## 1prtea2026007 ## Source details **Canonical URL:** [1prtea2026007](https://www.imf.org/-/media/files/publications/cr/2026/english/1prtea2026007.pdf) ## Other formats - [Markdown version](/-/media/files/publications/cr/2026/english/1prtea2026007.pdf.md) - [Structured JSON version](/-/media/files/publications/cr/2026/english/1prtea2026007.pdf.json) --- ### Executive summary — overview and macrofinancial context - FSAP coverage: - Full-fledged top-down solvency stress test under two severe adverse macrofinancial scenarios. - Targeted sensitivity analyses: debt revaluations at amortized cost, potential losses from common large exposures, quasi-reverse stress tests. - Cash flow-based liquidity stress tests for general risk aversion, sovereign market distress, and idiosyncratic funding shocks. - Targeted assessments of household, corporate, and real estate sector vulnerabilities; interconnectedness analysis. - Recent macrofinancial indicators: - Real GDP growth: 1.9 percent in 2025. - Real wage growth: 6.0 percent in 2025 (after 7.4 percent in 2024). - Inflation (2025): headline 2.2 percent; core 2.1 percent. - Public debt: below 90 percent of GDP, down 44 ppts from 2020. - Housing prices: surge of 180 percent since 2015. - Near-term outlook: projected growth 1.7 percent in 2026; projected inflation 3.4 percent in 2026. - External risks: slowdown in euro area growth, trade tensions, tighter global financial conditions, commodity price volatility, conflict in the Middle East. ### Bank solvency stress test — starting position and main outcomes - Starting position: - High profitability, low non-performing assets, ample capital buffers after balance-sheet clean-up. - Operating costs among the lowest in the EA. - Stress scenarios: - Two severe adverse macrofinancial scenarios: “recessionary” and “geopolitical” (see Methodology section). - House price shocks calibrated: 20 percent (recessionary) and 25 percent (geopolitical). - Aggregate outcomes: - No bank’s capital position falls below the regulatory requirement in the adverse scenarios. - Combined materialization of credit, market, and interest rate risks under extreme stress does not deplete banks’ capital buffers. - Quasi-reverse stress test using banking sector’s maximum past PD over the past two decades still shows resilience. - Heterogeneity exists across banks; first‑year first quartile CET1 drops to 14.2 percent (recessionary) and 15.7 percent (geopolitical), reductions of 350 bps and 200 bps respectively. ### Liquidity risk and cash flow stress tests — summary - Structural liquidity metrics: - Aggregate LCR: 260 percent (2025Q2) — comfortably above minimum. - NSFR: 156 percent in Q2 2025 (up from 143 percent at end‑2021). - Cash flow stress test (nine sample banks, covering 85 percent of system assets): - Scenarios: general risk aversion, market sovereign distress, idiosyncratic funding shock. - Under general risk aversion and market sovereign distress, net liquidity positions remain resilient across horizons. - Under extreme idiosyncratic funding shock, a few banks face marginal liquidity shortfalls ranging from 0.1 percent to 3.5 percent of total assets; three banks experience negative liquidity position by five to six months. - System-wide counterbalancing capacity (CBC): 38 percent of total assets. - System’s cumulative NLP declined to EUR79.3 billion over 1-year in market sovereign distress; idiosyncratic scenario ends year with EUR39.1 billion CBC and total net outflows of 39 percent of total assets. ### Private-sector balance sheet vulnerabilities — households - Deleveraging and current risks: - Household indebtedness: 55 percent of GDP at end-2024, edging up to 57 percent at end-2025. - Household savings rate: 12.1 percent in 2025 (average 7.7 over 2013-15). - Adjustable-rate mortgages prevalent: about 75 percent of outstanding mortgages reprice within a year. - Under FSAP geopolitical scenario: - Share of vulnerable households (DSECTI>70 percent) projected to increase by 40 percent, reaching roughly one-quarter of households. - “Debt at risk” could increase to 20 percent of banks’ household loan portfolio (a 60 percent increase from starting point). - Household loan default rate projected to rise by about 35 bps. - Under “cost-of-living” shock: - Vulnerable households could increase by up to 60 percent (potentially approaching 30 percent of all households). - Debt at risk could increase by 75 percent. - Distributional findings: - Geopolitical scenario: about 60 percent of low-income households become overburdened (vs. roughly 45 percent starting point); 17 percent of middle-income and 2 percent of high-income households become stressed. - Under geopolitical shock, 46 percent of at-risk debt is owed by low-income borrowers; 13 percent by high-income borrowers. ### Private-sector balance sheet vulnerabilities — corporates - Deleveraging and resilience: - NFC debt: 103 percent of GDP at end-2025 (from 176 percent at end-2012). - NFC internal equity funding: 45.7 percent of funding at end-September 2025 (from 31 percent at end-June 2009). - NFC EBITDA at end-September 2025: 9.2 percent of total assets (20-year average: 7.3). - Stress results: - Aggregate corporate PD peaks: 6.3 percent in 2027 (recessionary) and 5 percent in 2027 (geopolitical). - Shocks similar to the European debt crisis would result in markedly lower default rates than in the past; construction and real estate sectors are more vulnerable. - Static sensitivity: share of firms with ICR<1 increases by about 45 percent in adverse scenarios. ### Real estate market and banking exposures - Key metrics: - Residential property prices: up 16.4 percent in real terms y-o-y in 2025Q4; grew 19 percent y-o-y in 2025Q4 in another summary point. - House prices: 180 percent increase since 2015 (EA average 56 percent as of 2025Q4). - Price-to-income ratio: 27 percent above historical average; price-to-rent ratio: 52 percent above historical average. - Real house prices about 21 percent higher than predicted by real GNDI per household as of 2025Q2. - Bank exposures: - Bank exposures to domestic real estate (mortgages, other real estate collateralized loans, real estate holdings, funds): amount to a third of the banking system’s total assets. - Residential mortgage lending growth: 10 percent y-o-y in 2026Q1 (fastest since 2010). - Risk mitigants: two-thirds of house loans LTV<80 percent at end-2024; 92 percent of loans have DSTI<50 percent at end-2024; systemic risk buffer introduced Oct 2024; countercyclical capital buffer raised to 0.75 percent in January 2026. - Vulnerability assessment: - Tail house price corrections sizable but do not generate aggregate capital shortfalls in main FSAP stress tests; macrofinancial tail approach with 40 percent house price shock reduces CET1 by additional 50 bps (post-stress CET1 15.8 percent). ### Interconnectedness and cross-sectoral exposures - Trend: - Financial exposures among key sectors limited and moderated compared to a decade ago. - Bilateral links between banks and nonbanks declined significantly, mainly due to contraction in exposures to OFIs. - Domestic sovereign-bank nexus weakened: banks’ direct exposure to Portuguese public debt down (domestic sovereign allocation 22 percent of debt securities portfolio, down from 73 percent in 2015). - Cross-border banking modest but monitor Spanish banks’ presence and contagion channels. - Large exposures (pre- and post-CRM, percent of Tier 1 Capital): - Total LEs: Pre-CRM 127.0; Post-CRM 70.7. - Common LEs: Pre-CRM 40.1; Post-CRM 35.7. - Top 10: Pre-CRM 110.6; Post-CRM 56.9. - Top 3: Pre-CRM 69.1; Post-CRM 23.7. - Complete default of 11 identified common large exposures → loss ≈ 35.7 percent of Tier 1 capital post-CRM. ### Key quantitative findings (selected) - Housing prices: 180 percent surge since 2015. - Real GDP growth: 1.9 percent in 2025; projected 1.7 percent in 2026. - Real wage growth: 6.0 percent in 2025 (7.4 percent in 2024). - Inflation: headline 2.2 percent; core 2.1 percent in 2025; projected 3.4 percent in 2026. - Public debt: below 90 percent of GDP, down 44 ppts from 2020. - LCR: aggregate 260 percent (2025Q2); NSFR: 156 percent (Q2 2025). - NPL ratio: 2.1 percent at end-2025 (peak 17.9 percent in early-2016). - Liquidity shortfalls under extreme idiosyncratic shock: between 0.1 percent and 3.5 percent of total assets. - Estimated total market revaluation losses (FV securities): EUR2.5 billion or 1.35 percent of aggregate initial RWA. - Unrealized losses on debt securities held at AC under adverse scenarios: EUR10 billion; up to 7 percent of initial RWA in aggregate; potential unrealized losses up to 600 bps in aggregate and up to 1100 bps for LSIs. ### Methodology and scenarios (stress testing framework) - Balance sheet assumption: - Static balance sheet: exposures and liabilities constant at cut-off; maturing loans replaced by new performing loans; banks can build capital through retained earnings; no capital injections. - Hurdle rates: - Minimum capital ratio: Basel III CET1 4.5 percent + Pillar 2 requirements (P2R). - Minimum capital + Combined Buffer Requirement (CCoB and O-SII buffer). - Leverage ratio assessed against 3 percent Basel III minimum. - Adverse macro scenarios: - Recessionary: synchronized global slowdown, sovereign distress in EA, GDP growth drops to -4.4 percent (severity -1.9 sd). - Geopolitical: escalation of conflicts, supply disruptions, monetary tightening, GDP growth drops to -3.4 percent (severity -1.5 sd). - House price shocks included: 20 percent (recessionary) and 25 percent (geopolitical); alternative 40 percent tail shock analyzed in sensitivity. - Market risk scenarios: - Two short-term market stress scenarios using expected shortfall at 0.1 percent of market factors’ marginal distributions (Jan 2008–Oct 2024). - Credit risk measurement: - PDs: household PDs from HFCS (2021 aged to 2024); corporate PDs from Tressel and Ding (2021) model; sovereign/financial PDs via Merton on projected credit spreads, LGD fixed at 35 percent. - LGDs: collateralized via structural modelling using LTVs and bank-specific property price paths; unsecured via Frye and Jacobs (2012). - IFRS 9 provisioning implemented: 12-month for stage 1; lifetime for stages 2 and 3. - NII modelling: - Structural approach across 18 portfolio segments with repricing ladders, new-business rates from satellite models, and passthrough estimates (e.g., variable mortgage passthrough 1.1 first year, 1 second year). - Aggregate NIMs: baseline 2.7 percent; geopolitical scenario increases to 3.6 percent (about +30 percent); recessionary scenario falls to 2.4 percent by 2028 (about -10 percent). - Liquidity cash flow test: - June 2025 COREP C66 contractual cash flows; scenarios aligned with solvency narratives; key run-off/inflow and CBC parameters applied (detailed in Appendix VI). ### Sensitivity analyses and tail assessments - House price tail shock: - Macrofinancial approach (40 percent house price shock) → additional capital decline of 50 bps relative to main FSAP stress tests; post-stress CET1 15.8 percent. - Sensitivity approach: each 20 percent house price decline ≈ 20 bps losses in mortgage portfolio. - Revaluation of debt securities at amortized cost (AC): - AC holdings: 72 percent of debt portfolio. - Including potential unrealized AC losses → market losses reach EUR10 billion; up to 7 percent of initial RWA. - Impact: breach of combined buffer by two banks; average CET1 drops to as low as 11.3 percent (recessionary) and 13.0 percent (geopolitical) when AC losses included. - Large exposures: - Post-CRM top-10 exposures = 56.9 percent of Tier 1 capital. - Complete default of 11 common large exposures → loss ≈ 35.7 percent of Tier 1 capital post-CRM. - Quasi-reverse solvency stress test: - Applying households’ and corporates’ historical peak PDs (household 2.7 percent peak; corporate 10 percent peak historically) — under current adverse scenario PDs rise to 1.6 percent (household) and 6.3 percent (corporate), both below 2012 peaks. - Using highest historical PDs: no banks dip into buffers; average losses increase by 150 bps (geopolitical) and 240 bps (recessionary). - Doubling historical peak PDs: some banks dip into minimum buffers; average losses increase by additional 500 bps of initial RWA. - Middle East conflict sensitivity: - IMF April 2026 WEO severe downside: oil +100 percent from 2026; inflation expectations +100 bps; risk premiums +100 bps; lending rates +200 bps; food +25 percent. - Targeted analyses: tourism/transport/manufacturing credit losses and household micro-model. - Results: direct household loan credit losses contained at 30 bps; share of overburdened households doubles to 40 percent; total credit losses up to 140 bps — about half of FSAP adverse scenarios’ credit losses; capitalization remains above minimum requirements. ### Main policy recommendations on stress testing (selected) - Solvency stress testing: - Continue monitoring real estate exposures and sovereign debt portfolios, in particular potential unrealized losses in securities held at amortized cost. (ST) - Invest resources to further develop monitoring tools: refine NII models to better capture passthrough from mixed-rate mortgages, and develop reverse stress test capabilities. (MT) - Expand stress-testing models to account for macrofinancial feedback loops and enable assessment of policy intervention impacts. (MT) - Liquidity stress testing: - Continue monitoring demandable deposit outflows and risks associated with unexpected withdrawals, including from new bank and non-bank competitors and new products. (ST) - Timing legend: - I: Immediate = less than one year; ST: short term = 1-2 years; MT: medium term = 3–5 years. ### Conclusions and monitoring priorities - Overall assessment: - Banking system resilient to adverse scenarios and targeted sensitivities considered, supported by strong capitalization, profitability, low NPLs, and improved funding resilience. - Ongoing vigilance required given elevated house prices, sizable residential mortgage exposure, and significant holdings of EA sovereign debt with limited diversification. - Monitoring priorities: - Strengthen monitoring of demandable deposit risks. - Enhance modelling of NII dynamics and macrofinancial feedbacks. - Develop reverse stress testing capabilities to better capture emerging risks. *Source: EXECUTIVE SUMMARY and chapter content (FSAP) — Portugal (1prtea2026007).* ### EXECUTIVE SUMMARY __________________________________________________________________________ 7 ### EXECUTIVE SUMMARY ### Overview - The Financial Sector Assessment Program (FSAP) for Portugal conducted a comprehensive systemic risk analysis and assessment of the resilience of the banking sector, including: - A full-fledged top-down solvency stress test under two severe adverse macrofinancial scenarios. - Targeted sensitivity analyses of risks not fully captured in solvency tests (debt revaluations at amortized cost, potential losses from common large exposures, quasi-reverse stress tests). - Cash flow-based liquidity stress tests for general risk aversion, sovereign market distress, and idiosyncratic funding shocks. - Targeted assessments of household, corporate, and real estate sector vulnerabilities. - Interconnectedness analysis of cross-sectoral linkages in the financial sector. ### Macrofinancial Context - Recent macroeconomic performance and risks: - Real GDP grew at 1.9 percent in 2025, mainly driven by private consumption and supported by RRF funding. - Real wage grew by 6.0 percent in 2025 (after 7.4 percent in 2024). - Headline and core inflation were 2.2 and 2.1 percent respectively in 2025. - Public debt declined to below 90 percent of GDP, down 44 ppts from 2020. - Housing prices have surged by 180 percent since 2015, posing a key risk given banks’ substantial residential mortgage exposure. - Near-term outlook: growth projected at 1.7 percent in 2026; inflation expected to increase to 3.4 percent in 2026. - External risks include slowdown in euro area growth, trade tensions, tighter global financial conditions, commodity price volatility, and the conflict in the Middle East. ### Bank Solvency Stress Test Results - Starting position and resilience: - Banks enter scenarios from a position of strength supported by high profitability, low level of non-performing assets, and ample capital buffers following a decade of balance sheet clean-up and risk reduction. - Operating costs are among the lowest in the EA. - Stress scenarios and outcomes: - A combined materialization of credit, market, and interest rate risks under extreme stress does not lead to depletion of banks’ capital buffers. - No bank’s capital position falls below the regulatory requirement in the adverse scenarios. - Banks, in aggregate, display strong profitability and limited credit and market risk losses even under assumed extreme shocks, though heterogeneity exists across banks. - Quasi-reverse stress test using the banking sector’s maximum past probability of default over the past two decades still shows resilience. - Sensitivity analyses indicate the ongoing conflict in the Middle East does not change these solvency conclusions. ### Liquidity Risk and Cash Flow Stress Tests - Aggregate and structural liquidity metrics: - Aggregate LCR of all banks continues to comfortably exceed minimum regulatory requirements. - NSFR indicates strengthening of structural funding resilience since 2021. - Cash flow-based stress test of nine sample banks: - Net liquidity positions remain resilient across horizons under general risk aversion and market sovereign distress scenarios. - Under an extreme idiosyncratic funding shock, a few banks would face marginal liquidity shortfalls ranging from 0.1 percent to 3.5 percent of total assets. - Liquidity drain in the extreme idiosyncratic scenario is driven mainly by severe deposit outflows. - Monitoring priorities: - Emphasis on demandable deposit outflows and unexpected withdrawals, including risks from new bank and non-bank competitors and new products. ### Private-Sector Balance Sheet Vulnerabilities - Households: - Significant deleveraging over the past decade reduced vulnerabilities, but housing market risks have increased. - Low-income households remain vulnerable to rising living costs and interest rates due to prevalence of adjustable-rate mortgages. - Adverse conditions could sharply increase household “debt at risk”, impacting lower-income groups most; however, the very low share of mortgage credit allocated to lower-income groups should mitigate systemic risk. - Corporate sector: - Corporate debt levels are below euro area averages; resilience has strengthened through deleveraging and improved profitability. - Shocks of severity similar to the European debt crisis would result in markedly lower default rates than in the past, but sectors such as construction and real estate are more vulnerable. - Real estate: - Limited housing supply has kept house prices above levels implied by income and demographic trends, stressing the need to address supply bottlenecks. ### Interconnectedness and Cross-Sectoral Exposures - Financial exposures and contagion risk: - Financial exposures among key sectors are limited and have moderated compared to a decade ago. - Bilateral links between banks and nonbanks have significantly declined on both asset and liability sides, largely due to contraction in exposures to OFIs. - Domestic sovereign-bank nexus has weakened with lower direct bank exposure to Portuguese public debt. - Cross-border banking operations are modest but warrant closer monitoring given the notable presence of Spanish banks and possible contagion channels. ### Key Quantitative Findings (selected) - Housing prices: surge of 180 percent since 2015. - Real GDP growth: 1.9 percent in 2025. - Real wage growth: 6.0 percent in 2025; 7.4 percent in 2024. - Inflation: headline 2.2 percent and core 2.1 percent in 2025. - Public debt: below 90 percent of GDP, down 44 ppts from 2020. - Projected growth: 1.7 percent in 2026. - Projected inflation: 3.4 percent in 2026. - Liquidity shortfalls under extreme idiosyncratic shock: between 0.1 percent and 3.5 percent of total assets. ### Main Policy Recommendations on Stress Testing (selected from Table 1) - Solvency Stress Testing: - Continue to monitor risks related to real estate exposures and sovereign debt portfolios, in particular potential unrealized losses in the securities held at amortized cost. (ST) - Invest resources to further develop tools to monitor the banking sector, including refining models for net interest income to better capture changes in passthrough dynamics from the rise in mixed rate mortgages, and developing reverse stress test capabilities. (MT) - Expand stress-testing models to account for macrofinancial feedback loops and enable assessment of policy intervention impacts. (MT) - Liquidity Stress Testing: - Continue to monitor demandable deposit outflows and potential risks associated with unexpected deposits withdrawals. (ST) - Timing legend: - I: Immediate = less than one year; ST: short term = 1-2 years; MT: medium term = 3–5 years. ### Conclusions and Monitoring Priorities - The banking system is resilient to the adverse scenarios and targeted sensitivities considered, supported by strong capitalization, profitability, low NPLs, and improved funding resilience. - Ongoing vigilance is required given elevated house prices, banks’ sizable residential mortgage exposure, and significant holdings of EA sovereign debt with limited diversification. - Strengthen monitoring of demandable deposit risks, enhance modelling of NII dynamics and macrofinancial feedbacks, and develop reverse stress testing capabilities to better capture emerging risks. *Source: EXECUTIVE SUMMARY (FSAP) — Portugal* ### 3. Bank lending has accelerated after contracting in the 2010s, but loans continued to ### 3. Bank lending has accelerated after contracting in the 2010s, but loans continued to ### Credit dynamics and private sector deleveraging - Credit to domestic activities: - Grew by 6.8 percent in end-2025 (1.9 percent in December 2024). - After a 1.6 percent drop in 2023, is on the recovery path started in 2020 after a decade of negative growth. - Continued to decline relative to GDP, down to 90 percent of GDP at end-September 2025 from 191 percent at end-2010. - Segmental credit growth (end-2025): - Domestic mortgage and other household lending: up 8.9 percent. - Corporate credit: up 2.8 percent, after declining from June 2023 until September 2024. - Private sector deleveraging since 2013 (with brief interruption after 2020): - Debt of non-financial corporates (NFCs): declined to 103 percent of GDP at end-2025, from 176 percent at end-2012. - NFCs’ internal equity funding: 45.7 percent of NFCs’ funding at end-September 2025, up from 31 percent at end-June 2009. - NFCs’ EBITDA at end-September 2025: 9.2 percent of total assets (20-year average: 7.3). - Household indebtedness: declined from a peak of 94 percent of GDP at end-2009 to 55 percent at end-2024, edging up to 57 percent at end-2025. - Household savings rate: 12.1 percent in 2025 (average of 7.7 over 2013-15). ### Financial sector structure - Bank-centric system: - Banking sector total assets: 165 percent of GDP at end-September 2025, accounting for 65 percent of assets in the financial sector. - NBFI total assets: 91 percent of GDP in Q3 2025, compared to 368 percent in the Euro Area (EA). - NBFI composition highlights (Q3 2025): - Captive financial institutions and money lenders (including financial and non-financial holding companies): assets equivalent to 32 percent of GDP. - Insurance companies: assets at 20 percent of GDP. - Investment funds and pension funds jointly: 27 percent of GDP. - Banking sector concentration (end-June 2025): - Three significant institutions (SIs): account for 54 percent of banking system assets. - Foreign-owned subsidiaries: account for 24 percent of banking system assets. - Less significant institutions (LSIs): 14 percent of total assets. - Foreign-owned branches: 9 percent of total assets. ### Capitalization, liquidity, funding, and asset composition - Capital and liquidity (end-June 2025): - CET1 ratio: 17.9 percent. - Liquidity coverage ratio (LCR): 260 percent (EA median: 183 percent). - Net stable funding ratio (NSFR): considerably above regulatory minima (exact value not specified). - Funding structure (Q2 2025): - Deposits: 81 percent of total liabilities. - Within deposits: households 56 percent of total liabilities; corporate deposits 19 percent of total liabilities. - Term deposits: majority of household deposits. - Asset composition (end-June 2025): - Loans: 56 percent of total assets. - Mortgages: 27 percent of total assets. - Corporate loans: 18 percent of total assets. - Debt securities holdings: 29 percent of total assets (EA average: 14 percent). - Of debt securities, 72 percent are government debt securities at amortized cost. - Exposure to top three sovereigns: 48 percent of the total debt securities portfolio. - Domestic sovereign allocation: 22 percent (down from 73 percent in 2015). - Interest rate profile: - About 75 percent of outstanding mortgages and almost all corporate loans have a variable rate or short maturity. - Mixed-rate mortgages: 6 percent of total assets. - Repricing periods: - 73 percent of existing mortgages reprice within a year (95 percent at end-2022). - ECB interest rate cuts since June 2024: 200bps (noted potential pressure on net interest margins). - Interest rates (mid-2025): - Mortgage loans: 3.4 percent. - NFC loans: 4 percent. - Consumer loans: 7.3 percent. - Term deposits: 1.3 percent; sight deposits close to zero. ### Real estate exposures, property market, and credit quality - Real estate exposures: - Bank exposures to domestic real estate (mortgages, other loans collateralized against real estate, real estate holdings, real estate funds): amount to a third of the banking system’s total assets. - Residential market dynamics: - Residential mortgage lending growth: fastest since 2010 (10 percent y-o-y in 2026Q1). - Residential property prices: up 16.4 percent in real terms y-o-y in 2025Q4 (fastest y-o-y growth since 1989), outpacing household income. - Key demand drivers: supply constraints and non-resident and tourism demand. - Non-resident buyers: accounted for 10 percent of total transaction value in 2024. - Golden Visa program reference: granted residence to foreign buyers of properties valued over EUR500,000 (2013-23). - Risk mitigants for housing exposure: - Declining share of loans with high LTV and DSTI. - Systemic risk buffer for residential real estate exposures: introduced in October 2024. - Countercyclical capital buffer: raised to 0.75 percent in January 2026. - Household disposable income and savings rate rising. - Loan distributions (end-2024 / end-September 2018 comparisons): - Two-thirds of house loans had LTVs of less than 80 percent and virtually all less than 90 percent at end-2024 (55 percent and 79.5 percent, respectively, at end-September 2018). - 92 percent of loans associated with DSTI of less than 50 percent at end-2024 (80 percent at end-September 2018). - Credit quality improvements: - NPL ratio: declined to 2.1 percent at end-2025 from 17.9 percent peak in early-2016. - NPLs by portfolio (end-September 2025): housing loans 1 percent; consumption loans 6.1 percent. - Banks pursued large-scale NPL portfolio sales and asset management strategies since 2014. ### Bank profitability, efficiency, and balance sheet shifts - Profitability: - ROA: 1.4 percent at end-June 2025, 1.3 percent at end-December 2025. - Drivers: high interest income, reduced provisioning, stable fee income, low operational costs. - Interest-earning asset funding spread: large due to variable-rate loans and low pass-through to deposits. - Efficiency: - Portuguese banks among most efficient in EA with low cost-to-income ratios. - Balance sheet shifts since European debt crisis: - Banking sector assets: declined to 165 percent of GDP at end-September 2025 from 293 percent of GDP at end-2012. - Customer loans: declined to 56 percent of total assets at end-September 2025 from 70 percent in 2008. - Debt securities holdings: rose to 29 percent of total assets at end-September 2025 from 11 percent in 2008. - Rationale: deleveraging of households and NFCs, NPL disposals, capitalization requirements, and constrained access to wholesale funding during the crisis. ### Funding concentration and systemic considerations - Deposit dominance and implications: - Deposits represent 81 percent of total liabilities, with a large share protected by the deposit guarantee scheme. - Market-based funding plays a limited role; potential new competitors (including non-banks) and new products (such as stablecoins) could amplify deposit outflows in stress scenarios and warrant continuous monitoring. - Interlinkages and cross-border exposure: - Institutional interconnectedness has declined over the past decade. - Bilateral links between banks and nonbanks significantly declined, largely due to contraction in exposures to OFIs. - Sovereign-bank nexus has weakened with lower direct exposure to domestic public debt. - Cross-border banking operations modest relative to balance sheets; Spanish banks have notable presence in Portugal. - Foreign subsidiaries predominantly owned by Spanish parent institutions; Novo Banco subject to acquisition by French group BPCE (transaction noted). ### FSAP stress testing strategy and coverage - Stress testing approaches used: - Fully-fledged top-down solvency stress test against two severe, adverse macrofinancial scenarios. - Selected sensitivity analyses (including revaluation of debt securities at amortized cost, a quasi-reverse solvency stress test, large exposures analysis). - Wide range of liquidity stress tests for market-wide and idiosyncratic liquidity risks. - Assessment of private-sector balance sheet vulnerabilities for households, NFCs, and the housing market, with PD estimation and application in the bank solvency stress test. - Interconnectedness analysis for cross-sectoral structural assessment. - Alignment and data: - Stress test follows EA FSAP approaches and integrates Portugal-specific features using Banco de Portugal granular data. - New satellite models developed to estimate credit risk, NII, and market risk leveraging granular data. - Coverage and data sources: - Stress test used mid-2025 data for the nine largest banks, covering 85 percent of the banking system’s total assets. - Bank-level supervisory data included EBA FINREP and COREP, ECB STE templates, Portuguese credit register, and SHSG database. - Aggregation and grouping: - Results presented as aggregation for the full sample of nine banks. - Banks grouped into three categories: (i) SIs (3 banks), (ii) LSIs (3 banks), and (iii) foreign subsidiaries (3 banks). - Banco BIC acquisition by Abanca in 2024 noted; merger pro-forma not completed by Q2 2025 cutoff. *Source: 1prtea2026007 - 3. Bank lending has accelerated after contracting in the 2010s, but loans continued to (PDF chapter).* ### 22. Bank resilience was assessed against three criteria (“hurdle rates”): ### 22. Bank resilience was assessed against three criteria (“hurdle rates”): ### Hurdle rates - The “minimum capital ratio” or Supervisory Review and Evaluation Process (SREP) requirement: - Consisting of a Basel III regulatory minimum (CET1 ratio of 4.5 percent) plus Pillar 2 requirements (P2R). - The “minimum capital ratio” plus the “combined buffer requirement (CBR)”: - Consisting of two additional elements: a Capital Conservation Buffer (CCoB), and the Other Systemically Important Institution (O-SII) buffer. - The leverage ratio: - Assessed against the 3 percent Basel III minimum requirement. ### Methodology: static balance sheet assumption - A static balance sheet approach was adopted for the stress tests: - Exposures and liabilities remain constant and maintain the cut-off date level throughout the stress testing horizon. - Maturing loans are replaced by new performing loans; the maturity structure is kept constant. - Banks can build capital through retained earnings, but there are no capital injections. ### Scenarios: macrofinancial adverse scenarios - Two severe adverse macrofinancial scenarios were used: - “Recessionary scenario”: - Synchronized global slowdown, sovereign debt distress in the EA, widening of credit spreads, term-premium decompression, confidence losses. - Accommodative monetary policy reaction assumed, mitigating adverse impact on aggregate demand. - “Geopolitical scenario”: - Further escalation of geopolitical conflicts, heightened commodity price volatility, disrupted global production chains, large adverse trade, price, and tariff shocks (“trade wars”). - Countries with adequate fiscal space assumed to use fiscal policy to partly offset the demand shock. - Rise in inflation due to production chain disruptions assumed to prompt monetary policy tightening. - House price shocks included in both scenarios to capture real estate exposure risk: - House prices increased by 180 percent since 2015 in Portugal, compared to an EA average of 56 percent as of 2025Q4. - Price-to-income and price-to-rent ratios are 30 percent and 38 percent above estimated fundamental value, respectively. - ECB estimates overvaluation gap in Portugal at around 20 percent as of end-2024. - Calibration of house price shocks: - Recessionary scenario: 20 percent house price shock. - Geopolitical scenario: 25 percent house price shock. - Baseline scenario: - Reflected World Economic Outlook (WEO) projections in line with October 2025 WEO. - Severity and GDP growth outcomes: - Adverse scenarios calibrated to be severe and slightly more adverse than the 2025 EBA stress test scenario. - GDP growth drops to -4.4 percent in the recessionary scenario and -3.4 in the geopolitical scenario. - The GDP path corresponds to a severity of respectively -1.9 and -1.5 standard deviation. ### Market risk scenario - Two short-term market stress scenarios captured high-frequency market price and volatility movements (January 2008–October 2024) using expected shortfall at 0.1 percent of market factors’ marginal distributions. - Geopolitical market scenario: - Increase in interest rates (particularly at the short end of the EUR yield curve), commodity prices, and credit spreads, and a sharp contraction in equity prices. - Recessionary market scenario: - Drop in commodity and equity prices, larger increase in the credit spread of mid- and high-risk EA sovereigns, while the short end of the EUR yield curve remained muted. ### Credit risk: scope and measurement - Credit risk sources: - Losses from loans and advances to households, corporates, governments, credit institutions, other financial institutions, and default probability of debt securities issued by corporates, governments, credit institutions, and other financial institutions. - IFRS measurement for debt securities in the banking book: - Amortized cost (AC), fair value through profit and loss (FVPL), and fair value through other comprehensive income (FVOCI). - Modeling differentiation by accounting category: - AC securities: credit impairments estimated as for loans and advances. - FVPL securities: credit risk embedded in market risk methodology via price changes. - FVOCI securities: estimated through both market risk methodology and banking book credit impairment estimation. ### Probability of Default (PD) estimations - Household loans PDs: - Calibrated based on Household Financial and Consumption Survey (HFCS) completed in 2021. - HFCS: includes 83,000 households and 200,000 personal files across 22 countries; final dataset contains 6,107 households and 15,077 personal files for Portugal. - Corporate loans and bonds (AC and FVOCI) PDs: - Derived from the corporate PD projection model of Tressel and Ding (2021). - Sovereigns and financial institutions PDs: - Estimated using a Merton approach applied to credit spreads projected in the scenario, defined as the difference between bond yields and the risk-free rate (approximated by Germany’s 10-year government bond yield), assuming a fixed loss given default (LGD) of 35 percent. - Consequently, the PD of banks’ sovereign exposures to Germany, Austria and Netherland is 0. - Projected PD peaks: - Aggregate corporate PD rises to 6.3 percent in 2027 in the recessionary scenario and 5 percent in 2027 in the geopolitical scenario. - Household PD rises to 1.6 percent in the geopolitical scenario and to 1.3 percent in the recessionary scenario from a starting point of 1.1 percent. - Historical comparison: PD levels remain below the peak of 2.7 percent during the European debt crisis. - Implementation for each bank and lending segment: - Bank- and segment-specific PD path generated as a weighted average of aggregate, country-specific PDs based on the bank’s COREP (C09.02) lending exposure distribution. - Adjustment to bank- and segment-specific starting point-in-time (PiT) PD using the distance-to-default formula: - PD_{i,t} = N( G(PD_{i,0}) + G(aPD_t) − G(aPD_0) ) - Where G is inverse of cumulative distribution function of standard normal distribution and N is cumulative distribution function of standard normal distribution. - Adjusted PD path and initial transition matrix serve as input to beta-linking methodology to derive IFRS 9-compliant transition matrices. ### Loss Given Default (LGD) estimation - Collateralized lending LGDs: - Calibrated via structural modelling (Gross et al., 2020) using reported collateral value in loan-to-value (LTV) metrics and bank-specific property price growth paths. - Bank-specific property price growth paths are weighted averages of aggregate country-level property price paths based on each bank’s real estate exposures. - Substantial decline in property values under adverse scenarios materially affects LGD of secured lending. - Unsecured lending LGDs: - Calibrated via Frye and Jacobs (2012) method. - Starting LTVs and starting LGDs for all segments reported in STE templates; both secured and unsecured LGDs calibrated at bank-specific portfolio level. ### Credit risk impact through Profit and Loss (P&L) - IFRS 9 provisioning rules implemented: - Provisions calculated on a 12-month basis for stage 1 exposures and on a lifetime basis for stage 2 and stage 3 exposures; stage 3 PDs assumed to be 100 percent. - Exposure at default (EAD) includes on-balance sheet amounts and off-balance sheet items; off-balance sheet items weighted by segment characteristics. - Static balance sheet assumption: matured loans reinitialized at the same stage as when matured; no write-offs assumed. - Credit impairments in P&L reflect annual change in balance sheet stock of provisions. - Estimated loan impairment charges (sample of nine banks): - Baseline: provisions amount to 0.2 percent of total loans in 2026. - Baseline path: rise modestly to 0.3 percent in 2026 and 2027 (aggregate statement). - Recessionary scenario: - Provisions increase from 0.3 percent in 2026 to a peak of 1.2 percent in 2027, then moderate to 0.6 percent in 2028. - Geopolitical scenario: - Provisions rise to 0.3 percent in 2026, 1.1 percent in 2027, and 0.6 percent in 2028. - Bank-type patterns: - Similar overall pattern; foreign subsidiaries show a more modest reduction in 2028. - Timing: - 2027 spike driven by largest drop in property prices and sharpest migration into stages 2 and 3; 2028 decline follows GDP growth recovery and partial property price recovery. - Drivers of impairment charges: - Corporate and consumer loans drive estimated impairment charges; mortgage loan impairment charges are small. - Across all banks: - Corporate loans represent about 40 percent of total loans and account for more than 60 percent of additional provisions in 2027. - Mortgage loans can represent as much as 50 percent of loans but account for only 10 percent of impairment charges. - Variations by bank group: - LSIs: corporate loans account for as much as 80 percent of provisions; LSIs show no mortgage-related impairment charges in 2027 under adverse scenarios. - Foreign subsidiaries: face 17 percent impairment charges in the geopolitical scenario linked to mortgage loans. - Consumer loans (6 percent of total loans) account for 30 percent of LSIs’ aggregate impairment charges in the adverse geopolitical scenario and up to 40 percent for Sis. ### Capital requirements for credit risk and RWAs - IRB portfolios: - Asymptotic single risk factor model for unexpected losses implemented (Basel III). - RWAs adjusted for changes in PDs and LGDs, provisions, and credit conversion factors of off-balance sheet items. - Regulatory through-the-cycle (TTC) PDs calibrated as weighted average of PiT PDs for each year (weight 0.2) and previous year’s TTC PDs (weight 0.8), starting with TTC PDs reported in COREP. - Regulatory downturn (DT) LGD is the maximum between reported DT LGD at period zero and estimated PiT LGD. - Standardized approach banks: - Credit risk charges estimated using the density of credit RWA at the cut-off date (end of Q2 2025). - Non-performing exposures (NPEs) subtracted from RWA at time zero assuming a risk weight of 100 percent. - Performing RWA to performing exposures ratio used to project performing RWA each year; NPEs added back multiplied by 100 percent NPE RWA density. - RWA density evolution: - Baseline aggregate RWA density increases from 0.41 to 0.45. - Adverse scenarios aggregate RWA density increases up to 0.54 (an increase of 36 percent). - RWAs do not increase considerably for LSIs; RWAs increase by 50 percent for foreign subsidiaries. - RWA increases driven by higher LGD related to real estate shock and accumulation of stage 3 and stage 2 assets. *Source: IMF staff calculations and the IMF FSAP chapter content provided.* ### 43. The analysis covers all interest-earning assets and interest-bearing liabilities, with the ### 1prtea2026007 - 43. The analysis covers all interest-earning assets and interest-bearing liabilities, with the ### Scope and methodology - Analysis covers all interest-earning assets and interest-bearing liabilities, with the exception of derivatives. - Includes instruments classified as AC and fair value—both FVOCI and FVPL; only on-balance sheet items covered. - Derivatives excluded due to data limitations. - NPEs assumed to generate no interest income (income set to zero in the stress test). - NII not capped at the starting point level to capture transmission from short-term interest rate changes. - Interest income and interest expense projected using a structural approach requiring three inputs per portfolio segment: - (i) bank-level repricing structure at the starting point (“repricing ladder”); - (ii) time series for interest rates for newly originated exposures and liabilities (“interest rates for new business”); - (iii) interest rate levels at the starting point. - FSAP first projected interest rate for new business at country level using satellite models, then computed impact via structural model using bank-level repricing ladders and starting interest rates. ### Segmentation and balance sheet shares - Net interest income modeled across 18 bank portfolio segments. - Interest earning assets represent about 85 percent of Portuguese banks’ balance sheet. - Interest-bearing liabilities amount to 80 percent of total liabilities. - Analysis covers 10 portfolio segments for loans, two segments for debt securities and one for remaining assets. - Mortgages segmented by fixed, mixed, or variable rate; variable rate mortgages further broken into sub-segments to account for reference rates and repricing schedules. - For loans, besides Portugal, only other material geographical exposure is Poland (mostly due to one subsidiary of an SI). - For government debt securities, all EA countries are covered and explicitly modelled. ### Repricing ladder (starting dynamics) - Using bank-level data (BdP, COREP): 85 percent of loans reprice over the scenarios’ three-year horizon, and 75 percent reprice each year. - For debt securities, 40 percent reprice over the scenarios’ horizon. - Among interest-bearing liabilities, 90 percent reprice every year (large share of sight deposits and 6-month average maturity of term deposits). ### New business rates projection and passthroughs - New business rates estimated for all segments for each scenario year. - Passthroughs for loans and deposits estimated using country-level quarterly data (BdP for mortgage loans; MIR for corporate and consumer loans and deposits). - New business rates deltas estimated using specification (Equation 2): - IR_loan/deposit_t = α + β1 ∆ST_rate_t + β2 ∆ST_rate_{t−1} + β3 ∆LT_rate_t + β4 ∆LT_rate_{t−1} + β5 ∆infl_{t−1} (+β6 HPg_t + β7 GDPg_t) + ε_t - ∆ denotes quarterly changes; ST_rate = short-term rate; LT_rate = long-term rate; infl = yearly inflation. - Variable rate mortgages equation also includes HPg (house price yearly growth) and GDPg (annual GDP growth). - Debt securities’ interest rate projections taken from scenarios’ long-term rates. - Variable rate mortgages deltas derived from projected short-term and long-term rate projections. - Estimated passthrough for variable interest rates: 1.1 in the first year and 1 in the second year. - Passthroughs lower for fixed and mixed rate mortgages and for corporate and consumer loans, ranging between 0.1 to 0.6 after two years. - Passthrough for term deposits: - corporate deposits: 0.7 in the first year and 0.9 after two years; - term deposits for households: 0.2 in the first year and 0.6 in the second year. - Passthrough for sight deposits close to zero (slightly higher for corporate deposits). - Sight deposits’ interest rate almost flat at around zero due to low passthrough and zero-percent floor applied on rates. - Sight deposits represent about 30 percent of Portuguese banks’ liabilities. ### Projected Net Interest Margins (NIMs) - NIM projections follow short-term interest rate paths across three scenarios. - Aggregate NIM remains at 2.7 percent of interest earning assets over the three years in the baseline. - Geopolitical scenario: NIM increases sharply by about 30 percent, reaching 3.6 percent. - Recessionary scenario: system-wide NIM falls by 10 percent to 2.4 percent by 2028. - Business model heterogeneity limited; all banks exhibit dynamics closely following the policy rate path. - Foreign subsidiaries show a more pronounced NIM dynamic (larger increase and reduction). - Drop in NIM in recessionary scenario less pronounced than increase in geopolitical scenario due to sizeable portfolio of debt securities and rise in long-term interest rates in adverse scenarios. ### NFCI and other profit and loss items - NFCI projected using a satellite model with the same econometric specification as EA FSAP (IMF, 2025); LASSO used to select variables; projected at aggregate level using data on small EA banks from Haver (excluding G-SIBs). - Projected NFCI drops by about 20 percent on aggregate. - Baseline: NFCI follows a flat path of 72 basis points (bps). - Geopolitical scenario: NFCI falls to 59 bps by 2028. - Recessionary scenario: NFCI ends at 57 bps by 2028. - NTI (net trading income): baseline client revenues given by 5-year average of NTI/total assets; in adverse scenarios client revenues set to zero in first year and given a 20 percent haircut relative to baseline in remaining years. - NTI not a significant source of income for Portuguese banks. - Other P&L items projected by taking five-year average relative to total assets and scaling by the path of inflation. - Tax: historical effective-rate approach using each bank’s average effective tax rate over past five profitable fiscal years; applied only when bank reports positive PBT; no tax expense recognized when a bank records a loss. - Dividends calibrated with payout rate of 80 percent applied to total comprehensive income (TCI); dividends set to zero if bank incurs losses or if bank’s capital position breaches minimum regulatory requirements (Pillar 1 and 2). ### Market risk (debt securities revaluation) - Market risk assessed using modified duration to revalue debt securities held at fair value in two market stress scenarios; assessment covers only debt securities. - Debt securities on aggregate represent 30 percent of total assets; share as high as 40 percent for LSIs. - Only positions held at fair value (FVOCI and FVPL) revalued; AC portfolio excluded from revaluation. - Sensitivity analysis performed for securities held at AC. - Market-risk impact incorporated as a one-off overlay in the first year; estimated valuation losses not reversed in subsequent years. - No market shocks in baseline scenario. - Potential hedging not captured due to data limitations—estimated losses represent an upper band. - Shock derived from “delta shock” in Table 1 of Appendix III; combines shocks on risk-free interest rate (common depending on remaining maturity) and credit spread (depends on risk classification of country of exposure). - Data: seven out of nine banks used SHSG data; two banks used COREP reporting on debt securities (C33.00.a); for COREP banks duration approximated as mid-point bucket maturity; duration information provided by Banco de Portugal. - Revaluation formula (Equation 3): - FV_t = FV_{t−1} (1 − MD_t ⋅ ∆r_t) - where FV_t = fair value at time t; MD_t = modified duration; ∆r_t = shock on relevant interest rate (“delta shock”). - Estimated total losses: EUR2.5 billion or 1.35 percent of aggregate initial RWA. - Losses slightly higher in the recessionary scenario. - LSIs suffered less losses due to very low share of debt securities held at fair value. - SIs and foreign subsidiaries had larger losses due to higher share of fair value securities and concentration in higher-risk countries. ### Solvency stress test results - Initial capital ratios: CET1 = 17.7 percent; total capital ratio (TCR) = 20.2 percent. - Baseline path: CET1 rises to 18.8 percent in 2028; TCR reaches 21.2 percent. - Geopolitical scenario: CET1 falls first year to 17.1 percent, recovers to 17.8 percent in 2028. - Recessionary scenario: CET1 ends at 16.3 percent. - Aggregate results driven by SIs; LSIs experience a lower drop; Foreign Subsidiaries register an increase of their CET1 at end of third year in both adverse scenarios. - No bank breaches combined buffer requirements. - Heterogeneity of losses evident in quartile distribution: - First-year first quartile of CET1 drops to 14.2 percent under the recessionary scenario and to 15.7 percent for the geopolitical scenario. - Corresponds to capital ratio reductions of 350 basis points and 200 basis points, respectively (roughly double the sample average). ### Profitability dynamics - Banks remain profitable in all scenarios; most impacted in recessionary scenario. - Baseline: peak return on assets of 0.64 percent in 2026, softening to 0.35 percent by end of scenario horizon. - Recessionary scenario: profits fall to 0.57 in the first year (mainly due to a market-valuation shock) and decline to 0.47 in the following year as credit-risk charges increase; constant drop of NII also contributes. - Geopolitical scenario: profits increase compared to baseline despite significant credit impairment, thanks to a bump in net interest income from repricing of variable-rate loans and modest increase in loan impairments. *Source: IMF Staff calculations and Banco de Portugal data as presented in the provided chapter content.* ### 64. In the baseline scenario, strong NII and robust fee generation boost system-wide ### 64. In the baseline scenario, strong NII and robust fee generation boost system-wide ### Baseline and adverse scenario capital dynamics - Baseline scenario: - Strong NII and robust fee generation boost system‑wide capital despite moderate credit‑loss charges. - Large income streams partly offset by operating costs, taxes, and dividends. - Gradual growth in RWA is associated with a slight decrease in the ratio toward the end of the period. - Net effect: increase of 110 bps of capital buffers over the three‑year horizon. - Adverse scenarios: - Geopolitical scenario: - CET1 ratio remains broadly unchanged from its initial position. - Despite losses from market risk, a threefold increase in credit‑risk provisions, RWA rise due to NPL accumulation, and rise in LGD with the house price shock, revenues grow even more significantly. - Recessionary scenario: - NII is compressed. - Combined with slightly heavier credit and market losses, CET1 ratio ends up 140 bps below its initial position. - Note: Detailed results by bank type are provided in Appendix V. ### Sensitivity analyses overview (four analyses) - Purpose: explore risks not directly covered in the main solvency stress test; assess resilience; model extreme‑tail risks one factor at a time. - Four analyses: 1. Tail shock on residential real estate prices and banking sector losses. 2. Risks from unrealized losses in debt securities held at amortized cost (AC). 3. System’s large exposures and complete default of few counterparties. 4. Quasi‑reverse stress test: what level PDs would generate meaningful losses (other factors constant). ### Additional capital impact from tail shocks to residential real estate prices - Context: - House prices surged by 169 percent since 2015. - Residential real estate accounts for about 28 percent of total bank assets. - Price‑to‑income ratio increased to 170 percent in 2025 from 100 percent in 2015. - Current estimates point to an overvaluation of 20-40 percent (Cousin et al., 2025). - Stress test initially incorporates house price shocks of 20 and 25 percent under the recessionary and geopolitical scenarios, respectively. - Methods (two complementary approaches using aggregated banking‑system data): 1. Re‑run macrofinancial scenario with house price shock in the recessionary scenario adjusted to 40 percent (from 20 percent) and endogenously propagate spillovers; satellite models reused to translate stronger shock into higher PDs. - Illustration: the shock on GDP reaches -5.7 percent in that scenario and -4.4 in the recessionary scenario. 2. Sensitivity analysis estimating mortgage portfolio P&L channel via expected losses (PD×LGD×EAD), primarily via stressed LGDs (and to a lesser extent PDs); maps incremental credit losses to capital impacts. - Caveat: likely underestimates overall impact because it abstracts from spillovers to other portfolios. - Results: - Macrofinancial scenario‑based approach: capital declines by an additional 50 bps relative to primary FSAP stress tests, leading to a post‑stress CET1 ratio of 15.8 percent. - Sensitivity‑based approach: each 20 percent house price decline is associated with about 20 bps of losses in the mortgage portfolio. - Conclusion: resulting losses sizable but do not generate capital shortfalls at aggregate level; banking system appears sufficiently capitalized to absorb even extreme housing corrections, subject to caveats on simplifying assumptions and aggregated data. ### Revaluation of debt securities held at amortized cost (AC) - Holdings: - 72 percent of Portuguese banks’ debt holdings are held at AC. - Unrealized losses under adverse scenarios: - Market losses would reach EUR10 billion when including unrealized losses on debt securities held at AC. - Losses on debt securities portfolio including potential unrealized losses would be multiplied by five on aggregate compared to revaluation of debt securities held at FV only. - This represents up to seven percent of initial RWA in aggregate. - By bank group: - Losses reach up to six percent for SIs. - Up to 10 percent for foreign subsidiaries. - For LSIs, losses reach up to 12 percent. - Potential unrealized losses can reach up to 600 bps in aggregate and up to 1100 bps for LSIs. - Impact on solvency: - Addition of potential unrealized losses to the solvency stress test leads to a breach of the combined buffer by two banks. - Shortfall represents 50 bps of the banking system initial RWA. - CET1 ratio implications: - Average CET1 drops to a low of 13 percent in the geopolitical scenario and 11.3 percent in the recessionary scenario, compared with 17.9 and 17.2 percent respectively in the main solvency stress tests. - First quartile dips to 9 percent against 14.2 percent in the main stress test. ### Large exposures - Pre‑ and post‑CRM totals (In percent of Tier 1 Capital): - Total LEs: Pre‑CRM 127.0; Post‑CRM 70.7. - Common LEs: Pre‑CRM 40.1; Post‑CRM 35.7. - Top 10: Pre‑CRM 110.6; Post‑CRM 56.9. - Top 5: Pre‑CRM 85.0; Post‑CRM 38.6. - Top 3: Pre‑CRM 69.1; Post‑CRM 23.7. - Observations: - Total large exposures mainly to NFCs exceed 100 percent of Tier 1 capital pre‑CRM but fall to 70.7 percent after eligible collateral, guarantees, and other risk mitigants. - Top 10 groups account for 56.9 percent of Tier 1 capital post‑CRM. - Assuming complete default of 11 identified common large exposures, system could face loss of roughly 35.7 percent of Tier 1 capital post‑CRM. - Note: Group‑level totals exclude exposures to sovereigns and households. ### Quasi‑reverse solvency stress test (alternative PDs) - Method: re‑run solvency stress test using households’ and corporates’ historical peak PDs to assess resilience relative to 2012 peaks. - Under adverse scenarios PDs rise to 1.6 percent for households and 6.3 percent for corporates (both below 2012 peaks of 2.7 percent and 10 percent). - Results: - No banks dip into their buffers when applying highest historical PDs, but aggregate losses increase significantly. - On average, losses increase by 150 bps for the geopolitical and 240 bps for the recessionary scenario. - First quartile of CET1 distribution: 13.9 percent (geopolitical) and 11.9 percent (recessionary). - Only when applying PDs twice the historical peak do a few banks dip into minimum buffers; average losses rise by an additional 500 bps of initial RWA compared to main solvency results. - First quartile CET1 reaches 10.5 and 8.9 percent for geopolitical and recessionary scenarios respectively. ### Sensitivity: potential impact of the Middle East conflict - Framework: anchored in IMF’s April 2026 WEO severe downside scenario characterized by: - Oil prices increasing by 100 percent starting in 2026, relative to October 2025 baseline. - One‑year‑ahead inflation expectations ratcheting up by as much as 100 basis points in advanced economies. - Risk premiums in advanced economies up by 100 basis points; combined with inflation expectations translate into an increase of 200 basis points in lending rates relative to October 2025 baseline. - Food prices increase by 25 percent. - Two targeted sensitivity analyses (partial equilibrium, conservative): 1. Credit losses on banks’ loan exposures to tourism, transportation, and manufacturing sectors (jointly 11.9 percent of total lending) via PD×LGD×EAD; PDs stressed to historical peak default rates—up to about three times levels embedded in FSAP geopolitical adverse scenario. 2. Household stress test using micro model mapping WEO severe adverse macrofinancial paths into household budgets via higher living costs and higher mortgage repayments. - Results: - Direct potential credit losses from household loans (57 percent of bank loans) contained at 30 bps. - Share of overburdened households doubles in sensitivity assessment to reach 40 percent of households, putting 32 percent of bank debt at risk. - Corporate lending losses modest and largely concentrated in manufacturing. - Altogether, total credit losses for tourism, transportation, manufacturing, and rising household defaults assessed up to 140 bps (about half of credit losses from FSAP’s adverse stress test scenarios). - Banking sector capitalization remains comfortably above minimum requirements given sizeable buffers. ### Bank liquidity stress test — overview and key findings - Two themes: 1. Structural assessment of LCR and NSFR against regulatory thresholds. 2. Cash flow‑based stress test of nine banks (85 percent of sector assets) with three scenarios: general risk aversion, market sovereign distress, idiosyncratic funding; horizon overnight up to one year. - Key findings: - Banks entered stress scenarios from a strong starting liquidity position supported by large HQLA stock. - Sight deposits are a key funding source, but predominance of stable retail funding supports resilience. - Net liquidity shortfalls materialize only under idiosyncratic funding shock: three banks would experience a negative liquidity position by five to six months horizon, though shortfalls are marginal relative to banks’ total assets; driven mainly by stronger wholesale deposit outflows. ### Structural liquidity analysis (LCR, NSFR, HQLA, asset encumbrance) - LCR: - All banks comfortably exceeded minimum regulatory requirements. - Aggregate LCR reached 260 percent in 2025Q2. - By bank group: - LSIs held highest ratios; SIs broadly tracked system average. - Foreign subsidiaries LCRs: 182 percent in 2021 → 139 percent in 2022 → 161 percent in 2025Q2. - NSFR: - Aggregate NSFR increased to 156 percent in Q2 2025 from 143 percent at end‑2021. - Foreign subsidiaries’ NSFR levels closer to prudential thresholds. - Retail deposits constitute more than 70 percent of available stable funding of both SIs and LSIs. - HQLA: - Share of HQLA to total assets has steadily increased (aggregate LCR driven by HQLA increase while net outflow broadly constant). - Asset encumbrance: - Assessed relative to other EU countries (Q4 2024 metrics reported in figures). *Source: IMF Staff calculations.* ### 84. The asset encumbrance ratio of Portuguese banks is among the lowest in the EU ### 1prtea2026007 - 84. The asset encumbrance ratio of Portuguese banks is among the lowest in the EU ### Asset Encumbrance and Liquidity Buffers - The asset encumbrance ratio of Portuguese banks is among the lowest in the EU. - Low encumbrance indicates a solid liquidity cushion and flexibility to raise funding through secured operations, if needed. - High-quality liquid assets (HQLA) account for 74 percent of the sample banks’ total counterbalancing capacity (CBC), with more than half of HQLA comprising Level 1 assets. - Within Level 1 tradable assets, high-rated sovereign securities—i.e., credit quality steps (CQS) 2–3—account for the largest share of the portfolio. - The composition is relevant for stress-test results because these assets would face large haircuts under the market sovereign distress scenario (Appendix VI). ### Funding Structure - Banks’ funding structure is predominantly retail-based, reflecting a stable traditional funding model. - Deposits represent about 80-85 percent of total liabilities. - Market funding plays only a limited role. - Household deposits have been the dominant, stable form of deposit funding, averaging 58 percent of deposits in Q1 2015-Q2 2025. - Corporate deposits provide additional liquidity; interbank and other deposits remain small. - Term deposits still anchor household funding, but sight balances rose from 26 percent in 2014 to a peak of 52 percent in 2021, before easing to 45 percent in Q2 2025. ### Cash Flow-Based Liquidity Stress Test — Overview - The cash flow-based liquidity stress test assesses potential liquidity gaps from overnight up to one year. - The test used June 2025 COREP C66 data on contractual cash flows by maturity buckets. - Scenarios: three stress scenarios (see Appendix VI for key assumptions). - The analysis evaluates cumulative net contractual gaps (inflows minus outflows), cumulative counterbalancing capacity (CBC), and additional outflows from committed and uncommitted funding facilities. - The approach complements LCR and NSFR by capturing liquidity risks across interim maturity buckets, including potential shortfalls below 30 days and between 31 days and one year. ### Counterbalancing Capacity (CBC) - System-wide CBC amounts to 38 percent of total assets. - Over half of CBC is made up of Level 1 tradable assets and other CBC-eligible collateral. - By bank type: - SIs held the highest coverage at around 43 percent. - LSIs: 33 percent. - Foreign subsidiaries: around 29 percent. - Within Level 1 tradable portfolios, highly-rated sovereign debt (CQS 2-3) accounts for the largest share, most pronounced among foreign subsidiaries. ### Deposit-Based Funding Maturity Profile - Funding across all bank groups is broadly dominated by stable retail deposits; sight deposits account for a large share. - Share of deposits reported as “overnight” (could hypothetically be withdrawn overnight), by bank group: - Foreign subsidiaries: about 83 percent of their deposits. - SIs: around 60 percent. - LSIs: about 47 percent. - Foreign subsidiaries’ higher overnight share reflects a relatively higher share of sight deposits and notable non-operational wholesale funding. ### Results of the Cash Flow-Based Liquidity Stress Test - Under general risk aversion and market distress scenarios, all NLP quartiles remain comfortably positive, indicating system-wide resilience. - Under the idiosyncratic funding scenario (SC3), dispersion increases and liquidity pressures are limited to a few banks: - The 25th percentile is pushed slightly below zero at about 0.75 percent of total assets. - A time-bucket assessment shows three banks experiencing a negative liquidity position by the 5-6-month bucket onward (details on timing and number of banks with earlier marginal negatives are confidential). - Group-level net liquidity profiles under the idiosyncratic scenario: - LSIs and SIs maintain positive net liquidity balances across the horizon. - SIs start with the highest initial stock but see a sharper step-down at the overnight bucket, then a steady decline until the 9–12-month horizon. - Foreign subsidiaries start with a thinner buffer that is depleted over the exercise horizon; the shortfall reaches 2.5 percent of total group assets, but intragroup funding and parent support may mitigate strains. - A waterfall decomposition of cash-flow analysis: - Market sovereign distress scenario: the system’s cumulative NLP declined to EUR79.3 billion over the 1-year horizon, with net outflows totaling roughly 27 percent of total assets. - Idiosyncratic funding scenario: the system starts with a higher CBC but ends the year with EUR39.1 billion; total net outflows amount to 39 percent of total assets, driven mainly by greater deposit outflows, particularly wholesale deposits with higher run-off rates. ### Interconnectedness Analysis — Overview and Key Findings - The analysis maps cross-sectoral financial interlinkages using a network mapping approach and from-whom-to-whom financial exposures. - Cross-sectoral exposures include banks, NBFIs, corporates, government, households, and nonresidents. - Key finding: financial cross-sectoral exposures in Portugal are limited and have declined considerably since a decade ago. - Bank–NBFI interlinkages, which contributed to risk transmission during the GFC and European debt crisis, have weakened since then. - Banks’ financial exposures with the sovereign and the rest of the world have also declined; reliance on household funding (mostly deposits) has increased. - Aggregate net external financing to the domestic economy moderated, but network maps show considerable external funding to NFCs. ### Evolution of Sectoral Balance Sheets and Financial Exposures - Net financial positions show a gradual narrowing of sectoral asset-liability gaps since end-2013. - Government (GG) and NFC sectors—both net borrowers—have deleveraged since the European debt crisis. - Household sector and rest of the world (ROW) are net lenders; household net lending remained stable, ROW’s position steadily declined since 2013. - NBFI sector has been a small but steady net borrower. - Cross-sectoral shifts between 2013 and 2025: - Q4 2013: ROW was the largest net lender at 127 percent of GDP, channeling funding mainly to GG and NFCs. - Q1 2025: ROW’s net position declined to 69 percent of GDP. - NFCs’ cross-border net exposure increased to 53 percent of GDP in Q1 2025 from 48 percent in 2013, mainly via issuance of unlisted shares and other equity. - Domestic retrenchment: NFCs’ net financial position fell to 118 percent of GDP from 148 percent at end-2013; MFIs’ (banks) net financial position fell to 2 percent of GDP from 9 percent at end-2013. - Households’ creditor role strengthened, reflected in a rise in their net financial position. ### Bank–NBFI Interconnectedness - Banks’ gross exposures to NBFIs contracted sharply since end-2013: - Banks’ claims on NBFIs fell from EUR49.8 billion to EUR18.6 billion. - Banks’ liabilities to NBFIs declined from EUR60.5 billion to EUR18.7 billion in Q1 2025. - Reductions driven primarily by OFIs: - Banks’ claims on OFIs dropped from EUR35.9 billion to EUR12.1 billion. - Banks’ liabilities to OFIs declined from EUR39 billion to EUR9.8 billion. - Relative to banks’ total assets: - Banks’ exposures to NBFIs dropped from 10.1 percent at end-2013 to 3.9 percent in Q1 2025. - Funding from NBFIs fell to 3.9 percent of banks’ total liabilities from 12.6 percent. - Banks’ exposures to the government declined to 5.9 percent of banks’ total assets from 9.8 percent at end-2013. ### Private-Sector (Household) Balance Sheet Vulnerabilities - Household leverage has declined but vulnerabilities remain: - Household debt to disposable income ratio declined by about 40 percentage points over the past decade to below the EA average as of end-2024. - Bank credit to households fell from 81 percent of GDP in 2012 to 46 percent in 2025. - Credit growth accelerated to 7.4 percent in 2025 after subdued growth over the prior decade. - Risks to household finances: - A rise in cost of living and interest rates can pressure household finances, especially low-income households (which represent a small share of mortgage loans). - Prevalence of adjustable-rate outstanding mortgages increases sensitivity to rising interest rates. - Newly originated housing loans have initial rate fixation periods typically exceeding 2 years (mixed-rate mortgages), reducing short-term sensitivity of new loans’ debt-service costs to interest rate increases. *Based on content in the provided PDF chapter.* ### 98. The analysis uses micro-level data from the 2023 (latest) Household Finance and ### 1prtea2026007 - 98. ### Household-sector vulnerability assessment and stress-test methodology - Data and scenarios - Uses micro-level data from the 2023 Household Finance and Consumption Survey (HFCS) to simulate macrofinancial shocks on household finances and link household stress to bank credit risk. - Analysis run on the baseline and three adverse scenarios: the FSAP geopolitical scenario, the FSAP recessionary scenario, and a “cost-of-living” scenario. - The cost-of-living scenario combines a 200-basis-point increase in lending rates with an income shock consisting of a 10 percent decline in household disposable income and a 5-percentage-point unemployment shock (accounting for unemployment benefits), as well as a 20 percent increase in food and energy prices. All shocks are defined relative to baseline projections. - Note on survey ageing: The 2024 HFCS survey was conducted in 2021, aged to 2024 based on relatively aggregate assumptions; results should be taken with some caution. - Definitions and staging mapping to IFRS 9 - DSECTI: debt service and essential consumption (food, utilities, rents) to gross income ratio. - Overburdened households: DSECTI exceeding 70 percent. - IFRS 9 stage mapping (broadly matched with household balance-sheet metrics): - Stage 1: Households not in arrears and not struggling (DSECTI below 70 percent). - Stage 2: Households struggling (DSECTI above 70 percent, but not on arrears). - Stage 3: Households in arrears, irrespective of their financial ratios. - Approach is conservative relative to some bank practices that use arrears exceeding 90 days as a proxy for Stage 3. ### Key household findings and projections - Under the FSAP geopolitical scenario: - Share of vulnerable Portuguese households (debt service and essentials exceeding 70 percent of income) is projected to increase by 40 percent relative to initial levels—reaching roughly one-quarter of all households (vs. about one-fifth in the EA). - Share of “debt at risk” (portion of total household debt held by these vulnerable borrowers) could increase to 20 percent of banks’ household loan portfolio (a 60 percent increase from the starting point, same as Euro Area average). - Household loan default rate is projected to rise by about 35 basis points. - Roughly one-fifth of banks’ household credit portfolio would be at Stage 2 under stress, roughly doubling S2 loans relative to initial levels. (Model simulations estimate the starting share of S2 loans was 12 percent, compared to 10 percent reported by Portuguese banks at end 2021.) - Under the “cost-of-living” shock: - Rise in vulnerability is larger: up to 60 percent more households than baseline may become overburdened (potentially approaching 30 percent of all households). - Debt at risk could increase by 75 percent. - Caveat: conclusions should be interpreted with caution due to lags in the survey. - Distributional aspects (income terciles) - Under the geopolitical scenario, about 60 percent of low-income households in Portugal become overburdened (up from roughly 45 percent at the starting point), compared to 17 percent of middle-income and 2 percent of high-income households becoming stressed. - Under the geopolitical shock, 46 percent of at-risk debt is owed by low-income borrowers (versus 34 percent in the EA average); only 13 percent of at-risk debt is held by high-income borrowers (versus 30 percent in the EA). - Under the cost-of-living shock, up to 90 percent of bank debt is at risk within the low-income bucket (compared to 70 percent in the starting point). - Note: Debt profile structure of households has changed since the survey year, including the impact of BdP’s macroprudential recommendation on credit to consumers and the associated improvement in debtors’ credit risk profile (BdP, 2025a). ### Macro-financial and real-economy implications from household stress - Migration of exposures and provisioning - Rising household stress drives migration from Stage 1 to Stages 2 and 3, increasing bank provisions. - Many vulnerable borrowers are captured in Stage 2 rather than immediate default; persistent shocks could convert Stage 2 loans into Stage 3. - Consumption and GDP effects - Identifies “consumers at risk” whose total expenses (including debt service and rents) absorb their entire income and who would cut non-essential spending. - Under the FSAP recessionary (geopolitical) scenario, about 6 percent (7 percent) of aggregate consumption may need to be reduced, weighing down on GDP growth in Portugal, broadly in line with average projections for the EA. ### Corporate-sector vulnerabilities: methodology - Complementary approaches: dynamic scenario-based analysis and static sensitivity analysis (methodology similar to Tressel and Ding (2021)). - Dynamic analysis - Uses panel averages across 10 sectors and 4 firm sizes (40 categories) for 2006-24 to project corporate performance, including default rates, under baseline and two adverse scenarios. - Steps: estimate profitability, sales growth, and leverage with panel regressions (sector and size fixed effects); predict 2026-28 recursively holding 2024 structural firm characteristics constant and using scenario GDP growth and unemployment paths; predict other financial indicators (effective interest rate, ICR) from accounting identities; estimate probability of default as a function of firm-specific and macro variables. - Static sensitivity analysis - Uses firm-level data for 2024 to assess shares of NFCs that would experience cash shortfalls or difficulties servicing debt under calibrated shocks to sales and interest costs, with behavioral responses (reductions in cost of goods sold, other operating expenses, labor costs) calibrated at the national level. - Criteria: firms “at risk” when ICR < 1; firms “at risk” when existing cash balances and current earnings insufficient to cover interest payments. ### Corporate-sector findings and projections - Aggregate trends and resilience - Bank loans to non-financial corporates fell from a peak of 67 percent of GDP in 2009 to 24 percent in 2025. - Corporate sector has gone through significant deleveraging; NFCs’ total and net debt as a share of GDP have declined below EA levels since 2021. - Corporate profitability post-pandemic stronger than in past periods; average capital and interest coverage ratios increased, with heterogeneity across firm sizes. - Composition of corporate debt shifted: share of bank loans in total debt declined, particularly in construction and accommodation and food sectors; higher reliance on other loans and debt securities. - Dynamic scenario results (debt-weighted averages) - Shocks matching the severity of the European debt crisis would result in lower increases in default rates today than during that crisis because leverage remains significantly lower and pre-shock profitability is higher. - In the adverse scenarios, return on assets and sales growth decline by magnitudes similar to the European debt crisis; effective interest rate increases more, especially in the geopolitical scenario. - Probability of Default (PD) increases by 3 percentage points in the recessionary scenario and 2 percentage points in the geopolitical scenario compared to baseline, reaching 63 percent (recessionary) and 50 percent (geopolitical) of the 2012 level, respectively. - Sector heterogeneity: highest default rates predicted for the construction and real estate sector. - Static sensitivity results - Share of firms falling below the ICR threshold increases by about 45 percent in both adverse scenarios. - Geopolitical shock has a larger impact on debt at risk and bank loans at risk, increasing these ratios by 53 percent and 87 percent respectively. - Debt at risk increases most in professional services, trade, and transport sectors; bank loans at risk increase most in the utilities sector (from a relatively low base). - Vulnerability based on cash balances is lower and increases less in stress scenarios due to robust cash buffers. - Sectoral increases in debt at risk and bank loans at risk range between 10 and 30 percent for most sectors, with trade, accommodation and food, and professional services seeing increases from 50 to 100 percent. - In absolute terms: - Debt-weighted average ICR below threshold levels would increase from 23 percent to 34 percent (recessionary) and 36 percent (geopolitical). - Debt-weighted average cash below threshold levels would increase from 15 percent to 21 percent (recessionary) and 23 percent (geopolitical). - Historical benchmarks: ICR below threshold reached 43 percent in 2012 (European debt crisis); cash below threshold reached 33 percent in 2012. *Source: 2023 HFCS and IMF Staff calculations, Banco de Portugal, and IMF staff analysis as presented in the cited chapter.* ### 109. House prices have risen sharply in recent years. They grew by 19 percent y-o-y in 2025Q4 ### 1prtea2026007 - 109. House prices have risen sharply in recent years. They grew by 19 percent y-o-y in 2025Q4 ### Recent developments and key statistics - House prices grew by 19 percent y-o-y in 2025Q4. - House prices have grown by 180 percent since 2015, compared to an EA average of 56 percent. - The price-to-income ratio is currently 27 percent above its historical average. - The price-to-rent ratio is currently 52 percent above its historical average. - The ratio of new housing credit to total value of house purchase transactions has been less than 50 percent on average since 2012. - Nonresidents on average account for 6 percent of the number of transactions but buy houses that are on average 80 percent more expensive. - As of 2025Q2 real house prices were about 21 percent higher than predicted by real GNDI per household. - The 2024 ESRB report assessed vulnerabilities in the residential real estate sector of Portugal as medium (ESRB, 2024). ### Credit, foreign demand, and household behavior - Credit has not been a key driver of house prices historically: the stock of mortgage loans did not increase concurrently with house prices, unlike the EA overall. - Mortgage growth in Portugal has been picking up faster than in the EA in recent quarters, suggesting this dynamic may be changing. - Foreign demand remained strong after the discontinuation of the golden visa program for residential real estate investors in 2023, as evidenced by growing FDI in the real estate sector. - A significant share of Portuguese household savings flows into real estate, likely due to lack of alternatives. - High-level model estimates indicate recent government measures to support young homebuyers have likely contributed to house price growth and are unlikely to have led to a significant rise in house purchases (more data needed for deeper analysis). ### Short-term drivers and model results - Short-term drivers identified as important determinants of house price growth: unemployment, financial conditions, and foreign demand (real estate FDI). - A rise in the unemployment rate to 10 percent (from 6 percent currently) reduces real house price growth by 3 percentage points, ceteris paribus. - A reduction of real estate FDI to GDP ratio by 0.1 percentage points (equivalent to one standard deviation in the post-pandemic period) reduces real house price growth by 7 percentage points. - Tightening of financial conditions by 1 standard deviation reduces real house price growth by 4 percentage points. - Model results suggest house price growth does not consistently adjust to close the gap between real house prices and their long-term equilibrium levels as determined by real gross disposable income per household. - The coefficient on the error-correction term has the expected negative sign but is not always statistically significant across specifications and does not significantly improve explanatory power. ### Supply constraints and construction sector dynamics - Supply of new houses, while picking up in recent years, remains below the historical average. - Persistent house price deviations from long-term equilibrium are associated with slow supply responses in the housing market. - Structural shift: post-European debt crisis period saw lower price elasticity of housing supply relative to pre-crisis construction boom. - New house market: 2009–10 had very high levels of permits and sales, followed by a sharp decline during 2013–16; since 2017 higher permits were accompanied by slow growth of sales, which accelerated after the pandemic. - Existing house market: relationship between rehabilitation permits and sales is less direct but appears positively correlated after the European debt crisis. - Bottlenecks in the construction sector contributing to low supply elasticity: - Low and relatively stable productivity in construction over two decades, but productivity is relatively low compared to peers. - As of 2024, 86 percent of firms operating in the construction sector were micro-enterprises (BdP, 2025b). - Widespread skill shortages and mismatches. - Overly restrictive building regulation and lengthy building permit procedures. - Licensing for short-term tourist rentals is associated with lower sales of existing properties without necessarily boosting new construction. - Portugal has the highest share of dwellings not used as primary residence in OECD, yet the rental market remains small and inefficient, raising demand for house purchases. - The reliance on transaction taxes instead of recurrent property taxation disincentivizes the sale of underused properties (OECD, 2026). ### Assessment of vulnerabilities and links to financial sector stress - The 2024 ESRB report assessed residential real estate vulnerabilities as medium. - A rapid correction of house prices (§D2) is identified in the Risk Assessment Matrix as a domestic risk with medium overall level of concern: banks' high direct and indirect exposures to the real estate market make them vulnerable to a downward correction. - A rise in interest rates or an economic slowdown could impair borrowers’ repayment capacity and worsen banks' asset quality; contraction in consumption to support debt servicing could have second-round effects on growth. ### Analytical approach and model framework (selected methodological points) - Long-run relationship estimated between real house prices and real GNDI per household; using per household measure helps capture impact of declining household size on housing demand. - Short-term drivers estimated by regressing real house price growth on: residual from the long-run regression, unemployment rate, a measure of financial conditions, and a proxy for foreign demand for real estate. - Empirical analysis links LGD shocks for collateralized exposures to paths for real estate prices in stress-testing scenarios (smoothing factor applied to account for TTC regulatory approach). - Real estate analysis used publicly available data from INE, Banco de Portugal, and Eurostat; other inputs included household microdata (HFCS) and company-level data from Portugal’s Central Balance Sheet Database (CBSD). *Source: IMF staff analysis in the Portugal country chapter (figures and text as provided).* ### 4. Risks and ### 4. Risks and Buffers ### Risk Coverage and Sensitivity Analyses - Risks covered: credit (on loans and debt securities), market (valuation impact of financial instruments with respect to market risk factors such as interest rates, foreign exchange, credit spread, equity prices) and interest rate risk. - Sensitivity analyses performed: - Revaluation of debt securities portfolio held at amortized cost. - Quasi-reverse stress test on credit risk using more severe PDs. - Assessment of large exposures. ### Behavioral Adjustment (Modeling Assumptions) - Static balance sheet approach: size of portfolios (gross of NPLs) remains constant throughout the stress testing horizon (with no write-offs allowed). - RWAs projection: - Standardized and IRB portfolios differentiated. - Standardized portfolios: RWAs change due to shift in composition of performing and non-performing exposures. - IRB portfolios: through-the-cycle-PDs, downturn LGDs and EAD for each asset class/industry used to project risk weights. - Interest income from nonperforming loans is not accrued. - Dividends: - Paid out by banks that remain profitable and adequately capitalized. - Tax rate based on historical bank-level tax rate. - Dividend rate set at 80 percent to account for transfer of profits from foreign subsidiaries to the headquarter. ### Regulatory and Market-Based Standards and Parameters - Two hurdle rates considered: - “Minimum capital hurdle”: SREP capital requirements (the Basel 4.5 percent CET1 ratio) plus P2R. - “Combined buffers requirement”: minimum capital requirements plus the CCoB and O-SII capital buffers. - CCyB and sSyRB assumed to be zero in the adverse scenarios. - Leverage ratio assessed against the 3 percent Basel III minimum requirement during the stress test horizon. ### Reporting Form for Results — Solvency Outputs - Aggregate capital path for each scenario by groups of banks, categorized by bank groups. - Outputs include information on impact of different result drivers, including profit components. ### Banking Sector: Liquidity Test — Top-Down by FSAP Team Institutional Perimeter - Institutions included: 9 banks, of which 3 are SIs, 3 are LSIs and 3 are foreign bank subsidiaries. - Market share: 85 percent of the banking sector assets. Data and Scope - Data vintage: 2025 Q2. - Data source: Supervisory data from Implementing Technical Standards (ITS) files (COREP). - Scope of consolidation: highest level of consolidation for banks headquartered in Portugal and sub-consolidation level of all Portuguese activity for foreign subsidiaries. Methodology and Horizon - Structural analysis: evolution of LCR, NSFR, asset encumbrance, and funding structure. - Dynamic analysis: Cash flow-based stress test. - Stress test horizon for cash flow analysis: 1 year. Type of Analyses and Scenarios - Cash flow liquidity stress tests consider multiple stress scenarios including general risk aversion, market sovereign distress, and idiosyncratic funding to examine potential liquidity shortfalls from overnight up to one-year horizon. Buffers and Capacity - Buffers assessed: capacity of banks to generate liquidity from inflows and from assets under stress (i.e., counterbalancing capacity). Regulatory / Accounting and Market-Based Standards - LCR and NSFR hurdle rate set at 100 percent at the aggregate currency level (per Basel III and domestic regulation). - For cashflow analysis, outcomes of interest: Net Liquidity Position and the survival period. Reporting Format for Results — Liquidity Outputs - Output includes: 1. Descriptive analysis of LCR, NSFR, asset encumbrance, funding structure, and stock of CBC. 2. Net liquidity position over the horizon for the cash flow analysis. ### Appendix III — Scenarios (Macro and Market Shocks) - Severe scenario shocks (summary of assumptions): - Oil prices increasing by 100 percent starting in 2026, relative to the October 2025 baseline. - One-year-ahead inflation expectations ratchet up by as much as 100 basis points in advanced economies. - A significant risk-off episode pushes up risk premiums in advanced economies by 100 basis points. - The combined effect of the last two shocks results in an increase of 200 basis points in lending rates compared to the October 2025 baseline. - Food prices increase by 25 percent. - Appendix III. Table 1 (market scenarios) reports a range of market shock magnitudes across commodities, credit spread buckets (e.g., CR-low-5Y 85.5 bps, CR-high-5Y 206.8 bps), equity regional changes (e.g., EQ-US -38.2 percent with vol 56.0), and a set of absolute changes for EUR and USD tenors (e.g., EUR-1M 148.7 bps; EUR-1Y 178.7 bps; USD-1M 155.8 bps; USD-1Y 171.4 bps). (Appendix III. Figure 2 and Appendix III. Table 1 provide full scenario tables and series comparisons.) ### Appendix IV — Solvency Stress Test Specifications Net Interest Income (NII) — Data and Inputs - NII in the year before the cut-off date, 푁퐼퐼푏,푡=0, obtained from FINREP (F16.01). - Exposures and liabilities for each segment obtained from FINREP (F20.04 and F20.06). - Repricing ladder and starting interest rates for Portuguese loans provided by BdP using the credit register; loan exposures to Poland treated analogously to the EA FSAP (see IMF, 2025). - Projections for new business rate from satellite models using BdP and ECB “MFI Interest Rates Statistics” (MIR). Portfolio Segmentation and Repricing - Loans and deposits split across 18 bank portfolio segments. - Analysis covers 10 portfolio segments for loans, two for debt securities and one for remaining assets. - Mortgages segmented for fixed, mixed, or variable rate dynamics; variable rate mortgages further segmented by reference rates and repricing schedules. - Repricing ladder buckets used at T0: [0;1] yrs, [1;2] yrs, [2;3] yrs, [3;4+] yrs; shares denoted 휃0[...] = E0[...] / E0. NII Model — Three Steps (formulas and assumptions preserved) - Step 1: Simulate the law of motion of exposures across repricing buckets using: - Et[k-1,k] = Et-1[k,k+1] + It[k-1,k] - Shares of exposures across buckets assumed constant: 휃t[k,k+1] = 휃0[k,k+1] for all t,k. - Step 2: Simulate average interest rate per bucket: - rt[k-1,k] = ρt rt-1[k,k+1] + (1−ρt) rtnb where ρt = Et-1[k,k+1] / Et[k-1,k]. - Initial condition r0[k-1,k] = IIR0 (average interest income rate of the portfolio at T0). - Step 3: Calculate interest income IIt (case without NPEs): - II t = Σ_{k=0}^{3} rt-1[k,k+1] E0[k,k+1] + (1−ω)(rtnb − rt-1[0,1]) E0[0,1] - ω = avg days to repricing / 365. - For variable rates, rt = it + s and rtnb = itnb + s; spread s kept constant. - Mixed rates: loans with one year or less remaining maturity treated as variable rate; remaining use projected rtnb for mixed rates. - Incorporating NPLs: - Assumption: NPE ratio same across buckets. - Adjusted interest income: ĨIt = (1 − NPEr t) · II t. New Business Rates — Projection Method - New business rates projected via regressions estimating passthrough using quarterly macro data from WEO (2000Q1–2025Q2), BdP mortgage rates (2018Q4–2025Q2), and MIR deposit and loan data. - Regressions use short‑term policy rates, long‑term interest rates and their changes, inflation, and macro controls. - Empirical findings (high level): - Contemporaneous short-term rates and long-term rate changes generally positive (except mixed-rate mortgages). - Lagged interest rates usually insignificant or positive. - Inflation typically enters positively in loan regressions except corporate loans. - House prices negatively associated with consumer loan rates. - For deposit rates: short‑term rates and inflation main drivers; long‑term rates matter as proxy for alternative funding costs. Passthrough Estimations — Select Results - Appendix IV. Table 1 (Loans): key coefficient highlights include: - ∆ Short-term Rate coefficients: Mortgage Fixed 0.92***; Mortgage Variable 1.13***; Corporate Loans 0.56***. - Adjusted R2 values: Mortgage Fixed 0.96; Mortgage Mixed 0.85; Mortgage Variable 0.97; Corporate 0.85; Consumption 0.60. - Appendix IV. Table 2 (Deposits): selected coefficients: - ∆ Short-term Rate: Corporate Term Deposit Rate 0.65***; Households Term Deposit Rate 0.18**. - Adjusted R2 values: Corporate Sight 0.67; Households Sight 0.57; Corporate Term 0.89; Households Term 0.86. Net Fees and Commissions Income (NFCI) - Panel regression to project NFCI over total assets using small EA banks (excluding G-SIBs) 1995–2023 with bank fixed-effects. - Main drivers: real GDP growth and the EURIBOR one month delta (d.Euribor 1 month). - Starting NFCI: 0.72 percent of total assets. - Projected NFCI outcomes: drop of about 10 percent in the geopolitical scenario and 14 percent in the recessionary scenario. - Appendix IV. Table 3 (NFCI regressions): selected coefficients: - NFCI ratio (t-1): 0.723*** (All Regressors), 0.720*** (Lasso-selected), 0.469*** (Arellano-Bond). - Real GDP growth: 0.00599*** (All Regressors). - d.Euribor 1 month: -0.0201*** (All Regressors). ### Output Presentation and Reporting - Solvency outputs: aggregate capital paths, impact decomposition by result drivers, profit component breakdowns. - Liquidity outputs: descriptive metrics (LCR, NSFR, asset encumbrance, funding structure, stock of CBC) and Net Liquidity Position over the 1-year horizon. *Source: IMF staff compilation from "4. Risks and Buffers" (1prtea2026007 - 4. Risks and), 2025–2026 supervisory and staff data as presented in the chapter.* ### 7. For the House Price Tail shock sensitivity analysis, PDs were projected using a time ### 1prtea2026007 - House Price Tail Shock PD Projection and Stress Test Appendices ### PD projection methodology for the House Price Tail shock - PDs used in the House Price Tail shock sensitivity analysis are projected using a time series econometric approach (Appendix IV. Table 4 and Figure 3). - PD satellite models are estimated with quarterly year‑over‑year macroeconomic data from the WEO over 2000Q1–2025Q2, combined with sector‑level aggregated PD data from Banco de Portugal: - corporate PDs: 2000Q1–2025Q2 - mortgage and consumer loan PDs: 2010Q2–2025Q2 - Projections produced under baseline, recessionary, and geopolitical scenarios. - Model specifications favor parsimony. Estimated equations (variables preserved as in source): - (5) PD_t_corp = α + β1 GDP_g_t + β2 GDP_g_t * covid + β3 covid + β4 unempr_t−1 + β5 infl_t + β6 LT_rate_t−1 + ε_t - (6) PD_t_mort = α + β1 GDP_g_t + β2 GDP_g_t * covid + β3 covid + β4 unempr_t−1 + β5 infl_t + β6 LT_rate_t + ε_t - (7) PD_t_cons = α + β1 GDP_g_t + β2 GDP_g_t * covid + β3 covid + β4 unempr_t−1 + ε_t - Variable relevance by portfolio: - Mortgages: inflation and short‑term interest rates did not materially improve estimation or projections (attributed to growing share of fixed‑ and mixed‑rate contracts). - Corporates: inflation is informative, plausibly capturing nominal profit developments. - Consumer loans: PDs remain more sensitive to short‑term rates given prevalence of variable‑rate lending. - House prices (levels or growth) did not significantly affect mortgage PD estimates in the sample. ### Appendix IV. Econometric estimation key results (Table 4) - Corporate Loans PDs - Constant: -0.89 * - GDP Growth (annual): -0.11 - COVID (dummy): -0.60 - Unemployment (t-1): 0.52 *** - Inflation: -0.14 ** - Long-term Rate: 0.05 - GDP growth * COVID: 0.11 - Observations: 101 - Adjusted R^2: 0.83 - Residual Std. Error: 1.00 (df = 94) - F Statistic: 81.36 *** - Mortgage PDs - Constant: -0.03 - GDP Growth (annual): -0.02 - COVID (dummy): -0.17 * - Unemployment (t-1): 0.08 *** - Inflation: -0.02 ** - Long-term Rate: 0.04 ** - GDP growth * COVID: 0.02 - Observations: 61 - Adjusted R^2: 0.93 - Residual Std. Error: 0.13 (df = 54) - F Statistic: 142.87 *** - Consumption Loans PDs - Constant: 2.00 *** - GDP Growth (annual): -0.20 *** - COVID (dummy): -0.81 *** - Unemployment (t-1): 0.17 *** - GDP growth * COVID: 0.19 *** - Observations: 61 - Adjusted R^2: 0.80 - Residual Std. Error: 0.53 (df = 56) - F Statistic: 62.17 *** - Note on significance notation: * p<0.1, ** p<0.05, *** p<0.01 ### PD projections and sample considerations - Quarterly data provide sufficient observations despite shorter sample for mortgage and consumer PDs. - Across portfolios, unemployment and GDP are main drivers with expected signs; GDP is interacted with a COVID dummy to account for pandemic-related decoupling. - Appendix IV. Figure 3 contains historical and projected PD series for Mortgage PDs, Corporate PDs, and Consumer PDs. ### Solvency stress test—summary of detailed results (Appendix V) - Baseline (overall bank groups): - SIs: large NII and healthy fee generation with moderate credit-loss charges; gradual RWA growth reduces ratio; overall increase of 100 bps of capital buffers over the three‑year period. - LSIs: similar pattern; overall increase of 100 bps of capital buffers over the three‑year period. - Foreign Subsidiaries: overall increase of 130 bps of capital buffers over the three‑year period. - Adverse scenarios: - SIs: - Recessionary scenario: CET1 ratio ends up 220 bps below starting point (NII compressed, higher credit and market losses, large surge in RWAs). - Geopolitical scenario: CET1 capital ratio down by about 60 bps (revenues grow significantly but market risk losses and credit‑risk provisions multiplied by about three). - SIs suffer largest credit and market risk losses compared to LSIs and Foreign subsidiaries. - LSIs: - Recessionary scenario: NII compressed but sufficient to cover credit and market losses; resulting CET1 ratio ends at about same level as starting point. - Geopolitical scenario: CET1 capital ratio up by about 80 bps (revenues grow significantly; lower credit‑risk losses relative to SIs; stronger NII). - Foreign Subsidiaries: - Recessionary scenario: NII compressed but sufficient; CET1 ratio ends about same as starting point. - Geopolitical scenario: CET1 capital ratio up by about 190 bps (revenues grow significantly and more than compensate slightly increased credit‑risk losses). - Foreign subsidiaries display lowest starting level of CET1 ratio but generate more income over the scenario horizon than SIs and LSIs. ### Liquidity stress test—scenario design and parameters (Appendix VI) - Three cash flow stress scenarios aligned to solvency narratives: - Scenario 1: general risk aversion shock (mild tightening). - Scenario 2: market sovereign distress shock (larger haircuts, severe drawdown of liquidity facilities). - Scenario 3: idiosyncratic funding shock (severe deposit withdrawals across all funding sources). - Comparison with ECB 2019 LiST (Appendix VI. Table 1): - Term deposits (commercial/NFC): ECB 27-76% vs EA and PT FSAP 25-76% - Sight deposits (stable/commercial/NFC): ECB 18-74% vs EA and PT FSAP 5-49% - Sight deposits (financial counterparts): ECB 100% vs EA and PT FSAP 100% - Key assumptions underpinning scenarios: - (i) Constant interest rate (banks are not assumed to raise rates). - (ii) CBC components assumed fully monetizable through market or central bank facilities. - (iii) Open maturity outflows fully reflected overnight. - (iv) Contractual maturity outflows: same run-off rates across time buckets. - (v) Inflows: roll‑over assumptions follow EA FSAP; all inflows at 100 percent except loans to households, NFCs, and other borrowers set at 2-3 percent. - Detailed cash flow stress test parameters (Appendix VI. Table 2) — selected items (values shown in percent, by scenario): - Outflows examples: - Unsecured bonds due: 50 (Scenario 1), 100 (Scenario 2), 50 (Scenario 3) - Stable retail deposits: 5, 5, 8 - Other retail deposits: 10, 10, 40 - Operational deposits: 25, 15, 49 - Non-operational deposits from credit institutions: 100, 50, 100 - Non-operational deposits from other financial customers: 75, 50, 100 - Non-operational deposits from non-financial corporates: 25, 25, 76 - FX-swaps maturing: 50, 100, 100 - Inflows examples: - Level 1 central bank: 100, 100, 100 - Loans and advances granted to retail customers: 2, 2, 2 - Loans and advances granted to non-financial corporates: 2, 2, 2 - Loans and advances granted to credit institutions: 100, 100, 100 - Other inflows: 100, 100, 100 - Counterbalancing capacity examples: - Coins and bank notes: 100, 100, 100 - Withdrawable central bank reserves: 100, 100, 100 - Level 1 (CQS1): 98, 90, 98 - Level 1 (CQS2, CQS3): 95, 85, 95 - Level 1 (CQS4+): 75, 60, 75 - Non tradable assets eligible for central banks: 62, 50, 62 - Own issuances eligible for central banks: 62, 50, 62 ### Private‑Sector Balance Sheet Vulnerabilities—Household analysis (Appendix VII) - Data and aging: - Uses 2021 Household Finance and Consumption Survey (HFCS) for Portugal and other EA countries. - HFCS sample: 83,000 households and 200,000 personal files across 22 countries (20 EA countries, Czech Republic, and Hungary). - Households’ balance sheets, incomes, and expenditures are “aged” to end-2024. - Microdata-driven simulation framework: - Simulates macro‑financial shocks (interest rate increases, income declines, inflation surges, commodity price shocks) and transmits to household finances. - Incorporates Monte Carlo simulations of unemployment shocks at individual level, using Portugal Social Security microdata and Portugal Labor Force survey to model differential likelihood by contract type (fixed‑term vs permanent). - Link to bank credit risk: - Household financials mapped to loan performance using logistic regressions. - DSECTI metric defined: Debt Service plus Essential Consumption to Income (includes mortgage payments, rents, and essential expenses such as food, energy). - Critical threshold: 70 percent DSECTI noted as critical—when essential payments and debt obligations exceed 70 percent of disposable income, probability of default rises sharply. - This “overburden” benchmark is significantly above traditional 40 percent DSTI rule. - In Portugal, likelihood of falling into arrears roughly doubles from 4.5 percent to 9.2 percent once a household becomes overburdened. - A “stretched consumer” is flagged if essential expenses plus debt service exceed income. - Scenarios and shocks: - Baseline: July 2025 WEO projections. - Adverse scenarios: FSAP “geopolitical” scenario (escalation of geopolitical tensions causing higher interest rates and commodity prices), “recessionary scenario” (synchronized global slowdown, sovereign distress, rating downgrades). - Sensitivity scenarios: include a “cost‑of‑living” scenario with combined 200 bps interest rate increase and a 20 percent rise in food and energy prices relative to baseline. - Shocks applied over stress test horizon: 2026–2028. - Household-level indicators (notably DSECTI) computed under each scenario; share of households crossing 70 percent threshold projected. *Source: IMF staff calculations and Banco de Portugal data as presented in the supplied content.* ### 4. Using these estimates, the model projects the migration of banks’ loan exposures ### 4. Using these estimates, the model projects the migration of banks’ loan exposures across IFRS 9 stages ### Household-to-bank IFRS 9 migration and integration with bank stress testing - IFRS9 stage mapping to household balance sheet metrics: - Stage 1: Households not in arrears and not struggling (cost-adjusted-debt service-to-income ratio below 70 percent). - Stage 2: Households struggling (cost-adjusted-debt service-to-income ratio above 70 percent, but not on arrears). - Stage 3: Households in arrears, irrespective of their financial ratios. - Purpose: estimate how household shocks (higher rates and living costs) lead to migration from Stage 1 to Stages 2 and 3, producing increases in bank provisions and expected credit losses. - Methodological note: conservative approach—some banks use arrears exceeding 90 days as a proxy for Stage 3 rather than classifying exposures as Stage 3 based solely on any arrears. ### Corporate sector vulnerability analysis: data and coverage - Dataset: panel of 40 sector-size combinations over 2006-24 from Portuguese Central Balance Sheet Database (CBSD) and Central Credit Register. - Sector classification (10 sectors): (A) Agriculture, forestry, and fishing; (B+C) Manufacturing; (D+E) Electricity, gas, and water; (F+L) Construction and real estate activities; (G) Trade and repair of vehicles; (H) Transport and Storage; (I) Accommodation and food service activities; (J) Information and communication; (M-70100+N) Professional and administrative activities; (P+Q+R+S) Education, health and other activities. - Size categories (4): (1) Micro; (2) Small; (3) Medium; (4) Large, defined consistent with the methodology of the European Commission. - For each variable, the mean in each of the 40 categories is used—variation within a category is not considered. ### Dynamic analysis: fixed-effects model and prediction setup - Dependent variables (Y_i,t): profitability (return on assets), leverage (ratio of total debt to total assets), and year-on-year sales growth. - Model specification (fixed-effects, with notation preserved): - Y_i,t = α∙Y_i,t−1 + β∙{Firm}_i,t−1 + γ∙{Macro}_t + d_s + π_z + ν_i,s,z,t - {Firm}_i,t−1 includes asset tangibility (ratio of fixed assets to total assets) and cash flow generation capacity (ratio of sales to total assets). - {Macro}_t includes annual real GDP growth and unemployment rates and a dummy variable for covid. - d_s and π_z denote sector and size fixed effects; ν_i,s,z,t is the error term, clustered for within-group correlation. - Predictions: - Y_i,t predicted recursively for three scenarios differing by real GDP growth and unemployment paths for 2026-28. - Structural firm characteristics and estimated sector and size fixed effects are held constant at their 2024 levels. - Data limitation: as the data end in 2024, values for 2025 are predicted as in the baseline scenario. ### Derived financial indicators and interest rate pass-through - Effective interest rate prediction (notation preserved): - EIR_i,t = EIR_i,t−1 + [ (LTD/TD)_i,2024 ∙ ΔLTI_t + (1−(LTD/TD))_i,2024 ∙ ΔSTI_t ] - ΔLTI_t and ΔSTI_t are changes of long-term and short-term interest rates from their 2024 levels in year t under each scenario. - LTD/TD ratio is held constant at its 2024 level. - Approximation linking EBIT and ROA: - EBIT_i,t ≈ ROA_i,t ∙ TA_i,t−1 - Interest coverage ratio approximation: - ICR_i,t = EBIT_i,t / Interest Expense_i,t ≈ ROA_i,t [ (TD/TA)_i,t−1 ∙ EIR_i,t ] ### Probability of default estimation (zero-inflated beta regression) - PD model structure (preserving notation and conditions): - For PD_i,t > 0: - PD_i,t = α D_ICR_i,t ∗ (TD/TA)_i,t−1 + β (TD/TA)_i,t−1 + γ (Real growth_t) + τ (Unemployment_t) + π (Covid_t) + ω (Covid_t)∗(Real growth_t) + δ_s + μ_z + ε_i,s,z,t - For PD_i,t = 0: - PD_i,t = θ ROA_i,t + ρ (TD/TA)_i,t−1 + σ (Covid_t) + δ_s + μ_z + ε_i,s,z,t - Definitions and interpretation: - PD_i,t is probability of default in category i in year t (continuous between 0 and 1). - D_ICR_i,t is a dummy equal to one when ICR < 1. - Interaction of leverage and ICR dummy captures nonlinear increase in default risk when high leverage coincides with low debt servicing capacity. - Interaction between GDP growth and covid dummy captures effects of policy support during covid. - For zero defaults, profitability (ROA) is important. - Sector and size fixed effects are significant. ### Static sensitivity analysis: firm-level shocks, sample, and mechanics - Sample construction: - Firm-level data across same sectors as dynamic analysis. - Firms with no employees and/or no assets and/or no sales and/or no interest payments were dropped. - Resulting sample covers: 42 percent of all firms, 81 percent of employees, 70 percent of assets, 85 percent of sales, and 70 percent of total debt. - Caveat: firms treated as separate entities; intragroup financing treated equivalent to other financing. - Shock computation and behavioral parameters: - Cash impact of sales shock: - Cash impact of sales shock = x% ∙ [ Net sales − y% ∙ (COGS + SES) − z% ∙ Wages ] - Post-shock interest expense: - Interest Expense_post shock = Interest Expense_pre shock ∙ (1 + ie%) - Definitions: - x is share of sales lost; ie is change in the effective interest rate in adverse scenarios; y and z are shares of wage and other production costs firms adjust. - Behavioral calibrations from Bank of Portugal: y = 95% and z = 60%. - Shocks x and ie are calibrated as sectoral bank debt-weighted averages of peak annual impacts under each scenario estimated in the dynamic analysis. - Post-shock metrics: - Post-shock ICR: - ICR = (EBIT_pre shock − Cash impact of sales shock) / Interest Expense_post shock - Post-shock cash balance: - Cash Balance_post shock = Cash Balance_pre shock + EBIT_pre shock − Cash impact of sales shock − Interest Expense_post shock - Assumption: firms do not obtain new financing or retain more earnings to alleviate cash stress. ### Real estate market analysis: long-term and short-term drivers - Long-term relationship testing: - Two OLS models regressing real house price index (deflated by CPI) on: - real GNDI per household (logs), and - real GDP per household (logs). - Finding: both models show significant correlation; GNDI model has higher explanatory power. - Cointegration tests: - Real house prices cointegrated with real GNDI per household, but not with real GDP per household. - Current deviation: - Real house prices are about 21 percent higher than predicted by real GNDI per household. - Real house prices are about almost 50 percent higher than predicted by real GDP per household. - Short-term drivers regression setup: - Regress real house price growth on residual from long-run regression and: - lagged unemployment rate, - proxy for foreign demand (ratio of FDI in real estate sector to GDP, moving average over 3 periods), - measure of financial conditions (first principal component of several financial variables), - dummy = 1 starting 2024Q3 for government measures supporting young homebuyers. - Model implications and elasticities (results summary): - Lagged residual from long-term regression has expected negative sign but is not consistently significant; excluding it does not meaningfully reduce explanatory power. - Unemployment: - Increasing unemployment rate to 10 percent (compared to about 6 currently) leads to 3 percentage points lower real house price growth, holding other factors constant. - Foreign demand: - Reduction of real estate FDI to GDP ratio by 0.1 percentage points reduces real house price growth by 7 percentage points. - Financial conditions: - Tightening by 1 standard deviation reduces real house price growth by 4 percentage points. - Policy dummy: - Positive and significant coefficient suggests measures introduced were associated with higher house price growth. - Housing supply determinants (IV regressions on sales growth of houses): - Regressors: house price growth (instrumented), number of local accommodation licenses issued over past year, ratio of new housing construction cost index to CPI over past two years. - Instruments: lagged growth of GNDI per household and lagged ratio of FDI in real estate to GDP. - Findings: - For new houses, estimated price elasticity is lower and model fit is poorer. - Construction cost coefficient over entire sample not significant for new house sales. - Negative and significant coefficient on touristic licenses implies short-term rentals reduce supply of houses available for sale. - Policy dummy not significant—measures have not necessarily led to additional house purchases. *Source: IMF staff calculations and analysis as presented in the cited chapter.* --- _Source: https://www.imf.org/-/media/files/publications/cr/2026/english/1prtea2026007.pdf_