## htnea2021006

## Source details

**Canonical URL:** [htnea2021006](https://www.imf.org/-/media/files/publications/howtonotes/2021/english/htnea2021006.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/howtonotes/2021/english/htnea2021006.pdf.md)
- [Structured JSON version](/-/media/files/publications/howtonotes/2021/english/htnea2021006.pdf.json)

---

### Executive summary: key findings and context
- The coronavirus (COVID-19) crisis is likely to deteriorate banks’ balance sheets in Africa, with the largest threat pertaining to loan portfolios and a likely sharp increase in nonperforming loans (NPLs) in the short to medium term.
- Elevated NPLs generate macroprudential and financial stability risks and impair banks’ ability to support the economy during the recovery by:
  - Raising capital requirements;
  - Denting banks’ net interest margins;
  - Generating service and management costs;
  - Potentially weakening the ability of banks to grant new loans.
- Structural impediments in sub-Saharan Africa to NPL resolution include weak debt enforcement procedures, weak legal rights, and financial infrastructure gaps.
- The note and accompanying Excel template quantify the capital relief and increase in credit capacity from disposing of NPLs (sale to third parties), focusing on the capital relief channel while noting broader micro and macroeconomic benefits (reduced uncertainty, efficiency gains, restored profitability).
- In countries with high provisions (LLR are the cumulation (stock) of provisions over time), regulatory capital released by NPL disposal can be relatively limited because risk-weighted assets and capital requirements are based on net NPLs (NPLs minus LLR).

### NPL resolution strategies and empirical examples
- Broad categories of strategies for NPLs:
  - Keep NPL on balance sheet:
    - Legal enforcement (court, insolvency, liquidation, administrative processes).
    - Consensual solutions (cash settlement, traditional loan workout, out-of-court restructuring).
  - Remove NPL from balance sheet:
    - Write-off.
    - Sale to private entity.
    - Sale to public entity.
    - Sale to mixed public-private entity.
- Empirical examples of NPL disposal approaches and outcomes:
  - Accelerated write-offs:
    - Malawi: new regulation from 2017 led NPL ratio to decline from 15.7 percent at the end of 2017 to 3.6 percent in September 2019 (largely due to write-offs, loan recovery, and overall growth in bank lending).
    - Tanzania: 2018 Bank of Tanzania directive to write-off credit accommodations in loss category for more than four consecutive quarters (previously more than 12 quarters) contributed to NPL ratio decline from 11.5 percent at the end of 2017 to 9.8 percent at the end of 2019.
  - Securitization via a Special Purpose Vehicle (SPV):
    - Nigeria (2010): SPV acquired NPLs with an original book value of N4.02 trillion at a price of N1.76 trillion, equivalent to 1.7 percent of GDP (reflecting a 56 percent haircut). After transfer and securitization, NPL ratio dropped from 38 percent at the end of 2010 to below 5 percent at the end of 2012.
  - Centralized asset management companies (AMCs):
    - Angola (2016): Recredit, a state-owned AMC, purchased distressed assets to free up lending capacity. Recredit purchased NPLs from one bank associated with six large borrowers for a total of Kz480 billion, equivalent to about 3 percent of GDP.

### Analytical focus, framework, and template output
- Purpose and scope:
  - Estimate regulatory capital released by removing NPLs from bank balance sheets (recorded at net book value) and the potential quantum of fresh credit that could be provided from the capital released.
  - The template focuses on the capital relief channel while noting broader benefits and caveats (data quality, heterogeneity).
- Key definitions and equivalences:
  - Haircut (level) = NBV − sale price = capital loss (if positive) or capital gain (if negative).
  - Net book value (NBV) = Gross book value (GBV) − loan loss reserves (LLR).
  - Haircut (level) ≈ unprovisioned loan loss, where unprovisioned loss = total projected loss (NPV) − LLR.
- Model-based unprovisioned loss and loss-under-default (per unit of GBV):
  - Unprovisioned loss per unit of gross NPL = p * (1 − α) + (1 − p) * Loss under default − llr, where llr = LLR / GBV.
  - Loss under default per unit of gross NPL = uncollat/(1 + r)^t + [collat/(1 + r)^t − collat * (1 − d)^t/(1 + r)^t] + m_cost + l_cost.
  - Equivalent rearrangement: Loss under default per unit of gross NPL = 1/(1 + r)^t − [collat * (1 − d)^t]/(1 + r)^t + m_cost + l_cost.
  - Model-based haircut (level) = Unprovisioned loan loss per unit of Gross NPL sold * Gross NPL sold.

### Main simulation steps in the template
- Step 1: Calculation of the Tied-up Capital
  - Tied-up capital = Net NPL sold * (WNPL * reg. CAR%) * (dRWA / dCRWA).
  - Net NPL sold = Actual net NPL − Target net NPL (Net NPL sold = Actual Gross NPL − Gross NPL Target)*(1 − llr) under default template option.
  - Default parameters: reg. CAR% = 12 percent; default WNPL = 100 percent; default WPL = 100 percent.
  - dRWA / dCRWA options:
    - Default: assume other components of RWA fixed ⇒ dRWA / dCRWA = 1.
    - Alternative: assume RWA composition constant ⇒ dRWA / dCRWA = RWA / C RWA.
- Step 2: Calculation of the Capital Relief
  - Scenario 1 (no haircut): Capital relief = Tied-up capital.
  - Scenario 2 (ad hoc haircut ratio θ): Capital relief = Tied-up capital − θ * Net NPL sold.
  - Scenario 3 (model-based haircut): Capital relief = Tied-up capital − Unprovisioned loan loss per unit of Gross NPL sold * Gross NPL sold.
  - Haircut ratio in scenario 3 (percent of net NPL) = Unprovisioned loan loss per unit of Gross NPL sold * Gross NPL sold / Net NPL sold.
- Step 3: Use of the Freed-up Capital to Grant New Loans
  - Additional performing loans = Capital relief * (1 / (WPL * reg.CAR%)) * (dCRWA / dRWA).
  - dCRWA / dRWA options:
    - Default: assume other components fixed ⇒ dCRWA / dRWA = 1.
    - Alternative: dCRWA / dRWA = C RWA / RWA.

### Template scenarios, defaults, and calibration
- Three haircut scenarios simulated:
  - Scenario 1: no haircut (NPLs sold at NBV) — main scenario for cross-country comparison.
  - Scenario 2: fixed haircut (default θ = 10 percent positive haircut ratio).
  - Scenario 3: country-specific, model-based haircuts using the model formulas above.
- Common default assumptions:
  - Banks reduce NPL ratio to a desired level; default assumes NPL ratio is halved relative to its 2018 value.
  - Regulatory CAR default = 12 percent.
  - Default WNPL = 100 percent (alternative 150 percent available).
  - Default WPL = 100 percent (alternative 75 percent available).
  - Default dCRWA / dRWA = 1 (alternative = C RWA / RWA).
- Default calibration for model-based haircut (scenario 3) parameters:
  - Discount rate r = 10 percent.
  - collat = 0.8, uncollat = 0.2.
  - Collateral decay rate δ = 0.05 per year.
  - Management cost m_cost = 0.05.
  - Legal cost l_cost = proxied from World Bank Doing Business “enforcing contracts” indicator; 75 percent of reported value used.
  - Average remaining time to resolution t = sourced from World Bank Doing Business (converted to years and rounded to nearest 0.5 by template).
  - Probability of consensual recovery p = 0.67.
  - Recovery fraction under consensual route α = 0.35.
- Haircut scenario notes and empirical ranges:
  - Default uniform positive haircut = 10 percent, consistent with Jobst, Portier, and Sanfilippo (2015).
  - Historical haircuts in crisis contexts can be very high (close to 80 percent of gross loans); typical ranges noted include 50 percent of gross loans for mortgage/collateralized loans, two-thirds for retail NPLs, or higher for corporate debt in difficult times.
  - Haircuts can be nonexistent or negative in special circumstances (e.g., public purchasing entity, public support for systemically important banks).

### Options for computing provisioning on NPLs sold
- Option 1 (default): country average provision ratio from IMF FSI data: llr = aggregate specific provisions / aggregate gross NPLs.
- Option 2: weighted average of provision rates by NPL category (substandard, doubtful, loss) using World Bank Bank Regulation and Supervision Survey; weights from Fitch Connect; assumed disposal order: loss → doubtful → substandard.
- Option 3: same as Option 2 but disposal order reversed: substandard → doubtful → loss.
- Practical data caveats for Options 2 and 3: missing reporting by banks for NPL categories leads to imputations (remove banks with no data; distribute residuals; use SSA median when country data missing).

### Outputs, sensitivity analyses, and policy experiments
- Main outputs for each country and scenario:
  - Capital released by NPL disposal (capital relief).
  - Additional performing loans that can be generated from capital relief.
  - Descriptive statistics: NPL ratios (net and gross), breakdown by loan type (substandard, doubtful, loss).
  - For scenario 3: country-specific model-based haircuts (sensitive to parameter choices).
- Sensitivity analyses under scenario 3 vary parameters such as collateral depreciation rate, collateralized portion of NPL, probability of consensual resolution, target CAR, and others.
- Example policy experiments the template can simulate:
  - Measures that boost market value of NPLs (develop distressed-asset market, improve collateral valuation/registry, establish specialized NPL collection agencies) by inputting a negative haircut ratio.
  - Targeted disposal strategies that remove legacy (loss) NPLs first by selecting Option 2/3 and adjusting parameters accordingly.
  - Reforms to reduce time and costs of contract enforcement (e.g., lower legal duration by one year) to assess impact on haircuts, capital relief, and additional lending.

### Data issues, interpretation notes, and caveats
- Scenario 3 can produce extreme haircut ratios expressed as percent of net NPLs where net NPL figures are very low due to:
  - Stringent provisioning practices (legacy NPLs fully provisioned but kept on balance sheet because of tax/legal impediments).
  - Statistical issues (provisions reported against both performing and nonperforming loans; slow reclassification of loans causing provisional overstatement relative to recorded NPLs).
- Template results are sensitive to assumptions; model-based haircuts are more informative for policy experiments than absolute quantification.
- Important caveats highlighted:
  - Data quality and differing NPL/provision definitions across countries.
  - Sensitivity of outputs to assumptions (collateralization, collateral decay, haircut rates).
  - Credit scarcity can also stem from fiscal dominance, poor financial market infrastructure, financial inclusion bottlenecks, and lack of bankable projects.
  - The template is based on a stylized macroeconomic framework and cannot fully capture bank-/country-specific legal, regulatory, and portfolio heterogeneity.

### Selected median statistics and figure highlights (as reported)
- SSA median = 0.04
- SSA median gross NPLs = 11.8
- SSA median loan loss provisions = 8.4
- % of initial performing loans (RHS) SSA median = 1.9 (RHS)
- % of GDP, SSA median = 0.4
- SSA median = 0.01
- % of initial performing loans, SSA median = 0.3 (RHS)
- % of GDP, SSA median = 0.1
- SSA median reduction in gross NPLs = 5.9
- SSA median reduction in net NPLs = 1.7
- Chart/figure notes preserved from source:
  - The shock presented in panel charts is a 50 percent decline in the NPL ratios relative to 2018.
  - Chart examples include "3. SSA: Capital Relief from NPLs Disposal, No Haircut (Percent of GDP)" with vertical scale markings 0, 5, 10, 15, 20, 25, 30, 35, 40 and plotted small numeric values including 0.3, 0.0, 0.2, 0.1, 2.5, 0.0, 1.5, 1.0, 2.0, 0.5.
  - "2. SSA: Reduction in NPLs in Percent of Initial Total Gross Loans (Percentage points)" shows a prominent numeric label 14 corresponding to the 50 percent shock.
- Country-specific model-based haircut entries as presented (order and exact values preserved, including formatting anomalies from source):
  - Eswatini 47.4%
  - Seychelles 36.9%
  - Namibia 28.8%
  - Malawi 27.4%
  - Equatorial Guinea 21.0%
  - Central African Rep. 227%
  - Cameroon 229%
  - Gabon 231%
  - Mali 16.8%
  - South Africa 14.0%
  - Niger 9.5%
  - Lesotho 8.0%
  - Nigeria 216.6%
  - Côte d'Ivoire 21.8%
  - São Tomé & Príncipe 217%
  - Kenya 219%
  - Burundi 224%
  - Zambia 225%
  - Uganda 5.2%
  - Ethiopia 5.5%
  - Mauritius 5.5%
  - Madagascar 21.1%
  - Rwanda 21.6%
  - Sierra Leone 23.8%
  - Senegal 24.4%
  - Togo 212.3%
  - Ghana 213.6%
  - Cabo Verde 214.2%
  - Liberia 214.4%
  - Gambia, The 22.5%
  - Benin 3.2%
  - Burkina Faso 2.8%
  - Comoros 1.9%
  - Congo, Republic of 19.7%
  - Angola 21.9%
  - Mozambique 21.9%
  - Burundi 224% (duplicate/format anomaly)
  - Lesotho 8.0% (duplicate)
  - Namibia 28.8% (duplicate)
- SSA Median shown in the figure: 0.3%

*Source: htnea2021006 - references.*

### references.

### references.

### Executive summary: key findings and context
- The coronavirus (COVID-19) crisis is likely to deteriorate banks’ balance sheets in Africa, with the largest threat pertaining to loan portfolios and a likely sharp increase in nonperforming loans (NPLs) in the short to medium term.
- Elevated NPLs generate macroprudential and financial stability risks and impair banks’ ability to support the economy during the recovery by:
  - Raising capital requirements;
  - Denting banks’ net interest margins;
  - Generating service and management costs;
  - Potentially weakening the ability of banks to grant new loans.
- Structural impediments in sub-Saharan Africa to NPL resolution include weak debt enforcement procedures, weak legal rights, and financial infrastructure gaps.
- The note and accompanying Excel template quantify the capital relief and increase in credit capacity from disposing of NPLs (sale to third parties), focusing on the capital relief channel while noting broader micro and macroeconomic benefits (reduced uncertainty, efficiency gains, restored profitability).
- In countries with high provisions (LLR are the cumulation (stock) of provisions over time), regulatory capital released by NPL disposal can be relatively limited because risk-weighted assets and capital requirements are based on net NPLs (NPLs minus LLR).

### NPL resolution strategies and empirical examples
- Broad categories of strategies for NPLs:
  - Keep NPL on balance sheet:
    - Legal enforcement (court, insolvency, liquidation, administrative processes).
    - Consensual solutions (cash settlement, traditional loan workout, out-of-court restructuring).
  - Remove NPL from balance sheet:
    - Write-off.
    - Sale to private entity.
    - Sale to public entity.
    - Sale to mixed public-private entity.
- Examples of NPL disposal approaches and outcomes:
  - Accelerated write-offs:
    - Malawi: new regulation from 2017 led NPL ratio to decline from 15.7 percent at the end of 2017 to 3.6 percent in September 2019 (largely due to write-offs, loan recovery, and overall growth in bank lending).
    - Tanzania: 2018 Bank of Tanzania directive to write-off credit accommodations in loss category for more than four consecutive quarters (previously more than 12 quarters) contributed to NPL ratio decline from 11.5 percent at the end of 2017 to 9.8 percent at the end of 2019.
  - Securitization via a Special Purpose Vehicle (SPV):
    - Nigeria (2010): SPV acquired NPLs with an original book value of N4.02 trillion at a price of N1.76 trillion, equivalent to 1.7 percent of GDP (reflecting a 56 percent haircut). After transfer and securitization, NPL ratio dropped from 38 percent at the end of 2010 to below 5 percent at the end of 2012.
  - Centralized asset management companies (AMCs):
    - Angola (2016): Recredit, a state-owned AMC, purchased distressed assets to free up lending capacity. Recredit purchased NPLs from one bank associated with six large borrowers for a total of Kz480 billion, equivalent to about 3 percent of GDP.

### Analytical focus, framework, and template output
- The note defines key concepts and computational steps for valuing and simulating NPL disposal effects:
  - Definitions: haircut, capital loss, unprovisioned loan loss; net book value defined as gross book value net of specific loan loss reserves.
  - Equivalences: haircut ≡ capital loss ≡ unprovisioned loan loss (as developed in the note).
  - Computation elements covered: unprovisioned loan loss; loss under default; model-based haircut formula.
- Main simulation steps in the template:
  - Step 1: Calculation of the Tied-up Capital.
  - Step 2: Calculation of the Capital Relief.
  - Step 3: Use of the Freed-up Capital to Grant New Loans.
- Structure, calibration, and outputs:
  - The Excel template includes structure and calibration inputs and computes the outputs of interest (capital relief and additional lending capacity). The template builds on and expands earlier IMF and external work (for example, Jobst, Portier, and Sanfilippo (2015) in a European context).

### Box 1. Examples of NPL Sales and Write-Offs in Sub-Saharan Africa

### Box 1. Examples of NPL Sales and Write-Offs in Sub-Saharan Africa

### Overview of the analytical framework
- Purpose: Estimate regulatory capital released by removing NPLs from bank balance sheets (recorded at net book value) and the potential quantum of fresh credit that could be provided from the capital released.
- Two main effects of NPL disposal:
  - Capital requirement effect: sale reduces banks’ regulatory capital charge proportionately because high-risk-weight NPLs are exchanged for cash (zero risk weight).
  - Capital resource effect: sale can lower bank capital if sold below net book value (positive haircut) or raise capital if sold above net book value (negative haircut).
- Definition: Capital relief = change in bank capital resources (pre and post NPL sale) minus change in capital requirement (pre and post NPL sale).
- Important caveats highlighted:
  - Data quality and differing NPL/provision definitions across countries.
  - Sensitivity of outputs to assumptions (collateralization, collateral decay, haircut rates).
  - Credit scarcity can also stem from fiscal dominance, poor financial market infrastructure, financial inclusion bottlenecks, and lack of bankable projects.
  - The template is based on a stylized macroeconomic framework and cannot fully capture bank-/country-specific legal, regulatory, and portfolio heterogeneity.

### Key concepts: haircut, capital loss, unprovisioned loss
- Haircut (level) = NBV − sale price = capital loss (if positive) or capital gain (if negative).
- Net book value (NBV) = Gross book value (GBV) − loan loss reserves (LLR). (Footnote: GBV is calculated by amortized cost method using the original effective interest rate; NBV corrects GBV by value adjustment if borrower has difficulties.)
- Alternative expression: haircut (level) ≈ unprovisioned loan loss, where:
  - Unprovisioned loss = total projected loss (NPV) − LLR.
  - Haircut (level) = NBV − sale price = (NBV − GBV) − (sale price − GBV) ≈ −LLR + total projected loss = unprovisioned loan loss.

### Model-based unprovisioned loss and loss-under-default computation
- Two recovery options for an NPL (per unit of GBV):
  - Consensual recovery with probability p: recovery fraction α → loss per unit = (1 − α), probability p.
  - Legal enforcement with probability (1 − p): loss under default (see formula below).
- Unprovisioned loss per unit of gross NPL:
  - Unprovisioned loss per unit of gross NPL = p * (1 − α) + (1 − p) * Loss under default − llr
  - where llr = LLR / GBV.
- Loss under default per unit of gross NPL (NPV at initial period), with secured fraction collat and unsecured fraction uncollat (uncollat = 1 − collat), discount rate r, resolution time t, collateral decay rate δ, management cost m_cost, legal cost l_cost:
  - Loss under default per unit of gross NPL = uncollat/(1 + r)^t + [collat/(1 + r)^t − collat * (1 − d)^t/(1 + r)^t] + m_cost + l_cost
  - Equivalent rearrangement provided: Loss under default per unit of gross NPL = 1/(1 + r)^t − [collat * (1 − d)^t]/(1 + r)^t + m_cost + l_cost
- Model-based haircut (level) = Unprovisioned loan loss per unit of Gross NPL sold * Gross NPL sold.

### Main steps of the simulations (three stages)
- Step 1: Calculation of the tied-up capital
  - Tied-up capital = Net NPL sold * (WNPL * reg. CAR%) * (dRWA / dCRWA)
  - Net NPL sold = Actual net NPL − Target net NPL (Net NPL sold = Actual Gross NPL − Gross NPL Target)*(1 − llr) under default template option.
  - By default reg. CAR% = 12 percent; default WNPL = 100 percent; default WPL = 100 percent.
  - Two options for dRWA / dCRWA:
    - Default: assume other components of RWA fixed ⇒ dRWA / dCRWA = 1.
    - Alternative: assume RWA composition constant ⇒ dRWA / dCRWA = RWA / C RWA (computed from bank-level data).
- Step 2: Calculation of the capital relief
  - Scenario 1 (no haircut): Capital relief = Tied-up capital.
  - Scenario 2 (ad hoc haircut ratio θ): Capital relief = Tied-up capital − θ * Net NPL sold.
  - Scenario 3 (model-based haircut): Capital relief = Tied-up capital − Unprovisioned loan loss per unit of Gross NPL sold * Gross NPL sold.
  - Haircut ratio in scenario 3 (percent of net NPL) = Unprovisioned loan loss per unit of Gross NPL sold * Gross NPL sold / Net NPL sold.
- Step 3: Use freed-up capital to grant new loans
  - Additional performing loans = Capital relief * (1 / (WPL * reg.CAR%)) * (dCRWA / dRWA)
  - Two options for dCRWA / dRWA:
    - Default: assume other components fixed ⇒ dCRWA / dRWA = 1.
    - Alternative: dCRWA / dRWA = C RWA / RWA.

### Template structure, calibration defaults, and haircut scenarios
- Three haircut scenarios simulated in the template:
  - Scenario 1: no haircut (NPLs sold at NBV) — main scenario for cross-country comparison.
  - Scenario 2: fixed haircut (default θ = 10 percent positive haircut ratio).
  - Scenario 3: country-specific, model-based haircuts using the model formulas above.
- Common default assumptions:
  - Banks reduce NPL ratio to a desired level; default assumes NPL ratio is halved relative to its 2018 value.
  - Regulatory CAR default = 12 percent.
  - Default WNPL = 100 percent (alternative 150 percent available).
  - Default WPL = 100 percent (alternative 75 percent available).
  - Default dCRWA / dRWA = 1 (alternative = C RWA / RWA).
- Default calibration for model-based haircut (scenario 3) parameters:
  - Discount rate r = 10 percent.
  - collat = 0.8 (80 percent collateralized), uncollat = 0.2.
  - Collateral decay rate δ = 0.05 per year.
  - Management cost m_cost = 0.05.
  - Legal cost l_cost = proxied from World Bank Doing Business “enforcing contracts” indicator; 75 percent of reported value used.
  - Average remaining time to resolution t = sourced from World Bank Doing Business (converted to years and rounded to nearest 0.5 by template).
  - Probability of consensual recovery p = 0.67.
  - Recovery fraction under consensual route α = 0.35.
- Haircut scenario notes and empirical ranges:
  - Default uniform positive haircut = 10 percent, consistent with Jobst, Portier, and Sanfilippo (2015).
  - Historical haircuts in crisis contexts can be very high (close to 80 percent of gross loans); typical ranges noted include 50 percent of gross loans for mortgage/collateralized loans, two-thirds for retail NPLs, or higher for corporate debt in difficult times.
  - Haircuts can be nonexistent or negative in special circumstances (e.g., public purchasing entity, public support for systemically important banks).

### Options for computing provisioning on NPLs sold
- Option 1 (default): country average provision ratio from IMF FSI data: llr = aggregate specific provisions / aggregate gross NPLs.
- Option 2: weighted average of provision rates by NPL category (substandard, doubtful, loss) using World Bank Bank Regulation and Supervision Survey; weights from Fitch Connect; assumed disposal order: loss → doubtful → substandard.
- Option 3: same as Option 2 but disposal order reversed: substandard → doubtful → loss.
- Practical data caveats for Options 2 and 3: missing reporting by banks for NPL categories leads to imputations (remove banks with no data; distribute residuals; use SSA median when country data missing).

### Outputs, sensitivity analyses, and policy experiments
- Main outputs for each country and scenario:
  - Capital released by NPL disposal (capital relief).
  - Additional performing loans that can be generated from capital relief.
  - Descriptive statistics: NPL ratios (net and gross), breakdown by loan type (substandard, doubtful, loss).
  - For scenario 3: country-specific model-based haircuts (sensitive to parameter choices).
- Sensitivity analyses under scenario 3 vary parameters such as collateral depreciation rate, collateralized portion of NPL, probability of consensual resolution, target CAR, and others.
- Example policy experiments the template can simulate:
  - Measures that boost market value of NPLs (develop distressed-asset market, improve collateral valuation/registry, establish specialized NPL collection agencies) by inputting a negative haircut ratio.
  - Targeted disposal strategies that remove legacy (loss) NPLs first by selecting Option 2/3 and adjusting parameters accordingly.
  - Reforms to reduce time and costs of contract enforcement (e.g., lower legal duration by one year) to assess impact on haircuts, capital relief, and additional lending.

### Data issues and interpretation notes
- Scenario 3 can produce extreme haircut ratios expressed as percent of net NPLs where net NPL figures are very low due to:
  - Stringent provisioning practices (legacy NPLs fully provisioned but kept on balance sheet because of tax/legal impediments).
  - Statistical issues (provisions reported against both performing and nonperforming loans; slow reclassification of loans causing provisional overstatement relative to recorded NPLs).
- Template results are sensitive to assumptions; model-based haircuts are more informative for policy experiments than absolute quantification.

### Selected median statistics shown in template outputs (as reported)
- SSA median = 0.04
- SSA median gross NPLs = 11.8
- SSA median loan loss provisions = 8.4
- % of initial performing loans (RHS) SSA median = 1.9 (RHS)
- % of GDP, SSA median = 0.4
- SSA median = 0.01
- % of initial performing loans, SSA median = 0.3 (RHS)
- % of GDP, SSA median = 0.1
- SSA median reduction in gross NPLs = 5.9
- SSA median reduction in net NPLs = 1.7

*Source: IMF staff (from Box 1 in the PDF "HOW TO ASSESS THE BENEFITS OF NONPERFORMING LOAN DISPOSAL IN SUB-SAHARAN AFRICA USING A SIMPLE ANALYTICAL FRAMEWORK", June 2021).*

### 1. SSA: NPLs and Specic Provisions, 2018

### 1. SSA: NPLs and Specific Provisions, 2018

### NPLs and Specific Provisions (Percent of total gross loans)
- Chart title: "1. SSA: NPLs and Specific Provisions, 2018 (Percent of total gross loans)"
- Country labels included (order as in figure): GNQ, GNB, COM, BEN, GHA, NER, CAF, LBR, SEN, CMR, CPV, KEN, MOZ, TZA, SWZ, MDG, MUS, BWA, ZAF, NAM, GMB, TCD, STP, AGO, COG, TGO, COD, BDI, MLI, SLE, GAB, GIN, NGA, ZMB, CIV, BFA, ZWE, RWA, SYC, LSO, UGA, ETH, MWI, TCD, STP, AGO, COG, TGO, COD, BDI, MLI, SLE, GAB, GIN, NGA, ZMB, CIV, BFA, ZWE, RWA, SYC, LSO, UGA, ETH, GNQ, GNB, COM, BEN, GHA, NER, CAF, LBR, SEN, CMR, CPV, KEN, MOZ, TZA, SWZ, MDG, MUS, BWA, ZAF, NAM, GMB, MWI
- Note: The shock presented in these panel charts is a 50 percent decline in the NPL ratios relative to 2018.

### Capital Relief from NPLs Disposal — No Haircut (Percent of GDP)
- Chart title: "3. SSA: Capital Relief from NPLs Disposal, No Haircut (Percent of GDP)"
- Vertical scale markings implied by figure: 0, 5, 10, 15, 20, 25, 30, 35, 40 (percent of GDP axis labels shown in chart)
- Country order in chart (as presented): MUS, CPV, COG, BEN, AGO, TGO, SWZ, COM, NER, CIV, KEN, TZA, LBR, GIN, MDG, LSO, BDI, ETH, UGA, GMB, MLI, GNQ, ZAF, GNB, SEN, TCD, NAM, BWA, STP, BFA, SYC, GHA, RWA, GAB, MWI, NGA, SLE, CMR, ZMB, CAF
- Specific small numeric values shown within the figure (select examples as plotted): 0.3, 0.0, 0.2, 0.1, 2.5, 0.0, 1.5, 1.0, 2.0, 0.5
- Note: The shock presented in these panel charts is a 50 percent decline in the NPL ratios relative to 2018.
- Sources: IMF Financial Soundness Indicators; country authorities; and staff estimates.

### Additional Performing Loans from NPLs Disposal — No Haircut
- Chart title: "4. SSA: Additional Performing Loans from NPLs Disposal, No Haircut"
- Vertical axis markings (percent of initial total gross loans implied): 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 25, 30, 35, 40
- Country order repeated (as presented): GNQ, GNB, COM, BEN, GHA, NER, CAF, LBR, SEN, CMR, CPV, KEN, MOZ, TZA, SWZ, MDG, MUS, BWA, ZAF, NAM, GMB, TCD, STP, AGO, COG, TGO, COD, BDI, MLI, SLE, GAB, GIN, NGA, ZMB, CIV, BFA, ZWE, RWA, SYC, LSO, UGA, ETH, MWI
- Example numeric ticks/labels shown in chart body: 20, 0, 15, 10, 5, 25, 20, 35, 30, 20

### Reduction in NPLs (Percentage points)
- Chart title: "2. SSA: Reduction in NPLs in Percent of Initial Total Gross Loans (Percentage points)"
- Large numeric label shown: 14 (prominent on page)
- Page context: part of template outputs and panel charts showing reductions corresponding to a 50 percent decline in NPL ratios relative to 2018.

### Capital Relief from NPLs Disposal — 10% Positive Haircut (Percent of GDP)
- Chart title: "5. SSA: Capital Relief from NPLs Disposal, 10% Positive Haircut (Percent of GDP)"
- Vertical axis sample numeric values plotted in figure: 0.05, 0.00, 0.03, 0.04, 0.01, 0.02
- Country order as in figure: MUS, CPV, COG, BEN, AGO, TGO, SWZ, COM, NER, CIV, KEN, TZA, LBR, GIN, MDG, LSO, BDI, ETH, UGA, GMB, GNQ, MLI, ZAF, GNB, SEN, TCD, NAM, BWA, STP, BFA, SYC, GHA, RWA, GAB, MWI, NGA, SLE, CMR, ZMB, CAF

### Additional Performing Loans from NPLs Disposal — 10% Positive Haircut
- Chart title: "6. SSA: Additional Performing Loans from NPLs Disposal, 10% Positive Haircut"
- Vertical axis sample numeric values in figure: 0.4, 0.0, 0.2, 0.1, 0.3
- Secondary numeric scale examples shown: 3.5, 0.0, 1.0, 0.5, 2.5, 3.0, 2.0, 1.5

### Country-Specific Model-Based Haircuts (Percent of gross NPL sold)
- Table title: "Country-Specific Model-Based Haircuts (Percent of gross NPL sold)"
- Reported entries (country — haircut percent) as presented (order and exact values preserved):
  - Eswatini 47.4%
  - Seychelles 36.9%
  - Namibia 28.8%
  - Malawi 27.4%
  - Equatorial Guinea 21.0%
  - Central African Rep. 27% (presented as "Central African Rep. 227%")
  - Cameroon 29% (presented as "Cameroon 229%")
  - Gabon 31% (presented as "Gabon 231%")
  - Mali 16.8%
  - South Africa 14.0%
  - Niger 9.5%
  - Lesotho 8.0%
  - Nigeria 16.6% (presented as "Nigeria 216.6%")
  - Côte d'Ivoire 21.8% (presented under column)
  - São Tomé & Príncipe 217% (presented as "São Tomé & Príncipe 217%")
  - Kenya 219% (presented as "Kenya 219%")
  - Burundi 224% (presented as "Burundi 224%")
  - Zambia 225% (presented as "Zambia 225%")
  - Uganda 5.2%
  - Ethiopia 5.5%
  - Mauritius 5.5%
  - Madagascar 21.1%
  - Rwanda 21.6%
  - Sierra Leone 23.8%
  - Senegal 24.4%
  - Togo 12.3% (presented as "Togo 212.3%")
  - Ghana 13.6% (presented as "Ghana 213.6%")
  - Cabo Verde 14.2% (presented as "Cabo Verde 214.2%")
  - Liberia 14.4% (presented as "Liberia 214.4%")
  - Rwanda 21.6% (duplicate entry appears in table)
  - Gambia, The 22.5%
  - Benin 3.2%
  - Burkina Faso 2.8%
  - Comoros 1.9%
  - Congo, Republic of 19.7%
  - Angola 21.9%
  - Togo 12.3% (duplicate format)
  - Ghana 13.6% (duplicate format)
  - Cabo Verde 14.2% (duplicate format)
  - Liberia 14.4% (duplicate format)
  - Lesotho 8.0% (duplicate)
  - Namibia 28.8% (duplicate)
  - Mozambique 21.9% (presented under Angola column grouping)
  - Burundi 24% (presented as "Burundi 224%")
- SSA Median shown: 0.3%
- Notes on display: some country entries in the figure appear with formatting anomalies (e.g., numbers prefixed by "2" such as "217%", "219%", "224%") consistent with the source presentation.
- Legend indicator: "Positive Negative" (presentation in source)

### Sensitivity Analysis for Benin (Capital Relief from NPLs Disposal)
- Chart titles: "7� Sensitivity Analysis for Benin: Capital Relief from NPLs Disposal (Percent of GDP)" and "8� Sensitivity Analysis for Benin: Capital Relief from NPLs Disposal (Percent of GDP)"
- Axis and curve labels shown in figure (preserved from source):
  - Horizontal axis labels include ranges: –100 –80 –60 –40 –20 0 10 20 30 40 50 60 70 80 90 100 020408060100
  - Variables plotted: Haircut (percent of net NPL sold); Portion of NPL collateralized (percent)
  - Vertical scale sample values: 1.2, –0.8, –0.4, 0.0, 0.4, 0.2, –0.5, –0.4, –0.3, –0.2, 0.0, –0.1, 0.1

### Notes, Sources, and References
- Repeated note: The shock presented in these panel charts is a 50 percent decline in the NPL ratios relative to 2018.
- Sources cited on the page: IMF Financial Soundness Indicators; country authorities; and staff estimates.
- Selected references listed on the page (titles and authors preserved as shown): Aiyar et al. 2015; Alvarez & Marsal 2016; Basel Committee on Banking Supervision 2019; Bhatia et al. 2020; Ciavoliello et al. 2016; Croatian National Bank 2020; Dobler, Moretti, and Piris 2020; European Banking Authority 2016; Gandrud and Hallerberg 2014; Jobst, Portier, and Sanfilippo 2015.

*Source: IMF Financial Soundness Indicators; country authorities; and staff estimates (pages and figures from the PDF chapter "1. SSA: NPLs and Specific Provisions, 2018").*

---


_Source: https://www.imf.org/-/media/files/publications/howtonotes/2021/english/htnea2021006.pdf_