements are modeled by adjusting down asset values in line with the scenario.
  - Interest rate risk analysis focuses on “low-for-long” scenarios: long-term policies priced at historically higher interest rates face erosion of net interest income as higher-yielding bonds mature and are reinvested at lower yields.
  - Bottom-up sensitivity tests examine shocks to liabilities (e.g., morbidity, mortality, longevity risks for life insurers; natural disaster and cyber losses for non-life insurers).
  - Catastrophe risk insurance and re-insurance diversify property-insurance tail risks; solvency impacts may be limited as contracts and premiums are usually revised every year.
- Cross-sector interactions:
  - Some FSAPs explicitly consider interactions between insurers and banks, especially in financial conglomerates or where banking groups own insurance subsidiaries (example: 2018 Belgium FSAP stress-tested bank-insurance conglomerate models).

### Asset manager Liquidity Stress Test
- Asset-manager stress tests focus on knock-on effects on securities markets through liquidity stress rather than solvency, since many funds are equity-funded or pass on investment risk to clients (IMF, 2015).
- Key vulnerabilities and testing elements:
  - Liquidity mismatch: open-ended mutual and investment funds are often redeemable on demand; runs can trigger asset sell-offs at large discounts if assets are illiquid.
  - Fire-sale and contagion channels: asset-manager fire sales can dry up market liquidity, affect market funding for banks, other NBFIs, and NFCs, and cause excessive asset price declines.
- Methodologies used in FSAPs:
  - Redemption pressures: calibrate severe yet plausible redemption scenarios using historical redemption behavior at fund or fund-class level.
  - Fire sale pressures: estimate amounts asset managers would sell by considering pro-rata and waterfall selling strategies; estimate elasticities for different asset types and vary selling order.
  - Fire sale impact: measure market price impact of hypothetical sales by comparing to dealer inventories (2015 United States FSAP), to investment funds’ liquid assets (Luxembourg), or to market turnover (2016 Sweden FSAP); 2018 Brazil FSAP estimated price effects using elasticities from market liquidity measures.
  - Contagion effects: Luxembourg FSAP measured impacts on banks from liquidity stress to investment funds; Brazil FSAP introduced second-round effects where asset price falls lead to another round of redemptions and further price falls.
- COVID-19 lessons:
  - Severe market turbulence after the onset of COVID-19 re-emphasized asset-manager liquidity mismatch risks despite reforms.
  - Fund investments in higher-risk NFC bonds and leveraged positions contributed to market freeze and unprecedented central bank liquidity support in money and corporate bond markets in some jurisdictions.
  - Post-turbulence, linkages strengthened as many funds activated credit lines from banks.
  - The pandemic underscores evolving risks from the asset management industry and the need to adapt risk analysis.

### Pension Fund Solvency Stress Tests
- Pension funds are generally stable long-term institutional investors and too small/disconnected in most countries to be systemic, but in some countries they can contribute to systemic risk (examples of FSAP coverage: Mexico, Namibia, Netherlands).
- Linkages:
  - Life insurers often administer pension plans and sell annuities; health of insurers and pension funds is linked.
  - Defined contribution funds pass on market risk to members (similar systemic risks as asset managers).
  - Defined benefit funds can create contingent liabilities for sponsors; simultaneous asset price falls across many large defined benefit plans can pose systemic risk via government and firm contingent liabilities.
  - Unfunded (pay-as-you-go) pension plans lack balance sheet information and cannot be stress-tested in an FSAP.
- Methodology:
  - Pension fund stress tests follow approaches similar to insurers: analyze sudden asset price drops or the effect of a low-yield environment on net interest income over several years.

### Challenges and Work Going Forward
- Focus shift: Future FSAPs will emphasize assessing NBFIs’ contribution to systemic risk through contagion and interconnectedness—especially cross-sectoral interactions and impacts on banks and markets—rather than on individual NBFI failures.
- Scenario design challenges for insurers:
  - Key question: how closely to align insurer adverse scenarios with bank solvency stress tests—treatment of government bond yields is critical (yields may rise with capital flight or fall with central bank easing).
  - Insurers’ liabilities are usually longer-term than assets, so valuation effects of government bond shocks can go in opposite directions to banks’; additional insurer-focused scenarios or batteries of single-factor sensitivity shocks are often needed to ensure sufficient prudence.
- Data and standards gaps:
  - There are no Basel-like international standards on prudential requirements for NBFIs; definitions of capital and FSIs for insurers are not globally accepted.
  - The FSAP will benefit from the Statistics Department’s project on a methodology for FSIs for the insurance sector.
  - Data gaps remain for some NBFIs (e.g., hedge funds, new regulatory-arbitrage NBFIs such as wealth management products in China), limiting risk-analysis coverage.
- Policy implications:
  - Quantitative NBFI analysis could inform macroprudential policy and crisis management as market financing and NBFI footprints rise.
  - Macroprudential tools that only target banks and their borrowers may lose effectiveness when NFCs borrow more from markets; there may be a need to develop macroprudential tools for NBFIs and consider a liquidity provision framework in crises (as highlighted by COVID-19 market turbulence).
  - Caution is advised: avoid rushing to prudential measures without fully understanding systemic importance of vulnerabilities.

### Risks in Nonfinancial Sectors
- Growing incorporation of vulnerabilities in NFCs, households, and governments into FSAP risk analysis due to potential financial spillovers.
- Historical examples of nonfinancial sector vulnerabilities causing financial crises:
  - Asian crisis in the late 1990s linked to NPLs from sharply increased corporate lending.
  - High NPLs and bank weaknesses in Italy related to NFC debt burdens.
  - U.S. subprime crisis and banking crises in Ireland and Spain linked to household vulnerabilities and unsound mortgage lending.
- Implications:
  - Loans to NFCs and households are often largest bank portfolio items, making analysis of related vulnerabilities crucial to bank solvency stress tests and financial stability analysis.
  - Sovereign-bank linkages became central during the European sovereign debt crisis.
  - Distress in the nonfinancial sector that reduces valuations of securities issued by these entities can raise banks’ liquidity risks as liquid asset buffers deteriorate.

### Nonfinancial Corporations
- NFC leverage has risen in both advanced economies and emerging markets since the GFC; borrowing has been increasingly driven by global factors rather than firm-level characteristics.
- During the pandemic, NFC leverage sometimes rose further as firms facing earnings and liquidity shocks increased borrowing, supported partly by public support measures and central bank liquidity injections.
- Quantitative approaches in FSAPs:
  - Focus on the interest coverage ratio (ICR) as summarized in Chow (2016).
  - Distress is defined as risk of failing to repay any borrowing type (bank loans, payables, bonds, international borrowings).
  - Stress tests estimate impacts of shocks to interest rate, exchange rate, and profits (EBIT) on ICR (earnings divided by interest payments).
  - Shocks can be applied singly (sensitivity) or combined, or reflect an adverse macro scenario (IMF, 2016b, Figure 5).
  - An ICR of below two is often considered a sign of distress.
  - Debt-at-risk measures amounts of corporate debt issued by firms with an ICR below pre-specified thresholds, indicating potential extent of corporate debt distress.

*Source: 2021 FSAP REVIEW—BACKGROUND PAPER ON QUANTITATIVE ANALYSIS (excerpt).*

### 46.      Another approach is to work with the PD. A firm’s ICR is  related to credit risk, but it is,

### ppea2021041 - 46.      Another approach is to work with the PD. A firm’s ICR is  related to credit risk, but it is,

### Approaches to assessing corporate default risks
- Three approaches to assess corporate default risks: structural, empirical, and hybrid. Each has strengths and limitations. Current academic consensus: the hybrid approach has the best predictive performance (Campbell and others, 2008).
- Structural approach:
  - Builds on Merton-type asset pricing models for corporate debt based on option pricing models, including the contingent claims approach (CCA) developed by IMF staff (Gray and Malone, 2008) and Moody’s KMV.
  - Heavily relies on market-based indicators and is well suited for higher frequency monitoring.
  - Not applicable to firms without traded equity or bonds.
  - Forecasting performance tends to be weaker than other approaches.
- Empirical approach:
  - Reduced form empirical model that regresses indicators of actual defaults on various firm characteristics.
  - Historically focused on explaining cross-firm differences using firm-specific characteristics and indicators.
  - Some models incorporate macro-financial variables (Bruneau and others, 2012).
  - Can cover a broader sample of firms than structural approach but tends to be of lower frequency.
- Hybrid approach:
  - Empirical models that include outputs from structural models, in particular distance-to-default (Campbell and others 2008).
  - Bottom-up Default Analysis (BuDA) developed jointly by IMF staff and the National University of Singapore adopts this approach (Credit Research Initiative, CRI, 2019a and 2019b).
  - Tends to show the best out-of-sample forecasting performance.
  - Some FSAPs (e.g., 2017 Indonesia) have used BuDA.

### NFC (Nonfinancial Corporation) risk analysis and relationship to bank stress tests
- NFC risk analysis is a useful input and complement to bank stress testing.
- Differences between NFC analysis and bank stress tests:
  - Coverage: NFC analysis can include listed and unlisted companies irrespective of whether these firms have bank loans; bank stress tests reflect credit risk only from firms that have bank loans.
  - Concept of credit stress event/default: Bank stress tests consider loans classified as nonperforming; NFC analysis can conceptualize credit stress as bankruptcy, default on any loans or bonds, or key metrics (such as the interest coverage ratio) falling below specific thresholds.
- Uses of NFC analysis in FSAPs:
  - Can be treated as an independent exercise, as robustness checks of bank credit risk models, or as substitutes for credit risk models.

### Integrating corporate sector analysis into bank stress tests (COVID-19 implications)
- COVID-19 highlighted the need to further integrate corporate sector analysis into bank stress tests due to unusually high variance of pandemic impact across sectors and uncertainty over solvency and liquidity positions once pandemic-related policies are withdrawn.
- Multi-year macro scenario-based stress tests for NFCs integrated into bank credit risk modeling can provide more granular understanding of potential financial stability impact of corporate stress.
- Tressel and Ding (forthcoming) framework:
  - Complements ICR-based analysis with additional indicators such as cash and equity buffers.
  - Based on same macro scenarios used for bank stress tests (e.g., GST).
  - Maps stress indicators into aggregate bank PDs using historical relationship between corporate defaults and these indicators, providing input for bank credit risk modeling.
- Corporate stress tests are useful when supervisory data are incomplete or of low quality.

### Household balance sheet vulnerabilities and microdata use
- Main source of household vulnerability: debt, especially mortgages.
- Measures of household indebtedness:
  - Leverage: debt-to-income ratio (DTI) and debt-to-asset ratio.
  - Ability and willingness to repay: DSTI and LTV.
  - LTV affects loss-given-default (LGD) incurred by lenders.
- Risk drivers and mitigants:
  - Risks tend to rise with lower bank lending standards.
  - Characteristics of borrowers (income brackets) and loan purpose (primary residence vs. investments) are important.
  - Financial asset buffers mitigate risks; monitor saving ratios and financial asset allocations.
  - Residential real estate market prices and potential overvaluation are essential to determine household balance sheet vulnerabilities.
- Role of microdata:
  - Household surveys essential for assessing risks and calibrating borrower-based macroprudential tools.
  - Aggregate indicators can mask pockets of vulnerabilities concentrated in small segments.
  - FSAPs have used household survey data to calibrate LTV and DSTI ratios for mortgages.
- Examples of microdata-based stress tests and models:
  - Single factor stress tests of household balance sheets using microdata: Italy, Luxembourg, the Netherlands, Ireland, Brazil.
  - 2019 France FSAP and 2017 Luxembourg FSAP analyzed household vulnerabilities and estimated models of residential real estate prices.
  - 2019 Switzerland FSAP relied on a structural model calibrated on microdata to estimate PDs and LGDs for mortgages.
- Future plans:
  - Staff plan to increase use of microdata and better integrate household vulnerability analysis into system-wide stress testing.
  - Household model framework objectives: (i) assess effects of downturn scenarios on household risk parameters; (ii) gauge effects of policies including LTVs, DSTIs, DTIs; (iii) enhance assessment of household credit dynamics (mortgages, consumer credit).
  - Framework will consider second-round macro feedback effects and be integrated into bank stress tests where household risk parameters are modeled as a function of scenarios.
  - The framework is documented in Gross, M. and Población, J. (2017).

### Public sector / sovereign risk in FSAP stress tests
- Most FSAP stress tests have treated sovereign risk as market risk from valuation changes in sovereign securities (Jobst and Oura, 2019).
- Key challenges:
  - Size of sovereign shocks:
    - Sovereign distress is relatively rare in post-World War II period for advanced economies and many emerging and developing economies.
    - Historical data may not include sufficient distress events and may generate too small shocks compared to what could potentially happen.
    - Referencing cross-country experiences and using risk-sensitive market data (such as sovereign CDS spreads) when available could be useful.
  - Treatment of sovereign exposures:
    - Basel capital rules smooth out volatile short-term effects to avoid excessive pro-cyclicality; same sovereign exposures can be valued or provisioned differently depending on labeling (e.g., held-to-maturity account).
    - Such smoothing could reduce effectiveness of macroprudential stress tests; FSAP stress tests have often deviated from strict regulatory standards and applied stressed market valuations of sovereign portfolios when necessary.
  - Amplification and feedback mechanisms:
    - Sovereign distress can trigger wide-ranging spillovers across sectors, making overall impact highly uncertain.
    - Designing adequate macroeconomic scenarios is challenging; focus may be needed on a few country-relevant channels.
- Pandemic implications:
  - Pandemic-induced deficit-financed fiscal expansions increase debt burdens and could raise sovereign risks in coming years, notably in emerging and developing economies (EMDEs).
  - Sovereign distress in EMDEs is more likely to be outright default (explicit default on external debt, monetization of domestic debt, elevated bank loans to governments with evergreening, accumulation of arrears).
  - Unconventional monetary policy in many EMDEs poses new challenges to assessing sovereign risks.
  - FSAPs may need alternative techniques such as incorporating sovereign default and its macroeconomic impact in scenarios and accounting explicitly for credit risks from government exposures.

### Interconnectedness analysis: mapping, modeling, and policy recommendations
- Expansion since 2014 FSAP Review: greater focus on exposure-based interconnectedness in domestic interbank market and cross-border bank lending, and price-based interconnectedness; improved data for sectoral financial accounts enables broader cross-sectoral and cross-border linkages analysis.
- Interconnectedness analysis typically includes:
  - Mapping of financial system.
  - Analysis or modeling of interbank, cross-sectoral, and cross-border linkages.
  - Policy discussions.
- Mapping financial interlinkages improves understanding of shock amplification and spillovers:
  - Interbank: supervisory data mapping loans, bonds, capital participation, off-balance sheet exposures; comparison with intra-group exposures (e.g., Spain).
  - Cross-sectoral: balance sheet analysis approach (BSA) used to map cross-sectoral exposures from aggregated sectoral balance sheets (e.g., Romania); supervisory data to map ownership structure (e.g., Poland); security-level data to map cross-segment linkages (e.g., France).
  - Financial Market Infrastructures (FMIs): authorities sharing FMI data has allowed mapping settlement and clearing linkages (e.g., China FSAP used network analysis to map linkage between FMIs and banks).
  - Cross-border banking: use of BIS consolidated and locational banking statistics to map cross-border linkages and analyze types, destinations, origins of exposures (e.g., Spain).
- Modeling approaches used:
  - Exposure-based approaches:
    - Espinosa-Vega and Sole (2010) model for credit and funding shock propagation (e.g., Luxembourg).
    - Contagion Mapping (CoMap) by Covi, Gorpe, and Kok (2019) applied to euro area banking network in Euro Area FSAP; added to bank stress tests to capture second-round interbank contagion (Indonesia and Poland).
  - Price-based approaches:
    - Diebold and Yilmaz (2014) used to analyze interconnectedness from equity and other market prices (e.g., Finland and Spain).
    - CoVaR (Adrian and Brunnermeier, 2016) used in New Zealand FSAP.
    - SRISK (Acharya and coauthors, 2012) and SyRIN (Cortes and others, 2018) applied in United Kingdom FSAP for banking and insurance sectors.
  - Hybrid approaches:
    - Contingent claims analysis (CCA) and systemic CCA (Gray and Jobst, 2013) combine market and balance sheet data.
    - CCA combined with VAR and GVAR (Dees and others, 2017) measure cross-sector and cross-border interlinkages (Euro Area and United States FSAPs).
    - Combined CCA-GVAR applied in Euro Area FSAP involving banks, insurance, sovereigns, and economies (Gross, Kok, and Zochowski, 2016).
- Best practices and limitations:
  - Combining exposure- and price-based approaches provides more comprehensive analysis than either alone.
  - Exposure data do not reflect indirect linkages and market-perception amplification channels.
  - Price-data-based models are often not structural and cannot pinpoint contagion channels.
  - Interconnectedness and contagion results should be viewed alongside stress tests and nonbank/market analysis for a holistic assessment.
- Policy recommendations arising from interconnectedness analysis grouped into four main areas:
  - Strengthen monitoring of linkages among financial institutions and between financial institutions and the real economy.
  - Close major data constraints, especially cross-sectoral linkages at the entity level (e.g., between banks and insurers, conglomerates).
  - Enhance analytical tools and expand coverage of cross-sectoral and cross-border linkages.
  - Improve cooperation in cross-border supervision and resolution, including development of resolution plans for foreign subsidiaries and enhanced inter-agency and college collaboration and coordination.
- Desirable future coverage:
  - Broader institution and activity coverage, including institution-level interconnectedness analysis among banks and NBFIs, and both direct and indirect channels (common exposures).
  - Challenges when large institutions are lightly supervised or outside supervisors’ remit (e.g., NFCs) and in mapping interlinkages in markets with NFCs, government agencies, and foreign institutions.
- Data access needs:
  - FSAP teams often lack access to exposures across G-SIBs, financial conglomerate data, activity-based FMI data, and cross- and common-exposure data among banks and NBFIs.
  - Some national authorities use activity-based data from clearing and depository institutions.

### Systemic liquidity
- Systemic liquidity analysis is closely linked to interconnectedness assessments.
- Definition and distinction:
  - Systemic liquidity risk: risk that multiple institutions simultaneously face liquidity difficulties.
  - Difference from institution-level liquidity risk: amplification effect through interconnectedness across the financial system.
- Characteristics and sources:
  - Can emerge in certain markets (e.g., repos) involving a broad range of participants and require activity-based analyses.
  - A liquidity shock in one segment (e.g., investment funds suffering mass redemptions) can spill over to another segment (e.g., banks where funds keep liquid deposit assets).

*Source: 2021 FSAP REVIEW—BACKGROUND PAPER ON QUANTITATIVE ANALYSIS, INTERNATIONAL MONETARY FUND*

### 59.      A liquidity stress in a part of the financial system could turn into systemic shock

### 59.      A liquidity stress in a part of the financial system could turn into systemic shock

### Mechanisms of systemic liquidity risk
- Market dislocation:
  - Arises in financial systems that are primarily reliant on wholesale market instruments when institutions face difficulties obtaining funding (funding risk) because of widespread dislocations of money and capital markets (IMF, 2011).
  - Dislocation involves a wide range of institutions and financial instruments.
  - Interaction of market and funding liquidity stresses can amplify the effects of a relatively small trigger (Brunnermeier and Pedersen, 2008, and Adrian and Shin, 2010), causing institutions to fire-sell assets and further depress the market.
- System-wide liquidity shortage:
  - Occurs in bank-dominated financial systems with little market funding, driven by maturity mismatch and a system-wide loss of deposits (wholesale — e.g., government, corporate, and NBFI deposits — or retail).
  - Triggers include common underlying drivers such as capital outflows, commodity price shocks, sovereign distress, or other issues that lead to a spike in risk aversion and liquidity needs across sectors.

### Challenges in conducting system-wide liquidity stress tests
- Data availability and integration:
  - Complete system-wide liquidity stress testing remains a challenge due to significant gaps in collating data across different types of financial institutions and economic sectors.
  - A comprehensive stress test requires granular activity-based data, possibly through FMIs or by merging multiple databases collected by various financial regulatory agencies; such data have rarely been made available to FSAPs.
  - In some cases data volumes create confidential data processing demands that are hard to accomplish without longer, more intensive engagement.
- Modeling behavior under stress:
  - The need to model participants’ behavior in stress—similarly to bank liquidity stress tests—is an outstanding challenge.
  - Progress amongst national authorities has been slow, reflecting a need for significant collaboration across multiple regulatory agencies to integrate their extremely detailed databases.

### FSAP approaches and empirical work
- Mapping aggregate financial linkages:
  - Recent FSAPs have mapped main aggregate financial linkages to identify key funding and liquidity markets, interconnections, and roles of different participant types.
  - Ideal data include granular who-to-whom exposures (e.g., flow of funds by counterpart or balance sheet approach, BSA, data) and exposure data by instruments (examples include Romania, Nigeria, Thailand, and Philippines FSAPs).
- Examples of deeper analysis:
  - The 2017 Luxembourg FSAP conducted a detailed liquidity analysis of mutual funds; Article IV examined the link between banks and mutual funds through deposits.
  - MCM developed a new tool to assess system-wide liquidity stress caused by balance of payment shocks in small open economies and their spillovers across economic sectors using BSA data.
  - The 2020 Philippines FSAP applied this tool to assess potential liquidity stress spillovers between banks and NFCs under loan moratorium programs introduced to counter COVID-19.
- Illustrative studies (footnote examples in source):
  - Paddrik and others (2016) examined the U.S. Comprehensive Capital Analysis and Review assumptions and their impact on CDS market participants through margin calls.
  - Levels and others (2018) analyzed the impact of Brexit on drivers of CDS transactions in the Netherlands.

### Behavioral, operational, and regulatory factors affecting propagation
- Behavioral assumptions are critical:
  - Propagation depends heavily on participant behavior: fire selling of assets and their pecking order, hoarding cash, discontinuing market-making, etc.
  - Regulatory requirements are likely to drive parts of this behavior.
- Market infrastructure and FMIs:
  - FMIs and their operational frameworks differ across key markets, affecting their resilience and role in transmitting liquidity shocks across participants.
- Supervisory reporting:
  - Supervisory information—such as contingent financing plans of financial institutions—could help inform behavioral assumptions in stress tests.

### Role of crisis management frameworks and backstops
- Incorporating government and central bank support:
  - Typical liquidity stress tests do not incorporate central bank support, but judging system-wide resilience would be more appropriate if liquidity support is included in cases of systemic liquidity stress.
  - Financial institutions should hold sufficient liquidity buffers to counter institution-specific shocks, though not necessarily under system-wide distress.
- Perimeter of systemic liquidity support:
  - Extent of support—especially to NBFIs and certain markets—could affect liquidity stress test results.
  - Availability of deposit insurance, government backstops to emergency liquidity facilities, and FMIs would alter agents’ behavior in these markets.

### Cross-jurisdictional capacity differences and implications
- Reserve-currency jurisdictions:
  - Stress could be successfully mitigated in jurisdictions with reserve currencies because central bank backstop capacity is little constrained.
  - Major central banks during the GFC and early months of the COVID-19 crisis mitigated systemic liquidity stress successfully, expanding the perimeter of liquidity support to non-traditional counterparts and developing new instruments.
- Small open economies without reserve currencies:
  - The same mitigation capacity does not necessarily apply; when systemic liquidity shocks originate from balance of payment stress, such economies need external finance to mitigate systemic liquidity stress.

### Development agenda for FSAP systemic liquidity analysis
- Near-term priorities:
  - Develop targeted and manageable pilot exercises building on past FSAP experience.
- Components of a comprehensive systemic liquidity stress test:
  - Include liquidity stress tests of key institutions, incorporating spillover effects via direct exposures and major liquidity markets.
  - Use both institution-level liquidity position data and activity-based market transaction and positioning data.
- Data and technical capacity needs:
  - Conducting analysis at the contract and securities level could require substantial investments in big data processing capacity.
  - Explore collaboration options with regulatory agencies for data processing and analysis, given specific challenges.

*From: ppea2021041 - 59.      A liquidity stress in a part of the financial system could turn into systemic shock*

### 79.      These approaches point to the limitations of applying static and quasi-static

### These approaches point to the limitations of applying static and quasi-static assumptions of bank balance sheet growth and exogenous scenarios in standard bank stress tests

### Limitations of static and quasi-static stress test assumptions
- Standard stress tests that assume static or quasi-static bank balance sheet growth and exogenous scenarios can mis-rank banks: the relative performance (ranking of impact on capital ratios) changes with balance sheet growth assumptions and when macro-feedback effects are incorporated.
- Banks that appear resilient (vulnerable) under a static balance sheet and exogenous scenarios could turn out to be vulnerable (resilient) when assessed through a dynamic balance sheet approach with macro-feedback effects and endogenous scenarios.
- Stress-testing frameworks should incorporate dynamic balance sheets, macro-feedback effects, and endogenous scenarios to more accurately capture bank vulnerabilities.

### Agent-Based Models (ABMs) as a simulation-based approach (paragraphs 80–83)
- ABMs are a promising simulation-based approach to capture macro-financial linkages “from the bottom-up,” modeling individual agents (firms, households, banks, central banks, sovereigns) with heuristic behavioral rules and often bounded rationality.
- Compared to typical stress test models, ABMs can include:
  - richer behavioral rules;
  - comprehensive agent groups beyond banks (nonbanks, households, firms, sovereigns, central banks);
  - explicit financial contracts and markets determining price formation and fair valuation;
  - regulatory, market, and internal constraints (capital, liquidity ratios, leverage ratios, internal risk limits).
- ABMs enable assessment of complex interactions and systemic impact via granular agent behavior, network interconnectedness, and state dependence (non-linearities), which is essential for macro-financial stress tests focused on tail events and out-of-equilibrium dynamics.
- Current status: agent-based stress testing models have been developed based on simplified balance sheet structures and incorporate solvency-liquidity and macro feedback effects, but they remain experimental and have not yet been used in FSAPs.

### Prototype ABM developments and findings (paragraphs 84–85)
- Prototype model (Valderrama, forthcoming):
  - Simulations show temporary shocks can morph into long-lived shocks that erode banking sector solvency, depress credit growth, and undermine economic growth.
  - Findings indicate that tightening only bank regulatory requirements or restricting market access at the local level may be ineffective due to linkages between the banking sector, the securities market, and the credit market.
  - Policy implication: a system-wide perspective to prudential regulation, including banks and nonbanks, is needed.
- Second prototype (Eurace 2.0 model; Gross, Hilberg, Hoog, and Kohlweyer, forthcoming):
  - Larger-scale macro-financial ABM featuring banks, households, firms, a sovereign, and a central bank.
  - Designed to assess borrower-based and capital-based macroprudential policy measures within a rich bank loan-granting process for households (mortgages) and firms (investment).
  - Comprises an integrated balance sheet structure between micro agents within and across groups of agents.

### Macroprudential policy use of quantitative analysis (paragraphs 86–88)
- FSAPs provide comprehensive assessments of potential vulnerabilities: rapid overall credit growth, sectoral vulnerabilities (household and corporate indebtedness), liquidity and FX mismatches, and structural interconnectedness including between bank and nonbank financial systems.
- FSAPs can formulate macroprudential policy recommendations using early warning and leverage indicators (guidance note Table 1 in IMF, 2014) and FSAP risk assessment incorporating existing risk mitigants.
- Examples:
  - 2018 France FSAP: identified corporate sector vulnerabilities via interest coverage ratios (debt-at-risk) and recommended actions; found household vulnerabilities more contained due to shift towards fixed-rate mortgages and leveling off of house prices.
  - 2018 Romania FSAP: used credit register data to identify DSTI sensitivity thresholds guiding DSTI cap calibration.
  - 2017 Netherlands FSAP: used a DSGE model to assess macroeconomic impacts of a housing shock for different LTV ratios.
  - 2018 Peru FSAP: assessed impact on bank lending of tightening capital buffers using bank-level event-study analysis.
- Future directions:
  - More explicit use of solvency and liquidity stress test methods to inform macroprudential stance, with stress test severity linked to economic and financial buoyancy.
  - Stress test results can inform buffer sizes (e.g., countercyclical capital buffer or sectoral buffers).
  - Microdata-based analytical tools for calibrating borrower-based tools (LTV, DSTI) and quantitative assessment tools for nonbank financial system risks and policy calibration.

### Improving efficiency of core quantitative tools (paragraphs 89–94)
- Need: improve efficiency of core risk assessment tools to expand FSAP risk analysis within established resource envelope while preserving independent FSAP assessments and covering emerging risks, interconnectedness, and macro-financial linkages.
- Trends:
  - Use of quantitative tools expanded over past five years while overall FSAP cost remained broadly flat.
  - More FSAPs cover risks from NBFIs and interconnectedness; central banks have increased financial stability analysis resources.
- Standardization efforts:
  - Staff will standardize core quantitative tools for different data environments, with detailed guidance notes and files/codes on a refreshed webpage to support FSAP teams.
  - Developing a tool to efficiently estimate satellite models for stress tests and to check their performance.
  - Shift from excel-based tools to program codes to increase efficiency, accuracy, and validation.
- Constraints:
  - Need to tailor risk analysis to country-specific conditions: differing bank-level supervisory reporting formats, accounting standards, bank business models, loan-loss provisions, collateral valuation, LTV calculation, securities/derivatives valuation, off-balance sheet treatment, and supervisory intrusiveness.
  - Some jurisdictions follow Basel III for internationally active banks, others apply different rules for domestic banks, and many jurisdictions without global banks follow Basel II or I rules.
  - Standardized tools must be flexible to handle country-specific risks, transmission channels, regulations, and data.
- Ongoing projects and tools:
  - Bank stress test tools: MCM updating internal guidance notes and excel tools; collating into internal operational reference note for FSAP teams; projects to develop stress testing codes.
  - Stress test scenarios: creation of MCM modeling unit to improve efficiency and consistency across FSAPs.
  - Growth-at-Risk (GaR): user-friendly package with Excel interface and underlying Python codes; published on GitHub.
  - Corporate sector stress test tools: corporate risk assessment tool under development (Tressel and Ding forthcoming); access to BuDA tool for macro scenario corporate stress tests with limited human resources.
- External dissemination: publishing methodological notes will facilitate discussions with national authorities and improve communication.

### Emerging risks — Climate change (paragraphs 95–97 and Box 1)
- Climate analysis challenges for FSAPs:
  - Need to obtain data and build frameworks to assess which climate risks are material for members.
  - Need to design plausible scenarios at different horizons; choice of horizon matters (typical bank stress tests use the 3-5-year horizon; climate literature often considers pathways up to 2100).
  - Staff propose assessing both short- and long-term financial stability risks from climate change and mapping macro-financial consequences into FSAP stress testing frameworks.
- Two broad categories of climate-related financial stability risk:
  - Physical risks: damages from direct and indirect consequences of climate change, manifesting as acute physical risk (extreme events such as cyclones, floods) or chronic physical risk (sea-level rise, drought).
  - Transition risks: consequences of changes in public policies and technology aimed at mitigation and adjustment to a lower-carbon economy, often modeled as higher carbon price scenarios.
- Box 1 — Long vs. Near-Term Scenarios:
  - NGFS is developing scenarios over an 80-year horizon (2020-2100) based on global temperature targets and the “Middle of the road” Shared Socioeconomic Pathway.
  - Representative NGFS scenarios have been used by Banque de France and planned for Bank of England Biennial Exploratory Scenario (BES).
  - Other approaches use shorter horizons (e.g., 5-year) not explicitly tied to temperature outcomes (Dutch Central Bank, ESRB).
  - Complementarity: long-term scenarios capture pluri-decadal impacts and societal response; near-term scenarios focus on immediate dangers and operational indications relevant for FSAPs’ system-wide risk gauging and mitigation proposals.

*Source: 2021 FSAP REVIEW—BACKGROUND PAPER ON QUANTITATIVE ANALYSIS (excerpts).*

### 98.      FSAPs have been assessing the impact of climate-related natural disaster events on

### 98.      FSAPs have been assessing the impact of climate-related natural disaster events on

### Climate-related natural disasters and climate risk in FSAPs
- A textual analysis of 192 FSAP reports (up to 2019) found that 33 (17 percent) contained meaningful references to risk factors such as droughts, floods, and storms.
- Many instances were for small island states (such as the Bahamas, Jamaica, and Samoa), but assessments for advanced economies (such as the United States, France, Belgium, Denmark, and Sweden) have also covered natural catastrophe risks as part of insurance stress testing.
- More recent FSAPs have developed new approaches to assessing climate change risk, including:
  - Transition risk in the 2019 Norway FSAP (Box 1 referenced).
  - Physical risk in the 2021 Philippines FSAP.
- Significant effort and collaboration will be needed to develop stress testing tools for climate change risks and deploy them regularly in FSAPs. This work will require close inter-departmental collaboration within Fund departments and with partners such as the World Bank, the United Nations (UN), and the NGFS.

### Transition risk — Norway FSAP pilot (Box 2)
- The pilot explored three possible transmission channels for transition risk shocks to the financial system:
  - The impact of a substantial increase in domestic carbon pricing on banks’ credit exposures, such as loans, via its effect on corporates’ operating costs and profitability, under severe assumptions.
  - The impact of a drastic increase in global carbon prices on the domestic economy on banks’ loan losses via the fall in the revenues of domestic oil producers.
  - The impact of a forced reduction in the production of domestic oil firms on their share prices and, in turn, on the net wealth of domestic shareholders (such as households or financial and nonfinancial corporates).
- Results show that a sharp increase in carbon prices would have a significant but manageable impact on banks (IMF, 2020b and Grippa and Mann, 2020).

### Cyber risk: vulnerabilities, FSAP practices, and findings
- Cyber risk is a growing source of potential systemic risk to financial sectors.
- Financial systems are particularly vulnerable given reliance on ICT systems; attack entry points include in-house systems and systems of third-party vendors, contractors, clients, retail partners, or counterparties.
- Cyberattacks can spread quickly through ICT interconnection and financial system interconnectedness, and can be systemic if they target several financial institutions simultaneously or a systemically important financial institution (SIFI).
- Potential systemic impacts include collapse of key liquidity markets and FMIs, correlated reputational risks, stalled payments and settlement transactions, liquidity crunches to banks, and mass insurance claims.
- Quantitative analysis of fintech and cyber risks has been limited due to data constraints. Examples of FSAP approaches:
  - Namibia: gathered descriptive information on cybersecurity practices through interviews.
  - Poland: collected potential losses from cyberattacks through questionnaires.
  - Bouveret (2019): explores potential loss estimates as research work (not in FSAP’s financial stability risk assessment).
- Euro Area FSAP cyber risk-motivated liquidity stress test:
  - Simulated a scenario in which banks could not access their collateral at central counterparties for five business days.
  - Identified vulnerabilities, especially at internationally active banks.
  - Demonstrated that conventional liquidity, solvency, and interconnectedness toolkits can assess some aspects of cyber risk.
- Singapore FSAP cyber analysis highlights:
  - Authorities provided data on historical cyberattacks; banks provided cyber risk scenarios and associated loss estimates and management actions.
  - Scenarios were organized in a Cyber Risk Assessment Matrix (application of FSAP Risk Assessment Matrix to cyber risk).
  - Insurers were asked to estimate the losses they would incur if their ten largest policyholders of affirmative and silent cyber coverage experienced cyberattacks and claimed on their policies; losses were assessed to be manageable after reinsurance recoveries, but insurers identified a need to restrict implicit cyber coverage.
- Data constraints on cyberattacks and losses remain key:
  - Event frequency, distribution across sectors, and cybersecurity budgets are key indicators to track.
  - Sparse country-level cyberattack data mean models should be applied to country data rather than re-estimated.
  - Cyber risk mapping—listing financial and nonfinancial institutions and IT links and exposures—can support institution-level impact estimation using operational risk techniques.
  - Under the Basel rule, banks must set aside capital for operational risks, including cyber risks.

### Fintech: early-stage quantitative analysis and pilot findings
- Quantitative analysis of financial stability risks from fintech is still at an early stage; many nascent risks are operational in nature.
- Need for a conceptual framework to model incentives for financial innovation and risk-taking by incumbents and entrants, the role of market structure, and government policies balancing risk-return tradeoffs.
- Data gaps and a rapidly changing landscape hinder quantitative assessment.
- Pilot FSAP analyses (2019 Singapore and 2019 Korea) estimated:
  - Potential non-interest income reduction for incumbent banks conditional on increased competition by new entrants.
  - In Singapore, potential gains from fintech via changes in the unit cost of financial intermediation.
    - The analysis suggested that the unit cost of financial intermediation in Singapore has been around 1.5 to 2 percent for the past decade, similar to that of the United States, indicating both scope for eroding bank earnings and potential gains from fintech development.
  - In Korea, an income shock was introduced into the stress testing framework to illustrate potential impacts on bank capital.

### Data constraints: availability and access, and ways to improve
- FSAP missions face two sources of data constraints: availability (gaps) and access (confidentiality).
  - Some data do not exist—especially for interconnectedness, shadow banks, and emerging risks.
  - Rapidly changing landscapes (fintech, shadow banks) can render reporting formats obsolete.
  - Accessibility depends on authorities’ willingness to share; for global data involving multiple jurisdictions, permission from all is required.
- Improvements in conventional risk data since the GFC:
  - G-20 Data Gap Initiatives (G20 DGI) aim to close post-GFC data gaps, covering monitoring risk in the financial sector, international network connections, and sectoral and other financial and economic datasets.
  - As of 2020, progress has been made though gaps remain with data for systemic risks for insurers, sectoral accounts, currency composition of IIPs, and commercial property prices (FSB and IMF, 2020).
  - Ongoing discussion to extend the exercise beyond the end of the current initiative in 2021.
  - FSB now publishes an annual global report on NBFI.
  - Activity-based data from centralized FMIs and big data techniques allow monitoring of transactions and exposures that include shadow banks and NBFIs.
- FSAP access to confidential supervisory data:
  - For mandatory financial stability assessments, jurisdictions providing access to confidential supervisory data for quantitative analysis increased from 75 percent in the first assessments to 97 percent in the latest.
  - In some jurisdictions (including European countries under the Single Supervisory Mechanism), access has been allowed only in specific physical locations (secure “data room”), increasing FSAP team costs.
  - For voluntary financial stability assessments in the past five years, all jurisdictions have shared at least some confidential supervisory data, though breadth and quality varied.
- Cross-border interconnectedness and FMI data:
  - Publicly available country-level BIS cross-border banking statistics have been useful; country-aggregate ultimate-borrower-basis data are publicly available.
  - Locational data and institution-level G-SIB interconnectedness data can further strengthen FSAP contagion analysis when national authorities permit access.
  - The Fund typically has little access to activity-based data collected by FMIs (often private sector companies), though some recent FSAPs have made exceptions (e.g., 2019 France FSAP interconnection map effort).
  - Data aggregation from multiple agencies with different confidentiality protocols creates data management challenges even when authorities are willing.
- Data existence and access vary by emerging risk:
  - Climate change: relatively ample climate data and forecasts from climatology literature; catastrophe modeling for physical risk exists but tends to focus on advanced economies with high insurance coverage.
  - Cyber-risk: cyberattack databases are being constructed by national authorities and private platforms but may be incomplete; incident reports may omit monetary loss figures; FSAP access may require national security clearances beyond typical confidentiality agreements.
  - Fintech: suffers the most from data gaps due to novelty and rapid transformation of the industry landscape.
- Recommendations and approaches to improve FSAP data constraints:
  - Improving data existence: IMF’s statistics department and IT department provide technical assistance to develop many country-aggregate data in the context of the G20 Initiatives; FSSR for lower-income economies discusses financial data development and includes data components in TA roadmaps.
  - Improving access to data: staff welcomes Board support for encouraging national supervisory agencies to share data with FSAP teams as standard practice; electronic remote access could substantially improve efficiency and save costs.
  - Exploring alternative approaches: develop codes to implement analysis and ask counterparty authorities to run them and return results; while Excel is currently the core tool and requires direct data access, a code-based approach could help circumvent data access constraints though it might make FSAP-like analysis less accessible to Article IV teams.

*Source: ppea2021041 - 98.      FSAPs have been assessing the impact of climate-related natural disaster events on*

### 114.      FSAP quantitative tools can help strengthen macrofinancial analysis in Article IV

### 114.      FSAP quantitative tools can help strengthen macrofinancial analysis in Article IV

### Use of FSAP quantitative tools in Article IV reports
- Article IV reports have used a variety of vulnerability indicators, balance sheet analysis, and Growth-at-Risk tools.
- Vulnerability indicators
  - Indicators used in FSAPs have been used to analyze the links between the financial cycle and the business cycle.
  - The credit-to-GDP gap, defined as a deviation of credit from the simple HP trend or a trend estimated using a semi-structural model, has been used to inform recommendations on system-wide macro-prudential tools such as the countercyclical capital buffer.
- Growth at risk
  - Some Article IV teams have used the GaR as a forward-looking tool for the assessment of downside risks to growth and the identification of vulnerabilities that can trigger systemic risk.

### Planned expansion of FSAP quantitative tools for Article IV surveillance
- Staff plan to provide a broader menu of FSAP quantitative tools; desk economists could choose priority areas and corresponding tools depending on country financial vulnerabilities.
- Simple stress testing tools
  - The standardized GST and UST tools using publicly available data would allow Article IV teams to carry forward the bank solvency analysis.
  - The system-wide FX liquidity stress testing tool uses BSA data and links system-wide liquidity shortages from balance of payment shocks to bank liquidity stress test analysis.
  - The tool could enrich the reserve adequacy discussion in Article IVs.
- NFC and household vulnerability assessment tools
  - New macro scenario stress testing tools developed by Tressel and Ding (forthcoming) for NFCs could help strengthen NFC vulnerability assessment, especially in the context of COVID-19.
  - MCM is developing multiple household vulnerability assessment tools, and some are explicitly lined to calibrate borrower-based MPMs.
- Tools to analyze macro-feedback effects and inform macroprudential policy advice
  - Macro-financial linkages models under development could be used in Article IV surveillance to capture two-way macro-financial feedback effects.
  - Some of these models can help inform the calibration of broad-based macroprudential tools.
- Stress testing results informing other policies
  - Elements of FSAP stress testing can inform policies not covered by an FSAP but relevant for Article IV surveillance.
  - Examples:
    - Liquidity stress tests can be used to calibrate the reserve requirement as a prudential tool rather than a monetary policy tool when adequate.
    - Foreign currency liquidity stress tests could be used to inform the adequacy of international reserves and assessment of exchange rate misalignment.
    - Borrower-based MPMs could be informed by household and corporate sector analysis.
    - Cyclical capital measures could be calibrated using macro-financial feedback effect models that incorporate bank stress tests.
    - Analysis of the sovereign-bank nexus in solvency stress tests could inform public debt sustainability analysis.
    - Stress tests for climate risk could trigger policy advice on demographic, industrial, labor policies to limit the ultimate financial stability impact of climate change.

### Appendix I — Approach to Assessing Climate Change Risk in FSAPs
- Three-stage approach
  - Stage 1: diagnostic and assessment of principal sources of climate risk facing jurisdictions.
  - Stage 2: link identified risks to specific scenarios of the evolution of physical and transition climate risks.
  - Stage 3: map climate scenarios into banking resiliency using either a macro approach (map stage 2 scenarios into macro-financial scenarios and apply standard stress testing) or a micro approach (borrower-level assessments building up to bank-level stress tests), depending on data granularity and scope.
- Purpose and limitations
  - Climate risk scenario analysis is not a standard fail-or-pass stress test; objective is to assess pressures on capital to gauge magnitude of the challenge and the need for adaptation.
  - Developing credible climate scenarios is difficult due to unprecedented uncertainty and pronounced sectoral and geographical granularity of impacts.
- Stage 1 — Climate Financial Risk Diagnostic
  - MCM will build a heatmap to help FSAP teams decide scope and relevant physical and transition risks.
  - FSAP teams will develop a climate risk assessment matrix (C-RAM) informed by a global climate risk assessment in the G-RAM.
  - The heatmap will pull together data from identified datasets and consider spillovers (e.g., higher imported energy prices due to higher carbon taxation in trading partners).
- Stage 2 — Designing Climate Scenarios
  - Climate change is generally considered a long-term challenge.
  - Physical risk
    - Designing scenarios entails mapping emissions and global temperature pathways into projections of climate-related events typically over a 50-80-year horizon.
    - Use future projected distribution of disasters, compare with historical distribution, and apply near-term and long-term future distributions to assess damages and balance sheet impacts.
    - If modeling future distributions is not feasible, staff could go further into the tail of historical distributions to mimic climate shocks (e.g., as in catastrophe insurance models that consider once in 250-500-year events).
    - The approach taken in the 2021 Philippines FSAP is noted as an example.
  - Transition risk
    - Transition to a low-carbon economy is typically long-term; uncertainties exist over emissions and temperature trajectories.
    - NGFS scenarios will be leveraged initially to assess financial stability risks over the traditional 3-to-5-year horizon in FSAP work.
    - The NGFS scenarios include a range of carbon tax increases, from no increase in carbon taxes to $100 in 2025.
    - Staff will explore NGFS scenarios over the long term to analyze opportunities from higher carbon prices over the next 30 years and consider higher carbon price trajectories for sensitivity analysis.
  - Sudden transitions
    - Stress tests may consider up-front shocks to carbon prices to illustrate pressure points.
    - One-off shocks could simulate abrupt materialization of transition risk (a “Minsky moment”) affecting macro-financial variables and industry- or firm-specific asset valuations.
- Stage 3 — Mapping Climate Scenarios into Financial Stability
  - Map temperature paths, physical risk distributions, and transition risk materialization into impacts on the macroeconomy and bank capital.
  - Use projections of climate events, damages, and transition impacts as inputs in macro-financial models to estimate impacts on economic or sectoral growth and other macro and financial variables.
  - Once macro-financial scenarios are built, apply standard FSAP stress testing approaches for credit and market risks to assess risks and impacts on bank capital.

*Source: ppea2021041 - 114.      FSAP quantitative tools can help strengthen macrofinancial analysis in Article IV*

### 10.      The scope and depth of the analysis that maps climate scenarios into the banking

### 10.      The scope and depth of the analysis that maps climate scenarios into the banking sector’s health

### General approaches and trade-offs
- Two general approaches:
  - Macro approach: focuses on macro-financial transmission channels of climate scenarios and assesses the impact of macro scenarios on the banking sector (possibly using data on the distribution of financial exposures by industry or regions).
  - Micro approach: builds on geographical exposures of the financial sector, detailed sectoral analysis, industry and firm-level data, data on balance sheets of households and the government, and collateral values to link them to the earnings of financial institutions.
- Comparative observation:
  - A more micro approach will provide a more accurate assessment because the effects of physical and transition risk will vary across different sectors and firms.
  - The micro approach requires granular bank exposure data, macro models that account for differentiated shocks across industries, and analysis of corporate and household sectors integrated with bank stress tests.
  - Depending on specific climate risk and data granularity, a combination of the macro and micro approaches could be considered (including the impact of macrofinancial scenarios on sectors’ financial statements).
  - In practice, FSAPs may fall on a spectrum between these two approaches depending on the availability of data and models.
  - In the short term, the macro approach may be the more feasible option given current resource constraints.

### Macro modeling approaches (mapping climate scenarios into macro-financial shocks)
- Standard macro-financial models:
  - Standard IMF/FSAP macro-financial models with relatively less cross-industry detail could be extended to simulate climate shocks on macro and financial variables.
  - Research Department staff are developing extensions of the Fund’s multi-country models with climate features to support FSAPs once ready.
  - In the meantime, staff are exploring models from external vendors (also used by the NGFS) and a small macro model developed in-house in MCM.
  - For physical risks: empirical findings from disaster/climate economic models could help calibrate shocks (such as productivity and capital depreciation shocks).
  - For transition risks: models could incorporate economy-wide effects of carbon tax, technology and preference shocks, and modeling opportunities from transition risk (the rise of green industry).
- Macro models combined with cross-industry models:
  - Macro models can be used in combination with CGE models (such as the GTAP model maintained by the Research Department) to generate greater sectorization in scenario design.
  - Distinct disaster-impact models that separate infrastructure sector from sectors producing final goods and services can serve as inputs to macro models.
- Examples of carbon-price path and shock assumptions (Level of assumed carbon prices in scenario (2020 US$/ton CO2)):
  - NGFS: Orderly 0103132350; Disroderly 21017841; Hothouse 77711
  - Bank of France: Orderly transition 05075180; Delayed transition 000700; Sudden transition 00170900
  - Bank of Canada: 2°C (consistent) scenario 080190600; 2°C (delayed action) scenario (abrupt transition) 000800; Nationally determined contributions 02070190
  - ECB: Orderly 760114360; Disroderly 75468845; Hothouse 7141416
  - ESRB: Sudden transition USD 100 per tonne carbon price shock 100
  - Dutch National Bank: Sudden transition USD 100 per tonne carbon price shock 100
  - WEO: Adverse (average) 1/71927101
  - 1/ Assumes the baseline is the NGFS hothouse scenario.

### Micro approach (sectoral and firm-level analysis)
- Core features and requirements:
  - Directly examine financial performance of affected sectors to which banks are exposed.
  - Very high requirement for data granularity and industry-specific knowledge.
  - Requires micro-data on industry and geographical characteristics of financial institutions’ underlying sectoral exposures and assessment of borrowers’ capacity to pay, incorporating cross-industry effects.
  - Models using micro-data (e.g., cash flow models) estimate vulnerability of sectors (in terms of earnings and their volatilities) to physical or transition risks.
  - Vulnerability analysis used to revalue financial institutions’ exposures, for example by incorporating long-term losses into asset valuations.
- Feasibility:
  - Given complexity and resource constraints, the micro approach is likely feasible in only a few FSAPs where central banks have developed the needed data and modeling.

### Operational considerations and next steps
- Country specificity and experimentation:
  - The Annex lays out a general approach that will be country specific and reflect available data and modeling tools.
  - Staff will need to experiment and learn from experience; further details will be developed in an MCM paper planned for later this year.
- Collaboration:
  - Close collaboration required with country teams, functional departments, the World Bank and the NGFS.
  - Collaboration opportunities include: (i) judging materiality of different climate risks for individual country cases, (ii) scenario design including projecting hazards and estimating damages and losses, (iii) carbon price paths, (iv) implications for scenario design of the use of carbon tax proceeds, (v) modeling approaches (including coverage of all member countries), and (vi) challenge of physical climate risk for financial stability in smaller members.
  - Noted scope for collaboration with the World Bank on materiality assessment and deeper analysis of physical risks leveraging the Bank’s expertise in catastrophe insurance and financing.
- Data needs:
  - Climate change analysis requires novel and very granular data.
  - Staff will coordinate internally and with the World Bank, NGFS partners, and outside vendors to seek synergies and cost efficiencies on access to needed data.
- Applicability beyond banks:
  - The methodology could be applied to other financial sectors such as insurance companies and mutual funds.
  - For insurance companies and mutual funds, stage 1, 2 and part of stage 3 that pertains to macro scenarios would be the same while scenario analysis methodology for each sector would be different.
- Beyond FSAPs:
  - The framework for climate risk scenario analysis is intended to be of value to members through capacity development work including in support of FSSRs.
  - Resources permitting, there could be an opportunity to provide assessments of physical risk facing financial systems in fragile states.

### Examples of country/institutional practice (selected operational features)
- Bank of England:
  - Approach: Bottom up
  - Climate Scenarios: 3 scenarios provided by BoE (from NGFS): early, late, no policy action (carbon price, tech change, consumer preferences, emissions, temperature; frequency and severity of climate events; productivity)
  - Data granularity: Corporate exposures, household exposures (assessment done by banks)
  - Stress testing horizon: 30 years (60 years for physical risk)
  - Reporting frequency: Every 5 years
  - Static versus dynamic balance sheet: Static (dynamic responses captured through a qualitative questionnaire)
- Bank of Canada:
  - Approach: Top down
  - Climate Scenarios: 4 scenarios (from IPCC): no action, NDCs, 2C (consistent), 2C (delayed action)
  - Data granularity: 18 regions, 33 sectors
  - Stress testing horizon: 80 years
  - Reporting frequency: Every 5 years
- Bank of France, Dutch National Bank, ESRB, ECB (selected comparisons):
  - Data granularity examples:
    - Bank of France: 55 sector in CGE model (assessment done by banks)
    - Dutch National Bank and ESRB: bond and equity holdings at the level of individual securities; banks’ corporate loan exposures disaggregated by risk classes and industries
    - ECB: bond and equity holdings (at the level of individual securities); banks’ corporate loan exposures (at the level of individual firms), country-level assessment for HHs
  - Stress testing horizons in table entries: 30 years; 5 years; 1 year; 30 years (varies by institution)
  - Reporting frequency examples: Every 5 years; 1 year
  - Static versus dynamic balance sheet: entries include Static (2020-25) and Dynamic (2025-50); Static; Dynamic; Static and Dynamic (10 years)

*Source: ppea2021041 - 10.      The scope and depth of the analysis that maps climate scenarios into the banking sector’s health*

### References

### References

### Systemic risk, stress testing, and macroprudential frameworks
- Acharya, V., R. Engle, and M. Richardson. 2012. Capital Shortfall: A New Approach to Ranking and Regulating Systemic Risks. American Economic Review, 102, 3, 59-64.
- Adrian, Tobias and Markus Brunnermeier. 2016. CoVaR. American Economic Review, 106(7): 1705-41.
- Adrian, T., N. Boyarchenko, and G., Domenico. 2017. Vulnerable Growth. Federal Reserve Bank of New York. Staff Report No. 794 September 2016 Revised November 2017.
- Basel Committee on Banking Supervision (BCBS). 2015. Making Supervisory Stress Tests More Macroprudential: Considering Liquidity and Solvency Interactions and Systemic Risk. Working Paper 29, Bank for International Settlement.
- BCBS. 2018. Stress Testing Principles, Guidelines.
- Cont, R., A. Kotlicki, and L. Valderrama. 2020. Liquidity at Risk: Joint Stress Testing of Solvency and Liquidity. Journal of Banking and Finance, Vol. 118.
- Krznar, I., and T. Matheson. 2017. Towards Macroprudential Stress Testing: Incorporating Macro-Feedback Effects. IMF Working Paper 17/149.
- Catalan, Mario and Alexander Hoffmeister. 2020. When Banks Punch Back: Macrofinancial Feedback Loops in Stress Tests. IMF Working Paper No. 20/72.
- Gray D. and A. Jobst. 2013. Systemic Contingent Claims Anlysis – Estimating Market-Implied Systemic Risk. IMF Working Paper No. 13/54.
- Carabenciov, I., C. Freedman, R. Garcia-Saltos, O. Kamenik, D. Laxton and P. Manchev. 2013. GPM6: The Global Projection Model with 6 Regions. IMF Working Paper No. 13/87.
- Gross, M, D. Laliotis, M. Leika, and P. Lukyantsau. 2020. Expected Credit Loss Modeling from a Top-Down Stress Testing Perspective. IMF Working Paper No. 20/111.
- Gross. M. and J. Poblacion. 2017. Implications of model uncertainty for bank stress testing. Journal of Financial Services Research, Vol. 55.
- Han, F., Leika M. 2020. Integrating Solvency and Liquidity Stress Tests: The Use of Markov Regime-Switching Models. IMF Working Paper No. 19/250.
- Lipinsky, F. and M. Miescu. 2019. Macro-financial Risk and Capital Gap Analysis (Cap-Gap). Technical Note to France FSAP.

### Liquidity, leverage, market microstructure, and contagion
- Adrian, T. and Shin H. 2010. Liquidity and Leverage. Federal Reserve Bank of New York Staff Reports no. 328.
- Brunnermeier, M.K. and Pedersen, L.H. 2008. Market liquidity and funding liquidity. The review of financial studies, 22(6), pp.2201-2238.
- Brunnermeier, Markus, and Yuliy Sannikov. 2014. A Macroeconomic Model with a Financial Sector. American Economic Review 104 (2): 379–421.
- Allen, F. and D. Gale. 2000. Financial Contagion. Journal of Political Economy, 108(1).
- Allen, F. and D. Gale. 2004. Competition and financial stability. Journal of Money, Credit and Banking, pp.453-480.
- Huang, W., A. Menkveld, and S. Yu. 2019. Central Counterparty Exposures in Stressed Markets, BIS Working Paper No. 833.
- Paddrik, M., S. Rajan, and H. P. Young. 2016. Contagion in the CDS Market. Office of Financial Research Working Paper No. 16-12.
- Baranova, Y., J. Coen, P. Lowe, Jo. Noss, and L. Silvestri. 2017a. Simulating stress across the financial system: the resilience of corporate bond markets and the role of investment funds. Bank of England Financial Stability Paper, No. 4.
- Baranova, Y., Z. Liu, and T. Shakir. 2017b. Dealer intermediation, market liquidity and the impact of regulatory reform. Bank of England Working Paper, No. 6.
- Coen, J., C. Lepore, and E. Schaaning. 2019. Taking regulation seriously: fire sales under solvency and liquidity constraints. Bank of England, Staff Working Paper No. 793.
- Bruneau, C., O. de Bandt, and W. El Amri. 2012. Macroeconomic fluctuations and corporate financial fragility. Journal of Financial Stability. Vo. 8.

### Network analysis, interconnectedness, and systemic mapping
- Diebold, F. X., and K. Yilmaz. 2014. On the network topology of variance decompositions: Measuring the connectedness of financial firms. Journal of Econometrics, 182, No. 1: 119-134.
- Espinosa-Vega M. A., and J. Sole. 2010. Cross-Border Financial Surveillance; A Network Perspective. IMF Working Papers 10/105.
- Cortes F., P. Lindner, S. Malik, and M. A. Segoviano. 2018. A Comprehensive Multi-Sector Tool for Analysis of Systemic Risk and Interconnectedness (SyRIN). IMF Working Paper No. 18/14.
- Covi, G., M. Gorpe, and C. Kok. 2019. CoMap: mapping contagion in the euro area banking sector. IMF Working Paper, No. 19/102.
- Bricco, J. and T. Xu. 2019. Practical Guidance on Interconnectedness and Contagion Analysis. IMF Working Paper No. 19/220.
- Dees S., F. di Mauro, M. H. Pesaran and L. V. Smith. 2007. Exploring the international linkages of the euro area: A Global VAR Analysis. Journal of Applied Econometrics.

### Agent-based, computational models, and macro-financial simulation
- Bookstaber, R. 2012. Using Agent-Based Models for Analyzing Threats to Financial Stability. OFR Working Paper 3.
- Bookstabler, R. and M. Paddrik. 2015. An Agent-based Model for Crisis Liquidity Dynamics. OFR Working Paper 15-18.
- LeBaron, B. and L. Tesfatsion. 2008. Modeling macroeconomies as open-ended dynamic systems of interacting agents. American Economic Review, 98(2), pp. 246-50.
- Tesfatsion, L. 2006. Agent-based computational modeling and macroeconomics. In Colander, D., editor, Post-Walrasian Macroeconomics: Beyond the Dynamic Stochastic General Equilibrium Model, pages 175-202. Cambridge University Press, Cambridge, UK.
- Tesfatsion, L. 2006. Chapter 16 in Agent-based computational economics: A constructive approach to economic theory. In Tesfatsion, L. and Judd, K., editors, Handbook of Computational Economics, Vol. 2, pp. 831-880.
- Aymanns, C., J.D. Farmer, A.M. Kleinnijenhuis, and T. Wetzer. 2018. Models of financial stability and their application in stress tests. In Handbook of Computational Economics (Vol. 4, pp. 329-391). Elsevier.
- Valderrama, L. forthcoming. An Agent-Based Model for Stress Testing. IMF Working Paper.
- Gross, M., Mi. Leika, and L. Valderrama. forthcoming. The Macro-financial System Simulator (MASS). IMF Working Paper.
- Gross, M., B., Hilberg, Hoog, von der S., and D. Kohlweyer. forthcoming. The Eurace 2.0 model, IMF Working Paper.

### Climate-related financial risk and scenario analysis
- Allen, T., S. Dees, J. Boissinot, C. Mateo Caicedo Graciano, V. Chouard, L. Clerc, A. de Gaye, A. Devulder, S. Diot, N. Lisack, F. Pegoraro, M. Rabaté, R. Svartzman, and L. Vernet. 2020. Climate-Related Scenarios for Financial Stability Assessment: An Application to France. Banque de France. July 2020.
- Carney, M. 2015. Breaking the tragedy of the horizon - climate change and financial stability. Bank of England Speech. September 29.
- Carney, M., F. Villeroy de Galhau, and F. Elderson. 2019. Open Letter on Climate-Related Financial Risks.
- Network for Greening the Financial System (NGFS). 2018. WS2—Macro-financial Workstream: mandate and workplan from 2018 to April 2020. Manuscript.
- NGFS. 2020. Guide to climate scenario analysis for central banks and supervisors. June 2020.
- European Systemic Risk Board. 2016. Too late, too sudden - Transition to a low-carbon economy and systemic risk. Reports of the Advisory Scientific Committee.
- Grippa, P. and S. Mann. 2020. Climate-Related Stress Testing: Transition Risks in Norway. IMF Working Paper No. 20/232.
- Vermeulen, R., Schets, E., Lohuis, M., Kolbl, B., Jansen, D.-J. and Heeringa, W. 2018. An energy transition risk stress test for the financial system of the Netherlands. DNB Occasional Studies 1607.
- CO-Firm and 2° Investing Initiative. 2017. The Transition Risk-O-Meter – Reference Scenarios for Financial Analysis.
- Cambridge Institute for Sustainability Leadership. 2015. Unhedgeable risk - How climate change sentiment impacts investment.
- Fuss, S., J. Canadell, G. Peters, M. Tavoni, R. Andrew, P. Ciais, R. Jackson, C. Jones, F. Kraxner, N. Nakicenovic, C. Le Quere, M. Raupach, A. Sharifi, P. Smith, and Y. Yamagata. 2014. Commentary: Betting on Negative Emissions. Nature Climate Change, Vol. 4.

### Fintech, cyber risk, and data initiatives
- Adrian, T. 2018. Remarks in 2018 IMF Fintech Roundtable.
- Financial Stability Board. 2019. FinTech and market structure in financial services: Market developments and potential financial stability implications, Report.
- Financial Stability Board and the International Monetary Fund. 2020. G20 Data Gap Initiative (DGI-2): The Fourth Progress Report—Countdown to 2021 in light of COVID-19.
- International Monetary Fund (IMF). 2016a. Virtual Currencies and Beyond: Initial Considerations. IMF Staff Discussion Note, SDN/16/03.
- IMF. 2017a. Fintech and Financial Services: Initial Considerations. IMF Staff Discussion Note.
- IMF. 2019a. Fintech: The Experience So Far. IMF Policy Paper No. 19/024. (Joint with the World Bank)
- IMF. 2019b. Emerging supervisory practices to strengthen cyber resilience in the financial sector. MCM Departmental Paper.
- Healey, J., Mosser, P., Rosen, K. and A. Tache, 2018. The Future of Financial Stability and Cyber Risk. The Brookings Institution Cybersecurity Project, October.
- Bouveret, A. 2019. Estimation of losses due to cyber risk for financial institutions. Journal of Operational Risk, 14(2) pp. 1-20.
- Goh, J., Kang, H., Koh, Z., Lim, J., Ng, C., Sher, G., and C. Yao. 2020. Cyber risk surveillance: A case study of Singapore. MAS Staff Paper No. 57.
- Kamiya, S., J. Kang, J. Kim, A. Milidonis, R., Stulz. 2018. What is the Impact of Successful Cyberattacks on Target Firms? NBER Working Paper No. 24409.

### Sovereign risk, corporate credit, and default analysis
- Ams, J., R. Baqir, A. Gelpern, and C. Trebesch. 2018. Sovereign Default. In Sovereign Debt: A Guide for Economists and Practitioners. edited by S. Ali Abbas, Alex Pienkowski, and Kenneth Rogoff. Chapter 7. Oxford: Oxford University Press.
- Jobst, A. and H. Oura. 2019. Sovereign Risk in Macroprudential Solvency Stress Testing. IMF Working Paper, No. 19/226.
- Jacome, L., T. Sedik, and S. Townsend, 2011. Can Emerging Market Central Banks Bail Out Banks? Cautionary Tale from Latin America, IMFWP 11/258.
- Credit Research Initiative (CRI). 2019a. White Paper: Probability of Default. National University of Singapore.
- CRI. 2019b. White Papter: Bottom-up Default Analysis (BuDA v3.1.1). National University of Singapore.
- Giesecke, K., F. Longstaff. S. Schaefer, and I. Strebulaev. 2014. Macroeconomic effects of corporate default crisis: A long-term perspective. Journal of Financial Economics, vol. 111.
- Campbell, J., J. Hilscher, and J. Szilagy. 2008. In Search of Distress Risk. The Journal of Finance, Vol. 63. No. 6.

### IMF policy, guidance, and internal analysis
- International Monetary Fund (IMF). 2011. How to Address the Systemic Part of Liquidity Risk, Global Financial Stability Report, April 2011.
- IMF. 2012. Macrofinancial Stress Testing—Principles and Practices, IMF Policy Paper.
- IMF. 2013. Key Aspects of Macroprudential Policy, IMF Policy Paper.
- IMF. 2014a. Staff Guidance Note on Macroprudential Policy, IMF Policy Paper.
- IMF. 2014b. Shadow banking around the globe: How large, and how risky? Global Financial Stability Report, Chapter 2, pp. 65-104.
- IMF. 2015a. Balance Sheet Analysis in Fund Surveillance, IMF Policy Paper.
- IMF. 2015b. The Asset Management Industry and Financial Stability, Chapter 3 of the Global Financial Stability Report, April 2015.
- IMF. 2016b. Financial Stability Challenges in a Low-Growth, Low-Rate era. Chapter 1 of the Global Financial Stability Report, October 2016.
- IMF. 2017b. Financial Conditions and Growth at Risk. Chapter 3 of the Global Financial Stability Report, Fall 2017.
- IMF. 2018. Managing the Sovereign-Bank Nexus. Departmental Paper No. 18/16.
- IMF. 2019c. Sustainable Finance: Looking Farther. Chapter 6 of the Global Financial Stability Report, October 2019.
- IMF. 2020a. Physical Risk and Equity Prices. Chapter 5 of the Global Financial Stability Report, April 2020.
- IMF. 2020b. Norway : Financial Sector Assessment Program-Technical Note-Risk Analysis and Stress Testing. November 2020.
- Caprio, G. 2019. Assessing the FSAP: Quality, Relevance, and Value Added. Background paper for the Independent Evaluation Office’s Report on IMF Financial Surveillance.
- Independent Evaluation Office (IEO) of the IMF. 2019. Evaluation Report: IMF Financial Surveillance.

*Source: ppea2021041 - References (2021 FSAP REVIEW—BACKGROUND PAPER ON QUANTITATIVE ANALYSIS).*

---


_Source: https://www.imf.org/-/media/files/publications/pp/2021/english/ppea2021041.pdf_