## 1nldea2024010

## Source details

**Canonical URL:** [1nldea2024010](https://www.imf.org/-/media/files/publications/cr/2024/english/1nldea2024010.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2024/english/1nldea2024010.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2024/english/1nldea2024010.pdf.json)

---

### Executive summary — macrofinancial setting and FSAP scope
- Financial system size and composition:
  - System assets roughly eight times GDP.
  - Banks: about one third of system; 253 percent of GDP as of 2023Q2.
  - Occupational pension funds: 142 percent of GDP.
  - Insurance sector: about 43 percent of GDP.
- Recent macro developments:
  - GDP growth turned negative in mid-2023; external demand waned.
  - Headline inflation fell from double-digit levels; elevated core inflation persists.
  - ECB policy rates increased by 4.5 percentage points since July 2022; APP reinvestment ended.
  - DNB increased the Countercyclical Capital Buffer (CCyB) to 2 percent (effective May 2024).
  - House prices fell 6 percent through mid-2023 from July 2022 peak, with later recovery.
  - CET1 was 16.5 percent in 2023Q3.
- FSAP scope:
  - Quantitative analyses covering banks, insurers, and pension funds; household and corporate vulnerability analyses; CRE price-at-risk (CaR) analysis.
  - Four main risks considered: abrupt slowdown in growth; persistently high inflation; continued tightening of financial conditions; severe housing market correction.
  - Bank testing: solvency and liquidity stress tests, integrated solvency–liquidity tests, sensitivity analysis.
  - Insurers and pension funds: solvency and liquidity stress tests and sensitivity analysis.
  - Contagion analysis: fire-sale channel and cross-border interconnectedness.

### Banking sector — vulnerabilities, stress testing, and results
- Key vulnerabilities:
  - Mortgages constitute more than half of banks’ domestic loan book.
  - Residential real estate (RRE) exposures rose from ~27 percent of total loans in 2017 to over 40 percent.
  - Around 45 percent of fixed-rate mortgages are interest-only.
  - CRE loans: 7 percent of assets.
  - High concentration: four largest commercial banks account for 76 percent of banking system.
- Stress test design:
  - Six SIs, dynamic balance-sheet, three-year horizon 2023–2025, reference date June 2023 (COREP, FINREP, STE).
  - Adverse scenario features stagflation, de-anchored inflation expectations, higher interest rates, widened corporate spreads, large housing price correction.
  - Scenario numeric sequences (selected series from Figure 9):
    - House Price Growth (percent): 13.6 -6.0 -6.0 0.1 13.6 -16.0 -8.5 -4.5
    - Long-term Rate (percent): 1.2 2.5 2.8 2.4 5.7 4.7 3.0 0.0
    - Corporate Credit Spread (percent): 1.0 1.8 1.7 1.6 1.0 2.6 2.3 1.9
    - Inflation (percent): 11.6 3.9 4.2 2.1 8.2 6.7 1.7 0.0
    - Short-term Rate (percent): 0.0 3.2 3.6 2.9 0.0 6.2 5.5 2.8
    - Nominal Credit Growth (percent): 4.9 5.4 3.6 4.9 5.6 5.0 2.3 2
    - Real GDP Growth (percent): 4.5 1.0 1.2 1.5 -4.0 -3.0 -2.0 -1.0
    - Unemployment Rate (percent): 3.5 3.9 4.2 4.5 4.1 5.7 7.2 0.0
- Solvency results (SIs):
  - Baseline: SIs’ CET1 ratio drops by 1.0, 2.0, and 2.4 percentage points in 2023, 2024 and 2025 respectively.
  - Adverse: SIs’ CET1 ratio falls to 12.5 in 2024 from 15.7 in 2022.
  - Some SIs would need to draw down buffers: about 4.3 billion in 2024 and 9.7 billion in 2025.
  - One bank fails to meet minimum CET1 in the adverse scenario.
- Earnings and risk drivers:
  - Rising interest rates support net interest income (NII) but credit impairment and higher RWAs reduce capital.
  - Losses from market-risk factors generally low for the group; spread risk impact dominated by one bank (data quality caveat).
- Liquidity stress-testing:
  - LCRs and NSFRs well above 100 for all banks in aggregate.
  - Monte Carlo exercise: simulated LCRs centered around regulatory LCRs; simulated NSFRs centered left of regulatory NSFR but still above cutoff.
  - Retail deposit sensitivity: all banks have LCR above 100 if retail deposit runoff rates < 0.2; all but two have LCR below 100 if runoff rates > 0.28 (30-day horizon).
  - Cash-flow severe scenario: one bank illiquid at three months; two banks illiquid at nine months.
- HTM sales and USD funding vulnerabilities:
  - Median share of HTM assets in counterbalancing capacity ~50 percent; banks may need to sell HTM assets as short as a week into stress.
  - Some banks have relatively large short-term unsecured USD liabilities and low USD counterbalancing capacities; largest funding gap across banks ~10 percent of total counterbalancing capacity (using June 2023 average USD-Euro rate).
- Integrated solvency–liquidity sensitivities:
  - Sensitivity 1 (sustained sight deposit outflow): deposit shift leads to higher interest expense and capital deterioration.
  - Sensitivity 2 (forced HTM sales + term-deposit cost channel): additional banks draw down capital.
- Contagion and cross-border spillovers:
  - Fire-sale systemic stress test (88 agents; MSR holdings of 3,590 securities):
    - Bank solvency losses lead to fire-sale induced losses ~5 percent of initial equity; can reach 10 percent.
    - >60 percent of losses occur in second- and higher-order rounds.
    - Securities-issuing agents causing largest losses are not necessarily SIs.
  - Network simulations using BIS consolidated data:
    - Dutch foreign claims USD 1,531 billion as of end-June 2023.
    - Netherlands susceptible to shocks from U.S. and Germany banking systems.
    - At LGD = 50 percent, Dutch banking system faces immediate failure following collapse of U.S. banking system.
    - At LGD ≥ 75 percent, immediate failure follows collapse of German and/or U.S. banking systems.

### Insurance sector — solvency, liquidity, and key stress-test results
- Sample and scope:
  - 16 large insurers: 5 life insurers (≈93 percent of life assets), 5 P&C, 6 health (≈70 percent coverage in P&C and health).
  - Aggregated assets EUR 390bn; EUR 340bn attributed to life insurers.
  - Before-stress solvency ratios: all 16 participants above regulatory threshold of 100 percent.
- Adverse scenario and sensitivities:
  - Instantaneous market shocks at reference date 30 June 2023; key market shocks include risk-free rate, equity, property, corporate and sovereign spreads, mortgage spreads; EUR depreciation -10.7 percent external value.
  - Single-factor sensitivities: EUR term structure parallel downward shift -100 bps; EUR depreciation/appreciation ±10 percent; default of largest banking counterparty.
- Solvency stress-test findings:
  - Median asset value declines: Life 19 percent; P&C 11 percent; Health 3 percent.
  - Median asset-liability ratio declines: Life -5.0 percentage points to 102.7 percent; P&C -0.6 points to 126.3 percent; Health -2.8 points to 151.2 percent.
  - Median life insurer: Eligible own funds (EOF) decline by 62 percent; SCR declines by 25 percent; SCR coverage after stress 100 percent (from 189 percent pre-stress).
  - Aggregate capital shortfall: two life insurers below 100 percent SCR totaling EUR 2.9bn (≈0.3 percent of GDP).
  - P&C median SCR after stress: 128 percent (from 158 percent). Health median SCR after stress: 126 percent (from 134 percent).
- Liquidity analysis (variation margin on interest rate swaps):
  - Interest rate swaps notional ≈ EUR 360bn (almost exclusively life insurers).
  - BU reported total margin call: EUR 10.1bn; TD cross-check EUR 10.3bn.
  - Share of margin calls needing cash: 94 percent.
  - Cash due Day 1: EUR 8.4bn; Day 2: EUR 1.1bn.
  - Insurers’ planned Day 1 liquidity (five insurers): EUR 10.2bn composed of: Cash/deposits 22 percent; Repo facilities 53 percent (mainly uncommitted); Money-market funds 16 percent; Short HQLA 7 percent.
- Policy implication for insurers:
  - Close monitoring of repo market conditions due to insurers’ reliance on repo facilities and simultaneous liquidity needs across market participants.

### Pension funds — solvency, liquidity, and stress-test findings
- Solvency results (top-down, ten largest funds ~70 percent assets):
  - Adverse scenario: higher interest rates lower liabilities by 27 percent on average; asset values decline by 23 percent on average (EUR -235bn for the sample); liability decline totals EUR 245bn for the sample.
  - Own funds increase on average by 7 percent (from EUR 136bn to EUR 145bn).
  - Average funding ratio increases by 7 percentage points to 122 percent (from 115 percent pre-stress).
- Liquidity (BU survey of five largest funds ~60 percent assets):
  - Total variation margin call: EUR 25.0bn; margin call related to interest rate swaps: EUR 14.5bn.
  - Share of margin calls met in kind: 36 percent.
  - Cash collateral call: EUR 18.4bn; 60 percent due within one day (Day 1).
  - Day 1 liquidity composition: Cash/deposits 18 percent; Reverse repos/uncommitted repo lines 34 percent; Liquidation of MMFs & short HQLA 12 percent.
  - Day 2 composition: sale of MMFs and repo transactions each ~39 percent.
- Policy implication for pension funds:
  - Close monitoring of market conditions and repo market reliance; many smaller funds use bilateral swaps allowing settlement in kind, reducing cash liquidity needs.

### Households, nonfinancial corporates (NFCs), and CRE vulnerabilities
- Households:
  - Household debt-to-income decreased by almost 80 percentage points from 2011Q1 peak to 2023Q2 but remains among highest in advanced economies.
  - Banks provide 69 percent of mortgages (>€550 billion outstanding) as of 2023Q2.
  - Pension entitlements = 25 percent of household assets; real estate = 50 percent of household assets (end-2022).
  - IO mortgages: <44 percent of total stock in 2022Q1 (versus >57 percent in 2013Q1); 2.78 million households have partial or full IO mortgages; 29 percent of IO stock matures 2034–2039.
  - If house prices drop 20 percent: 13 percent of homeowners LTV > 90 percent; 8 percent negative equity.
  - Mortgage default rate: 0.7 percent as of 2022Q1 (2013–2021 avg 0.5 percent). NHG applies to 25 percent of mortgage market.
  - Interest-rate fixation: 75 percent of outstanding mortgages fixed >5 years; <3 percent flexible in 2023; 25 percent flexible within next 5 years.
- Household stress-test outcomes:
  - Aggregate: proportion of borrowers with high LTV almost triples in adverse vs baseline for certain joint metrics.
  - Aggregate changes (adverse vs baseline):
    - Proportion with high LTV +11 percentage points higher.
    - Proportion with high DTI +4 percentage points higher.
    - Proportion with high DSTI +3 percentage points higher.
  - Joint LTV >80 and DTI >4: increases from 4.3 percent (baseline) to 11 percent (adverse).
  - NIBUD DSTI-limit based high-risk borrowers: increase from 6.6 percent to 8.2 percent over three years in adverse; lowest-income borrowers from 19.3 percent to 22.5 percent; borrowers <35 years from 7.1 percent to 11.3 percent.
  - Recommendations: calibrate model by borrower groups; estimate PD elasticities to DSTI and LTV for sub-groups; consider alternative DSTI thresholds.
- Nonfinancial corporates (NFCs):
  - NFC debt-to-GDP fell relative to GDP but remains elevated vs pre-pandemic; dependence on short-term securities and loans increased since 2021.
  - Corporate stress-test (dynamic, firm-level):
    - Share of firms with ICR < 1: 8.2 percent in 2022; 31.5 percent in 2023 under adverse scenario.
    - Proportion of firms with negative cash balances doubles from 2022 to 2023 under adverse, peaks 20.6 percent in 2024.
    - Drivers: weakened earnings (lower sales), higher short- and long-term interest rates.
    - Regression selected coefficients (examples):
      - ROA(-1): 0.581*** (0.020)
      - Real GDP Growth (-1): 0.187*** (0.061)
      - Observations: 18,056 (ROA); R-squared ROA 0.387.
    - Policy: monitor firms with shorter maturity profiles; improve access to bank loan-level data.
- Commercial Real Estate (CRE):
  - Total CRE exposures (banks, pension funds, insurers): ≈€360 billion.
  - Bank CRE loans: 7 percent of balance sheet; 10.7 percent of total loans.
  - CRE NPL ratio: 2.8 percent in 2022Q4.
  - Average CRE loan LTV: below 60 percent in 2023Q2; average recovery rate for foreclosed CRE: 70 percent.
  - CRE prices doubled 2015Q1–2022Q2, then declined sharply.
  - CaR methodology and findings:
    - CaR: 5th percentile of conditional distribution of future CRE price growth via panel quantile local projections.
    - A 1-standard-deviation increase in CRE price misalignment (24 bps) → cumulative 3.4 percentage point increase in downside risk over 4 quarters.
    - Netherlands CaR profile: hovered around -2 percent since 2012; moved from -5.4 percent in 2008Q4 to -1.6 percent in 2019Q4; 2022Q4 CaR -5.88; 2023Q1* CaR -3.21.
  - Recommendation: close gap between CBS CRE index and private market data; access granular CRE loan-level data for stress testing.

### Systemic interaction, contagion, and cross-sector amplification
- Interactions:
  - Solvency and liquidity stresses and cross-sector contagion can amplify initial losses beyond first-round impacts.
  - Fire-sale channel analysis shows significant amplification via asset price declines across rounds.
- Cross-border channels:
  - Large foreign claims and exposures emphasize susceptibility to external banking-system shocks, notably U.S. and Germany.
  - Network simulations show immediate Dutch banking-system failure under certain LGD and origin-country failure combinations (e.g., U.S. collapse at LGD 50 percent).

### Key policy recommendations and priorities (selected)
- Strengthen data and stress-testing frameworks:
  - DNB — ST, M: Tap alternative data sources (e.g., SHSG and EMIR trade repository) to complement market-risk analyses based on bank-reported sensitivities.
  - DNB — ST, M: Adopt IRRBB template when integrated into COREP in 2024 to improve interest-rate risk analysis.
  - DNB — ST, M: Continue collecting additional risk parameters and conduct granular credit-risk analyses for LSIs.
  - DNB — MT, M: Develop system-wide stress-testing methodologies to assess contagion effects across banks and NBFIs.
  - DNB — C, H: Monitor pension funds’ repo transactions, amend supervisory reporting as needed, and perform liquidity stress tests incorporating repo market drying-up scenarios.
  - DNB — C, M: Improve supervisory reporting quality for insurers and pension funds; prioritize data relevant for systemic risk analysis.
  - DNB — ST, M: Refine household stress-test model by calibrating equations by borrower group and using alternative DSTI thresholds.
  - DNB — ST, M: Develop risk assessment models for corporate and CRE sectors using granular bank-firm level loan data and better CRE measurement.
- Market infrastructure and contingency planning:
  - Monitor repo market depth and concurrent liquidity needs of insurers and pension funds.
  - Assess USD funding vulnerabilities and operational readiness for currency conversions and access to liquidity facilities.
- Data and reporting enhancements:
  - Improve timeliness and granularity of interest-rate-derivative reporting (e.g., EMIR) to inform daily margin-call analyses.

_Italic: Source: IMF staff summary of FSAP material contained in 1nldea2024010._

### EXECUTIVE SUMMARY __________________________________________________________________________ 8

### EXECUTIVE SUMMARY

### Introduction and macrofinancial setting
- The Netherlands has a large financial system: the system’s assets are roughly eight times the Gross Domestic Product (GDP) of the Netherlands.
- Banks account for about one third of the financial system and 253 percent of GDP as of 2023Q2.
- Occupational pension funds stand at 142 percent of GDP.
- The insurance sector, in particular life insurance, stands at about 43 percent of GDP and has been undergoing consolidation.
- Other financial institutions have grown significantly to surpass banks in size, reflecting responses to Brexit and financial innovation.
- The Dutch financial system is deeply interconnected domestically and with the rest of the world.

### FSAP scope and scenarios
- The FSAP conducted quantitative analysis of selected segments of the Dutch financial system covering banks, insurers, and pension funds, complemented by household and corporate vulnerability analysis and a commercial real estate (CRE) price-at-risk analysis.
- Four main risks were considered in the macrofinancial scenario analysis:
  - an abrupt slowdown in growth;
  - persistently high inflation;
  - continued tightening of financial conditions;
  - a severe correction in the housing market.
- Bank testing included solvency and liquidity stress tests, integrated solvency-liquidity tests, and sensitivity analysis.
- Insurers and pension funds underwent solvency and liquidity stress tests and sensitivity analysis.
- Contagion analysis considered a fire sale channel across institutions and sectors, and an interconnectedness analysis of cross-border banking sector exposures was conducted.

### Banking sector: findings and vulnerabilities
- Aggregate solvency: A solvency stress test on Significant Institutions (SIs) indicates the sector would remain sufficiently capitalized even if all main risk factors materialize simultaneously.
- Institution-level outcomes: Some SIs might see capital buffers erode since earnings weaken over several years in the adverse scenario.
- Contagion: An institution-level cross-sector contagion analysis shows bank capital losses could lead to substantial additional losses in the banking, insurance, and pension fund sectors.
- Liquidity: Liquidity stress tests on SIs show the sector would be able to withstand severe funding pressures.
- Cross-border exposure: Dutch banks could be susceptible to shocks from U.S. and German banking systems.

### Insurance sector: findings and vulnerabilities
- Broad resilience: The insurance solvency stress test shows broad resilience of the Dutch insurance sector, particularly for property & casualty (P&C) and health insurers.
- Life insurers: Vulnerabilities exist for some life insurers. Large Dutch life insurers strive for a close match of assets and liabilities through the cycle, and in times of rising interest rates they benefit less than many of their European peers.
- Shock mechanics: Aggregated interest rate effects on assets and liabilities are almost balanced, so other asset-side shocks cause a significant decline in own funds for life insurers.
- Liquidity resilience: Life insurers are broadly resilient to liquidity shocks despite large interest rate swap positions. Assuming a euro interest rate increase of 100 basis points, margin calls are sizable, but sampled entities apply heterogeneous strategies and draw on a variety of liquidity sources, including cash and deposits, uncommitted repo facilities, and the sale of money-market funds.

### Pension funds: findings
- Solvency: Pension funds’ solvency positions are strong; funding ratios have improved considerably since 2021 due to rising interest rates which lower the value of long-dated pension liabilities and compensate for asset declines.
- Liquidity: The sector appears resilient to liquidity risks, including from margin calls.
- Bottom-up analysis: De Nederlandsche Bank (DNB) BU analyses show resilience of pension funds to liquidity risks from margin calls, even when restricting access to the repo market. Margin calls are more sizable than for insurers, but a larger share of the call could be met in kind. Many smaller pension funds use bilateral swap transactions allowing settlement in kind, lowering liquidity risks.

### Household, corporate, and CRE vulnerability analysis
- Households:
  - Most impacted groups: Young, low-income borrowers, and borrowers with high loan-to-value (LTV) and debt-to-income (DTI) ratios.
  - Stress outcome: The proportion of borrowers with high LTV and DTI almost triples in the adverse scenario compared to the baseline.
  - Methodology: Loan-level data analysis assessed the impact of income, interest rate, and housing price shocks on borrowers’ debt-service-to-income (DSTI) and LTV ratios.
- Nonfinancial corporations (NFCs):
  - Vulnerabilities: Susceptible to lower growth and tighter financial conditions, particularly firms with shorter debt maturity profiles.
- Commercial real estate (CRE):
  - Price-at-risk analysis suggests downside risks in the CRE market remain elevated.

### Systemic interaction and contagion
- Integrated effects: Interactions between solvency and liquidity stresses and cross-sector contagion can amplify losses beyond initial shock impacts.
- Fire-sale channel: Contagion analysis explicitly considers a fire sale channel across institutions and sectors.
- Cross-border interconnectedness: Analysis highlights exposure channels to foreign banking systems, notably the U.S. and Germany, that could transmit stress to Dutch banks.

### Policy focus and recommendations
- The FSAP recommendations aim to address observed gaps and further strengthen the Netherlands’ systemic risk analysis framework.
- Key policy priorities include strengthening monitoring and mitigation of vulnerabilities identified in banks, life insurers, pension funds, households, NFCs, and the CRE market, and enhancing cross-sector and cross-border contagion preparedness.

*Prepared by Wei Sun, Max Sebastian Dovì, Romain Bouis, Junghwan Mok (all IMF MCM), and Timo Broszeit (external expert).*

### INTRODUCTION

### INTRODUCTION

### Macrofinancial Setting

- GDP growth held up to the effects of the war in Ukraine following the post-pandemic recovery, but turned negative in mid-2023 as external demand wanes and consumption growth slows.
- Inflation developments:
  - Headline inflation has fallen from double digit levels as energy price shocks subsided, with headline inflation approaching target.
  - Elevated core inflation has persisted amid a still-tight labor market.
- Financial conditions and lending:
  - The ECB increased its policy rates by 4.5 percentage points since July 2022.
  - The ECB has ended reinvestment in its Asset Purchase Programme (APP).
  - Banks are repaying amounts borrowed under TLTROs.
  - Financial conditions have tightened, though at a decelerating pace and even easing somewhat recently.
  - Bank lending to households and nonfinancial corporations has slowed in 2023.
  - The financial cycle has turned.
  - DNB increased the Countercyclical Capital Buffer (CCyB) to 2 percent (effective May 2024), a level it considers neutral.
- Housing market:
  - House prices had fallen 6 percent through mid-2023 from their July 2022 peak, before recovering lately.
  - Mortgage rates have risen, and home sales have been stable recently.
  - DNB imposed a macroprudential floor on risk weights on Dutch mortgages in 2022.
- Banking system:
  - Bank profitability improved as increases in lending rates outpaced those of deposit rates.
  - Core Equity Tier 1 (CET1) was 16.5 percent in 2023Q3.
  - Systemwide nonperforming loans (NPLs) stayed low, and the Liquidity Coverage Ratio (LCR) was adequate during this period.
- Pensions and insurers:
  - Higher interest rates have improved occupational pension funds’ funding ratios since 2021 as liability values declined.
  - The new collective defined contribution (DC) regime will shift investment risks to pension fund members and beneficiaries.
  - Solvency ratios of life insurers were stable over the last two years, though below the European Union (EU) average.
- Households:
  - Household debt as a share of disposable income peaked in the early 2010s and has declined since then.
  - Deleveraging accelerated since 2022 on the back of strong nominal disposable income growth.
  - Banks’ real estate exposures, housing valuations, and household debt are high relative to peers.
  - Household assets are high but mainly illiquid and tied up in pension savings.
  - The boom in housing prices has increased vulnerabilities for the most recent borrowers.
- Corporate sector:
  - Nonfinancial corporate debt increased with the recovery from the pandemic, while the debt-to-GDP ratio declined.
  - Bankruptcies have increased but remain below pre-pandemic levels.
  - Within the Euro Area, the Netherlands ranks the fourth highest in debt-to-surplus ratio.

- Key recommendations (Table 1 — summarized by addressee, timing, priority):
  - DNB — ST, M: Tap alternative data sources, e.g., Securities Holdings Statistics Group (SHSG) and trade repository of derivatives required by EMIR, to complement market risk analyses based on bank-reported sensitivities.
  - DNB — ST, M: Adopt the Interest Rate Risk for the Banking Book (IRRBB) template to conduct the interest rate analysis as it becomes available in Common Reporting (COREP) IN 2024.
  - DNB — ST, M: Continue efforts in collecting additional risk parameters and conducting granular credit-risk analyses for LSIs.
  - DNB — MT, M: Develop system-wide stress-testing methodologies to assess contagion effects of price and funding shocks across banks and NBFIs.
  - DNB — C, H: Closely monitor pension funds’ repo transactions, amend supervisory reporting where necessary, and perform liquidity stress tests which incorporate a drying-up of repo markets.
  - DNB — C, M: Continue efforts to improve the quality of supervisory reporting of insurers and pension funds, and prioritize data items relevant for systemic risk analysis.
  - DNB — ST, M: Refine the household stress test model by calibrating some equations by group of borrowers and by using alternative DSTI thresholds to identify borrowers at risk, complemented by sensitivity analyses.
  - DNB — ST, M: Develop risk assessment models for both the corporate sector and CRE sector using more granular bank-firm level loan data and more accurately measured data on the CRE.
  - Timing legend: C = Continuous; I = Immediate (within one year); ST = Short Term (within 1-3 years); MT = Medium Term (within 3-5 years).
  - Priority legend: H = High; M = Medium; L = Low.

### Financial System Landscape

- Size and composition:
  - Total system assets are roughly eight times GDP, with banks accounting for one third of the system as of 2023Q2.
  - The banking sector is highly concentrated: the four largest commercial banks (ING, Rabo, AMRO, and de Volksbank) account for 76 percent of the banking system.
  - Two policy banks (BNG and Nederlandse Waterschapsbank) are state-owned and AAA-rated.
- Systemically important banks (SIs):
  - Dutch SIs have a loan-dominant business model funded mostly by deposits.
  - The two policy banks mainly lend to corporate borrowers.
  - De Volksbank specializes in domestic mortgages.
  - Other large banks are active in international markets, including Australia, Belgium, Germany, UK, and U.S.
  - SIs have been reducing holdings of non-interest-earning assets and excess reserves at the central bank as financial conditions tightened.
  - Some outflows of sight deposits occurred, partially offset by inflows into term deposits.
  - SIs are repaying TLTROs, reducing outstanding balance to the ECB.
- Less significant institutions (LSIs):
  - 23 LSIs hold about 8 percent of total banking assets.
  - Business models among LSIs are diverse: financial conglomerates (FICO), universal banks, corporate and emerging market subsidiaries, custodian and specialized banks.
- Nonbanks and markets:
  - Occupational pension funds are among the largest globally, at 142 percent of GDP.
  - Leverage in pension funds is considerably lower than UK peers.
  - A pension reform adopted in 2023 moves the system from defined benefit (DB) toward defined contribution (DC), due for completion by 2028.
  - Insurance sector remains sizeable at 43 percent of GDP; life insurers undergoing consolidation.
  - Securities trading shifted materially from London to Amsterdam since Brexit:
    - Dutch platforms’ share of European trading (including UK) increased to over 30 percent from 5-10 percent pre-Brexit.
    - Daily turnover volumes of EU-listed shares exceed those in London.
  - Amsterdam hosts significant fixed-income trading venues, including repo trading venues, and is a significant hub for trading Title Transfer Facility (TTF) gas futures.

### Vulnerabilities and Risks

- Structural and concentration risks:
  - High concentration in mortgage lending exposes banks to larger credit losses if activity slows, financial conditions tighten, and housing prices decline.
- Market and valuation risks:
  - Banks holding significant marketable securities are susceptible to valuation losses when market conditions shift quickly if not adequately hedged.
- Interconnectedness:
  - The interconnected nature of the Dutch financial system can allow idiosyncratic shocks to propagate and cause collateral damage.
- Climate risks:
  - The Dutch financial system is prone to material climate-related risks (discussed in a separate Technical Note).

*Source: INTRODUCTION (1nldea2024010) — IMF staff.*

### 13.      Banks and NBFIs are significantly exposed to real estate. Mortgages constitute more than

### 1nldea2024010 - 13.      Banks and NBFIs are significantly exposed to real estate. Mortgages constitute more than

### Exposure to real estate and financial sector composition
- Mortgages constitute more than half of banks’ domestic loan book.
- RRE exposures have risen from around 27 percent of total loans in 2017 to over 40 percent.
- Most mortgages are fixed-rate, with around 45 percent of these being interest-only.
- Insurers are active in mortgage lending, accounting for 15 percent of their assets.
- Banks’ commercial real estate (CRE) loans make up 7 percent of assets.
- Investments in CRE account for 7 percent and 8 percent of the balance sheets of pension funds and insurers, respectively.
- Residential mortgage holdings by sector are depicted in the source chart (Banks, Insurers, Pension funds, Investment funds, Other financial institutions).

### LSI (Less Systemic Institutions) vulnerabilities and heterogeneity
- LSI performance varies and could be vulnerable to more diverse macrofinancial shocks.
- LSIs have more diverse business models relative to SIs.
- Capital, asset quality, liquidity, and profitability metrics for LSIs exhibit more variability (Figure 8, Table 5).
- Vulnerability channels for LSIs include:
  - Large lending portfolios concentrated in certain foreign countries → susceptibility to credit risk driven by domestic conditions in those countries.
  - Large securities portfolios → adverse effects from higher interest rates and lower asset valuations.
  - Reliance on foreign or wholesale funding → liquidity strains during funding market dislocations.
- Contagion analysis found that failure of LSIs could cause significant damage to the wider financial system.

### CRE and housing market developments and risks
- CRE prices in the Netherlands doubled in value between 2015Q1 and 2022Q2, with significant declines thereafter.
- CRE price growth rebounded quickly after the pandemic but was abruptly reversed in the second half of 2022 under tightened financial conditions.
- Sudden drops in CRE prices pose credit risks to banks and other financial institutions because many companies use CRE as collateral for borrowing.
- The boom in housing prices increased vulnerabilities for the most recent borrowers:
  - Higher housing prices decreased average LTV ratios but increased the proportion of households at borrowing limits.
  - Increased debt-to-income ratios of new mortgagors, especially younger borrowers.
  - Around 60 percent of households under 36, and 45 percent of older households have a debt-to -income ratio above 450 percent.
  - The stock of IO mortgages has declined in the past decade but remains high; applications for IO mortgages recently resurged.
  - A large volume of IO mortgages will mature between 2034 and 2039, though the authorities’ analysis indicates that this will not pose any systemic risk.
- Authorities’ response:
  - AFM multi-year campaign launched in 2018 to monitor vulnerabilities from IO mortgages.
  - Banks have been taking actions to address financial concerns associated with maturing IO mortgages of some customers.

### Main macrofinancial risks (Risk Assessment)
- Key risks stem from an abrupt slowdown in growth combined with persistently high inflation, continued tightening of financial conditions, and a severe correction in the housing market.
- Such a scenario could be accompanied by lower external demand and financial spillovers affecting liquidity and funding conditions of certain financial institutions.
- FSAP analysis focused on:
  - Housing: Banks vulnerable to higher interest rates and a severe house price correction; house prices show signs of overvaluation.
  - NBFIs: Occupational pension funds and life insurers face market and liquidity risk; P&C insurers face higher inflation risk.
  - Higher interest rates have exposed vulnerabilities associated with margin calls for interest-rate derivatives.
  - Pension reform may lead pension funds to shift investment strategies gradually during the transition.
  - Higher inflation risks for insurers—claims inflation related to higher building costs, wages and medical costs strain insurers’ profitability.

### Bank stress testing framework, scenarios, and key scenario values
- Stress tests covered six SIs with a baseline and an “extreme but plausible” adverse scenario spanning a three-year horizon from 2023 to 2025, using supervisory statistics as of June 2023 (COREP, FINREP, STE).
- Stress test features:
  - Dynamic balance sheet approach; banks assumed to keep business models unchanged.
  - Credit, market, and interest rate risks assessed via econometric and accounting models.
  - Credit RWAs projected separately for standardized and IRB portfolios; other RWAs follow a predetermined path.
  - No extraordinary management action, policy support, or change of business models assumed.
  - Foreign exposures explicitly considered where possible, including country-level shocks.
- Scenario generation:
  - Baseline scenario aligned with April 2023 WEO projections; captures dynamics of housing price, wage growth, credit growth, and corporate credit spread.
  - Adverse scenario generated by the IMF’s Global Macrofinancial Model (GMM, Vitek 2018) and captures simultaneous materialization of macrofinancial risks, featuring stagflation, de-anchored inflation expectations, higher interest rates, faster nominal bank credit growth in 2023, widened corporate credit spreads, and a large housing price correction.
- Figure 9 (Macrofinancial Scenarios) presents scenario numeric sequences as shown in the source figure:
  - House Price Growth (percent): 13.6 -6.0 -6.0 0.1 13.6 -16.0 -8.5 -4.5
  - Long-term Rate (percent): 1.2 2.5 2.8 2.4 5.7 4.7 3.0 0.0
  - Corporate Credit Spread (percent): 1.0 1.8 1.7 1.6 1.0 2.6 2.3 1.9
  - Inflation (percent): 11.6 3.9 4.2 2.1 8.2 6.7 1.7 0.0
  - Short-term Rate (percent): 0.0 3.2 3.6 2.9 0.0 6.2 5.5 2.8
  - Nominal Credit Growth (percent): 4.9 5.4 3.6 4.9 5.6 5.0 2.3 2
  - Real GDP Growth (percent): 4.5 1.0 1.2 1.5 -4.0 -3.0 -2.0 -1.0
  - Unemployment Rate (percent): 3.5 3.9 4.2 4.5 4.1 5.7 7.2 0.0

### Credit, interest rate, and market risk methodology highlights
- Credit risk module:
  - Projects credit impairment using future trajectories of PD, LGD, and IFRS 9 provisioning rules.
  - Authorities provided historical PD and LGD data by portfolio and country of exposure.
  - Securities at amortized cost credit losses not considered; mark-to-market losses assessed in sensitivity analysis.
  - Used a suite of “satellite models” to project 216 future PD paths for six banks, six portfolios, and six economies.
  - Four portfolios modeled with panel regressions: mortgage, other retail, qualifying revolving, corporate.
  - Structural model applied to government and institutional portfolios.
- LGD estimation:
  - Two structural models: secured portfolio LGD derived from bank-specific LTV projections and cost factors; unsecured LGDs modeled as a function of future PDs.
- IFRS 9 provisioning:
  - Bank-specific transition matrices estimated by sector (household, corporate, government, institution).
  - Adjusted transition probabilities with scenario-conditional PDs using “beta-linking”.
  - Computed 12-month provision for stage 1 loans and lifelong provision for stage 2 and 3 loans.
  - Write-off rate assumed to be zero.
- Interest rate risk module:
  - Used IRRBB data from STE, categorizing repricing schedules into eight time-buckets and tracing movements across buckets.
  - Estimated pass-through from risk-free rates to sight deposit, term deposit, corporate, and mortgage loan rates via satellite models.
  - Assumed 100 percent pass-through for market-based instruments; TLTRO deposits treated as banks issuing debt securities to cover lost funding from repayment.
  - Nonperforming loans cease to generate interest income in future periods.
  - Included net interest income (NII) from the trading book, with trading book income/expense growing with the balance sheet.
- Market risk analysis:
  - Combined bank-specific market risk sensitivities (“delta” from STE) with scenario shocks to interest rate, spread, FX, equity, and commodity prices.
  - “Delta” represents gains/losses in euro caused by one unit of shock; presented separately for trading and banking books and decomposed across FVPL, FVOCI, and amortized cost accounts.
  - Caution: “deltas” are marginal, net sensitivities; valuation changes from multiple shocks are not necessarily additive and extreme “deltas” may reflect reporting quality issues.
- Other income statement and RWA assumptions:
  - Fees and commissions, other income, and non-interest expense grow with the balance sheet based on historical observations.
  - Banks pay the effective tax rate as of 2022 on future income.
  - Banks pay out 50 percent of their net income as dividends provided total comprehensive income stays positive post payment and banks continue to meet the overall capital requirement (OCR).
  - Credit RWAs estimated separately for STA and IRB portfolios and by country of exposure; special adjustments for mortgages per macroprudential minimum risk-weight rule:
    - Mortgage risk weight constructed from bank-specific LTV distribution: 12 percent weight for LTV under 55 percent; 45 percent weight for portion exceeding 55 percent.
    - Declining housing prices increase LTV and therefore credit RWA.
  - Market, operational, and other RWAs assumed to grow at the same rate as the balance sheet.

*INTERNATIONAL MONETARY FUND*

### 37.      The Sis as a group appear resilient to severe macrofinancial shocks, but some might

### 37.      The Sis as a group appear resilient to severe macrofinancial shocks, but some might 

### Solvency stress-test results (Sis)
- Consecutive years of weak earnings under the adverse scenario and the increase in countercyclical capital buffer (CcyB) to 2.0 percent from 2024 both contribute to capital pressure.  
- The stress test assumes the buffer requirement (including the CcyB of 2 percent in both baseline and adverse scenarios) and the mortgage RWA floor to stay the same over the horizon.  
- Baseline scenario:
  - Sis’ CET1 ratio drops by 1.0, 2.0, and 2.4 percentage points in 2023, 2024 and 2025, respectively.
  - Credit impairment is the main driver of the CET1 decline (Figure 10, Panels 1, 3, and 4).
  - Increase in RWA (from a dynamic balance sheet and mortgage risk weight adjustment) also contributes.
  - Some Sis need to restrict dividend payouts to stay above the OCR.
- Adverse scenario:
  - Sis’ CET1 ratio falls to 12.5 in 2024, the trough of the recession, from 15.7 in 2022 (Figure 10, Panels 3 and 4).
  - Increase in net interest income is offset by higher credit impairment, higher RWA, and lower other comprehensive income.
  - The group as a whole maintains a solid buffer position, but some Sis need to draw down buffers of about 4.3 billion in 2024 and 9.7 billion in 2025 to weather the severe shocks.
- Data and uncertainty:
  - Solvency analysis results on two banks could be prone to uncertainties due to data-reporting issues.
  - Reported prudential parameters for credit risk are sporadic; extreme values are reported for market risk sensitivities and large exposures.
  - One bank fails to meet the minimum CET1 capital requirement in the adverse scenario; additional information is needed to assess fiscal implications given these banks’ special nature.

*Sources: DNB and IMF staff estimates. Notes: OCI=other comprehensive income; TI=trading income; SREP=supervisory review; OCR=overall capital requirement. The Sensitivity 1 and 2 represent two exercises incorporating additional stress from solvency-liquidity interactions.*

### Income and market-risk drivers
- Interest-rate effects and net interest income (NII):
  - Rising interest rates support bank profitability through higher NII (Figure 11, Panels 1 and 2).
  - Empirical analysis shows incomplete pass-through of the reference rate to term deposit rates, reflecting the “deposit channel” of monetary policy.
  - Transmission is slower to sight deposits after years of near-zero rates, but faster to lending rates, widening loan-deposit rate differential and increasing NII (Annex A).
  - Cash positions at the ECB’s deposit facility, short-term corporate loans, floating rate mortgages by internationally active banks, and interest rate hedges reduce the maturity gap and allow faster benefit from higher lending rates.
  - Several Sis hold more securities for funding than investing: some securities-related income losses expected.
  - Several Sis face higher funding cost replacing expiring TLTROs, but outstanding TLTRO due for repayment is insignificant.
  - Net hedging costs remain relatively stable as paying and receiving legs of hedges are both affected by higher rates.
  - Overall, NII of the banking book appears higher in the adverse scenario; NII of the trading book does not change much given the balance sheet assumption (¶32).
- Credit impairment, RWAs, and market risk:
  - Higher credit impairment and RWAs reduce bank capital ratios (Figure 11, Panels 3 and 4).
  - PDs increase as the economy slows, debt servicing costs rise, housing prices decline, and wage growth fails to catch up with inflation.
  - LGDs for secured and unsecured portfolios generally follow housing prices and PDs patterns; aggregate LTV increases as housing prices decline, increasing credit RWA as more mortgages face higher risk weights.
  - Losses from market-risk factors are generally low (Figure 11, Panels 5 and 6); Sis as a group do not hold significant positions subject to interest rate, spread, FX, equity, and commodity price shocks and appear to hedge these risks well.
  - Materialization of spread risk impact is dominated by one bank, which could reflect incorrectly reported data.

### LSI credit-risk analysis (corporate and EM LSIs)
- Coverage and methodology:
  - Team analyzed credit risk of five corporate and four EM LSIs.
  - About 65 percent of their corporate exposures are outside of the EA and distributed across Africa, Asia, non-EA Europe, Japan, Latin America, Turkey, UK, and U.S.
  - Due to limited borrower information, the team used a PD database of 92,000 publicly listed companies to proxy borrower credit risk.
  - PDs grouped into nine economic groups: U.S., UK, EA, non-EA Europe, Turkey, Japan, Asia excluding Japan, Africa, and Latin America.
  - An econometric model (with Bayesian Model Averaging, BMA) established historical relationships between average PD of each group and selected macrofinancial variables, including interest rates of major advanced economies.
  - Simulated future PDs under baseline and adverse global scenarios were produced using BMA-estimated coefficients.
- Findings:
  - LSIs could experience rising default probabilities under severe global macrofinancial conditions.
  - Default probabilities of corporate borrowers in advanced markets are more vulnerable to interest-rate rises; those in EMs are more sensitive to economic slowdowns and FX fluctuations.
  - The stagflationary global scenarios (slow growth and high interest rates) produce a sharp increase in corporate credit risk, surpassing levels during the global financial crisis.
  - Corporate and EM banks could suffer significant credit losses if the slow-growth-high-interest-rate scenario materializes.
  - Once bank-specific PDs become available, this analysis can be re-anchored to infer borrower PD levels and expected credit losses more precisely (Annex A).

### SI liquidity stress-test methodology
- Three-pronged approach:
  - LCRs: evaluate whether stocks of HQLAs cover net outflows over a 30-day horizon.
  - NSFRs: evaluate whether available stable funding covers required stable funding over a one-year period.
  - Cash-flow analyses: use maturity ladders to analyze funding pressures across horizons from overnight to up to one year. June 2023 data used.
- Sensitivity and reverse-stress testing:
  - Monte Carlo exercise: LCR and NSFR parameter values drawn from a uniform distribution centered at regulatory values with ±25 percent bounds to test sensitivity to scenarios similar to, but distinct from, regulators’ scenarios.
  - Targeted LCR-based reverse stress test for large retail deposit outflows: grid of deposit runoff rates between zero and one; compute LCR for each bank across grid.
  - Reverse-stress testing for extreme scenarios: increase severity of all parameters by a scalar factor (capped at 0 and 1) to identify scenarios that cause failure to meet LCR and NSFR requirements.
- Cash-flow specifics:
  - Banks can use existing cash and cash value of securities (after valuation adjustments) to cover negative funding gaps; collateral cash values from repos included in counterbalancing capacity after netting rolled-off reverse repos.
  - Baseline calibrations similar to regulatory LCR; severe calibrations include moderate-to-severe deposit runs and more severe valuation effects on Level 1 and Level 2 assets.
  - Sensitivity assessed via reverse-stress testing increasing parameter severity by a scalar multiple of baseline.
  - Assessment of solvency risks from selling HTM assets: compute number of banks forced to sell HTM assets as a function of hypothesized shares of HTM assets in counterbalancing capacities; solvency risks then assessed using those shares.

### Liquidity stress-test results (Sis)
- LCRs and NSFRs:
  - Banks’ all-currencies LCRs and NSFRs are strong; all banks have regulatory LCRs and NSFRs well above 100 (Figure 12, Panels 1 and 2).
  - LCRs and NSFRs remain above 100 across a wide range of stress scenarios.
  - Monte Carlo exercise:
    - Aggregate distribution of simulated LCRs is roughly centered around regulatory LCRs; no evidence that regulatory LCRs overestimate resilience to similar liquidity stress (Figure 12, Panel 3).
    - Some evidence that regulatory NSFRs overestimate funding resilience (distribution centered to the left of regulatory NSFR), but all simulated NSFR values are still well above regulatory cutoff (Figure 12, Panel 4).
  - Bank-level Monte Carlo exercises produce similar results.
- Sensitivity to retail deposit outflows:
  - All banks have an LCR above 100 if retail deposit runoff rates are below 0.2.
  - All banks but two have an LCR below 100 if retail deposit runoff rates are above 0.28 (Figure 13) over a 30-day horizon.
- Cash-flow analysis:
  - Aggregate cash-flow analyses confirm banks are generally liquid in both baseline and severe scenarios (Figure 14, Panels 1 and 2).
  - Severe scenario outcomes:
    - One bank becomes illiquid at a horizon of three months or more.
    - Two banks become illiquid at a horizon of nine months or more (Figure 17).
  - Several banks remain liquid under substantially more severe scenarios than the baseline (e.g., large revaluations of Level 1 CQS1 assets).
  - Overall, results confirm strong liquidity positions for banks considered.

*Sources: COREP; June 2023 data; DNB and IMF staff estimates. Notes: Retail deposits include: ‘retail deposits exempted from the calculation of outflows’; ‘retail deposits where the payout has been agreed within the following 30 days’; ‘retail deposits subject to higher outflows, category 1’; ‘retail deposits subject to higher outflows, category 2’; ‘stable deposits’, ‘derogated stable deposits’; ‘other retail deposits’.*

*IMF staff analysis based on DNB and supervisory data.*

### 55.      Some banks would need to sell HTM assets in the scenarios considered. The FSAP team

### 1nldea2024010 - 55.      Some banks would need to sell HTM assets in the scenarios considered. The FSAP team

### Liquidity pressures and sale of HTM assets
- The FSAP team estimated the median share of HTM assets in the counterbalancing capacity to be around 50 percent.
- This suggests that banks would need to sell HTM assets, at horizons as short as a week.
- Figure 15 (cash-flow analysis) indicates banks are likely to have to sell HTM assets to meet funding shortfalls, given the ~50 percent share of HTM in counterbalancing capacity.

### USD funding vulnerabilities
- Several banks have relatively large short-term unsecured wholesale USD liabilities but relatively small USD counterbalancing capacities.
- Some banks have a USD LCR and NSFR below 100.
- Some low USD LCRs arise from hedging strategies combined with the LCR formula mechanics that cap inflows at 75 percent of outflows.
- Even in the cash-flow analysis (which does not impose caps on inflows), some banks would need to liquidate parts of their non-USD counterbalancing capacity and exchange the proceeds for USD to meet funding gaps in both the baseline and severe scenario.
- Drivers:
  - Large share of unsecured wholesale funding in banks’ USD-denominated liabilities.
  - Relatively low share of USD-denominated assets in their counterbalancing capacities.
- The largest funding gap across banks and all scenarios is approximately 10 percent of the total counterbalancing capacity (using the average USD-Euro exchange rate in June 2023).
- Implications:
  - Even during severe USD-funding distress and absent additional liquidity facilities, banks would only have to liquidate a relatively small part of their non-USD-denominated counterbalancing capacity.
  - USD funding shortfalls could materialize early, causing operational challenges when accessing liquidity facilities or liquidating and converting proceeds of non-USD denominated assets.

### Integrated solvency–liquidity analysis (sensitivity analyses)
- Sensitivity analysis 1 (deposit behavior shift based on sight deposit outflow observed over past year):
  - Assumption: If the rate of outflow persists throughout 2023, about 60 percent of sight deposits would remain in the sight account, 20 percent move to the term account, and another 20 percent flow out of the banking system.
  - Result: Loss in deposits requires SIs to issue securities at a higher cost to fund a growing balance sheet; interest expense sharply increases, more so for banks with a large sight deposit base.
  - Outcome: SIs’ capital positions further deteriorate from the adverse scenario (referred to as “Sensitivity 1”).
- Sensitivity analysis 2 (additional liquidity–solvency interactions):
  - Assumptions:
    - Banks sell HTM securities and realize associated losses whenever funding gaps in the cash-flow scenario cannot be closed using other counterbalancing capacity; HTM securities sold at prices specified in the cash-flow scenario.
    - Regression analysis indicates banks’ CET1 ratios explain some variation of term deposit rates; lower CET1 ratios increase the cost of attracting term deposits for banks with larger term deposit bases.
  - Result: These two additional cost channels lead more banks to draw down capital (referred to as “Sensitivity 2”).
- Note: Footnoted caveat — since SIs considered are likely able to use HTM securities to obtain liquidity from the ECB, forced sale of HTM is somewhat improbable unless a bank is not operationally ready to do so.

### Contagion: fire-sale systemic stress test (methodology and results)
- Data and scope:
  - December 2022 balance-sheet data for 88 agents: 14 banks, 27 insurers, 47 pension funds, combined with Monthly Securities Reporting (MSR) on holdings of 3,590 marketable securities.
- Methodology highlights:
  - Adaptation of Cetorelli et al. (2023) to allow institution-specific selling decisions and default conditions.
  - Agents experiencing balance-sheet losses seek to keep leverage ratio constant, translating losses into proportional sales of securities (sales proportional to market value).
  - Buyers assumed to be “rest of the world”; security-specific prices depend on quantity sold and a security-specific elasticity parameter.
  - Nonlinearities: prices of securities issued by agents in default set to zero.
  - Two sets of initial shocks considered:
    - Losses of the six SIs in the adverse solvency scenario (considered jointly).
    - Losses generated when each securities-issuing agent defaults in turn.
- Key results:
  - Losses from bank solvency stress-test exercise lead to fire-sale induced losses of about 5 percent of initial equity.
  - In early rounds, losses concentrated in banking sector; spread to insurance and pension sectors in later rounds.
  - Losses can be as high as 10 percent of initial equity.
  - More than 60 percent of losses occur in second- and higher-order rounds.
  - Individual-agent defaults can cause substantial losses amplified by fire-sale channels, but no agent’s default leads to defaults of other agents.
  - Banks suffer the highest total loss as a share of initial equity following defaults of securities-issuing agents.
  - The securities-issuing agents causing the five largest losses are not SIs, indicating contagion can be caused by default of relatively small agents.

### Interconnectedness and cross-border spillovers
- Data: BIS consolidated banking statistics as of end-June 2023.
- Dutch banks’ foreign claims amount to USD 1,531 billion as of end-June 2023.
- Major counterparties: non-financial private sector (including non-bank financial institutions), foreign official sector, and banks.
- Largest foreign exposures are to the non-financial private sector and to other advanced economies.
- Network simulation approach (Espinosa-Vega and Solé (2010) applied to BIS data):
  - Simulates propagation of a single country banking system failure through credit and funding channels across 12 countries (Belgium, Canada, France, Germany, Ireland, Italy, Japan, Netherlands, Spain, Switzerland, UK, U.S.).
  - Key assumptions:
    - A country’s banking system fails when losses from credit and funding shocks surpass aggregate bank capital.
    - In creditor country, baseline credit loss equals half of its exposure to a failed borrower.
    - Funding risk: when a creditor fails, only 50 percent of funding withdrawn is assumed to be replenished by other creditors; borrower incurs a loss equivalent to 25 percent of initial amount borrowed from failed creditor (per example formulation).
    - Path simulations assume countries with highest contagion index fail first; simulations run at different LGD levels.
- Findings on Dutch susceptibility:
  - Dutch banking system is susceptible to shocks originating from the U.S. and Germany, and to a lesser extent the UK.
  - Capital levels in the Dutch system would be depleted following collapse of the U.S. banking system, highlighting substantial U.S. exposures.
  - Path simulations show susceptibility of the Netherlands to failure of U.S. and Germany banking systems.
  - At an LGD assumption of 50 percent, the Dutch banking system faces an immediate failure following collapse of the U.S. banking system.
  - At an LGD assumption of 75 percent or higher, the Dutch banking system experiences immediate failure following collapse of the German and/or U.S. banking systems.
  - The Dutch banking system demonstrates vulnerability in the second round after failure of the UK banking system.

### Recommendations (to strengthen bank solvency stress testing)
- Continue to develop market risk analysis based on sensitivities, or “deltas”, reported to the Single Supervisory Mechanism (SSM).
  - Complement delta-based analysis with alternative, more granular data sources (examples: SHSG and trade repository of derivatives recorded by EMIR) to better pinpoint underlying vulnerabilities.
- Adopt the IRRBB template in interest rate risk analysis.
  - Current reliance on COREP 66 maturity ladder requires additional assumptions to capture repricing; the IRRBB template contains repricing information and is expected to be integrated into COREP in early 2024.
  - IRRBB template is more desirable for analyzing net interest income, the main source of Dutch banks’ profitability.
- Continue efforts to develop credit risk analysis for LSIs.
  - Authorities have progressed in requiring LSIs to report additional risk parameters, such as PDs by portfolio and country of exposure.
  - Given LSIs’ diverse exposures to advanced and EM economies, enhanced reporting will improve understanding of LSIs’ borrower risk profiles and support tools to analyze LSI resilience to abrupt macrofinancial changes.

*Source: IMF staff summary of FSAP material contained in the supplied content.*

### 70.      The authorities can strengthen their system-wide stress testing by developing

### 1nldea2024010 - 70.      The authorities can strengthen their system-wide stress testing by developing

### INSURANCE SECTOR RISK ANALYSIS — Overview
- FSAP team conducted:
  - a top-down solvency stress test,
  - a liquidity risk analysis,
  - an analysis of physical climate risks (covering up to 16 insurers).
- Solvency stress test used the macrofinancial adverse scenario from the banking stress test, with added granularity for market and interest rate shocks.
- Sensitivity analyses covered interest rate and currency shocks and default of the largest banking counterparty.
- Liquidity analysis focused on variation margin calls for interest rate swaps held by life insurers.
- Climate risk analysis included a bottom-up flood risk exercise and parametric increases in severity and frequency of weather-related loss events (results in a separate Technical Note).

### A. Scope and Sample of the Solvency Stress Test
- Sample: 16 large insurers on a solo-entity basis:
  - 5 life insurers — coverage of around 93 percent of assets in the life market.
  - 5 P&C insurers and 6 health insurers — market coverage around 70 percent in these sectors.
- Aggregated balance sheet assets of participants: EUR 390bn, of which EUR 340bn attributed to life insurers.
- Before-stress solvency ratios: all 16 participants record before-stress solvency ratios above the regulatory threshold of 100 percent.
- Capital composition and regulatory measures:
  - 91 percent of eligible own funds are unrestricted Tier 1 capital.
  - 3 percent of eligible own funds are Tier 3.
  - 27 insurers have permission to use the Volatility Adjustment (VA).
  - For VA users, SCR at end-2022 would have been around 70 percentage points lower without VA (but still above 100 percent).

### B. Scenarios for the Solvency Stress Test
- Adverse scenario features: significant supply shocks, synchronized growth slow-down, de-anchoring of inflation expectations, globally sharp rise in interest rates, domestic house price decline.
- For insurance test, all shocks were assumed instantaneous at the beginning of the first year (reference date 30 June 2023).
- Market shocks included shocks to:
  - risk-free interest rate,
  - equity prices,
  - property prices,
  - credit spreads of corporate and sovereign bonds,
  - spreads of mortgage loans.
- Interest rate shock applied to interest rate swap portfolio (notional value of more than EUR 300bn).
- VA increases with credit spread increases, partially offsetting credit spread shock for VA users.
- Comparable benchmark: EIOPA 2018 “Yield Curve Up” scenario (10-year EUR swap rates +86 bps, equity -39 percent EU average, spreads of 10-year Dutch government bonds +42 bps, A-rated EU financials +164 bps, Dutch commercial real estate -35 percent).
- Additional single-factor sensitivity tests:
  - Interest rates: parallel downward shift of EUR risk-free term structure by 100 basis points.
  - Currencies: Euro depreciation by 10 percent (increase external value by 10 percent).
  - Counterparty risk: default of largest banking counterparty with haircuts: equity exposures and subordinated bonds 100 percent haircut; secured bonds LGD 10 percent; other on-balance sheet exposures LGD 70 percent.

### C. Capital Standard and Modeling Assumptions
- Solvency II is the basis of the stress test; assets and liabilities valued mark-to-market but LTG and transitional measures can affect discount rate.
- Main output: effect on own funds eligible for the coverage of the SCR; SCR partially recalculated after stress.
- Data sources: Solvency II QRTs used include S.02.01, S.06.02, S.08.01, S.13.01, S.18.01, S.22.01, S.23.01, S.25.01/S.25.02/S.25.03, S.26.01, S.26.03.
- Stress application:
  - Haircuts applied to market value of directly-held assets per adverse scenario.
  - No look-through to individual securities held via investment funds; fund holdings stressed by underlying asset-class shocks.
  - Fixed-income assets re-valued with stressed term structure for each major currency.
  - Technical provisions (except unit-linked) approximated with stressed term structure including VA where applicable.
  - Unit-linked liabilities decline mirrored market value loss of underlying assets.
- SCR re-calculation limited to selected risk modules:
  - Market risk module: equity, spread, and property risk capital charges adjusted proportionately to change in exposures.
  - Equity risk capital charge corrected for symmetric equity adjustment changing from +1.2 to -10.0 percentage points after equity price fall.
  - Life underwriting risk capital charge assumed to change proportionately with technical provisions.
  - Counterparty default risk, non-life underwriting risk, operational risk assumed unchanged.
  - Internal model users: relative change in SCR approximated via simplified approach building on standard formula including aggregation and diversification effects.
  - Loss-absorbing capacity of deferred taxes re-calculated based on modeled valuation losses.

- Limitations and scope:
  - Stress tests are macrofinancial diagnostics, not pass-fail regulatory exercises.
  - Top-down test cannot fully capture financial hedges, stop-loss arrangements, financial reinsurance, or reactive management actions (e.g., underwriting changes, reinsurance, withholding profits, de-risking asset allocation, group capital support).
  - Static balance sheet assumption: management actions not modeled; results likely less severe with management actions included.

### D. Results of the Solvency Stress Test — Key Findings and Statistics
- Valuation impacts:
  - Asset values decline more than liability values across sub-sectors.
  - Median asset value declines:
    - Life: 19 percent
    - P&C: 11 percent
    - Health: 3 percent
  - Asset-liability ratio median declines:
    - Life: decline of 5.0 percentage points to 102.7 percent
    - P&C: decline of 0.6 percentage points to 126.3 percent
    - Health: decline of 2.8 percentage points to 151.2 percent
- Drivers:
  - Life insurers: most asset value decline stems from higher interest rates affecting fixed-income instruments and interest rate swaps; liabilities also decline due to higher discount rates, reflecting close asset-liability matching.
  - Euro depreciation partially offsets other adverse shocks because insurers hold more foreign-denominated assets than liabilities.
- Solvency levels:
  - Median life insurer:
    - Eligible own funds (EOF) decline by 62 percent.
    - SCR declines by 25 percent.
    - SCR coverage after stress: 100 percent (down from 189 percent before stress).
    - Heterogeneous results across sample; aggregate capital shortfall for two life insurers below 100 percent SCR equals EUR 2.9bn (about 0.3 percent of GDP).
  - Median P&C insurer:
    - SCR after stress: 128 percent (down from 158 percent).
  - Median health insurer:
    - SCR after stress: 126 percent (down from 134 percent).
  - P&C and health insurers display lower sensitivity to market and credit risks and are more resilient.

### E. Sensitivity Analyses — Findings
- Interest rates 100 bps lower:
  - Effect heterogenous in life sector; for median life insurer, SCR coverage changes marginally by +2 percentage points due to close duration matching.
- Euro appreciation by 10 percent:
  - Slightly negative impact, most pronounced for life sector; median SCR declines by 12 percent.
- Default of largest banking counterparty:
  - Would not have a significant direct impact on insurance sector as a whole; some insurers with concentrated exposures could be affected but no widespread solvency shortfalls.
- Default of second-largest banking counterparty:
  - Results more homogeneous in life sector; outliers in health sector possible.

### F. Liquidity Risk Analysis — Scope and Key Quantities
- Derivatives usage:
  - Market value of asset-side derivatives: 6.3 percent of assets (sector-wide).
  - Liability-side derivatives: 8.4 percent of total liabilities (sector-wide).
  - Interest rate swaps: largest derivative type with total notional value of around EUR 360bn as of end-2022 — almost exclusively used by life insurers.
- Liquidity analysis focus: impact of variation margin calls for interest rate swaps held by life insurers.

*Source: IMF staff calculations based on DNB data and Solvency II QRTs.*

### 91.      The FSAP team carried out a BU and TD analysis of potential liquidity risks from

### The FSAP liquidity and solvency risk analyses: insurers, pension funds, and households

### Insurance sector: approach and data
- BU and TD analysis of potential liquidity risks from variation margin calls for interest rate swaps; scenario assumes a one-day interest-rate increase of 100 basis points.
- Life insurers reported margin calls split into cash calls and collateral provided in kind, spread over the first two days after the shock, and reported planned sources to meet cash calls (cash, repo facilities, liquidation of assets re-valued with the shocked interest rate).
- A top-down analysis cross-checked reported margin calls using the S.08.01 template of Solvency II reporting.
- Data quality in QRTs was sufficient to re-value almost all interest rate swap positions reported by the life insurers in the sample; supervisory reporting quality has improved since Solvency II implementation in 2016.
- Note: cross-checks with EMIR data could enable further analysis given daily availability versus quarterly QRTs.

### Insurance sector: results of the liquidity risk analysis
- Total margin call reported by insurers in the BU analysis: EUR 10.1bn.
- Share of margin calls needing to be met in cash: 94 percent.
- Cash due within one day (Day 1): EUR 8.4bn.
- Cash due on Day 2: EUR 1.1bn.
- Insurers’ planned Day 1 liquidity raising (five insurers in sample): total EUR 10.2bn.
  - Day 1 liquidity composition (planned):
    - Cash and deposits: 22 percent.
    - Repo facilities: 53 percent (mainly uncommitted repos).
    - Money-market funds: 16 percent.
    - High-quality government bonds with short remaining maturity: 7 percent.
  - Liquidation of assets expected to be limited; remaining Day 1 liquidity intended to meet Day 2 margin calls, with possible further divestments in government bonds on Day 2.
- Heterogeneity in collateral management across insurers reduces systemic risk from overreliance on a single source; however, repo market amounts need assessment alongside simultaneous needs of other market participants.
- TD analysis cross-check:
  - TD margin call: EUR 10.3bn.
  - BU margin call: EUR 10.1bn.
  - Difference within expected modeling uncertainty, confirming reporting quality.

### Insurance sector: recommendation
- Close monitoring of repo market conditions remains warranted, because insurers’ reliance on repo markets should be assessed together with simultaneous liquidity needs by other market participants including Dutch pension funds.

### Pension funds: approach and scope
- TD solvency analysis covered the ten largest pension funds, representing 70 percent of the market in assets; analysis adopted the current DB regime with funding ratio as the performance metric.
- Adverse scenario aligned with insurance sector; interest-rate shock application used a more approximate duration-based approach for pension funds due to limited reporting on future cash flows and interest rate swap details for some funds.
- Liquidity analysis: DNB surveyed the five largest pension funds (covering about 60 percent of the market) in a BU approach. Data requested on potential margin calls and sources of liquidity (cash, repo facilities, liquidation of assets), re-valued under shocked interest rates.
- Four scenarios used for pension funds’ liquidity analysis; the most severe scenario presented assumed:
  - 36-bps increase in EUR interest rates.
  - 4.4 percent appreciation of the Euro against the U.S. dollar.
  - Note: additional currency-specific interest rate shocks include 44 bps for USD and 77 bps for GBP in the broader analysis.

### Pension funds: solvency analysis results
- Interest rate effects:
  - Higher interest rates lower the value of pension fund liabilities by 27 percent on average.
  - Asset values decline by 23 percent on average (EUR -235bn for the sample).
  - Decline in liabilities amounts to EUR 245bn for the sample.
- Own funds (assets minus liabilities):
  - Own funds increase on average by 7 percent, from EUR 136bn to EUR 145bn.
- Funding ratios:
  - Average funding ratio increases by 7 percentage points to 122 percent (from 115 percent pre-stress).
  - Several pension funds record increases greater than 10 percentage points, especially those with larger duration gaps between assets and liabilities.
- A large effect on own funds stems from the adverse scenario’s shock to share prices; valuation loss on equities is almost equal to pre-stress own funds, and remaining asset-side shocks are substantially offset by liability value declines and FX effects.

### Pension funds: liquidity risk analysis results
- Total variation margin call in the scenario for pension funds in sample: EUR 25.0bn.
  - Margin call related to interest rate swaps: EUR 14.5bn.
- Share of margin calls met in kind (non-cash) among pension funds: 36 percent.
- Cash collateral call amount: EUR 18.4bn.
  - Cash collateral timing: 60 percent due within one day (Day 1), remainder due on Day 2.
- Settlement characteristics:
  - Smaller pension funds using bilateral swaps can often settle in kind, lowering liquidity needs.
  - Clearing obligation for larger pension funds increases liquidity risk with each new swap or derivative transaction.
- Sources of liquidity and Day 1 composition:
  - Cash and deposits: 18 percent of sourced liquidity on Day 1.
  - Reverse repos and uncommitted repo lines: 34 percent.
  - Liquidation of money-market funds and short-term high-quality bonds: 12 percent.
- Day 2 composition:
  - Further sale of money-market funds is most relevant, followed closely by further repo transactions (39 percent each).

### Pension funds: recommendation
- Close monitoring of market conditions and pension funds’ liquidity risk management practices remains crucial, given the importance of repo markets for large pension funds and the relatively shallow domestic repo market where large transactions can be difficult to absorb.

### Household sector: recent developments and vulnerabilities
- Household debt trends:
  - Household debt-to-income ratio decreased by almost 80 percentage points from the peak of 2011Q1 to 2023Q2.
  - Despite the decline, the Dutch household debt-to-income ratio remains one of the highest among advanced economies (debt ratios as of 2023Q1, except noted exceptions).
- Mortgage market structure:
  - Banks provide 69 percent of mortgages (more than 550 billion euros outstanding) as of 2023Q2.
  - Mortgages by insurers and investment funds increased over the last decade; activity by other financial institutions declined.
- Household assets and illiquidity:
  - Pension entitlements represent 25 percent of households’ total assets as of end-2022.
  - Real estate assets represent 50 percent of households’ total assets as of end-2022.
  - Pension entitlements dropped by almost 25 percent from 2020 to 2022.
  - Real estate asset values more than doubled from 2013 to 2022 and could experience a sharp correction.
- Borrower vulnerabilities:
  - More than 40 percent of mortgages contain an interest-only loan (IOL) element, exposing many households to refinancing risk that could become systemic around 2036 when a large volume of loans matures.
  - Interest rates increased sharply in the past two years; housing prices started to decrease by 6 percent from their peak of mid-2022 to mid-2023.
  - Average LTV ratios declined from 82 percent in 2013Q1 to 65.3 percent in 2022Q1 following LTV cap introduction and tightening.
  - Pockets of vulnerability exist among younger first-time buyers who purchased more at the end of the boom and pushed borrowing limits.
    - If house prices dropped by 20 percent:
      - 13 percent of homeowners would see their LTV ratio rise to over 90 percent.
      - 8 percent of homeowners would fall into negative equity.
- Cost-of-living pressures:
  - A CPB cost-of-living stress test suggests between 540,000 and 860,000 Dutch households face affordability problems due to high inflation (CPB, 2022).
  - Reports of household defaults have been increasing for several months.

*Source: IMF staff calculations and supervisory data as presented in the FSAP chapter.*

### 112.      Mortgage defaults are very low on aggregate, but loans with higher LTV and loan-to-

### 1nldea2024010 - 112. Mortgage defaults are very low on aggregate, but loans with higher LTV and loan-to-

### Key findings on mortgage performance and vulnerabilities
- Mortgage default rate on banks’ mortgages in the Netherlands: 0.7 percent as of 2022Q1 versus an average of 0.5 percent over 2013-2021.
- The Nationale Hypotheek Garantie (NHG) applies to a quarter of the total mortgage market.
- Full recourse of lenders on borrowers’ income and assets reduces incentives to default.
- Mortgages with higher LTV and LTI at origination, higher current DSTI ratios, and interest-only (IO) status have historically been more fragile (de Haan and Mastrogiacomo (2020) cited).

### Interest-rate pass-through and debt-service impacts
- 75 percent of outstanding mortgages have rates fixed for more than 5 years.
- Less than 3 percent of Dutch mortgages have interest rates which are flexible in 2023 and 25 percent within the next 5 years.
- DNB Autumn 2022 FSR scenario: following a 3-percentage-point increase in interest rates,
  - average DSTI ratio of homeowners whose fixed-interest period expires in the short term would increase from 12 to 17 percent;
  - proportion of households spending more than a quarter of their disposable income on monthly mortgage payments would rise from 12 to 26 percent.
- Compared to some other European countries (Austria, Estonia, Finland, Lithuania, Luxembourg, Spain), the average interest rate of outstanding mortgages has moderately increased in the Netherlands so far.

### Interest-only (IO) mortgages: stock, maturity concentration, and risks
- IO mortgage stock:
  - IO mortgages represented less than 44 percent of the total stock of mortgages in 2022Q1 versus more than 57 percent in 2013Q1.
  - About 2.78 million households (out of 3.82 million households with mortgage debt) still have either a partial IO mortgage or a 100-percent IO mortgage.
  - 29 percent of the stock of IO debt matures between 2034-2039.
- Maturity peak and associated concerns:
  - The maturity peak for IO loans has significantly decreased due to early repayment, refinancing, and conversion into amortizing loans but remains high.
  - Mortgage providers have little prior experience with large-scale maturities of IO loans and therefore limited data on related credit and refinancing risk.
  - Potential borrower outcomes by the 2030s include negative equity, higher refinancing interest rates, and lower incomes in retirement, reducing qualification for new mortgages and possibly forcing home sales.
- AFM basic scenario findings (AFM, 2021):
  - Assumptions: property values increase by 2.5 percent annually and the interest rate is at 3 percent.
  - Estimated 78,000 households out of the 3 million households with an IO mortgage (2.6 percent) are at risk of having to sell their home during 2020-2050.
  - Of these, 18,000 households would not be able to use the sale to pay off their mortgage debt in full and would be left with residual debt.
- Alternative AFM adverse scenarios:
  - Housing prices drop by -22.6 percent in 5 years or interest rates increase by +2 percentage points to 5 percent in 5 years → 44,000 households left with residual debt.
  - Residential market crisis with housing prices decreasing by -34.7 percent over 5 years → 147,000 households left with residual debt.
  - Note: AFM study does not consider simultaneous increase in interest rates and drop in housing prices.
- IMF back-of-the-envelope estimate (comparable scenario):
  - A drop of house prices of 35 percent today, followed by an annual increase of 2 percent subsequently, translates into a drop of 18 percent by mid-2030 (that is of 72.000 euros in the average value of houses).
  - With 18,000 households left with residual debt, total loss could amount to 1.3 billion euros.
  - Comparison: 110 billion euros of Tier 1 capital held by the four largest Dutch banks at end-2022.
  - Sale of houses by mortgage providers could trigger second-round effects on prices, but these are expected to be limited given the structural shortage of housing supply.

### Authorities’ assessments and lender responses
- Authorities (AFM, DNB, FSC) view systemic risk from maturing IOLs as limited under baseline scenarios; DNB internal analysis and IMF back-of-the-envelope suggest cumulated losses remain well below banks’ capital.
- Policy and supervisory actions:
  - November 2021: Financial Stability Committee (FSC) called on banks to take further steps to reduce the risks of interest-only mortgages.
  - DNB has called on mortgage lenders to inform customers about IO risks and to strengthen risk management.
  - AFM campaign: since rollout in 2019, mortgage providers have contacted a total of 1.68 million customers with an IO mortgage or part IO mortgage.
    - 370,000 customers have tested whether they would still be able to afford the IO mortgage loan in the future.
    - 744,000 customers have taken action to improve their financial situation (incidental or periodic repayments, conversion to part repayment mortgages, additional savings).
    - The number of customers at increased risk has dropped by 63 percent during this process, partly due to actions taken by customers and partly reflecting the increase in house prices.

### Simulation analysis methodology and data (DNB Real Estate Vulnerability Assessment Model)
- Model scope and mechanics:
  - Scenario-based tool assessing the impact of shocks to nominal interest rate, nominal wage growth, inflation, unemployment, and housing prices on borrowers’ DSTI and LTV ratios over a 3-year horizon.
  - Changes in interest rate, inflation, nominal wage growth, and unemployment affect DSTI via changes in debt service (installments) and disposable income.
  - LTV depends negatively on the change in housing prices.
  - Changes in DSTI and LTV are used to compute changes in PD and expected loss (EL) for banks’ mortgage portfolios using estimated elasticities ε_DSTI and ε_LTV (estimated with loan-level data 2012-2018).
- Key model assumptions (conservative):
  - Absence of new loans over simulation horizon; existing exposures decline only through amortization.
  - Fixed net replacement rate for unemployment at 73 percent of income (OECD estimate); unemployment transition is absorbing (once unemployed, remain unemployed for the scenario horizon).
  - No endogenous borrower behavior such as pre-payments beyond modeled actions.
  - PDs and LGDs reported at aggregate level; changes in shares of borrowers reported for granular groups.
  - High-risk borrowers defined as borrowers with current DSTI ratios breaching 90 percent of DSTI limits defined by NIBUD (2023 DSTI limits applied).
  - Simulations consider equivalent LTI limits corresponding to DSTI limits under a 30-year amortization.
- Loan-Level Data (LLD) used by DNB:
  - Quarterly administrative panel dataset with information on six million loans and three million borrowers.
  - Contains about 75 variables related to loans and borrowers (mortgage provider, loan types, interest rates, NHG participation, origination and maturity, current property evaluation, borrower information at origination).
  - Latest version used for simulations: 2022Q1.
  - Data quality: low measurement error due to administrative nature; checked annually by households during tax form approval/correction.
- LLD sample characteristics:
  - Almost three quarters of mortgages in the LLD sample are held by borrowers older than 45.
  - Age buckets 45-55 and 55-65 each represent 23 percent of borrowers.
  - 97 percent of loans are outside Amsterdam.
  - Almost 96 percent of loans have an LTV ratio under 90 percent (higher than 2013 sample).
  - Distribution of LTI ratios shifted right: 87 percent of mortgages having LTI ratios above 3 versus 34.5 percent in the 2013 sample.

### Vulnerable borrower profiles and aggregate risks
- Vulnerable groups include:
  - Most recent buyers (higher purchase prices), buyers with bridge loans, young first-time buyers, and lower-income households.
- Debt-to-income concentration:
  - Around 60 percent of households under the age of 36 and 45 percent of older households have a debt-to-income ratio above 450 percent.
- Aggregate mitigants and concerns:
  - Low share of mortgages with flexible interest rates in the short term mitigates immediate rate pass-through.
  - High nominal wage growth in some sectors (reaching 10 percent in annual terms) mitigates some DSTI pressures.
  - Reports of increasing household defaults and rising proportion of households facing financial difficulties (Nationaal Instituut voor Budgetvoorlichting NIBUD) raise risk of growing losses on lenders’ mortgage portfolios over time.

*Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1nldea2024010.pdf*

### 122.      In aggregate, the proportion of borrowers at risk increases moderately in the adverse

### 1nldea2024010 - 122.      In aggregate, the proportion of borrowers at risk increases moderately in the adverse

### Aggregate borrower vulnerability: baseline vs adverse
- Proportion of borrowers with a high LTV is almost 11 percentage points higher in the adverse scenario than in the baseline.
- Housing prices fall by -29 percent in cumulative terms over three years in the adverse scenario, driving the LTV increase.
- Proportion of borrowers with a high DTI is 4 percentage points higher in the adverse scenario than in the baseline.
- Proportion of borrowers with a high DSTI is 3 percentage points higher in the adverse scenario than in the baseline.
- These aggregate changes mask significant heterogeneity across borrower groups.

### Vulnerability by income
- Lower-income borrowers are most vulnerable according to high-DTI and high-DSTI criteria.
- Lowest-income households have a higher proportion of borrowers with a DSTI ratio above 25 percent.
- Low-income households show a high proportion of borrowers with high DTI ratios.
- Low-income households do not show a higher proportion of borrowers with high LTV ratios compared to low-income households (as reported).

### Joint LTV and DTI distribution
- Proportion of borrowers with LTV above 80 and DTI above 4 increases from 4.3 percent in the baseline scenario to 11 percent in the adverse scenario.
- Proportion of the safest borrowers (by the study’s definition) decreases from 73.6 percent in the baseline to 58.5 percent in the adverse scenario as the distribution of risky loans shifts right.

### Vulnerability by age
- Young borrowers have the highest proportion of borrowers with a high LTV ratio (younger buyers more likely to have purchased at high valuations).
- Young households are not the group with the highest proportion of high DTI ratios in the adverse scenario; older households exhibit more IOLs (interest-only loans), which are associated with higher debt due to absence of repayment.
- Young borrowers are more likely to have a high proportion of individuals with a high DSTI as they tend to borrow at the limit.
- DSTI-based risk for older borrowers barely changes in the adverse scenario, reflecting that older borrowers are relatively shielded from the income shock because most are retired.

### NIBUD DSTI-limit based results
- Using breach of the NIBUD DSTI limit (90-percent of the DSTI limit) to define high-risk borrowers:
  - Aggregate proportion of high-risk borrowers increases from 6.6 percent to 8.2 percent over three years in the adverse scenario.
  - Borrowers with the lowest incomes: proportion increases from 19.3 percent to 22.5 percent.
  - Borrowers under 35 years: proportion increases from 7.1 percent to 11.3 percent.
  - High-risk borrowers with initially high LTV and DTI ratios see the largest increases.
  - Proportion of high-risk households with a large share of IOLs barely increases (households with high IOL shares are older and less impacted by the income shock).

### Recommendations for the household stress test model
- Calibrate model equations by group of borrowers to reflect heterogeneity in elasticities.
- Estimate elasticities of the probability of default (PD) to DSTI and LTV shocks for various sub-groups of borrowers to fully use outputs from the Real Estate Vulnerability Assessment Model.
- Consider DSTI thresholds that better reflect probability of borrower distress, not solely a uniform 25-percent DSTI limit.
- Complement with sensitivity analysis using alternative DSTI thresholds.

### Corporate sector: overview and stylized facts
- The non-financial corporate (NFC) sector recovered from the Covid-19 downturn; value-added constitutes 93.4 percent of total value-added as of the third quarter of 2023.
- Bankruptcies have risen with tightened financial conditions, approaching pre-pandemic levels; notable surges in construction and service sectors.
- NFC debt-to-GDP has fallen relative to GDP but NFC debt increased with the recovery and remains elevated versus pre-pandemic levels.
- Corporate dependence on short-term securities and loans has increased since 2021, raising interest-rate sensitivity.
- NFCs rely on CRE/RRE as collateral.
- Within the euro area, the Netherlands ranks fourth highest in debt-to-surplus ratio.
- Dutch firms maintained gross profit rates above 45 percent for the past decade, reaching 48.3 percent in 2022 (surpassing the EA average).

### Corporate sector vulnerabilities (firm-level data)
- Orbis database sample: initial universe over 750,000 firms reduced to approximately 338,000 after excluding financial corporations and restricting to firms with recent data (2019–2021); further reduced to 3,369 firms due to missing interest/payment data.
- The final sample represents 72 percent of total NFC assets but may not be representative across sectors (overrepresentation: mining (BDE), manufacturing (C), wholesale and accommodation (GHI); underrepresentation: professional (MN), public (OPQ), other (RSTU) services).
- Two financial indicators used:
  - Interest coverage ratio (ICR) = EBIT / interest payments; ICR < 1 indicates weak debt-servicing capability.
  - Cash balance = initial cash balance + EBIT net of interest payments and taxes; negative cash balance implies need to liquidate assets or borrow.
- Since 2018 the proportion of firms with repayment trouble (ICR < 1) has risen, notably in wholesale, transportation and accommodation (GHI), real estate service (L), and professional service (M).
- Need for external financing generally declined over past decade as larger firms accumulated cash, but rising interest rates could force some firms to increase leverage.
- Vulnerable sectors with low ICRs continue to have lower cash balances; accommodation and transportation sectors show greatest liquidity needs.

### Dynamic scenario-based corporate stress test (2022–2025)
- Methodology: dynamic scenario-based approach (Tressel and Ding, 2021); firm-level panel regressions for 2001–2021 link financial indicators to structural/cyclical characteristics, industry fixed effects, and macro-financial conditions.
- Under baseline and adverse scenarios, shocks to GDP growth and short- and long-term interest rates affect firms’ ICRs and cash balances through accounting identities and regression projections.
- Short- and long-term interest rates increase and remain elevated in both scenarios.
- GDP experiences negative growth rates in the adverse scenario, with the adverse effects most pronounced in 2023 and gradual recovery toward steady state; even in 2025 GDP growth rates remain below 2022 levels while both interest rates exceed 2022 levels.

*Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1nldea2024010.pdf*

### 139.      Both scenarios trigger a substantial increase in the proportion of firms facing debt-

### 1nldea2024010 - 139.

### Corporate stress-test results: incidence of debt-servicing difficulties and cash shortfalls
- Both scenarios trigger a substantial increase in the proportion of firms facing debt-servicing difficulties and borrowing needs, with the adverse scenario having significantly more severe repercussions.
- Share of firms with ICR below 1:
  - 8.2 percent in 2022.
  - 31.5 percent in 2023 under the adverse scenario.
  - Rises by only 5.3 percentage points under the baseline scenario (2023).
- Proportion of firms with negative cash balances:
  - Doubles from 2022 to 2023 under the adverse scenario, reaching a peak of 20.6 percent in 2024.
  - In the baseline scenario, the proportion is higher than in 2022, hovering around 10 percent.

### Drivers of corporate vulnerability
- Primary drivers:
  - Weakened earnings due to subdued GDP growth that impedes sales growth.
  - Elevated debt burdens from increases in both short- and long-term interest rates.
- Consequence: the compounding effect of lower earnings and higher interest rates prolongs the recovery process and makes it significantly slower under the adverse scenario.

### Regression results (selected coefficients and statistics)
- Dependent variables reported: ROA, Leverage, Sales Growth (median or regression context implied).
- Coefficients (with robust standard errors in parentheses):
  - ROA(-1): 0.581*** (0.020)
  - Leverage(-1): -0.000 (0.006)
  - Sales Growth(-1): -0.007** (0.002)
  - Asset Tangibility Ratio (-1): 0.009** (0.003)
  - Cash Flow Generation Ratio (-1): 0.004*** (0.001)
  - Size (-1): 0.052 (0.077)
  - Real GDP Growth (-1): 0.187*** (0.061)
- Observations and fit:
  - Observations: 18,056 (ROA), 16,472 (Leverage), 18,047 (Sales Growth)
  - R-squared: 0.387 (ROA), 0.966 (Leverage), 0.084 (Sales Growth)
- Significance notation:
  - *** p<0.01, ** p<0.05, * p<0.1

### Policy recommendations for corporate sector monitoring
- Vigilant monitoring of the corporate sector, particularly firms with shorter maturity profiles.
- Conduct a more granular examination of corporate balance sheets and financial conditions.
- Improve access to bank loan-level data, since NFCs often use RRE and CRE as collateral; loan-level data would clarify borrowing terms and maturities faced by individual firms.

### Commercial Real Estate (CRE) market: exposures and current metrics
- Total CRE exposures of Dutch banks, pension funds and insurers: approximately 360 billion euros (domestic and international).
- Bank loans to the CRE sector:
  - Account for 7 percent of the balance sheet.
  - Account for 10.7 percent of total loans.
- NPL ratio for CRE loans:
  - Reached 2.8 percent in 2022Q4 (historically low).
- Average LTV ratio of CRE loans:
  - Remains below 60 percent in 2023Q2.
  - Average recovery rate for foreclosed CRE: 70 percent.
- Trend: recent growth in proportion of CRE loans with LTV ratios below 60 percent.

### CRE price dynamics and valuation indicators
- CRE prices:
  - Doubled in value between 2015Q1 and 2022Q2, with significant declines thereafter.
  - Residential and office properties experienced the most substantial price increase until early 2022, then underwent a pronounced downturn.
- Capitalization rate (net operating income yield) trends:
  - Declined from 6.5 percent in 2008Q4 to 3.4 percent in 2019Q4 for office sector / overall trend.
  - Capitalization rate rose with reduced rental income during/after the pandemic and as CRE price growth turned negative.

### CRE vulnerability assessment: price misalignment and CaR methodology
- Price misalignment measured via capitalization rate using HP filter: positive from 2016Q4 to 2022Q3 for the Netherlands, indicating vulnerability to adverse shocks (e.g., higher interest rates).
- CaR definition and method:
  - CaR defined as the 5th percentile of the conditional distribution of future CRE price growth.
  - Estimation uses panel quantile estimation methodology (Machado and Santos Silva (2019)) and a cross-country panel sample of 12 European countries due to limited Dutch time series.
- Quantified effect:
  - A one-standard deviation increase in CRE price misalignment (equivalent to 24 basis points) is associated with a cumulative 3.4 percentage point increase in downside risks to CRE prices (or decrease in CaR value) over four quarters.
- Netherlands CaR profile:
  - CaR has hovered around -2 percent since 2012.
  - CaR moved from -5.4 percent in 2008Q4 to -1.6 percent in 2019Q4.
  - Post-pandemic: downside risks increased after the pandemic, manifested in the sharp price drop in 2022; low CaR value in 2023Q1 indicates downside risks remain elevated compared to pre-pandemic levels.

### CRE policy recommendations
- Close monitoring of CRE market developments is warranted given elevated downside risks.
- Data issues:
  - Public data on CRE is limited compared to RRE.
  - CBS provides a CRE price index based on Land Registry and registry data, but there is a significant discrepancy between the CBS price index and market data from private vendors.
- Recommendation: explore how to close the gap between market data and the CBS price using the ample granular data sources available to authorities.

*Source: IMF staff calculations and analysis as presented in the provided content.*

### 156.      As the CRE market is closely linked to both corporate and financial sectors, stress

### 1nldea2024010 - 156. As the CRE market is closely linked to both corporate and financial sectors, stress

### Commercial Real Estate (CRE) linkages and data caveat
- As the CRE market is closely linked to both corporate and financial sectors, stress testing exercises with granular loan level CRE data could be also considered.
- Caveat: absence of a connection between CRE prices and the financial sector’s exposure to CRE sector.
- Recommendation: Access to bank loan-level data, coupled with detailed information on properties, could enhance the efficiency of the credit risk assessment.

### Realized Average Growth Rate and CaR (quarterly averages)
- Realized Average Growth Rate (quarterly average over future four quarters, except 2023Q1 which is the averaged value over recent three quarters):
  - 2019Q4: 1.53
  - 2020Q4: 1.82
  - 2021Q4: 2.71
  - 2022Q4: -2.89
  - 2023Q1*: -2.24
- CaR5:
  - 2019Q4: -1.63
  - 2020Q4: -1.45
  - 2021Q4: -2.11
  - 2022Q4: -5.88
  - 2023Q1*: -3.21
- Note: *All values are the quarterly average over future four quarters, except for 2023Q1 which is the averaged value over recent three quarters.

### Financial Soundness Indicators (selected highlights and exact values)
- Regulatory Capital to Risk-Weighted Assets:
  - 2017: 22.0
  - 2018: 22.3
  - 2019: 22.9
  - 2020: 22.8
  - 2021: 22.4
  - 2022: 21.0
  - 2023Q3: 21.2
- Regulatory Tier 1 Capital to Risk-Weighted Assets: 18.4; 18.8; 18.9; 19.3; 19.3; 18.0; 18.5
- Common Equity Tier 1 Capital to Risk-Weighted Assets: 16.8; 17.4; 17.4; 16.1; 16.5 (selected entries)
- Nonperforming loans to total gross loans: 2.3; 2.0; 1.8; 1.9; 1.7; 1.6; 1.6
- Return on Assets: 0.7; 0.7; 0.6; 0.3; 0.7; 0.6; 1.1
- Return on Equity: 12.8; 11.7; 7.6; 3.3; 8.8; 8.0; 12.8
- Interest Margin to Gross Income: 73.5; 54.8; 54.9; 54.9; 54.1; 47.7; 49.6; 55.6 (selected entries)
- Liquidity Coverage Ratio: 138.6; 169.1; 163.3; 153.6; 159.9 (selected recent entries)
- Net Stable Funding Ratio: 168.6; 150.2; 135.6; 133.6; 135.7 (selected entries)
- Customer deposits to total (non interbank) loans: 62.3; 62.2; 74.0; 75.6; 75.3 (selected entries)
- Residential real estate loans to total gross loans: 27.3; 24.4; 37.5; 35.8; 43.2; 44.1; 43.3 (selected entries)
- Commercial real estate loans to total gross loans: 8.1; 8.0; 9.0; 10.1; 9.9 (selected entries)

### Financial sector structure (selected figures)
- Banks total assets (billions of euro) and percent of financial system / percent of GDP (selected snapshots):
  - Dec-16, Dec-19, Jun-23 (multiple entries presented in source)
- Globally systemic institutions: 184; 513.2; 119.3; 189; 212.5; 109.7; 11,029; 12.7; 9.6 (table entries reflect counts and asset figures across periods)
- Pension funds: 297; 1,266; 19.8; 1,778; 2,201; 1,554; 21.8; 1,191; 2,131; 1,469; 18.1; 1,142.3 (table entries reflect counts and asset figures across periods)
- Investment funds and other financial institutions: multiple entries with asset levels across periods (source table).

### Financial Soundness Indicators for SIs and LSIs (2018–2022)
- Return on equity:
  - SI: 8.71; 7.62; 3.27; 8.35; 8.00
  - LSI: 4.63; 9.08; 1.97; 8.24; 5.80
- Return on assets:
  - SI: 0.52; 0.46; 0.19; 0.49; 0.48
  - LSI: 0.54; 0.93; 0.22; 0.92; 0.63
- Net interest margin:
  - SI: 1.44; 1.42; 1.28; 1.23; 1.34
  - LSI: 1.26; 1.15; 1.05; 0.96; 1.08
- Cost-to-income ratio:
  - SI: 53.70; 50.32; 49.39; 49.74; 49.12
  - LSI: 53.57; 49.51; 58.03; 48.38; 52.41
- CET1 to RWA:
  - SI: 16.45; 16.48; 17.03; 16.97; 15.68
  - LSI: 21.36; 20.08; 25.20; 24.99; 23.35
- Leverage ratio:
  - SI: 4.61; 4.88; 5.23; 6.54; 5.80
  - LSI: 9.87; 8.53; 10.25; 10.34; 9.18
- NPL ratio:
  - SI: 1.95; 1.86; 2.24; 1.73; 1.60
  - LSI: 1.88; 1.63; 2.29; 1.98; 1.80
- RWA density:
  - SI: 30.46; 31.19; 28.68; 29.76; 33.23
  - LSI: 48.06; 44.30; 38.79; 38.99; 40.16
- Loan to deposit ratio:
  - SI: 123.45; 124.52; 106.46; 104.01; 108.04
  - LSI: 104.27; 102.76; 97.46; 94.09; 95.04

### Risk Assessment Matrix: key risks, likelihood, and expected impacts
- Global conjunctural risks:
  - Abrupt global slowdown or recession — Likelihood: Medium.
    - Expected impact: sharp drop in economic activity, domestic and external demand; spillovers from depressed activity in key trading partners (e.g., Germany) would affect the Netherlands and exacerbate credit risks.
  - Europe: Intensifying fallout from the war in Ukraine, supply disruptions, tight financial conditions, and real estate market corrections — Likelihood: Medium.
- Intensification of regional conflict(s) — Likelihood: High.
  - Expected impact: trade, energy, food, tourism, supply chains disruptions; increased refugee flows; payment system and financial flow disruptions.
- Monetary policy miscalibration — Likelihood: Medium.
  - Expected impact: may require policy reversal leading to demand cooling, house price declines, pressures on borrowers given elevated private debt; tightened conditions could reduce value of marked-to-market securities.
- Systemic financial instability — Likelihood: Medium.
  - Expected impact: sharp swings in asset prices and risk premia; could affect capital positions and cause failures of individual banks/NBFIs and fire sales.
- Structural risks:
  - Deepening geo-economic fragmentation — Likelihood: High.
    - Expected impact: supply disruptions, weaker investor confidence, impacts on bank asset quality and non-bank investment asset valuations.
  - Extreme climate events — Likelihood: Medium.
    - Expected impact: loss of lives, infrastructure damage, supply disruptions; the Netherlands vulnerable to sea level rise; nitrogen deposition policy actions could have adverse macroeconomic effects.
- Netherlands-specific risks:
  - Rapid correction of house prices — Likelihood: Medium.
    - Expected impact: Dutch banks highly exposed to highly indebted households; downward housing correction would worsen asset quality and reduce household consumption to service debts.
  - Adverse change in direction of economic and climate policies in context of political fragmentation — Likelihood: Medium.
    - Expected impact: supply disruptions, stranded assets, affecting investment and growth.

### Stress testing approach: Banks (Table 7)
- Institutional perimeter:
  - 6 significant institutions—over 90 percent of the banking system.
- Methodology and risk drivers:
  - Scenario-conditional simulation of P&L drivers: credit risk (credit impairment), interest rate risk (interest income and expense), market risk (mark-to-market revaluation of marketable securities).
  - Credit risk model linking macrofinancial shocks with default probabilities of loan portfolios by country of exposure.
  - Interest rate models linking risk free rates to lending and borrowing rates.
  - Market-to-market valuation linking sensitivity factors (“delta”) with shocks to interest rate, spread, foreign exchange rate, equity, and commodity prices.
- Scenarios:
  - Baseline scenario aligned with April 2023 IMF WEO.
  - Bespoke adverse scenarios based on RAM (Table 6) addressing the most relevant risks confronting the Dutch financial system.
- Sensitivity analyses:
  - Alternative interest rate paths: simulation on bank capital through interest income and expense assuming sight deposits move to term accounts and flow out of the banking system.
  - Liquidation of HTM securities: estimation of losses when banks are forced to liquidate held-to-maturity securities to cover cash shortfalls as funding runs off under stress scenario.
  - LSI analysis: credit risk analysis on foreign credit exposures of corporate and EM banks using publicly available default probabilities as proxy to stress test creditworthiness of foreign corporate exposures against macrofinancial scenarios of 40 economies.
- Bank Liquidity Stress Test:
  - Institutional perimeter: 6 significant institutions—over 90 percent of the banking system.
  - Methodology: regulatory liquidity stress test (evaluation of LCRs and NSFRs); cash-flow-based liquidity stress test; sensitivity analysis to model assumptions.
- Interconnectedness and Contagion Analysis:
  - Institutional perimeter: 14 banks, 27 insurers, 47 pension funds, 3,590 different marketable securities making up more than 50 percent of total assets for the median institution.
  - Methodology: institution-level contagion analysis based on a fire-sale channel (selling assets by institutions in distress affects other institutions through the price channel).
- Banking Sector Climate Risk Analysis:
  - Institutional perimeter: the six Dutch banks designated as systemically important.
  - Methodology and risk drivers:
    - Physical risk from floods mapped into economic damage.
    - Flood damages impact banks’ credit risk (domestic and international loans).
    - Macro approach mapping climate scenarios into macrofinancial scenarios; standard stress testing methodologies to assess implications of climate risks for banking system resiliency.
  - Scenarios:
    - Multiple flood scenarios across regions, different climate conditions and return periods.
    - Extreme flood scenarios including floods in unembanked and embanked areas.
    - Macrofinancial scenarios including impact of floods on Dutch economy and neighboring countries (Belgium and Germany).

### Stress testing approach: Insurers (Table 8)
- Insurers Solvency Stress Test (Top-down):
  - Institutional perimeter:
    - 5 life insurers; 5 P&C insurers; 6 health insurers.
    - Market share: Life: 93 percent, P&C: around 70 percent, Health: around 70 percent.
    - Consolidation level: Unconsolidated.
    - Data: Statutory returns.
    - Reference Date: June 30, 2023.
  - Channels of risk propagation / Methodology:
    - Investment assets: market value changes after price shocks.
    - Liabilities: valuation change due to interest rate shock.
    - Impact on available capital (net assets = stressed assets minus liabilities).
    - Recalculation of the SCR.
    - Time horizon: Instantaneous shock.
  - Adverse scenario (granular market and interest rate risks):
    - Risk-free rate: full Solvency II term structure incl. extrapolation towards the ultimate forward rate, EUR +147 bps (1y) and +158 bps (10y); USD +5 bps (1y) and +203 bps (10y).
    - Equity: -40.7 percent (Netherlands), -42.4 percent (Euro Area), -42.1 percent (U.S.), -41.2 percent (other advanced economies).
    - Property: -13.0 percent (domestic RRE), -15.0 percent (domestic CRE), -10.0 percent (foreign RRE), -12.0 percent (foreign CRE).
    - Sovereign bond spreads: +55 bps (Netherlands), +60 bps (Euro Area, United States).
    - Corporate bond spreads: +45 bps (AAA, non-financials) to 400 bps (CCC and lower).
    - Mortgage loan spreads: +45 bps.
    - Currency: -10.7 percent depreciation of the EUR external value.
  - Sensitivity analysis:
    - Parallel decline of the EUR interest term structure: -100 bps.
    - Appreciation of the EUR external value: +10 percent.
    - Default of largest banking counterparty.
  - Risk factors assessed: Market risks (equity, property); interest rate risks; credit risks (bond spreads, (mortgage) loan spreads, default of largest banking counterparty).
  - Regulatory/accounting standards: Solvency II, National GAAP.
  - Reporting formats / Output presentation:
    - Change in valuation of assets and liabilities.
    - Solvency ratios.
    - Aggregated capital shortfall.
    - Dispersion across companies.
    - Contribution of individual shocks.
- Insurer Liquidity Stress Test (Bottom-up and Top-down):
  - Institutional perimeter: 5 life insurers; Market share Life: 93 percent.
  - Data: company submissions and statutory returns; Reference date: June 30, 2023.
  - Channels of risk propagation / Methodology:
    - Variation margin call on interest rate swap positions after a sudden increase in interest rate.
    - Time horizon: Two days.
    - Scenario: Increase in short-term EUR interest rates by 100 bps.
    - Risk factor assessed: Liquidity risks.
    - Regulatory/accounting standards: Solvency II, National GAAP.
    - Output presentation:
      - Amount of margin call (per day).
      - Share of margin calls which could be met in kind.
      - Liquid assets.
      - Sources of liquidity to meet margin calls.

### Stress testing approach: Pension Funds (Table 9)
- Pension Funds: Solvency Stress Test (Top-down):
  - Institutional perimeter:
    - 10 occupational pension funds (defined benefit).
    - Market share: 70 percent of assets.
    - Data: Statutory returns.
    - Reference date: June 30, 2023.
  - Channels of risk propagation / Methodology:
    - Investment assets: market value changes after price shocks.
    - Liabilities: valuation change due to interest rate shock.
    - Impact on own funds (net assets = stressed assets minus liabilities).
    - Time horizon: Instantaneous shock.
  - Adverse scenario (same granularity as insurers):
    - Risk-free rate: full Solvency II term structure incl. extrapolation towards the ultimate forward rate, EUR +147 bps (1y) and +158 bps (10y); USD +5 bps (1y) and +203 bps (10y).
    - Equity: -40.7 percent (Netherlands), -42.4 percent (Euro Area), -42.1 percent (United States), -41.2 percent (other advanced economies).
    - Property: -13.0 percent (domestic RRE), -15.0 percent (domestic CRE), -10.0 percent (foreign RRE), -12.0 percent (foreign CRE).
    - Sovereign bond spreads: +55 bps (Netherlands), +60 bps (Euro Area, United States).
    - Corporate bond spreads: +45 bps (AAA, non-financials) to 400 bps (CCC and lower).
    - Mortgage loan spreads: +45 bps.
    - Currency: -10.7 percent depreciation of the EUR external value.
  - Sensitivity analysis:
    - Parallel decline of the EUR interest term structure: -100 bps.
    - Appreciation of the EUR external value: +10 percent.
    - Default of largest banking counterparty.

*Italic: Source: IMF.*

### 11. Risk factors assessed   • Market risks (equity, property)

### 11. Risk factors assessed   • Market risks (equity, property)

### Market and interest-rate risk overview
- Risk factors assessed:
  - Market risks (equity, property)
  - Interest rate risks
  - Credit risks: bond spreads, (mortgage) loan spreads, default of largest banking counterparty

### Pension Funds: Liquidity Risk (Bottom-up; Conducted by DNB)
- Institutional perimeter:
  - Number of institutions: 5 occupational pension funds (defined benefit)
  - Market Share: ~60 percent of assets
  - Data: Statutory returns
  - Reference Date: December 31, 2022
- Channels of risk propagation / Methodology:
  - Combination of interest rate (EA, U.S., UK, JP) and FX shocks (USD, GBP, JPY) leading to margin calls on pension funds’ derivative positions
  - Time horizon: Two days
- Scenario analysis — Four adverse scenarios:
  1. Parallel interest rate shock between 17 and 38 bps; EUR appreciation between 2.2 and 3.8 percent
  2. As scenario 1, with limited access to the repo market
  3. Parallel interest rate shock between 33 and 77 bps; EUR appreciation between 4.4 and 7.5 percent
  4. As scenario 3, with limited access to the repo market
- Reporting formats for results (Output presentation):
  - Change in values of assets and liabilities
  - Funding ratios
  - Dispersion across companies
  - Contribution of individual shocks

### Insurance / Investment fund stress specification (selected calibration and asset shocks)
- Asset allocation and shocks (selected items as presented):
  - Equity: Netherlands-40.7%, Equity-42.0%
  - Euro Area-42.4% (Debt***)
  - United States-42.1% (Money-market-0.5%)
  - Other advanced economies-41.2%
  - Emerging economies-33.6%
  - Real estate-12.0%
  - Unlisted-12.0%
  - Alternative-8.0%
  - Private equity-12.0%
  - Infrastructure-4.0%
  - RRE, domestic-13.0%
  - CRE, domestic-15.0%
  - RRE, other countries-10.0%
  - CRE, other countries-12.0%
  - Structured notes-6.0%
  - Collateralised securities-4.0%
  - Other investments-8.0%
  - EUR external value-10.7%
- Credit spread shocks (in bps) by issuer/country (selected rows):
  - Netherlands+55
  - Euro Area+60
  - United States+60
  - Other advanced economies+65
  - Emerging economies+100
  - Supranational+/-0
- Corporate bonds (non-financials, in bps) by Credit Quality Step (selected):
  - Credit Quality Step 0+45
  - Credit Quality Step 1+65
  - Credit Quality Step 2+85
  - Credit Quality Step 3+120
  - Credit Quality Step 4+180
  - Credit Quality Step 5+300
  - Credit Quality Step 6+400
  - Unrated+120
- Interest rates (in bps) (selected):
  - EUR 1Y+147
  - EUR 10Y+158
  - USD 1Y+5
  - USD 10Y+203
- Note: *** company-specific shock based on rating breakdown and duration of direct holdings

### Liquidity risks (Pension funds and banking liquidity calibration)
- Risk factors assessed: Liquidity risks
- Regulatory/accounting standards: National GAAP
- Reporting formats for results (Output presentation):
  - Aggregated margin calls (absolute amount, relative to liquid assets)
  - Cashflows and liquidity position
  - Dispersion across companies

- Appendix I. Table 1 — Calibration for Cash-flow Analyses (selected outflow/inflow/counterbalancing capacity rates)
  - OUTFLOWS (Baseline / Severe):
    - Unsecured bonds due: 100% / 100%
    - Regulated covered bonds: 100% / 100%
    - Securitisations due: 100% / 100%
    - Stable retail deposits: 5% / 20%
    - Other retail deposits: 10% / 30%
    - Operational deposits: 25% / 40%
    - Non-operational deposits from central banks: 40% / 40%
    - Non-operational deposits from non-financial corporates: 40% / 75%
  - INFLOWS (Baseline / Severe):
    - Level 1 central bank: 0% / 100%
    - Level 2A corporate bonds (CQS1): 15% / 100%
    - Level 2B ABS (CQS1): 50% / 100%
    - Retail customers: 50% / 25%
    - Non-financial corporates: 50% / 25%
    - Credit institutions: 100% / 100%
    - Other counterparties: 50% / 25%
    - FX-swaps maturing: 100% / 100%
  - COUNTERBALANCING CAPACITY (Baseline / Severe):
    - Coins and banknotes: 100% / 100%
    - Withdrawable central bank reserves: 100% / 100%
    - Level 1 central bank: 100% / 93%
    - Level 2A corporate bonds (CQS1): 85% / 80%
    - Level 2B ABS (CQS1): 50% / 50%
    - Level 1 facilities: 100% / 100%
    - Level 2B restricted use facilities: 100% / 100%
  - CONTINGENCIES:
    - Considered as Level 2B by the receiver: 6% / 6%
    - Liquidity facilities: 38% / 38%
    - Outflows due to downgrade triggers: 100% / 100%

### Residential mortgage and loan-level statistics (based on 2022Q1 loan-level database, DNB)
- Table 11: Share of Mortgages in Arrears by Originating LTV and LTI Ratios (values presented in percent; tabulated by originating LTV and LTI)
- Table 12: Proportion of Interest-only Loans by Age Bracket (percent of mortgage loan production) — selected values by quarter:
  - 2013Q4, 2018Q4, 2021Q1, 2021Q2, 2021Q3, 2021Q4, 2022Q1
  - Age 18-25: 2.2 0.4 1.7 1.6 2.6 4.1 4.5
  - Age 26-35: 20.4 8.1 9.5 9.8 12.0 14.2 14.7
  - Age 36-45: 41.8 26.2 26.5 27.6 29.2 31.7 31.1
  - Age 46-55: 56.1 38.5 42.0 43.8 44.8 47.8 47.6
  - Age 56-65: 74.3 59.0 61.3 63.7 65.0 67.7 67.9
  - Age 66+: 90.5 81.8 82.8 83.8 83.6 85.4 85.9
- Table 13: Composition of Residential Mortgages (percent, based on 2022Q1):
  - Location: Amsterdam 3.0, Non-Amsterdam 97.0
  - Current LTV ratio distribution:
    - LTV under 60: 61.9
    - LTV 60-70: 16.9
    - LTV 70-80: 11.2
    - LTV 80-90: 5.8
    - LTV 90-100: 3.0
    - LTV 100-110: 0.7
    - LTV 110-120: 0.2
    - LTV 120+: 0.3
  - Current LTI ratio distribution:
    - LTI 0-2 times: 8.3
    - LTI 2-3 times: 14.7
    - LTI 3-4 times: 38.3
    - LTI 4-5 times: 30.2
    - LTI 5-6 times: 11.5
    - LTI over 6 times: 7.0
  - Gross Income (threshold in euros) by percentile:
    - Income under 20-percentile: 29,372
    - Income 20-40-percentile: 43,673
    - Income 40-60-percentile: 59,218
    - Income 60-80-percentile: 81,871
    - Income 80-100-percentile: 221,788

### Sample coverage by sector (Orbis, 2021)
- Total number of firms: 3,369
- Number of firms by NACE sector (selected):
  - A Agriculture, Forestry and Fishing: 40
  - C Manufacturing: 577
  - F Construction: 207
  - GHI Wholesale and retail trade, transportation and storage, accommodation, and food service: 1,046
  - L Real estate activities: 185
  - MN Professional, scientific, technical, administration and support service: 665
  - OPQ Public administration, defense, education, human health, and social work: 283
  - RSTU Other services: 69

### Appendix I — Satellite models for bank solvency stress test (credit risk and PD projection)
- Credit Risk: Scenario-conditional PD trajectories
  - Portfolios modeled: mortgage, other retail, qualifying revolving, corporate, government, financial institution
  - Countries of exposure: Netherlands, Germany, Belgium, UK, United States, Australia, and the rest of the world
  - For mortgage, other retail, qualifying revolving, and corporate portfolios: Panel Autoregressive Distributed Lag (ARDL) model deployed; logit-transformed probability of default equation:
    - ln(PD_{i,t}/(1−PD_{i,t})) = α_i + λ∙ ln(PD_{i,t−1}/(1−PD_{i,t−1})) + ∑_{s=0}^P β_{i,s}∙ z_{t−s} + u_{i,t}
  - Explanatory variables z_t include economic growth, interest rate, housing price growth, and real wage growth; selection may differ across portfolios
  - For government and financial institution PDs, structural model used:
    - PD_{i,t} = Φ(Credit Spread_{i,t}/(1−Recovery Rate))
  - PD projections by portfolio and economy transformed to bank-specific level by assuming constant differential of riskiness between system aggregate and a bank holding the same portfolio; difference in distance-to-default as of 2022 is held constant through the horizon

### Interest rate risk — pass-through coefficients and rate projection
- Two-part analysis:
  1. Estimate how fast monetary condition shifts “pass through” to lending and funding rates of various instruments
  2. Compute interest income and expense based on repricing schedule
- Time-series model for pass-through:
  - i_t = α_i + λ∙ i_{t−1} + ∑_{s=0}^P β_{i,s}∙ z_{t−s} + u_t
  - i_t denotes nominal interest of instruments: household sight deposit, corporate sight deposit, household term deposit, corporate term deposit, mortgage loan, consumer loan, and corporate loan rates
  - z_{t−s} includes short-term risk-free rates, inflation, GDP growth; regressions on loan rates include long-term risk-free rates to capture term premium
- Estimation approach:
  - Bayesian Model Averaging (BMA) used to account for model uncertainty and allow sign constraints
  - Market-based instruments (debt securities and derivatives): assumed 100 percent “pass-through”
  - For TLTRO deposits: assumed banks issue debt securities to cover any repayment amount; interest expenses increase as banks move from paying negative to positive rates on outstanding TLTRO amount

### Contagion methodology (network sales and price impact)
- Agents, assets, and rounds: N, M, T
- p_t: M×1 vector of asset prices in round t; p_0 normalized to ones
- q_{i,t,j}: number of assets j owned by agent i in round t; x_{i,t,j}: number of asset j sold by agent i in round t
- Default rules:
  - Banks and insurers: default if capital is negative
  - Pension funds: do not default
- Fire-sale rule (as in Cetorelli and others (2023)):
  - Agent i sells assets of value R_i Euros for every Euro of market-value losses, where R_i = 1/(LR_i−1), and LR_i = (A_{i,0}−L_{i,0})/A_{i,0}
  - Agents sell proportionally to the value of their holdings
  - Market prices evolve according to: p_{j,t+1} = p_{j,t} (1 − δ_{j,t−1} ∑_{i=1}^N x_{i,t,j})
  - Price elasticities δ_{j,t−1} given by: δ_{j,t−1} = δ_j−1 = (1−γ) / ∑_{i=1}^N q_{i,0,j}, with γ ∈ (0,1)
  - Non-linearities: prices of securities issued by agents in default are set to zero
  - Interpretation: δ_j−1 captures market depth; fewer initial holders → larger δ_j−1 → larger price impact for a given sale

### Dynamic scenario-based corporate stress test methodology
- Data: Orbis firm-level data, 2001–2022, excluding financial firms
- Modeling approach:
  - Firm-level dynamic OLS regressions with industry fixed effects:
    - Y_{i,s,t} = α Y_{i,s,t−1} + Γ⋅X_{i,s,t−1} + Φ⋅Z_t + δ_s + ε_{i,s,t}
  - Dependent variables: ROA, leverage (debt-to-asset ratio), sales growth
  - Firm-level regressors: lagged ROA, lagged leverage, size (total assets relative to average), asset tangibility, cashflow generation ratio (sales/total assets), growth opportunities (sales growth)
  - Some structural characteristics (size, asset tangibility, cashflow generation ratio) set at 2021 values
- Accounting identity linkages used for projections:
  - Cash = Initial Cash and eq. + EBIT − Taxes − Interest Expense
  - Debt Increase = − Cash, if Cash < 0
  - Interest Expense_t = [ Interest Rate_{t−1} + (LTD/TD)×ΔLT Rate_t + (1−LTD/TD)×ΔST Rate_t ] × TD_{t−1}
  - LTD/TD assumed constant
  - Macro inputs: Real GDP growth rate, ΔLT Rate_t, and ΔST Rate_t provided by macro baseline and adverse scenarios

_Italic: Source: IMF staff (material extracted from "11. Risk factors assessed • Market risks (equity, property)" section of the Netherlands FSAP chapter)._

### 20. This annex describes the methodology used in the CaR analysis. This analysis uses

### Annex: Methodology used in the CaR analysis

### Data
- Cross-country unbalanced panel data for 12 European economies from 2000Q1 to 2023Q1.

### Panel quantile model and estimation approach
- A panel quantile regression model, following Canay (2011), is used to assess impacts on the τ-th quantile of future changes in CRE prices.
- Baseline panel quantile model:
  - ΔhYi,t+h,τ = αi,τh + βτh Xi,t + εi,t,τh
  - ΔhYi,t+h,τ denotes the quarterly average percentage change in real CRE prices in country i from the base time t to t + h (h = 1, 2, ..., H), at a specific quantile τ ∈ (0,1).
  - Xi,t is a k×1 vector of control variables including:
    - lagged real CRE price growth,
    - CRE price misalignment,
    - financial condition index purged of residential real estate price,
    - changes in credit-to-GDP ratio,
    - real GDP growth rate,
    - capital inflows to GDP ratio.
  - αi,τh are country fixed effects.
  - All control variables except for lagged real CRE price growth rate are normalized by their standard deviation for the purpose of comparison of the size of coefficients across variables.
- Quantile local projections estimated via:
  - β̂τh ≡ argminβτh∈Rk Σt=1^T [ τ × I(ΔhYi,t+h,τ ≥ Xi,t βτh) × |ΔhYi,t+h,τ − Xi,t βτh| + (1 − τ) × I(ΔhYi,t+h,τ < Xi,t βτh) × |ΔhYi,t+h,τ − Xi,t βτh| ]
  - I(·) denotes the indicator function.
  - The estimated coefficient β̂τh measures the effect of Xi,t on the τ-th quantile of the conditional distribution of ΔhYi,t+h,τ.

### Definition and interpretation of CaR
- CaR values capture downside risks to future CRE price growth.
- CaR on the τ-th quantile in country i over the next h periods is defined as the τ-th quantile of ΔhYi,t+h conditional on Xi,t:
  - CaRi,τh ≡ Qτ(ΔhYi,t+h | Xi,t) = Xi,t β̂τh
- Both β̂τh and CaRi,τh depend on the quantile τ and the horizon h.
- The analysis mainly focuses on the 5th percentile and the 4-quarter horizon.

### Post-estimation: smoothing and conditional distribution
- The estimated quantile distribution is smoothed every quarter by interpolating between the estimated quantiles using kernel density estimation.
- This transforms the empirical quantile distribution into an estimated conditional distribution of CRE price growth.

### Key technical references cited in the methodology
- Canay, I., 2011. “A Simple Approach to Quantile Regression for Panel Data.” The Econometrics Journal 14(3), pp. 368-386.
- Deghi, A., J. Mok, and T. Tsuruga, 2021. “Commercial Real Estate and Macrofinancial Stability During COVID-19.“ IMF Working Paper, WP/21/264.
- IMF 2021. “Commercial Real Estate – Financial Stability Risk during the COVID-19 Crisis and Beyond.” Global Financial Stability Report.
- (Additional references listed in the source material.)

*Source: Annex describing the methodology used in the CaR analysis (text provided).*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1nldea2024010.pdf_