## stssaea ## Source details **Canonical URL:** [stssaea](https://www.imf.org/-/media/files/publications/dp/2020/english/stssaea.pdf) ## Other formats - [Markdown version](/-/media/files/publications/dp/2020/english/stssaea.pdf.md) - [Structured JSON version](/-/media/files/publications/dp/2020/english/stssaea.pdf.json) --- ### Executive summary — adoption, methods, disclosure, and TA - Adoption and scope: - Supervisory stress tests conducted in more than half of sub‑Saharan African (SSA) countries, concentrated in western and southern Africa. - The number of individual stress tests grew exponentially since the early 2010s to about 50 in 2017. - Recommendation: all countries should run some form of stress test regardless of the state of financial development. - Risks assessed and methodologies: - Most countries assess credit risk and liquidity risk using straightforward sensitivity analyses based on shock sizes generally in line with historical events. - Market risk (interest rate and exchange rate) and operational risk analyses are less common. - Few countries change shock sizes over time; a few apply lower credit-quality stress than past crisis episodes would suggest. - Few central banks publish macro‑financial linkage assessments; publications focus more on financial‑sector trends than external/domestic transmission risks. - Disclosure and communications: - Disclosure detail mirrors coverage: credit and liquidity disclosure more detailed than market and macro‑financial risk disclosure, with wide cross‑country variation. - Regression analysis finds financial deepening (credit to GDP ratio) has a significantly positive effect on developing stress test frameworks and publishing them. - No significant effect found for size of the economy (GDP), GDP per capita, financial account openness, or exchange rate regime in the reported models. - Disclosure approach should consider state of financial market development, financial literacy, and authorities’ communications capacity; severity of revealed risks should also guide disclosure level. - IMF technical assistance (TA) and capacity building: - Since 2003, IMF delivered more than 100 TA missions solely focusing on stress testing; African countries received more than half and SSA countries more than 40. - Common TA requests: adopting/extending stress testing, scenario design, satellite models, and building supervisory models. - Common TA findings: lack of detailed and consistent data, methodological shortcomings, institutional weaknesses (for example, scenarios formulated without management involvement; infrequent model updating; limited use of stress tests as an early‑warning/supervisory tool). - High demand for IMF TA likely to continue. ### Key analytical conclusions and policy guidance - More-developed financial sectors tend to have more complex and comprehensive stress testing frameworks and communications strategies. - Stress test shock sizes to nonperforming loans (NPLs) and liquidity differ greatly but are largely in line with historical experience. - More sophisticated stress testing approaches should be employed only after addressing data and methodological issues. - The paper provides examples of good practices in macro‑financial and banking sector risk analysis and examples of quality disclosure of stress test outcomes. ### Notable numeric coverage and timing details - Data collection cut-off date: March 31, 2019. - Of the 45 countries in SSA, 24 published stress test results through the cut-off date. - About one-third of SSA countries routinely publish financial stability reports. - All published tests are “top-down”; South Africa additionally requires “bottom-up” tests by banks. - More than half of countries have received TA on stress testing and three‑fourths of those do conduct stress tests. - In about one‑third of countries the survey could not confirm any stress testing being conducted from public sources. ### Disclosure quality indicators (scoring: 0 = non‑existent, 1 = system‑level, 2 = bank‑level) - Credit: Obs 45, Mean 0.93, Std. 0.96, Min 0, Max 2 - Liquidity: Obs 45, Mean 0.87, Std. 0.94, Min 0, Max 2 - Market: Obs 45, Mean 0.33, Std. 0.67, Min 0, Max 2 - Total: Obs 45, Mean 2.13, Std. 2.14, Min 0, Max 6 - Macrofinancial: Obs 45, Mean 0.47, Std. 0.73, Min 0, Max 2 ### Econometric highlights on drivers of stress testing and disclosure - Privately provided credit to GDP (log) associations (coefficients shown with standard errors in parentheses): - Probit Stress test: coefficient 1.362*** (0.479) - Probit ST public: coefficient 1.108*** (0.361) - OLS Information disclosed: coefficient 1.219*** (0.299) - OLS Macro financial: coefficient 0.433*** (0.120) - GDP (log) finding: - OLS Macro financial: coefficient 0.0870* (0.0510) - Regression sample: Observations: 40 in each reported regression; example R-squared values shown include 0.186 and 0.239 for OLS columns. - Interpretation: financial deepening is positively and significantly associated with adoption, publication, and richer disclosure; no robust effect for GDP size, GDP per capita, financial openness, or exchange rate regime; no significant predictors for likelihood of having received TA in models shown. --- ### Coverage, methodologies, and typical outcomes - Geographic and institutional scope: - Stress tests nearly always national; cross‑border spillovers and group‑level assessments for pan‑African banks typically not captured. - IMF (2015) recommendation: establish a Pan‑African Bank Supervisory Oversight Committee to assess spillovers and stress‑test cross‑border banking groups. - Typical methodologies and tools: - Many SSA authorities use simple Excel-based tools (often original/modified IMF Stress Tester) with a one‑period accounting approach that does not model changes in risk‑weighted assets (RWAs). - Common extensions requested: RWAs calculations, multi‑period setups, granular market‑risk decompositions, profit‑and‑loss detail. - Most authorities apply ad hoc shocks to existing NPL stocks or assume migration of a share of performing loans to NPLs (outright percentage increases in NPLs). - Advanced approaches used in a few countries: macro/satellite models (South Africa, Mauritius), reverse stress tests (Eswatini, Uganda), multi‑factor tests (Madagascar, Malawi). - Shock sizing and provisioning: - Selected NPL shock sizes tend to exceed average five‑year changes but sometimes fall short of historical maximum one‑period shocks since 2006. - Nearly half of countries kept shock sizes identical over time; only five of 24 countries both running and publishing stress tests changed shock sizes over time. - Many authorities do not disclose or justify provisioning rates applied to additional NPLs; some assume full provisioning; only two countries (Malawi, Madagascar) assume different provisioning rates across moderate and severe scenarios. - Disaggregation and reporting: - About three‑quarters of publishing countries provide disaggregated credit‑test results (number of banks failing, grouped by size; combined market share of failing banks; total capital shortfall). - For liquidity tests, most countries publish detailed bank‑by‑bank results including number of banks failing minimum liquidity requirements and daily increments. ### Macro‑financial linkages and banking vulnerabilities - Macro‑financial linkage analysis is rarely published though it helps calibrate risk factors and scenario design. - Countries providing detailed external/domestic risk discussion include Kenya, Mauritius, Nigeria, South Africa, Malawi, Namibia, Rwanda, Uganda, Seychelles. - Commonly identified banking vulnerabilities: - Asset quality: elevated NPLs, write‑offs, high credit concentration. - Low solvency/profitability: thin capital buffers, weak operational efficiency. - Liquidity/funding: lack of longer‑term funding, maturity mismatches, high deposit concentration. - Market risk: large net open FX positions. - Other risks: cyclical indicators (credit‑to‑GDP gap, loan‑to‑deposit ratio), market concentration (Nigeria), contagion risk (Madagascar), cybersecurity (Kenya, South Africa). ### Credit concentration and liquidity testing - Credit risk: - Of 24 SSA jurisdictions publishing stress tests through March 31, 2019, all but one run credit risk tests in some form. - A majority (17) assess name concentration risk. - Two‑thirds of countries running credit risk tests also assess concentration by shocking largest one to five exposures. - Liquidity risk: - As common as credit tests; typically direct shocks to deposits (demand, time, savings) over a five‑day horizon with multiple shock sizes. - Most countries use single‑factor deposit outflow shocks; maturity ladders and detailed cashflow run‑offs are rare. - Several countries have introduced Basel III LCR (Mauritius, Rwanda, South Africa, Uganda). - Where historical data published, assumed demand deposit withdrawals exceeded both maximum month‑on‑month percentage drops since 2000 and highest outflow rates during previous five years. - Most countries provide detailed bank‑level liquidity outcomes; system‑level‑only reporting is uncommon. ### Market risk and other tests - Market risk: - About one‑fourth of SSA countries run market‑risk tests (interest rate risk, FX risk). - Approaches diverse: shocks to net interest income, rise in funding costs, gap analysis across maturity buckets, valuation impacts on trading book. - FX tests typically assume depreciation vs. US dollar and other world currencies; approximately three‑quarters of countries running FX tests have non‑floating exchange rate regimes. - Essential information often not reported: which assets/liabilities are shocked, whether net open FX positions shocked, transmission into capital. - Other risks: - Operational risk tested in a few countries (Madagascar, South Africa). - Interbank contagion and cross‑border default tests used in Nigeria, Angola, Madagascar. - Sovereign risk in banks’ holdings of government bonds not analyzed systematically despite high reliance on such investments. --- ### Communications, disclosure strategies, and recommendations - Definitions: - Disclosure = making information publicly available. - Transparency = amount/detail of disclosed information relative to institution’s information set. - Communications = proactive, strategic messaging (press releases, conferences, social media). - Typology and trade‑offs: - Stress test publications typically embedded in FSRs or annual reports; no SSA practice of dedicated stress‑testing reports observed. - International context (BCBS 2017): 62 percent of authorities publish stress test results; 50 percent disclose aggregated results; 19 percent report bank‑specific results. - Theoretical trade‑offs: excessive transparency may reduce private information acquisition or trigger coordination failures/self‑fulfilling runs; disclosure modalities should be adjusted accordingly. - Recommendations on disclosure and communications: - Establish dedicated channel to communicate financial stability issues and stress test outcomes, separate from monetary policy communications. - Tailor disclosure to authorities’ communications capacity, sophistication of banking/financial markets, national media capacity, and population financial literacy. - Two broad disclosure approaches: - More‑advanced markets/media and higher financial literacy: consider higher transparency and proactive communications; publishing institution‑level data may be appropriate. - Less‑developed markets/media and lower financial literacy: greater restraint; publish aggregate impacts and avoid excessive detail. - Communications should reflect stress test outcomes: - If no significant risks: routine disclosure (FSRs, press releases, social media) and explanation of routine corrective actions. - If substantial risks: elevated communications combining problem recognition with a convincing plan to address them; coordination across agencies and political levels may be required. - Tiered communications by audience: - Tier 1 technical: detailed technical material on website for experts. - Tier 2 familiar: blogs, op‑eds, press briefings for business/media. - Tier 3 less familiar: social media, video, radio, infographics for broader public. - Crisis vs. normal times: - Crisis‑period stress tests should be coordinated with crisis response and crisis communications; a dedicated stress‑test publication and planned media campaign often appropriate. - Normal‑time stress tests typically disclosed as part of FSRs, but generating media interest may be challenging. --- ### IMF TA sequencing, data and institutional guidance, and country examples - Typical TA sequence for implementing/upgrading stress testing: 1. risk identification; 2. assessing data availability and needs; 3. calibration of risk factors; 4. identification of shock transmission channels; 5. quantification of shocks; 6. estimation of satellite models; 7. designing balance sheet models; 8. recommendation for policy use of stress testing models; 9. assessment of staff availability and resource constraints. - Data issues and TA recommendations: - Data challenges: limited collection on banks’ exposures, liquidity, market, operational risk; limited private source data or central credit registers; macro data gaps; inconsistent reporting. - Typical TA recommendations: automated data reporting with validation rules; start collecting real estate prices, loan write‑offs and recoveries, banks’ cross‑border exposures, financial situation of debtors; check collected data for consistency; explore credit registry use. - Improved data enables PD/LGD estimation, loan transition matrices, and adequacy checks of loan‑loss provisioning. - Methodological and institutional issues and guidance: - Stress parameters often ad hoc; lack clear separation between baseline and shock scenarios; parameters often unchanged over time. - Recommend a simple macro stress testing model requiring minimum data and credible for policy use; pursue advanced modeling only after data and capacity improvements. - Institutional recommendations: designate framework ownership, coordinate across supervision/research/statistics, involve management in scenario formulation, use results in supervisory/risk‑based processes, prioritize riskier banks given staff constraints. - Communications capacity: - IMF Logical Framework used to strengthen financial stability communications at institutional, national, and supranational levels. - Mozambique example: IMF helped develop a Logical Framework, held a financial stability communications workshop for Bank of Mozambique in 2017, and provided operational advice on communicating bank distress. - Namibia staged upgrading example: - IMF TA phases during 2013–18 moved Namibia from simple sensitivity tests to a state‑of‑the‑art scenario‑conditional dynamic balance sheet model with multi‑year conditionality and satellite NPL models. --- ### Findings, gaps, and final outlook - Progress and limitations: - Stress testing in SSA has advanced but simple methods dominate; macro‑financial assessment is uneven; results are sparingly used; communications are concise in many cases. - Majority of SSA authorities now conduct stress tests for credit and liquidity risk to assess bank resilience and inform the public. - Fewer countries conduct rigorous macro‑financial linkage analysis; straightforward sensitivity checks dominate. - Severity of assumed shocks is broadly appropriate; single‑factor tests can convey realistic bank performance under stress. - Few countries change shock sizes over time, risking procyclicality in scenario design. - Use and heterogeneity: - Stress tests are routinely used by many central banks but often do not feed into supervisory processes or policy instruments. - Heterogeneity: more‑developed financial systems have more complex frameworks; nearly 40 percent of SSA countries do not appear to be using stress test methodologies at all; 15 percent of the remaining countries do not report on the stress tests they run. - Role of TA and outlook: - Technical assistance can help develop and refine methodologies across levels of financial system development. - IMF TA has helped implement stress testing and address data and capacity constraints in about half of the SSA region. - Remaining countries likely to continue seeking IMF TA to adopt or upgrade stress testing and to develop communications strategies for stress test operations and outcomes. *stssaea - conclusions and an outlook on further development of stress testing in SSA.* ### Executive Summary ������������������������������������������������������������������������������������������������������ ### Executive Summary ### Adoption and scope of stress testing in sub-Saharan Africa - Supervisory stress tests are conducted in more than half of sub-Saharan African countries, particularly in western and southern Africa. - The number of individual stress tests has grown exponentially since the early 2010s. - The paper recommends that all countries run some form of stress test regardless of the state of financial development. ### Types of risks assessed and methodological patterns - Most countries assess credit risk and liquidity risk, typically using straightforward sensitivity analyses based on shock sizes generally in line with historical events. - A few countries apply lower stress to credit quality (that is, increases in the stock of nonperforming loans) than past crisis episodes would suggest. - Few countries change the shock size over time. - Analysis of market risk (that is, interest rate and exchange rate risk) and operational risk are less common in sub-Saharan African countries. - Few central banks publish assessments of macro-financial linkages; the focus leans more toward discussing trends and weaknesses within the financial sector than on outside risks that may negatively affect its performance. ### Disclosure, communications, and determinants of publication - The level of detail in communications mirrors coverage: disclosure for credit and liquidity risk is more detailed than for market risk and macro-financial risk assessment, with wide variations across countries. - A regression analysis finds that financial deepening (as measured by the credit to GDP ratio) has a significantly positive effect on developing stress test frameworks and publishing them. - No significant effect was found for size of the economy (using GDP), economic development (GDP per capita), financial account openness, or exchange rate regime. - The paper concludes that the authorities’ disclosure and communications approach should consider the state of financial market development and literacy as well as authorities’ communications capacity, which may suggest limited or more aggregated disclosure of stress test outcomes where constraints still exist. - The decision of whether and what level of detail to publish should also take the severity of financial stability risks revealed by the stress tests into account. ### Technical assistance (TA) by the IMF and capacity building - Taking stock of IMF technical assistance for stress testing since 2003, most recipient authorities asked for help with adopting or further extending the stress testing tool for their own reporting, designing stress test scenarios, and creating satellite models given lack of specific expertise. - Many TA missions found a lack of detailed data inhibiting more sophisticated stress testing as well as issues with the quality of existing data, particularly consistency of reported data. - Missions also found methodological and institutional shortcomings, including: - insufficient differentiation of stress test scenarios or their formulation without management involvement; - infrequent updating of models; - lack of use of stress test results as an early warning instrument to address credit and liquidity risks in a risk-based supervisory process. - The paper gives a recent example of IMF technical assistance for strengthening communications capacity, notably in financial stability communications. - High demand for IMF technical assistance will likely continue for the remaining sub-Saharan African countries striving to adopt stress testing and for those already conducting stress testing and aiming to upgrade their approaches. ### Key analytical conclusions and policy guidance - More-developed financial sectors tend to have more complex and comprehensive stress testing frameworks and communications strategies. - Stress test shock sizes to nonperforming loans (NPLs) and liquidity differ greatly across countries but are largely in line with historical experience. - More sophisticated stress testing approaches should be employed only by authorities that have managed to address, including with IMF assistance, data and methodological issues common in the region. - The paper provides examples of good practices in analyzing macro-financial and banking sector risks and examples of quality disclosure of stress test outcomes. *Source: Executive Summary, STRESS TESTING IN SUB-SAHARAN AFRICA* ### conclusions and an outlook on further development of stress testing in SSA. ### conclusions and an outlook on further development of stress testing in SSA. ### Overview of stress testing adoption and practice - Stress testing in SSA is often grounded in ad hoc single-factor shocks rather than in comprehensive macro-financial risk assessment. - Survey coverage and timing: - The latest-available publications used in the survey were in most cases published during 2017–18 (a few in 2016). - The data collection cut-off date was March 31, 2019. - Of the 45 countries in SSA, 24 published stress test results through the cut-off date. - About one-third of SSA countries routinely publish financial stability reports. - Growth in practice: - Authorities started adopting stress testing and reporting in the early 2010s. - The number of stress tests (credit, market, liquidity) grew exponentially to about 50 in 2017. - All published tests are “top-down”; South Africa additionally requires “bottom-up” tests by banks. - Disclosure and technical assistance: - In slightly more than half of the countries stress tests are disclosed in publicly accessible documents. - More than half of the countries have received technical assistance (TA) on stress testing and three-fourths of those do conduct stress tests. - In about one-third of countries the survey could not confirm any stress testing being conducted from public sources. ### Coverage, methodology, and outcomes of stress tests - Geographic and institutional scope: - Stress tests nearly always take a purely national perspective; cross-border spillovers and group-level assessments for pan‑African banks are typically not captured. - IMF (2015) recommendation: establish a Pan-African Bank Supervisory Oversight Committee to assess spillovers and stress-test cross-border banking groups. - Typical methodologies and tools: - Many SSA countries use simple, Excel-based tools (often original/modified IMF Stress Tester) with a one-period accounting approach that does not model changes in risk-weighted assets (RWAs). - Common extensions requested: RWAs calculations, multi-period setups, more granular market-risk decompositions, and profit-and-loss detail. - Most authorities apply ad hoc shocks to the stock of existing NPLs or assume migration of a share of performing loans to NPLs; common assumption is an outright percentage increase in NPLs. - A few countries use advanced approaches: macro/satellite models (South Africa, Mauritius), reverse stress tests (Eswatini, Uganda), multi-factor tests (Madagascar, Malawi). - Shock sizing and provisioning: - The selected size of NPL shocks tends to exceed average five-year changes but sometimes falls short of historical maximum one-period shocks since 2006. - Only a few countries explicitly link shock sizes to past shocks (for example, Rwanda). - Nearly half of countries kept shock sizes identical over time; only five of 24 countries both running and publishing stress tests changed shock sizes over time. - Many authorities do not disclose or motivate the provisioning rates applied to additional NPLs; some assume full provisioning; only two countries (Malawi, Madagascar) assume different provisioning rates across moderate and severe scenarios. - Disaggregation and reporting of results: - About three-fourths of publishing countries provide disaggregated credit-test results (e.g., number of banks failing the test, grouped by size; combined market share of failing banks; total capital shortfall). - For liquidity tests, most countries publish detailed bank-by-bank results including number of banks failing minimum liquidity requirements and daily increments. ### Macro-financial linkages and banking sector vulnerabilities - Macro-financial risk assessment: - The IMF mainstreamed macro-financial linkage analysis in surveillance; such assessment helps calibrate risk factors, shock sizes, and scenario design for stress tests. - Few supervisory authorities in SSA publish macro-financial risk assessments; emphasis in publications is often on financial-sector trends rather than external/domestic risks and transmission channels. - Examples of countries providing detailed external and domestic risk discussion: Kenya, Mauritius, Nigeria, South Africa, Malawi, Mauritius, Namibia, Rwanda, Uganda, Seychelles. - Banking sector vulnerabilities commonly identified: - Asset quality issues: elevated NPLs, write-offs, high credit concentration. - Low solvency/profitability: thin capital buffers and weak operational efficiency. - Liquidity and funding issues: lack of longer-term funding, maturity mismatches, high deposit concentration. - Market risk: unbalanced foreign currency exposures (large net open FX positions). - Other risks: cyclical risks (credit-to-GDP gap, loan-to-deposit ratio, household debt indicators), market concentration (Nigeria), contagion risk from interconnected banks (Madagascar), and cybersecurity (Kenya, South Africa). ### Credit risk and credit concentration - Prevalence: - Of the 24 SSA jurisdictions that published stress test results through March 31, 2019, all but one run tests for credit risk in some form. - A majority (17) assess name concentration (impact of deterioration of exposures to the very largest clients). - Typical concentration tests: - Two-thirds of countries running credit risk tests also assess credit concentration risk, typically by shocking default/deterioration of the largest one to five exposures and measuring capital impact. - Sectoral concentration tests are less common but used in some countries (Nigeria, South Africa). - Disaggregation: - All countries running credit concentration stress tests disclose disaggregated results (e.g., number of banks failing; identification of affected banks by size). ### Liquidity risk - Prevalence and typical design: - Liquidity stress testing is as common as credit stress testing and typically involves direct shocks to deposits (demand, time, savings), commonly over a five-day horizon with multiple shock sizes. - Most countries use single-factor deposit outflow shocks; detailed cashflow run-off approaches and maturity ladders are rare. - Several countries have introduced Basel III Liquidity Coverage Ratio (LCR) (Mauritius, Rwanda, South Africa, Uganda). - Appropriateness of shocks: - In countries publishing shock sizes with historical data, assumed withdrawals of demand deposits exceeded both the maximum month-on-month percentage drop since 2000 and the highest outflow rate during the previous five years. - Reporting: - Most countries provide detailed bank-level liquidity test outcomes; system-level-only reporting is uncommon for liquidity tests. ### Market risk - Prevalence: - About one-fourth of SSA countries run market-risk stress tests (interest rate risk, foreign currency risk). - Interest rate risk approaches: - Diverse: outright shock on net interest income, assumed rise in funding costs, gap analysis across maturity buckets (e.g., Namibia, Nigeria, Tanzania), or valuation impacts on trading book securities. - FX risk approaches and reporting gaps: - Most authorities assume depreciation of the national currency against the US dollar and other world currencies; in some cases (Burundi) indirect credit effects of depreciation are assessed. - Approximately three-fourths of countries running FX stress tests have non‑floating exchange rate regimes (managed/stabilized or pegs). - Essential information often not reported: which assets/liabilities are shocked, whether net open FX positions are shocked, and the transmission mechanism into capital. ### Other risks tested - Operational risk and other tests: - A few countries test operational risk impacts (Madagascar: destruction of share of financial assets; South Africa: cyber risks). - Interbank contagion and cross-border default tests are employed in some countries (Nigeria, Angola, Madagascar). - Notable omission: - Sovereign risk in banks’ holdings of government bonds is not analyzed systematically despite high reliance on such investments in many jurisdictions. ### Disclosure, transparency, and communication practices - Definitions: - Disclosure: making information available to the public. - Transparency: amount and detail of information disclosed relative to the institution’s overall information. - Communications: proactive, strategic messaging (press releases, conferences, social media). - Typology of publications: - Stress test publications appear embedded in FSRs or annual reports; no SSA practice of dedicated stress-testing reports was observed in the survey. - Trade-offs and international context: - Global survey (BCBS 2017): 62 percent of authorities publish stress test results; 50 percent disclose aggregated results; 19 percent report bank-specific results. - Theoretical trade-offs: too much transparency may crowd out private information acquisition or trigger coordination failures/self-fulfilling runs; supervisors may therefore adjust disclosure modalities. - Current SSA practices: - Proactive communications on stress testing/financial stability are limited in SSA. - The press in SSA shows interest in banks’ condition; there is room to strengthen financial stability communications and separate them from monetary policy communications. - Quality indicators and findings: - Disclosure quality scoring (0 = non-existent, 1 = system-level, 2 = bank-level): - Credit: Obs 45, Mean 0.93, Std. 0.96, Min 0, Max 2 - Liquidity: Obs 45, Mean 0.87, Std. 0.94, Min 0, Max 2 - Market: Obs 45, Mean 0.33, Std. 0.67, Min 0, Max 2 - Total: Obs 45, Mean 2.13, Std. 2.14, Min 0, Max 6 - Macrofinancial: Obs 45, Mean 0.47, Std. 0.73, Min 0, Max 2 ### Econometric findings on drivers of stress testing and disclosure - Regression highlights (sample sizes and model types noted in source): - Priv credit to GDP (log) findings: - Probit Stress test: coefficient 1.362*** (0.479) - Probit ST public: coefficient 1.108*** (0.361) - OLS Information disclosed: coefficient 1.219*** (0.299) - OLS Macro financial: coefficient 0.433*** (0.120) - GDP (log) findings: - OLS Macro financial: coefficient 0.0870* (0.0510) - Other coefficients in probit/OLS models not statistically significant as presented. - Constant terms shown (for example, Probit Stress test constant 27.583* (4.002)). - Observations: 40 in each reported regression; R-squared reported for OLS columns (for example, R-squared 0.186 and 0.239 as shown). - Interpretation: - Financial deepening (privately provided credit to GDP) is positively and significantly associated with adoption of stress test frameworks and publishing them, and with better disclosure and macro-financial analysis. - No robust effect found for size of the economy (GDP), GDP per capita, financial account openness, or exchange rate regime on these outcomes in the reported models. - No significant predictors found for the likelihood of having received TA on stress testing in the models shown. ### Recommendations and examples of good practices - Broad recommendation: - Conduct stress testing in all countries, even if only for internal purposes, tailored to the complexity of the financial system. - Key elements of good practice identified: - Macro-financial risk analysis examples: - Eswatini: spider web chart of seven risk categories. - Namibia: detailed assessment of external developments’ impact on banking (for example, monetary tightening in neighboring South Africa). - Rwanda (2016–17 FSR): box on external shocks and domestic shocks and their likely impact on the banking sector. - South Africa: risk assessment matrix feeding into macro-financial assessment. - Analysis of bank vulnerabilities/systemic risks: - Cabo Verde: flags maturity mismatches, deposit and credit concentration, peer comparisons. - Lesotho and Tanzania: use of indicators for cyclical/credit and liquidity risks. - Uganda: detailed bank-specific systemic risk analysis across multiple risk types. - Quality of stress test disclosure: - Malawi: detailed disclosure with several credit-risk scenarios, number of banks failing tests, sensitivity analyses, and funding shocks. - Nigeria: system-level and bank-size breakdowns for multiple risk types (moderate and severe scenarios). - Lesotho, Madagascar, Malawi, Namibia: explicit mention of provisioning rates assumed for additional NPLs under stress. - Recommendations on disclosure and communications: - Establish a dedicated channel to communicate financial stability issues and stress test outcomes, separate from monetary policy communications. - Tailor disclosure and transparency to: - authorities’ communications capacity, - sophistication of banking system and financial markets, - national media capacity, - financial literacy of the population. - Two broad country categories and corresponding disclosure approaches: - More-advanced financial markets/media and higher financial literacy: consider higher transparency and proactive communications; publishing institution-level data may be appropriate. - Less-developed financial markets/media and lower financial literacy: exercise greater restraint; publish aggregate impacts and avoid excessive detail that could be misreported or misunderstood. - Communications modalities should reflect stress test outcomes: - If no significant risks uncovered: routine disclosure (FSRs, press releases, social media) and explanation of routine corrective actions. - If substantial risks uncovered: elevated communications combining problem recognition with a convincing plan to address them; coordination across agencies and political levels may be required. - Emphasize tiering communications by content and channels: - Tier 1 technical (detailed technical material on website) for experts. - Tier 2 familiar (blogs, op-eds, press briefings) for business/media. - Tier 3 less familiar (social media, video, radio, infographics) for the broader public. - Crisis vs. normal-time distinctions: - Crisis-period stress tests should be closely coordinated with crisis response and crisis communications practices; a dedicated stress-test publication and a planned media campaign are often appropriate. - Normal-time stress tests are typically disclosed as part of FSRs, but generating media interest may be challenging. *stssaea - conclusions and an outlook on further development of stress testing in SSA.* ### Box 1. Communications on Stress Testing: Crisis Response versus Normal Times ### Box 1. Communications on Stress Testing: Crisis Response versus Normal Times ### Communications practices and principles - Ahead of presentation, technical experts are available to respond to questions from the press under embargo. - Example: For the November 2018 edition, a media advisory from the Treasury flagged the upcoming publication and associated media opportunities, as per established practice. - Results were presented concisely in the FSR—both in the executive summary and in the main body—with the stress tests methodology explained in detail in an annex. - Media coverage focused on key risks to financial stability as identified in the FSR, even though the stress tests were not discussed. - Key communication principle: - At all times, it is important for communications to explain that scenarios used for stress testing are hypotheticals, not forecasts. ### IMF Technical Assistance (TA) for stress testing: scope and uptake - From 2003 to the first quarter of 2019, the IMF delivered more than 100 TA missions solely focusing on stress testing, many of which to SSA countries. - African countries received more than half of such missions, and SSA countries more than 40. - Most of the countries in the region requesting such TA received more than one mission. - IMF also provided technical support via regional initiatives, such as within the East African Community (EAC). - SSA authorities consider stress testing an important tool to enhance risk analysis frameworks, including when moving to risk-based supervision. - TA focus in SSA differs from other regions: supervisory stress testing is emphasized in SSA, whereas other regions (Europe, Asia) often use advanced methodologies (for example, network/contagion analysis) and macroprudential stress testing. - Countries request TA for: - adoption or further extension of stress testing for reporting; - help with scenario design and satellite models; - building models suitable for supervisory purposes given transitions to Basel II/III or other supervisory changes. - Ideal model sought by authorities enables comprehensive stress testing of various risks: credit quality and concentration risk, market risk, liquidity risk, contagion risk, and other risks affecting bank income and profitability. - Tests can be run using a single-factor (sensitivity analysis) or combination of factors (scenario analysis). - TA missions focused on creating satellite models when authorities lacked econometric skills; sometimes local efforts were prompted by IMF activity (for example, FSAP in Mauritius). ### Typical TA sequence of steps for implementing/upgrading stress testing frameworks - (1) risk identification; - (2) assessing data availability and needs for stress testing; - (3) calibration of risk factors; - (4) identification of shock transmission channels; - (5) quantification of shocks; - (6) estimation of satellite models; - (7) designing balance sheet models for calculation of impact; - (8) recommendation for policy use of stress testing models; - (9) an assessment of staff availability and resource constraints. ### Data issues identified by TA missions - Many SSA central banks and supervisors have adequate data for simple sensitivity and scenario-based stress tests but face challenges assembling data for genuine macroprudential stress testing. - Data availability challenges stem from limited data collection on banks’ balance sheet exposures and on liquidity, market, and operational risk. - Limited availability of data from private sources or central credit registers hinders risk analysis and model-building efforts. - Acute data limitations encountered in macro data; banks’ balance sheets; and information about credit, market, and liquidity risk. - Even when data exist, quality and consistency remain a challenge: old reporting systems, manual submissions, inconsistent data (for example, sum of large exposures larger than total exposures, provisioning ratios exceeding 100 percent). - Typical TA recommendations for data improvement: - employ automated data reporting systems with data point validation rules flagging inconsistent reporting; - start collecting data on real estate prices, loan write-offs and recoveries of written-off loans, banks’ cross-border exposures, and the financial situation of bank debtors; - check collected data for consistency; - explore use of credit registry data. - Collecting such data would allow central banks and supervisory authorities to do more advanced credit risk analysis (such as the calculation of probability of default and loss given defaults, and loan transition matrices), estimate imbalances in housing and commercial real estate markets that may give rise to credit risk, and check the adequacy of loan-loss provisioning. - Supervisory authorities need to cooperate with other entities: financial institutions, statistical agencies, and private sector firms (for example, private credit bureaus, real estate agencies). ### Methodological and institutional issues - Stress parameters are often formulated on an ad hoc basis; no clear separation may exist between the baseline and hypothetical shock scenarios. - TA missions emphasized need to develop a simple macro stress testing model that requires minimum data and is credible enough for policy use—particularly for credit risk—whereas liquidity risk models are already straightforward. - Little modeling is used to assess income risk. - Trade-offs noted: - More-advanced models (for example, multi-period, dynamic balance sheet models) may better represent reality but are subject to multiple model-related uncertainties and require a range of assumptions. - Upgrading models must be assessed against potential use, reliability, and authorities’ capacity to maintain and update them. - Human capital constraints: SSA central banks may lack enough staff with relevant modeling skills to periodically update advanced models. - Consequently, most TA missions recommended establishing a basic stress test framework that staff can operate and use to inform decision makers about idiosyncratic/systemic risks. - Institutional deficiencies found: - stress test scenarios are formulated by technical staff and usually not discussed with management before the exercise; - stress test results may be used for monetary policy purposes but typically do not influence the supervisory process; - stress tests are not used as an early warning instrument to address identified credit and liquidity risks nor to support risk-based supervision by focusing on more risky banks. ### Communications capacity: IMF approach and country example - IMF takes a holistic approach to strengthening communications capacity, drawing on the IMF’s Logical Framework for strengthening financial stability communications at the institutional, national, and supranational levels. - The framework covers communications (and coordination where needed) by authorities responsible for financial stability: central banks, supervisory authorities, deposit insurance funds, and ministries of finance. - There is value in building communications capacity in parallel with policy and analytical capacity; communications capacity should not fall behind and ideally should be somewhat ahead of policymaking to respond to unexpected developments. - Mozambique example: - As part of a central bank modernization project begun in 2017, the IMF Communications Department developed a Logical Framework for building capacity to communicate on financial stability policies at a national level. - Key workstream: build capacity to communicate on financial stability policies in parallel with development of a macroprudential department, a financial stability report, and stress testing capacity. - A workshop on financial stability communications was held at the Bank of Mozambique (BoM) in 2017. - IMF provided operational advice to the BoM on strengthening financial stability communications in bank distress to help preserve central bank credibility. - Few SSA TA missions have focused on communications of supervisory components; such capacity development would include: - communicating stress test results to the public; - explaining use of stress tests for policy purposes, notably microprudential and macroprudential policy; - explaining the use of stress tests in day-to-day supervision, for example when setting capital and liquidity requirements. ### Namibia—example of staged methodology upgrading (Box 2 summary) - Namibia progressed from a simplistic approach with limited publication to coherent single-factor sensitivity tests with more detailed reporting and ultimately to a state-of-the-art macro stress testing model (scenario-conditional dynamic balance sheet structure). - IMF TA was delivered in three phases during 2013–18: - 2013–14: TA provided options for solvency and liquidity stress testing and addressed data issues. IMF suggested focusing on migration of performing loans to nonperforming loan (NPL) status and adjusting risk weights rather than PD/LGD given Basel II Standardized Approach usage. Bank of Namibia (BoN) conducted stress tests and published abridged results in its FSR. Mission noted BoN difficulty retaining qualified staff and suggested staff rotation and internal documentation for business continuity. - Early 2015: mission performed solvency and liquidity stress tests and provided more training in the IMF Stress Tester tool. - Mid-2015: TA advised on model-based stress testing linking risk factors to macro-financial variables; results tentative due to limited data (for example, short NPL time series). - 2016–2018: BoN published more-detailed stress test results—2016: credit risk; 2017/18: market and liquidity risk. An IMF FSAP mission evaluated the framework in late 2017. - 2018: mission implemented a customized solvency stress model with a simple dynamic balance sheet structure allowing gross credit stocks to grow in projections and a multiyear scenario-conditionality of up to five years; model core includes a scenario-conditional credit risk module (satellite models for NPL ratios at bank level) and a scenario-conditional market risk module for trading and banking book (including interest rate risk in the banking book). ### Findings, gaps, and conclusions - Stress testing in SSA has advanced but limitations remain: simple methods dominate, macro-financial assessment is uneven, stress test results are sparingly used, and communications are concise in some cases. - Majority of SSA authorities now conduct stress tests for a range of banking risks, most prominently credit and liquidity risk, to assess bank resilience and inform the public. - Fewer countries conduct proper assessment of macro-financial linkages and banking sector vulnerabilities; straightforward sensitivity checks dominate. - Severity of assumed shocks to risk factors is found to be broadly appropriate; single-factor stress tests can convey a realistic picture of bank performance under stress. - Few countries change the size of shocks over time, which may lead to procyclicality in scenario design (for example, keeping shock sizes unchanged in a recession). - The routine use of stress tests by many SSA central banks indicates elevated importance, but results do not necessarily feed into supervisory processes or other policy objectives. - Evidence of heterogeneity: - more-developed financial systems tend to have more complex and comprehensive stress testing frameworks and communications strategies; - nearly 40 percent of SSA countries do not appear to be using stress test methodologies at all; - 15 percent of the remaining countries do not report on the stress tests that they run in their official publications. - Technical assistance can help develop or refine stress testing methodologies across all levels of financial system development. - IMF TA has helped implement stress testing techniques and address data and capacity constraints in about half of the SSA region. - Remaining countries are likely to continue seeking IMF TA to adopt or upgrade stress testing; TA can also be instrumental in developing a strategy for communicating stress test operations and outcomes. *Source: Box 1. Communications on Stress Testing: Crisis Response versus Normal Times — from the IMF PDF "stssaea - Box 1. Communications on Stress Testing: Crisis Response versus Normal Times".* ### 1. Risk identification. Except for a few cases (for example, South Africa), ### 1. Risk identification. Except for a few cases (for example, South Africa), ### Risk identification - Financial systems in most SSA countries are “plain vanilla,” with banks as the main providers of credit and limited intermediation by nonbank financial institutions (NBFIs). - Risk identification therefore focuses on risks to banks (for example, exposure to specific economic sectors).1 - Banks are mostly funded by retail deposits but may face concentration risk from large institutional depositors (pension funds, public entities). - Cross-border funding and interbank market activity are often low. - Countries tend to be reluctant to properly identify risks stemming from public sector entities, banks’ exposures to bonds of their own sovereign, as well as state-owned enterprises.2 ### Data availability and needs for stress testing - TA missions recommend collecting additional data where gaps exist in: - Global data (for example, commodity prices). - Supervisory data. - Macroeconomic data (for example, aggregate balance sheets of corporates/households, cross exposures between sectors). - Real estate data. - Technical assistance missions encountered acute data limitations in: - Macro data and availability of macro forecasts; asset price information (for example, construction, commercial/residential real estate, land) is often not available or is lagged.1 - Credit risk: credit risk concentration (by sector, single borrowers); macroprudential information (for example, loan-to-value, debt-service-to-income, collateral); NPLs by currency and written-off/restructured/cured loans; cure rates of NPLs. - Balance sheet information: banks’ compliance with provisioning requirements, collateral valuation, sectoral composition of loans not regularly checked for consistency. - Market risk: duration of securities held by banks; exposures to equities by company; sensitivity to interest rate changes (for example, parallel shift in yield curve); repricing gap; behavioral data on loan prepayments and deposit mix. - Liquidity risk: cash flow/maturity ladder data (contractual and behavioral), cash flow data by significant currencies in financially dollarized systems; asset encumbrance; largest funding sources and providers (funding concentration reports). - Other data: nonbank financial sector data and data about banks’ borrowers (corporates, households). - Even where banking sector data exist, data for firms, households, and NBFIs are often missing; e.g., no accurate estimate of household debt ratios due to lack of disposable income and debt data. ### Risk factor calibration (scenario design) - Systemic financial crises have been rare in SSA as of late (Mlachila, Park, and Yabara 2013),2 making historical calibration of shocks to anchoring variables (GDP, unemployment, interest rates) challenging. - Scenarios are often designed by applying a reasonable shock to NPLs (for example, largest historic annual/quarterly increase in NPLs overall/by sector), thereby bypassing transmission channels, shock quantifications, and satellite models.3 ### Transmission channels - Identification of shock transmission channels is crucial to build the narrative of how shocks affect banks via exposures to borrowers. - Purpose of stress tests: not only quantify losses and impact on bank capital, but assess how shocks affect borrowers and then banks. ### Shock quantification - Aim: produce macro-financial variable paths consistent with initial shocked risk factors (GDP growth, unemployment, commodity price shocks). - Only a few SSA countries have macro forecasting models suitable for internally-consistent macro scenarios (for example, vector autoregression or structural models). - Limitations stem from lack of reliable and sufficiently-long time series and econometric skills. - Expert judgment may suffice for sensitivity analysis but may not produce consistent variable paths. ### Satellite models - Satellite models link macro-financial variables with risk indicators (for example, NPLs or probability of default, provisioning) and profit-and-loss items (for example, interest income and expenses, fees and commissions, administrative costs). - TA missions advise against designing/maintaining satellite models when data and capacity limitations make maintenance difficult. ### Balance sheet models - Final step: design a balance sheet model to calculate income, losses, changes in balance sheet positions, risk weights and capital adequacy. - Users of the Stress Tester tool requested modifications and extensions, including multi-year stress testing capability. ### Policy use - Some TA missions provided recommendations on the actual and prospective use of stress testing. - SSA countries have used stress testing more for monetary policy3 than for micro-/macroprudential purposes. - Senior management has questioned supervisors’ ability to use stress test results to force banks to increase capital and/or to de-risk. - Using stress tests for macroprudential policy presents larger challenges in SSA due to lack of adequate macroprudential policy frameworks, such as decision-making bodies, legislation, and experience in systemic risk identification and prevention.4 - Evidence indicates stress test results are often not used as a basis for formal discussions with banks, do not affect inspection schedules or supervisory priorities, and rarely lead to extra capital requirements (for example, countercyclical capital/systemic risk buffer). - TA reports emphasize using stress tests as an early warning tool to: - Identify banks with inadequate capital (based on expected losses). - Force preparation of plans to increase capital buffers or to de-risk (decrease risk-weighted assets). - Engage in risk-based supervision by focusing on more risky banks.1 ### Staff availability and resource constraints - A small number of staff working on stress testing and risk assessment hinders full utilization of TA advice and leads to multiple follow-up TA requests.5 - High staff turnover because stress testers are reassigned or leave central banks due to uncompetitive remuneration. - Other resource constraints: software for econometric modeling, programming experience, and limited allocation of time for nontraditional projects. ### Methodological and institutional issues - Methodological issues noted by TA missions: - Stress parameters often ad hoc; no clear separation between baseline and shock scenarios. - Scenario parameter values often unchanged over time; same scenarios reused for multiple years (five or more) from last IMF mission (for example, FSAP, TA). - Currency risk stress tests need improvement in financially dollarized economies; tests often use net open position and largest historical monthly shock without assessing indirect credit-risk effects from unhedged borrowers. - Liquidity stress tests are simple (assume deposit withdrawals over a timespan, apply outflow rates to deposit stocks, compare liquid assets to deposit outflow). - Income risk tests are rare and simple; typically assume ad hoc decreases in yields/interest income without modeling behavioral responses (loan prepayments, deposit shifts) or including fees, commission income, trading income. - Institutional issues: - Stress testing work is distributed across banking supervision and research/statistics departments, creating coordination issues. - Need clarity on who is proprietor of the framework, who produces scenarios, who collects/checks data, and who determines use/publication of results. - Scenarios are usually formulated by technical staff and not discussed with management before the exercise. - Where scenarios are crafted, research provides real sector inputs and financial stability/supervisory staff choose stress factors and shock sizes. - Scenarios can be common across banks or bank-specific. - Formal top-management approval is typically not obtained. - Stress test results are often treated as a research task rather than a supervisory/macroprudential tool. ### Recommendations and practical guidance from TA missions - Develop a simple macro stress testing model that: - Requires minimum data and is credible for policy use. - Uses minimum macro and financial data (for example, NPLs/NPL ratios or provisions/write-offs regressed on real GDP, interest and exchange rates, and, if available, unemployment). - Can be based on single-equation OLS regression or simultaneous equation systems (vector autoregression, vector error correction models); dynamic panel models or Bayesian approaches are occasionally used. - Advance modeling should be pursued only when more complete and reliable data become available. - Improve coordination within central banks/supervisory agencies to designate framework ownership, scenario production, data collection/validation, and use of results. - Prioritize analysis of riskier banks given limited staff resources; focus on developments over time rather than absolute values.1 *Source: IMF staff.* --- _Source: https://www.imf.org/-/media/files/publications/dp/2020/english/stssaea.pdf_