## cr18344 — EXECUTIVE SUMMARY ## Source details **Canonical URL:** [cr18344 — EXECUTIVE SUMMARY](https://www.imf.org/-/media/files/publications/cr/2018/cr18344.pdf) ## Other formats - [Markdown version](/-/media/files/publications/cr/2018/cr18344.pdf.md) - [Structured JSON version](/-/media/files/publications/cr/2018/cr18344.pdf.json) --- ### Financial system resilience and structure - System dominated by banks and investment funds. - Banking sector: - Continued to be well-capitalized, profitable, and liquid. - Profitability supported by prudent lending standards, high interest margins and robust fee income, despite record loan losses. - Outstanding nonperforming loans increased marginally during the recession largely because banks actively wrote off bad loans. - CET1 capital ratio for the banking sector: 12.7 percent at end-2017Q2. - System-wide return on assets: 1.3 percent as of 2017Q3. - Nonperforming loans (NPLs): 3.7 percent of total loans at end-2017Q3. - Impairment cost: 4.2 percent of total assets. - Provisioning coverage: 160 percent. - Deferred tax assets (DTAs) account for almost 30 percent of CET1 capital. - External funding exposures low (around 10 percent of total funding). - Net foreign exchange positions are small. - Investment fund industry: - Solid, steady growth of assets under management; open-ended funds account for about 78 percent of total net assets (TNA). - Fixed-income funds dominate (about 82 percent of open-ended TNA on average). - Aggregate cash liquidity position: R$922 billion. ### FSAP focus and macrofinancial risk drivers - Main FSAP modules: - Bank solvency stress tests for 12 largest banks (accounting for 90 percent of banking sector assets). - Bank liquidity stress tests: LCR-, NSFR-based and cash-flow analyses. - Interbank contagion and interconnectedness analyses. - Corporate and household sector vulnerability assessments. - Investment fund redemption and government bond market impact analysis. - Cross-border spillovers using market data. - Three principal macrofinancial risks guiding analysis: - Weak domestic growth due to loss of confidence. - Tighter and more volatile global financial conditions. - A significant slowdown in China with spillovers. ### Macrofinancial scenarios underpinning the financial stability analysis - Two scenarios used: - Baseline: October 2017 WEO projections; real GDP growth reaching 2 percent during 2019–20; credit growth around 8 percent annually during 2018–20; interest rates on downward trend. - Adverse: simulated with the Global Macrofinancial Model (GFM); real GDP falls 11.8 percent below baseline by 2018 (3-year cumulative growth 2 standard deviation below baseline); features sudden stop of capital inflows, 50 percent real exchange rate depreciation, long-term government bond yield increase by 700 basis points, real equity price decline by 30 percent, real house price fall by 7.5 percent, Selic path of 11.0 (2018), 8.1 (2019), 4.5 (2020). ### Bank solvency stress tests — baseline and adverse scenario results - Exercise scope: - Data reference date: end-2017Q3. - Scenario horizon: three years, 2017Q4–2020Q4. - Sample: 12 largest banks. - Key baseline projections: - CET1 capital ratio for the 12 largest banks falls from 13.2 percent at end-2017Q3 to 11.4 percent at end-2020. - One public D-SIB may fall below hurdle rates; additional capitalization need R$6 billion (0.1 percent of GDP). - Drivers: phase-in deductions (-0.2 percentage points), dividend payout (-2.1 percentage points), balance sheet growth (-2.7 percentage points); banks generate net income boosting capitalization by 3.4 percentage points. - Adverse scenario (benchmark, no adjustments): - CET1 capital ratio of the 12 largest banks declines by 4.7 percentage points to 8.5 percent at end-2020. - Net interest income falls by 21 percent relative to baseline; credit loss increases by 23 percent relative to baseline. - Credit loss alone roughly equals pre-impairment income (both about 11.2 percent of risk-weighted assets). - Market loss (including unrealized loss on debt securities) amounts to 1.5 percent of risk-weighted assets over the horizon; market loss is 3.6 percent of risk-weighted assets in 2018. - Three banks’ capitalization fall below hurdle rates at end-2020; another bank fails temporarily but recovers by end-2020. - Additional capital required: R$46 billion (0.7 percent of GDP), nearly all for two public banks. - Treatment of DTAs: - DTAs ≈ 30 percent of CET1 capital. - Deduction of all existing DTAs reduces CET1 by 3.1 percentage points (assumes risk weight 2.5). - Under DTA deduction, CET1 would be 5.8 percent at end-2020 in adverse scenario (vs 8.5 percent with no adjustment). - Combined shocks (DTA deduction + downgrading performing problem loans) reduce CET1 by 3.3 percentage points immediately; adverse scenario CET1 falls to 5.5 percent at end-2020. - Under combined shocks: seven banks fail to meet hurdle rates at some point; additional capitalization need R$136 billion (2 percent of GDP), about three quarters for public banks. - Integrated solvency + interbank contagion: - Adverse scenario with interbank contagion: CET1 falls to 7.1 percent at end-2020. - Interbank contagion adds R$68 billion loss (1.5 percent of risk-weighted assets). - Six banks’ capitalization fall below hurdle rates at some point; additional capitalization need R$81 billion (1.2 percent of GDP). - Sensitivity findings: - Downgrading all performing problem loans to "E" reduces CET1 by 0.2 percentage points immediately. - Market shocks: bond yield spike by 700 basis points reduces CET1 by 1.1 percentage points (market loss only); equity price fall of 50 percent reduces CET1 by 1.2 percentage points. - Interest-rate sensitivity: bond yield spike by 700 basis points could reduce CET1 by 3.4 percentage points (All), with unrealized AFS losses and trading losses being key contributors. - Exchange rate depreciation by 50 percent reduces CET1 by 1.8 percentage points. - Comparison with BCB top-down tests: - BCB baseline: CET1 projected to 15.4 percent at end-2020 (differences due to no dividend payout, lower credit loss and taxes). - BCB adverse: CET1 projected to 13.5 percent at end-2020; two banks failing with capital shortfalls of R$15 billion. ### Bank liquidity resilience and funding risks - Short-term liquidity: - System-wide LCR above 100 percent since end-2015. - Aggregate LCR (September 2017): 203 percent for 10 largest commercial banks; 248 percent for all 12 banks. - Aggregate IL (BCB internal indicator) (September 2017): 271 percent for 10 largest commercial banks; 275 percent for all 12 banks. - LCR-based adverse scenario: - Assumptions: retail deposit losses between 10 and 30 percent in one month; unsecured wholesale funding reduced by 71 percent; secured wholesale funding reduced by 10 percent; haircut of 9.35 percent on government securities. - One D‑SIB would see LCR below 80 percent with liquidity shortfall R$31 billion (less than 1 percent of the bank’s assets and less than 1 percent of GDP); aggregate LCR falls to 122 percent for 10 banks and 145 percent for 12 banks. - ELA access (not modeled) would be required to cover remaining gap for that D‑SIB. - FX LCR: - Baseline USD LCR below 35 percent for three out of eight largest commercial banks; FX shortfalls small (equivalent to less than 1 percent of foreign reserves; R$9.3 billion). - NSFR-based results: - Available stable funding: R$4,027 billion; required stable funding: R$3,486 billion; aggregate NSFR: 116 percent (baseline). - Adverse NSFR: aggregate NSFR 95 percent; stable funding gap R$288 billion (4.6 percent of GDP; 4.7 percent of banks’ assets). - NSFR of 5 D-SIBs falls to 91 percent (four out of five in range 80–91 percent); stable funding gap attributable to D‑SIBs: R$227 billion. - Cash-flow analysis: - Maturity buckets from overnight up to over 6 months; counterbalancing capacity includes access to standard BCB facilities and conditional ELA. - In adverse scenario, three banks face liquidity shortfalls within 180 days: one D‑SIB (shortfall within 30 days given group-wide deposit outflow of 21 percent in one month and 25 percent in six months) and two non D‑SIBs for 180-day horizon. - If drawdown of required reserves allowed, these banks could remain liquid for 180 days; if reserves kept, ELA alone insufficient. - FX cash-flow: seven banks would deplete FX liquidity buffers within six months; FX shortfall after sales R$52 billion (4.3 percent of foreign reserves). - Reputational crisis scenarios: - Localized crisis: aggregate LCR falls to 119 percent for 10 largest commercial banks; one D‑SIB with shortfall R$25 billion. - System-wide crisis (fire sales): two D‑SIBs shortfalls R$56 billion; aggregate LCR falls to 107 percent for 10 largest commercial banks. - Policy implications: - Liquidity stress tests confirm strong buffers overall but identify some banks at risk in extreme scenarios. - Strengthen BCB’s liquidity provision framework including ELA. - Perform regular cash-flow based liquidity stress tests in local and foreign currencies and align data sources across liquidity metrics. ### Interbank contagion and interconnectedness - Interbank exposures: - Interbank exposures among 12 largest banks: 2.4 percent of banking system assets and 2.8 percent of GDP. - About 67 percent of interbank exposures are secured (mostly overnight repos collateralized by government securities). - Six large banks combined are net borrowers; smaller banks are net lenders. - Stand-alone network simulations: - Default of one large bank could trigger failure of another bank in limited cases (e.g., bank 1 default triggered default of bank 9 in simulation). - Hypothetical default of large bank 5 reduces system-wide total capital by around 11 percent; large bank 1 reduces system-wide capital by around 10 percent. - Index of vulnerability up to 18 percent (bank 9). - Sensitivity: - Baseline LGD = 50 percent; if LGD = 100 percent, contagiousness and capital impairment increase substantially (e.g., large bank 5 failure could impair 25.5 percent of other banks’ capital). - Funding replacement assumptions materially affect contagion outcomes. - Integrated solvency + interbank contagion exercise increases number of banks failing hurdle tests and additional capitalization needs (see solvency section). - Cross-sector interconnectedness (flow-of-funds, 2016Q4): - Total assets/liabilities: 43,084,344 BRL Million. - Equity and investment funds shares/units: 15,001,190 BRL Million (34.8 percent). - Debt securities: 8,604,486 BRL Million (20.0 percent). - Currency and deposits: 7,256,700 BRL Million (16.8 percent). - Loans: 6,335,062 BRL Million (14.7 percent). - Government plays a central role via public banks, repos collateralized by government securities, and government securities as main liquid instrument. - Banks are most interconnected sector; investment funds link strongly with banks via repos and with insurers/pension funds via funding and investment channels. ### Investment funds and potential government bond market effects - Dataset: universe of 555 funds; runnable open-ended sample accounts for 52 percent by count and 78 percent of TNA (September 2017). - Three calibration approaches for extreme redemptions: - Fund-homogeneity: common extreme shock per fund family (1st percentile of family flow rates). - Fund-heterogeneity: each fund faces its own 1st percentile outflow simultaneously. - Fund-family: 1st percentile of combined outflow of all funds within family. - Calibrated redemption rates (1st percentile examples): - Fund-homogeneity: 25 percent (fixed-income), 14 percent (multi-market), 8 percent (stocks). - Fund-family: 1.9 percent (fixed-income), 2.2 percent (multi-market), 1 percent (stocks). - Simulated asset sales to meet redemptions (waterfall schedule): - Funds would use reverse repos to meet extreme redemptions: - Fund-homogeneity: R$437 billion of reverse repos used. - Fund-heterogeneity: R$215 billion of reverse repos used. - Fund-family: R$77 billion of reverse repos used. - Followed by total sale of government bonds: - Fund-homogeneity: R$194 billion (waterfall) — pro rata R$298 billion. - Fund-heterogeneity: R$34 billion (waterfall) — pro rata R$88 billion. - Fund-family: R$10 billion (waterfall) — pro rata R$39 billion. - Cash liquidity coverage of total redemptions (waterfall): - Fund-homogeneity: 67 percent; Fund-heterogeneity: 83 percent; Fund-family: 85 percent. - NAV at risk (share of funds with insufficient cash liquidity): - Fund-homogeneity: 52 percent; Fund-heterogeneity: 21 percent; Fund-family: 20 percent. - Price effects (first-round, Amihud-based), waterfall sales: - Under 2017 market conditions, average rise in government bond yields: - Fund-homogeneity: 62 basis points. - Fund-heterogeneity: 21 basis points. - Fund-family: 8 basis points. - Under 2015 market conditions (more severe liquidity): - Fund-homogeneity: 99 basis points. - Fund-heterogeneity: 28 basis points. - Fund-family: 11 basis points. - Waterfall sales concentrate sales into more liquid short-term bonds and could significantly raise short-term yields (example: NTN-B short-term declines of 186, 88 and 36 basis points under fund-homogeneity, fund-heterogeneity and fund-family calibrations, respectively). - Regression on redemption sensitivity to returns: - Flow rate equation results show insignificant sensitivity of investors to short-term returns once controls included; adjusted R^2 = 0.07 percent; lag RETURN and RETURN coefficients not significant. - Interpretation: second-round redemption effects from return changes are economically weak in the Brazilian investment fund industry. - Conclusion: - Investment funds could contribute to short-term price effects in government bond market under large-scale redemptions; magnitude critically depends on calibration assumptions and market liquidity conditions. ### Corporate sector vulnerabilities and analysis - Corporate sector shows moderate recovery: - Median ROE improved to 8.5 percent in 2017Q3 (down from pre-recession levels). - Median ICR increased to 2 during the recession from 1.5 but remains below pre-recession ICR of 3. - Median total corporate debt relative to equity: 0.7 in 2017Q3 (peaked at 0.8 in 2015). - 40 percent of firms in 2017Q3 could not generate sufficient EBITDA to cover interest expenses. - 1/3 of firms remain highly leveraged with total debt exceeding twice equity. - Sectoral debt-at-risk findings: - Services sector: debt at risk (ROE below Selic) exceeded 80 percent; more than 2/3 of debt in services sector have ICR below 1.5; more than half of services-sector debt has net debt to EBITDA > 5. - Manufacturing and energy show lower debt-at-risk shares (manufacturing < 50 percent; energy 5–10 percent by some metrics). - Sensitivity analysis to shocks consistent with bank adverse scenario: - Interest rate shock (8 percent increase in market interest rate) substantially increases share of debt with ICR < 1. - FX shock (43 percent) increases debt at risk by 4 (inclusive hedge) or 2 (conservative) percentage points. - Profitability shock (10 percent decline) increases debt at risk by 14 percentage points — largest impact. - Policy implication: - Review corporate insolvency regime to facilitate efficient debt restructuring and recovery. - Require banks to assess sensitivity of borrowers’ debt servicing capacity to macrofinancial shocks. ### Household sector vulnerabilities and debt-at-risk analysis - Household debt and composition: - Household debt to disposable income: 41¼ percent as of November 2017 (peak 46½ percent in April 2015). - Mortgage borrowing accounts for about 45 percent of total household debt. - Flexible-rate loans account for 43 percent of bank lending to households. - Households spend about 20 percent of income on principal and interest payments. - Debt-at-risk metrics (as of 2017Q3, BCB credit registry): - Debt service-to-income above 40 percent: Share of financially weak households 20.2; Share of debt at risk 52.1; Share of debt at risk not covered by assets 24.5. - Debt-to-income above 3: Share of financially weak households 3.6; Share of debt at risk 25.8; Share of debt at risk not covered by assets 3.4. - Combined metric: Share of financially weak households 2.9; Share of debt at risk 22.5; Share of debt at risk not covered by assets 2.9. - Trends and scenario sensitivity: - Share of financially weak households edged up by 0.3 percentage points since 2013 (combined metric). - Scenario (declining income 8 percent, interest rates +250 basis points, house price −10 percent): - Combined metric: Share of financially weak households rises to 3.9; Share of debt at risk 27.7; Share of debt at risk not covered by assets 3.7. - Even under adverse scenarios estimated bank credit loss expected to be relatively limited given collateral coverage and ability to execute collaterals. - Policy recommendations: - Consider implementing limits on debt-to-income and debt service-to-income for mortgage lending. - Require banks to perform borrower-level stress testing for debt servicing capacity. ### Cross-border spillovers and market-based connectedness - Brazil experiences both inward and outward spillovers in equity, sovereign credit risk, and FX markets. - Patterns: - Equity and sovereign CDS: regional spillovers within Latin America; inward spillovers from Europe and the United States. - FX: concentrated among large, financially integrated emerging market economies; Brazil generates outward spillovers to smaller Latin American economies. - Diebold-Yilmaz based connectedness: - VAR with 10-day forecast horizon, sample 2007Q4–2017Q3, countries include Argentina, Chile, China, Colombia, France, Germany, India, Indonesia, Japan, Mexico, Russia, South Africa, Spain, Turkey, UK and US. - Brazil is linked to advanced economies, with US acting as global hub. ### Key policy recommendations (selected; Immediate to Near-term unless specified) - Use Pillar 2 capital requirements to address bank-specific risks and concentration (Immediate). - Encourage banks with stress-test capital shortfalls to build capital buffers (e.g., restrict dividend distribution) (Immediate). - Ensure accuracy and consistency of information across data sources for reliable analyses (Immediate). - Conduct regular cash-flow analyses to assess banks’ liquidity risk in local and foreign currencies (Immediate). - Continue vigilant monitoring of corporate and household vulnerabilities (Immediate). - Implement limits on debt-to-income and debt service-to-income for mortgage lending (Near-term). - Validate results produced by banks in annual bottom-up stress testing and improve banks’ risk management capacity (Near-term). - Adopt integrated approach to assess financial-sovereign linkages, including large holdings of government bonds and investment fund redemption effects (Near-term). - Review corporate insolvency regime to ensure efficient debt restructuring and recovery processes (Near-term). - Require banks to assess borrower sensitivity to macrofinancial shocks (Near-term). - Develop solvency-liquidity contagion module in stress-testing framework (Medium-term). - Improve handling of overseas operations and estimate credit loss via default probability and loss given default in stress testing (Medium-term). - Enhance data validation and consistency across supervisory and market data sources (Near-term to Medium-term). *Source: EXECUTIVE SUMMARY (cr18344) — Technical Note prepared for the FSAP analysis.* ### EXECUTIVE SUMMARY __________________________________________________________________________ 8 ### EXECUTIVE SUMMARY ### INTRODUCTION - Structure of the FSAP chapter: - A. Financial System Landscape - B. Macrofinancial Environment - C. A Bird’s-Eye View on Systemic Risk - D. Scope of the Financial Stability Analysis in the FSAP ### MACROFINANCIAL RISKS UNDERPINNING THE FINANCIAL STABILITY ANALYSIS - Major components: - A. Overview of Key Macrofinancial Risks - B. Macrofinancial Scenarios Underpinning the Financial Stability Analysis - Related exhibits: - Figure 1. Macrofinancial Developments - Figure 2. A Bird’s-Eye View on Systemic Risk - Figure 3. Key Macrofinancial Variables in the Baseline and Adverse Scenarios - Table 2. Risk Assessment Matrix (RAM) - Appendix I. Characterization of Baseline and Adverse Scenarios ### BANK SOLVENCY STRESS TESTS - Coverage: - A. Overview - B. Recent Banking Sector Performance - C. IMF Top-down Stress Tests: Methodology - D. IMF Top-down Stress Tests: Results - E. Top-down Stress Tests by the Authorities - F. Assessment of the BCB’s Solvency Stress Testing Framework - G. Policy Recommendations - Key supporting items: - Figure 4. Banking Financial Soundness Indicators - Figure 5. Selected Emerging Market Economies: Banking Financial Soundness Indicators - Figure 6. Estimated Average Interest Rates for Bank Solvency Stress Tests - Figure 7. Estimated Impairment Cost for Bank Solvency Stress Tests - Figure 8. Bank Solvency Stress Test Results - Figure 9. Comparison of Results of Bank Solvency Stress Tests Conducted by the FSAP Team and the BCB - Table 3. Banking Financial Soundness Indicators - Table 4. Hurdle Rates of the Bank Solvency Stress Tests - Table 5. Stylized Impact on Capital Given the Existence of DTCs - Table 6. Bank Solvency Stress Test Results – Scenario Analysis - Table 7. Bank Solvency Stress Test Results—Sensitivity Analysis - Appendix II. Stress Testing Matrix (STeM) - Appendix III. Detail on Balance Sheet and Income Statement Projections for the Bank Solvency Stress Tests ### BANK LIQUIDITY STRESS TESTS - Coverage: - A. Overview - B. Current Liquidity Conditions and Banks’ Liquidity Profiles - C. LCR-Based Stress Tests - D. NSFR-Based Stress Tests - E. Cash-Flow Analysis - F. Reputational Crisis Scenario - G. Policy Recommendations - Key supporting items: - Figure 10. Liquidity Conditions - Figure 11. Funding Structure of Brazil’s Banking System - Figure 12. The Regulatory LCR Test by Currency - Figure 13. The LCR and NSFR Tests - Figure 14. Short-Run Liquidity Results: Brazil’s IL and LCR - Figure 15. LCR in Reputational Crisis - Table 8. Brazil: Summary of the LCR-Based Stress Test Results - Table 9. Short-Run Liquidity Results: Brazil’s IL Versus Regulatory LCR - Table 10. Summary of the NSFR-Based Stress Test Results - Table 11. Cash-flow Stress Test Result in Domestic Currency - Table 12. Cash-flow Stress Test Result in Foreign Currency - Appendix IV. Calibration of Parameters for Bank Liquidity Stress Tests ### BANK CONTAGION ANALYSIS - Coverage: - A. Interbank Network - Key supporting items: - Figure 16. Interbank Assets and Liabilities - Figure 17. Interbank Exposures - Figure 18. Result of the Simulation Study of 12 Largest Brazilian Banks: Sensitivity to the LGD Assumption - Figure 19. Result of the Simulation Study of 12 Largest Brazilian Banks: Sensitivity to the Funding Replacement Assumption - Figure 20. Result of the Simulation Study of 135 Brazilian Banks: Sensitivity to the Funding Replacement Assumption - Figure 21. Result of the Simulation Study of 135 Brazilian Banks: Sensitivity to the LGD Assumptions - Appendix V. Detailed Methodology for the Interbank Network Analysis - Appendix VI. Network Analysis Measures and Concepts ### CORPORATE SECTOR ANALYSIS - Coverage: - A. Overview - B. Recent Developments - C. Vulnerability Assessment by Sector - D. Sensitivity Analysis - Key supporting items: - Figure 22. Corporate Financial Soundness - Figure 23. Vulnerabilities in Corporate Sector - Figure 24. Sensitivity Analysis of Debt Servicing Capacity in the Corporate Sector ### HOUSEHOLD SECTOR ANALYSIS - Coverage: - A. Overview - B. Recent Development - C. Household-Level Balance Sheet Analysis - D. Policy Recommendations - Key supporting items: - Figure 25. Household Financial Soundness - Figure 26. Vulnerabilities of Household Balance Sheets - Table 13. Sensitivity and Scenario Analysis of Household Debt at Risk ### ANALYSIS OF POTENTIAL IMPACT ON SOVEREIGN BOND MARKET STEMMING FROM INVESTMENT FUNDS - Coverage: - A. Introduction - B. Data, Methodology, and Results - Key supporting items: - Figure 27. Share of TNA of Open-Ended Funds (By Fund Families) - Figure 28. Median Weekly Returns (By Fund Family) - Figure 29. Historical Distribution of Flow Rates (By Fund Family) - Figure 30. Historical Distribution of Redemption Rates Per Individual Fund (By Fund Family) - Figure 31. Portfolio (By Fund Family) - Figure 32. Historical Distribution of Weekly Price Changes of Government Bonds - Table 14. Exclusion of Funds From the Sample - Table 15. Asset Composition (Panel A) and Asset Sales (Panel B) (R$ Billion) - Table 16. Comparison of the Scale of Government Bond Sales by Investment Funds Relative to Weekly Turnover (R$ Billion) - Table 17. Price Effects from Sale of Government Bonds by Funds to Meet Redemptions - Table 18. Regression Estimate Modeling Investor Redemption Behavior in Terms of Returns ### INTERCONNECTEDNESS ANALYSIS - Coverage: - A. Overview - B. Cross-Sectoral Linkages - C. Cross-Border Spillovers Based on Market Data - Key supporting items: - Figure 33. Cross-Sector Network in 2016 Q4 - Figure 34. Selected Cross-Sector Subnetworks in 2016 Q4 - Figure 35. Diebold-Yilmaz Outward Spillover Matrices - Figure 36. Summary Diebold-Yilmaz Outward Spillover Matrices for Individual Banks - Table 19. Financial Accounts by Financial Instruments in 2016 Q4 - Table 20. Financial Accounts by Institutional Sector in 2016 Q4 - Appendix VII. Data for Market-based Spillover Analysis ### SUPPORTING MATERIALS, BOX, AND APPENDICES - Box 1. Operational Hedging in Brazilian Corporate Sector - Appendices I–VII covering scenarios, stress testing matrices, balance sheet projection details, liquidity parameter calibration, network methodology, network concepts, and data for market-based spillover analysis. - Glossary of abbreviations includes: AFS, AOCI, BCB, BCBS, BIS, BNK, CCB, CCP, CDI, CDS, CET1, CVM, D-SIB, DSC, DTA, DTC, EBITDA, ECB, ELA, EMBI, ETF, FGC, FSAP, FSB, FX, GDP, GFC, GFM, GOV, HH, HQLA, HTM, IBGE, ICR, IFM, IFNM, IL, ILE, IMF, INSPF, LCR, LTV, MSMEs, NFCs, NPL, NSFR, PD, RAM, ROE, ROW, SELIC, SLR, SME, SNA, SOE, T1, TNA, VAR, WEO *Source: cr18344 - EXECUTIVE SUMMARY (PDF chapter)* ### EXECUTIVE SUMMARY ### EXECUTIVE SUMMARY ### Financial system resilience and structure - Banks and investment funds dominate Brazil’s financial system. - Banking sector characteristics: - Continued to be well-capitalized, profitable, and liquid. - Profitability supported by prudent lending standards, high interest margins and robust fee income, despite record loan losses. - Outstanding nonperforming loans have increased marginally during the recession largely because banks have actively written off bad loans. - Investment fund industry: - Solid, enjoying steady growth of assets under management without experiencing net outflows, in aggregate, during the recession. - Market-based indicators point to relatively low levels of systemic risk in 2017. - Risk to nonbank sector as interest rates fall: - Lower returns will affect investment income. - A search for yield may increase risk-taking. ### FSAP focus and macrofinancial risk drivers - FSAP activities (in cooperation with BCB and CVM): - Bank stress tests on Brazil’s 12 largest banks (accounting for 90 percent of banking sector assets). - Analysis of interbank contagion, financial interconnectedness in Brazil, and cross-border spillovers. - Evaluation of financial soundness of corporates and households. - Assessment of implications from large-scale redemptions in the investment funds’ industry. - Three main macrofinancial risks guiding analysis: - Weak domestic growth due to loss of confidence. - Tighter and more volatile global financial conditions. - A significant slowdown in China and its spillovers. ### Bank solvency: baseline and adverse scenarios - Overall resilience: - Banks appear generally resilient to severe macrofinancial shocks. - Baseline scenario: - Envisages slightly compressed net interest margins and moderate credit growth. - One domestic systemically important bank (D-SIB) may struggle to meet the capital regulatory requirements. - Adverse scenario: - Banks would be able to absorb sizeable credit and market losses owing to existing capital and strong profitability buffers. - Four banks could face some solvency difficulty, but capital shortfalls relative to the hurdle would be modest. - Deferred tax assets (DTAs) concern: - DTAs account for a large share of common equity tier-1 (CET1) capital. - With a full deduction of DTAs from CET1 capital, capital shortfalls would still be manageable although seven banks (two-thirds in terms of total assets in the sample) would find themselves below capital requirements. - BCB stress tests: - Broadly confirmed these results, though showed smaller capital shortfall in the adverse scenario. ### Bank exposure to concentration, exchange rate, and market risks - Concentration risk: - Default of the largest borrowers of individual banks and of the system could have a significant impact on capital. - Exchange rate risk: - Banks have small net open foreign exchange positions, but large exchange rate depreciation could increase risk-weighted assets and create losses. - Market risk: - Two banks could experience significant losses when equity prices fall sharply. - Most banks could face a large decline in their capital following an increase in bond yields. - Banks’ significant exposure to Brazilian sovereign risk creates a strong bank-sovereign nexus. ### Pillar 2 capital and bank-specific measures - Use of Pillar 2 capital requirements: - Could help mitigate identified risks and support building additional capital buffers at some banks. - Capital add-ons based on individual banks’ risk profiles should be considered given concentration risk and large holdings of debt and equity securities. - BCB is already taking steps in this direction. - Banks with capital shortfalls in stress tests: - Should be encouraged to build capital buffers, for example, through restrictions on dividend distribution. ### Liquidity resilience and funding risks - Short-term liquidity: - Banks are well-positioned to manage short-term liquidity pressures. - Regulatory Liquidity Coverage Ratio (LCR) results show Brazilian banks have ample liquidity buffers. - Even under a stressed scenario of larger deposit outflows than the Basel III standards and significant stress in the bond market, only one D-SIB could face some difficulty in handling short-term liquidity pressures; a drawdown on all its required reserves combined with access to emergency liquidity assistance would enable it to cover liquidity shortfalls. - Stable funding and NSFR: - Banks seem well prepared to implement the Net Stable Funding Ratio (NSFR). - Many banks (including all D-SIBs) under extreme stressed conditions might not be able to maintain the required stable funding profile within a one-year horizon. - Cash-flow analysis: - Confirmed liquidity results by assessing liquidity by maturity buckets. ### Interbank contagion and interconnectedness - Interbank contagion: - Interbank exposures are relatively small and a large share are secured. - Only a default of one large bank could trigger another bank to fail; contagion would be limited in other cases. - In the adverse solvency stress scenario, more banks would not meet the hurdle due to credit loss related to interbank exposures; additional capital shortfalls would still be manageable. - Interconnectedness underpinned by government role: - Central role arises from presence of public banks and state-owned firms, importance of repos collateralized by government securities, and government securities as main liquid financial instrument. - Banks are the most interconnected sector by size of exposures and presence across all financial instruments. - Investment funds have strong links with banks via repo transactions, and with insurance and pension fund sectors by lending to the former and receiving funds from the latter. ### Investment funds and potential government bond market effects - Large-scale redemptions and calibration approaches: - Three calibration approaches used: fund-homogeneity, fund-heterogeneity and fund-family. - Under these approaches funds would use R$437, R$215 and R$77 billion of their reverse repos to meet extreme redemptions, respectively. - Followed by a total sale of R$194, R$34 and R$10 billion of government bonds, respectively. - Impact on yields under 2017 market liquidity: - Average rise in government bond yields of 62, 21 and 8 basis points, respectively. - Impact under more severe market liquidity conditions of September 2015: - Average increase in bond yields of 99, 28 and 11 basis points, respectively. - Conclusion: - Investment funds could contribute to short-term price effects in the government bond market in the event of large-scale redemptions; effects crucially depend on calibration assumptions. ### Corporate and household sector vulnerabilities - Corporate sector: - Shows moderate signs of recovery from the recession. - Profitability has improved and leverage has fallen since 2015. - Shares of firms with negative equity, high leverage, and low debt servicing capacity remain elevated. - Corporate sector vulnerable to negative macrofinancial shocks that could substantially increase amount of debt at risk. - Given large amount of debt belonging to financially weak firms, a review of corporate insolvency regime is useful to ensure efficient debt restructuring and recovery processes. - Household sector: - Household debt relative to income has declined since early 2015. - Household balance sheets have weathered the recession fairly well. - Households have large debt servicing obligations: 22 percent (share of income), among the highest in the world, reflecting high interest rates. - Shares of financially weak households and corresponding debt at risk have increased marginally since 2013; negative macrofinancial shocks would bring these figures above previous peaks. - Financial stability implication limited given most debt at risk is covered by assets. - Growing share of mortgage-related debt and high share of mortgage borrowers who are relatively indebted and have large debt servicing obligations suggest BCB should consider implementing limits on debt to income and debt service to income for mortgage lending. ### Cross-border spillovers - Brazil experiences both inward and outward spillovers in: - Equity market. - Sovereign credit risk market. - Foreign exchange market. - Patterns: - Equity and sovereign credit risk: spillovers both inward and outward within Latin America; inward spillovers from Europe and the United States. - Foreign exchange: inward and outward spillovers concentrated among large and financially integrated emerging market economies, with Brazil generating outward spillovers to other smaller Latin American economies. ### Risk monitoring, stress testing framework, and suggested enhancements - BCB vigilance: - Continue vigilant risk monitoring given existing vulnerabilities. - Close monitoring of vulnerabilities in corporate and household sectors warranted. - Enhance monitoring of bank liquidity risk using cash-flow analysis considering varied degrees of stress in both local and foreign currencies. - Adopt integrated approach to assess financial-sovereign linkages, particularly implications of large holdings of liquid government bonds and effects of large-scale redemptions in investment fund industry for government bond market liquidity and consequences for banks’ solvency and liquidity. - Stress testing framework: - BCB’s framework is sound and the plan to develop solvency-liquidity contagion module is welcomed. - Framework could benefit from better handling of overseas operations and estimation of credit loss based on default probability and loss given default. - Improve data validation given use of various data sources to ensure accuracy and consistency. - Validate results produced by banks to reap full benefit of annual bottom-up stress testing exercise and improve banks’ risk management capacity. ### Key recommendations (selected from Table 1) - Main recommendations (timeframe: Immediate to Near-term unless specified): - Use Pillar 2 capital requirements to handle bank-specific risk profiles to mitigate concentration risk, including large holdings of debt and equity securities. — Immediate - Encourage banks to build capital buffers, especially those with capital shortfalls in the stress tests, for example, through restrictions on dividend distribution. — Immediate - Ensure the accuracy and consistency of information to ensure the reliability of analyses. — Immediate - Conduct a cash-flow analysis to assess banks’ liquidity risk on a regular basis, considering varied degrees of stress in both local and foreign currencies. — Immediate - Continue vigilant risk monitoring given existing vulnerabilities in the corporate and household sectors. — Immediate - Implement limits on debt to income and debt service to income for mortgage lending. — Near-term - Validate the results produced by participating banks in the annual bottom-up stress testing exercise, in part to ensure banks’ risk management capacity. — Near-term - Adopt an integrated approach to assess the financial-sovereign linkages. — Near-term - Other recommendations (timeframe: Near-term to Medium-term): - Review the corporate insolvency regime to ensure efficient debt restructuring and recovery processes. — Near-term - Require banks to assess the sensitivity of borrowers’ debt servicing capacity to macrofinancial shocks. — Near-term - Develop the solvency-liquidity contagion module in the bank stress testing framework as planned. — Medium-term - Enhance the stress testing framework through improvement in handling of overseas operations and estimation of credit loss based on default probability and loss given default. — Medium-term *Source: EXECUTIVE SUMMARY (cr18344) — Technical Note prepared for the FSAP analysis.* ### 7. Despite the severe recession, the banking sector has remained resilient. The banking ### 7. Despite the severe recession, the banking sector has remained resilient. The banking ### Banking sector resilience - The banking sector has continued to be well-capitalized, profitable, and liquid. - Profitability supported by prudent lending standards, high interest margins and robust fee income, despite record loan losses. - Strong profits and deleveraging have ensured that banks’ capitalization remains above regulatory requirements. - Public banks have lower capital ratios than private counterparts. - External funding exposures are low (around 10 percent of total funding). - Net foreign exchange positions are small. ### Credit intermediation and lending trends - Signs that credit intermediation has started to recover following the large contraction during the recession. - The aggressive monetary easing that started at end-2016 have had its effects. - Financial conditions have eased; surveys showed more favorable lending terms through 2017. - Lending-to-funding spreads have fallen by 380 basis points (as of January 2018) since end-2016. - Overall bank lending still contracted in 2017, mainly due to the continued decline in lending to corporates. - Lending to households accelerated to 5.6 percent in 2017, supported by earmarked credit. ### Nonbank financial sector performance and risks - Assets under management of investment funds have grown steadily, led by fixed-income funds, the dominant investment vehicle in the industry. - Growth of insurance premiums has fallen, but the sector’s profitability and other key credit metrics have improved, reflecting higher investment returns supported by higher interest rates. - Pension funds have enjoyed steady growth as tax benefits support growing contributions. - Some pension schemes, especially of state-owned enterprises, have experienced poor performance of investment portfolios given losses on equity investment. - As interest rates approach historical lows, outlook becomes more challenging: lower returns will affect investment income and a search for yield may increase risk-taking. ### A Bird’s-Eye View on Systemic Risk (as of 2017Q3) - Market data suggested systemic risk was low in 2017Q3. - The probability that several financial institutions experience distress simultaneously, despite rising since early 2016, was about a half of the level prevailing during the stress episode in 2015. - Expected loss in the event of distress remained at a historically low level. - Findings should be interpreted with caution given market information contains noises. - Large banks were the main drivers of systemic risk but some appeared to play a stabilizing role in periods of stress. - Systemic importance of the four-listed domestic systemically important banks (D-SIBs) is illustrated by greater potential to generate cascade effects than other banks and nonbanks in absolute terms. - The central clearing party (CCP) operator also appeared to be systemically important. - Contribution to systemic risk of most of these D-SIBs was smaller than their size, suggesting they tended to mitigate rather than amplify shocks at time of stress. - Systemic risk arising from interconnectedness was largely concentrated among the four listed D-SIBs. - The probability that other D-SIBs would become distressed given that one D-SIB became distressed was relatively high. ### Scope of the FSAP financial stability analysis - The FSAP focused on the two largest financial sectors—banks and investment funds—and included analysis of corporates, households, interconnectedness, and cross-border linkages. - Modules undertaken: - Bank solvency stress tests: - Top-down bank solvency stress test for Brazil’s 12 largest banks, which account for 90 percent of banking sector assets, using supervisory data. - Banco Central do Brasil (BCB) carried out a parallel top-down test for the same sample using its own methodology. - All tests used the same baseline (October 2017 WEO projections) and adverse scenarios. - Bank liquidity stress tests: - Top-down liquidity stress testing for the 12 largest banks using supervisory data. - Three modules: (i) Liquidity Coverage Ratio (LCR)-based test, (ii) Net Stable Funding Ratio (NSFR)-based test, and (iii) cash flow analysis. - LCR- and NSFR-based tests conducted based on regulatory and stressed parameters; cash-flow analysis considered liquidity for domestic and foreign currencies separately. - FSAP team analyzed BCB’s internal short-term liquidity indicator (IL) for a larger set of banks. - Bank contagion analysis: - Interbank network analysis covering the 12 largest banks, including standalone and as part of solvency stress tests to gauge additional capital impact from exposures to defaulting banks. - Corporate sector analysis: - Firm-level balance sheets of listed and non-listed companies using publicly available data. - Adopted debt-at-risk approach based on metrics such as interest coverage ratio (ICR) to assess “debt at risk.” - Household sector analysis: - BCB carried out household-level balance sheets using the debt-at-risk approach and credit registry data; identified financially weak households using debt service to income and debt to income metrics. - Analysis of potential vulnerabilities from investment funds: - Evaluated consequences of large-scale asset sales by investment funds due to a tail-event redemption shock for government bond market liquidity. - Covered open-end money market funds, bond funds and mixed funds, which account for 78 percent of the industry’s total net assets (TNA). - Interconnectedness and spillover analysis: - Analyzed financial interlinkages among different sectors using flow-of-funds data and examined cross-border linkages using market data. ### Key macrofinancial risks - Brazil faces significant downside risks driven by domestic and external factors. - Major domestic risks: political instability, spillovers from the corruption investigation, risk to the reform agenda, dilution or delay of reforms to maintain the constitutional expenditure ceiling. - Main external risks: tighter or more volatile global financial conditions associated with unwinding unconventional monetary policy, retreat from cross-border integration, and a significant slowdown in China. - Risk Assessment Matrix highlights: - Weak domestic economic growth due to loss of confidence — Likelihood: High; Impact: High. - Tighter and more volatile global financial conditions — Likelihood: High; Impact: Medium. - Significant China slowdown and its spillovers — Likelihood: Medium; Impact: Medium. - Potential consequences of weakened confidence and prolonged uncertainty include: - Increase in funding costs across the board, impacting banks’ net interest income and causing liquidity squeeze for corporates. - Impaired debt servicing capacity of corporates and households, raising banks’ nonperforming loans. - Losses on holdings of government bonds for banks, and redemptions at investment funds. - Reduced bank lending to the economy, adversely affecting growth and public finance. ### Macrofinancial scenarios underpinning the financial stability analysis - Two scenarios used: baseline (October 2017 WEO projections) and an adverse scenario simulated with the Global Macrofinancial Model (GFM). - Baseline: - Envisages a moderate recovery, with real GDP growth reaching 2 percent during 2019–20. - Interest rates expected on a downward trend, resulting in some compression in net interest margins. - Credit will grow at around 8 percent annually during 2018–20. - Adverse scenario: - Designed based on key macrofinancial risks outlined in the RAM; Global environment also challenging. - Real GDP would fall 11.8 percent below the baseline by 2018, with its 3-year cumulative growth 2 standard deviation (based on real GDP growth in the past 30 years) below the baseline. - Scenario features a sudden stop of capital inflows, large depreciation, collapsing stock prices, spiking bond yields, and initial monetary policy tightening to contain inflationary pressures. - Corporate and household balance sheets would weaken due to rising funding costs and falling income; unhedged foreign-currency borrowers see a sharp rise in the debt burden. - Banks could face large credit and market losses given impaired borrower repayment capacity and holdings of government bonds. - Liquidity risk could materialize with deposit funding outflows and investment fund redemptions, potentially prompting asset sales that further increase bond yields. - Tightening of financial conditions could be amplified by intra-system linkages, while constrained bank lending capacity could further exacerbate the recession. ### Bank solvency stress tests (introductory note) - The two scenarios (baseline and adverse) were used for the bank solvency stress tests. - The GFM was used to simulate the adverse scenario and informed stress magnitudes and other analysis modules. *Source: IMF staff technical note (excerpt).* ### 18. The bank solvency stress testing exercise comprised both scenario and sensitivity ### 18. The bank solvency stress testing exercise comprised both scenario and sensitivity analyses ### Overview of the exercise - Data reference date: end-2017Q3. - Scenario analysis horizon: three years, 2017Q4–2020Q4. - Sensitivity analysis: assessed impact of single risk factors (exchange rate risk, interest rate risk, concentration risk, and some other aspects of market risk) on existing capital buffers; carried out only by the FSAP team. - FSAP top-down (TD) exercise used the methodology described below; Appendix II contains the stress testing matrix (STeM). ### Recent banking sector performance (key statistics and observations) - CET1 capital ratio for the banking sector: 12.7 percent at end-2017Q2 (up from 11.1 percent about two years earlier). - Capital buffers relative to regulatory requirements: about 6.3 percent of risk-weighted assets, as of 2017Q3 (buffers calculated based on T1 capital). - Composition and caveats: - About three-quarters of capital are in the form of CET1 capital. - Deferred tax assets (DTAs) account for almost 30 percent of CET1 capital. - Profitability: - System-wide return on assets: 1.3 percent as of 2017Q3. - Profitability supported by large net interest income; reduced net interest income has driven a long-term downward trajectory in profitability. - Public banks are on average less profitable than private banks. - Asset quality and loss provisioning: - Nonperforming loans (NPLs): 3.7 percent of total loans at end-2017Q3 (down by 0.3 percentage point from its peak a year earlier; pre-recession level was 2.9 percent at end-2013). - Alternative measures of problem loans: would increase to 5 percent if restructured loans are included, and to 8.1 percent if restructured loans and loans in the “E” to “H” categories are included. - Impairment cost: 4.2 percent of total assets (up from 2.6 percent prior to the economic downturn). - Provisioning coverage: 160 percent. - Within the sector, private banks and D-SIBs have higher NPL ratios and larger impairment costs. ### IMF top-down stress test methodology: main features - Regulatory and accounting basis: - Applies the Basel III capital standard per the national framework; all banks use the standardized approach for credit risk. - No initial adjustment to the starting position for DTAs; FSAP assessed sensitivity to (i) deduction of all existing DTAs from CET1 capital and (ii) downgrading of performing problem loans. - FSAP deviated from the regulatory framework by estimating unrealized gain/loss associated with holdings of debt securities in both available-for-sale (AFS) and held-to-maturity (HTM) accounts. - Balance sheet approach and projection scope: - Focus on balance-sheet items, particularly loan portfolios (49 percent of total assets) and securities holdings (20 percent of total assets). - Other important items: cash in vault and lending to the central bank (13 percent of total assets); other investment and credit claims (12 percent of total assets); tax assets and nonfinancial assets (6 percent of total assets). - Key items projected on annual frequency for 2018, 2019, and 2020. - Quasi-static balance sheet assumption: - Growth of gross exposures (total gross loans and gross holdings of debt securities) assumed identical to overall credit growth in macrofinancial scenarios. - Balance sheet composition assumed constant. - Banks can build capital only through retained earnings; cannot raise new capital; maturing tier-2 instruments not renewed. - Dividend policy: banks allowed to pay dividends if positive net income after taxes and no additional capital needed, with dividend payout rate similar to the 2016 level. - Banks expected to raise additional funding as needed at the same prevailing funding costs. - Pre-impairment income as loss-absorbing buffer: - Pre-impairment (net) income comprises net interest income, other (net) income, and operating expense. - Net interest income estimation accounts for balance sheet size, interest rates, and existing NPLs; new NPLs during horizon generate no interest income. - Average interest rates estimated via fixed-effects panel regressions with explanatory variables: real GDP growth, system-wide bank lending and funding rates, and EMBI spread. - Other (net) income and operating expense account for balance sheet size changes; fee income modeled with dependence on real GDP growth, Selic, and equity price; other income components assumed constant relative to total assets. - Assumes zero non-operating net income and no extraordinary items. - Capital and hurdle metrics: - Tests measure resilience via CET1, tier-1 (T1), total capital ratios, and leverage ratio. - For the adverse scenario, hurdle rates combine Basel minimums with applicable D-SIB capital surcharges. - For the baseline scenario, hurdle rates include phased-in capital conservation buffer; countercyclical capital buffer (CCB) set at zero throughout the horizon. - Table of hurdle rates (in percent of risk-weighted assets): - Minimum levels (CET1, T1, Total capital): 4.5; 6.0; 8.0 (for 2017, 2018, 2019, 2020). - Capital surcharges for D-SIBs: 1.0 (2017–2020). - Capital conservation buffer (baseline scenario): 1.25 (2017), 1.875 (2018), 2.5 (2019), 2.5 (2020). - Leverage ratio (T1 capital, in percent of total exposures): 3.0 (2017–2020). - Credit risk modeling and treatment of HTM securities: - Credit loss captured for loan portfolios, securities holding, and other credit claims. - Suite of satellite credit risk models used to project provisions for credit loss associated with lending and non-lending activities. - Provisioning approaches explored: - Reported results based on estimating lending impairment and other impairment separately (lending impairment projected using a macrofinancial-indicator-based model; other impairment set at 2017Q1–Q3 level). - Alternatives: (i) estimate total impairment, or (ii) use estimated problem loans to derive provisions. - HTM account exposures treated economically like AFS exposures; potential credit loss captured by unrealized gain/loss based on market value. - Market risk and valuation adjustments: - Market risk from existing open positions and valuation adjustments for debt and equity securities assessed. - Market loss/gain driven by fluctuations in exchange rate, bond yields, and commodity and equity prices. - Financial positions adjusted in line with balance sheet growth; changes in bond yields estimated based on exposure by interest-rate type (fixed-rate, flexible-rate, inflation-linked) and maturity buckets. - Yield curves derived from changes in short-term and long-term interest rates; inflation changes used for inflation-linked securities (value affected by change in real yields). - Exercise did not consider impact of interest rate movements on other financial derivatives (e.g., interest rate swaps). - Risk-weighted assets and tax treatment: - Change in risk-weighted assets (RWA) assumed to primarily reflect size of total assets; RWA increase proportionate with balance sheet size. - RWA for foreign-currency exposures adjusted to reflect changes in amount of exposures in Brazilian reais as currency moves. - Income tax computed based on pre-tax net income; tax rate used: 45 percent. - Rationale: provisions for loan loss are not tax deductible; banks receive deferred tax credits (DTCs) from provisions for loan loss, effectively adding 45 percent of provisions for loan loss to CET1 capital as DTCs in each period; exercise follows typical accounting (taxes on pre-tax net income) to avoid modeling evolution of DTCs, with a stylized example illustrating neutrality on capital. - Contagion and network effects: - Exercise does not consider second-round macroeconomic effects from bank deleveraging or reduced credit growth impacting NPLs. - Contagion via interbank cross-exposures is incorporated: at the end of each year, additional credit loss from failure of other banks is calculated; capital after contagion becomes starting capital for the next period. *Source: cr18344 - 18. The bank solvency stress testing exercise comprised both scenario and sensitivity* ### 32. This subsection discusses the stress test results conducted by the FSAP mission. Given ### 32. This subsection discusses the stress test results conducted by the FSAP mission ### Projection of Average Interest Rates by Satellite Net Interest Income Models - Method: - Estimated average interest rates for lending activity, non-lending activity and funding using the fixed-effects panel regression approach. - Estimation based on bank-level quarterly information on average interest rates for the period of 2000Q1–2017Q3. - Explanatory variables include system-wide lending and funding rates, Selic, EMBI spread, and real GDP growth. - Key dynamics (adverse scenario vs baseline): - Interest margin would be compressed significantly during the early phase of the adverse scenario, largely driven by a sharp increase in funding cost. - Subsequently, interest margin would widen as banks earn at a higher rate from lending and non-lending activities and funding cost declines due to monetary policy accommodation. - In the baseline, the interest rate margin would decline marginally over the stress testing horizon, reflecting loosening financial conditions. - Note: - “In reality, the interest rate margin could be much more compressed given recent monetary policy loosening that is not envisaged in the baseline scenario analyzed in this stress testing exercise.” ### Projection of Credit Loss by Satellite Credit Risk Models - Method: - Suite of satellite credit risk models estimated using fixed-effects panel regression. - Bank-level quarterly information on total and lending impairment costs for 2000Q1–2017Q3 and problem loans in total and by portfolios for 20004Q1–2017Q3. - Problem loans include NPLs and other loans in the “E” to “H” categories. - Three approaches considered: - First approach (used to present results): estimate lending impairment and other impairment separately. Lending impairment uses regression with real GDP growth and EMBI spread; other impairment follows 2017Q1–Q3 level. - Second approach: total impairment cost estimated via a regression model. - Third approach: estimate problem loans via regression to derive provisions; then estimate relationship between lending impairment cost and outstanding problem loans by regression. Problem loans can be estimated in total and by portfolios. - Note: estimating impairment cost from change in NPLs is challenging due to banks’ prevalent practice of writing off NPLs; increase in outstanding NPLs represents a fraction of newly defaulted exposures. - Key projections: - Lending impairment cost relative to total loans would increase to 5.3 percent in the adverse scenario; lending impairment would peak in 2019, though still lower than the level after the global financial crisis (6.2 percent of total loans). - Total impairment cost would reach 2.9 percent of total assets in the adverse scenario (higher than during the last recession). - Total impairment cost is projected to remain broadly stable relative to total assets in the baseline. - Over the stress testing horizon, total impairment cost would be cumulatively larger in the adverse scenario by about 1.6 percent of total assets than in the baseline. - Dynamics under different approaches are broadly similar; the third approach yields a smaller amount of total impairment cost in the adverse scenario. - Drivers: - Asset quality mainly influenced by real GDP growth or unemployment, EMBI spread, and system-wide lending interest rate. - Other macrofinancial indicators generally do not influence impairment cost or problem loans either in total or by portfolios. ### Benchmark Results of the Scenario Analysis (12 largest banks sample) - Baseline: - CET1 capital ratio for the 12 largest banks would fall from 13.2 percent at end-2017Q3 to 11.4 percent at end-2020. - Contributors to reduction: phase-in of deductions from CET1 capital (-0.2 percentage points), dividend payout (-2.1 percentage points), growth of balance sheet (-2.7 percentage points). - Banks generate net income boosting capitalization by 3.4 percentage points over the period. - One public D-SIB’s capitalization would fall below the hurdle rates, with additional capitalization need of R$6 billion (0.1 percent of GDP). The bank would struggle to meet regulatory capital requirements under the assumption of no new equity issuance. - Adverse scenario (benchmark, no adjustments): - CET1 capital ratio of the 12 largest banks would decline by 4.7 percentage points to 8.5 percent at end-2020. - Aggregated losses and unrealized loss associated with holding of debt securities reduce CET1 by 0.2 and 1.4 percentage points, respectively. - Reduction drivers also include dividend payout (1.1 percentage points) and growth of balance sheet (1.8 percentage points). - Quantified impacts and dynamics: - Relative to the baseline, net interest income would fall by 21 percent and credit loss would increase by 23 percent. - Credit loss alone is roughly equal to pre-impairment income (both about 11.2 percent of risk-weighted assets). - Market loss (including unrealized loss associated with holding of debt securities) would amount to 1.5 percent of risk-weighted assets over the horizon; market loss is 3.6 percent of risk-weighted assets in 2018. - Three banks’ capitalization would fall below the hurdle rates at end-2020; another bank would fail temporarily but recover by end-2020. - Additional capital required: R$46 billion (0.7 percent of GDP), nearly all for two public banks. ### Additional Results and Sensitivity to DTAs and Problem Loans - Treatment of DTAs: - About 30 percent of CET1 capital are in the form of DTAs. - Test assuming deduction of all existing DTAs (including DTCs) from CET1 capital immediately reduces CET1 by 3.1 percentage points (calculation assumes risk weight for all these DTAs is 2.5). - Under this adjustment, CET1 capital ratio would be 5.8 percent at end-2020 in the adverse scenario (compared with 8.5 percent with no adjustments). - DTCs eligible for CET1 account for 85 percent of these DTAs; the remaining are DTAs allowed in CET1 up to a certain limit. - Downgrading of performing problem loans: - Test downgrades all performing problem loans to the “E” category (based on regulatory framework where loans are classified into nine categories; “E” is best class among NPLs). - Immediate reduction in CET1 capital ratio by 0.2 percentage points. - With more restrictions on dividend distribution, CET1 would be 8.3 percent at end-2020 (slightly higher than 7.8 percent with no adjustments under that dividend restriction assumption). - Combined shocks (DTAs deduction + downgrading problem loans): - Immediate reduction in CET1 by 3.3 percentage points. - In the adverse scenario, CET1 would fall to 5.5 percent at end-2020. - Seven banks would fail to meet hurdle rates at some point, with one bank recovering by end-2020. - Additional capitalization need: R$136 billion (2 percent of GDP), about three quarters for public banks. ### Interbank Contagion and Integrated Solvency Exercise - Integrated exercise (bank solvency stress tests + interbank network analysis) considers credit loss due to interbank cross-exposures. - Adverse scenario with interbank contagion: - CET1 capital ratio would fall to 7.1 percent at end-2020. - Additional loss amounting to R$68 billion (1.5 percent of risk-weighted assets). - Six banks’ capitalization would fall below hurdle rates at some point, resulting in additional capitalization need of R$81 billion (1.2 percent of GDP). - Interbank contagion causes two additional banks to fail the hurdle tests relative to standalone exercise. - Exercise specifics: - Only the credit loss channel is considered. - Trigger for bank default set at the minimum regulatory capital requirements; results not sensitive to trigger threshold (additional credit loss unchanged if trigger set at half the minimum regulatory capital requirements, though default timing shifts). ### Banking–Sovereign Linkages and Hypothetical Government Default - The adverse scenario does not model a government default, but implications are discussed for a hypothetical default event: - Adverse scenario features an increase in government bond yields by 700 basis points. - A spike in bond yields could result in market loss of R$90 billion (2.3 percent of risk-weighted assets), about two-thirds of which are unrealized losses. - If flexible-rate government debt securities are subject to a similar haircut (16 percent), the loss could be R$34 billion larger. - Government inability to recapitalize banks or guarantee liabilities would prevent use of DTAs to absorb losses, implying larger capital shortfalls. - Given public banks’ large share of credit to nonfinancial sectors and their weaker resilience, government inability to support weak public banks would have a significant impact on economic activity. ### Results of the Sensitivity Analysis (summary) - Key sensitivities: - CET1 outcomes are highly sensitive to treatment of DTAs and downgrading of performing problem loans. - Interbank contagion materially increases losses and capital shortfalls. - A sovereign default would substantially amplify market losses and capital shortfalls, particularly for public banks. *IMF staff estimates; Banco Central do Brasil data used in estimations.* ### 43. The sensitivity analysis considers four types of risk—concentration risk, market risk, ### The sensitivity analysis considers four types of risk—concentration risk, market risk, exchange rate risk, and interest rate risk. ### Sensitivity analysis: scope and transmission channels - The sensitivity analysis assesses the impact of single-risk factors individually with the magnitude broadly in line with those in the scenario analysis. - Transmission channels captured: - Concentration risk: credit loss generated by the default of large borrowers from an individual bank as well as the banking sector. Loans to five largest borrowers of individual banks account for 3.6 percent of banking sector assets, while loans to five largest system-wide borrowers account for 2.9 percent of banking sector assets. - Market risk: market loss related to movements in bond yields, equity prices, exchange rates, and commodity prices. Impact related to banks’ holding of debt securities only covers net exposures (including future positions) in the trading account. For other risk factors, the impact is assessed based on gross holding of equity securities, net open foreign exchange positions and net exposures of commodities in the entire balance sheet. - Exchange rate risk: credit loss associated with foreign-currency lending (credit risk induced by exchange rate movements), market loss due to existing net open positions, and additional risk-weight assets related to foreign-currency exposures. - Interest rate risk: market loss due to fluctuations in bond yields, change in net interest income following the adjustment of interest rates, and credit loss due to changes in borrowers’ debt servicing capacity. ### Key findings — concentration and market risk - Concentration risk and market risk appear significant, with public banks especially vulnerable. - Default of five largest borrowers of individual banks could reduce the CET1 capital ratio by 5.3 percentage points. - Note: for assessing concentration risk, the exercise assumes loss given default of 100 percent and individual banks’ average risk weight for credit risk. - Default of five largest system-wide borrowers could reduce the CET1 capital ratio by 4.3 percentage points. - Market loss via bond yields and equity prices: - CET1 capital ratio would fall by 1.1 percentage points if bond yields increase by 700 basis points. - Footnote context: the exercise assumes that the nominal yield curve has a parallel upward shift and the real nominal yield curve is also affected the increase in near-term inflation by 4.5 percentage points and in long-term inflation by about 2 percentage points. - CET1 capital ratio would fall by 1.2 percentage points if equity prices fall by 50 percent. - Market losses associated with net open foreign exchange positions and net exposures of commodities are very limited. ### Key findings — exchange rate and interest rate risk - Exchange rate depreciation: - CET1 capital ratio would decline by 1.8 percentage points following a large exchange rate depreciation. - Despite banks’ small net open foreign exchange positions, the impact would largely arise from additional credit loss and increased risk-weighted assets as value in reais of foreign-currency exposures increases. - Interest rate risk: - The impact on banks’ solvency could amount to 3.5 percentage points of risk-weighted assets following a spike in bond yields by 700 basis points. - The reduction in the CET1 capital ratio from a bond yield spike is mainly attributed to: unrealized loss associated with the AFS account (54 percent) and market loss through income statement (31 percent). Unrealized loss associated with the HTM account would be relatively limited. - A decline in Selic by 200 basis points would reduce net interest income by 0.5 percent of risk-weighted assets. With some benefit from reduced credit loss, the impact on net income would be 0.4 percent of risk-weighted assets. - Public banks appear more vulnerable to falling interest rates than private banks given that the reduction in their net interest income would likely be larger. - The impact is estimated based on simple satellite models that project average interest rates and lending impairment cost based on the dynamics of Selic. ### Hypothetical sovereign risk-weight exercise - Applying a 0.2 risk weight for exposures to the Brazilian government would reduce the CET1 capital by 0.5 percentage points. - Under the assumption that banks rebuild capital buffers to maintain the same level of capital ratios, such additional capital buffers would be 0.5 percent of risk-weight assets. - These additional buffers would only cover losses related to Brazilian government debt securities if bond yields increase by about 250 basis points (using the same assumption about the yield curve referenced above). ### Quantified impacts from Table 7 — Impact on CET1 capital ratio (in percentage points) - Concentration risk — Default of individual bank's largest borrowers: - The largest borrower: All -1.8, D-SIBs -1.5, Public -2.9, Private -0.9 - Three largest borrowers: All -3.9, D-SIBs -3.1, Public -6.4, Private -1.8 - Five largest borrowers: All -5.3, D-SIBs -4.1, Public -8.7, Private -2.4 - Ten largest borrowers: All -7.7, D-SIBs -5.9, Public -12.6, Private -3.5 - Concentration risk — Default of banking system's largest borrowers: - The largest borrower: All -1.6, D-SIBs -1.3, Public -2.8, Private -0.5 - Five largest borrowers: All -4.3, D-SIBs -3.3, Public -8.0, Private -1.0 - Ten largest borrowers: All -6.2, D-SIBs -5.0, Public -11.7, Private -1.7 - Twenty-five largest borrowers: All -9.4, D-SIBs -7.4, Public -17.1, Private -3.0 - Market risk (market loss only): - Bond yield spike by 700 basis points: All -1.0, D-SIBs -1.1, Public 0.0, Private -1.9 - Equity price decline by 50 percent: All -1.2, D-SIBs -0.3, Public -2.0, Private -0.5 - Exchange rate depreciation by 50 percent: All 0.1, D-SIBs 0.0, Public 0.2, Private 0.0 - Commodity price decline by 50 percent: All 0.0, D-SIBs 0.0, Public 0.0, Private 0.0 - Exchange rate risk: - Exchange rate depreciation by 50 percent: All -1.8, D-SIBs -2.0, Public -1.5, Private -2.2 - Due to open position: All 0.1, D-SIBs 0.0, Public 0.2, Private 0.0 - Due to risk-weighted assets: All -0.8, D-SIBs -0.7, Public -0.6, Private -1.0 - Due to depreciation-induced credit loss: All -1.1, D-SIBs -1.4, Public -1.0, Private -1.2 - Interest rate risk: - Bond yield spike by 700 basis points: All -3.4, D-SIBs -3.6, Public -2.7, Private -4.0 - Market loss for the trading account: All -1.0, D-SIBs -1.1, Public 0.0, Private -1.9 - Unrealized loss associated with AFS securities: All -1.8, D-SIBs -1.9, Public -1.8, Private -1.8 - Unrealized loss associated with HTM securities: All -0.5, D-SIBs -0.6, Public -0.8, Private -0.2 - Selic increase by 350 basis points: All 0.6, D-SIBs 0.5, Public 0.9, Private 0.4 - Related to interest income: All 1.0, D-SIBs 0.9, Public 1.2, Private 0.8 - Related to credit loss: All -0.4, D-SIBs -0.5, Public -0.4, Private -0.4 - Selic decrease by 200 basis points: All -0.4, D-SIBs -0.3, Public -0.5, Private -0.3 - Related to interest income: All -0.5, D-SIBs -0.5, Public -0.6, Private -0.4 - Related to credit loss: All 0.1, D-SIBs 0.1, Public 0.1, Private 0.1 - Asset quality — Downgrading of problem loans to "E" category: - Total loans: All -0.2, D-SIBs -0.3, Public -0.4, Private -0.1 - Renegotiated loans: All -0.2, D-SIBs -0.2, Public -0.3, Private -0.1 - Restructured loans: All -0.2, D-SIBs -0.2, Public -0.3, Private -0.1 - Quality and quantity of capital: - Derecognition of existing deferred tax assets: All -3.1, D-SIBs -3.6, Public -2.8, Private -3.3 - Assume risk weight of 0.2 for sovereign exposures: All -0.5, D-SIBs -0.5, Public -0.5, Private -0.6 ### Top-down stress tests by the authorities (BCB) — methodology and results comparison - BCB framework: balance sheet approach with projections of credit growth, average interest rates for lending activity, non-lending activity and funding, other income and operating expense components, credit loss derived from dynamics of provisions, and market loss based on marking-to-market. Balance sheet growth is driven only by credit portfolio growth. Unrealized loss/gain in the AFS account is included in market loss. Impact on capital accounts for evolution of prudential filters and risk-weighted assets; market risk component estimated per regulatory framework. - Results comparison (IMF vs BCB): - Baseline scenario: BCB projects CET1 capital ratio increasing to 15.4 percent at end-2020; BCB's higher estimate mainly due to no dividend payout (2.1 percentage points), lower credit loss (1.9 percentage points), and taxes (1.9 percentage points). BCB test envisages smaller balance sheet expansion and lower pre-impairment income. Two banks would fail to meet the hurdle, resulting in capital shortfalls of R$12 billion. - Adverse scenario: BCB projects CET1 capital ratio improving to 13.5 percent by end-2020; differences largely due to lower credit loss (3.8 percentage points), lower market loss (1.3 percentage points; 0.4 percentage points due to different treatment of debt securities in the HTM account), and no dividend payout (1.1 percentage points). Two banks would fail the hurdle, with capital shortfalls of R$15 billion. ### Assessment of the BCB’s solvency stress testing framework — strengths and areas for enhancement - Strengths: - Granular approach to market risk related to debt securities holdings and net foreign exchange positions. - Estimation of the market risk component of risk-weighted assets. - Accounting for Brazil-specific tax and prudential treatments. - Use of various data sources enabling timely stress tests (monthly information with two-months lag) from supervisory reports, credit registry, payment systems, and trading repositories. - Areas for enhancement: - Balance sheet growth: ensure aggregated bank-level credit projections are consistent with overall credit growth in macrofinancial scenarios; adjust other exposures in line with macrofinancial conditions. - Overseas operations: differentiate domestic and overseas operations; estimate net interest income and credit loss of key overseas subsidiaries. - Net interest income module: review linkages (currently average interest rates in each component solely linked to Selic) to ensure sensible projections given low estimated net interest income in outer years. - Credit loss estimation: improve by projecting default probability and loss given default rather than relying solely on provisions dynamics; default probability could be proxied by newly recognized NPLs and/or outstanding NPLs adjusted for loan recovery and write-offs; loss given default could be estimated based on provisioning practices. - Market loss: include market loss associated with equity securities holdings given existing large exposures at some banks. - Capitalization: ensure market risk component of risk-weighted assets is updated through the horizon (market risk component appeared to be calculated once at the beginning of the horizon), and continue to account for Brazil-specific tax and prudential treatments (particularly related to DTAs). The amount of CET1 capital is adjusted to reflect the cap on ineligible DTAs (at 15 percent of CET1 capital). ### Scenario design and governance - The BCB regularly conducts stress tests based on three scenarios: baseline (using Consensus Economics forecasts), adverse (targeting real GDP growth at the 5th-percentile) and structural break (targeting all macrofinancial indicators at the 5th-percentile), with adverse and structural break scenarios designed using a Bayesian vector auto regression (VAR) model. - Recommendation: conduct top-down stress tests based on macrofinancial scenarios that reflect the BCB’s view of the baseline and relevant financial stability risks in adverse scenarios; the BCB plans to use scenarios consistent with its own macro view in the upcoming annual bottom-up exercise. - BCB plans and recent initiatives: - Further enhancements including development of a solvency-liquidity contagion module. - Expanding satellite credit risk models to increase granularity beyond the current three portfolios. - Estimating all income components at the bank level for the largest banks. - Annual bottom-up stress testing exercise initiated in 2017 with plans to validate participating banks’ results in upcoming exercises. *Source: IMF staff estimates, excerpts from the IMF Financial Sector Assessment Program chapter on Brazil (sensitivity analysis, top-down tests, and assessment of BCB framework).* ### 55. The use of Pillar 2 capital requirements could help mitigate identified risks, supporting ### 55. The use of Pillar 2 capital requirements could help mitigate identified risks, supporting ### Pillar 2 capital requirements and supervisory actions - Pillar 2 capital requirements could help mitigate identified risks, supporting the need to build additional capital buffers at some banks. - Given concentration risk, including large holdings of debt and equity securities, capital add‑ons based on individual banks’ risk profiles should be considered. - The BCB is rightly taking steps in this direction. - The supervisory process, including grading of banks’ financial soundness, already covers the abovementioned risks and takes stress testing results into account. - Banks that would experience capital shortfalls in the stress tests should be encouraged to build capital buffers, for example, through the restriction on dividend distribution, to improve their resilience to macrofinancial shocks. ### Stress testing framework: strengths and recommended enhancements - The BCB’s stress testing framework is generally sound and could benefit from further enhancements. - The recent upgrade in the framework led to a significant improvement, but its performance should be reviewed to ensure that all modules generate sensible results. - Areas for improvement: - Handling of overseas operations. - Estimation of credit loss based on default probability and loss given default. - Development of a solvency‑liquidity contagion module to better capture potential risk amplification. - Ensuring accuracy and consistency across various data sources used by the BCB. - Validation of results produced by banks to reap the full benefit of the annual bottom‑up stress testing exercise, including improvement in banks’ risk management capacity. ### BANK LIQUIDITY STRESS TESTS — Overview - Liquidity stress tests comprised LCR‑based tests, NSFR‑based tests, and a cash‑flows analysis. - Purpose: assess banks’ ability to manage potential funding outflows during a period of significant financial stress. - Tests were conducted jointly by the FSAP mission and the BCB. - LCR test: measures ability to meet liquidity needs in a 30‑day stress scenario using stock of unencumbered HQLA. - NSFR test: complementary view of funding profile over a one‑year horizon. - Cash‑flow analysis: granular analysis by maturity buckets (from 1 day to more than six months). - Data: supervisory data as of 2017Q3 covering the 12 largest banks (11 largest commercial banks—of which five are D‑SIBs and four are publicly‑owned—and one publicly‑owned development bank). - BCB’s IL ratio also computed for 132 small and medium sized banks. - NSFR‑based tests performed in all currencies together; LCR‑based tests and cash‑flow analysis analyzed liquidity risks in local currency and foreign currency separately. ### Current liquidity conditions and banks’ liquidity profiles — key findings - Liquidity risk was low despite the large recession. - Brazilian banks increased holdings of liquid assets in a period of low credit demand. - Brazilian banks have one of the highest liquid assets to short‑term liabilities ratio amongst its peers. - BCB’s IL ratios have remained comfortably above the internal monitoring warning level despite the recession. - System‑wide LCR has been above 100 percent since end‑2015. - Long‑term: BCB’s Long‑Term Structural Liquidity Ratio does not suggest large maturity transformation at the aggregate level. - Composition of HQLA: - Almost all of LCR’s HQLA is composed of Level 1 assets (99 percent). - Government securities: 70 percent for all 12 banks and 66 percent excluding the development bank. - For the development bank, Level 1 assets constitute 95 percent of its total HQLA and are only government securities. - Reserves: 23 percent. - Cash: 4 percent. - Foreign currency‑denominated assets are not a significant part of HQLA and overall Level 2 assets are almost zero. - More than seventy‑five percent of secured lending portfolios constitute reverse repos secured by government securities. - Funding structure: - More than sixty percent of commercial banks’ funding comes from wholesale sources. - Forty percent of wholesale funding for the 11 largest commercial banks comes from short‑term overnight repo funding secured by government securities (constituting 25 percent of total commercial bank funding). - Remaining 60 percent of wholesale funding comes from longer‑term unsecured wholesale sources. - Retail deposits comprise around a quarter of total funding. - Brazil’s largest development bank obtains almost all of its funding from unsecured long‑term wholesale funding in the form of borrowing (maturity beyond six months) from the Brazilian government. - Structural liquidity operations: banks intermediate liquidity between investment funds and the BCB via repos; BCB sterilizes Brazil’s structural liquidity surplus (around 25 percent of GDP) via OMOs. - Funding differences across bank types: - Share of short‑term funding larger for D‑SIBs compared to non‑SIBs. - Close to 60 percent of D‑SIBs’ overnight funding comes from repos backed by government securities versus 25 percent for non‑SIBs. - Development bank holds mainly long‑term assets funded by long‑term liabilities. - Reliance on offshore funding remains low: - Foreign funding accounts for 8 percent of total funding for the 12 largest banks. - Composition of foreign funding: 65 percent from issuance of securities in foreign currency, 15 percent from other unsecured wholesale funding, 9 percent from retail time deposits and secured borrowing. - On the assets side, loans to clients, banks and other financial institutions account for 86 percent of total foreign‑currency (FX) assets. - Overseas exposures concentrated at D‑SIBs: 90 percent of total FX loan portfolio and 83 percent of total foreign funding. ### LCR‑Based Stress Tests — scenarios and calibration - LCR tests conducted in line with national implementation of Basel III. - Regulatory requirement: Brazilian banks with total assets above R$100 billion are required to maintain an LCR above 80 percent, increasing to 100 percent in January 2019. - LCR is self‑declared: calculated daily by banks and sent to BCB monthly via the Report on Liquidity Risk; calculation accuracy assessed by on‑site supervision. - Limits on central bank reserves included in HQLA: BCB uses a conservative approach (three components determine reserves forming HQLA: (a) excess reserves; (b) automatic releases of required reserves due to deposit outflow; and (c) remaining required reserves up to 15 percent of the stock of Level 1 HQLA). - Two scenarios in LCR‑based tests: - Baseline scenario: standard LCR applying regulatory parameters set out by the Brazilian LCR requirement. - Adverse scenario: broadly consistent with solvency stress tests and historical funding shocks (assumptions include government bond yields underpinning calibration of haircut rates; run‑off rates for deposits calibrated based on historical bank‑level behavior during 2007–17; differentiated by deposit types and counterparties). - Adverse scenario calibration: uses historical data for run‑off parameters of retail deposits, operational deposits and funding from nonfinancial corporates, sovereigns, central banks, international organizations, public entities; run‑off parameters for deposits from SME are more conservative than historical data; other parameters include an additional markup over the regulatory scenario. - In the tests banks can use cash holdings, reserves, and standard BCB lending facilities to cover negative funding gaps; ELA access was not taken into account. ### LCR‑Based Stress Tests — sample grouping - Sample of 12 banks grouped into 4 categories for analysis: - (i) the 12 largest commercial and development banks in the solvency stress test; - (ii) the largest 10 commercial banks; - (iii) the 5 D‑SIBs (subset of the 10 largest commercial banks); - (iv) the 5 largest non‑SIBs. - Two banks excluded from some aggregate calculations because they have very ample short‑run liquidity (LCR comfortably above 1000 percent) and their inclusion would inflate aggregate liquidity measures. ### LCR‑Based Stress Tests — key results - Baseline LCR results: - Aggregate LCR is 203 percent for 10 largest commercial banks as of September 2017. - Aggregate LCR is 248 percent for all 12 banks as of September 2017. - Under baseline scenario each bank has LCR above the current regulatory requirement at 80 percent, and even above the 100 percent hurdle rate effective in 2019. - Strength drivers: large stable retail deposit base, large holdings of government securities and reverse repos, large share of secured funding collateralized by government securities (not subject to haircuts in baseline). - Adverse scenario results: - Assumptions: banks lose between 10 and 30 percent of their retail and small business deposits in a month; unsecured wholesale funding reduced by 71 percent; secured wholesale funding reduced by 10 percent; a haircut of 9.35 percent applied to entire portfolio of government securities (consistent with spike in government bond yields and inflation assumptions in solvency tests). - One D‑SIB (out of 10 banks) would see its LCR below the regulatory requirement at 80 percent, with a small liquidity shortfall of R$31 billion (equivalent of less than 1 percent of the bank’s assets and less than 1 percent of GDP). - Another D‑SIB would set its LCR slightly above 100 percent. - Aggregate LCR would fall to 122 percent for the 10 banks and 145 percent for the 12 banks. - For the D‑SIB with liquidity shortfalls, the use of all required reserves in HQLA would not be sufficient to bring the LCR above 100 percent; liquidity support through ELA would be required to cover the remaining gap. - Haircut rationale: haircut takes into account (i) average market loss on entire government securities portfolio from a spike in bond yield by 700 bps; and (ii) BCB’s haircut on standard lending facilities. The market loss varies by bond type and maturity (e.g., fixed‑rate bonds with maturity greater than 5 years facing 49 percent market loss). - FX LCR findings: - The (baseline) LCR in U.S. dollar (USD) is below 35 percent for three out of eight largest commercial banks. - For these three banks the share of net cash outflow in USD constitutes more than 13 percent of total cash outflow. - FX liquidity shortfalls are small and equivalent to less than 1 percent of foreign reserves (R$9.3 billion). - Overall LCRs for these three banks are well above 100 percent. - Foreign currency LCR is required as a monitoring tool. - Foreign funding constitutes 12 percent of total funding maturing within 30 days for one of these banks and less than 3 percent for the remaining two banks. ### Comparison of LCR and IL measures and short‑run liquidity outcomes - IL ratio overview: - IL ratio measures whether banks have enough liquid assets to cover short‑term (1 month) cash‑flow needs in a simulated stress scenario defined and calibrated by the BCB (since 2002). - IL ratio is computed based on granular data and is a complementary risk measure to the LCR. - Aggregate comparison (September 2017): - Aggregate IL was 271 percent for the 10 largest commercial banks. - Aggregate IL was 275 percent for all 12 banks. - Aggregate IL is higher than the baseline LCR ratio, implying a larger liquidity buffer by the IL metric. - Individual bank outcomes: - At individual bank level the LCR can be higher or lower than IL; results do not always align. - Based on the IL ratio, 19 small and micro banks would face liquidity shortfalls within a 30‑day horizon, mainly driven by large outflows of wholesale and retail deposits. - These 19 banks would face a liquidity shortfall of R$10 billion. - Methodological differences explaining divergences: - IL is more conservative in coverage of liquid assets, mainly reserves, in the numerator. - Net cash outflow in the LCR is larger than in the IL owing to: - a) assumptions on wholesale deposits run‑off which are not as strict in IL (particularly for D‑SIBs); - b) run‑off on retail deposits and net contractual outflows in IL are based on observed historical outflows (last 100 days), which are not as severe. *Source: cr18344 — IMF staff report excerpt.* ### 72. The NSFR-based tests assessed the maturity mismatch, utilizing the BCB’s Structural ### cr18344 - 72. The NSFR-based tests assessed the maturity mismatch, utilizing the BCB’s Structural ### NSFR-based tests and methodology - The NSFR-based tests utilized the BCB’s Structural Liquidity Ratio (ILE) developed based on the NSFR methodology and supervisory data using BCB’s internal definitions for available stable funding for calculating the SLR. - Banks are not yet required to meet the NSFR—it will be effective from October 2018 onwards—but the tests provide a complementary view of banks' funding profile at a one-year horizon. - The exercise was carried out based on regulatory and stress parameters (Appendix IV, Table 3). ### Scenarios considered - Two scenarios consistent with the LCR-based tests: - Baseline scenario: assumes regulatory parameters. - Adverse scenario: consistent with the adverse scenario used in the LCR-based tests and bank solvency stress tests to assess capacity to maintain stable funding while continuing credit operations. - In the adverse NSFR scenario parameters were adjusted by: (i) using the haircuts and the one-year cumulative run-off rates from the cash-flow-based liquidity analysis for comparable categories; or (ii) 10 percentage points for other categories. ### Baseline NSFR results (aggregate and individual) - Available stable funding for the twelve largest banks: R$4,027 billion (September 2017). - Required stable funding for the twelve largest banks: R$3,486 billion. - Aggregate NSFR: 116 percent (above minimum requirement of 100 percent). - Two non-SIB private banks would experience a small stable funding gap (individual NSFRs just slightly below 100 percent). - Drivers: high regulatory capital and high reserve requirements of Brazilian banks. ### Adverse NSFR results and impacts - Aggregate NSFR in adverse scenario: 95 percent. - Stable funding gap in adverse scenario: R$288 billion. - Stable funding gap as share of GDP: 4.6 percent of GDP. - Stable funding gap as share of banks’ assets: 4.7 percent of banks’ assets. - NSFR of the 5 D-SIBs would fall to 91 percent (four out of five in the range of 80-91 percent). - Stable funding gap attributable to the 5 D-SIBs: R$227 billion. - Main cause: reliance on redeemable deposits—particularly less stable retail deposits—assumed to face a cumulative outflow of 35 percent within one year. ### Cash-flow analysis: structure and assumptions - Maturity buckets used: overnight; 2 days to 1 week; 1 week to 2 weeks; 2 weeks to 1 month; 1 month to 3 months; 3 months to 6 months; over 6 months. - Time horizons analyzed: from 1 day to up to 6 months, in local and foreign currencies (foreign largely denominated in USD, close to 90 percent). - Net cash balance = existing cash position + counterbalancing capacity + net funding inflows. - Counterbalancing capacity includes access to standard BCB facilities (government bonds only as eligible collateral) and, if no eligible collateral, possible access to ELA under more stringent conditions and/or waiver of reserve requirement. - At the banking system level, eligible assets for BCB’s ELA (after haircuts) make up 5 percent of total assets and 20 percent of counterbalancing capacity. - ELA in the exercise only considered loans with maturity beyond 6 months to avoid double-counting with credit inflows. - Haircuts on liquid assets eligible as central bank collateral (government securities only) range up to 49 percent market loss for fixed-rate bonds with maturity greater than 5 years; Selic-indexed bonds face no market loss. - Exercise based on going concern assumption; banks assumed to keep a static hedging structure. ### Cash-flow analysis: findings - Three banks would face liquidity shortfalls within the 180-day horizon under the adverse scenario: - One D-SIB would face a liquidity shortfall within a 30-day period given a cumulative group-wide deposit outflow of 21 percent in one month and 25 percent in six months. - Two additional non D-SIBs would have insufficient liquidity for a 180-day horizon. - If banks draw down required reserves (before access to ELA), all above-mentioned banks would be able to remain liquid for a 180-day horizon. - If banks maintain required reserves instead of drawing them down, access to ELA alone is not sufficient to confront withdrawals of funding for these banks. - The same D-SIB would face short-term liquidity problems in the adverse scenario if required reserves are not included in HQLA. - Five banks could face liquidity problems for longer than 6 months, consistent with NSFR finding that six banks will not meet NSFR at 100 percent. ### FX cash-flow analysis - Seven banks would deplete all their FX liquidity buffers within six months in the adverse scenario. - Four D-SIBs and three other banks would face an FX liquidity shortfall of R$52 billion after sale of securities within 180 days. - R$52 billion shortfall is about 4.3 percent of foreign reserves. - Cause: banks do not hold sufficient FX assets that could be used to counterbalance negative FX net cash outflows. ### Reputational crisis scenarios and LCR outcomes - Two reputational scenarios analyzed with LCR-based tests: - Localized reputational crisis: run-off rates calibrated to match volatility during recent historical reputational events; Basel III regulatory haircuts applied. - System-wide reputational crisis: hypothetical broader loss of confidence, fire sales imply higher haircuts. - Localized reputational crisis results: - Aggregate LCR would fall to 119 percent for the 10 largest commercial banks (142 percent for all 12 banks). - One D-SIB would fall below the regulatory LCR hurdle rate of 80 percent with a small liquidity shortfall of R$25 billion (less than 0.5 percent of banks’ assets). - Counting all required reserves in HQLA would still not be sufficient to cover the liquidity gap in this bank. - System-wide reputational crisis (fire sales) results: - Two D-SIBs would experience liquidity shortfalls of R$56 billion (3.2 percent of failing banks’ assets); one of them would need ELA (after using all required reserves). - Aggregate LCR would fall to 107 percent for the 10 largest commercial banks (127 percent for all 12 banks). - Comparison with BCB internal study: - IMF and BCB identify the same two banks that would face a liquidity shortfall in the localized scenario. - BCB estimated a considerably lower liquidity shortfall driven by both a higher value of liquid assets and a lower value of net cash outflow. ### Policy recommendations on liquidity stress testing and frameworks - Key observation: liquidity stress test results confirm strong liquidity buffers overall, but liquidity risk may arise in some banks in extreme scenarios. - Strengthen BCB’s liquidity provision framework, including ELA (see Technical Note on Crisis Management). - Enhance stress testing methodology: - Perform liquidity stress tests using the structure of cash flows at various horizons; adopt a maturity ladder exercise including features of the FSAP cash-flow analysis and consider varied degrees of stress in both local and foreign currencies. - NSFR: continue efforts to implement NSFR requirements on a consolidated basis later in 2018; SLR is a proxy NSFR measure. - Ensure consistency across data sources: - Align data submitted by banks in the Report on Liquidity Risk (details of LCR calculation) with data from Trade Repository and other sources used to calculate the proxy NSFR and BCB’s IL ratio to improve calculation and reporting accuracy and comparability across liquidity tests. ### Interbank network contagion analysis (overview and methodology) - Analysis uses a network model (Espinosa-Vega and Solé, 2010) based on supervisory data of bilateral interbank gross exposures among the 12 largest banks as of 2017Q3, including exposures to the remaining 123 banks as a group. - Stand-alone interbank network analysis: - Simulates hypothetical default of each bank on all interbank obligations and assesses cascading defaults. - A bank is assumed to default when CET1 ratio < 4.5 percent of risk-weighted assets. - Recovery rates assumed: secured interbank lending recovery rate 90 percent; unsecured lending recovery rate 0 percent; overall LGD on total interbank exposure 50 percent (weighted average). - Funding shock assumption: a bank must replace 50 percent of funding lost due to a default by selling other assets with a 50 percent haircut, causing fire-sale losses. - Caveat: contagion assumed instantaneous; no time for banks to mitigate between rounds. - Solvency-interbank contagion interaction: - Direct credit losses from interbank exposures added to solvency stress-test impacts sequentially each year; post-contagion capital position becomes starting point for next year’s solvency test. - Does not consider funding loss channel in this integrated analysis. - Caveat: assumes bilateral exposure matrix does not change through the stress horizon. *Source: IMF staff calculations and BCB supervisory data as presented in the IMF FSAP excerpt.* ### 88. The six large banks combined are net borrowers in the interbank market, while smaller ### The six large banks combined are net borrowers in the interbank market, while smaller banks are net lenders ### Interbank structure and key statistics - By September 2017, about 42 percent of the banking system in terms of assets are net borrowers. - Interbank exposures among 12 largest banks: 2.4 percent of banking system assets and 2.8 percent of GDP. - Exposures of the 12 largest banks represent only half of their interbank exposures; the remaining half comes from interbank exposures to the 123 small and micro banks. - On average, 67 percent of interbank exposures are secured, almost all of which are overnight repos collateralized by government securities. - The large banks have the highest interbank assets relative to their CET1 capital, followed by the small banks. ### Stand-alone contagion analysis — results and indexes - Six simulations where each of the large banks defaulted: only one large bank (bank 1) triggered the default of one medium-sized bank (bank 9). - With the threshold at 4.5 percent of the CET1 ratio: - Hypothetical default of large bank 5 would reduce system-wide total capital by around 11 percent. - Hypothetical default of large bank 1 would reduce system-wide total capital by around 10 percent. - Index of vulnerability (percentage point of average capital loss due to the default of all other institutions) is at most 18 percent (in bank 9). - Note: banks 1–6 are the large banks and banks 7–12 are the medium-sized banks. ### Sensitivity to LGD (loss-given-default) assumptions - Baseline LGD used: 50 percent. - If LGD rate is assumed to be 100 percent (versus 50 percent in baseline): - Large bank 5 failure would impair 25.5 percent of the other banks’ capital. - Large bank 1 failure would impair 12 percent of the other banks’ capital. - In the former case (LGD = 100 percent and bank 5 fails), large bank 1’s capital would be completely wiped out. - Default of bank 1 would completely impair medium-sized bank 9’s capital in both LGD cases described. - Medium-sized bank 9’s vulnerability by LGD: - If LGD = 0 percent (i.e., funding shock only), average impairment of bank 9’s capital is 2.6 percent. - If LGD = 100 percent, average impairment of bank 9’s capital is 32 percent. - Caveat: LDG of 100 percent is noted as too extreme given that 60 percent of exposures on average are secured. ### Sensitivity to funding replacement assumptions - With increasing magnitude of the funding shock: - Contagiousness of bank 1 increases from 3 percent under 0 percent funding replacement to 17 percent under higher funding replacement. - Vulnerability of medium-sized bank 12 is very sensitive to the magnitude of the funding shock. - Failing medium-sized bank 9 is vulnerable even when funding stress is moderate: - Under 0 percent funding replacement, bank 9 would suffer 16 percent capital loss (as opposed to 18 percent in the baseline). - Overall, vulnerability in domestic interbank linkages among the 12 largest banks arises from: - Large bilateral exposures for some banks (notably exposures of bank 9 against bank 1, and bank 1 against bank 5). - Funding-side vulnerabilities for other banks. ### Failure of small and micro banks — system-wide effect - Hypothetical default of all small and micro banks simultaneously constitutes an 18 percent fall in the system’s CET1. - In the baseline, simultaneous default of all small and micro banks could, on average, reduce the capital of the rest of the banks by 23 percent through interbank exposures. - Sensitivity to LGD for the 12 largest banks’ exposure to small and micro banks: - If LGD = 75 percent, average impairment of bank capital is 36 percent and would eventually induce default of 3 medium-sized banks. - As a group, small and micro banks have a low index of vulnerability; however, at an individual bank level this might not hold (could not be tested owing to data constraints). ### Corporate sector analysis — overview and recent developments - Overview finding: corporate sector shows moderate signs of recovery from the recession, but vulnerabilities linger among financially weak firms that constitute a large fraction of the sector. - Specific sector vulnerabilities: low profitability especially in the energy sector where return on equity fell below interest rates in 2017Q3; about two-thirds of firms in the services sector faced difficulty servicing debt relying solely on operating income; firms in services were highly leveraged. - Recommendation: authorities should closely and continually monitor trends in the corporate sector and enhance the insolvency regime. ### Corporate financial soundness — recent metrics (2005–2017Q3) - Profitability: - Median ROE dropped to about 5percent by 2015, then improved to 8.5 percent in 2017Q3. - Compared with the Selic rate, 2/3 of all firms were unable to produce comparable returns on equity. - Debt servicing capacity: - Median ICR increased from 1.5 to 2 during the recession period but remains below pre-recession ICR of 3. - 40 percent of firms in 2017Q3 could not generate sufficient EBITDA to cover interest expenses. - Accounting for interest income through financial positions, 21 percent of firms in 2017Q3 failed to cover net interest expenses. - Leverage: - Median total corporate debt relative to equity rose to 0.8 by 2015 (almost doubled from 2006), then deleveraged to 0.7 in 2017Q3. - 1/3 of all firms remained highly leveraged with total debt exceeding twice equity. - Maturity composition: - Short-term borrowing accounted for 8percent of total debt in 2005, peaked at 12percent in 2013, and was about 9percent in 2017Q3. - 1/3 of firms used short-term debt that exceeded 35percent of their total debt, exposing them to roll-over risk. - Net interest coverage: - 1/5 of firms were unable to cover net interest expenses relying on earnings. ### Sectoral vulnerabilities (debt-at-risk findings) - Sector definitions: - Manufacturing: materials and industrials. - Energy: utilities and energy. - Consumer: consumer staples and consumer discretionary. - Services: telecommunication services, information technology, healthcare, and real estate. - Profitability-based debt-at-risk: - Debt at risk based on ROE falling below the Selic rate exceeded 80percent for services companies. - Debt at risk was less than 50percent for manufacturing companies. - Services sector had substantial fraction of negative ROE accounting for more than half of the debt in the sector. - Energy companies had many firms with low positive returns; negative ROE less of a concern. - Debt servicing capacity: - More than 2/3 of all debt in the services sector belong to firms with ICR falling below 1.5. - Debt at risk by this ICR threshold was only 10percent for companies in the energy sector. - Leverage by sector: - Based on net debt to EBITDA, more than half of the debt in the services sector had a leverage ratio exceeding 5. - Corresponding debt at risk for the energy sector was 5percent. ### Corporate sensitivity analysis to macro-financial shocks - Shocks modeled (consistent with bank solvency adverse scenario): - i) 8percent increase in the market interest rate. - ii) 43percent spike in foreign exchange rate. - iii) 10percent drop in profitability of businesses. - iv) Downturn scenario incorporating all shocks simultaneously. - Interest rate shock: - ICR specification accounts for interest-bearing assets and reduces net impact. - Interest rate shock almost doubles the share of total debt with ICR less than 1. - The shock adds an additional 6percent to an already 8percent debt at risk. - FX shock (43percent): - Firms appear reasonably hedged (natural and financial hedges). - Two hedging approaches considered: inclusive (operational + financial hedges) and conservative (financial hedges only). - In response to a 43percent FX shock, debt at risk would increase by 4 and 2 percentage points, respectively (inclusive and conservative approaches). - Conclusion: not a material change compared to other shocks; firms appear well-hedged against FX shocks. - Profitability shock (10percent): - ICR effect modeled as EBITD A*(1−profitability shock)/Interest Expense. - A 10percent profitability shock would increase debt at risk by 14 percentage points (more than 2.5 times). - Profitability shock has the largest impact on financially weak firms. - BCB alternative ICR measure: - Using net interest expenses (accounting for interest income from interest-bearing securities) lessens the impact of the profitability shock, while other shocks’ effects remain similar. ### Operational hedging (Box 1) — components and suggested formula updates - BCB defines operational hedges beyond financial hedges: - Exporters: classified as full natural hedge if debt linked to export revenues exceed half of total debt. Caveat: export revenues may cover import costs; net exports could be a better measure. - Assets overseas: value of overseas assets comoves with FX rate; liquidity of such assets matters (PPE less liquid). - Parent company: subsidiary of a foreign parent considered fully hedged if parent is likely to cover; rating of parent should be considered. - Suggested updated hedging expression (capped at 1 if larger than 1 implying a full hedge): - (Net Export ratio) + (effective liq value of foreign assets) + 1[high-rated foreign parent] + (fin hedge) - Where 1[high-rated foreign parent] = 1 if parent rating is high (e.g. above Baa) and 0 otherwise. *Source: IMF staff calculations and Brazil (BCB) data as presented in the supplied content.* ### 106. This section assesses financial strength of the household sector. It first reviews the ### 106. This section assesses financial strength of the household sector. It first reviews the ### Recent Development - Household debt relative to disposable income: - Fell to 41¼ percent as of November 2017, down from its peak at 46½ percent in April 2015. - Composition of household debt: - Mortgage borrowing accounts for about 45 percent of total household debt (up from 14 percent a decade ago). - Among the 12 largest banks, the higher share of mortgage lending increased by 14 percentage points since end-2012; the share of non-payroll-deducted secured lending declined by 8 percentage points; unsecured consumer credit declined by 7 percentage points; payroll-deducted lending increased by 3 percentage points. - Lending growth and underwriting: - Banks’ lending to households slowed early 2015 and has been growing at a positive rate, more strongly than banks’ overall lending. - Loan-to-value (LTV) requirements have become tighter; only a small fraction of mortgages feature high LTV ratios. - Nonperforming loans (NPLs) and write-offs: - NPL ratio for lending to households increased from 3.6 percent in June 2015 to 4.3 percent in January 2016, then trended downward. - As of December 2017, NPL ratio for lending to households was 3.5 percent; NPL ratio for overall lending was 3.2 percent. - Shares of loans written-off for lending to households and for overall lending were similar at almost 4 percent as of June 2017; dynamics of loans written-off lag NPLs by about a year. - Resilient performance attributed to banks’ active portfolio management (including debt restructuring) and relatively prudent lending practices. - Financial distress exposures: - Flexible-rate loans account for 43 percent of bank lending to households. - Households need to spend about 20 percent of their income on principal and interest payments; the level of debt servicing payments has remained elevated since 2012. - In 2017, about half of debt servicing payments were to cover interest charges. - High unemployment and weak real earnings continue to weigh on household balance sheets. - Since 2016, banks have been more active in restructuring loans—especially real estate financing; share of loans with 90 days past due and restructured loans had not yet declined as of June 2017. ### Household-Level Balance Sheet Analysis (debt-at-risk approach) - Methodology and data: - Debt-at-risk analysis conducted by BCB staff using credit registry data as of end-2017Q3 based on methodology agreed with the FSAP team. - Households considered financially weak if: debt service to income above 40 percent and/or debt to income above 3. - Debt at risk decomposed into amounts covered by assets (only housing assets serving as collaterals of mortgage loans considered) and those not covered by assets. - Key cross-sectional outcomes (as of 2017Q3): - Share of households with debt service to income above 40 percent: 20.2 percent. - Share of households with debt to income above 3: 3.6 percent. - Share of financially weak households (combined metrics): 2.9 percent. - Corresponding share of debt at risk (combined metrics): 22.5 percent. - Share of debt at risk not covered by assets (combined metrics): 2.9 percent. - Trends since 2013: - Share of financially weak households edged up by 0.3 percentage points since 2013 (based on combined metrics). - Share of debt at risk increased from 18.1 percent to 22.5 percent since 2013. - Share of debt at risk not covered by assets has been on a declining trend (influenced by increasing share of mortgage borrowing and considering only housing assets as collateral). - Groups more exposed to financial distress: - Low-income households (earnings up to minimum wage; ~20 percent of households), retirees, unsecured borrowers, multiple-loans borrowers, and mortgage borrowers face higher liquidity squeeze risk due to large debt servicing obligations. - Amount of debt at risk not covered by assets is relatively large for these groups, except mortgage borrowers whose borrowings are collateralized by housing assets. - When combined metrics are used, households with earnings between 1 and 3 times minimum wage (about 50 percent of households), employees, multiple-loans borrowers, and mortgage borrowers appear relatively weak. - Low-income households (earnings up to minimum wage) associated with the highest share of debt at risk not covered by assets. - Sensitivity analysis (single-factor shocks): - Declining income, rising interest rates, and falling house prices individually could drive the share of financially weak households and share of debt at risk above post-2013 peaks, regardless of metric used. - Impact of rising interest rates relevant only for borrowing with flexible interest rates; flexible-rate loans use reference rate (Taxa Referencial) set by the BCB and currently at a historical low level (note from source). - Scenario analysis (adverse scenario): - Scenario features: declining income by 8 percent, rising interest rates by 250 basis points, and falling house price by 10 percent. - Under combined metrics (debt service-to-income above 40 percent and debt-to-income above 3): - Share of financially weak households could increase from 2.9 percent to 3.9 percent. - Corresponding share of debt at risk could rise to 27.7 percent. - Share of debt at risk not covered by assets would increase to 3.7 percent (still below previous peak of 4.8 percent), reflecting larger housing assets pledged as collateral. - Estimated credit loss that banks may incur could be relatively limited, contingent on the ability to execute collaterals efficiently. - Selected sensitivity and scenario table entries (exact values replicated): - Latest (as of 2017Q3) - Debt service-to-income above 40 percent: Share of financially weak households 20.2; Share of debt at risk 52.1; Share of debt at risk not covered by assets 24.5. - Debt-to-income above 3: Share of financially weak households 3.6; Share of debt at risk 25.8; Share of debt at risk not covered by assets 3.4. - Combined metric: Share of financially weak households 2.9; Share of debt at risk 22.5; Share of debt at risk not covered by assets 2.9. - Highest level since 2013 (selected): - Combined metric: Share of financial weak households 2.9; Share of debt at risk 22.6; Share of debt at risk not covered by assets 4.8. - Sensitivity analysis highlights: - Declining income by 5 percent: Combined metric — Share of financially weak households 3.2; Share of debt at risk 24.4; Share of debt at risk not covered by assets 3.2. - Declining income by 10 percent: Combined metric — Share of financially weak households 3.6; Share of debt at risk 26.5; Share of debt at risk not covered by assets 3.5. - Rising interest rates by 250 basis points: Combined metric — Share of financially weak households 3.4; Share of debt at risk 24.6; Share of debt at risk not covered by assets 2.9. - Rising interest rates by 500 basis points: Combined metric — Share of financially weak households 3.6; Share of debt at risk 25.3; Share of debt at risk not covered by assets 2.9. - Falling house price by 10 percent: Debt service-to-income above 40 percent metric — Share of debt at risk 25.0; Combined metric — Share of debt at risk not covered by assets 3.2. - Falling house price by 20 percent: Debt service-to-income above 40 percent metric — Share of debt at risk 25.9; Combined metric — Share of debt at risk not covered by assets 3.8. - Scenario analysis (declining income by 8 percent, rising interest rates by 250 basis points, and falling house price by 10 percent): - Debt service-to-income above 40 percent metric — Share of financially weak households 24.1; Share of debt at risk 62.4; Share of debt at risk not covered by assets 27.7. - Debt-to-income above 3 metric — Share of financially weak households 4.1; Share of debt at risk 28.9; Share of debt at risk not covered by assets 4.2. - Combined metric — Share of financially weak households 3.9; Share of debt at risk 27.7; Share of debt at risk not covered by assets 3.7. ### Policy Recommendations - Implement macroprudential limits: - The BCB should consider implementing limits on debt to income and debt service to income to contain buildup of systemic risk in the mortgage lending segment. - Debt-at-risk analysis indicated that the share of mortgage borrowers who were relatively indebted and had large debt servicing obligations was at 30 percent. - Given banks’ lending to households increasingly geared towards mortgage lending, instituting these measures now could contain systemic risk when the credit cycle upturn starts. - Bank borrower stress testing: - The BCB should require banks to assess the sensitivity of borrowers’ debt servicing capacity to macrofinancial shocks such as rising interest rates. ### Introduction to Investment Funds Analysis (related section start) - Objective: - Test resilience of government bonds’ market to liquidity shocks in the investment fund industry absent action by the BCB and the National Treasury. - Evaluate consequences of large-scale asset sales by investment funds due to hypothetical tail-event redemption shock for government bonds’ market liquidity. - Approach: - Calibrate redemption shocks using three approaches based on historical observations. - Evaluate size of first-round asset sales to meet redemptions under waterfall and pro rata asset sales for each calibration approach; quantify associated price effects in government securities’ market. - Estimate regression model to assess second-round redemptions due to decreased market returns implied by first-round asset sales. *Source: IMF staff compilation based on Banco Central do Brasil; credit registry data as of end-2017Q3; IMF staff calculations.* ### 119. The analysis followed and extended the liquidity analysis undertaken in recent FSAP of ### cr18344 - 119. The analysis followed and extended the liquidity analysis undertaken in recent FSAP of ### Analysis overview and caveats - The analysis followed and extended liquidity analysis undertaken in recent FSAP of the U.S., Ireland, Sweden and Luxembourg and built on recent work by Baranova and others (2017). - This was the first time price effects were estimated and analyzed in greater depth for this context, using three calibration approaches that took fund heterogeneity into consideration. - Important caveats: - The analysis did not assume any reactionary role of the Central Bank and the Treasury. - It did not account for general equilibrium implications or provide a full picture of forces affecting the government bond market such as the role of monetary policy, the unwinding of repurchase agreements with the Central Bank, reverse auctions by the Treasury, or the role of banks. ### Data, sample selection, and preprocessing - Dataset coverage and timing: - Data provided by BCB and CVM covered the whole universe of 555-funds. - Dataset contained historical information for fund flows, NAV, returns, composition of funds’ asset holdings, and fund families and subfamilies. - September 2017 data was used for asset composition. - Open-ended funds and runnable sample: - Open-ended funds accounted for about 78 percent of total TNA of the entire investment fund industry by the end of September 2017. - Exclusions from full sample: (i) closed-ended funds, (ii) funds with pension funds as dominant investor, (iii) funds with non-deposit institutions as investor, (iv) funds with redemption notice date exceeding a week. - The remaining runnable sample accounted for 52 percent of funds in frequency and 78 percent of total TNA. - Table 14 (selected figures reproduced from source table): - Closed-ended: Count 2,280 (Fraction 30.0 percent); TNA 0.284 (Fraction 7.7 percent) - Pension fund Investor: Count 799 (Fraction 10.5 percent); TNA 0.408 (Fraction 11.0 percent) - Non-Deposit Investor: Count 14 (Fraction 0.2 percent); TNA 0.063 (Fraction 1.7 percent) - Long Redemption Notice: Count 527 (Fraction 6.9 percent); TNA 0.063 (Fraction 1.7 percent) - Open-ended: Count 3,976 (Fraction 52.3 percent); TNA 2.893 (Fraction 77.9 percent) - Total: Count 7,596; TNA 3.711 - Fund family composition (open-ended funds, historical averages): - Fixed-income family: 82 percent of total TNA (on average) - Multimarket family: 13 percent - Stocks family: 4 percent - Foreign investment: <1 percent (0.5 percent noted as tiny share) - Data cleaning and consistency rules: - NAV was used instead of TNA for calculating redemption rates per authorities’ guidance. - Missing NAV entries (0.14 percent of the sample) were imputed assuming an affine relationship. - Missing data for flows and returns were substituted with zero. - Observations with margin of error exceeding 5 percent (0.4 percent of the sample) were removed; the sample contained 2.7 million weekly observations. - NAV values that fell below the total sum of assets were overridden by the total sum of assets (16 percent of observations). - A new variable “other assets” defined as NAV minus total sum of assets for the rest of the sample. ### Calibration of redemption shocks (three approaches) - Flow rate definition: ratio of sum of flows during week relative to NAV at the beginning of the week. - Calibration approaches: - Fund-homogeneity calibration: - Uses the first percentile of distribution of flow rates of all fund-week observations in each fund family. - Assumes a common size of redemption shock impacting each fund regardless of differences. - Tests potentially stressful situations funds may not have experienced historically. - Fund-heterogeneity calibration: - For each individual fund, assumes outflow equal to the 1st percentile of that fund’s own historical flow rates. - Assumes fund-specific extreme shocks happen simultaneously. - More realistic in shock size but does not test beyond historical extremes. - Fund-family calibration: - Uses the 1st percentile of combined outflow of all funds within a fund family. - Assumes inflows and outflows net out within a fund family (transferred across funds). - Key calibrated redemption rates: - Fund-homogeneity (lowest 1 percent redemption rates): 25 percent (fixed-income), 14 percent (multi-market), 8 percent (stocks). - Fund-family (lowest 1 percent redemption rates): 1.9 percent (fixed-income), 2.2 percent (multi-market), 1 percent (stocks). - Worst combined outflow dates and magnitudes (fund-family calibration): - Fixed-income: 2.9 percent (Aug 1, 2014) - Multi-market: 4.2 percent (March 18, 2005) - Stocks: 3.3 percent (August 17, 2007) ### Asset holdings and simulated asset sales (portfolios and liquidation rules) - Aggregate portfolio composition (Panel a, September 2017, NAV in R$ billion): - Fixed Income NAV (total): 2292 - Cash Liquidity: 812 - Short-term Bonds: 131 - Medium-term Bonds: 713 - Long-term Bonds: 261 - Other Assets: 375 - Multimarket NAV (total): 464 - Cash Liquidity: 105 - Short-term Bonds: 24 - Medium-term Bonds: 97 - Long-term Bonds: 28 - Other Assets: 210 - Stocks NAV (total): 95 - Cash Liquidity: 4 - Short-term Bonds: 1 - Medium-term Bonds: 1 - Long-term Bonds: 1 - Other Assets: 88 - Total NAV (all families): 2852 - Cash liquidity position (sum of cash, reverse repo, deposits, offshore sovereign bonds): R$922 billion (aggregate liquidity position reported elsewhere) - Asset sale schedules modeled: - Waterfall approach: - Funds sell assets in descending order of liquidity as captured by LCR haircut hierarchy: i) cash, ii) reverse repo, iii) bank deposits, iv) sovereign bonds, v) short-term government bonds, vi) medium-term government bonds, vii) long-term government bonds, viii) Index of Bovespa (stocks), ix) stocks offshore, x) derivatives, xi) NFC debt. - Used as benchmark. - Pro-rata approach: - Funds sell a common fraction of all assets to preserve portfolio composition. - Simulated sales (Panel b, Table 15) — waterfall sales totals and pro rata totals (R$ billion) by calibration: - Totals to be liquidated to meet redemptions: - Fund-homogeneity: - Waterfall Total: 575 (Fixed Income) + 64 (Multi-market) + 8 (Stocks) = 648 - Pro rata Total: 575.3 (Fixed Income) + 64.1 (Multi-market) + 7.6 (Stocks) = 647.7 - Fund-heterogeneity: - Waterfall Total: 209.1 (Fixed Income) + 43.7 (Multi-market) + 4.4 (Stocks) = 257.7 - Pro rata Total: 209.1 (Fixed Income) + 43.7 (Multi-market) + 4.4 (Stocks) = 257.7 - Fund-family: - Waterfall Total: 67.2 (Fixed Income) + 19.5 (Multi-market) + 3.1 (Stocks) = 90.1 - Pro rata Total: 67.2 (Fixed Income) + 19.5 (Multi-market) + 3.1 (Stocks) = 90.1 - Government bond sales required to meet redemptions (R$ billion) under waterfall schedule: - Fund-homogeneity: R$194 billion - Fund-heterogeneity: R$34 billion - Fund-family: R$10 billion - Government bond sales required under pro rata schedule: - Fund-homogeneity: R$298 billion - Fund-heterogeneity: R$88 billion - Fund-family: R$39 billion - Cash liquidity coverage of total redemptions (aggregate cash liquidity used to meet redemptions): - Waterfall schedule: cash liquidity sufficient to cover 67 percent (fund-homogeneity), 83 percent (fund-heterogeneity), 85 percent (fund-family) of total redemptions. - Pro rata schedule: cash liquidity covers 34 percent (fund-homogeneity), 49 percent (fund-heterogeneity), 32 percent (fund-family) of total redemptions. - NAV at risk (sum of NAV of funds with insufficient cash liquidity position relative to aggregate NAV): - 52 percent (fund-homogeneity) - 21 percent (fund-heterogeneity) - 20 percent (fund-family) - Aggregated liquidity position vs redemptions: - Aggregate liquidity position of investment funds (R$ 922 billion) exceeded total value of redemptions under any calibration. - However, heterogeneities in individual cash positions would force some funds to sell assets (including government bonds). ### Scale of sales relative to historical turnover - Comparison of simulated government bond sales to historical weekly market turnover (Table 16 highlights percentiles): - Under benchmark calibration with waterfall sales: - Asset sales are at the higher end (~60th percentile) for some instruments and short-term NTN-B near its maximum (98th percentile in one cell). - Pro rata sales result in more extreme sales of less liquid assets (above 80th percentile for some categories). - Alternative calibration scenarios generally imply an insignificant size of government bond sales relative to historical turnover (with the exception of some NTN-B bonds). - Selected figures from Table 16 (R$ billion and percentile in distribution of weekly market turnover): - LTN Short-term: - Waterfall: (1) 22 (58th), (2) 15 (35th), (3) 3 (0th) - Pro rata: (1) 5 (1st), (2) 0 (0th), (3) 2 (0th) - NTN-B Short-term: - Waterfall: (1) 10 (98th), (2) 4 (85th), (3) 4 (88th) - Pro rata: (1) 1 (33rd), (2) 2 (57th), (3) 0 (21st) - LTN Medium-term: - Waterfall: (1) 29 (63rd), (2) 50 (99th), (3) 0 (0th) - Pro rata: (1) 9 (4th), (2) 0 (0th), (3) 6 (2nd) - NTN-B Medium-term: - Waterfall: (1) 20 (63rd), (2) 26 (81st), (3) 3 (0th) - Pro rata: (1) 7 (9th), (2) 0 (0th), (3) 4 (1st) - NTN-B Long-term: - Waterfall: (1) 12 (55th), (2) 22 (89th), (3) 2 (12th) - Pro rata: (1) 6 (31st), (2) 0 (1st), (3) 3 (18th) ### Assessing price sensitivity and market liquidity - Price sensitivity measurement: - Used the Amihud illiquidity measure evaluated for each government security type and maturity as the equally weighted average over a year of |R_t|/V_t where R_t is weekly return and V_t is weekly market turnover; length T = one year used in benchmark. - Extreme historical Amihud measures were also used to assess price effects under distressed market conditions. - First-round price effect: - Computed as Amihud illiquidity measure multiplied by the size of assets sold following the tail redemption event. *Source: BCB data and IMF staff calculations (excerpts from cr18344).* ### 136. Different calibration approaches lead to various degrees of price effects. Table 17 ### Different calibration approaches lead to various degrees of price effects. Table 17 ### Price effects from fund asset sales (summary findings) - Under 2017 market conditions (waterfall sales), average drops in government bond prices: - fund-homogeneity calibration: 62 basis points - fund-heterogeneity calibration: 21 basis points - fund-family calibration: 8 basis points - Under 2015 market conditions (waterfall sales), average relative drops: - fund-homogeneity calibration: 99 basis points - fund-heterogeneity calibration: 28 basis points - fund-family calibration: 11 basis points - Waterfall sales concentrate sales into more liquid/shorter-maturity bonds and could significantly raise short-term government rates: - Under 2017 market conditions, asset-sales based on fund-homogeneity, fund-heterogeneity and fund-family calibrations could lower NTN-B yields down from their median by 186 basis points, 88 basis points and 36 basis points, respectively. - Pro rata sales produce smaller price effects on short-term government bonds because short-term bonds are not prioritized relative to less liquid assets. - Medium-term government bond depressions under 2017 market conditions (average): - fund-homogeneity: 29 basis points - fund-heterogeneity: 2 basis points - fund-family: 1 basis point ### Table 17 — Price Effects (exact reported figures) - Notes: The left number under each asset sale schedule shows the amount of sales in R$ billion and the right number in percent shows the percentile in historical distribution of weekly trading volume relative to the median historical price changes. (1), (2) and (3) correspond to fund-homogeneity, fund-heterogeneity and fund-family calibrations. ST: Short-Term, MT: Medium-Term, LT: Long-Term. a. Based on 2017 market conditions - LTN (ST) - Waterfall: 0.008 — 18 (5th) 3 (35th) 0 (50th) - Pro rata: 12 (10th) 5 (27th) 2 (41th) - NTN-B (ST) - Waterfall: 0.177 — 186 (0th) 88 (0th) 36 (0th) - Pro rata: 82 (0th) 19 (5th) 12 (15th) - LTN (MT) - Waterfall: 0.011 — 32 (16th) 1 (48th) 0 (50th) - Pro rata: 55 (8th) 10 (37th) 7 (41th) - NTN-B (MT) - Waterfall: 0.013 — 26 (23th) 4 (44th) 1 (48th) - Pro rata: 35 (16th) 10 (36th) 5 (42th) - NTN-B (LT) - Waterfall: 0.043 — 52 (28th) 10 (46th) 4 (48th) - Pro rata: 99 (19th) 27 (40th) 13 (45th) b. Based on 2015 market conditions - LTN (ST) - Waterfall: 0.007 — 16 (6th) 3 (35th) 0 (50th) - Pro rata: 11 (11th) 4 (31th) 1 (44th) - NTN-B (ST) - Waterfall: 0.199 — 209 (0th) 99 (0th) 40 (0th) - Pro rata: 92 (0th) 22 (3th) 13 (13th) - LTN (MT) - Waterfall: 0.022 — 65 (7th) 1 (48th) 1 (48th) - Pro rata: 110 (1th) 20 (26th) 14 (32th) - NTN-B (MT) - Waterfall: 0.042 — 86 (4th) 13 (34th) 3 (45th) - Pro rata: 113 (3th) 32 (17th) 17 (32th) - NTN-B (LT) - Waterfall: 0.097 — 118 (15th) 23 (41th) 9 (47th) - Pro rata: 223 (6th) 62 (27th) 29 (38th) ### Sensitivity of redemptions to fund returns (regression findings) - Regression model: - F_{i,t} = β_i + β_1 R_{i,t−1} + β_2 R_{i,t} + β_3 X_{i,t} + ε_{i,t} - β_i captures fund-specific fixed effects; F_{i,t} denotes fund flows to fund i; R_{i,t−1} and R_{i,t} denote weekly returns of fund i in previous and current periods; X_{i,t} includes year fixed effect and fund family fixed effects. - Main result: Insignificant sensitivity of investors to changes in returns once lagged returns, fund size (NAV), year fixed effects and family fixed effects are included. - Table 18 (regression coefficients; Dependent Variable: Flow Rate): - (Intercept): 0.0051*** ; Std. error 0.0005 ; t-value 7.664 ; Pr(>|t|) 0.000 - Lag RETURN: 0.0000 ; Std. error 0.0002 ; t-value 0.01 ; Pr(>|t|) 0.990 - RETURN: 0.0011 ; Std. error 0.0048 ; t-value 0.22 ; Pr(>|t|) 0.825 - Year fixed effect: Yes - Fund family fixed effect: Yes - Adjusted R^2 = 0.07 percent - Interpretation: The economically weak relationship between fund flow rates and returns implies investors are insensitive to short-term return fluctuations and that second-round effects (redemptions-triggered additional sales) are insignificant in the Brazilian investment fund industry. ### Interconnectedness analysis — key findings - Data and scope: - Analysis uses non-consolidated financial accounts (flow-of-funds) as of end 2016 from a BCB-IBGE joint project covering ten institutional sectors and 22 financial instruments. - Institutional sectors include: Government (GOV), Non-financial Corporations (NFC), Households and Non-profit Institutions Serving Households (HH), Rest of the World (ROW), Central Bank of Brazil (BCB), Other Deposit-taking Institutions (BNK), Insurers and Pension Funds (INSPF), Other Financial Institutions (FIN), Monetary Investment Funds (IFM), Non-Monetary Investment Funds (IFNM). - Aggregate instrument shares (2016 Q4, totals across sectors): - Total Assets/Liabilities: 43,084,344 BRL Million - Equity and investment funds shares/units: 15,001,190 BRL Million (34.8 percent) - Equity: 9,633,247 BRL Million (22.4 percent) - Investment funds shares/units: 5,367,943 BRL Million (12.5 percent) - Debt securities: 8,604,486 BRL Million (20.0 percent) - Currency and deposits: 7,256,700 BRL Million (16.8 percent) - Loans: 6,335,062 BRL Million (14.7 percent) - Sectoral balances (Total assets; Total liabilities; Net Financing/Investment): - BCB: 2,969,017 ; 2,928,889 ; 40,128 - BNK: 8,358,636 ; 8,943,456 ; -584,819 - INSPF: 1,930,291 ; 1,848,216 ; 82,075 - FIN: 3,296,252 ; 3,269,111 ; 27,141 - IFM: 4,942,426 ; 4,924,068 ; 18,358 - IFNM: 526,429 ; 572,767 ; -46,337 - GOV: 3,876,539 ; 6,167,831 ; -2,291,292 - NFC: 6,235,561 ; 9,383,692 ; -3,148,131 - HH: 6,238,167 ; 2,353,143 ; 3,885,024 - ROW: 4,678,815 ; 2,660,962 ; 2,017,853 - Composition and channels: - Nearly 60 percent of cross-sectoral exposures are accounted for by equity, cash and currency deposits, and repos collateralized with government securities. - Long-term government securities represent 18 percent of total instruments; equity instruments (mainly issued by NFC) are nearly 25 percent. - Currency and deposits (deposits at banks and repos) are almost 17 percent and involve government, BCB, banks and investment funds. - Role of the government: - Government plays a central role driven by public banks/state-owned firms, the prevalence of repos collateralized with government securities for monetary policy and interbank transactions, and government securities as the main liquid instrument. - Banks and investment funds: - Banks are the most interconnected sector across instrument layers; NFC and HH are banks’ largest borrowers and lenders. - Investment funds attract primarily institutional investors (pension funds and insurance companies); retail investors (corporates and households) represent around 20 percent of investment funds’ funding. - Investment funds mainly invest in government securities and repos with banks as main counterparties. - By centrality measures, investment funds connect multiple sectors and layers and can propagate shocks across the system. - Pension funds and insurance companies: - Around 85 percent of their funding comes from households. - Almost 60 percent of their assets are invested in investment funds’ shares and an additional 10 percent in government securities. - Non-financial corporates: - Large bilateral exposures with ROW, banks and within NFCs. - Around 30 percent of corporates’ funding originates from the rest of the world. - NFCs act as central nodes and potential shock propagators across layers. - Households: - Large holdings of pensions and insurance products, investment funds’ shares and bank deposits; these holdings explain households’ large net investment position. - Intra-sectoral exposures: - Important for NFCs, investment funds and banks — implying intra-sector feedback channels (e.g., NFCs holding equity and receivables from other NFCs; investment funds holding own products; banks interconnected via interbank deposits and repos). *Source: BCB data and IMF staff calculations; figures and tables as reported in the source document.* ### 153. This section analyzes cross-border linkages between Brazil and its 16 most relevant ### cr18344 - 153. This section analyzes cross-border linkages between Brazil and its 16 most relevant ### Methodology and model specification - Methodology based on Diebold and Yilmaz (2014, 2015) and Demirer et al. (2017), using variance decompositions of market-based variables in VAR models with adjustments for large datasets and parametrization. - Produces weighted, directed and generally complete networks; yields pairwise and total directional connectedness measures and each country’s relative contribution to connectedness (returns and return volatilities). - VAR model details: - Sample period: 2007Q4 to 2017Q3. - Forecast horizon: 10-day forecast from VAR of daily returns and range volatilities. - VAR lag length: three lags. - Estimation uses elastic net shrinking/selection procedure combining LASSO and Ridge estimators (Demirer et al., 2017). - Variance decomposition invariant to ordering via generalized Koop-Pesaran-Potter-Shin identification. - Returns computed as the average of two days to adjust for different trading hours across time zones. - Range volatility: estimator of latent volatility using intraday information (high, low, close, opening prices). Garman and Klass (1980) formula used for most banks; Parkinson (1980) formula used for some Argentina and Chile banks due to data availability. ### Data, scope, and country coverage - Daily market information used on: - Stock price indices (total market and banking sector indices), - Sovereign CDS, - Dollar FX rates, - Additional exercise: bank stock return volatilities for 86 large banks headquartered in sample countries, including three largest private Brazilian banks and one public bank. - Country list (Brazil’s most important trade partners, regional economies, main financial centers, peer countries): Argentina, Chile, China, Colombia, France, Germany, India, Indonesia, Japan, Mexico, Russia, South Africa, Spain, Turkey, United Kingdom and the United States. - For sovereign CDS, Argentina, India and USA omitted due to lack of data (as noted in figures). For US dollar exchange rates, data for Germany and Spain omitted for redundancy; euro area represented by France. ### Connectedness measures and interpretation - Connectedness outputs: - Pairwise contributions to a country/bank’s variance due to shocks in another country/bank. - “From” directional spillover: extent to which the system explains total expected variation of a country/bank (inward spillover). - “To” directional spillover: contributions of an individual country/bank to the rest of the system (outward spillover). - Visual presentation: - Pairwise directional connectedness matrices presented as heatmaps (columns → rows). Main diagonal domestic spillovers placed next to heatmap to emphasize large domestic share. - Rationale: - Returns and volatility connectedness informative of financial stress episodes; volatility connectedness tends to be asymmetric and spikes in stress episodes. ### Key empirical findings on markets and regional patterns - General cross-market exposure: - Brazil is exposed to shocks from advanced economies across markets, notably the U.S. and euro area countries; cross-market spillovers can originate in, or be transmitted through, these economies. - Stock markets and sovereign CDS: - Clear regional clustering pattern for stock market indices, especially strong for sovereign CDS (credit-risk spillovers less uniformly distributed). - Brazil’s stock market and sovereign credit-risk inward and outward spillovers are likely to spread mainly within Latin America and to come from the U.S., and also from advanced European countries directly in stock markets. - Brazil shows very tight inward and outward spillover connections to its neighborhood in CDS markets and, to a lesser extent, to other large emerging markets (BRICs and MINTs). - FX markets: - Spillovers show different structure with less geographical clustering. - Volatility spillovers in FX are more concentrated among fewer countries; Brazil is more closely linked to other large and financially integrated emerging economies such as Mexico, South Africa or Turkey (mutual spillovers) than to Argentina or Chile. - Brazil is subject to shocks from the euro area and the U.K. and is a net transmitter of shocks in FX markets due to its regional weight. - U.S. role: - The U.S. is a source of shocks in stock markets to most countries in the sample and functions as a global hub connecting regions; shocks in other countries can be spread with delay through the U.S. ### Individual bank network and major bank-level statistics - Geographic clustering: - Individual bank network shows strong geographical clustering within and between countries, forming tightly connected regions—consistent with country-level stock market findings. - Four large Brazilian banks (volatility spillover characteristics and key figures): - Domestic volatility shocks from other Brazilian banks to each Brazilian bank are large and range between 63.4 to 88.1 percent for Itaú and Santander, respectively. - Examples: - Itaú Unibanco: domestic shocks sum to 63.4 percent; U.S. shocks are 24.4 percent. - Santander (Brazil): domestic shocks sum to 88.1 percent; U.S. shocks around 6 percent. - Residual shocks to Brazilian banks also come from Spain or France. - Brazilian banks are prone to shocks from U.S. banks; however, Brazilian banks do not produce significant volatility spillovers outward (limited reciprocity); spillovers from third countries to Brazil are generally not sizable. - Figure 36: Summary pairwise directional connectedness from countries (sum of banks per country) to four large Brazilian banks — sample values (selected): - BR01 (Banco do Brasil) domestic share: 0.523; contributions from USA column: 0.092. - BR02 (Itaú Unibanco) domestic share: 0.397; contributions from USA column: 0.244. - BR03 (Bradesco) domestic share: 0.466; contributions from USA column: 0.108. - BR04 (Banco Santander) domestic share: 0.857; contributions from USA column: 0.060. - (Matrix entries expressed as shares presented in the heatmap; larger shocks shown in red, smaller in green.) ### Additional notable model/data choices and practical considerations - High-dimensional generalized variance decompositions used to handle many variables. - Shrinkage methods (LASSO and Ridge) introduced to handle large datasets and parametrization; in practice LASSO shrinkage not relevant for country-level applications (small number of countries vs. time-series length); for individual banks some shrinkage penalizes large geographic distance. - Range-based volatility estimates for individual banks require high, low, opening and closing prices each day. *Source: IMF Staff analysis in the provided content unit.* ### Appendix I. Characterization of Baseline and Adverse Scenarios ### Appendix I. Characterization of Baseline and Adverse Scenarios ### Overview - The baseline and adverse scenarios are characterized by a set of global and domestic, macro and financial variables (Table 1). - Global indicators include global real GDP growth, U.S. interest rates and commodity prices. - Domestic indicators include real GDP growth, nominal GDP growth, inflation rate, unemployment rate, interest rates, exchange rate, equity price, house price, EMBI spread, and bank credit growth. ### Baseline scenario construction - The baseline scenario is based on the October 2017 WEO projections. - Macro and financial variables for the baseline were either obtained from WEO projections or projected using variables available from WEO projections. - Variables that are part of WEO projections: - Projected by either IMF’s Research Department (global WEO assumptions) or the country team. - Quarterly figures for outer years were interpolated based on the (average) annual figure available from WEO projections. - Variables not part of WEO projections: - Projected by the FSAP team using a set of simple regression models. Specific estimation methods are given in Table 2. ### Estimation methods for selected baseline variables (Table 2) - Bank funding rate: - Projection based on the change in the estimated bank funding rate: - Bank funding rate_t = 2.3 + 0.6 Selic_t. - Bank lending rate: - Projection based on the change in the estimated bank lending rate: - Bank lending rate_t = 15.3 + 0.9 Selic_t -0.5 Real GDP growth_{t-2}. - Real equity price: - Projection based on the estimated real equity price growth: - Real equity price growth_t = -4.0 + 1.7 Real GDP growth_t -1.3 ∆Selic_t -9.3 ∆EMBI spread_t. - The projection is being smoothed to remove some volatility. - Real house price: - Projection based on the estimated real house price growth: - Real house price growth_t = -3.5 + 0.7 ∆Real house price growth_{t-4} + 0.9 Real GDP growth_t -1.8 ∆Selic_t. - The projection is being smoothed to remove some volatility. - EMBI spread: - Projection based on the estimated change in EMBI spread: - ∆EMBI spread_t = -0.4 + 0.1 Exchange rate depreciation_t. - The estimated EMBI spread is subject to the historical minimum level. - Credit to GDP: - Projection based on the estimated change in credit to GDP: - ∆Credit/GDP_t = 2.0 + 0.2 Real GDP growth_{t-1} +0.2 Real house price growth_{t-4}. - The projection is being smoothed to remove some volatility. ### Adverse scenario design and simulation approach - The adverse scenario was simulated using the GFM to capture key domestic and external macrofinancial risks faced by Brazil. The GFM provides simulations as deviations from the baseline; combining the GFM deviations with the baseline yields the complete adverse scenario. - The adverse scenario features a double-dip recession with a sudden stop of capital flows driven by negative macrofinancial shocks. - The adverse scenario is largely triggered by domestic developments (loss of confidence) and occurs in a challenging global environment. - Policy simulation constraints: - Conventional monetary policy responds endogenously with nominal policy interest rate cuts subject to effective zero lower bound constraints worldwide. - No additional unconventional monetary policy responses are envisaged. - For the stress testing exercise, automatic relaxation of regulatory capital requirements in Brazil is assumed so that the simulation abstracts from any increase in banks’ lending interest rate in response to deterioration in banks’ capital position (relative to the regulatory requirement). ### Global layers of the adverse scenario (summary of modelling) - Asset price corrections globally: - Declines in house prices and equity prices across countries; flight to quality compresses term premium in safe-haven economies and raises long-term bond yields in other financially open economies. - Market corrections trigger persistent credit cycle downturns, captured by increasing default rates (Layer 1 and Layer 2, Table 3). - China financial stress and slowdown: - Real house price declines by 12.5 percent and real equity price declines by 25 percent. - Depreciation of the Chinese renminbi by 12.5 percent in real terms against the U.S. dollar. - Short-term money market spread rises by 150 basis points and long-term government bond yield rises by 75 basis points (Layer 3, Table 3). - Suppressed global economic risk-taking: - Domestic demand falls: private consumption, residential investment and business investment decline by 1, 4 and 4 percent, respectively, across all countries (Layer 4, Table 3). - Brazil faces inward spillovers via trade and financial channels: - Negative spillovers from asset price co-movements, worsening terms of trade due to falling commodity prices, and lower export demand. ### Domestic layers of the adverse scenario (summary of modelling) - Loss of confidence and sudden stop of capital flows: - Sudden stop causes capital flight and massive pressure on the Brazilian real, with a 50 percent real depreciation against the U.S. dollar. - Long-term government bond yield increases by 700 basis points. - Short-term money market spread (relative to the policy rate) edges up 50 basis points. - Real equity price declines by 30 percent; real house price falls by 7.5 percent (Layer 5, Table 3). - Financial stress causes credit losses, higher lending interest rates, constrained credit supply (Layer 6, Table 3). - Weak domestic demand: - Private consumption, residential investment and business investment decline by 4, 16, and 16 percent, respectively (Layer 7, Table 3). ### Adverse scenario shock magnitudes (Table 3, selected entries) - Layer 1: Asset price corrections globally, 2018 - House price; Housing risk premium shocks: -2,-4,-6 percent - Equity price; Equity risk premium shocks: -9,-12,-15 percent - Term premium; Duration risk premium shocks: Safe havens -25 basis points; Other financially open economies +50 basis points - Layer 2: Credit cycle downturn globally, 2018-20 - Mortgage loan default shocks: +0.1,+0.2,+0.3 percentage point - Corporate loan default shocks: +0.3,+0.4,+0.5 percentage point - Layer 3: Financial stress and growth slowdown in China, 2018 - Real house price; Housing risk premium shocks: -12.5 percent - Real equity price; Equity risk premium shocks: -25 percent - Money market interest rate spread; Credit risk premium shocks: +150 basis points - Long-term government bond yield; Duration risk premium shocks: +75 basis points - Real bilateral exchange rate; Current risk premium shocks: +12.5 percent - Layer 4: Suppressed economic risk taking globally, 2018-20 - Private consumption; Consumption demand shocks: -0.5 percent - Residential investment; Residential investment demand shocks: -2 percent - Business investment; Business investment demand shocks: -2 percent - Layer 5: Sudden stop of capital flows in Brazil, 2018-19 - Real house price; Housing risk premium shocks: -7.5 percent - Real equity price; Equity risk premium shocks: -30 percent - Money market interest rate spread; Credit risk premium shocks: +50 basis points - Long-term government bond yield; Duration risk premium shocks: +700 basis points - Real bilateral exchange rate; Current risk premium shocks: +50 percent - Layer 6: Credit risk materialization in Brazil, 2018-20 - Mortgage loan default shocks: +1 percentage point - Corporate loan default shocks: +4 percentage points - Layer 7: Weak domestic demand in Brazil, 2018-20 - Private consumption; Consumption demand shocks: -4 percent - Residential investment; Residential investment demand shocks: -16 percent - Business investment; Business investment demand shocks: -16 percent ### Adverse scenario outcomes and key macrofinancial results (Table 4 highlights) - Cumulative and annual macro outcomes: - Real GDP falls 11.8 percent below the baseline by 2018. - Annual real GDP growth under the adverse scenario: - 2018: -4.3 - 2019: -3.9 - 2020: 1.6 - Inflation rate (CPI-based) under the adverse scenario: - 2018: 6.1 - 2019: 5.8 - 2020: 1.3 - Unemployment rate under the adverse scenario: - 2018: 13.8 - 2019: 16.1 - 2020: 15.9 - Exchange rate (real per U.S. dollar) under the adverse scenario: - 2018: 4.6 - 2019: 4.8 - 2020: 4.4 - Equity price (2016=100) under the adverse scenario: - 2018: 86 - 2019: 88 - 2020: 103 - House price (2016=100) under the adverse scenario: - 2018: 91 - 2019: 94 - 2020: 97 - Selic under the adverse scenario: - 2018: 11.0 - 2019: 8.1 - 2020: 4.5 - Long-term government bond yield under the adverse scenario: - 2018: 15.8 - 2019: 16.4 - 2020: 15.0 - Bank funding rate under the adverse scenario: - 2018: 10.4 - 2019: 8.4 - 2020: 5.3 - Bank lending rate under the adverse scenario: - 2018: 26.7 - 2019: 26.6 - 2020: 23.2 - EMBI spread under the adverse scenario: - 2018: 5.0 - 2019: 4.9 - 2020: 4.0 - Bank credit growth under the adverse scenario: - 2018: 0.1 - 2019: -2.6 - 2020: 5.9 - Additional features and dynamics: - The economy enters another recession with negative real GDP growth for 2018 and 2019 (-4.3 and -3.9 percent, respectively). - CPI-based inflation accelerates during 2018-19 due to large exchange rate depreciation before moderating in 2019. - The unemployment rate rises above 16 percent, the highest level ever. - Financial stress manifests across markets: exchange rate depreciates more than 50 percent, bond and stock markets hit severely (share price falling by almost 50 percent and government bond yield increases by 800 basis points by end-2018), house price declines by about 7 percent. - Monetary policy initial tightening followed by loosening: - Selic is raised to 11 percent on average in 2018 to contain inflationary pressures from exchange rate depreciation. - Selic reduced to 8.1 and 4.5 percent on average in 2019 and 2020, respectively. - Spread between bank funding rate and bank lending rate is compressed; spread between bank lending rate and policy rate reaches 19 percentage points. - Banks’ credit supply becomes constrained during 2018-19: - Bank credit remains essentially flat in 2018 and contracts by 2.6 percent in 2019. - Bank credit growth turns positive in 2020, broadly in line with the economic recovery. *Source: Appendix I. Characterization of Baseline and Adverse Scenarios (from the supplied content).* ### 5. Regulatory and ### cr18344 - 5. Regulatory and ### Regulatory/accounting and market-based standards — Calibration of risk parameters - Point-in-time credit risk proxies - Marking-to-market for assessing market risk - National regulatory framework based on Basel III, with all banks under the standardized approach - Hurdle rates based on minimum capital requirements (CET1, T1 and total capital) and applicable surcharges for D-SIBs; for the baseline scenario, hurdle rates also including prevailing conservation capital buffer ### Reporting format for results (regulatory/accounting and market-based) - System-wide capital shortfalls - Number of banks that fail to meet the hurdle and their assets share in the banking sector ### BANKING SECTOR: LIQUIDITY RISK — Institutional perimeter and data - Institutions included: - Cash-flow analysis: 12 largest banks - LCR test: 12 largest banks (including the 8 banks currently subjected to the requirement) - NSFR test: 12 largest banks - Market share: 90 percent of banking sector assets - Data and baseline date: - Supervisory data - Data as of September 2017 - Scope of consolidation: perimeter of individual banks ### BANKING SECTOR: LIQUIDITY RISK — Channels of risk propagation and methodology - The FSAP liquidity stress testing exercise comprise three types of tests: - (i) Cash-flow analysis using maturity buckets; - (ii) LCR test; and - (iii) NSFR test ### BANKING SECTOR: LIQUIDITY RISK — Risks and buffers - Risks: - Funding liquidity risk (short-term liquidity outflows) - Market liquidity shock (asset price shocks and fire-sales) - Buffers: - Counterbalancing capacity includes liquidity obtained from markets through asset sales (subject to certain haircuts) and from BCB’s standing facilities - Expected cash inflows are also included ### BANKING SECTOR: LIQUIDITY RISK — Tail shocks - Size of the shock: - The haircuts are calibrated consistent with the increase in government bond yields in the solvency stress test. - The run-off rates are calibrated to reflect scenarios of system-wide deposit runs and dry-up of wholesale funding. Parameters are chosen to match volatility of deposits and reputational risk events. ### BANKING SECTOR: LIQUIDITY RISK — Regulatory standards and calibration of risk parameters - Calibration of risk parameters: - Regulatory: haircuts and run-off rates based on regulatory LCR/NSFR parameters. - Stressed: more severe haircuts similar to the adverse scenario in the solvency stress tests and larger run-off rates to reflect more severe episodes of market and funding risks based on historical events. - Regulatory standards: - For the LCR and NSFR tests, the hurdle is set to 100 percent. - For the cash-flow analysis, the hurdle is to have a non-negative cash balance. ### BANKING SECTOR: LIQUIDITY RISK — Reporting format for results - Number of banks that fail to meet the hurdle and their assets share in the banking sector - Bank-level survival period in days, number of banks that can still meet their obligations ### BANKING SECTOR: CONTAGION RISK — Institutional perimeter and data - Institutions included: 12 largest banks - Market share: 90 percent of banking sector assets - Data and baseline date: - Supervisory data - Data as of September 2017 ### BANKING SECTOR: CONTAGION RISK — Channels of risk propagation and methodology - Interbank network analysis based on Espinosa-Vega and Solé (2010) ### BANKING SECTOR: CONTAGION RISK — Risks and buffers - Risks: Credit and funding losses related to interbank cross-exposures as a result of some bank defaults - Buffers: Banks’ own capital ### BANKING SECTOR: CONTAGION RISK — Tail shocks (size of the shock) - Default threshold: banks would default if their CET1 capital ratios fall below 4.5 percent (regulatory minimum) - For the standalone exercise, each individual bank is assumed to default at the time - For the extended solvency stress testing exercise, the starting point is based on the solvency stress test results in the same period ### BANKING SECTOR: CONTAGION RISK — Reporting format for results - System-wide capital shortfall, bank-level capital shortfall - Number of banks that become undercapitalized - Evolution and direction of spillovers ### INTERCONNECTEDNESS ANALYSIS — Cross-sectoral linkages - Data and methodology: - Approach: “Flow-of-funds” type of data collected BCB for the disaggregation of ten institutional sectors in the System of National Accounts of Brazil: non-financial corporations, government, financial sector, families and NPISH, rest of the world, BCB, other deposit-taking institutions, insurers and pension funds, other financial institutions, monetary investment funds and non-monetary investment funds. - Financial instruments include deposits, debt securities, derivatives, equity and credit. - Sample: quarterly data between 2015Q1 and 2016Q4. - Methodology: Dees, Henry and Martin (eds) (2017), Castrén and Rancan (2014) and e Silva (2016). ### INTERCONNECTEDNESS ANALYSIS — Cross-border spillovers - Data and methodology: - Approach: Examine spillover risks between Brazil and the main relevant countries from a financial stability perspective across equity and FX markets and across sovereign CDS and bond/EMBI spreads. - Sample: Weekly observations between 2007Q4 and 2017Q4 in each series for Argentina, Brazil, Chile, China, Colombia, France, Germany, India, Indonesia, Japan, Mexico, Russia, South Africa, Spain, Turkey, United Kingdom and the United States. - Methodology: Demirer, Diebold, Liu and Yilmaz (2016) and Diebold and Yilmaz (2015). ### Appendix III — Detail on Balance Sheet and Income Statement Projections for the Bank Solvency Stress Tests: Overview of the Balance Sheet Approach - Balance sheet exposures: - Growth of gross exposures by types, such as total gross loans and gross holding of debt securities, identical to overall credit growth, with the balance sheet composition unchanged - Total assets not affected by exchange rate movements - Net interest income components: - Interest income from lending activity: Amount of loans (adjusted for balance sheet growth and existing NPLs) multiplied by estimated interest rate for lending activity - Bank-level interest rate for lending activity estimated using a panel regression model, including its lag, real GDP growth, system-wide lending rate, and EMBI spread - Interest income from other activities: Amount of other claims (adjusted for balance sheet growth) multiplied by estimated interest rate for non-lending activity; bank-level rate estimated using panel regression with its lag and EMBI spread - Interest expense: Amount of funding (adjusted for balance sheet growth and accounting for potential additional required funding) multiplied by estimated interest rate for funding; bank-level rate estimated using panel regression including its lag and system-wide funding rate - Pre-impairment income, other items: - Trading income: Total assets (adjusted for balance sheet growth) multiplied by average trading income (relative to total assets); average trading income based on the historical benchmark during 2007-17 - Fee and investment income: - Fee income: Total assets (adjusted for balance sheet growth) multiplied by estimated fee income (relative to total assets); bank-level fee income estimated using a panel regression model including its lag, real GDP growth, Selic and equity price movement - Investment income: Total assets (adjusted for balance sheet growth) multiplied by average investment income (relative to total assets); average investment income based on the 2017Q1-Q3 level - Other operating income and operating expense: Total assets (adjusted for balance sheet growth) multiplied by average ratios; averages based on 2017Q1-Q3 level - Impairment cost: - Combination of credit loss and market loss - Credit loss: Three approaches - First approach: based on lending impairment and other impairment; lending impairment cost (relative to total loans) estimated using a panel regression model including its lag, real GDP growth, and change in EMBI spread; other impairment cost based on the 2017Q1-Q3 level - Second approach: based on credit loss due to total impairment; total impairment cost (relative to total assets) estimated using a panel regression model including its lag, real GDP growth, and change in EMBI spread - Third approach: based on nonperforming loans (NPLs); total NPL ratio estimated using a panel regression model including its lag, real GDP growth, and change in system-wide lending rate; estimated NPL ratio then translated into lending impairment cost; total impairment cost derived from sum of lending impairment and other impairment (other impairment based on the 2017Q1-Q3 level) - Market loss: Marking-to-market related to (i) bond yield movements for holding of debt securities in the trading account, (ii) equity price movements for holding of equity securities, (iii) exchange rate movements for the net open foreign-exchange position, and (iv) commodity price movements for the net open commodity position - Net income and its effect on capital: - Net income before taxes: Sum of pre-impairment income and impairment cost - Taxes: Income tax based on the applicable tax rate at 45 percent; Tax credit allowed up to the amount of credit loss - Net income after taxes: Sum of net income before taxes and taxes - Dividend payout: Based on the 2016 level for positive net income and zero otherwise, subject to restrictions on dividend distribution - Profits attributed to capital: Sum of net income after taxes and dividend payout - Capital: Affected by retained profits, and unrealized gain/loss associated with the available-for-sale (AFS) and HTM accounts - Risk-weighted assets: Adjusted for balance sheet growth, and exchange rate movements (for foreign-currency exposures) - Total exposures: Adjusted for balance sheet growth, and exchange rate movements (for foreign-currency exposures) ### Regression specifications — Satellite pre-impairment income and credit risk models (high-level) - Satellite interest rate models: - Dependent variables: InterestRate_Lending_A, InterestRate_NonLending_A, InterestRate_Funding_A (all based on the average over four quarters) - Explanatory variables include Selic, lending rate, funding rate, EMBI spread (all annual average), and real GDP growth (year-on-year); models estimated via fixed-effects panel regression - Main stress test results based on Model L.1, Model NL.1 and Model F.1 - Satellite models for fee and trading income: - Dependent variables: FeeIncome_A and TradingIncome_A over four quarters - Explanatory variables include real GDP growth, change in exchange rate, change in equity price (year-on-year), Selic and long-term bond (annual average) - Main stress test results based on Model FI.1 for fee income; trading income estimated based on historical benchmark 2007-17 - Satellite credit risk models: - Dependent variables: CreditCost_A, TotalImpairment_A, and NPL_Total (logit-transformed) - Explanatory variables include real GDP growth, change in unemployment rate, change in lending rate, and change in EMBI spread (year-on-year) - Main stress test results based on Model I.1; alternative results based on Model II.1 and Model III.1 - Additional modelling notes: - Under the third approach, problem loans used to estimate lending impairment cost; relationships estimated via fixed-effects panel regression - Problem loans can be estimated in total and by loan portfolios (persons vs entities, by currencies, by economic sectors) - Important explanatory variables for problem loans by portfolios include change in house price and change in exchange rate for property-related and other business sectors; change in commodity price does not appear to influence problem loans dynamics for the commodity sector *International Monetary Fund — cr18344 — 5. Regulatory and* ### Appendix IV. Calibration of Parameters for Bank Liquidity Stress ### Appendix IV. Calibration of Parameters for Bank Liquidity Stress Tests ### Funding run-off principles and time profile - Funding pressures captured through time profiles of run-off rates for different funding sources. - Guiding principles: - Wholesale funding sources have higher run-off rates than retail funding sources. - Secured funding run-off rates are lower than unsecured funding run-off rates. - Sight deposits and/or deposits protected by deposit insurance are withdrawn at a slower pace than time deposits and/or uninsured deposits. - Cash outflows are assumed to be protracted—lasting for up to one year—with run-off rates declining over time. - Highest rates correspond to short-term maturity buckets (0–30 days). - Medium-size rates applied to intermediate maturity buckets (30–180 days). - Lowest rates correspond to longer-term time buckets (after 180 days). - Run-off rates on foreign currency funding are more stringent: - Insured deposits in foreign currency: run-off rates higher by 10 percentage points. - Uninsured deposits in foreign currency: run-off rates higher by 30 percentage points. ### Asset roll-off (conversion of maturing loans into cash proceeds) - Specific roll-off rates applied by asset/maturity: - Performing loans to retail and nonfinancial corporate customers: 50 percent. - Maturing loans to other entities (including financial institutions), non-operational deposits held at other financial institutions, and cash flows from debt securities: 100 percent. - For secured lending, roll-off rates aligned with valuation losses implied by haircuts on the underlying security. - Note on definition: A roll-off rate is the fraction of the asset amount maturing in a given period that is converted into a cash inflow by the bank. Exception: contingent credit/liquidity lines (see Table notes). ### LCR / Cash-flow stress parameters (selected items) - Cash outflows: run-off rates applied by type and maturity buckets as guided by the principles above (time buckets: 0–30 days, 30–180 days, after 180 days). - Cash inflows: roll-over rates applied to maturing assets to estimate cash inflows (see roll-off rates above). - Asset price haircuts applied to HQLA and other asset classes (see Tables referenced in source for full calibration). ### NSFR proxy calibration (selected ASF and RSF factors from Table 3) - ASF Factor 100percent: - (a) Total regulatory capital: 100percent / 100percent (Baseline / Adverse) - (b) Other capital instruments with effective maturity of one year or more: 100percent / 100percent - (c) Liabilities with effective residual maturity of one year or more: 100percent / 100percent - ASF Factor 95percent: - Stable non-maturity (demand) deposits and term deposits with residual maturity of less than one year provided by retail and SME customers: 95percent / 81percent - ASF Factor 90percent: - Less stable non-maturity (demand) deposits and term deposits with residual maturity of less than one year provided by retail and SME customers: 90percent / 55percent - ASF Factor 50percent: - (a) Funding (secured or unsecured) with residual maturity of less than one year provided by non-financial corporate clients: 50percent / 45percent - (b) Operational deposits: 50percent / 45percent - (c) Funding with residual maturity of less than one year from sovereigns, public sector entities (PSEs), and multilateral and national development banks: 50percent / 45percent - (d) Other funding (secured or unsecured) with residual maturity of not less than six months and less than one year not included in the above categories, including funding provided by central banks and financial institutions: 50percent / 45percent - ASF Factor 0percent: - (a) All other liabilities and equity categories not included above, including liabilities without stated maturity: 0percent / 0percent - (b) Derivatives payable net of derivatives receivable, if payables are greater than receivables: 0percent / 0percent - RSF Factor 0percent: - (a) Coins and banknotes: 0percent / 0percent - (b) All central bank reserves (including required and excess reserves): 0percent / 0percent - (c) Unencumbered loans to banks subject to prudential supervision with residual maturity of less than six months: 0percent / 0percent - RSF Factor 5percent: - Unencumbered Level 1 assets - excluding coins, banknotes and central bank reserves (above): 5percent / 10percent - RSF Factor 15percent: - Unencumbered Level 2A assets: 15percent / 23percent - RSF Factor 50percent: - (a) Unencumbered Level 2B assets: 50percent / 55percent - (b) HQLA encumbered for a period of six months or more and less than one year: 50percent / 55percent - (c) Loans to banks subject to prudential supervision with residual maturity six months or more and less than one year: 50percent / 55percent - (d) Deposits held at other financial institutions for operational purposes: 50percent / 55percent - (e) All other assets not included in the above categories, including loans to non-bank financial institutions, loans to non-financial corporate clients, loans to retail and small business customers, and loans to sovereigns, central banks and PSEs: 50percent / 55percent - RSF Factor 65percent: - (a) Unencumbered residential mortgages with a residual maturity of one year or more and with a risk weight of less than or equal to 35percent: 65percent / 75percent - (b) Other unencumbered loans not included in the above categories, excluding loans to financial institutions, with a residual maturity of one year or more and with a risk weight of less than or equal to 35percent under the Standardized Approach: 65percent / 75percent - RSF Factor 85percent: - (a) Other unencumbered performing loans with risk weights greater than 35percent under the Standardized Approach and residual maturity of one year or more, excluding loans to financial institutions: 85percent / 95percent - (b) Unencumbered securities that are not in default and do not qualify as HQLA including exchange-traded equities: 85percent / 95percent - (c) Physical traded commodities, including gold: 85percent / 95percent - RSF Factor 100percent: - (a) All assets encumbered for a period of one year or more: 100percent / 100percent - (b) Derivatives receivable net of derivatives payable, if receivables are greater than payables: 100percent / 100percent - (c) All other assets not included in the above categories, including non-performing loans, loans to financial institutions with residual maturity of one year or more, non-exchange-traded equities, fixed assets, pension assets, intangibles, deferred tax assets, retained interest, insurance assets, subsidiary interests, and defaulted securities: 100percent / 100percent - Off-Balance Sheet categories: - RSF Factor 5percent: Irrevocable and conditionally revocable credit and liquidity facilities to any client: 5percent / 10percent ### Cash-flow stress tests: notes - Run-off rates and roll-off rates definitions: - Run-off rate: fraction of the liability amount maturing in a period that is withdrawn by claim holders (sight deposits and contingent liabilities are exceptions). - Roll-off rate: fraction of the asset amount maturing in a given period that is converted into a cash inflow by the bank. Exception: cash inflows from contingent credit/liquidity lines (treated differently as withdrawals of initial contracted amount). - Run-off rates highlighted in the original tables may be defined with respect to the initial outstanding balance for certain categories. *Italic: Source — Appendix IV, “Calibration of Parameters for Bank Liquidity Stress Tests.”* ### Appendix V. Detailed Methodology for the Interbank Network Analysis ### Balance-sheet identity and contagion channels - Balance sheet identity for bank i in a network of N banks (notation preserved in source). - Credit loss channel: - Near-default of bank h induces credit losses λ x_h i at bank i. - A bank fails if its capitalization falls below a threshold T (in percent of risk-weighted assets). - Loss given default parameter: λ governs severity of credit loss. - Funding loss channel: - If bank i cannot fill a fraction of funding ρ x_i h obtained from failed bank h, it is forced to sell assets at discount rate δ. - Loss of funding parameter: ρ governs severity of funding loss. - Discount rate parameter: δ governs price impact of asset sales. - Failure condition formulations (as in source): - Credit shock simulation: (k_i − sum_{h ∈ F_t} λ x_h i) < T * RWA_i - Credit-funding shock simulation: (k_i − sum_{h ∈ F_t} (λ x_h i + δ ρ x_i h)) < T * RWA_i - Simulation continues until no more bank failures occur. ### Benchmark parameter calibration - Benchmark parameter values used in analysis: - λ = 0.5 for total interbank exposures (approximation based on weighted share of λ = 1 for unsecured exposures and λ = 0.1 for secured exposures). - ρ = 0.5. - δ = 0.5. *Italic: Source — Appendix V, “Detailed Methodology for the Interbank Network Analysis.”* ### Appendix VI. Network Analysis Measures and Concepts ### Network basics and components - Network: nodes (vertices) connected by links (edges); links determined by exposures or flows at a given time. - Financial networks are weighted and directed. - Adjacency matrix: matrix representation; symmetric in undirected networks, asymmetric in directed/weighted networks. - Giant component: largest connected component used for metrics like diameter or average path length when network is not fully connected. - Leaves: nodes with only one link; may be shock originators (source) or sinks. ### Topological measures and interpretation - Density: portion of potential links that are actual connections; low density means few connections. - Assortativity: tendency of nodes to attach to similar nodes (by degree). Positive assortativity: highly connected nodes attach to highly connected nodes. Negative assortativity: low-degree nodes attach to highly connected nodes. Relevant for core-periphery inference. - Diameter: - Greatest distance between any pair of nodes (based on shortest paths). - Empirical financial networks: diameter low if between 3 and 8. Fully connected network has diameter 1. - Average path length: - Average number of steps along shortest paths for all pairs of nodes. - Financial networks: average path lengths tend to be low, below 3. - Clustering coefficient: - Measures degree to which nodes cluster together (range 0–1). Disconnected node has 0. - High clustering + low diameter + low average path length → small-world properties. ### Centrality and related measures - Degree centrality: sum of in-coming and out-going links; in- and out-degree; in-strength and out-strength for weighted networks. - Average neighbor degree: mean degree centrality of one-step neighbors; for degree k = 1 called average nearest neighbor degree. - Betweenness centrality: number of shortest paths between all pairs that pass through a node; captures intermediary role beyond degree. - PageRank centrality: variant of Eigenvector Centrality; assigns score based on number and weight of incoming links; higher PageRank denotes more central role. - Hub and authority scores: - Hubs score = sum of authorities scores of successors. - Authorities score = sum of hubs scores of predecessors. - Sum of all hubs scores = 1; sum of all authorities scores = 1. - PageRank is similar to authority score. ### Structures and similarity measures - Core-periphery structure: core entities serve as hubs between periphery entities; core institutions interact intensely among themselves and can be systemically important. - Jaccard similarity index (J): probability of observing a given connection in one network conditional on observing it in another; intersection over union of ordered vectors. - Cosine similarity index (C): for directed weighted networks; measures cosine of angle between two vectors via normalized dot product; used as proximity measure in multilayer network analysis. *Italic: Source — Appendix VI, “Network Analysis Measures and Concepts.”* ### 19. The chord diagrams of all financial instruments subnetworks show that a multilayer ### 19. The chord diagrams of all financial instruments subnetworks show that a multilayer network approach is highly informative of interconnectedness in the system and channels of transmission of potential shocks with financial stability implications ### Summary of approach and insights - A multilayer network (multiplex) approach is used to analyze cross-exposures across financial instruments and sectors, making visible: - the strongest links, - the concentration of cross-exposures among specific sectors in each financial instrument, - the relative role of each sector in terms of the number and size of links. - Chord diagrams visualize the different cross-exposures in a granular way. - Table 1 complements the diagrams with sectoral centrality measures to quantify and compare leading centrality roles of sectors across subnetworks. ### Centrality measures — key statistics (2016 Q4) - Sector labels used in the table (columns): BCB, BNK, INSPF, FIN, IFM, IFNM, GOV, NFC, HH, ROW. - Selected centrality metrics presented include: In-strength, Out-strength, Betweeness, PageRank, Multiplex PageRank, Hub score, Authority score. - Example values (first block of centrality rows shown in the table): - In-strength: 2,928,889; 8,943,456; 1,848,216; 3,269,111; 4,924,068; 572,767; 6,167,831; 9,383,692; 2,353,143; 2,660,962 - Out-strength: 2,969,017; 8,358,636; 1,930,291; 3,296,252; 4,942,426; 526,429; 3,876,539; 6,235,561; 6,238,167; 4,678,815 - Betweeness: 0.13; 0.75; 0.75; 0.75; 0.13; 0.13; 0.75; 0.75; 0.25; 0.63 - PageRank: 0.09; 0.11; 0.11; 0.11; 0.10; 0.10; 0.11; 0.11; 0.10; 0.09 - Multiplex PageRank: -; -; -; -; -; -; -; -; -; - - Hub score: 0.07; 0.11; 0.11; 0.11; 0.10; 0.10; 0.11; 0.11; 0.11; 0.10 - Authority score: 0.09; 0.11; 0.11; 0.11; 0.10; 0.10; 0.11; 0.11; 0.10; 0.09 - Additional instrument-specific rows in the table report analogous numeric vectors for In-strength, Out-strength, Betweeness, PageRank, Multiplex PageRank, Hub score, and Authority score across the same sectors. Example additional numeric vectors found in the table: - In-strength (second block): 2,729,068; 3,835,808; 0; 15,161; 0; 0; 434,486; 0; 0; 242,177 - Out-strength (second block): 94,457; 2,608,704; 28,926; 366,140; 842,401; 13,534; 1,208,361; 712,380; 1,379,755; 2,043 - Betweeness (second block): 0.33; 4.17; 0.00; 0.67; 0.00; 0.00; 0.50; 0.00; 0.00; 2.33 - PageRank (second block): 0.24; 0.25; 0.02; 0.11; 0.02; 0.02; 0.02; 0.02; 0.02; 0.31 - Multiplex PageRank (second block): 0.07; 0.14; 0.06; 0.11; 0.08; 0.06; 0.13; 0.08; 0.21; 0.06 - The table covers multiple financial instruments including (as labeled at the table foot): Long-term securities, Short-term loans, Long-term loans, Total, Currency and deposits, Short-term securities, Equity, Investment funds shares/units, Insurance, pension and standardized guarantee schemes, Financial derivatives, Other accounts receivable/payable. - Representative numeric entries for various instruments appear throughout the table; examples include: - A row showing In-strength: 0; 1,418,510; 0; 594,471; 0; 0; 4,121,139; 398,520; 1,125; 1,071,067 - A row showing Out-strength: 2,449,103; 1,179,013; 231,762; 141,213; 1,692,328; 33,116; 167,101; 537,306; 533,156; 640,734 - A row showing In-strength: 139,675; 1,058,874; 170,057; 2,285,629; 0; 0; 0; 4,897,615; 0; 1,081,398 - A row showing Out-strength: 0; 712,612; 256,228; 1,945,971; 82,801; 302,287; 611,826; 1,763,239; 1,671,737; 2,286,546 - A row showing In-strength: 0; 0; 0; 0; 4,826,577; 541,366; 0; 0; 0; 0 - A row showing Out-strength: 0; 103,390; 1,191,979; 313,501; 2,000,068; 84,830; 208,850; 522,860; 773,795; 168,670 - A row showing In-strength: 60,035; 758,525; 115,246; 67,783; 17,698; 29,751; 198,444; 1,761,669; 674,324; 133,646 - A row showing Out-strength: 231,436; 343,213; 156,631; 142,839; 17,749; 81,411; 9,407; 2,327,204; 140,654; 366,577 ### Methodological notes and data - In-strength corresponds to total liabilities and Out-strength corresponds to total assets. - Multiplex PageRank is the combined version to allow sectors to gain centrality if they are connected to central nodes in other subnetworks; it was computed using a standard 0.85 damping factor. It is not available for the aggregated network. - The largest three values of selected centrality measures are highlighted in red in the original table to check consistency across indicators. - Table source: IMF staff estimates and BCB. ### Applications and implications for financial stability - The multilayer network approach and chord-diagram visualization help identify key channels of transmission of potential shocks across instruments and sectors. - Sectoral centrality metrics quantify which sectors are most central (by number and size of links) in specific instrument subnetworks and in the aggregate network, informing where shocks could propagate most strongly. - Combining visual (chord diagrams) and metric (centrality measures) information supports targeted monitoring and potential policy interventions focused on highly central sectors or concentrated cross-exposures. *Source: IMF staff estimates and BCB.* --- _Source: https://www.imf.org/-/media/files/publications/cr/2018/cr18344.pdf_