## Content unit 1luxea2024006

## Source details

**Canonical URL:** [Content unit 1luxea2024006](https://www.imf.org/-/media/files/publications/cr/2024/english/1luxea2024006.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2024/english/1luxea2024006.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2024/english/1luxea2024006.pdf.json)

---

### Executive summary — vulnerabilities and recent developments
- Strong policy support and high financial buffers helped the financial sector weather consecutive shocks, but pre-pandemic vulnerabilities continued to rise.
- Ultra loose financial conditions, in part as a consequence of ECB’s monetary policy, contributed to increased households’ indebtedness and stretched asset prices.
- Real estate prices grew rapidly over 2018–22 with signs of overvaluation.
- Over 2016–22, house prices increased by 78 percent.
- Following the tightening of monetary policy:
  - GDP has dropped by 1.1 percent.
  - Unemployment has risen to 5.5 percent in December 2023.
  - Credit growth turned negative for households and NFCs.
  - New mortgages plummeted.
  - House prices initiated an orderly correction, declining year on year by about 14½ percent in 2023Q4.
  - The CRE market is in the midst of a downturn.
- Banking system metrics:
  - CET1 at 23 percent.
  - LCR above 150 percent.
- Profitability has risen sharply benefiting from higher interest rates; asset quality has deteriorated (especially some segments) albeit from low levels; funding costs are rising.

### Household and residential real estate (RRE) vulnerabilities
- House price and indebtedness:
  - Prices grew by 78 percent over 2016–22.
  - Household debt to disposable income exceeded 180 percent in 2022.
- Mortgage risk profile deterioration over 2018H2–2022H2:
  - 60 percent of the new mortgages granted over 2018H2–2022H2 have a debt service-to-income (DSTI) higher than 40 percent.
  - More than 50 percent have a DTI higher that 900 percent.
- Lending standards and borrower metrics (as of end-2022):
  - about 60 percent of new mortgages have a DSTI higher than 40 percent and a DTI exceeding 900 percent.
  - about 45 percent have an LTV higher than 80 percent.
  - Share of outstanding mortgages with variable rate is at 40 percent.
- Regulatory and supervisory actions:
  - CSSF required banks to perform an interest rate sensitivity analysis of 200 basis points at loan origination; requirement contributed to reduced share of variable rate new mortgages.
  - CSSF required a 200 bp interest rate stress test at loan assessment end-2022.
- Recent developments and effects:
  - New government measures (tax bracket indexation equivalent to 4 indexation tranches; housing demand support measures) and automatic wage indexation could bolster resilience short term but could impede price adjustment and feed moral hazard.
  - House prices dropped by 14½ percent year-on-year in 2023Q4 with heterogeneity across segments.
  - Share of households with DSTI >40% increases from 17 percent in 2023 to 38 percent in 2026 in the baseline stress test.
  - Average (indicative) probability of default (PD) rises from 5 percent to 7.7 percent in the baseline.
  - In the adverse scenario: share of households at risk increases to 14 percent; debt at risk increases to 30 percent.
- Household-level stress test key assumptions:
  - Data: LU-HFCS wave 4 (2021).
  - Mortgage rates in the adverse scenario: 8.8 percent in 2024, 7.7    percent in 2025 and 5.4 percent in    2026.
  - Adverse scenario mortgages adjusted for 40 percent (current share of variable rate mortgages) and for another 25 percent per year corresponding to current share of new mortgages reaching reset.
  - Micro-simulations include fiscal measures: Energiedësch package, Solidarity packages (notably inflation adjustment of tax brackets of income tax: 4 indexation tranches in January 2024), increased caps for interest payment deductibility.
- Distributional findings:
  - PD and share of households at risk higher for bottom quintiles: one out of five households in bottom quintile has a positive PD.
  - In adverse scenario, share of households at risk in fourth and fifth quintiles increases to 11 and 6 percent respectively.
  - Contribution of the fourth and fifth quintiles to exposure at default is close to 50 percent; excluding concentrated risky exposures, upper quintiles’ contribution is 31 percent.
  - More affluent cohorts concentrated in recent risky mortgages but their high net wealth would reduce potential losses.

### Corporate sector and CRE vulnerabilities
- Aggregate and composition:
  - NFCs debt expanded from 341 percent of GDP in 2011 to 382 percent of GDP in 2022.
  - More than 55 percent of NFC debt is intra-sectoral lending, debt owed to captive financial institutions, and credit by nonresidents.
  - Debt vis-à-vis domestic financial institutions (excluding OFIs) is 57 billion (74 percent of GDP); half of these loans are granted by domestic banks.
- Firm-level financial positions:
  - Median debt to asset ratio has continuously dropped; interest coverage ratio (ICR) improved but close to 40 percent of firms report ICR below 1.
  - One out of five firms faces both liquidity and solvency challenges.
  - Real estate activities and hospitality sectors show weakest conditions (high indebtedness, low liquidity buffers; more than 40 percent of firms in hospitality with negative equity).
- Recent strains:
  - Cash buffers dropped 14 percent in September 2023.
  - Credit quality deteriorated especially for SMEs and CRE-collateralized loans.
  - Bankruptcies increased by close to 40 percent for construction and real estate activities (from low levels).
  - Banks tightened credit standards and increased loan rejection rates.
- NFC stress-test framework and results:
  - Method: firm-level panel regressions of ROA with firm controls and macro variables; dynamic projections of EBI, interest expenses, cash balances, debt and equity under scenarios; dataset STATEC central balance sheet office 2011–2022.
  - Adverse scenario outcomes:
    - Share of firms with liquidity shortfall exceeds 50 percent.
    - Debt at risk doubles compared to baseline to reach 40 percent.
    - Share of firms with negative equity rises to 37 percent and debt at risk to 25 percent.
  - Caveats: results conservative because cash excludes receivables; large firms may be underrepresented.

- CRE market specifics and bank exposures:
  - Office sector prices estimated to have fallen by -21 percent (private sector estimates).
  - Vacancy rate remains at historically low level; demand for climate-friendly buildings will support medium-term demand.
  - CRE loans share in Luxembourgish banks’ loan portfolios: relatively low at 6-7 percent, with significant heterogeneity across banks.
  - Weighted average LTV on outstanding CRE loans averages 54 percent; leverage on real estate developers at origination ranges from 65 to 80 percent.
  - Coverage ratio is high at 94.5 percent overall (98 percent for domestically oriented banks), though may not fully reflect recent valuation effects.
- CRE exposures outside banks:
  - Real estate funds’ direct investment reached EUR335 billion in 2023H1, of which 3.3 billion invested in the Luxembourgish CRE market.
  - Funds’ leverage moderate (around 130 percent); liquidity risks appear manageable for domestically active funds.
  - ESMA reports subset of funds with significant liquidity mismatch (1-year shortage of 9 percent); calculation does not incorporate redemption deferrals.
  - Insurance sector holds 2.2 billion in real estate assets (2.7 percent of total assets), of which 0.96 billion invested in Luxembourgish market.

### Institutional framework, governance, and operational issues
- CdRS composition and decision-making:
  - Chaired by the Minister of Finance (MoF); members: BCL (secretariat), CSSF, and CAA.
  - Decisions require a unanimous vote.
  - Except MoF, all members have a clear financial stability mandate.
- Identified institutional weaknesses contributing to inaction bias:
  - Potential influence of affordability concerns and de facto veto power of MoF due to unanimity requirement.
  - Lengthy legislative process for introduction of BBMs (more than 2 years).
  - High uncertainty around indicators (large revisions) and difficulty assessing macroprudential policy impact increased weight of judgement, contributing to limited or delayed action.
  - No systematic communications on systemic risks and macroprudential policy decisions for the general public beyond low-frequency publications.
- Legal constraints on BBMs:
  - State Council amendments imposed corridors (floors and caps) for BBMs and sequencing requirements; corridors specified include:
    - LTV maximum limit: 75 percent to 100 percent.
    - LTI maximum limit: 400 percent to 1200 percent.
    - DTI maximum limit: 400 to 1200 percent.
    - DSTI maximum limit: 35 to 75%.
    - Initial maturity maximum limit: 20 to 35 years.
  - Law requires that CRR/CRD tools be ineffective to address identified systemic risk before activating BBMs.
- Recommendations to reduce inaction bias and strengthen governance (high-level):
  - Uphold primacy of financial stability objective in CdRS by reducing role of MoF in macroprudential decisions (options: lower threshold to three out of four votes; designate MoF as non-voting member; transfer chairmanship to BCL).
  - Strengthen communication and accountability: systematically communicate rationale for macroprudential decisions, including inaction; publish records/summaries; adapt communications for wider audience.
  - Enhance operational agility: consider special fast procedures for macroprudential laws; use CSSF semi-hard powers when legislative delays occur; reassess BBM corridors and review law where legally feasible.
  - Enhance coordination with housing and other government policies; consider establishing a consultation mechanism with CdRS.

### Systemic risk monitoring, data gaps, and recommended data actions
- Progress and data initiatives:
  - Harmonized semi-annual data on LTV and DSTI collected per CSSF circulars N°18/703 and N°20/737.
  - AnaCredit transposed via BCL circular 2017/240 provides granular loan-level information; CSSF has faced challenges accessing AnaCredit data.
  - STATEC working on a CRE price indicator and participates in Eurostat Task Force on Commercial Real Estate Indicators.
  - CSSF launched semi-annual CRE loan data collection via circular N° 22/803 in 2022 and a CRE data collection for funds in September 2021.
- Key data gaps and recommended actions (Recommendation 54 summary):
  - Expedite implementation of a household credit register with clearly defined timelines.
  - Improve information on housing supply: expedite plans to collect data on vacant dwellings/rental stock; construct land price index and increase frequency of completed buildings data.
  - Improve micro data on households’ balance sheets: consider increasing HFCS coverage and adapt questionnaire regularly.
  - Improve corporate liabilities data quality: better isolate special purpose entities (SPEs) and improve reporting on balance sheet and P&L items in the central balance sheet office database.
- Monitoring and modeling enhancements:
  - BCL developed stock-flow model to assess housing supply-demand gap and econometric models for bankruptcies and NPLs by sector.
  - Stress-testing capacities: BCL MVAR top-down stress test for solvency; CSSF top-down sensitivity analysis and liquidity stress-testing frameworks for funds and banks; network analysis and bank–fund interconnectedness stress testing (LAREX).

### Policy tools, sequencing, and calibration recommendations
- Recommended sequencing:
  - Short term: preserve resilience against real estate vulnerabilities through targeted capital-based measures.
  - Then address structural indebtedness early in the recovery cycle through borrower-based measures (BBMs).
- Capital-based measures:
  - Use capital headroom to increase macroprudential requirements through targeted capital-based measures to real estate exposures, preferably a sectoral systemic risk buffer (SSyRB) for precise targeting and limited effects on credit supply.
- Borrower-based measures and calibrations:
  - Introduce income-based measures early in the recovery cycle with immediate preparations on calibrations and targeting.
  - FSAP analysis suggests:
    - Introducing a stressed-DSTI around 45-50 percent, possibly tied to the current interest rate stress test required by CSSF for new mortgages.
    - Consider gradually reducing the maximum LTV limit of 100 percent.
  - Additional guidance:
    - Apply a gradual approach with a speed limit to give flexibility to banks and tighten as recovery firms up.
    - Evidence suggests nonlinear complementarity between DSTI and LTV limits in reducing probability of default.
- CCyB framework:
  - Evaluate benefits of introducing a positive neutral CCyB; by keeping CCyB at 0.5 percent since 2019 CdRS/CSSF are de facto following that logic.
  - Recommendation: modify CCyB framework if introduction of positive neutral buffer deemed beneficial.
- Risks from housing demand support:
  - Untargeted tax incentives and buy-to-let measures risk impeding orderly market rebalancing and could encourage moral hazard and risk-taking; recommended to redesign help-to-buy policies over the medium term and consider frontloading public projects.

### Selected cross-country BBM calibrations (reported figures)
- Loan-to-value (LTV) examples:
  - Austria: 90% (20% speed limit).
  - Belgium: 90% for BTL (with 10% allowed up to 90%); 90% for FTB (with 35% allowed up to 100% and 5% over 100%); 90% for SSB (with 20% allowed up to 100%).
  - Canada: 95% for owner occupied with government guarantee up to 500,000; 80% for others.
  - Cyprus: 80% for FTB, 50% for construction of luxury properties, 70% for others.
  - Luxembourg: 100% for FTB, 90% for owner occupied (15% speed limit), 80% for others.
  - Netherlands: 100%.
  - Norway: 60% for second homes in Oslo, 85% for others.
  - Sweden: 85%.
  - United Kingdom: 90% for own and permanent residence, 80% for others.
- Debt service to income (DSTI) or stressed DSTI examples:
  - Austria: 40% (10% speed limit).
  - Belgium: 50%.
  - Czechia: 80%.
  - Denmark: 50% for those under age 36, 45% for others.
  - Finland: 50% stressed DSTI.
  - France: 35%.
  - Iceland: 40% stressed DSTI for FTB; 35% for others (5% speed limit).
  - Luxembourg (reported in table): 13-36% depending on income and interest rate.
  - Netherlands: 50% (with 10% allowed up to 60%).
  - New Zealand: 60% (with 5% allowed up to 70%).
  - United Kingdom: 40% unstressed; 50% stressed DSTI; 5% allowed up to 60%.
- Loan-to-income (LTI) examples:
  - Austria: LTI above 4 requires stress testing.
  - Belgium: 3.5 for FTB (20% speed limit); 3.5 for SSBs (20% speed limit).
  - Sweden: 4.5 times (15% speed limit).

### Summary of prioritized macroprudential recommendations (extract)
- A. Institutional Arrangements
  - 1. Reduce the role of the MoF in macroprudential policy decisions. Agency: Government. Time: MT.
  - 2. Strengthen communication and accountability by systematically communicating on macroprudential decisions, including in case of inaction, and enhancing accessibility to the general public. Agency: CdRS and its members. Time: I.
  - 3. Enhance coordination with housing and other government policies that have financial stability implications and consider establishing a consultation mechanism with the CdRS. Agency: Government/CdRS. Time: NT.
- B. Systemic Risk Monitoring and Articulation with Policy Decisions
  - 4. Continue to improve the quality and availability of data, especially in housing supply and households and corporate sector balance sheet. Agency: STATEC/BCL. Time: MT.
  - 5. Ensure a better mapping between objectives, indicators and instruments especially related to households’ indebtedness and real estate vulnerabilities. Agency: CdRS. Time: NT.
- C. Tools and Calibration
  - 6. Increase macroprudential capital requirements to preserve resilience against vulnerabilities in the real estate sector preferably through a sectoral systemic risk buffer. Agency: CdRS. Time: I.
  - 7. Activate income-based macroprudential measures—such as stressed DSTI—early in the recovery cycle and consider gradually reducing the maximum-LTV ratio from 100 percent. Agency: CdRS. Time: Early in the recovery cycle.

*Source: EXECUTIVE SUMMARY and Chapter 8 (Content unit 1luxea2024006) — IMF staff technical analysis.*

### EXECUTIVE SUMMARY

### EXECUTIVE SUMMARY

### Key findings on vulnerabilities and recent developments
- Strong policy support and high financial buffers have helped the financial sector weather consecutive shocks, but pre-pandemic vulnerabilities have continued to rise.
- Ultra loose financial conditions, in part as a consequence of ECB’s monetary policy, contributed to increased households’ indebtedness and stretched asset prices.
- Real estate prices grew rapidly over 2018–22 with signs of overvaluation.
- Households’ indebtedness continued to rise, although partly mitigated by high households’ net wealth.
- The average LTV has dropped but the impact on DSTI and DTI has been more muted.
- Following the tightening of monetary policy:
  - GDP has dropped by 1.1 percent.
  - Unemployment has risen to 5.5 percent in December 2023.
  - Credit growth turned negative for households and NFCs.
  - New mortgages plummeted.
  - House prices initiated an orderly correction, declining year on year by about 14½ percent in 2023Q4.
  - The CRE market is in the midst of a downturn.
- With a CET1 at 23 percent and a LCR above 150 percent, Luxembourg’s banking system capital and liquidity remained well above minimum requirements.
- Profitability has risen sharply recently, benefiting from higher interest rates; asset quality has deteriorated (especially for some segments), albeit from low levels, and funding costs are rising.

### Household and residential real estate (RRE) vulnerabilities
- Over 2016–22, house prices increased by 78 percent.
- Households’ debt to disposable income exceeded 180 percent.
- Risk profile of mortgages deteriorated over 2018H2–2022H2:
  - 60 percent of the new mortgages granted over 2018H2–2022H2 have a debt service-to-income (DSTI) higher than 40 percent.
  - More than 50 percent have a DTI higher that 900 percent.
- CSSF required banks to perform an interest rate sensitivity analysis of 200 basis points at loan origination, which helped reduce the share of variable rate new mortgages.
- New government measures to stimulate housing demand and purchasing power, together with automatic wage indexation, could bolster resilience in the short term but could impede price adjustment, feed moral hazard, and put debt on an upward trend with potential unintended distributional and financial stability consequences.

### Corporate and CRE vulnerabilities
- NFCs and CRE exhibit heightened vulnerabilities amid higher interest rates and slowing activity.
- The CRE market is experiencing a downturn and prompted stepped-up monitoring (regular data collection, thematic surveys, industry dialogue).
- Stock vulnerabilities stem mainly from real estate exposures, which are concentrated in few banks.
- NPLs have been recently increasing for both households and corporates, albeit from low levels.

### Institutional framework and operational issues
- Authorities extended macroprudential tools to include borrower-based measures (BBMs).
- BCL and CSSF have advanced systemic risk monitoring with a focus on real estate and non-bank financial institutions and made substantial progress filling data gaps.
- Institutional weaknesses remain that may contribute to inaction bias:
  - Potential influence of affordability concerns and a de facto veto power of the Ministry of Finance (MoF) due to unanimous vote requirement in the systemic risk committee (CdRS).
  - Lengthy legislative process for introduction of BBMs constrained timely action.
  - High uncertainty around indicators (large revisions) and difficulty assessing macroprudential policy impact increased the weight of judgement and contributed to limited or delayed action.
- Recommendations to reduce inaction bias include:
  - Uphold the primacy of the financial stability objective in CdRS decisions by reducing the role of the MoF in macroprudential decisions; potential institutional changes include revoking unanimity by lowering threshold to three out of four votes or designating MoF as a non-voting member, and transferring chairmanship of CdRS to the central bank (BCL).
  - Strengthen communication on macroprudential policy decisions, including systematic communication in case of inaction and enhancing public accessibility.
  - Strengthen operational agility for macroprudential authorities outside the CRR/CRD (including for BBMs), regularly reassess constraints on ability to act, review law as needed, avoid delays in legislative processes, and use CSSF semi-hard powers where legally feasible.
  - Enhance coordination with housing policies and other government policies that have financial stability implications; consider establishing a consultation mechanism with CdRS.

### Systemic risk monitoring and data gaps
- Authorities should continue to improve availability and quality of data:
  - Expedite plans to establish a household credit register.
  - Improve information on housing supply.
  - Improve quality and coverage of corporate balance sheets, better identify special purpose entities, and analyze their role and linkages with investment funds.
- Improve mapping of objectives, indicators, and instruments behind policy decisions; decisions not to act should be sufficiently motivated while keeping scope for judgement.

### Policy tools and calibration (recommended sequencing and calibration)
- Short term: preserve resilience against real estate vulnerabilities through targeted capital-based measures, then address structural indebtedness early in the recovery cycle through borrower-based measures.
- Use capital headroom to increase macroprudential requirements through targeted capital-based measures to real estate exposures, preferably a sectoral systemic risk buffer (SSyRB) for precise targeting and limited effects on credit supply.
- Introduce income-based measures early in the recovery cycle with immediate preparations on calibrations and targeting. FSAP analysis suggests:
  - Introducing a stressed-DSTI around 45-50 percent, possibly tied to the current interest rate stress test required by CSSF for new mortgages.
  - Consider gradually reducing the maximum LTV limit of 100 percent.

### Table of key macroprudential recommendations (summary)
- A. Institutional Arrangements
  - 1. Reduce the role of the MoF in macroprudential policy decisions. Agency: Government. Time: MT.
  - 2. Strengthen communication and accountability by systematically communicating on macroprudential decisions, including in case of inaction, and enhancing accessibility to the general public. Agency: CdRS and its members. Time: I.
  - 3. Enhance coordination with housing and other government policies that have financial stability implications and consider establishing a consultation mechanism with the CdRS. Agency: Government/CdRS. Time: NT.
- B. Systemic Risk Monitoring and Articulation with Policy Decisions
  - 4. Continue to improve the quality and availability of data, especially in housing supply and households and corporate sector balance sheet. Agency: STATEC/BCL. Time: MT.
  - 5. Ensure a better mapping between objectives, indicators and instruments especially related to households’ indebtedness and real estate vulnerabilities. Agency: CdRS. Time: NT.
- C. Tools and Calibration
  - 6. Increase macroprudential capital requirements to preserve resilience against vulnerabilities in the real estate sector preferably through a sectoral systemic risk buffer. Agency: CdRS. Time: I.
  - 7. Activate income-based macroprudential measures—such as stressed DSTI—early in the recovery cycle and consider gradually reducing the maximum-LTV ratio from 100 percent. Agency: CdRS. Time: Early in the recovery cycle.

*Source: EXECUTIVE SUMMARY (1luxea2024006) — IMF technical note.*

### 8.      The authorities have taken measures to contain mounting vulnerabilities in the RRE

### 8.      The authorities have taken measures to contain mounting vulnerabilities in the RRE sector

### Policy measures taken to date
- 2016: CSSF increased the minimum risk weights on residential real estate exposures for internal ratings-based (IRB) banks to 15 percent.
- 2019: CCyB has been gradually raised to 0.5 percent.
- November 2019 / December 2020:
  - Adoption of a legal framework for borrower-based limits and issuance of Regulation N°20-08 introducing differentiated LTV limits for mortgage loans (80% LTV for RRE in LU, exceptions: 100% for first-time buyers (FTBs); 90% for non-FTBs acquiring primary residence with up to 15 percent of annual new mortgage portfolio allowed between 90–100 percent LTV).
- 2021: Implementation of legally binding differentiated LTV limit for mortgage loans and fiscal measures to reduce speculative demand (reduction of the accelerated depreciation rate; new special 20 percent tax—“prélèvement immobilier”—on all income derived by tax-opaque Luxembourg Funds from domestic real estate assets; announcement of a tax reform to lands and empty dwellings).
- End-2022: CSSF required banks to perform sensitivity analysis including a debt servicing test buffer of 2 percentage points increase in rates at loan origination drawing on the EU Mortgage Credit Directive principles.
- Resulting impacts noted: bolstered banks’ resilience; mixed impact on mortgage risk profile (average LTV dropped; stress test requirement contributed to reduced share of variable loans; DSTI continued to increase).

### Government measures responding to the recent downturn
- Cost-of-living support packages implemented by former government complement automatic stabilizers (automatic wage indexation, social safety nets).
- Newly elected government measures include:
  - Retroactive adjustment of personal income tax brackets (equivalent to 4 indexation tranches).
  - Package to support housing demand: purchase of new affordable residential projects; temporary tax incentives mostly directed to buy-to-let segment (increase in accelerated depreciation rate, increase in cap of VAT payment deductions, tax credit on registration tax for acquisition of buy-to-let dwellings).
  - Announced structural measures to support affordability and housing supply (public housing supply, reduced red tape).

### Assessment of credit and macrofinancial risks
- Credit to the nonfinancial private sector growth turned negative: -3 percent year on year in November 2023.
- Drivers: higher interest rates from monetary policy tightening, uncertain economic outlook, substantial decline in demand for loans, and banks tightening credit standards.
- Household credit: mortgages contracted due to lower housing demand and higher bank rejection rates.
- NFCs: loans growth turned negative as firms postpone capital expenditures and rely on internal funding.
- Latest indicators suggest tightening cycle close to end with forward-looking indicators pointing to positive dynamics, especially for households.
- CdRS decision: CCyB kept unchanged (0.5 percent) despite negative credit to GDP gap, motivated by uncertainty and need to bolster banks’ loss absorptive capacity—de facto a positive neutral CCyB.

### Household and housing market vulnerabilities
- House price growth:
  - Prices grew by 78 percent over 2016–22.
  - Significant overvaluation signaled by statistical and model-based indicators, especially since 2019, with significant uncertainty around estimates.
- Household indebtedness and leverage:
  - Household debt to disposable income ratio exceeded 180 percent in 2022.
  - Share of outstanding mortgages with variable rate is at 40 percent.
  - Debt service-to-income has been rising even as interest rates dropped due to reduced maturities.
- Asset position:
  - About 70 percent of households are homeowners.
  - Financial assets are more than twice the value of financial liabilities.
  - Asset gains reversed in 2022 and 2023; most assets are illiquid or market-sensitive and unevenly distributed across households.
- Lending standards and borrower metrics:
  - 2018–2020: LTV, DTI and maturities loosened across income levels.
  - CdRS recommended legally binding differentiated LTV limit in November 2020, reducing average LTV by 3 percentage points within a year.
  - LTV declined further in 2022–23 due to banks tightening lending standards.
  - As of end-2022:
    - about 60 percent of new mortgages have a DSTI higher than 40 percent and a DTI exceeding 900 percent,
    - about 45 percent have an LTV higher than 80 percent.
  - CSSF required a 200 bp interest rate stress test at loan assessment, helping reduce appetite for variable rate mortgages.
- Recent rate spike effects:
  - House prices dropped by 14½ percent year-on-year in 2023Q4 (with heterogeneity across segments).
  - Adjustment incomplete; price-at-risk estimates suggest further downward pressures, moderated by lower supply and government support.
  - Households’ debt servicing capacity decreasing despite automatic wage indexation, government support measures, and excess savings.
  - Early signs of asset quality deterioration: rising NPLs and loan renegotiations (from low levels).

### Household-level stress tests and scenario assumptions
- Data and methodology:
  - Use of LU-HFCS wave 4 (2021) household-level data.
  - Definition of “financially vulnerable households” follows Ziegelmeyer and Giordana (2019): probability of default based on financial margin (FM) and availability of liquid financial assets (LIQ), with M set to 3 months.
- Micro-level shock assumptions:
  - Mortgages and interest rate shock:
    - Population of indebted households in 2022–2026 assumed to share 2021 demographic and socio-economic characteristics.
    - Average mortgage rates in the adverse scenario set at EA long-term interest rate plus a spread of 200 bp (100bps for the baseline).
    - The spread distributed uniformly across households by quantiles of debt-to-income.
    - Mortgages adjusted for 40 percent (current share of variable rate mortgages) and for another 25 percent per year corresponding to current share of new mortgages (including adjustable-rate mortgage loans reaching reset).
    - For other loans, interest rate moves with euro area short-term rate.
    - The mortgage rates in the adverse scenario are as follows: 8.8 percent in 2024, 7.7    percent in 2025 and 5.4 percent in    2026.
  - Unemployment shock/benefits:
    - Logistic model links employment status to individual and household characteristics to generate stochastic draws matching population-level unemployment path and persistence.
    - Simulated unemployed income mimics unemployment benefits; pre-existing unemployed are assumed to receive no new benefits (long-term unemployed) and if income below minimum subsistence, household receives Social Inclusion Income (“REVIS”).
  - Income shock and fiscal support:
    - Compensation per employee steers disposable income path; automatic wage indexation assumed to hold but real labor income drops moderately in adverse scenario.
    - Micro-simulations include fiscal measures: Energiedësch package, Solidarity packages (notably inflation adjustment of tax brackets of income tax: 4 indexation tranches in January 2024), increased caps for interest payment deductibility.
  - Consumption and savings:
    - Consumption assumed constant in real terms; households with lower propensity to consume accumulate savings and increase capacity to service debt.

### Stress test results and key projections
- Baseline scenario:
  - Share of households with DSTI higher than 40 percent increases from 17 percent in 2023 to 38 percent in 2026.
  - Average (indicative) probability of default (PD) rises from 5 percent to 7.7 percent.
  - Automatic wage indexation and fiscal support help cushion lower-propensity-to-consume households.
- Adverse scenario:
  - Share of households at risk increases to 14 percent.
  - Debt at risk (debt with a positive indicator of default probability) increases to 30 percent.
  - Automatic stabilizers (unemployment benefits, social safety nets, interest deductibility) materially absorb unemployment/interest rate shocks.

*Source: Content unit 1luxea2024006 — Chapter 8, as provided.*

### 21.      While the lower income households are the most vulnerable, the more affluent

### 21.      While the lower income households are the most vulnerable, the more affluent 

### Household-level vulnerabilities and distributional findings
- The “PD” and the share of households at risk are higher for the bottom quintiles, with one out of five households in this cohort having a positive “PD.”
- In the adverse scenario, more affluent households are relatively more affected (higher multiple “PD”).
- The share of households at risk in the fourth and fifth quintiles increases respectively to 11 and 6 percent in the adverse scenario.
- The contribution of the fourth and fifth quintiles to the exposure at default is close to 50 percent.
- Excluding concentrated risky exposures (that could be outliers), the contribution of the upper quintiles to the exposure at default is 31 percent.
- Higher credit risk is concentrated on recent mortgages that were mostly dominated by the higher income cohorts; however, the high net wealth of these cohorts would reduce potential losses in case of default.

### Banks’ potential losses and concentration risks
- A sharp correction in house prices combined with a 20 percent haircut linked to transaction costs could increase banks losses assuming they do not wait to sell their collateral.
- This risk is compounded by a significant concentration of loans to households (and mortgages) in few domestically oriented banks.
- Overvalued house prices could expose banks to material, though manageable, losses.

### Caveats and limitations of household analysis
- The HFCS data sample is relatively small and granular analysis at the income quintile levels are subject to uncertainty.
- HFCS wave IV was conducted during the pandemic and some critical information (such as the type of interest fixation) was not collected, requiring simplifying assumptions.
- Behavioral responses are not considered (for example, banks could preemptively restructure loans to ease household payment burdens).
- Despite caveats, results suggest high and rising household indebtedness is a growing concern for middle income and high-income households, and that in case of shocks households may need to cut consumption especially where DSTI increases materially.

### Nonfinancial corporate (NFC) debt and composition
- NFCs debt expanded from 341 percent of GDP in 2011 percent to 382 percent of GDP in 2022.
- More than 55 percent of NFC debt is intra-sectoral lending, debt owed to captive financial institutions, and credit by nonresidents.
- Focusing on debt vis-à-vis domestic financial institutions (excluding OFIs), the debt is 57 billion (74 percent of GDP).
- Half of these loans are granted by domestic banks.
- Since 2018, the debt owed to non-money market funds (non-MMF) has risen substantially; this might reflect securitization of bank loans to NFCs rather than direct market financing.

### Corporate financial positions and sectoral heterogeneity
- Median debt to asset ratio has continuously dropped; higher profitability and lower interest rates enhanced the interest coverage ratio (ICR).
- Close to 40 percent of firms report an interest coverage ratio below 1.
- One out of five firms faces both liquidity and solvency challenges.
- Real estate activities stand out as having the weakest conditions: high indebtedness and low liquidity buffers.
- The hospitality sector has high debt and low ICR, with more than 40 percent of firms with negative equity, although firms have accumulated significant cash buffers in recent years.

### Recent developments straining firms
- Cash buffers dropped 14 percent in September 2023.
- Credit quality deteriorated especially for SMEs and CRE-collateralized loans.
- Manufacturing, construction, and real estate activities have seen a sharp deterioration in the NPL ratio.
- Bankruptcies increased by close to 40 percent for construction and real estate activities (from low levels).
- Banks’ credit standards have tightened and the loans rejection rate increased.
- The government has set up dedicated support mechanisms to help construction firms overcoming temporary financial difficulties should the situation further deteriorate.

### Stress-testing methodology for NFCs
- Framework based on firm-level panel regressions of profitability (ROA) with firm-level controls (size, leverage, tangibility) and macro variables (real GDP growth, short term interest rate, house price growth, inflation, oil price growth).
- Two strategies: panel regression for all firms with industry fixed effects, and sectoral level regressions to capture sector-specific shocks.
- ROA is defined as the ratio of earning before interests (EBI) expenses to total assets (taxes and depreciation not reported).
- Dynamic projection of ROA; EBI derived assuming total assets remain constant.
- Interest expenses(t) = Total debt (t-1) * effective interest rate (t); effective interest rate evolution mimics euro area short term rate plus an additional risk premium in the adverse scenario (170 basis point ± 50 bp depending on firm level indebtedness).
- Cash Balance(t+1) = EBI (t+1) – Interest Expenses (t+1) + Cash Balance(t); negative projected cash balance indicates liquidity shortage and need to rely on external credit.
- Debt(t+1) = Debt(t) - Cash Balance(t+1) if projected cash balance is negative; = Debt(t) otherwise.
- Interest Coverage Ratio(t) = EBI(t)/Interest Expenses(t).
- Equity(t+1) = Total assets (t+1) - Total debt (t+1) assuming no new equity and no dividends during analysis years.
- Firms at risk defined as firms with cash balances or equity below zero; debt at risk is debt of firms at risk as a fraction of total NFC debt.
- Dataset: STATEC central balance sheet office covering 10 industries annually over 2011–2022; potential SPEs cleaned using criteria (exclude NACE K and M701; financial assets share ≤ 90 percent; debt owed to affiliated companies ≤ 75 percent; SOPARFI tax indicators used).

### NFC stress-test results and projections
- Under the baseline, the share of firms with negative cash balance increases to 35 percent compared to 20 percent during the pandemic; associated debt at risk remains contained.
- Under the adverse scenario:
  - The share of firms with liquidity shortfall exceeds 50 percent.
  - Debt at risk doubles compared to the baseline to reach 40 percent.
  - Impact on solvency: share of firms with negative equity rises to 37 percent and debt at risk to 25 percent.
- Liquidity risk increases significantly under the adverse scenario.
- Interpretation caveats:
  - Analysis does not consider natural hedges such as cuts in other expenses.
  - Cash refers to currency and deposits and does not include account receivables and other cash equivalents, possibly making results conservative.
  - Large firms may be underrepresented in the sample.

### Sector-specific shocks and vulnerabilities
- Construction (NACE F) and real estate activities (NACE L) account for 62 percent of NFC loans granted by domestically oriented banks and would be more affected by a sharp house/real estate price correction.
- Real estate activities have much weaker initial conditions—higher leverage and lower liquidity buffers—making them more exposed to liquidity and solvency risk even under the baseline.
- Real estate developers often create SPVs with high leverage and land as collateral; parent companies may provide support but also have incentives to transfer liquidity risk to banks or let SPVs go bankrupt if conditions worsen.
- Construction sector’s pressure on debt-servicing partly absorbed by high cash buffers accumulated in recent years.

### CRE market developments and bank exposures
- Office sector prices estimated to have fallen by -21 percent based on private sector estimates.
- Vacancy rate remains at historically low level; demand for climate-friendly buildings will support medium-term demand.
- Share of CRE loans in Luxembourgish banks’ loan portfolios is relatively low at 6-7 percent of total loans, but heterogeneity across banks is significant.
- Weighted average LTV on outstanding CRE loans averages 54 percent; leverage on real estate developers at origination ranges from 65 to 80 percent.
- Coverage ratio is high at 94.5 percent overall (98 percent for domestically oriented banks) though this may not fully reflect valuation effects of recent price drops.

### CRE exposures outside banks and mitigating factors
- Real estate funds’ direct investment reached EUR335 billion in 2023H1, of which 3.3 billion invested in the Luxembourgish CRE market.
- Domestic investments concentrated in the office market and prime locations.
- Real estate funds’ leverage is moderate (around 130 percent); liquidity risks appear manageable for funds active in domestic market.
- Mitigants: predominance of institutional investors, many funds are closed-ended, and widespread use of redemption deferrals as a liquidity management tool.
- ESMA reports a subset of funds with a significant liquidity mismatch (1-year shortage of 9 percent); this calculation does not incorporate redemption deferrals.
- Insurance sector holds 2.2 billion in real estate assets (2.7 percent of total assets), of which 0.96 billion are invested in the Luxembourgish market.

*Source: IMF staff analysis and STATEC/central balance sheet office data as presented in the chapter.*

### 36.      Robust institutional arrangements are essential for the effectiveness of

### 36.      Robust institutional arrangements are essential for the effectiveness of

### Willingness to Act
- The CdRS serves as the Luxembourg macroprudential authority, tasked to identify, monitor, and assess risks to financial stability, and issue opinions, warnings, and recommendations of corrective actions.
- CdRS composition and governance:
  - Chaired by the Minister of Finance (MoF).
  - Members: BCL (also acting as the secretariat), CSSF, and CAA.
  - Except for MoF, all CdRS members have a clear financial stability mandate.
  - Decisions are taken based on a unanimous vote.
  - CdRS is accountable to the national parliament and is subject to a transparency framework.
- CRD IV implementation:
  - CSSF is the national designated authority; when acting in that capacity it consults the BCL to find a common position and, where applicable, requests the opinion of the CdRS or takes the latter’s recommendations into account.
- Assessment and past recommendations:
  - 2017 FSAP recommended: switch from unanimous to majority voting, enshrine in law the lead role of BCL, award BCL powers to issue recommendations, and publish periodic analysis of systemic risks including “notes de surveillance” and risk dashboards.
- Recent practice and issues:
  - Institutional framework remains unchanged; concerns about potential inaction bias persist.
  - Macroprudential measures focused on strengthening resilience but tended to come somewhat late and were only partially sufficient to tackle rising households’ indebtedness and risk-taking by borrowers and banks.
  - Legislation and implementation of borrower-based limits took more than four years.
  - Calibration: maximum LTV kept at 100 percent for first-time homebuyers in context of significant overvaluation—the highest among peers.
- Communications:
  - No systematic communications on systemic risks and macroprudential policy decisions for the general public.
  - Periodic publications exist: BCL financial stability report, CdRS and CSSF annual reports; these are low frequency.
  - CdRS issues systematic justifications for quarterly CCyB decisions, reciprocation of other countries’ measures, and annual G-/O-SII designations; other communications are not systematic and are technical.
  - Public misperceptions have occurred (e.g., LTV-maximum limits interpreted as actual LTV allowed by banks).
- Recommendations (Willingness to Act):
  - Reduce the role of MoF to uphold the primacy of the financial stability objective; options include:
    - Lowering the threshold to three out of four votes for CdRS decisions.
    - Designating MoF as a non-voting member.
    - Transferring CdRS chairmanship to the central bank (BCL).
  - Strengthen accountability via systematic communication on rationale for macroprudential decisions, including inaction.
  - Extend CdRS protocol on communicating CCyB and G-/O-SII buffers to all macroprudential decisions.
  - Publish records of CdRS meetings soon after they take place (excluding sensitive/confidential information); provide summaries of discussions and divergence in views.
  - Regularly publish the risk assessment (note de surveillance) and risk dashboards; increase use of warnings.
  - Adapt communications for a wider audience: press conferences, events, social media, non-technical summaries, chart packs, heatmaps.
  - Update macroprudential policy strategy periodically (strategy published in 2016 and not updated since).

### Ability to Act
- Legal and informational powers:
  - CdRS empowered to make opinions, recommendations, and warnings to its member institutions on tools entrusted to them by European or national legislation, with a “comply or explain” mechanism.
  - CdRS monitors implementation of recommendations and can take corrective actions if needed.
  - Organic Law gives CdRS wide-ranging powers to request information required to contain systemic financial sector risks.
  - Law of December 4, 2019 strengthened BCL access to aggregated information from public entities for CdRS analysis.
  - Information sharing has been relatively fluid, although not always systematic; no formal arrangements in place to ensure it.
- Challenges adding new tools:
  - Presence of MoF should facilitate adoption of new laws, yet introduction of borrower-based measures (BBMs) in 2019 revealed a lengthy (more than 2 years) and complex legislative process.
  - BCL in 2019: concerns that delays prevented authorities from having essential macroprudential tools for residential real estate market risks.
  - State Council (Conseil d’Etat) introduced amendments imposing corridors (floors and caps) to BBMs and sequencing for activation, potentially limiting flexibility and proportionality.
  - Corridors specified in law:
    - LTV maximum limit: 75 percent to 100 percent.
    - LTI maximum limit: 400 percent to 1200 percent.
    - DTI maximum limit: 400 to 1200 percent.
    - DSTI maximum limit: 35 to 75%.
    - Initial maturity maximum limit: 20 to 35 years.
  - Law requires that other CRR/CRD tools are ineffective to address identified systemic risk before activating BBMs.
- Recommendations (Ability to Act):
  - Enhance operational agility of macroprudential framework; reduce unnecessary hurdles and enable rapid adaptation.
  - Consider special (faster) procedures for macroprudential laws.
  - CSSF should deploy semi-hard and soft powers as needed when legislative delays occur.
  - Reassess whether BBM corridors constrain ability to act and review the law as needed where legally feasible.

### Effective Coordination and Cooperation
- Domestic coordination:
  - CdRS coordinates between members to contribute to financial stability through cross-fertilization of views and exchange of information.
  - Preparatory technical meetings support coordination; “notes de surveillance” benefit from inputs across institutions.
  - A working group formed to follow up on ESRB recommendation on CRE vulnerabilities.
  - Exceptional CdRS meetings were called during stress periods for rapid assessment and coordinated response.
- International and European coordination:
  - ESRB collects and communicates national macroprudential policy decisions and fosters mutual recognition to reduce leakages.
  - BCL and CSSF participate regularly in international fora and contribute to regulatory/policy debates and research.
- Recommendations (Coordination):
  - Enhance coordination between macroprudential policy and other policies, particularly housing policies.
  - Consider establishing a consultation mechanism between the government and the CdRS to incorporate financial stability considerations in the design/calibration of government policies (especially housing and fiscal policies).

### Systemic Risk Monitoring and Articulation of Policy Decisions — Data Issues
- Importance:
  - Monitoring and evaluating systemic vulnerabilities is crucial for timely calibration of macroprudential policy; assessment covers data availability, quantitative methods, and articulation between assessment and policy.
- Progress since 2017 FSAP:
  - Households and residential real estate:
    - Harmonized data on LTV and DSTI ratios and additional relevant indicators collected on semi-annual basis (CSSF circular N°18/703 and N°20/737), following ESRB recommendation ESRB/2016/14.
    - Data collection instrumental for assessing risk profile of new loans; more granularity (richer bucketing of risks) encouraged.
    - BCL uses triennial Household Finance and Consumption Survey (HFCS) to stress test households’ balance sheet.
  - Commercial real estate (CRE):
    - CSSF issued circulars for regular data collection from banks and investment funds.
    - STATEC working with Eurostat to develop a price indicator for CRE; faces technical hurdles due to lack of transactions and price discovery especially in downturns.
    - ESRB assessed Luxembourg as fully compliant with respect to CRE.
  - Non-financial corporations:
    - AnaCredit EU regulation 2016/867 transposed via BCL circular 2017/240 “introduction of granular credit risk”; database offers granular loan-level information.
    - CSSF has faced challenges accessing AnaCredit data.
  - Investment funds and other reporting:
    - New EU reporting for MMFs implemented in accordance with article 37 of regulation 2017/1131 of 14 June 2017.
    - Article 4 of Regulation 2015/2365 introduced EU-wide reporting on SFTs and reuse.
    - Efforts to improve data quality in prudential reporting: automating controls on UCITS Risk Reporting; implementing EU data quality engagement frameworks for AIFMD and MMFR; EU data quality action plans on EMIR and SFTR.
    - Some AIFMD reporting fields changed from optional to mandatory to increase coverage; CSSF engaged with ESMA to provide guidance on AIFMD risk indicators lacking standardization.
- Use of qualitative information:
  - Ad hoc surveys and dialogues with market participants complement data.
  - Example: CSSF survey of Luxembourg real estate developers’ portfolio covered domestic banks representing 92 percent of total Luxembourg real estate developers’ portfolio.
  - Purpose of thematic review: (i) overview risks from slowdown of real estate development project prices and increases in interest rates/construction costs, (ii) benchmark banks on lending standards and monitoring, (iii) propose remedial actions based on findings.

*Source: IMF staff assessment in the referenced chapter.*

### Box 1. Efforts in Filling CRE Data Gaps

### Box 1. Efforts in Filling CRE Data Gaps

### Progress on closing CRE data gaps
- The ESRB issued recommendations on closing real estate data gaps (ESRB/2016/14 & ESRB/2019/3) and a Recommendation on vulnerabilities in the commercial real estate sector in the EEA (ESRB/2022/9).
- Luxembourg (LU) was assessed by the ESRB as fully compliant with respect to commercial real estate following authorities' efforts.
- The CdRS established a working group on real estate data gaps (“Groupe de travail sur les données immobilières”) in 2019, chaired by the BCL, composed of representatives of the BCL, the CSSF, the CAA, the Ministry of Finance and STATEC, to coordinate and develop a harmonized set of real estate indicators for banks, funds, and insurers and a set of physical CRE market indicators.

### New data collections and scope
- Banks
  - In 2022, the CSSF published circular N° 22/803, launching a semi-annual data collection on bank loans aimed at acquiring a CRE property or secured by a CRE property.
  - Data collection designed after extensive discussions with main banks and industry representatives and to meet expectations of ECB, ESRB, etc.
  - CSSF can collect CRE exposure data with granularity on:
    - purpose of CRE loans,
    - property type associated to these loans,
    - location of the CRE property.
  - Collected indicators include LTV ratio (aggregate and by CRE purpose), interest coverage and debt service coverage ratios.
  - Non-performing loans for acquiring CRE and loan loss provisions on loans for acquiring CRE are collected with the same granularity.
  - Indicators are collected on both the stock of CRE loans and newly granted CRE loans.
- Funds
  - In September 2021, the CSSF introduced a data collection on commercial real estate for investment fund managers via an online questionnaire developed with funds industry professionals.
  - This gathers data on CRE direct and indirect domestic and foreign investment, by location (prime, non-prime), type (office, retail, etc.), and country where the building is located.
  - Indicators collected on both the stock of CRE loans and newly granted CRE loans.

### Physical CRE market indicators and challenges
- STATEC is compiling a CRE price index and participates in the Eurostat Task Force on Commercial Real Estate Indicators to develop physical CRE market indicators and a European-wide legal framework.
- Practical challenges for Luxembourg in constructing a CRE price index:
  - very low number of transactions per year,
  - difficult price measurement in multi-use properties,
  - definition of “prime” and “non-prime” real estate.
- STATEC engaged with national statistical institutes from other small CRE market jurisdictions and Eurostat to address challenges related to few market transactions in the new legal framework.
- STATEC expected a new European regulatory framework for the collection of physical CRE market indicators to enter into force by end-2025, followed by a first data collection in the first quarter of 2026.
- In the interim, STATEC signed an agreement with a private actor to obtain aggregated commercial real estate data.

### Key recommendations from the Box
- Recommendation 54: Continue improving the quality, granularity, and availability of data, especially on housing supply as well as households and corporate balance sheets.
  - STATEC and relevant data sources should fill data gaps on housing supply:
    - Data on completed buildings are available with a considerable lag (latest available is 2019).
    - Estimates of stock of residential dwellings are only available at very low frequency through census data.
    - Plans to collect data on vacant dwellings/rental stock should be expedited.
    - The construction of a land price index and collection of associated transactions data by the “observatoire de l’habitat” is welcome; higher frequency would be highly beneficial.
  - Micro data on households’ balance sheets should be improved:
    - Expedite implementation of a credit register for households with clearly defined timelines.
    - BCL should consider increasing the coverage of the HFCS (sample and response rate) and regularly adapting the questionnaire to emerging macroprudential questions.
  - Corporate sector liabilities data quality:
    - Data are subject to large revisions and high volatility partly due to large inter-company lending from special purpose entities and multinational companies.
    - STATEC should better isolate these entities in national and financial accounts.
    - Improve reporting quality on balance sheet and P&L items in the central balance sheet office database (e.g., taxes, turnover, granular components of debt).

### Quantitative methods, monitoring, and stress-testing enhancements
- The BCL and CSSF have continued to develop a comprehensive systemic risk monitoring framework in line with the 2017 FSAP recommendations, including strengthening monitoring of residential real estate, non-banks, investment fund–bank interconnectedness, and the corporate sector.
- Residential real estate and household vulnerability monitoring:
  - BCL and CSSF frameworks include analysis of house price growth, overvaluation indicators (statistical and model-based), house price-at-risk modeling, mortgage credit developments, household debt, domestic bank exposures and mortgage risk-weights.
  - BCL developed a stock-flow model to assess the gap between housing supply and demand.
  - Participation in SSM and ECB/ESRB working groups and BCL participation in the BIS CGFS Study Group on housing policies.
  - Current house price models are largely partial equilibrium and should be further developed to capture interplay among labor, housing, and credit markets and the impact of government and macroprudential policies.
- Corporate vulnerability assessment:
  - BCL developed econometric models to project bankruptcies and NPLs by sector.
  - BCL developed risk monitoring indicators for non-financial companies using AnaCredit.
  - Following ESRB Recommendation of December 1, 2022, BCL and CSSF stepped up monitoring of banks’ and funds’ CRE exposures and lending standards.
  - BCL obtained access to nonpublic balance sheet data to help develop corporate balance sheet stress tests and analysis.
- Stress-testing capacities:
  - Solvency: BCL developed a top-down stress test using mixture VAR (MVAR) model projecting probability of default of banks’ counterparts and impact on capital buffers. CSSF relies on top-down stress tests for loss-absorptive capacity and conducts sensitivity analysis/stress-tests based on real estate shocks.
  - Liquidity and market risks: BCL compiles liquidity indicators for investment funds. CSSF performs annual maturity transformation assessments at banks and monitors debt securities portfolio sensitivities on a quarterly basis using modified duration. CSSF monitors UCITS debt securities portfolios on a semi-annual basis via univariate stress tests in UCITS Risk Reporting. MMFs are monitored via quarterly EU reporting including WAM and WAL and standardized stress tests.
  - CSSF developed a liquidity stress testing (LST) framework for investment funds considering second round effects and extended asset classes and published a working paper on the effectiveness of LMTs.
  - Interconnectedness: BCL developed network analysis. CSSF simulates annual bank interconnectedness stress tests including second and third round effects (using LAREX) and conducts bi-annual stress testing on bank–investment fund interlinkages.
  - CSSF approaches to calibrate outflow rates for funds’ deposits at Luxembourg banks:
    - (i) outflow rates vary between 40% and 100%, depending on fund type,
    - (ii) deposit outflow rates depend on a redemption shock and liquid assets holdings of each fund; deposit outflows are compared to banks’ counterbalancing capacity, including short-term intra-group placements.

### Challenges in assessing macroprudential stance and systemic risks
- Lack of historical time series on asset quality, credit conditions and history of macroprudential policy complicates ex-ante assessment of macroprudential measures and may contribute to an inaction bias.
- Indicators and models are subject to substantial uncertainty due to data revisions and volatility; many rely on GDP or income measures that are highly volatile and subject to large revisions.
- BCL mitigates risks by relying on average measures and a signal approach (AUROC) based on multiple indicators.

### Recommendations on policy framework and calibration
- Recommendation 61:
  - Better map objectives, indicators and instruments.
  - Increase reliance on indicators less subject to revisions and coordinate with STATEC to isolate sources of volatility in national and financial accounts.
  - Set internal thresholds for household indebtedness and real estate related indicators with structural and cyclical components to trigger discussion about potential action; decisions not to act when thresholds are breached should be sufficiently motivated.
  - Use gradual calibration of tools and combined measures to reduce uncertainty about macroprudential policy impact.
- Recommendation 62 (short term):
  - Lock-in bank capital headroom using targeted capital-based measures.
  - Preserve resilience by increasing macroprudential capital requirements, preferably through sectoral systemic risk buffers on real estate exposures, allowing more precise targeting of stock vulnerability compared to the CCyB.
  - Continue close monitoring of collateral valuation effects and refinancing/liquidity risks in the real estate sector while maintaining sound lending practices.
- Recommendation 63–65 (income-based measures and calibration):
  - Introduce income-based measures early in the recovery cycle to address rising household indebtedness.
  - Existing measures: LTV and stress test requirement of 200bp of borrowers’ capacity to repay were introduced but fell short of tackling rising housing indebtedness.
  - Evidence suggests nonlinear complementarity between DSTI and LTV limits in reducing probability of default.
  - The maximum loan-to-value limit of 100 percent is very high and should be reduced gradually over time.
  - FSAP analysis calibrated a continuous-time heterogeneous agent model suggesting income-based limits reduce house prices in the medium term under supply rigidity, reducing household indebtedness and preserving consumption after shocks; affordability costs for low-income households could be compensated by social housing and targeted policies.
  - Authorities should immediately start working on calibration of income-based limits:
    - FSAP suggests calibrating stressed-DSTI around 45-50 percent, possibly tied to the current interest rate stress test required by CSSF for mortgages.
    - The stressed interest rate could be rule-based (e.g., maximum between a flat rate and the current mortgage rate augmented by a margin).
    - Apply a gradual approach with a speed limit giving flexibility to banks and tightening as recovery firms up.
  - Figure 14 findings: “optimal” level for all households is 45 percent, while by income level most peaks are around 50 percent.

*Source: Box 1. Efforts in Filling CRE Data Gaps, 1luxea2024006*

### 66.      Over the medium term, the CdRS should evaluate the benefits of a positive neutral

### Over the medium term, the CdRS should evaluate the benefits of a positive neutral CCyB and modify its CCyB framework accordingly if the introduction of such a buffer is deemed beneficial

### Macro-financial rationale and recent practice
- Shocks can be asymmetric and an unexpected sudden credit contraction could occur even when there are no signs of excessive credit built-up.
- A positive neutral rate could provide releasable buffers against unexpected macro-financial shocks that are unrelated to the credit cycle.
- By keeping the CCyB at 0.5 since 2019, the CdRS and the CSSF is de facto following the same logic.

### Policy recommendations on macroprudential buffers
- Evaluate the benefits of introducing a positive neutral CCyB.
- Modify the CCyB framework accordingly if the introduction of such a buffer is deemed beneficial.

### Housing demand support and purchasing power — risks and recommended calibrations
- Untargeted tax incentives and buy-to-let measures, even if temporary, could support demand but risk impeding the orderly rebalancing of the housing market and could lead to moral hazard and encourage risk-taking by borrowers and lenders.
- Assistance to viable real estate developers could be envisaged under strict conditions; support in the completion of unfinished projects (e.g., in the form of guarantees under strict underwriting standards) could be considered.
- In line with past year article IV recommendations, the authorities should consider frontloading public projects, with greater involvement of the private sector.
- Over the medium term, consider redesigning help-to-buy policies, including by gradually phasing out interest payment deductibility, and expediting measures to unlock supply bottlenecks.

### Cross-country borrower-based measure calibrations — selected reported figures from Table 2
- Loan-to-value (LTV) examples:
  - Austria: 90% (20% speed limit).
  - Belgium: 90% for BTL (with 10% allowed up to 90%); 90% for FTB (with 35% allowed up to 100% and 5% over 100%); 90% for SSB (with 20% allowed up to 100%).
  - Canada: 95% for owner occupied with government guarantee up to 500,000; 80% for others.
  - Cyprus: 80% for FTB, 50% for construction of luxury properties, 70% for others.
  - Luxembourg: 100% for FTB, 90% for owner occupied (15% speed limit), 80% for others.
  - Netherlands: 100%.
  - Norway: 60% for second homes in Oslo, 85% for others.
  - Sweden: 85%.
  - United Kingdom: 90% for own and permanent residence, 80% for others.
- Debt service to income (DSTI) or stressed DSTI examples:
  - Austria: 40% (10% speed limit).
  - Belgium: 50%.
  - Czechia: 80%.
  - Denmark: 50% for those under age 36, 45% for others.
  - Finland: 50% stressed DSTI.
  - France: 35%.
  - Iceland: 40% stressed DSTI for FTB; 35% for others (5% speed limit).
  - Luxembourg (reported in table): 13-36% depending on income and interest rate.
  - Netherlands: 50% (with 10% allowed up to 60%).
  - New Zealand: 60% (with 5% allowed up to 70%).
  - United Kingdom: 40% unstressed; 50% stressed DSTI; 5% allowed up to 60%.
- Loan-to-income (LTI) examples:
  - Austria: LTI above 4 requires stress testing.
  - Belgium: 3.5 for FTB (20% speed limit); 3.5 for SSBs (20% speed limit).
  - Sweden: 4.5 times (15% speed limit).

*Source: IMF staff analysis and Table 2 as provided in the source content.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1luxea2024006.pdf_
