## 1colea2022007 ## Source details **Canonical URL:** [1colea2022007](https://www.imf.org/-/media/files/publications/cr/2022/english/1colea2022007.pdf) ## Other formats - [Markdown version](/-/media/files/publications/cr/2022/english/1colea2022007.pdf.md) - [Structured JSON version](/-/media/files/publications/cr/2022/english/1colea2022007.pdf.json) --- ### Executive summary — background and macrofinancial context - FSAP risk analysis conducted after the initial COVID shock and lockdowns, during a strong economic recovery underway in 2021; high uncertainty around the evolution of the COVID-19 virus and the recovery led to significant downside risk and revisions. - Banking system entered the COVID-19 pandemic from a position of relative strength; authorities implemented fiscal and monetary measures, SFC allowing release of countercyclical provisions, temporary borrower grace periods and loan modifications, and the Program to Support Debtors (PAD) (phased out at end-August 2021). - Funding structure: “Banks’ funding is largely deposit-based, with almost 80 percent of funding coming from the wholesale sources.” - System concentration: five largest conglomerates own about 60 percent of system assets (later noted as controlling 80 percent of banking sector assets in another section); increasing cross-border exposures concentrated in Central America. ### Solvency stress test — design, methodology, and scenarios - Coverage and horizon: - Sample: 12 banks covering 93.8 percent of banking sector assets (all four D-SIBs, five non-systemic domestically owned banks, and three non-systemic foreign-owned banks). - Solvency stress scenarios span three years from mid-2021 to mid-2024. - Adverse scenario calibration (deviations from baseline): - Adverse scenario GDP path: Year 1: -1.2 percent; Year 2: -1.3 percent; Year 3: 4 percent; averaging 0.5 percent real GDP growth over three years. - Housing prices drop: 21 percent. - Exchange rate depreciation: about 25 percent. - Long-term sovereign yields rise: 190 basis points. - Interbank spread widening: 125 basis points. - Subsidiaries P&L shocks: 50 percent in year 1 and 25 percent in year 2. - Modeling and key assumptions: - Quasi-static balance-sheet growth tied to nominal GDP (with lower bound of zero); structure/composition unchanged ex ante. - RWAs operational risk increase: 5 percent per year in the first two years of the adverse scenario. - Income tax calibration: sample average of 20 percent. - Dividend payout ratios: baseline sample average about 48 percent; adverse sample average about 37 percent (years 1–2). - Hurdle rates and buffers: CAR 9 percent; CET1 4.5 percent; T1 currently 4.9 percent rising to 6 percent by January 2024; leverage ratio 3 percent; CCoB 1.5 percent by January 2024; DSIB buffer 1 percent by January 2024. ### Solvency stress test — key results and sensitivities - Aggregate resilience: - Aggregate Tier 1 ratio: Before stress 15.1 percent; Adverse scenario: 12.8 percent in year 1, 13.0 percent in year 2, almost fully recovering in year 3. - Total CAR: Before stress 20.2 percent; Adverse scenario: 17.6 percent in year 1, 17.4 percent in year 2, returning to 19.1 percent after year 3 recovery. - Distributional effects: - DSIBs suffer about 180 basis points drop in Tier 1 ratio; non-DSIBs about 360 basis points. - Two small banks breach minimum capital requirements under the adverse scenario, producing a maximum capital shortfall of 0.01 percent of GDP. - One bank would see leverage ratio fall below the regulatory minimum. - Drivers of capital change by year 2: - Credit impairments: cumulative impact of -4.1 percentage points. - Interest rate risk (NII effect): cumulative impact of -2.1 percentage points. - Market valuation losses on all portfolios: -0.7 percentage points. - Decline in profit shares from subsidiaries: -0.5 percentage point (concentrated effect). - Concentration sensitivity: - Default of the 10 largest nonfinancial corporate exposures (with financial guarantees): 5 banks undercapitalized totaling about 0.2 percent of GDP. - Without financial guarantees: undercapitalization about 0.7 percent of GDP involving nine banks. - Counterparty overlap: above 40 percent of the 10 largest exposures involve the same counterparty across multiple banks (up to 7 banks). ### Liquidity stress tests — conditions, methodology, and results - Coverage and metrics: - Top-down LCR-based and cashflow-based tests covering 95 percent of the banking system (12 banks; Col$708 trillion in assets). - National implementation IRL (LCR equivalent) and NSFR assessed on individual basis for all currencies combined. - Current liquidity positions (top 12 banks, end-June 2021): - Aggregate IRL: 188 percent (pre-COVID-19 level returned). - NSFR: 112 percent. - Liquid assets compose 16 percent of total assets; Level 1 securities are 96 percent of liquid assets. - Liquid assets denominated in foreign currency: about 13 percent (one DSIB exception: 41 percent). - Funding structure: unsecured wholesale funding constitutes 75 percent of total funding; demand deposits 41 percent, term deposits 17 percent, bonds 9 percent, bank credits and other financial obligations 8 percent; retail demand deposits 16 percent; retail term deposits 7 percent. - LCR/IRL stress scenarios and outcomes: - Retail adverse scenario: Aggregate IRL falls to 143 percent; all banks remain >100 percent. - Wholesale adverse scenario: Aggregate IRL falls to 123 percent; two banks fall below 100 percent. - Combined severe scenario: Aggregate IRL falls to 102 percent; five banks fall below 100 percent. - Aggregate liquidity shortfall in extreme scenario: 1.7 percent of total assets of sample; individual bank shortfalls do not exceed 3 percent of total assets. - Cashflow-based (up to 90 days) results: - Number of banks with shortfalls: Up to 30 Days = 2 banks; Up to 90 Days = 6 banks. - Liquidity shortfall (percent of sample assets): Up to 30 Days = 0.2; Up to 90 Days = 1.6. - Mitigants: BR’s ability to provide liquidity; partial drawdown of foreign credit lines and reduction in loan extension can materially reduce shortfalls (example: drawdown scenario reduces cumulative shortfall to 0.6 percent). - NSFR distribution (selected percentiles, 2019M12–2021M6): - Median NSFR: 105 | 107 | 112 | 111 | 109 (2019M12 | 2020M6 | 2020M12 | 2021M3 | 2021M6). - Counts of banks with NSFR < 100 percent: 5 | 2 | 0 | 1 | 0 (same date sequence). ### Interconnectedness and contagion analysis — structure, experiments, and findings - Network and model: - CoMap-based framework extended to include liquidity dynamics, nonbank financials, nonfinancials, and selected foreign counterparts; dataset calibrated to December 2020. - Active nodes: all 25 banks, top five pension funds, top five insurers, top five investment funds; passive nodes: aggregated residuals and nonfinancial sectors. - Contagion experiments — key results: - Bank failures can induce up to three contagion failures in other banks and up to six failures in nonbank entities in modeled simulations. - Government default scenario (LGD assumed 20 percent) produces 16 entity failures in total: First round 11 entities (8 banks) fail; Second round 5 entities (4 banks) fail. - Household shock (default of riskiest quartile of loans) produces 8 entity failures in total: First round 4 entities (4 banks) fail; Second round 2 entities (1 systemic bank) fail; Third round 2 entities (1 bank) fail. - Microcredit shock induces failure of three banks with large microcredit exposures. - Cross-border stress scenarios modeled did not induce failure of any modeled entity. - Heterogeneity and entities of concern: - Vulnerability varies: some banks (B12, B14, B23, B6) more vulnerable to government shocks; others (B2, B20) to household shocks. - Entities meriting close monitoring for combined high contagion and vulnerability: Banks B11, B12, B20; Insurer INS2. - Model and data caveats: - CoMap does not model market-perception-driven runs or partial default thresholds; single-default threshold can overstate losses. - Several transmission channels and foreign funding details are incompletely observed; trusts, securities firms, and certain foreign creditor identification need better data. ### Corporate sector vulnerability analysis — data, pandemic impact, and stress test - Data and coverage: - Firm-level data from Supersociedades (SSC), 2016–2020; panel covers 21 industries and about 80 percent of GDP. - Sample in 2020: Corporates 3,346; SMEs 22,720. - Sample captures 38 percent of gross loan portfolio of banks (25 percent corporates; 13 percent SMEs). - 2020 pandemic effects (summary): - Repayment capacity and profitability worsened; firms accumulated more cash; leverage slightly declined. - Sectoral deterioration notable in services; oil sector suffered pandemic plus sharp oil price decline. - Adverse scenario firm-level stress results (second year of adverse unless noted): - Share of firms with ICR<1: - Corporates: increases by 3 percentage points (from 39 percent to 42 percent) by year 2. - SMEs: increases by 3 percentage points (from 25 percent to 28 percent) by year 2. - Note: 2020 observed increase was about 5 percentage points (pandemic year with support policies). - Share of firms with Cash Available < 0 (potential borrowing needs): - Corporates: increases by about 9 percentage points to 45 percent by year 2. - SMEs: increases by about 9 percentage points to 42 percent by year 2. - Modeling approach: - Dynamic regressions for ROA and leverage; Probit models for ICR and cash-availability; structural characteristics kept at 2019 values; macro drivers: GDP growth and a financial conditions index. - Out-of-sample checks: - Actual vs projected 2020 averages: Corporates ROA Actual 2.1, Projected 2.2; Corporates ICR<1 Actual 0.39, Projected 0.37; SMEs ROA Actual 2.7, Projected 3.1. ### Key quantitative facts (select system-level figures) - Credit institutions’ assets: 768,724 billion Col$ (76.2 percent of GDP; 34.9 percent of total system); number of entities: 46. - Commercial banks: 701,990 billion Col$ (69.6 percent of GDP; 31.9 percent of total system); number of entities: 24. - State-owned bank: 27,851 billion Col$ (2.8 percent of GDP; 1.3 percent of total system); number of entities: 1. - Pension funds total: 383,761 billion Col$ (38.1 percent of GDP; 17.4 percent of total system); pension funds: 325,138 billion Col$ (32.2 percent of GDP; 14.8 percent of total system); other retirement funds: 47,927 billion Col$ (4.8 percent of GDP; 2.2 percent of total system); prima media: 10,696 billion Col$ (1.1 percent of GDP; 0.5 percent of total system). - Mutual funds: 135,027 billion Col$ (13.4 percent of GDP; 6.1 percent of total system); number of entities: 279. - Insurance: 94,820 billion Col$ (9.4 percent of GDP; 4.3 percent of total system); number of entities: 45; life: 58,358 billion Col$ (5.8 percent of GDP; 2.7 percent of total system); general: 33,137 billion Col$ (3.3 percent of GDP; 1.5 percent of total system). - Trust services: 660,546 billion Col$ (65.5 percent of GDP; 30.0 percent of total system); number of entities: 23,680. - Total financial system assets: 2,200,188 billion Col$ (218 percent of GDP; 100 percent of total system). - Banking sector credit exposures: 69 percent of total banking assets; composition: commercial 54 percent, consumer 29 percent, mortgage 15 percent. - Peak NPL increases in adverse scenario: commercial from about 3.7 percent to 4.8 percent; consumer from 3 percent to 5 percent; mortgage from 3.4 percent to a little over 10 percent. - Credit losses over first two years in adverse scenario: 3 percent of total banking system assets. - Baseline three-year provision reimbursement: about 0.7 percent of total banking system assets. ### Key recommendations (select, as presented) - Strengthen monitoring of cross-border exposures and conduct fully consolidated stress-testing by filling data gaps on exposures and risk metrics of ultimate subsidiaries. — Agency: SFC — Time: ST - Extend data collection to monitor liquidity risks by currency; collect more granular data on assets and liabilities generating cashflows, including those related to cross-border exposures. — Agency: SFC — Time: ST - Readjust or determine some parameters used in computation of local LCR and NSFR ratios to further align with Basel III requirements. — Agency: SFC — Time: MT - Develop network analysis tools and improve data coverage to bolster Early Warning Systems (EWS) for domestic and cross-border contagion. — Agency: SFC/BR — Time: ST - Broaden the sample of firms for the early warning exercise (“Alerta Temprana”) for the nonfinancial corporate sector and enhance cross-agency information sharing to reinforce feedback for the early warning exercise. — Agency: SSC/SFC — Time: ST *Italic: Source: IMF staff compilation from the Colombia FSAP chapter and EXECUTIVE SUMMARY (content unit 1colea2022007).* ### EXECUTIVE SUMMARY __________________________________________________________________________ 6 ### EXECUTIVE SUMMARY ### Background and macrofinancial context - FSAP risk analysis conducted after the initial COVID shock and lockdowns, during a strong economic recovery underway in 2021; high uncertainty around the evolution of the COVID-19 virus and the recovery led to significant downside risk and revisions. - Workstreams used the latest macroeconomic and supervisory data available, but test results should be interpreted with caution given high uncertainty around central projections and downside risks. - The banking system entered the COVID-19 pandemic from a position of relative strength; authorities mounted a strong policy response to avoid a tightening of credit supply, including fiscal and monetary measures, SFC allowing the release of countercyclical provisions, temporary borrower grace periods and loan modifications, and the Program to Support Debtors (PAD) (phased out at end-August 2021). - Risk indicators had broadly recovered from the 2014–16 oil price shock; banks had strong profitability buffers and adequate capital, further boosted by migration to Basel III. - Funding structure: “Banks’ funding is largely deposit-based, with almost 80 percent of funding coming from the wholesale sources.” - Financial system dominated by large and complex financial conglomerates (FC); five largest conglomerates own about 60 percent of system assets; increasing cross-border exposures and growing systemic footprint in several Central American countries suggest monitoring intragroup risks and contagion channels is essential. ### Solvency stress test: design and results - Coverage: solvency and liquidity stress tests conducted on four domestic systemically important banks (DSIBs) and eight non-systemic banks covering the vast majority of banking system assets. - Stress scenario: severe—combines global resurgence of the pandemic, tightening of financial conditions, political uncertainty, and fiscal deterioration. - Aggregate result: system-wide high starting capital and strong profit buffers allow most banks to absorb a large shock under the adverse scenario, leaving very small capital shortfall at the system level. - Distributional effects: - Non-DSIBs and foreign-owned banks are impacted somewhat more than DSIBs and domestically owned banks, respectively. - Credit impairments and lower net interest income are the key contributing factors to stress losses. - Aggregate conclusion: broad resilience at the aggregate level, albeit with pockets of vulnerabilities in a few banks. ### Liquidity stress tests: conditions and results - Post-pandemic liquidity boost from central bank support, flight-to-stability reallocation, and slower loan growth. - LCR-based and cash-flow analyses show banks are well positioned to manage significant liquidity pressures; only modest liquidity shortfalls in the extreme funding shock. - Caveat: some banks would lack liquid assets to support credit intermediation if faced with long-lasting, very severe liquidity shocks. - Policy-relevant vulnerability: reliance on wholesale unsecured funding warrants attention. ### Interconnectedness and contagion analysis - Interbank contagion appears limited; DSIBs are sources of heightened contagion risk to the rest of the system, but cascade effects remain contained. - Low vulnerability levels among DSIBs limit potential amplification from idiosyncratic bank shocks. - Important vulnerabilities identified: - Intersectoral linkages and cross-border exposures can lead to substantial losses. - Contagion to nonbank financial entities and cross-border exposures is meaningful. - Some vulnerability of insurers due to exposures to banks. - Recommendation emphasis: strengthen monitoring of cross-border exposures and conglomerate structures to better identify contagion channels. ### Corporate sector vulnerability analysis - Firm-level analysis using corporate and SME financial statements indicates pandemic led to deterioration in repayment capacity and profitability in 2020, but policy actions prevented widespread systemic failures. - Under the adverse scenario: - Increase in the share of firms with lower repayment capacity. - Increase in the share of firms with potential borrowing needs. - Increase in the share of firms under solvency pressure, though to a lesser extent than other indicators. - Monitoring priority: continue to monitor the nonfinancial corporate sector, especially the services sector. - Early warnings: ongoing work could be broadened to increase firm sample size for better monitoring. ### Key quantitative facts (select) - Credit institutions’ assets: 768,724 billion Col$ (76.2 percent of GDP; 34.9 percent of total system); number of entities: 46. - Commercial banks: 701,990 billion Col$ (69.6 percent of GDP; 31.9 percent of total system); number of entities: 24. - State-owned bank: 27,851 billion Col$ (2.8 percent of GDP; 1.3 percent of total system); number of entities: 1. - Pension funds total: 383,761 billion Col$ (38.1 percent of GDP; 17.4 percent of total system); number of entities: 5. - Pension funds: 325,138 billion Col$ (32.2 percent of GDP; 14.8 percent of total system). - Other retirement funds: 47,927 billion Col$ (4.8 percent of GDP; 2.2 percent of total system). - Prima media: 10,696 billion Col$ (1.1 percent of GDP; 0.5 percent of total system). - Mutual funds: 135,027 billion Col$ (13.4 percent of GDP; 6.1 percent of total system); number of entities: 279. - Insurance: 94,820 billion Col$ (9.4 percent of GDP; 4.3 percent of total system); number of entities: 45. - Life: 58,358 billion Col$ (5.8 percent of GDP; 2.7 percent of total system). - General: 33,137 billion Col$ (3.3 percent of GDP; 1.5 percent of total system). - Trust services: 660,546 billion Col$ (65.5 percent of GDP; 30.0 percent of total system); number of entities: 23,680. - Total financial system assets: 2,200,188 billion Col$ (218 percent of GDP; 100 percent of total system). ### Key recommendations (Table 1: Colombia: Risk Analysis Recommendations) - 1. Strengthen the ability to monitor cross-border exposures and conduct a fully consolidated stress-testing by filling data gaps on the exposures and risk metrics of ultimate subsidiaries. — Agency: SFC — Time: ST - 2. Extend data collection to monitor liquidity risks by currency. Collect more granular data on assets and liabilities generating cashflows, including those related to cross-border exposures. — Agency: SFC — Time: ST - 3. Readjust or determine some parameters used in the computation of the local LCR and NSFR ratios to further align with Basel III requirements. — Agency: SFC — Time: MT - 4. Develop network analysis tools and improve data coverage to bolster EWS for domestic and cross-border contagion. — Agency: SFC/BR — Time: ST - 5. Broaden the sample of firms for the early warning exercise (“Alerta Temprana”) for the nonfinancial corporate sector and enhance cross-agency information sharing to reinforce the feedback channel to strengthen the early warning exercise. — Agency: SSC/SFC — Time: ST *Source: EXECUTIVE SUMMARY, FSAP risk analysis (Colombia).* ### 12. Colombia’s financial sector has been operating in an environment of strong ### 12. Colombia’s financial sector has been operating in an environment of strong macroeconomic policies and policy framework, but in an economy with still relatively high commodity export dependence and high exposure to capital flow shocks. ### Macroeconomic and external exposure - Commodity exports amount to about half of total exports. - High nonresident participation in the local bond market at about 8 percent of GDP and close to 23.7 percent of the total holdings before the pandemic. - External vulnerabilities are limited by adequate foreign reserves buffers and access to a flexible credit line (FCL) from the Fund. ### Pandemic impact and policy response - GDP fell about 7 percent in 2020 in response to lockdowns and spillovers from lower oil prices and the collapse of global growth. - Economic growth rebounded by about 10.6 percent in 2021. - The pandemic led to a substantial widening of the budget deficit and suspension of the fiscal rule. - Labor market effects: job losses, record-high unemployment, and a drop in labor force participation. - As restrictions eased in Q3 2020, recovery led by manufacturing, retail trade, and public services; continued service sector weakness. Policy and supervisory measures - The SFC allowed the release of countercyclical provisions built up in the past. - Banks were allowed to provide temporary grace periods, extensions, and other loan modifications on a case-by-case basis without affecting the debtor’s credit rating or triggering loan reclassification as nonperforming; provisioning rules were not relaxed. - The SFC launched PAD to support viable borrowers; PAD was phased out at end-August 2021 and triaged borrowers by degree of impact with higher provisioning for severely affected debtors. - Uptake of relief: above 40 percent of all gross loans benefited from payment deferrals at peak in the first phase; uptake dropped to about 10 percent in the second, more targeted phase. - Overall NPLs picked up, reaching 5.6 percent by end-2020; banks increased provisions. - By end of PAD, loans under PAD represented 6.8 percent of total gross loans (Col$37 billion associated with 2.2 million debtors). - Data as of November 2021: nearly 0.4 percent of gross loans still covered by first-wave measures (EC 007 and 014 of 2020); PAD accounts for 5.3 percent of total gross loans. ### Credit trends, household and corporate balance sheets - Pre-pandemic credit recovered after the 2014–16 oil price shock. - High competition in banking reduced costs and boosted consumer credit growth. - Subsidized housing loans increased from 24.1 percent of total mortgages at end-2019 to 26.2 percent at end-2020. - Consumption (especially unsecured) and microcredit loan growth were hit in H1 2020 but began recovering in Q3 2020. - During 2021, high-value mortgage was the most dynamic segment; total mortgage loans surpassed pre-pandemic growth rates. - Commercial credit ended 2021 with 3.6 percent in real growth. - FX lending by banks is low. Corporate FX debt rise relative to GDP is mostly due to peso depreciation following the 2014–16 oil shock. - Household indebtedness about 25 percent of GDP at end-2020. - Leverage of large corporates increased significantly after 2013, largely due to depreciation. FX-denominated debt is about 35 percent of private sector corporate debt. - Data caveat: shortcomings in household indebtedness data limit precise calculation of total household debt-to-income and total debt-service-to-income ratios. ### Financial system structure and concentration - Financial conglomerates are central and interconnected; the five largest conglomerates control 80 percent of banking sector assets. - Banks are the main financing source for households and nonfinancial corporations and are significant assets on balance sheets of investment funds, insurers, securities firms, and trusts. - Limited development of domestic capital markets: market capitalization about 40 percent of GDP, restricting domestic investment alternatives beyond government and bank securities for institutional investors. - Conglomerates’ foreign presence and exposures: - Colombian conglomerates operate in 14 countries in banking, insurance, and asset management. - Number of entities abroad rose from 46 to 288 from 2009 to 2020. - Corresponding exposures rose from US$11 billion to US$93 billion during that period. - Exposures have more than doubled since the last FSAP. - About 83 percent of these assets are held in Central American countries. - Specific entities can be highly exposed to Panama, Costa Rica, El Salvador, and Guatemala. - Investments in foreign jurisdictions comprise over one-fifth of total banking sector assets. - Wholesale funding accounts for almost 80 percent of funding sources; unsecured portion is the vast majority: - Demand deposits 41 percent - Term deposits 17 percent - Bonds 9 percent - Bank credits and other financial obligations 8 percent - Retail funding: retail demand deposits 16 percent; retail term deposits 7 percent of total funding. - Liquidity indicators: NSFR and IRL increased during 2020 due to higher banks’ holdings of liquid assets; IRL stabilized around pre-pandemic level by mid-2021. ### Nonbank financial institutions - Collective investment funds (CIFs) faced investor withdrawals of Col$24.6 trillion during March 2020, causing a 32 percent decline in total assets. - Extraordinary liquidity facilities from the central bank alleviated fire-sale pressure; CIFs’ assets under management returned to pre-pandemic levels by end-2020. - This led to an increase in retail deposits at banks. ### Systemic soundness and vulnerabilities - Banking sector soundness indicators improved toward pre-pandemic levels as of June 2021. - Banks entered 2021 with markedly higher capital ratios compared to pre-pandemic levels, aided by convergence with Basel III and strengthening via subordinated debt and retained earnings. - As of August 2021, all but two smaller banks have total capital ratios higher than pre-COVID-19. - Bank profitability bounced back; banks entered the pandemic with relatively strong buffers. - Main vulnerability: credit risk. NPLs initially declined (partly due to debtor support and relief programs), peaked later in 2020, and began declining in 2021. - Risks if growth slows or pandemic prolongs: larger-than-expected risk revelation, especially in consumer portfolios and the most affected sectors. - Net interest margins have remained healthy but could be compressed if global financial tightening occurs. - Decline in NPLs also reflects increasing write-offs (16.2 percent year-on-year increase in June 2021) and loan collection. ### Cross-sectoral interconnectedness and contagion - Banks are central nodes in cross-sectoral interlinkages; exposure mapping shows significant interconnectedness among credit institutions, pension funds, insurers, investment funds, and nonfinancial corporates. - Concentrated cross-border exposures and over one-fifth of banking assets held abroad present cross-border contagion risks, warranting intensified monitoring. ### FSAP stress test: scope, methodology, and coverage - Forward-looking risk assessment components: - (i) solvency; - (ii) liquidity risk analysis for the banking system; - (iii) domestic and cross-border interconnectedness and contagion analysis; and - (iv) corporate stress test. - Solvency stress test uses IMF internal tools (core balance sheet and capital engine, satellite models) to project P&L and balance sheet items under macro scenarios, focusing on credit risk, market risk, and interest rate risk; sensitivity tests assess concentration risk. - Sample: 12 banks covering 93.8 percent of banking sector assets, including all four D-SIBs, five non-systemic domestically owned banks, and three non-systemic foreign-owned banks. One D-SIB is foreign-owned. - Exercise conducted on a solo-reporting basis; cross-border exposures included via shock factor on gains/losses from associated entities and via interconnectedness/contagion analysis. - Stress test reference frameworks: national implementation of Basel III came into force on January 1, 2021. - Alignments: definitions of CET1, Tier 1, and total capital; credit risk weights aligned with latest standardized approach; leverage ratio with a minimum of 3 percent incorporated. - Phasing: transitional arrangements remain for internal loss indicator for operational risk, phasing in capital conservation and systemically important bank buffers, and phasing in the Tier 1 ratio. - No applicable countercyclical capital buffer in Colombia; banks are required to build up countercyclical provisions instead. - All banks use the same risk weights according to the standardized model by the Financial Regulation Agency (URF). - Caveat on stress-test assumptions: standard FSAP assumptions may produce more pessimistic results than actual outcomes because exercises are designed to assess resilience in tail events and assume "neutral behaviors" (e.g., not assuming aggressive bank deleveraging or other management actions). *Source: IMF staff compilation from the Colombia FSAP chapter.* ### 27. The solvency risk analysis is based on the assessment of current conditions, ### 1colea2022007 - 27. The solvency risk analysis is based on the assessment of current conditions, ### Risk assessment and scenarios - Current conditions are characterized by continuing pandemic-related risks and other external and domestic risks exacerbated by the impact of the pandemic. - Risks to the ongoing recovery include: - uncontrolled COVID-19 local outbreaks and global resurgence of the pandemic; - de-anchoring of U.S. inflation expectations and rise in global risk premia; - disorderly structural transformations triggered by COVID-19; - increasing fiscal pressures. - These risks can lead to: a downturn in economic activity, higher risk premia and financing difficulties for leveraged firms and banks, capital outflows, depreciation and inflation pressures, higher cost for sovereign financing, and a sharp fall in asset prices. - Stress testing exercise structure: - Scenarios span a three-year horizon from mid-2021 to mid-2024. - Baseline scenario: aligned with the October 2021 World Economic Outlook (WEO) projections; reflects a strong 2021 rebound followed by a more gradual recovery, declining unemployment through 2022, and inflation remaining within target as the policy rate is steadily raised. - Adverse scenario: protracted recession forming a U-shaped GDP level profile; incorporates global and Colombia-specific macrofinancial risks from the Risk Assessment Matrix. - Adverse scenario GDP path (three-year horizon): - Averaging 0.5 percent real GDP growth over three years. - Year 1: -1.2 percent - Year 2: -1.3 percent - Year 3: 4 percent - Cumulative GDP decline under adverse over three (two) years is equivalent to a 1.8 (2.5) standard deviation shock. - Additional adverse scenario shocks (deviations from baseline): - Drop in housing prices by 21 percent - Depreciation in the exchange rate by about 25 percent - Rise in long-term sovereign yields by 190 basis points - Interbank spread widening by 125 basis points ### Transmission channels and sovereign risks - Main transmission channels of external shocks: external demand, capital flows, and sovereign distress. - Global pandemic resurgence can disrupt activity and depress growth prospects globally; weak external demand (including for oil) adversely affects domestic confidence and public finances. - Signals of U.S. Fed policy tightening can tighten financial conditions and raise risk premia for credit and equities, and pressure the Colombian peso. - Increased reliance on external financing raises vulnerability to financial tightening, potentially triggering capital outflows, depreciations, and sovereign distress. - Feedback loops and spillovers: - Decline in housing prices hits consumer confidence and domestic demand via wealth effects. - Stress in funding markets raises corporate borrowing costs and reduces incomes, hurting banks’ asset quality and profitability. - Financial cycle downturn increases mortgage rates and depresses incomes, adding pressure on housing prices and creating a negative feedback loop with domestic demand and deflationary pressures. - Colombia-specific adverse amplifiers: - Political uncertainty and weakening debt profile can lead to further downgrades and foreign investor sell-offs. - Stress in public finances affects financial entities due to strong home bias in sovereign exposures. - Expiration of regulatory measures and prolonged adverse conditions can result in a larger-than-expected share of NPLs and a wave of bankruptcies. - Regional spillovers exacerbate conditions because Colombian financial conglomerates have significant exposures to several Central American countries facing similar risks. ### Stress test methodology — balance sheet, income projections, and hurdle rates - Balance sheet projection approach: - "Quasi-static" approach: balance sheet grows proportionally with nominal GDP under baseline and adverse scenarios, with a lower bound of zero. - In the first two years of the adverse scenario with nominal GDP contraction, balance sheet size remains constant. - Structure and composition of assets and liabilities are assumed unchanged ex ante; ex-post adjustment made to overnight retail deposits to cover funding gaps. - RWAs projection adjustments: - Account for balance sheet growth, new provisions for credit losses, exchange rate movements for foreign currency exposures, and a triggered portion of undisbursed credit lines (off-balance sheet). - RWAs for operational risks increase by 5 percent per year in the first two years of the adverse scenario. - Income projections: - Noninterest income items projected in line with balance sheet growth (net fee and commission income, other noninterest income, noninterest expenses). - Extraordinary income and losses assumed not to occur during the projection period. - Income tax reflected using median of historical tax expense ratios per bank (sample average of 20 percent). - Dividend policy assumptions: - After-tax income is distributed only if net income is positive. - Baseline: historical averages of dividend payout ratios per bank (excluding years with greater than 100 percent), sample average about 48 percent. - Adverse: historical non-zero minimum ratios during the first two years (sample average about 37 percent), reverting to baseline in year 3. - Banks’ indirect exposures through subsidiaries: - Solo-bank coverage with shocks to gains/losses from associated entities. - P&L item from subsidiaries shocked by 50 percent in year 1 and 25 percent in year 2 of the adverse scenario. - Stress test hurdle rates (current capital adequacy framework and phase-in): - Total capital adequacy ratio (CAR) hurdle: 9 percent (includes an additional 1 percent above Basel III requirement). - Minimum CET1 required: 4.5 percent (in effect since 2013). - Tier 1 (T1) hurdle: currently 4.9 percent, gradually increased to 6 percent by January 2024 per phase-in schedule. - Minimum leverage ratio (Tier 1 capital to total assets): 3 percent throughout. - Capital conservation buffer (CCoB) and DSIB buffer phased in to reach 1.5 percent and 1 percent of RWAs by January 2024. - Gradual phase-out of eligible capital instruments from additional Tier 2 incorporated identically under both scenarios. ### Credit risk - Credit exposures: - Credit exposures account for 69 percent of total banking assets. - Composition of credit exposures: commercial 54 percent, consumer 29 percent, mortgage 15 percent (collectively 98 percent of bank lending). - NPL modeling: - Satellite models estimate NPL ratios for each portfolio segment using Bayesian Model Averaging (BMA) methodology subject to sign constraints on long-run multipliers; panel fixed effect regressions on pooled equations. - NPL projections conditional on macrofinancial variables under baseline and adverse scenarios. - Provisioning ratios: - Historical provisioning coverage used for each portfolio segment. - In baseline, aggregate ratio at cut-off date is used. - In adverse, the ninetieth and ninety-fifth percentiles of aggregate provisioning coverage ratios are applied for each portfolio type. - New provisions calculated by applying the year 1 (T1) provisioning coverage ratio in year 1 (T1) to starting point data at cut-off date (T0). - Procyclical provisions not utilized (conservative approach). - NPL and provisioning projections (adverse scenario peaks): - Commercial NPL ratio: from about 3.7 percent to 4.8 percent at the peak. - Consumer NPL ratio: from 3 percent to 5 percent at the peak. - Mortgage NPL ratio: from 3.4 percent to a little over 10 percent at the peak. - Credit loss magnitudes: - Credit losses over the first two years amount to 3 percent of total banking system assets in the adverse scenario (similar ratios for DSIBs and non-DSIBs), recovering in year 3. - In the baseline scenario, banks register overall provision reimbursements over three years of about 0.7 percent of total banking system assets. - Treatment of HTM debt securities: - Treated under the credit risk module using a haircut approach linked to shocks on portfolio durations and credit spreads. - Component is a very small fraction of credit loss provisions. ### Market risk and net interest income (interest rate risk) - Market risk assessment: - Banks face losses due to changes in risk-free interest rates, exchange rates, and equity prices, in addition to credit-spread-related losses. - No allowance made for macro hedges. - Marked-to-market valuation methodologies: - Sovereign and corporate debt classified as available for sale (AFS) or held for trading (HFT) revalued using a modified-duration formula based on residual duration and bond yield assumptions under scenarios. - Foreign exchange risk and equity risk estimated from net open positions and scenario paths for equity index and exchange rate. - Net interest income (NII) assessment: - Time-to-repricing buckets used to measure gaps between interest-sensitive assets and liabilities. - Outstanding volumes and effective repricing schedules established from regulatory reports at cut-off date. - Asset side breakdown: portfolio (commercial, consumer, mortgage, microcredit) and rate type (fixed, money-market-linked, deposit-linked). - Liability side breakdown: overnight and term, retail and wholesale deposit, repos, unsecured, government, other. - Effective interest rates estimated using satellite models and pass-through assumptions. - Deposit groups modeled: less than 3 months, between 3 and 12 months, longer than 1 year. - Paths of deposit rates per bank projected under baseline and adverse scenarios. - For liability segments not modeled, shocks on corporate credit spreads used. - Shifts in effective loan rates calculated using pass-through rate assumptions (details in Appendix IV). ### Key quantitative outcomes and calibrations referenced - Scenario horizon: mid-2021 to mid-2024. - Adverse scenario GDP path: -1.2 percent, -1.3 percent, and 4 percent (years 1–3 respectively). - Adverse scenario average real GDP growth over three years: 0.5 percent. - Cumulative GDP decline equivalence: 1.8 (three years) and 2.5 (two years) standard deviation shocks. - Housing prices decline: 21 percent (adverse scenario calibration). - Exchange rate depreciation: about 25 percent (adverse scenario calibration). - Long-term sovereign yields rise: 190 basis points (adverse scenario calibration). - Interbank spread widening: 125 basis points (adverse scenario calibration). - RWAs operational risk increase: 5 percent per year in first two years of adverse scenario. - Income tax calibration: sample average of 20 percent. - Dividend payout ratios: baseline sample average about 48 percent; adverse sample average about 37 percent (years 1–2). - Subsidiaries P&L shocks: 50 percent in year 1 and 25 percent in year 2 (adverse scenario). - Stress test hurdle rates and buffers: CAR 9 percent; CET1 4.5 percent; T1 currently 4.9 percent rising to 6 percent by January 2024; leverage ratio 3 percent; CCoB 1.5 percent by January 2024; DSIB buffer 1 percent by January 2024. - Credit exposures composition: commercial 54 percent, consumer 29 percent, mortgage 15 percent (of total bank lending). - Credit exposures share of total banking assets: 69 percent. - Peak NPL increases in adverse scenario: commercial 3.7 percent to 4.8 percent; consumer 3 percent to 5 percent; mortgage 3.4 percent to a little over 10 percent. - Credit losses over first two years in adverse: 3 percent of total banking system assets. - Baseline three-year provision reimbursement: about 0.7 percent of total banking system assets. *Source: IMF staff (FSAP chapter).* ### 48. The projection of net interest income (NII) was, accordingly, based on a ### 1colea2022007 - 48. The projection of net interest income (NII) was, accordingly, based on a ### Net Interest Income (NII) Projection Methodology - Projection based on a semi-structural approach. - New business repricing assumption: maturing instruments replaced by identical new instruments (of the same segment and with the same initial maturity) but at reference and margin rates implied by the scenario and the pass-through assumptions. - NII projection equals the difference in interest income and expense for each segment after: - accounting for the repricing impact for the new business each year, and - accounting for changes in the amount of performing interest-bearing assets due to credit risk. ### Solvency Stress Test Results (Key Findings) - Aggregate resilience: Broad resilience at the aggregate level with pockets of vulnerabilities in a few banks. - Starting buffers: High starting levels of system-wide capital and strong profit buffers allow most banks to absorb a large shock under the adverse scenario. - Differential impacts: - non-DSIB banks and foreign-owned banks impacted somewhat more than DSIBs, and domestically owned banks, respectively. - Aggregate capital ratios under scenarios (using full Basel III regulatory requirements): - Aggregate Tier 1 ratio: - Before stress: 15.1 percent - Adverse scenario: drops to 12.8 percent in year 1, minor improvement in year 2 to 13.0 percent, almost fully recovering in year 3. - Total CAR: - Before stress: 20.2 percent - Adverse scenario: declines to 17.6 percent in year 1 and to 17.4 percent in year 2, before returning to 19.1 percent after year 3 recovery. - DSIBs vs non-DSIBs: - DSIBs experience about 180 basis points drop in the Tier 1 ratio versus 360 basis points for non-DSIBs. - DSIBs keep comfortable buffers given a much higher Tier 1 ratio at the outset; DSIB capital adequacy would recover by year 3, non-DSIBs do not fully recover within the projection period. - Bank-level breaches and leverage: - For two small banks, the Tier 1 ratio and CAR ratio would be lower than the minimum requirements, resulting in a maximum capital shortfall of 0.01 percent of GDP under the adverse scenario. - For one of these banks the leverage ratio would fall below the regulatory minimum. - Contributions to change in capital (drivers by year 2): - Credit impairments: cumulative impact of -4.1 percentage points by year 2. - Interest rate risk: cumulative impact of -2.1 percentage points by year 2. - Market valuation losses on all portfolios: further -0.7 percentage points fall by year 2. - Sharp decline in profit shares from subsidiaries: 0.5 percentage point drop, mostly impacting one DSIB and two small banks. ### Sensitivity Analysis: Concentration Risk - Default of the 10 largest nonfinancial corporate exposures (after application of financial guarantees): - Would lead 5 banks to be undercapitalized by about 0.2 percent of GDP collectively with regard to the Tier 1 or CAR minimum requirements. - Without financial guarantees (very conservative assumption): - Overall impact: undercapitalization by about 0.7 percent of GDP, involving nine banks. - Counterparty overlap: - Above 40 percent of the 10 largest exposures among banks involve the same counterparty across multiple banks (up to 7 banks). ### Liquidity Stress Tests — Overview and Coverage - Tests conducted: top-down LCR-based and cashflow-based tests covering 95 percent of the banking system. - Regulatory liquidity metrics imposed by SFC: the Liquidity Risk Index (IRL, national implementation of Basel III LCR) and the NSFR, implemented on an individual basis and for all currencies combined. - IRL requirements: liquid assets must be above net liquidity requirements for the 7- and 30-day horizons. - NSFR: estimates the ratio of available stable funding relative to required stable funding. - Data and horizon: - LCR-based test aligned with IRL over a 30-day period. - Cashflow-based test over a 3-month period on an individual basis for all currencies combined, using data as of end-June 2021. - Sample for liquidity tests: - 12 banks with total assets of Col$708 trillion, covering about 95 percent of total assets of banks in Colombia. - Sample composition: four DSIBs and eight non-DSIBs, one state-owned bank; one DSIB and three non-DSIBs foreign headquartered. ### Current Liquidity Conditions and Banks’ Liquidity Profiles - Aggregate IRL (top 12 banks): - Pre-COVID-19 level: well above the regulatory minimum of 100 percent. - Returned close to pre-pandemic level of 188 percent at end-June 2021. - NSFR: - Increased since start of pandemic and reached 112 percent. - Liquidity composition: - Level 1 securities (HQLA) constitute 96 percent of liquid assets. - Composition of Level 1 securities: - required reserves: 26 percent of liquid assets - other cash items: 16 percent - securities admitted by the BR, mainly Colombian sovereign debt: 54 percent of liquid assets - Liquid assets denominated in foreign currency: about 13 percent (exception: one DSIB has 41 percent of liquid assets in foreign currency). - Liquid assets comprise 16 percent of total assets. - Funding structure: - Almost 80 percent of commercial banks’ funding comes from wholesale sources (nonfinancial corporates and financial institutions). - As of end-March 2021, unsecured wholesale funding of the 12 largest banks constitutes 75 percent of total funding and consists of: - demand deposits: 41 percent of total funding - term deposits: 17 percent of total funding - bonds: 9 percent of total funding - bank credit and other financial obligations: 8 percent of total funding - remaining wholesale funding (3 percent) from passive money market operations with the BR and financial entities - Retail funding: - retail demand deposits and term deposits comprise about 16 percent and 7 percent of total funding, respectively. - overall, demand and term deposits constitute 57 percent and 24 percent respectively. - Main holders of deposits: nonfinancial companies, followed by natural persons, and non-credit-institutions financial entities. - Deposit concentration risks: - For top 12 banks, on average: - top 1 depositor holds about 4 percent of total liabilities - top 20 depositors hold about 15 percent of total liabilities - Top depositors mainly hold current and savings deposits that might be withdrawn immediately; most are not related counterparties. - Pandemic-related developments: - Demand deposits in the top 12 banks increased by 31 percent between end-2019 and June 2021. - DSIBs: 33 percent increase - non-DSIBs: 28 percent increase - Increase in liquid assets was main contributor to rise in IRL, partially offset by predicted cash outflows attributed to higher balances of demand deposits. - ASF and RSF developments affecting NSFR: - Available stable funding (ASF) composition: - total regulatory capital: 20 percent - demand deposits: 50 percent - term deposits: 49 percent - ASF increase: from 422 at end-December 2019 to Col$498 trillion at end-June 2021. - Required Stable Funding (RSF) increased slightly by 5.6 percent due to rise in holdings of assets with high RSF factor, such as non-exchange-traded equities; benefited from slower loan growth. ### Foreign Exchange Liquidity Exposure - Exchange rate risk exposure: banks do not demonstrate significant exposures to exchange rate risks overall. - Foreign currency net liquidity requirements (NLR): - Foreign currency NLR comprise only 11 percent of total NLR and are dominated by US dollars (97 percent). - At aggregate level of the top 12 banks, liquid assets in US dollars comfortably exceed NLR. - Some non-DSIBs have liquid assets in US dollars significantly below NLR (as percent of US dollar NLR) in June 2021, but excess liquidity in COL compensates after an 11 percent regulatory haircut. ### LCR-based Stress Tests: Scenarios and Results - LCR-based stress test aligned with national implementation of Basel III framework; assesses ability to use unencumbered liquid assets over 30 days. - Scenarios: - Baseline scenario: reflects IRL assumptions. Cash outflows include contractual outflows subject to 100 percent run-off and noncontractual outflows related to demand deposits segmented by client type. Cash inflows include contractual inflows subject to 100 percent factors (except inflows related to loan portfolios adjusted by half of latest past-due portfolio index). Liquid assets include reserves subject to haircuts including foreign exchange risk haircuts. - Adverse scenarios: simulate more abrupt withdrawal of funding with more severe run-off rates for demand deposits; include: - retail funding stress scenario: increases run-off rates from about 3 percent to 13 percent on average across banks to 5 percent to 20 percent (IRL methodology differences across banks apply). - wholesale funding stress scenario: simulates abrupt withdrawal of wholesale demand deposits. - severe scenario: combines stressed run-off parameters of retail and wholesale scenarios. - Adverse scenarios chosen based on expert judgement, international benchmarks, and regulatory parameters representing the fifth percentile of historical changes in demand deposit balances for some banks. - Baseline (regulatory) scenario results: - Aggregate IRL: 188 percent; all individual banks pass 100 percent hurdle comfortably. - IRL by bank type: non-DSIBs 241 percent vs DSIBs 163 percent. - Components: - cash outflows account for 15 percent of total assets (above level of liquid assets after haircuts). - contractual inflows: 8 percent of total assets, one-fourth due to maturing loans. - DSIBs hold smaller amount of liquid assets relative to net outflows, implying stronger impact under very severe liquidity shocks. - Aggregate IRL for foreign headquartered banks: 243 percent vs local banks: 170 percent; possible reasons include sample being subject to Basel III LCR and parent company containment of liquidity contingency plans. *Source: 1colea2022007 (IMF staff calculations).* ### 63. The “retail” adverse scenario tests the banks’ resilience to retail demand deposits ### 63. The “retail” adverse scenario tests the banks’ resilience to retail demand deposits ### LCR-/IRL-Based Stress Test Results (retail, wholesale, combined) - Retail adverse scenario: - Aggregate IRL decreases to 143 percent. - All banks maintain IRL above 100 percent. - Wholesale adverse scenario (dry-up of wholesale funding): - Aggregate IRL falls to 123 percent. - Two banks fall below the 100 percent hurdle rate. - Combined retail and wholesale withdrawal scenario: - Aggregate IRL decreases to 102 percent. - Five banks’ IRL fall below 100 percent. - Differential impacts by bank type: - DSIBs: aggregate IRL falls from 163 percent in the regulatory scenario to 88 percent in the severe scenario. - Foreign headquartered banks: aggregate IRL declines from 243 percent in the baseline scenario to 141 percent in the severe scenario. - Local banks: aggregate IRL declines from 170 percent to 91 percent in the severe scenario. - Magnitude of shortfalls: - Aggregate liquidity shortfall in the extreme scenario is 1.7 percent of total assets of banks in the sample. - Liquidity shortfalls for individual banks do not exceed 3 percent of total assets. - Regulatory note on run-off rates: - Regulatory run-off rates are already high for retail deposits relative to those in the adverse retail scenario. - According to IRL methodology, for each bank, run-off rates for demand deposits are computed as the median from: (i) the zero percentile of historical changes in balances of demand deposits for each client category for that bank up to the previous month; (ii) the fifth percentile of the same series; and (iii) regulatory estimates on the historical data up to between 2011 and 2017 (red dots). ### Cashflow-Based Analysis (up to 90 days) - Methodology highlights: - Assesses maturity structure of cash inflows and outflows up to a three-month horizon (buckets: 1–7 days, 8–15 days, 16–30 days, 30–90 days). - Net cash balance for each bucket = existing cash position + counterbalancing capacity + net funding inflows. - Counterbalancing capacity includes cash, unencumbered holdings of debt securities, required reserves, and separately treated potential drawdown of banks’ credit lines with foreign entities. - Cash inflows: 100 percent roll-off rates applied to all maturing assets, except for maturing loans and maturing secured lending. - Assumptions on run-off behavior: - More informed/sophisticated investors withdraw a larger amount of funding than less informed depositors. - Run-off rates on secured funding sources are lower than on unsecured funding sources. - Liquidity pressure is assumed protracted, lasting up to three months, with run-off rates declining after 30 days. - Partial utilization of off-balance sheet credit or liquidity commitments by clients is included. - Results (Table 7 summary): - Number of banks experiencing shortfalls by horizon: - Total: Up to 7 Days = 0; Up to 15 Days = 0; Up to 30 Days = 2; Up to 90 Days = 6. - DSIBs: Up to 7 Days = 0; Up to 15 Days = 0; Up to 30 Days = 2; Up to 90 Days = 3. - Non-SIBs: Up to 7 Days = 0; Up to 15 Days = 0; Up to 30 Days = 0; Up to 90 Days = 3. - Liquidity shortfall (percent of banks’ assets in sample): - Up to 7 Days = 0.0 - Up to 15 Days = 0.0 - Up to 30 Days = 0.2 - Up to 90 Days = 1.6 - Mitigants and sensitivities: - Even in the very extreme case, the combined shortfall of 1.6 percent of assets is modest and should be manageable given the BR’s ability to provide liquidity to the system. - If a drawdown of 15 percent of foreign credit lines by banks experiencing liquidity shortfall is exercised, one non-DSIB covers its liquidity shortfall, and the cumulative liquidity shortfall decreases to 0.6 percent of total assets of banks in the sample. - Liquidity shortfall could also potentially be covered by a reduction in loan extension. ### NSFR-Based Exploratory Analysis - Implementation status: - NSFR was already above 100 percent for most banks at end-June 2021. - Authorities started gradual implementation with the level depending on systemic importance; 100 percent limit for banks with assets greater than 2 percent of total system assets. - Implementation scheduled to be finalized by end-March 2022; by June 2021 NSFR was above 100 percent for all except one bank in the sample. - NSFR distribution by date (in percent): - 2019M12 | 2020M6 | 2020M12 | 2021M3 | 2021M6 - Minimum value: 87 | 89 | 93 | 95 | 97 - First quantile: 96 | 99 | 103 | 104 | 105 - Median value: 105 | 107 | 112 | 111 | 109 - Third quantile: 114 | 121 | 128 | 129 | 133 - Maximum value: 193 | 179 | 183 | 189 | 198 - Counts of banks below thresholds (as presented): - No. of banks with NSFR < 100 percent: 5 | 2 | 0 | 1 | 0 - No. of banks with NSFR < 90 percent: 4 | 1 | 0 | 0 | 0 - No. of banks with NSFR < 80 percent: 2 | 0 | 0 | 0 | 0 ### Conclusions and Policy Recommendations - Data collection enhancements for liquidity risk monitoring: - Collect more granular data to reflect diversity of sources and uses of funds for refined quantitative exercises. - Suggested additional data items: term deposit inflows by counterparty; operational demand deposits and parent funding in unsecured wholesale funding; secured funding and lending by type of security; off-balance sheet commitments. - Additional data on time structure of cashflows beyond 90 days to assess effects of more protracted liquidity shocks. - Given large foreign exposures of some banks, extended data collection should ensure availability of cashflow information by significant currency and cross-border exposures. - Regulatory alignment: - Regulatory parameters used for computation of the IRL and the NSFR could be re-adjusted or further aligned with Basel III requirements. - Closer convergence to Basel III standards would improve international comparability of liquidity indexes and clarify why foreign headquartered banks show higher liquidity indexes relative to local banks (IRL standards are less stringent than Basel III LCR in some respects). - Extended data collection would support assessment of consequences of possible adjustments in regulatory parameters. ### Interconnectedness and Contagion Mapping (overview) - Dataset and mapping: - Wide anonymized dataset of sectoral and entity-level exposures across equities, short-term debt, long-term debt, loans, deposits, derivatives, interbank exposures, and money market operations covering banks, pension funds, insurers, investment funds, trusts, securities firms, nonfinancial corporates, households, and government entities; foreign debtor counterparts mapped where identifiable. - Sectoral and entity-level findings: - Banks lie at the core of the Colombian financial network as dominant creditors and providers of funding to households and NFCs. - Pension funds, insurers, and investment funds are prominent asset holders with significant exposures to government securities. - At the entity level, banks are connected to all main entities and sectors, producing complex interbank and cross-sector contagion channels; most banks have significant loan exposures to households and hold government securities. - Financial conglomerates: - The three largest conglomerates control about half of assets in the financial system and two-thirds of banking sector assets. - Conglomerate banks sit at the core of the interbank network and create multiple transmission channels across the system through their networks of trusts, insurers, pension funds, securities firms, and other institutions. - This complexity motivates holistic conglomerate supervision and understanding of inter-conglomerate contagion channels. *Source: SFC and IMF staff calculations as presented in the IMF Colombia FSAP chapter excerpt.* ### 77. The composition of conglomerate exposures—particularly in Central America—is ### 1colea2022007 - 77. The composition of conglomerate exposures—particularly in Central America—is ### Composition of conglomerate cross-border exposures - Total foreign assets: Col$54 trillion. - Equity exposures in subsidiaries: Col$25 trillion (46 percent) of total foreign assets. - Loans directly: Col$8 trillion (15 percent) of total foreign assets. - Note: Indirect loan exposures on a consolidated basis are higher because some loan exposures in non-Panama Central American countries occur through equity exposures in Panama. - Foreign liabilities appear dominated by: - Debt securities: Col$23 trillion. - Deposits: Col$18 trillion. - Implication: Foreign funding transmission channel is more dispersed but remains viable. ### Cross-border exposures beyond conglomerates - Banks: At least two banks with significant relative positions in Central America; several banks hold exposures to other countries. - Nonbank financial institutions (NBFIs): Exposures dominated by asset allocations to U.S. and other international securities; exposures also occur through trusts and connecting financial entities. - Implication: Holistic monitoring of the network is needed to capture these channels. ### Contagion risk modeling approach - Model used: Expanded CoMap model from Covi, Gorpe, and Kok (2019), extended to incorporate: - Bank- and exposure-specific parameters. - Liquidity dynamics reflecting parametric heterogeneity. - Nonbank financials, non-financials, and selected foreign counterparts. - Stylized shock transmission: - Entities (or aggregated sectors) default one at a time at simulation start. - Credit channel: default induces losses on creditors. - Liquidity channel: entities may conduct “fire sales” depending on surplus, liquidity constraints, and asset quality; haircuts applied by exposure type. - Iteration produces potential cascade effects and capital losses. - Modeled entities as shock absorbers/amplifiers: - 25 banks modeled as both shock absorbers and amplifiers. - 5 pension funds, largest 5 investment funds, largest 5 insurers modeled as both shock absorbers and amplifiers. - Other entities aggregated by industry and traced as shock sources but not as amplifiers. ### Key contagion model findings - Systemic banks are the most contagious entities. - Bank-only network effects understate total potential losses because banks induce liabilities in pension funds, investment funds, and other financial entities. - Systemic banks identified by supervisors align with model-identified contagion potential. - Overall contagion outcomes from bank failures: - Bank failures can lead to three contagion failures in other banks. - Bank failures can lead to six failures in nonbank entities. - Certain non-systemic banks (B6, B11, B7) can induce entity failures without causing the largest system-wide capital losses. - No bank failure scenarios led, by themselves, to second-round failures. - Pension funds, investment funds, and insurers do not induce failures in the model — caveat: data limitations and unmodelled behavioral responses mean some transmission mechanisms are not considered. ### Cross-border shock experiments - Assumed loss-given-default parameters for aggregated foreign corporates: - Equities: 25 percent loss-given-default. - Short-term debt, long-term debt, other non-loan exposures: 20 percent loss-given-default. - Loans: default of the upper quartile of riskiest loans. - Country-specific composition differences: - C1, C3, C4 shocks expand mainly through pension funds holding significant foreign securities. - C2 contagion occurs mainly through banks and equity exposures concentrated in certain conglomerates. - Result: None of the cross-border stress scenarios induced the failure of any entity modeled in the network. ### Domestic shock experiments and induced failures - Government shock: - Loss-given-default parameter assumed: 20 percent. - Government default scenario produces 16 entity failures in total. - Strong government shock induces: - First round: 11 entities (8 banks) fail. - Second round: 5 entities (4 banks) fail (cascade effects). - Household shock: - Shock: default of the riskiest quartile of loans. - Household default scenario produces 8 entity failures in total. - Household shock induces: - First round: 4 entities (4 banks) fail. - Second round: 2 entities (1 systemic bank) fail. - Third round: 2 entities (1 bank) fail — only shock to produce a third round. - Microcredit shock: - Induces failure of three banks with large microcredit exposures. - Summary: Government and household shocks produce the most contagion (government: 12 bank failures and 16 entity failures in total; households: 6 bank failures and 8 entities in total as summarized elsewhere). ### Heterogeneity of vulnerabilities across entities - Vulnerability index: aggregate losses to an entity from failures of other entities divided by a scaled measure of its own capital. - Examples of bank-specific vulnerability composition: - B12, B14, B23, B6: more vulnerable to government shocks. - B2, B20: more vulnerable to household shocks. - B18, B23: significant vulnerability to other banks. - Systemic banks generally not particularly vulnerable; only B1 failed to the household shock in the second round after joint failures. - Nonbank entities: - 12 insurer failures across simulations (mainly 2 firms failing in multiple simulations). - 1 investment fund failure across simulations. - Insurers and investment funds are significantly vulnerable due to exposures as creditors to banks relative to their excess solvency capital. ### Entities meriting close monitoring - Entities with both high contagion and high vulnerability: - Banks: B11, B12, B20. - Insurer: INS2. - Recommendation: Careful monitoring of these entities is warranted. ### Modeling and data caveats - Model limitations: - CoMap does not account for market perception shifts that could accelerate contagion. - Single solvency threshold and full-default assumptions can overstate capital losses. - Banks could be supported by policy interventions or internal responses not modeled. - Data limitations: - Exercise rules out amplification through NFCs, government entities, households, large investors, trusts, securities firms, nonfinancial corporates, and foreign entities as amplifiers due to data gaps. - Not all contagion routes within conglomerates or via foreign funding channels are modeled. - Limited detailed information on foreign holdings of Colombian obligations constrains modeling of foreign funding contagion. ### Summary and policy implications - Main conclusions: - Contagion risks concentrated around large domestic exposures to households and government entities. - Cross-border risks are relatively contained in modeled scenarios. - Vulnerability to government securities: produces 12 bank failures and 16 entity failures in total (per summarized findings). - Vulnerability to households: produces 6 bank failures and 8 entities in total (per summarized findings). - Household stress can cause a systemic bank to fail via second-round effects. - Insurers present vulnerabilities due to concentrated exposure to banks relative to solvency capital. - Policy implications and recommendations: - Further development of monitoring tools is needed to capture evolving network complexity and multiple contagion channels (trusts, securities firms, other financial institutions). - Data and methodology improvements to bolster network mapping and contagion assessment: - Incorporate individual trusts and securities firms into contagion risk analysis. - Advance identification of foreign creditors on Colombian claims to monitor cross-border funding risks. - Gather further data on composition of conglomerate subsidiary operations in Central America. - Goal: Bolster early-warning systems (EWS) while vulnerabilities are building. - Roles and responsibilities: - Data coverage improvements lie with the SFC. - The Banco de la República’s (BR) ability to improve network analysis tools depends on access to necessary data. - FSAP Technical Note recommendation: - Improved coordination in systemic risks monitoring and leveraging the BR’s technical expertise, including critical access to necessary data for these tasks. *Italic: Sources: SFC; and staff calculations; CoMap model calibrated to December 2020 data.* ### 91. This analysis is conducted using firm-level data on Colombian firms from the ### Firm-level analysis of Colombian firms from Supersociedades (2016–2020) ### Data and sample - Source: Firm-level data from the Superintendency of Companies (SSC, Supersociedades); annual financial (end-December) and income statements for corporates and SMEs for the period 2016–2020. - Coverage: 21 industries (SIC classification) and represents about 80 percent of GDP. - Number of firms in database: - Corporates: ranges between 3,000 and 4,000 per year (specific yearly counts in Table 9: 2016: 3,469; 2017: 3,397; 2018: 3,556; 2019: 3,706; 2020: 3,346). - SMEs: ranges between 16,000 and 23,000 per year (specific yearly counts in Table 9: 2016: 19,681; 2017: 16,410; 2018: 15,938; 2019: 21,638; 2020: 22,720). - Panel persistence: - About 60 percent of corporates appear for four or five years. - About 40 percent of SMEs appear for four or five years. - Above 40 percent of SMEs appear only one or two years (reflecting higher SME entry and exit). - Banking exposure: sample entities capture 38 percent of gross loan portfolio of banks (25 percent by corporates and 13 percent by SMEs). - Note: SSC classification uses financial accounting (IFRS or NIIF in Colombia) to classify firms as corporates (plenas) or SMEs (pymes). ### Sectoral concentration - Corporates (by number): concentrated in industry, wholesale and retail, and construction. - Corporates (by assets): concentrated in industry, wholesale and retail, other financial services, and construction. - SMEs (by number): concentrated in wholesale and retail, real estate activities, industry, and construction. - SMEs (by assets): mainly concentrated in construction and real estate activities. ### Pre-pandemic financial conditions (2017–2019) - Median firm trends: - Liquidity and profitability: relatively stable; median firm showed increases in liquidity and profitability in 2019. - Leverage ratios (debt-to-assets and debt-to-equity): showed a slight decline in 2019. - Interest coverage ratio (ICR): - ICR was above 1 for median firms. - Corporates exhibited a lower median ICR compared to SMEs. - Corporates had a higher debt share associated with ICR<1 (slight decline in 2019). - Cash available ratio (defined as cash and equivalent plus receivables minus short-term liabilities (excluding short-term debt); data limitations did not allow to exclude the portion of long-term debt paid in the current year): - Relatively stable before the pandemic. - Share of debt of firms with cash available ratio below zero was relatively stable between 2017 and 2019. - Pre-pandemic sectoral vulnerabilities (largest debt shares of firms with ICR<1): - Corporates: real estate, wholesale and retail, other financial services, and energy. - SMEs: transport and energy. ### Pandemic impact (2020) - General effects in 2020: - Repayment capacity and profitability worsened for both corporates and SMEs. - Firms accumulated more cash, increasing the cash available ratio. - Leverage indicators (debt-to-assets and debt-to-equity) exhibited a slight decline. - Profitability (return on assets and return on equity) declined due to lower firm revenues. - Sectoral effects: - Many sectors—especially services—showed deterioration in repayment capacity and profitability. - Oil sector experienced a dual shock: pandemic effects plus a sharp oil price decline. - Debt share of firms with ICR<1 increased across most economic sectors for both corporates and SMEs. - Table 10: Change in indicators from 2019 to 2020 (select sectoral changes; "Change of the indicator between 2020 and 2019"): - Corporates: - Agriculture: Cash 0.3; ICR 0.1; Debt-to-assets 1.8; ROA 0.6; ROE 1.1 - Oil/Gas/Mining: Cash -0.1; ICR -2.6; Debt-to-assets 0.4; ROA -1.6; ROE -4.8 - Industry: Cash 1.6; ICR -0.1; Debt-to-assets -1.6; ROA 0.0; ROE -0.4 - Construction: Cash -0.5; ICR -0.5; Debt-to-assets 0.0; ROA -0.3; ROE -0.7 - Wholesale/retail: Cash 2.0; ICR -0.1; Debt-to-assets -1.9; ROA 0.1; ROE 0.0 - Transport: Cash 4.6; ICR 0.8; Debt-to-assets -0.9; ROA 0.5; ROE -2.4 - Real estate: Cash 0.3; ICR -0.3; Debt-to-assets -1.3; ROA 0.2; ROE -0.1 - Telecomunications: Cash 3.0; ICR -0.8; Debt-to-assets -0.8; ROA -0.9; ROE -6.9 - Energy: Cash -0.7; ICR -0.3; Debt-to-assets 4.9; ROA -3.0; ROE -2.6 - Other financial: Cash 0.6; ICR -1.0; Debt-to-assets 1.5; ROA -0.7; ROE -0.4 - SMEs: - Agriculture: Cash 0.6; ICR 0.3; Debt-to-assets -0.4; ROA 0.3; ROE 0.4 - Oil/Gas/Mining: Cash 0.2; ICR -1.1; Debt-to-assets 2.0; ROA -0.6; ROE -1.1 - Industry: Cash 1.1; ICR -0.1; Debt-to-assets 0.3; ROA -0.4; ROE -1.2 - Construction: Cash 0.5; ICR 0.1; Debt-to-assets -0.3; ROA -0.3; ROE -1.4 - Wholesale/retail: Cash 1.4; ICR -0.1; Debt-to-assets -0.2; ROA -0.1; ROE -0.7 - Transport: Cash 1.4; ICR -1.0; Debt-to-assets -0.6; ROA -0.7; ROE -1.8 - Real estate: Cash 0.2; ICR -0.5; Debt-to-assets 0.4; ROA -0.3; ROE -0.6 - Telecomunications: Cash 1.1; ICR 0.8; Debt-to-assets 0.4; ROA -0.7; ROE -1.7 - Energy: Cash 4.2; ICR 1.1; Debt-to-assets 5.7; ROA -1.2; ROE -0.7 - Other financial: Cash 0.3; ICR -1.1; Debt-to-assets -0.8; ROA -0.3; ROE -0.7 - Note: Data limitations do not allow distinguishing firms that obtained COVID-related support measures. ### Stress test — Adverse scenario (dynamic scenario-based analysis) - Methodology: Follows Tressel and Ding (2021, forthcoming); focuses on firm-level indicators to assess ability to service debt (ICR) and external borrowing needs (cash available ratio) given shocks to macrofinancial variables (GDP growth and an index of financial conditions). Annex I references methodology; adverse scenario implies a moderate two-year recession and a two-year increase of the index of financial conditions. - Financial conditions index: calculated following IMF (2017) Global Financial Stability Report methodology; for Colombia includes: Spread of local currency sovereign debt, spread of US dollar sovereign debt, spread of US dollar corporate debt, equity prices, exchange rate, and house prices. - Key drivers of distress: - Probability a firm is unable to pay interest (ICR<1) is: - Negatively correlated with profitability. - Positively correlated with leverage. - Negatively correlated with GDP growth. - Probability cash available ratio < 0 is: - Negatively correlated with profitability. - Positively correlated with leverage. - Negatively correlated with GDP growth. - Adverse scenario impacts (second year of adverse scenario unless noted): - ICR<1: - Corporates: share increases by 3 percentage points (from 39 percent to 42 percent) by the second year, then declines by the third year as GDP growth and financial conditions recover. - SMEs: share increases by 3 percentage points (from 25 percent to 28 percent) by the second year, then declines by the third year. - For 2020 (pandemic year), the share increased by about 5 percentage points (outcome included pandemic-related support policies; adverse scenario impacts do not account for potential implementation of support policies). - Cash available ratio < 0 (firms with potential borrowing needs): - Corporates: share increases by about 9 percentage points to 45 percent by the second year of the adverse scenario. - SMEs: share increases by about 9 percentage points to 42 percent by the second year. - In 2020, the share declined between 4 and 5 percentage points (due to firms increasing cash positions at beginning of pandemic). The adverse scenario reflects a more typical recession; shares increase with GDP decline. - Caveat: Impacts from the adverse scenario do not account for potential implementation of support measures that could ameliorate negative impacts on firms. ### Conclusions and policy recommendations - Monitoring priorities: - Continue monitoring the nonfinancial corporate sector, especially the services sector, to identify potential pockets of vulnerability. - Continue assessing firms in sectors benefiting from the strong economic rebound to identify firms for which the 2020 shock may not be temporary. - Early warning (Alerta Temprana): - Ongoing work on early warnings could be broadened to increase the number of firms in the sample. - Assess whether information from early warning could inform preparation of contingency (policy) plans by supervisory authorities, given sensitivity of repayment capacity, cash available ratio, and solvency to adverse shocks. - To the extent legally feasible, sharing parts of the early warning information with other supervisory institutions is recommended to complement monitoring and receive feedback to strengthen the early warning exercise. - Note: Alerta Temprana is a newly implemented SSC project using descriptive and predictive analytics based on granular firm-level data. *Source: Supersociedades; IMF staff calculations; IMF chapter content (firm-level analysis, Colombia).* ### Appendix II. Stress Test Matrix ### Appendix II. Stress Test Matrix ### Banking Sector: Solvency Risk — Institutional perimeter and data - Institutions included: The top 12 commercial and retail banks. - Market share: Nearly 93.8 percent of total assets in the banking system. - Data and baseline date: - Cut-off date: June 2021. - SFC bank-by-bank supervisory data including credit risk-sensitive exposures, market risk-sensitive exposures, and interest rate sensitive assets and liabilities, as well as historical information on NPLs by portfolio, and details of lending and funding rates. - Other market and publicly available data. - Scope of consolidation: banking activities on an individual basis. ### Banking Sector: Solvency Risk — Methodology and horizon - Methodology: - FSAP team satellite models and methodologies. - Balance-sheet approach with quasi-static balance sheet assumption. - All banks follow the standardized approach. - Satellite models for macrofinancial linkages: - FSAP team’s own model for credit losses from bank lending portfolios. - NPL models for each loan segment (commercial, consumer, and mortgage) with bank-specific fixed effects. - Models for funding costs with bank-specific fixed effects. - Stress test horizon: Three years (mid-2021-mid-2024). ### Banking Sector: Solvency Risk — Scenarios and sensitivities - Scenario analysis: - Macrofinancial scenario analysis. - Baseline scenario based on the projections of October 2021 WEO. - Adverse stress scenario: deviation from baseline forecasts, triggered by a series of global and domestic shocks, capturing the key risks in the RAM. - Sensitivity analysis: - Single-factor sensitivity test for concentrations risk, where the banks’ top single, five and ten largest exposures are assumed to fail simultaneously. ### Banking Sector: Solvency Risk — Risks, modeling assumptions, and behavioral adjustments - Risks/factors assessed: - Credit losses from lending: - Modeled NPLs for three portfolios (commercial, consumer, and mortgage) at a single geography level. - Estimated panel regression coefficients with fixed effects using the Bayesian Model Averaging (BMA) approach subject to sign constraints. - Provisioning ratios for the NPLs calculated based on historical data for each portfolio segment. - Specific procyclical and countercyclical provisions were treated uniformly due to data limitations; countercyclical provisions were not treated as additional buffer. - Market risk: - Losses due to changes in market variables: risk-free interest rates, exchange rate and equity prices, making no allowance for macro hedges. - For sovereign and corporate debt securities (under AFS or HFT), a modified-duration formula was used to revalue exposure as function of their residual duration and the relevant bond yield assumptions under the scenarios. - Foreign exchange risk and equity based on respective net open positions and the paths of equity index and exchange rate under the scenarios. - Non-interest income: - Impact of funding and lending interest rate shocks on net interest income estimated by measuring the gap between interest sensitive assets and liabilities using time-to-repricing buckets. - Effective interest rates estimated using satellite models and pass-through assumptions. - Other P&L components: - Net fee and commission income, other non-interest income and non-interest expenses projected in line with growth of each bank’s balance sheet. - Extraordinary income and losses assumed not to incur during the projection period. - Banks’ indirect exposures through subsidiaries: shocked by 50 percent and 25 percent in the first and second years, respectively, of the adverse scenario. - Income tax: calibrated as the median of historical tax expense ratios per bank. - Counterparty concentration risk: losses from simultaneous failure of top 1, 5, and 10 largest single-counterparty exposures. - Behavioral adjustments: - Quasi-static balance sheet assumption: balance sheet grows proportionally with nominal GDP under baseline and adverse scenarios, with a lower zero-bound (no ex ante deleveraging allowed). - Dividends can only be paid by banks that remain adequately capitalized and have positive profits. - New business repricing: maturing instruments replaced by identical new instruments at reference and margin rates implied by the scenario and pass-through assumptions. - Stress test did not take into account loan write-offs and cures due to data unavailability. ### Banking Sector: Solvency Risk — Regulatory parameters - Calibration of risk parameters: Based on credit models estimated by IMF staff. - Regulatory/accounting and market-based standards: - Hurdle rate based on the current capital adequacy framework accounting for phase-in of several components during the projection period. - A minimum total CAR of 9 percent applies. - Tier 1 (T1) hurdle rate: currently 4.9 percent (in 2021) and gradually increased to 6 percent by January 2024 until full convergence. - Minimum leverage ratio (tier 1 capital to total assets) of 3 percent applies throughout. - Capital conservation buffer (CCoB) and systemically important bank buffer (SIB) phased in gradually to reach 1.5 percent and 1 percent of RWAs by January 2024. - Capital projection incorporates gradual phase-out of eligible capital instruments from additional tier 2, with identical impact under baseline and adverse scenarios. ### Banking Sector: Solvency Risk — Reporting format - Output presentation: - Capital ratio decline of the banking system. - Number of banks and the percentage of banking assets (or GDP) in the system that fall below a hurdle rate. - Decomposition of the reduction in capital ratio by drivers (credit risk, market risk, interest rate risk, etc.). --- ### Banking Sector: Liquidity Risk — Institutional perimeter and data - Institutions included: The top 12 commercial and retail banks. - Market share: Nearly 93.8 percent of total assets in the banking system. - Data and baseline date: - Baseline date: June 2021. Data for December 2019, June 2020 and December 2020 will be used for comparison. - Source: Data compiled based on national reporting templates 458, 531, and 238 (regulatory returns monitoring the LCR and the NSFR), additional off-balance sheet data capturing balance of credit/liquidity facilities, and reporting template 474 covering the biggest depositors. - Scope of consolidation: individual bank basis. ### Banking Sector: Liquidity Risk — Methodology and tests - Methodology: - Exercise based on LCR-based and NSFR-based tests, and cash-flow analysis. - LCR-based test aligned with IRL—the SFR’s implementation of Basel’s LCR—on individual basis and for all currencies combined. May be complemented by CB’s liquidity metric (IEI) focusing on exchange rate risk. - Scenarios to produce stressed LCR exploring retail and wholesale funding shocks, complemented by large depositors’ concentration analysis. - Cashflow analysis based on data corrected by the SFC for IRL for the period up-to 90 days on individual basis and for all currencies combined; scenarios with run-off and roll-over rates and liquid assets haircuts implemented. - NSFR requirements gradually implemented: 100 percent limit for banks with assets greater than 2 percent of total system assets and 80 percent limit for other institutions with credit portfolio being primary activity by March 31, 2022, and informative for the rest. NSFR-based test uses national regulatory parameters, individual basis and for all currencies combined, with the hurdle at 100 percent. ### Banking Sector: Liquidity Risk — Risks, buffers, and shocks - Risks: - Funding liquidity. - Market liquidity. - Depositor concentration risk (withdrawal of largest depositors). - Buffer: - Counterbalancing capacity, including liquidity obtained from markets and/or regular central bank facilities (excluding the central bank’s emergency liquidity assistance (ELA) or other temporary measures). - Expected cash inflows included in cashflow and LCR-based analysis. - Tail shocks: - Stress scenarios on liability outflows reflecting retail outflows, wholesale funding outflows, and large depositors’ withdrawals, and scenarios on liquid assets shock. ### Banking Sector: Liquidity Risk — Regulatory standards and calibration - Regulatory standards: - LCR per Basel III—the hurdle at 100 percent. - NSFR per Basel III—the hurdle at 100 percent. - Cashflow analysis—a non-negative net cash balance, where the balance reflects net funding outflows and counterbalancing capacity. - Calibration of risk parameters: - Run-off rates, roll-over rates, and asset haircuts based on historical events, empirical evidence, and IMF expert judgment. ### Banking Sector: Liquidity Risk — Reporting format - Output presentation: - Distribution of liquidity positions by groups of banks and aggregated (system wide). - Number of institutions with LCR/NSFR below 100 percent and/or negative net cash balance. - Amount of liquidity shortfalls. --- ### Banking Sector: Interconnectedness Analysis and Contagion Risk — Institutional perimeter and data - Institutions included: - Active nodes: - All 25 banks (DSIBs and non-DSIBs); - Top five pension funds; - Top five insurance companies; - Top five investment funds; - Passive nodes: - Residual pension funds aggregated; - Residual insurance companies aggregated; - Residual investment funds aggregated; - Other financial sectors aggregated: trusts, securities firms, and other financial intermediaries; - Nonfinancial corporates aggregated at industry level (20 industries); - Other aggregated sectors: government (Colombia), households (Colombia), top five countries aggregated at country level, rest of the world foreign counterparts. - Market share: Active nodes account for 100 percent of total assets in the banking system. - Data and baseline date: - Date: December 2020. - Source: supervisory data on bilateral linkages within the banking system and with other entities to construct a large-exposures matrix, and solvency and liquidity related reports to calibrate model parameters. - Scope of consolidation: active nodes (banks and other financial entities) on individual basis. Sector residuals and nonfinancial entities aggregated at sector level. Exposures among members of a conglomerate are mapped in the network to analyze intra-conglomerate linkages. ### Banking Sector: Interconnectedness — Methodology and risk channels - Methodology: - Application of CoMap framework (Covi, Gorpe and Kok, 2019), with extensions to model contagion through nonbank financial entities. - Risks/factors assessed: - Credit default channel: impact of an entity defaulting on its obligations to other entities; creditor incurs losses on a share of its claims depending on nature and counterparty; exposure-specific loss-given default rates reflect risk mitigation and collateralization. - Funding shock and liquidity dynamics: impact of a funding withdrawal or non-rolling over by a failed entity; entities can use surplus high-quality liquid assets as collateral to borrow; once depleted, deleveraging by selling marketable securities at discount incurring haircut losses. - Cascading defaults due to insolvency and/or illiquidity failures: insolvency default when solvency threshold compromised; illiquidity-driven default when an entity runs out of marketable assets that can be sold. ### Banking Sector: Interconnectedness — Scenarios and reporting - Tail shocks: - Idiosyncratic shocks: hypothetical failure of each entity as a potential trigger event. Exercise comprises as many simulations as there are nodes in the network, where each simulation considers the hypothetical failure of a single node. - Output presentation: - Number of cascade defaults and amplification effects. - Contagion losses induced (contagion index) and experienced (vulnerability index) by each node. - Decomposition of losses by entity types. --- ### Nonfinancial Corporate Stress Test — Institutional perimeter and data - Institutions included: - 26,066 firms in cut-off year: 3,346 corporates and 22,720 SMEs. - Covers 21 industries. - Market share: Represent about 80 percent of GDP. - Data and baseline date: - Cut-off date: December 2020. - Source: firm-level time-series balance sheet and income statement data from the Superintendency of Companies (SS, Supersociedades). - Scope of consolidation: individual reporting basis (unconsolidated). ### Nonfinancial Corporate Stress Test — Methodology and horizon - Methodology: - Following methodology developed by Tressel and Ding (2021) in Global Corporate Stress Test: COVID-19 Impact and Medium-term Implications (forthcoming). - Assessment of: - (i) the ability to service debt (ICR); - (ii) external borrowing needs based on cash balance (cash available); - (iii) solvency position. - Models: - Dynamic regressions for return on assets (ROA) and leverage. - Probit models for cash balance (cash available) and ICR. - Stress test horizon: Three years (mid-2021-mid-2024). ### Nonfinancial Corporate Stress Test — Scenarios and reporting - Scenario analysis: - Macrofinancial scenario analysis: same scenario as the bank solvency stress-test. - Baseline scenario based on the projections of October 2021 WEO. - Adverse stress scenario designed as deviation from baseline forecasts, triggered by a series of global and domestic shocks, capturing the key risks in the RAM. - Output presentation: - Share of firms with lower repayment capacity (ICR<1). - Share of firms with potential borrowing needs (Cash Available<0). --- ### Appendix III. Technical Details of the Stress Test Macro Scenario Calibration — Model and adverse scenario design - Macromodel: - Adverse scenario projections simulated using the IMF’s Global Macrofinancial Model (GFM). - GFM: dynamic stochastic general equilibrium model disaggregated into forty national economies, with nominal and real rigidities, extensive macrofinancial linkages, and diverse spillover transmission channels (Vitek. 2018). - Colombia included among advanced and emerging market economies in the sample. - GFM parameterization based on mix of calibration and estimation; subsets of parameters jointly estimated by full information maximum likelihood conditional on calibrated values and observed variables. - Adverse scenario characteristics: - Projected average real GDP growth of 0.5 percent per annum over 2021Q3–2023Q3 under the adverse scenario. - Year-by-year GDP growth under the adverse scenario: -1.2 percent, -1.3 percent, and 4 percent in the first, second, and third years of the horizon, respectively. - Cumulative decline of GDP relative to the baseline over three (two) years would be about 11 (11.3) percentage points, equivalent to a 1.8 (2.5) standard deviation shock taking the 1995–2020 period as the benchmark. - The adverse scenario reflects a very severe deterioration relative to baseline, with recession more prolonged due to continuing outbreaks, social unrest and tightening of financial conditions; global factors have a slightly heavier weight in deviation from baseline. - The scenario is within FSAP protocols and is in the higher range of shocks used in FSAPs; it is more severe than the 2.5 standard deviation shock applied in the first two years compared with a 1.6 standard deviation shock in the June 2021 BR stress test. - Calibration of market and asset-price shocks: - Shock to the Colombia 10-year government bond yields in the first year aligned with the nearly 190 basis points average increase in 2008 (largest recent change). - Real housing prices fall by 21 percent over the three-year horizon (based on the fifth percentile of three-year cumulative changes in residential real estate prices). - Colombian peso depreciates bilaterally against the US dollar by about 25 percent over the stress-test horizon (one-standard-deviation shock based on historical exchange rate movements). - Equity prices (COLCAP stock price index) drop by 18 percent in the first year (fifth percentile of single-year changes in the stock index). - Interbank rate diverges from the policy rate with the spread reaching 125 basis points in year 2 of the horizon as liquidity dries up and counterparty risks heighten. *Source: Appendix II. Stress Test Matrix and Appendix III. Technical Details of the Stress Test Macro Scenario Calibration (cut-off and baseline dates and model specifications as provided).* ### Appendix III. Figure 1. Colombia: Adverse Scenario Severity ### 1colea2022007 - Appendix III. Figure 1. Colombia: Adverse Scenario Severity ### Scenario overview and historical perspective - Figure panel a: Scenario Severity from Historic Perspective (Real GDP (T0=100)). - Figure panel b: Cumulative Deviation from Baseline in Percent (equivalent standard deviations next to columns). - Sources cited in figures: World Economic Outlook; IMF staff calculations. ### FSAP macro projections — Baseline and Adverse scenarios (quarterly periods) Periods: - 2020Q3-2021Q2 (T0) - 2021Q3-2022Q2 (T1) - 2022Q3-2023Q2 (T2) - 2023Q3-2024Q2 (T3) Key projection rows (in percent unless otherwise indicated): - Real GDP growth (percent) - Baseline: T0 1.2, T1 5.0, T2 3.6, T3 3.3 - Adverse: T1 -1.2, T2 -1.3, T3 4.0 - Nominal GDP growth (percent) - Baseline: T0 4.2, T1 6.3, T2 8.1, T3 6.1 - Adverse: T1 -1.1, T2 -0.4, T3 3.7 - Inflation rate (percent) - Baseline: T0 2.0, T1 4.1, T2 3.0, T3 3.0 - Adverse: T1 3.8, T2 2.0, T3 1.8 - Unemployment rate (percent) - Baseline: T0 15.5, T1 14.2, T2 13.4, T3 12.8 - Adverse: T1 16.9, T2 18.2, T3 17.8 - Interbank lending rate (percent) - Baseline: T0 1.8, T1 2.5, T2 3.9, T3 4.4 - Adverse: T1 3.4, T2 4.3, T3 4.6 - 1-year government bond yield (percent) - Baseline: T0 2.3, T1 3.8, T2 4.7, T3 5.0 - Adverse: T1 4.2, T2 4.8, T3 4.7 - 10-year government bond yield (percent) - Baseline: T0 6.5, T1 7.8, T2 7.6, T3 7.2 - Adverse: T1 9.0, T2 10.1, T3 8.7 - Housing price (index) - Baseline: T0 100.0, T1 105.1, T2 108.9, T3 112.5 - Adverse: T1 88.2, T2 80.2, T3 78.8 - Equity price (index) - Baseline: T0 100.0, T1 106.3, T2 114.9, T3 121.9 - Adverse: T1 83.0, T2 79.0, T3 87.7 - Energy commodity prices (index) - Baseline: T0 100.0, T1 143.0, T2 124.9, T3 115.3 - Adverse: T1 112.9, T2 60.6, T3 68.3 - Non-energy commodity prices (index) - Baseline: T0 100.0, T1 110.3, T2 107.7, T3 107.2 - Adverse: T1 100.9, T2 79.5, T3 81.9 ### Credit risk — satellite model design and findings - Data and scope: - Longest available time-series: stock of NPLs. - Quarterly data starting 2002; estimations exclude the pandemic period (2020). - NPL definitions: consumer and commercial portfolios capture loans 90 days due or longer; mortgage loans capture loans 120 days due or longer. - Modeling approach: - Bayesian Model Averaging (BMA) used to develop satellite models. - Models estimated as panel regressions with fixed effects; LHS variables are NPLs for commercial loans, consumer loans, and mortgages. - No autoregressive lags of LHS variables were included. - Potential RHS variables: GDP growth, unemployment rate, exchange rate, lending rate, VIX, term premium, consumer price index, and real housing price index, including lags. - Model selection and weighting criteria: sign criteria, R-square, AIC, and size of impact in the forecasting period (adverse-baseline gap). - RHS variables with posterior inclusion probabilities equal or higher than prior inclusion probabilities are indicated with an asterisk (in source tables). - Empirical findings: - NPL ratios are generally more sensitive to unemployment rate and GDP. - Unemployment rate plays the most significant role for consumer and mortgage NPLs. - For mortgages: housing prices are the next most important factor. - For consumer loans: lending rates are an important determinant. - For commercial portfolio: GDP contributes more strongly to NPL projections in year 1; term spreads and VIX have heavier weight in year 2. - Expected losses and provisioning: - Expected losses calculated using historically calibrated provisioning ratios. - Provisioning ratio calculated as loan provision stocks divided by NPLs per portfolio (provision stocks include both procyclical and countercyclical individual provisions). - Countercyclical provisions could not be separately identified historically; therefore countercyclical provisions were treated uniformly with procyclical provisions and a single ratio was used to estimate credit losses without considering countercyclical buffers. ### Interest rate risk on banking book (Net Interest Income analysis) - Two-pronged approach for net income projections: - Funding costs estimated using satellite models for different liability segments. - Granular bank-specific deposit rate data by maturity available from 2014 onwards on a quarterly frequency. - Deposit segments modeled: - Deposits with less than 3 months maturity. - Deposits with maturity between 3 months and 12 months. - Deposits with maturities over 1 year. - Empirical drivers: - Deposit rates explained partly by their own lags; interbank rate consistently significant for all deposit rates. - Long-term risk-free rates significant for longer-maturity deposit rates. - Other liability segments: where modeling was not possible, shocks on corporate credit spreads were used under each scenario. - Loan effective interest rate adjustments: - Shifts calculated using pass-through rates. - Pass-through rates lower under the adverse scenario. - Effective pass-through rates for each bank depend on proportion of fixed vs. floating loans and time-to-repricing (portfolio composition and time-to-repricing distribution shown in Figure 1 in source). ### Liquidity stress tests — technical assumptions and parameters - Appendix V contains LCR-based test assumptions, run-off rates for cashflow analysis, roll-off rates for cash inflows, and haircuts for cashflow analysis (tables and notes in source). - Run-off rates note: - A run-off rate indicates the fraction of the liability amount maturing in each time bucket that is withdrawn (and not rolled over) by the claim holders. - Run-off rates for demand deposits and credit lines and revolving facilities defined as fraction of the initial outstanding balance withdrawn in each time bucket. - Reported data on cash outflows due to maturing term deposits are not disaggregated by client type; disaggregation in the adverse scenario approximated using data on balances of term deposits held by different counterparties. - Roll-off rates note: - For every type of asset and maturity bucket, roll-off rates indicate the fraction of the amount maturing that is converted into a cash inflow (and not rolled over) by the bank. ### Interconnectedness and contagion analysis — methodology - Framework: - Balance sheet-based network analysis follows the CoMap framework (Covi, Gorpe, and Kok 2019), extending Espinosa-Vega and Sole (2011). - Incorporates entity- and exposure-specific parameters and liquidity dynamics, nonbank financials, nonfinancials, and selected foreign counterparts. - Credit shock: - Losses from counterparty defaults computed using exposure-specific loss-given-default rates, 휆휆푖푗푘, applied to claims 푥푖푗푘 across defaulting counterparties. - 휆휆 parameters constrained to [0,1] for non-defaulted entities in formula (2) in source. - Funding shock and liquidity dynamics: - Funding shortfall aggregated across defaulting counterparties using entity-specific funding shortfall rate 휌휌푖 (formula (5) in source). - Entities can offset funding shortfalls with central bank credit lines up to liquidity surplus 훾훾푖 (HQLA in excess of net liquidity outflows); remaining shortage computed as in formula (7). - Liquidity constraint 휃휃푖 sets upper threshold of assets available for fire sales; deleveraging amount accounts for haircuts 훿훿푖 (formula (8)). - Solvency and illiquidity conditions: - Insolvency: entity i fails if capital surplus (푐푐푖−푐푐푖푎) insufficient to cover combined credit and fire-sale losses (formula (12)). - Illiquidity: entity i becomes illiquid if remaining assets insufficient to match liquidity shortage after exhausting liquidity surplus (formula (13)). - Simulation procedure: - For each simulation, the system is tested for a given entity default (initial default of entity 1 triggered by design to study cascades). - Subsequent rounds proceed if additional failures occur; stop when no more failures. - Sequence repeated for each entity as the trigger default event. - Output indicators generated: - Contagion index (CCI): captures each entity’s potential contagion by weighted average of losses of all other entities in percent of their capital (formula provided in source). - Vulnerability index (VVI): gauges each entity’s degree of vulnerability averaged across all individual default events with identical probability (formula provided in source). - Contagion default: counts number of entities experiencing severe distress associated with the triggered default of entity i (discrete indicator). - Default frequency: tally of simulations under which entity i falls below the capital distress threshold (discrete indicator). ### Interconnectedness analysis — data calibration and implementation - Data scope and node coverage: - Individual exposures as of December 2020 at entity level connecting banks, pension funds, insurers, investment funds, trusts, securities firms, nonfinancial corporates, households, and public entities. - Exposure types: listed and unlisted equities, short-term debt, long-term debt, loans, deposits, derivatives, interbank exposures, money market operations. - For supervised entities: balance sheet data obtained and linked to exposures. - Final domestic network nodes included: - All 25 banks operating in Colombia - 5 pension funds (top five; others aggregated) - Largest 5 insurers (plus aggregated small insurers node) - Largest 5 investment funds (plus aggregated small investment funds node) - 20 nonfinancial aggregated industries - Aggregated sectoral nodes for trusts, securities firms, financial intermediation, public government entities, and households - Cross-border exposures identification: - Three methods: loan exposures linked to foreign debtor identifiers; exposures to entities classified as foreign subsidiaries; entities associated to a foreign conglomerate identified by country. - Country-specific labeling available for latter two methods only. - Country categories created including one aggregate for non-Panama Central America. - Top five debtor country categories included as separate network nodes plus ROW (Rest of the World). - Limitation: only part of potential foreign debtor shock channels considered. - Parameter calibration highlights: - Loss-given-default rate (휆휆푖푗푘): calibrated at exposure level by type. - For loans: based on risk classification and impaired assets information. - For short-term and long-term debt: calibrated to .8. - For equities: calibrated to 1 (listed or unlisted shares). - For interbank exposures, money market operations, derivatives, and deposits: set to ratio of net vs gross exposures. - For aggregated sectors: partial default parameter of .2 for public entities; for other aggregated sectors only top quartile loan exposures (by LGD) considered. - Funding shortfall rate (휌휌푖): calibrated based on data on remaining maturity and set to 1 for maturities below 30 days. - Liquidity surplus (훾훾푖): defined by entity type: - Banks: HQLA minus net outflows. - Pension funds and insurers: total cash liquidity. - Investment funds: HQLA or equivalent assets. - Liquidity constraint (휃휃푖): calibrated as total amount of unencumbered eligible assets. - Fire-sale discount rate (훿훿푖): estimated per entity based on portfolio of unencumbered non-central bank eligible assets weighted by relevant haircuts; respective haircut rates applied. - Default/distress threshold (푐푐푖푎): defined by entity type: - Banks: based on total CAR minimum requirement plus systemically important institutions buffer (applicable to DSIBs). - Insurance companies: based on available solvency capital (own funds) and minimum capital requirement. - Pension and investment funds: based on difference between total assets and total liabilities with minimum set at zero. ### Sectors included in network analysis (as listed) - Financial sectors at entity level: - Banks (25) - Pension funds (Top five, others aggregated) - Insurers (Top five, others aggregated) - Investment funds (Top five, others aggregated) - Trusts - Securities firms - Other financial intermediation - Microcredit debtors - Other aggregates - Households - Foreign counterparts (Five aggregated country categories) - Aggregated nonfinancial industries: - Accommodation and meal services - Activities of extraterritorial organizations - Administration and support services - Agriculture and livestock - Art and entertainment - Communications - Construction - Education - Employees - Manufacturing - Mining and quarrying - Other community, social and personal services - Professional, scientific, and technical services - Private Household Activities - Real estate, rental, and business activities - Social and health services - Transport, storage, and communication - Wholesale and retail - Other nonfinancial corporates - Note: Aggregated nodes are sources of shocks in the model, but not shock transmitters. *Source: IMF staff compilation from the content unit "Appendix III. Figure 1. Colombia: Adverse Scenario Severity" and accompanying appendices in the provided document.* ### Appendix VI. Table 2. Summary Statistics of Calibrated Model Parameters ### Appendix VI. Table 2. Summary Statistics of Calibrated Model Parameters ### Calibrated parameters by investor type - Banks - Loss-given default - Average 0.17 - Min 0.00 - Max 0.54 - Funding shortfall - Average 0.36 - Min 0.06 - Max 0.99 - Discount rate - Average 0.48 - Min 0.08 - Max 0.67 - Pension funds - Loss-given default - Average 0.47 - Min 0.32 - Max 1.00 - Funding shortfall - Average 0.00 - Min 0.00 - Max 0.00 - Discount rate - Average 0.22 - Min 0.21 - Max 0.22 - Investment funds - Loss-given default - Average 0.52 - Min 0.27 - Max 0.72 - Funding shortfall - Average 0.13 - Min 0.00 - Max 0.56 - Discount rate - Average 0.27 - Min 0.22 - Max 0.33 - Insurers - Loss-given default - Average 0.39 - min 0.32 - Max 0.46 - Funding shortfall - Average 0.46 - min 0.00 - Max 1.00 - Discount rate - Average 0.20 - min 0.18 - Max 0.27 *Appendix VII. Corporate Risk Analysis—Technical Details* ### Methodology and model structure - Framework follows Tressel and Ding (2021, forthcoming): dynamic regressions for ROA and leverage; Probit models for cash available and ICR. - Dependent variables: - OLS: ROA and leverage ratio. - Probit: indicator =1 if ICR < 1; indicator =1 if cash available ≤ 0. - Basic dynamic OLS specification (firm i, industry s, year t): - Y_i,s,t = α·Y_i,s,t−1 + Δ·fiem_chae_i,s,t−1 + Φ·Macro_t + a_s + ν_i,s,t - where firm-level explanatory variables are lagged one period; macro variables enter at year t; industry fixed effects included; residuals clustered at industry-year level. - Probit specification: - P[R_i,s,k = 1] = β·fiem_chae_i,s,t−1 + δ·Macro_t + a_s + ε_i,s,t - Indicator R = 1 if firm is assessed risky (below threshold), = 0 otherwise. - Firm-level explanatory variables lagged by one period include: return on assets (ROA), leverage (debt-to-asset ratio), size (total assets), tangibility (fixed assets / total assets), turnover (sales / total assets). - Structural characteristics (size, tangibility, cashflow generation ratio) kept at 2019 values; ROA and leverage projected dynamically given macro-scenarios. - Macrofinancial determinants: annual real GDP growth and an index of financial conditions (Spread of local currency sovereign debt, spread of US dollar sovereign debt, spread of US dollar corporate debt, equity prices, exchange rate, house prices). ### Estimation details and sample - Dynamic OLS regressions for ROA and leverage estimated for 2016–2020, pooling corporates and SMEs after trimming extremes (10 percent of distribution eliminated; 5 percent in each tail). - Tests indicate coefficient differences between corporates and SMEs: ROA specification — size coefficient differs; leverage specification — coefficients for lagged leverage and turnover differ. SME-specific coefficients incorporated for those variables. ### Key empirical findings - Persistence - Leverage: coefficient close to 0.8 on the lagged dependent variable. - ROA: lower degree of persistence than leverage (shocks to profits more immediate; shocks to leverage more persistent and feed through over time). - Cross-sectional correlations (after controlling for country and industry fixed effects and persistence) - Less profitable and larger firms tend to be more indebted. - No statistically significant effects for turnover and tangibility on leverage in general results (though country-specific results for Colombia show firms with a larger share of fixed assets tend to be less profitable). - Profitability (ROA) strongly positively correlated with real GDP growth. - Leverage not correlated with the index of financial conditions in the Colombia sample. - Probit results (cash shortage and ICR risk) - Likelihood that a firm does not generate enough earnings to pay expenses without additional borrowing: - Negatively correlated with profitability (ROA). - Positively correlated with leverage, tangibility, turnover, and size. - Higher GDP growth reduces this likelihood. - Probability that a firm is unable to pay interest from earnings (ICR < 1): - Negatively correlated with profitability, turnover, and tangibility. - Positively correlated with leverage. - Negatively correlated with GDP growth. - Out-of-sample / ex post prediction - The ex post prediction for 2020 provides results close to actual values. ### Colombia: Actual and Projected Average Values for 2020 - Corporates - ROA: Actual 2.1; Projected 2.2 - Leverage (debt to assets): Actual 19.1; Projected 20.4 - ICR < 1 (share): Actual 0.39; Projected 0.37 - Cash availability <0 (share): Actual 0.34; Projected 0.34 - SMEs - ROA: Actual 2.7; Projected 3.1 - Leverage (debt to assets): Actual 18.0; Projected 18.5 - ICR < 1 (share): Actual 0.25; Projected 0.23 - Cash availability <0 (share): Actual 0.30; Projected 0.31 ### Colombia: Estimation Results (Appendix VII. Table 2) - Dependent variables and reported coefficients (standard errors in parentheses; clustered by industry-year). Significance: *** p<0.01, ** p<0.05, * p<0.1. - (1) ROA - ROA (t-1): 0.508*** (0.0338) - Leverage (t-1): -0.0132*** (0.00247) - Tangibility (t-1): - (not in ROA column) - Size (t-1): 0.0225 (0.0324) - Turnover (t-1): 0.00673*** (0.000864) - GDP growth: 0.342*** (0.0680) - Year2020: 2.912*** (0.733) - Constant: -0.387 (0.728) - Industry fixed effects: Yes - Observations: 32,815 - R-squared: 0.298 - (2) Leverage - ROA (t-1): -0.0573*** (0.00726) - Leverage (t-1): 0.806*** (0.00930) - Tangibility (t-1): -0.00478*** (0.00130) - Size (t-1): 0.243*** (0.0350) - Turnover (t-1): 0.00110 (0.00133) - Financial conditions: 1.640 (1.284) - Year2020: -1.175 (1.185) - Constant: -0.775 (0.696) - Industry fixed effects: Yes - Observations: 30,977 - R-squared: 0.633 - (4) ICR Risk - ROA (t-1): -0.103*** (0.00757) - Leverage (t-1): 0.00343*** (0.000648) - Tangibility (t-1): 0.00384 (0.00290) - Size (t-1): 0.0287** (0.0118) - Turnover (t-1): - (not in this column) - GDP growth: -0.0444*** (0.0171) - Year2020: -0.183 (0.168) - Constant: -0.629** (0.251) - Industry fixed effects: Yes - Observations: 26,101 - (5) Cash Available Risk - ROA (t-1): -0.0475*** (0.00565) - Leverage (t-1): -0.00270 (0.00240) - Tangibility (t-1): -0.00235*** (0.000736) - Size (t-1): 0.0608*** (0.00770) - Turnover (t-1): 0.00107* (0.000643) - GDP growth: -0.0299*** (0.00967) - Year2020: -0.392*** (0.0891) - Constant: -1.148*** (0.143) - Industry fixed effects: Yes - Observations: 32,494 *Italic: Appendix VI. Table 2 and Appendix VII contents from the source PDF.* --- _Source: https://www.imf.org/-/media/files/publications/cr/2022/english/1colea2022007.pdf_