## 1mdvea2024001

## Source details

**Canonical URL:** [1mdvea2024001](https://www.imf.org/-/media/files/publications/cr/2024/english/1mdvea2024001.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2024/english/1mdvea2024001.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2024/english/1mdvea2024001.pdf.json)

---

### EXECUTIVE SUMMARY — Overview and risks
- A systemic vulnerability analysis and stress tests were conducted using quarterly aggregate balance sheet supervisory data for the eight banks in Maldives as of December 2022.
- Risks analyzed: credit risk (materialized as non-performing loans and pressure on pre-provision income), liquidity risk (deposit outflows), and market risk (changes in interest and exchange rates).

### Macrofinancial developments (key facts)
- Real GDP growth projected at 7.2 percent in 2023 (IMF April 2023 WEO).
- Fiscal and external vulnerabilities driving risk:
  - High public debt and increasing fiscal expenditure on debt service and price subsidies.
  - Widening current account deficit.
  - Continued financial support to SOEs (notably HDC), increasing domestic fiscal financing needs.
  - Persistent FX shortage in official markets, fueling a large parallel FX market and reducing tax revenue.
- Domestic debt issuance and monetary financing have increased sovereign risk exposure of MMA and the domestic banking system.

### Systemic vulnerabilities identified
- Intensified sovereign–bank nexus: growing holdings of sovereign securities and sharply rising lending to SOEs.
- Prudential/regulatory incentives: zero-risk weight on domestic sovereign paper, including FX-denominated issues, encouraging bank accumulation of sovereign debt.
- FX shortages could deplete reserves and prompt exchange range realignment, affecting SOEs and corporates with currency mismatches.
- Shadow banking growth (including unregulated leasing companies) and underreported household recurring payment obligations (lease/hire purchase; government rent-to-own).
- Weak liquidity management; need to finetune reserve requirements and adopt draft regulation to address parallel FX market drivers.
- Large corporate client concentration: individual banks close to single exposure limits.

### Stress testing approach and data
- Solvency stress tests: top-down and bottom-up; banks provided macro projections and applied own methods in bottom-up exercise.
- Data: supervisory quarterly bank-by-bank panel data for 2010–2022; overall adequate though mixed in coverage and granularity.
- Scenarios: three macrofinancial scenarios (baseline, moderate, severe), increasing in severity; severe scenario augmented with sovereign risk sensitivity analysis.
- Top-down solvency methodology highlights:
  - Estimated long-run relationship between NPLs and macro variables by currency (LC and FC).
  - Projected NPL ratios, additional provisioning needs, PPI, taxes, dividends to derive after-tax income and changes in bank capital.
  - Projection of RWA to obtain projected capital adequacy ratios.
- Additional analyses: credit concentration sensitivity (default of five largest exposures); alternative capital adequacy with 100 percent RW on sovereign FX securities; liquidity stress tests (cashflow-based and Basel III LCR in LC and FC); interest rate and FX risk sensitivity.
- No interconnectedness stress tests (interbank market virtually non-existent).

### Key solvency stress test findings (summary)
- Banking system broadly resilient to macroeconomic shocks but vulnerable to sovereign shocks and concentration risk.
- Severe scenario (including a sovereign domestic debt exchange) and credit concentration shock (default of five largest exposures) were main capital stress drivers.
  - Required recapitalizations in either case amounted to less than 1 percent of GDP.
- Imposing a 100 percent RW on domestic securities in FX substantially increased RWA and lowered CAR but did not trigger required recapitalization in the aggregate baseline-to-severe comparisons.
- Sovereign sensitivity analysis and credit concentration materially reduce aggregate capitalization in severe scenarios.

### Liquidity and market risk findings
- Liquidity stress tests: a couple of banks faced difficulties but only in specific time buckets.
- Basel III LCR calculation indicated sector compliance.
  - Basel III Liquidity Coverage Ratio, MVR: 253.8
  - Basel III Liquidity Coverage Ratio, USD: 235.1
  - Lowest ratios for any bank: 157.1 percent (LC) and 79.9 percent (FC); one bank fails the FC LCR.
- Deposit concentration: system sensitive to withdrawals from each bank’s five largest depositors; seven banks (49.9 percent of system assets) could not withstand full withdrawal by five largest depositors.
- Market risk (interest rate repricing and FX risk) assessed as moderate.

### Climate risk analysis — approach and headline results
- Micro analysis shocking banks’ immovable asset–related loans under three climate scenarios (SSP2-4.5, SSP3-7.0, SSP5-8.5); coastal flood hazard modeled via sea level rise and storm surge (windspeed used for storm surge).
- Exposure: geographical disaggregation by atoll for country exposure; bank exposures not geographically differentiated.
- Damage estimation: damage functions and elevation used to convert coastal flood depth and atoll exposure into damage rates by atoll; Monte-Carlo simulation (10,000 simulations) to compute country loss distribution; shock metric TVaR ninety-nine of the country distribution.
- Findings:
  - Mid-century climate effects on the banking system mild at the 99th percentile of the country loss distribution.
  - End-century impacts could be significantly exacerbated, mainly due to sea level rise; end-century outcomes show fatter tails and very large increases in damages.
  - Rising reinsurance premiums could lead to limited or no reinsurance availability for climate-related events.
  - Direct mid-century (no insurance) impact: increase in provisions leads to a 0.5 percentage point decline in system CAR (up to 0.9 percentage points at individual banks).
- Data limitations: end-century impacts and insurability issues require more granular data; improved data granularity and coverage would enhance analysis.

### Policy recommendations (condensed, Responsible Authority / Time)
- Improve integrity and granularity of supervisory data, including data compiled by the Credit Information Bureau (CIB). MMA / ST
- Develop methodologies for solvency, liquidity, and market risk stress tests and engage banks in dialogue about stress test procedures and results, including banks’ own stress tests. MMA / ST
- Implement scenario-based solvency stress tests. MMA / MT
- Implement cash flow-based liquidity stress tests. MMA / MT
- Improve data granularity to identify stable vs. less-stable deposits for NSFR calculation. MMA / ST
- Improve granularity and coverage of climate data, geographical exposures of the country and financial system, and climate-related damages; foster intra-agency collaboration to support access to existing data. MoE, MBS, MMA / ST
- Initiate climate risk analysis in collaboration with other agencies to assess the effect of actual and future climate conditions on the financial sector and the economy. MMA / MT
- Time definitions: ST = short term = 1-2 years; MT = medium term = 3-5 years

---

### Banking system structure and key metrics (2022)
- Banking system assets about 92.1 percent of GDP.
- Banks hold more than three-fourths of system assets; non-bank sector nascent (MRPS, insurance, leasing).
- Largest institution (domestic state-owned bank) holds 50.1 percent of banking system assets.
- Ownership composition of banking system assets, 2022:
  - state-owned: 75%
  - private domestic: 11%
  - branch: 9%
  - subsidiary: 5%
- Breakdown by major banks (Percent of Banking System Assets, 2022): BML 50%, SBI 18%, BOC 7%, MIB 7%, HSBC 6%, MCB 5%, HBL 4%, CBM 3%.
- Asset allocation shifts (2010→2022):
  - Loans: 60.5% → 37% of total system assets.
  - Investments (mostly government securities): 16.1% → 30% of total assets (over 50% at some banks).
- Loan growth 2022: around 5% (down from 18% in 2016); by currency: 9.9 percent in local currency and 1 percent in foreign currency.
- Loan composition changes: personal loans and real estate up; tourism and commercial loans down; personal loans growing at 27 percent per year.
- Loan dollarization: 45 percent of total loans.
- Deposit dollarization: 49 percent of total deposits.
- Asset quality:
  - NPLs 5.9 percent of total loans.
  - Sectoral loan shares: tourism 34 percent, construction 18.5 percent, personal loans 13.7 percent.
  - NPLs concentrated in tourism (60 percent) and commerce (17.3 percent); NPL ratios tourism 10.6 percent, commerce 8.9 percent.
  - NPL ratios by currency: foreign currency 9.0 percent of foreign currency loans; local currency 3.4 percent of local currency loans.
- Capitalization:
  - Systemwide CAR 50.9 percent of RWA (up from 46.8 percent in December 2021).
  - High capitalization concentrated in three largest banks (75 percent of banking system assets); some banks CAR as high as 90 percent.
  - Government securities holdings are zero risk-weighted, biasing capital ratios upward.
  - FSAP team’s adjustment: imposing RW=100% on domestic sovereign securities in FX would reduce aggregate CAR to 41.1 percent.

### Data and scope notes for tests
- Tests applied: solvency (top-down and bottom-up), sovereign sensitivity (domestic debt exchange), cashflow liquidity stress, Basel III LCR, deposit concentration, interest rate and FX sensitivity.
- Tests not performed: interconnectedness (no interbank market), duration-based mark-to-market losses (very short tenor of sovereign paper), NSFR (lack of clarity on stable vs. non-stable funding), NBFI sector excluded (small size, data issues).
- Securities duration: 56 percent ≤ 6 months; 83 percent ≤ 12 months.
- Supervisory data gaps: loan data by sector only from 2015; some banks had data gaps; two banks excluded from estimations for spotty data.

---

### Macrofinancial stress-test scenarios (selected variables, in percent unless otherwise stated)
- Real GDP growth (Est. 2022; Proj. 2023 2024 2025)
  - Baseline: 12.3 7.2 5.7 6.5
  - Moderate: 12.3 -14.3 13.4 10.4
  - Severe: 12.3 -24.4 8.8 10.5
- Inflation rate (eop)
  - Baseline: 3.3 5.1 2.5 2.1
  - Moderate: 3.3 5.8 4.2 2.4
  - Severe: 3.3 9.5 5 2.6
- Bank lending rate
  - Baseline: 10.9 10.9 10.9 10.9
  - Moderate: 10.9 12.5 11.2 10.9
  - Severe: 10.9 16.3 13.2 10.9
- Exchange rate (MVR per USD)
  - Baseline: 15.4 15.4 15.4 15.4
  - Moderate: 15.4 15.4 15.4 15.4
  - Severe: 15.4 18.5 18.5 18.5
- Credit growth
  - Baseline: 11.7 9.8 8.7 8.9
  - Moderate: 11.7 6.6 13.9 10.7
  - Severe: 11.7 2.2 10.4 9.5
- Fiscal balance (percent of GDP)
  - Baseline: -11.1 -9.1 -8.5 -5.7
  - Moderate: -11.1 -18.2 -13.9 -9.6
  - Severe: -11.1 -22.1 -19.5 -15.1

### Severe scenario sovereign sensitivity (domestic debt exchange) assumptions (example)
- Amount of T-bills to be replaced: 100
- Maturity: 20
- New Coupon: 2.5%
- Discount factor: 4.4%
- Sum of discounted interest payments: 32.8
- Discounted Principal: 42.3
- NPV Haircut: 24.9
- Applied as a 24.9 percent NPV haircut to book value of banks’ securities holdings and recognized as a loss.

---

### Top-down solvency results (selected figures and exact paths)
- NPL projection foreign currency model (baseline scenario, NPL ratio):
  - Dec. 22 9.0; Dec. 23 10.5; Dec. 24 11.8; Dec. 25 12.0
  - Highest NPL ratio of any bank (foreign currency): Dec. 22 34.4; Dec. 23 41.6; Dec. 24 46.1; Dec. 25 46.8
- NPL projection local currency model (baseline scenario, NPL ratio):
  - Dec. 22 3.4; Dec. 23 4.5; Dec. 24 5.1; Dec. 25 5.4
  - Highest NPL ratio of any bank (local currency): Dec. 22 27.5; Dec. 23 36.1; Dec. 24 40.4; Dec. 25 42.1
- Solvency Stress Tests (CAR, percent of RWA; Dec. 22 / Dec. 23 / Dec. 24 / Dec. 25)
  - CAR baseline scenario: 50.9 / 52.3 / 53.9 / 55.0
  - CAR moderate scenario: 50.9 / 53.1 / 50.6 / 50.2
  - CAR severe scenario: 50.9 / 39.3 / 44.0 / 45.1
  - Number of banks failing: 1
  - Pct. of banking system <12% CAR: 3.3
  - Capital shortfall in % of GDP: 0.1
- CAR adjusted for 100% RW on domestic securities in FX (Dec. 22 / Dec. 23 / Dec. 24 / Dec. 25)
  - Baseline adjusted: 41.1 / 43.3 / 45.4 / 47.0
  - Moderate adjusted: 41.1 / 43.6 / 42.6 / 43.0
  - Severe adjusted: 41.1 / 32.7 / 36.3 / 38.1
- Sovereign risk impact:
  - Drop in CAR (ppt.) due to domestic debt exchange: 14.8 / 8.6 / 7.8 (reported in context of Table 5)
  - Standalone sovereign sensitivity analysis: drop in CAR at the system level by 13.5 percent compared to end-2022 in the simulated domestic debt exchange.
- Credit concentration risk (impact of five largest borrowers)
  - CAR two-notch downgrade of 5 largest borrowers (Dec. 22 / Dec. 23 / Dec. 24 / Dec. 25): 50.9 / 46.7 / 48.0 / 49.0
  - CAR downgrade to loss of 5 largest borrowers: 50.9 / 20.6 / 21.1 / 21.6
  - Number of banks failing under loss of 5 largest borrowers: 5
  - Pct. of banking system <12% CAR: 27.1
  - Capital shortfall in % of GDP: 0.7
- Observations:
  - Severe scenario domestic debt exchange causes system CAR to drop by as much as 12.1 percentage points (narrative).
  - One bank accounting for 3.3 percent of system assets becomes undercapitalized in that severe scenario.
  - Four banks jointly accounting for 30 percent of system assets would breach single exposure limits in the hypothetical crisis.

### Interest rate and FX sensitivity (selected results)
- Interest rate repricing shock: symmetric increase of 200 basis points on interest-bearing assets and liabilities maturing ≤ one year → moderate drop in capitalization of 1.2 percentage point; CAR after shock: 51.1 (percent of RWAs).
- Foreign currency risk shock: 20 percent devaluation → two banks breach Net Open Position limits; a large bank short in FX would see its NOP become more negative. Number of banks breaching NOP limits after 20% devaluation: 2.

---

### Liquidity stress testing — cashflow-based methodology and results
- Cashflow-based liquidity stress test uses roll-off/run-off rates across maturity buckets and currencies; methodology modeled after largest monthly deposit withdrawals since the GFC.
- Counterbalancing capacity includes cash, bank reserves (assumed excess reserves and half of required reserves available) and shorter-term securities; sales of other securities subject to fire-sale haircuts.
- Selected run-off/roll-off rates (exact values taken from Table 6):
  - Assets (Loans & Advances run-off/roll-off across buckets): 50%, 50%, 50%, 30%, 30%, 30%, 10%, 10%, 10% (Less than 7 days → Unclassified).
  - Liabilities (Demand Deposits run-off): 10%, 5%, 1%, 1%, 0%, 0%, 0%, 0%, 0%.
  - Liabilities (Savings Deposits run-off): 40%, 30%, 20%, 10%, 5%, 5%, 5%, 5%, 5%.
  - Import Deposits and Margin Accounts: 100% across all maturity buckets.
  - Borrowings: 100% across all maturity buckets.
- Cashflow-based liquidity results (aggregate and currency-specific):
  - LC: one bank may have insufficient liquidity buffers in the very short term (< 7 days); funding gap 182 million MVR (3.1 percent of total assets).
  - FC: one bank may have insufficient buffers in the long run (> 5 years); funding gap 605 million MVR (4.2 percent of total assets).
  - Aggregate: banks resilient to aggregate cash outflows; sufficient counterbalancing capacity overall (Table 7).
- Deposit concentration sensitivity analysis:
  - Full withdrawal of five largest depositors per bank modeled with use of cash inflows within 7 days.
  - Seven banks (49.9 percent of total banking sector assets) would not be able to withstand runs on either LC or FC deposits.
  - Summary metrics:
    - Number of banks failing cashflow-based test in either currency: 2
    - Banks failing the cashflow test in either currency in % of system assets: 25.3
    - Banks failing the cashflow test in either currency in % of GDP: 23.3
    - DEPOSIT CONCENTRATION RISK: Number of banks failing test in either currency: 7
    - Assets of banks failing the test in % of banking system assets: 49.9
    - Assets of banks failing the test in % of GDP: 46

---

### Climate risk analysis — methodology details and quantification
- Physical vulnerability context:
  - 80 percent of Maldives’ islands are less than one meter above sea level.
  - 90 percent of the islands report flooding annually.
  - 97 percent of the islands report shoreline erosion.
  - 64 percent of the islands report severe erosion.
  - 50 percent of housing structures are within 100 meters of the coastline.
- Hazard modeling:
  - Coastal flood hazard = storm surge tail distribution + sea level rise.
  - Sea level rise projections consider all GCMs; storm surge uses three GCMs and downscaled grids (~10x10 km for surge; ~100x100 km for sea level rise).
  - Sea level rise mid-century median differences small across scenarios; end-century levels can reach increases above 1.5 meters in some scenarios.
  - Mid-century sea level rise ranges reported between 0.14 to 0.62 and an average median of 0.25 (change in mm; location Male) across scenarios.
- Exposure and damage estimation:
  - Islands considered: 345 inhabited and resort islands used.
  - Exposure disaggregated by industry using MBS and MoT proxies (tourism ~26%, transportation ~11%).
  - Damage functions: Huizinga et al. (2017) for Asia used to estimate damage rates by flood depth and atoll.
- Transmission to banks:
  - Focus on immovable asset-related loans (construction, real estate, tourism) representing 60 percent of total loan portfolio in December 2020.
  - Assumptions: no insurance coverage; recovery rate 5 percent; impacted performing loans provisioned to 95 percent; non-performing loans provisions adjusted to target 95 percent coverage.
- Quantitative results (mid-century, 99th percentile, selected numbers)
  - Mid-century direct coastal flood damages could reach 3.24 percentage damage to capital stock for SSP2 4.5 and SSP5 8.5 in the 99th percentile with the 95th percentile of sea level rise.
  - Mid-century damages could reach up to 3.37 percent of the capital stock (range reported 3.24 to 3.37 percent).
  - 95th percentile sea level rise increases damage by around 0.5 percentage points compared with the median sea level rise.
  - End-century damages (not used for banking-system impact due to uncertainty) suggest increases from 130-734 percent in SSP5 8.5 and 80-153 percent in SSP2 4.5 relative to mid-century baselines.
  - Banking-sector capital impact (direct, mid-century, no insurance): increase in provisions leads to a 0.5 percentage point decline in system CAR (up to 0.9 percentage points at individual banks).

### Climate recommendations (selected)
- Improve data granularity and coverage: collect standardized geographical exposure data for financial institutions; MoE to expand hazard coverage beyond coastal floods; enhance data access across agencies.
- Develop geographical granularity for country and financial system exposures; involve MBS and MMA.
- Initiate climate risk analysis collaboration: MMA to develop a framework for climate risk analysis using global datasets while prioritizing local sources as they become available; prioritize industries with larger effects on the economy and financial system.

---

### Risk Assessment Matrix — selected systemic risks, likelihood and impact
- Systemic regional instability: Likelihood Medium; Impact High.
  - Drivers: financial market dislocations, widening parallel market exchange rate spread, sovereign-bank nexus.
  - Policy response: reduce domestic financing needs; adjust RW on banks’ sovereign debt exposure; reduce excess liquidity and dollarization.
- Intensification of military conflicts / commodity price shock / supply chain disruptions: Likelihood Medium; Impact High.
  - Policy response: asset purchase program, liquidity support to banks, fiscal consolidation, adjust banks’ RWs on sovereign debt exposures.
- Abrupt global slowdown or domestic tourism shock: Likelihood Medium; Impact High.
  - Policy response: targeted temporary fiscal support; keep financial conditions accommodative if necessary.
- Extreme climate events: Likelihood Medium; Impact High.
  - Policy response: contingent financing plans with development partners; targeted temporary assistance with clear criteria; prioritize disaster-resistant infrastructure; identify financial sector exposures to physical risks.

*Source: EXECUTIVE SUMMARY and chapter content, 1mdvea2024001*

### EXECUTIVE SUMMARY __________________________________________________________________________ 6

### EXECUTIVE SUMMARY

### Overview
- A systemic vulnerability analysis and stress tests were conducted as part of the Maldives FSAP using quarterly aggregate balance sheet supervisory data for the eight banks in Maldives as of December 2022.
- Risks analyzed: credit risk (materialized as non-performing loans and pressure on pre-provision income), liquidity risk (deposit outflows), and market risk (changes in interest and exchange rates).

### Macrofinancial developments
- The Maldives’ economy rebounded strongly from the pandemic-induced contraction; real GDP growth projected at 7.2 percent in 2023 (IMF April 2023 WEO).
- Fiscal and external vulnerabilities remain elevated due to:
  - High public debt and increasing fiscal expenditure on debt service and price subsidies.
  - Widening current account deficit.
  - Continued financial support to SOEs (notably HDC), increasing domestic fiscal financing needs.
  - Persistent FX shortage in official markets, fueling a large parallel FX market and reducing tax revenue.
- Domestic debt issuance and monetary financing have increased sovereign risk exposure of MMA and the domestic banking system.

### Systemic vulnerabilities identified
- Main vulnerabilities:
  - Intensified sovereign–bank nexus driven by growing holdings of sovereign securities and sharply rising lending to SOEs.
  - Prudential and regulatory incentives (zero-risk weight on domestic sovereign paper, including FX-denominated issues) encouraging bank accumulation of sovereign debt.
  - FX shortages that could deplete reserves and prompt exchange range realignment, affecting SOEs and corporates with currency mismatches.
  - Shadow banking growth (including unregulated leasing companies) and underreported household recurring payment obligations from lease and hire purchase programs and government rent-to-own scheme.
  - Weak liquidity management, need to finetune reserve requirements, and adopt draft regulation to address parallel FX market drivers.
- Large corporate client concentration: individual banks close to single exposure limits.

### Stress testing approach
- Solvency stress tests: top-down and bottom-up approaches, banks given macro projections and asked to apply their own methods in bottom-up exercise.
- Data: supervisory quarterly bank-by-bank panel data for 2010–2022; overall adequate though mixed in coverage and granularity.
- Scenarios: three macrofinancial scenarios (baseline, moderate, severe), increasing in severity; severe scenario augmented with sovereign risk sensitivity analysis.
- Methodology (top-down solvency test):
  - Estimated long-run relationship between NPLs and macro variables for each bank.
  - Projected NPL ratios for loan portfolios in local and foreign currency; calculated additional provisioning needs.
  - Projected pre-provision income, taxes, dividends to derive after-tax income and changes in bank capital.
  - Applied projection of RWA to obtain projected capital adequacy ratios.
- Additional analyses:
  - Credit concentration sensitivity (default of five largest exposures).
  - Alternative capital adequacy calculation assuming non-zero risk weights on sovereign securities in foreign currency.
  - Liquidity stress tests: cashflow-based methodology with long-term outflow estimates by deposit type; Basel III LCR also calculated, both in local and foreign currency.
  - Other sensitivity analyses: interest rate risk, foreign currency risk, deposit concentration risk.
  - No interconnectedness stress tests (interbank market virtually non-existent).

### Key solvency stress test findings
- Banking system is broadly resilient to macroeconomic shocks but less so to sovereign shocks and concentration risk.
- Severe scenario (including a sovereign domestic debt exchange) and the credit concentration shock (default of five largest exposures) were the main drivers of capital stress.
  - Required recapitalizations in either case amounted to less than 1 percent of GDP.
- A Basel risk weights adjustment of 100 percent on domestic securities in FX substantially increased RWA and lowered capital adequacy ratios, but did not trigger required recapitalization.
- Sovereign sensitivity analysis and credit concentration materially reduce aggregate capitalization in severe scenarios (see figures referenced in source for detailed paths).

### Liquidity and market risk findings
- Liquidity stress tests: a couple of banks faced difficulties but only in specific time buckets.
- LCR calculation indicated sector compliance with Basel III.
- Deposit concentration is a vulnerability: system shows sensitivity to withdrawals from each bank’s five largest depositors.
- Market risk (interest rate repricing and foreign currency risk) assessed as moderate.

### Climate risk analysis
- Approach: micro analysis shocking banks’ immovable asset–related loans under three climate scenarios; coastal flood hazard modeled via sea level rise and storm surge (windspeed used for storm surge modeling).
- Exposure: geographical disaggregation of economic activity by atoll; not differentiated by banks.
- Method: used damage functions and elevation to convert coastal flood depth and atoll exposure into damage rates by atoll; damage rates used to compute interactions across atolls and aggregate country losses.
- Findings:
  - Mid-century climate effects on the banking system found to be mild when considering the 99th percentile of the country loss distribution.
  - End-century impacts could be significantly exacerbated, mainly due to sea level rise.
  - Rising reinsurance premiums could lead to limited or no reinsurance availability for climate-related events.
- Data limitations: assessment of end-century impacts and insurability issues requires more granular data; improved data granularity and coverage would enhance climate risk analysis.

### Policy recommendations (Table 1 condensed)
- Stress Testing (Responsible Authority / Time*):
  - Improve integrity and granularity of supervisory data, including data compiled by the Credit Information Bureau (CIB). MMA / ST
  - Develop methodologies for solvency, liquidity, and market risk stress tests and engage banks in dialogue about stress test procedures and results, including banks’ own stress tests. MMA / ST
  - Implement scenario-based solvency stress tests. MMA / MT
  - Implement cash flow-based liquidity stress tests. MMA / MT
  - Improve data granularity to identify stable vs. less-stable deposits for NSFR calculation. MMA / ST
- Climate Risk Analysis (Responsible Authority / Time*):
  - Improve granularity and coverage of climate data, geographical exposures of the country and financial system, and climate-related damages; foster intra-agency collaboration to support access to existing data. MoE, MBS, MMA / ST
  - Initiate climate risk analysis in collaboration with other agencies to assess the effect of actual and future climate conditions on the financial sector and the economy. MMA / MT
- * ST: short term = 1-2 years; MT: medium term = 3-5 years

*Source: EXECUTIVE SUMMARY, 1mdvea2024001*

### 3. The Maldivian financial system is large relative to the economy and rather

### 3. The Maldivian financial system is large relative to the economy and rather

### Overview and banking structure
- Banks hold more than three-fourths of system assets; the non-bank financial sector is nascent and comprised overwhelmingly of the government retirement scheme (MRPS), a insurance and leasing companies.
- Banking system assets are about 92.1 percent of GDP.
- The largest institution, a domestic state-owned bank, holds 50.1 percent of banking system assets.
- Ownership composition of banking system assets, 2022:
  - state-owned: 75%
  - private domestic: 11%
  - branch: 9%
  - subsidiary: 5%
- Breakdown by major banks (Percent of Banking System Assets, 2022): BML 50%, SBI 18%, BOC 7%, MIB 7%, HSBC 6%, MCB 5%, HBL 4%, CBM 3%.

### Asset allocation and implications
- Loan placements decreased to 37 percent of total system assets in 2022 (down from 60.5 percent in 2010).
- Investments (mostly in government securities) grew to 30 percent of total assets in 2022 (up from 16.1 percent in 2010) and are over 50 percent at some banks.
- Loan growth in 2022: around 5 percent (down from 18 percent in 2016).
- By currency, loan growth: 9.9 percent in local currency and 1 percent in foreign currency.
- Recent credit composition shifts:
  - Personal loans and real estate loans have grown, tourism and commercial loans have declined.
  - Personal loans are growing at 27 percent per year, seven times the yearly growth of tourism loans.
- Consequence: growing sovereign-bank nexus via larger sovereign securities holdings; banks face risk of hitting high exposure limits with large foreign-currency infrastructure projects.

### Asset quality
- Non-performing loans (NPLs) are 5.9 percent of total loans, consistently declining and, in recent years, amply provisioned.
- Sectoral loan shares (outstanding loans): tourism 34 percent, construction 18.5 percent, personal loans 13.7 percent.
- NPL composition and ratios:
  - NPLs mostly in tourism (60 percent) and commerce (17.3 percent).
  - NPL ratios: tourism 10.6 percent, commerce 8.9 percent.
- Substantial loans over 720 days past due exist; some are legacy high value loans to tourism in foreign currency (FC).
- Loan dollarization: 45 percent of total loans.
- Currency breakdown: Most individuals borrow in local currency; nonfinancial corporations borrow in foreign currency.
- NPL ratios by currency: foreign currency 9.0 percent of foreign currency loans; local currency 3.4 percent of local currency loans.

### Liquidity and funding
- Liquidity is ample, especially in local currency.
- Excess liquidity is placed in government securities, mostly at short maturities.
- Foreign currency is scarce; availability subject to central bank allocation or to a premium in the parallel market.
- Most securities are held to maturity; lack of a secondary market effectively blurs HFT, AFS and HTM distinctions.
- Short investment maturities match very short-term funding.
- Deposit dollarization: 49 percent of total deposits.

### Capitalization
- Systemwide capital adequacy ratio (CAR) is 50.9 percent of RWA (up from 46.8 percent of RWA in December 2021).
- High capitalization concentrated in the three largest banks (75 percent of banking system assets); some banks have CAR as high as 90 percent.
- Government securities holdings are zero risk-weighted, biasing capital ratios upward.
- FSAP recommendation: imposing a risk weight (RW) of 100 percent on domestic sovereign securities in FX would reduce aggregate CAR to 41.1 percent (FSAP team’s analysis of initial aggregate CAR under that measure).

### Systemic risk analysis: scope and data quality
- Stress tests applied:
  - (i) solvency stress tests (top-down using dedicated credit risk models and bottom-up by banks) and a credit concentration sensitivity analysis;
  - (ii) sovereign risk sensitivity analysis based on a domestic debt exchange of Maldives T-bills into longer-dated, lower yielding bonds;
  - (iii) cashflow mismatch liquidity stress test and calculation of the Basel III Liquidity Coverage Ratio and a deposit concentration risk test;
  - (iv) market risk sensitivity analysis: interest rate risk (maturity mismatches impact on net interest income under a symmetric increase in loan and funding rates) and foreign currency risk (assumed devaluation impact on banks’ FX net open position).
- Tests not performed and reasons:
  - No interconnectedness stress test due to no functioning interbank market.
  - No duration-based mark-to-market losses test due to very short tenor of sovereign paper.
  - Basel III Net Stable Funding Ratio not calculated because of lack of clarity on stable vs. non-stable funding.
  - NBFI sector excluded from stress tests due to small size and data availability issues.
- Securities portfolio characteristics: 56 percent have a duration of up to 6 months and 83 percent of up to 12 months.
- Supervisory data quality:
  - Loan data by major economic sector only from 2015; certain recipient designations unavailable.
  - Panel used for credit risk modeling extended back to 2010 by segmenting loans into local currency (LC) and foreign currency (FC).
  - Funding and assets data sufficiently granular for cashflow-based liquidity tests and LCR calculation.
  - Some banks had data gaps (two banks with gaps for three quarters each during rising NPLs; two banks excluded from estimations for spotty data).

### Macrofinancial stress-test scenarios
- Three scenarios for solvency stress testing:
  - Baseline: based on IMF’s WEO projections as of April 2023.
  - Moderate stress: assumes external demand shock similar to the pandemic; decline in real GDP growth from baseline is three times the pre-pandemic standard deviation, reaching -14.3 percent in 2023.
  - Severe stress: builds on moderate scenario with more severe downturn, additional global and domestic tightening, domestic debt exchange of T-bills for long-dated bonds at lower yield (implying a 25 percent reduction in face value), and a one-step devaluation of the exchange rate by 20 percent. Real GDP contracts by -24.4 percent (less severe than the -33.4 percent contraction in 2020).
- Econometric modeling found only inflation and import growth (proxy for economic activity) to be significant drivers of banks’ NPL ratios.
- Two additional simulated shocks in severe scenario (from Global Projection Model tailored work):
  - global financial shock: 200 bps increase in the US treasury term premium;
  - domestic financial shock: 20 percentage points increase in sovereign spreads.

### Solvency stress-test methodology and key modeling assumptions
- NPL projection approach:
  - Used bank-by-bank quarterly panel data for 2010-2022.
  - Initial sectoral regressions were not significant; final estimation by currency (LC and FC) given tourism, dollarization, and FX availability importance.
  - Models used autoregressive specification with logit-transformed NPL ratio and fixed effects.
- NPL model findings:
  - Only inflation and import growth were significant drivers.
  - FC NPL Ratio Model: R2 = 0.91; coefficients reported in Table 2 (e.g., lagged NPL ratio 0.888 (0.039)’’’ ; inflation 1.963 (0.540)’’’ ; import growth -0.138 (0.054)’’ ).
  - LC NPL Ratio Model: R2 = 0.92; lagged NPL ratio 0.867 (0.042)’’’ ; lagged inflation in FC model 1,435 (0.556)’’ ; import growth -0.098 (0.028)’’’.
- Projection steps:
  - NPL ratios in LC and FC estimated for each bank and scenario for 2023–2025.
  - Projected NPL amounts calculated using projected credit growth and NPL ratios; system NPL ratio obtained by aggregating NPL amounts relative to total credit.
- Loan loss provisioning assumptions:
  - Baseline scenario: 20 percent provisioning on projected additional NPLs.
  - Stress scenarios: 100 percent additional NPL provisioning.
  - No deleveraging of assets assumed (no fire sales of assets; credit growth assumed).
  - Justification: LGD close to 100 percent due to protracted foreclosure delays; during early pandemic coverage ratios reached around 150 percent.
- Income, tax, dividend and other balance-sheet assumptions:
  - Pre-provision income (PPI = net interest income, net fee income and net other income):
    - Baseline: increases proportionally with credit growth (about half of credit growth).
    - Moderate scenario: changes broadly with evolution in PPI during the pandemic.
    - Severe scenario: recedes further with an even larger adjustment.
  - Corporate tax rate: 25 percent statutory rate applied.
  - Dividend payout rate: 0 percent assumed (MMA discouraged dividend payouts during the pandemic; several banks do not plan to pay dividends).

*Source: IMF staff estimates and supervisory data as presented in the chapter.*

### 18.      The projected capital adequacy ratios are obtained through the aforementioned

### 18.      The projected capital adequacy ratios are obtained through the aforementioned

### Projection methodology and drivers
- Projected capital adequacy ratios (CAR) are computed from:
  - Changes in banks’ retained earnings obtained from post-tax profits less dividends (dividends assumed to be zero in the stress test).
  - A projection of risk-weighted assets (RWA).
  - A projection of net interest income.
- Change in capital ratio = projected change in numerator (retained earnings) divided by projected RWA.
- RWA growth paths:
  - Baseline scenario: broadly in line with the average growth rate of RWA during the past decade.
  - Stress scenarios: somewhat lower RWA paths due to lower credit growth.
- RWA growth projections are derived autonomously within the stress test calculation template and are subject to model uncertainty.

### PPI and RWA growth paths (Table 3)
- In percent per annum:
  - Dec-23 / Dec-24 / Dec-25
  - Baseline scenario
    - PPI growth rate 0.9% / 1.2% / 2.9%
    - RWA growth rate 11.8% / 10.2% / 10.2%
  - Moderate scenario
    - PPI growth rate -4.3% / -3.0% / -0.1%
    - RWA growth rate 7.7% / 16.5% / 12.2%
  - Severe scenario
    - PPI growth rate -8.7% / -6.6% / 1.1%
    - RWA growth rate 17.8% / -0.1% / 10.9%

### Credit concentration sensitivity test
- Assumption: the five largest loan exposures in pass category of each bank migrate to:
  - Substandard (implying an additional provision of 19.5 percent, calculated as 20 percent for substandard minus the existing provision of 0.5 percent on normal loans, assuming no extra provisions made), or
  - Loss category (implying additional provisions of 99.5 percent in the standard case).
- Additional provisions are added to projected baseline provisions and profits/changes in capital re-calculated.

### Integration of sovereign risk in solvency stress test
- Sovereign sensitivity analysis assumes exchange of T-Bills for longer-dated T-Bonds with lower coupon:
  - Example assumptions (Table 4):
    - Amount of T-bills to be replaced. 100
    - Maturity 20
    - New Coupon 2.5%
    - Discount factor 4.4%
    - Sum of discounted interest payments 32.8
    - Discounted Principal 42.3
    - NPV Haircut 24.9
- Under these assumptions:
  - Reduction in net present value (NPV) amounts to 24.9 percent.
  - This haircut is applied to the book value of banks’ securities holdings and recognized as a loss, added to additional provisions in the severe scenario.
  - The sovereign sensitivity analysis depicts the drop of CAR in the severe scenario due to the NPV loss from the debt exchange.

### Potential adjustment to risk-weighted assets (RW on sovereign securities)
- FSAP team alternative assessment: impose a RW of 100 percent on all government securities in FX (instead of current zero RW), per a recommendation of the BCP assessment.
- Consequence: banks holding large amounts of treasury bills and bonds denominated in foreign currency see markedly increased RWA.
- Projected CARs for 2023-25 are adjusted downward by increasing RWA in all three scenarios (presented as a hypothetical impact adding to scenario impacts).

### Top-Down stress test results — credit risk, capitalization, and sovereign impact
- NPL projections (in percent of RWAs and percent of total loans reported elsewhere):
  - NPL projection foreign currency model (NPL ratio stress baseline scenario): Dec. 22 9.0; Dec. 23 10.5; Dec. 24 11.8; Dec. 25 12.0
  - Highest NPL ratio of any bank (foreign currency) Dec. 22 34.4; Dec. 23 41.6; Dec. 24 46.1; Dec. 25 46.8
  - NPL projection local currency model (baseline scenario): Dec. 22 3.4; Dec. 23 4.5; Dec. 24 5.1; Dec. 25 5.4
  - Highest NPL ratio of any bank (local currency) Dec. 22 27.5; Dec. 23 36.1; Dec. 24 40.4; Dec. 25 42.1
- Pre-provision income:
  - Resulting additional loan loss provisions are easily offset by ample pre-provision income due to wide interest rate margins.
- Solvency stress test summary (Table 5):
  - Solvency Stress Tests
    - CAR baseline scenario Dec. 22-Dec. 25: 50.9 / 52.3 / 53.9 / 55.0
    - CAR moderate scenario Dec. 22-Dec. 25: 50.9 / 53.1 / 50.6 / 50.2
    - CAR severe scenario Dec. 22-Dec. 25: 50.9 / 39.3 / 44.0 / 45.1
    - Number of banks failing 1
    - Pct. of banking system <12% CAR 3.3
    - Capital shortfall in % of GDP 0.1
  - CAR baseline scenario adjusted for 100% RW on domestic securities in FX: 41.1 / 43.3 / 45.4 / 47.0
  - CAR moderate scenario adjusted for 100% RW on domestic securities in FX: 41.1 / 43.6 / 42.6 / 43.0
  - CAR severe scenario adjusted for 100% RW on domestic securities in FX: 41.1 / 32.7 / 36.3 / 38.1
- Sovereign risk impact:
  - Drop in CAR (ppt.) due to impact of domestic debt exchange: 14.8 / 8.6 / 7.8 (context implies across years or measures reported in Table 5)
  - Standalone sovereign sensitivity analysis: drop in CAR at the system level by 13.5 percent compared to end-2022 in the simulated domestic debt exchange.
- Credit concentration risk:
  - CAR baseline scenario in 2023: 50.9 / 52.3 / 53.9 / 55.0 (reiterated)
  - CAR two-notch downgrade of 5 largest borrowers: 50.9 / 46.7 / 48.0 / 49.0
  - CAR downgrade to loss of 5 largest borrowers: 50.9 / 20.6 / 21.1 / 21.6
  - Number of banks failing under loss of 5 largest borrowers 5
  - Pct of banking system <12% CAR 27.1
  - Capital shortfall in % of GDP 0.7

### Observations on vulnerabilities and regulatory limits
- Severe scenario outcomes:
  - The assumed domestic debt exchange in the severe scenario causes the system’s CAR to drop by as much as 12.1 percentage points (narrative), with one bank accounting for 3.3 percent of system assets becoming undercapitalized.
  - As a result of sharply falling capital positions under the severe scenario, several banks would breach the prudential single exposure limit of 15 percent of regulatory capital (30 percent for exposures backed by collateral valued at least at 150 percent of the loan amount, 40 percent for corporate groups).
  - In the hypothetical crisis event, four banks accounting jointly for 30 percent of system assets would breach the applicable single exposure limit.
- Non-zero RW on government securities in FX:
  - Imposing a 100 percent RW on domestic FX securities would reduce system CAR by about 9 percentage points, with one additional bank exhausting much of its excess capital buffer when applied to the severe scenario.
  - A non-zero RW on government securities in MVR would also have a considerable impact given banks' large holdings.

### Credit concentration and indirect FX credit risk
- Credit concentration:
  - Migration of the five largest debtors from pass to substandard did not make any bank violate capital requirements.
  - Migration of the five largest debtors from pass to loss caused five banks jointly accounting for 27.1 percent of system assets to become undercapitalized.
- Indirect credit risk from FX loans:
  - Considerable but not quantified due to lack of borrower-level data.
  - Historical note: Parallel market rate rose to around 20 MVR/USD from its usual range of 17-17.5 MVR/USD at the beginning of the pandemic, causing potential unhedging of FX borrowers and requiring forbearance measures (e.g., payment moratorium).

### Bottom-Up stress test results and data/format issues
- Bottom-up submissions received from seven of eight banks.
- Findings:
  - Several banks expect a pronounced drop in CAR in the severe scenario, but all banks would remain adequately capitalized under their own projections.
  - Projected NPL increases broadly align with top-down estimates; in the severe scenario most banks’ NPL ratio edges up to a similar degree, although two banks expect their rate to more than double.
  - Banks expect continued high profitability under the baseline and a deteriorating bottom line under stress; return on assets may be cut about in half at most banks and two banks expect negative profits.
- Issues affecting comparability and analysis:
  - Presentation differences (metrics and formats).
  - Inconsistent data submissions (missing levels, partial scenarios).
  - Units inconsistent across banks (units, millions, thousands).
  - Initial levels for 2022 differed from supervisory data provided by MMA.

### Liquidity stress tests — cashflow-based methodology
- Cashflow-based liquidity stress test assesses resilience to sudden deposit outflows using liquid assets.
- Run-off rates applied across funding sources and maturity buckets, broken down by currency, modelled after the largest monthly deposit withdrawals since the Global Financial Crisis (GFC).
- Roll-off rates applied to assets by type and maturity in line with standard liquidity stress test practice:
  - Example: only 50 percent of loans maturing within 7 days were assumed to be available, with roll-off rates progressively declining for longer maturities.

*Source: IMF staff estimates (content unit: 1mdvea2024001 - 18.      The projected capital adequacy ratios are obtained through the aforementioned).*

### 33.      Banks also dispose of counterbalancing capacity that can be used if liquid assets are

### 33.      Banks also dispose of counterbalancing capacity that can be used if liquid assets are

### Counterbalancing capacity and stress test assumptions
- Cash, bank reserves (we assume excess reserves and half of required reserves to be available) and shorter-term securities were included in banks’ counterbalancing capacity at current value.
- Sales of other securities were subjected to different haircuts that may materialize in fire sale conditions (especially as   there is no active secondary market).
- The run-off and roll-off rates for the different categories and maturity buckets are displayed in Table 6.
- During the pandemic, MMA lowered the rate of required reserves in MVR from 10 percent to 7.5 percent and that on required reserves in FX from 10 percent to 5 percent. For simplicity, it is assumed that half or required reserves in either currency would be available as counterbalancing capacity under stress.

### Cashflow-based liquidity stress test results (key findings)
- Overall: Maldivian banks have ample liquidity, but a couple of banks might face difficulties.
- Aggregate: Banks are resilient to overall cash outflows and have sufficient counterbalancing capacity to meet aggregate deposit outflows (Table 7).
- Local currency (LC) results:
  - Only one bank may have insufficient liquidity buffers in the very short term (less than 7 days).
  - Funding gap in that bucket: 182 million MVR (or 3.1 percent of total assets).
  - The outflow under the LC liquidity stress is counterbalanced mostly by cash and reserves.
- Foreign currency (FC) results:
  - Liquidity buffers at one bank are found to be insufficient to meet potential foreign exchange-related liquidity pressures in the long run (over 5 years).
  - Funding gap in that bucket: 605 million MVR (or 4.2 percent of total assets) stemming from long-term borrowing pressures.
  - The outflow under the FC liquidity stress is counterbalanced mostly by cash and reserves and securities.
  - Methodology applies a 100 percent run-off rate to long-term borrowings, even if these are from the same banking group.
- Caveats:
  - More granular categories by maturity buckets would help differentiate the funding sources for a more precise identification of the gap mismatches.
  - A similar caveat applies to the asset side.

### Cashflow run-off and roll-off rates (selected exact rates from Table 6)
- Assets (example rows, run-off/roll-off across maturity buckets):
  - Debt Securities: 100% across all maturity buckets including Unclassified.
  - Loans & Advances: 50%, 50%, 50%, 30%, 30%, 30%, 10%, 10%, 10% (across Less than 7 days to Unclassified).
  - Repurchase Agreements: 100% across all maturity buckets including Unclassified.
  - Accrued Interest Receivable: 100% across all maturity buckets including Unclassified.
  - Shares and Other Equity Investments: 100% across all maturity buckets including Unclassified.
- Liabilities (example rows, run-off/roll-off across maturity buckets):
  - Demand Deposits (run-off): 10%, 5%, 1%, 1%, 0%, 0%, 0%, 0%, 0% (Less than 7 days to Unclassified).
  - Savings Deposits (run-off): 40%, 30%, 20%, 10%, 5%, 5%, 5%, 5%, 5%.
  - Time Deposits (run-off): 40%, 30%, 20%, 10%, 5%, 5%, 5%, 5%, 5%.
  - Import Deposits and Margin Accounts: 100% across all maturity buckets including Unclassified.
  - Borrowings: 100% across all maturity buckets including Unclassified.
  - Demand Deposits (roll-off): 20%, 10%, 5%, 5%, 0%, 0%, 0%, 0%, 0%.
  - Savings Deposits (roll-off): 30%, 15%, 15%, 10%, 5%, 5%, 5%, 5%, 5%.
  - Time Deposits (roll-off): 30%, 15%, 15%, 10%, 5%, 5%, 5%, 5%, 5%.
  - Import Deposits and Margin Accounts (roll-off): 100% across all maturity buckets including Unclassified.

### Liquidity Coverage Ratio (LCR) and related results
- Method:
  - Basel III LCR calculated using government securities as proxies for HQLA.
  - Standardized Basel stress scenario factors for haircuts, outflows and inflows were applied.
  - Because Maldives has no liquid secondary markets, HQLAs were proxied by central government securities.
  - Since 98 percent of bank securities investments are in central government securities (and the remaining 2 percent in securities issued by non-residents), all of those investments (both in local currency and FX) were deemed HQLA.
  - Only securities that mature within 30 days were included as HQLA.
  - The Basel III Net Stable Funding Ratio (NSFR) could not be calculated due to insufficient information at MMA for the computation of stable funding.
- Results:
  - Banking sector as a whole would be compliant with Basel III LCR.
  - LCR ratios:
    - Basel III Liquidity Coverage Ratio, MVR: 253.8
    - Basel III Liquidity Coverage Ratio, USD: 235.1
  - Dispersion:
    - The lowest ratios calculated for any bank stood at 157.1 percent (LC) and 79.9 percent (FC), indicating substantial dispersion across the banking system.
  - One bank fails the test in FC.

### Deposit concentration sensitivity analysis
- Method:
  - Analysis of full deposit withdrawals of each bank’s five largest depositors.
  - Banks are allowed to use their cash inflow within 7 days to cover deposit withdrawals by large depositors.
  - Analysis done separately on banks’ local currency and FX balance sheets due to potential conversion difficulties.
- Results:
  - Seven banks (49.9 percent of total banking sector assets) would not be able to withstand runs on either local currency or FX deposits by five largest depositors.
  - This suggests an inordinate deposit concentration in the system.
  - Some banks contend that certain deposits may not be withdrawn as they represent cash collateral that must be maintained for as long as the loan exposure exists.
- Summary metrics (from Table 8 / Figure):
  - Number of banks failing cashflow-based test in either currency: 2
  - Banks failing the cashflow test in either currency in % of system assets: 25.3
  - Banks failing the cashflow test in either currency in % of GDP: 23.3
  - DEPOSIT CONCENTRATION RISK: Number of banks failing test in either currency: 7
  - Assets of banks failing the test in % of banking system assets: 49.9
  - Assets of banks failing the test in % of GDP: 46

### Other sensitivity analyses — Interest rate and foreign currency risk
- Interest rate risk:
  - Assumed symmetric increase on interest-bearing assets and liabilities with residual maturities of up to one year by 200 basis points.
  - Result: moderate drop in capitalization of 1.2 percentage point (compared to the baseline scenario projection for end-2023) on account of a maturity mismatch between assets and liabilities.
  - Maturity profile: 58 percent of interest-bearing assets maturing within one year; 91 percent for liabilities.
  - INTEREST RATE REPRICING RISK: CAR after increase of interest rates by 200 basis points: 51.1 (in percent of RWAs).
- Foreign currency risk in balance sheets:
  - Assumed devaluation by 20 percent in a severe scenario.
  - Most banks: moderate changes in the Net Open Position (NOP).
  - Two banks breach the upper limit to the NOP after the shock.
  - A large bank that is short in FX would see its NOP become even more negative and close in on the lower single currency limit.
  - FOREIGN CURRENCY RISK: Number of banks breaching NOP limits after 20% devaluation: 2

### Climate risk analysis — physical risk context, scenarios, and methodology
- Physical climate vulnerability (key facts and historical events):
  - 80 percent of Maldives’ islands are less than one meter above sea level.
  - 90 percent of the islands report flooding annually.
  - 97 percent of the islands report shoreline erosion.
  - 64 percent of the islands report severe erosion.
  - 50 percent of the housing structure are within 100 meters of the coastline.
  - Historical events table (selected entries):
    - 1987 Flood: Persons affected 300; Damages 15,454 ('000 USD, Affected Atolls not fully listed).
    - 1991 Storm: Persons affected 23,849; Damages 64,465.
    - 2007 Flood: Persons affected 1,649; Affected Atolls Dh, GDh, L, Th.
    - 2019 Flood: Persons affected 1,800; Affected Atolls GA, HDh, L, Th.
    - 2021 Storm: Persons affected 1,320.
  - Most common climate-related events are floods, with coastal-related events as the most destructive and extreme precipitation events the most frequent.
  - Future climate emergencies likely to occur in an environment of reduced buffers (fiscal and FX), higher interest rates, reduced economic prospects, and much higher levels of public and external debt.
- Climate scenarios used:
  - IPCC AR6 scenarios referenced: SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5.
  - Analysis corresponds to SSP2 (Middle of the Road), SSP3 (Regional Rivalry), and SSP5 (Fossil-fueled Development).
  - Sea level rise data considers all General Circulation Models (GCM); storm surge data considers three GCMs selected to better depict windspeed conditions in the Maldives.
- Sea level rise projections (key points):
  - Sea level rise could reach critical levels by the end of the century with no adaptation.
  - Median sea level rise for mid-of-century does not have significant differences between scenarios; end-of-century levels can reach increases above 1.5 meters.
  - For mid-century, sea level rise ranges reported between 0.14 to 0.62 and an average of the median between the SPP2, SSP3, and SSP5 scenarios of 0.25 (change in mm; location Male).
  - Low-likelihood high-impact scenarios: SSP5 RCP8.5 between 0.16 - 0.62; SSP1 RCP2.6 between 0.12-0.46.
  - With 80 percent of the land mass below a 1-meter elevation and newly developed areas with a 2-meter height, low-confidence, high-impact scenarios for the end of the century could be catastrophic for the country.
- Methodology for climate risk analysis:
  - Estimates a shock to physical capital from coastal floods under different climate scenarios, using hazards projections, exposure, and vulnerability aggregated to administrative Atolls.
  - Hazards projections built by combining storm surges (Biffis & Wang, 2022) and sea level rise.
  - Exposure uses country-level GDP disaggregated by industry, with proxy variables from the Maldives Bureau of Statistics (MBS) and the Ministry of Tourism (MoT).
  - Vulnerability considers damage functions and the flooded fraction of the Atoll.

*Source: IMF staff estimates.*

### 42.      The analysis considers a micro approach that captures the sensitivity of banks'

### Physical Risk Analysis: Banks' Sensitivity to Coastal Flooding

### Methodology
- Micro approach capturing sensitivity of banks' portfolios to coastal flooding tail events through direct industry damages; analysis considers only physical risk.
- Due to unavailability of standardized geographical exposure data for the banking sector, the impact is modeled via a by-sector damage rate applied to banks' physical immovable assets exposure (not geographically differentiated).
- Indirect impacts (through households' and corporates' income and reduced capacity to pay creditors) are noted but not quantified in this direct-impact analysis.
- Monte-Carlo simulation used to compute joint country distribution from individual Atolls by sampling from damage-rate distributions and multiplying by corresponding exposure (Appendix 8).
- Shock metric: TVaR ninety-nine of the country distribution (Table 10); simulations: 10,000 simulations.

### Climate Scenarios and Hazard Modeling
- Three future climate scenarios: SSP2-4.5, SSP3-7.0, and SSP5-8.5.
- Hazard parameter for coastal floods = storm surge tail distribution + sea level rise.
- Sea level rise percentiles considered: 5th and 95th; future impacts set to mid-century (2050) and end-century (2100) (end-century not used for banking system impact).
- Storm surge component from downscale data (Biffis and Wang (2022)), with windspeed-height relation from Lin et al. (2010); storm surge grids ~10x10 km (0.1 degrees).
- Sea level rise grids ~100x100 km (1 degree); aggregation: sea level rise aggregated through the mean; storm surge parameters fit a Generalized Extreme Value distribution (GEV) using maximum likelihood (Appendix 7).
- Estimation corresponds to the annual maxima series; associated exceedance probability is annual; impact considered annual.
- Leading risk factor across the three scenarios: sea level rise.

### Exposure and Damage Estimation
- Islands considered: start with all islands from Maldives Land and Survey Authority (MLSA); filter to inhabited and resort islands using The President's Office (2023) — total of 345 islands (22.18 percent of the total islands).
- Climate and elevation data assigned to inhabited and resort islands; flooded fraction parameter built using gridded elevation data (NASA, 2019) by estimating percentage of elevation grids within an Atoll lower than coastal flood depth.
- Economic exposure disaggregated by industry using spatial proxies:
  - National GDP disaggregated by economic activity/industry from MBS; major components: Tourism (~26%) and Transportation (~11%).
  - Geographic proxies: Census 2014 island-level population and Atoll-level industrial-islands population (MBS); accommodation variable from MoT = sum of resort and hotel beds by Atoll + guesthouses (guesthouses considered as a unit).
  - GDP disaggregation used as percentage of capital stock exposure.
  - Greater Male area and Kaafu Atoll represent main exposures.
- Damage functions: Huizinga et al. (2017) for Asia used to estimate differentiated damage rates by coastal flood height for each Atoll; estimation done for three scenarios and six return periods.
- Resulting estimation produces five differentiated damage rate distributions by Atoll and scenario for mid and end-century, mapped to industries in the exposure.

### Transmission to the Banking System
- Focus on immovable asset-related loans: construction, real estate, and tourism categories used; these categories represent 60 percent of the total loan portfolio in December 2020.
- Assumptions for loss recognition:
  - No insurance coverage assumed.
  - Recovery rate assumed at 5 percent.
  - Impacted performing loans provisioned to a level of 95 percent.
  - For affected non-performing loans, provisions correspond to the excess over banks’ current coverage ratio relative to the target of 95 percent.
- Industry damage rates weighted by loan categories related to immovable assets to derive bank-level immovable asset damage rates and to adjust performing and non-performing loans per institution.

### Results
- Mid-century (2050) direct coastal flood damages:
  - Could reach 3.24 percentage damage to capital stock for SSP2 4.5 and SSP5 8.5 in the 99th percentile of the loss distribution with the 95th percentile of sea level rise.
  - The 95th percentile sea level rise increases damage by around 0.5 percentage points compared with the median sea level rise.
  - Mid-century estimates show an increase in damages, with respect to historical conditions, from 53-60 percent under different climate scenarios for the 95th percentile sea level distribution.
  - Damages could reach up to 3.37 percent of the capital stock (mid-century coastal flood range reported as 3.24 to 3.37 percent).
- End-century projections (not used for banking-system impact due to high uncertainty):
  - Damages suggest an increase from 130-734 percent in SSP5 8.5 and 80-153 percent in SSP2 4.5.
  - Loss distributions display fatter tails as climate parameters become more severe from mid to end-century (Appendix 9 and 10).
- Banking-sector capital impact (direct, mid-century, no insurance):
  - Increase in loan loss provisions leads to a 0.5 percentage point decline in the banking system’s CAR (up to 0.9 percentage points at individual banks).
  - Overall impact on the financial sector from direct damages to immovable-asset–backed loans: mild.
- Additional risks and caveats:
  - High uncertainty about climate data and estimates.
  - Limited data to model geographical exposures.
  - Possible indirect damages and spillover effects in Maldives’ tourism-dependent economy could lead to higher impacts when considered.
  - Local insurance market may face higher reinsurance prices in short term and limited or no reinsurance access in the long run, potentially leading insurers to limit or not offer coverage; absence of reinsurance would force banks to bear climate-related losses directly.

### Recommendations
- Improve data granularity and coverage for climate risk analysis:
  - Authorities should collect further data to better understand implications of climate-related events for the economy and financial system and enhance data access across agencies.
  - Ministry of Environment (MoE) should expand hazard coverage beyond coastal floods and promote access to locally available data and country-specific studies.
  - Develop geographical granularity for country and financial system exposures; involve MBS for country statistics and Maldives Monetary Authority (MMA) for the financial sector (banking, insurance, other supervised institutions).
  - MBS, MMA, and other relevant agencies should collaborate to deepen understanding of vulnerability (relationship between event intensities and losses).
  - Authorities should consider effects on households, government relief transfers, insurance claims, and reinsurance conditions to better quantify direct and indirect impacts.
- Initiate climate risk analysis collaboration:
  - MMA should begin efforts to develop a framework for climate risk analysis as further data become available.
  - Leverage global datasets while prioritizing local sources as they become available.
  - Prioritize analysis for industries with larger effects on the Maldivian economy and financial system.

*Source: IMF staff estimates.*

### Appendix III. Risk Assessment Matrix

### Appendix III. Risk Assessment Matrix

### Systemic regional instability
- Likelihood: Medium
- Impact of Risk: High
- Key risk drivers and transmission:
  - Financial market dislocations and adverse cross-border spillovers.
  - Widening of the parallel market exchange rate spread and increase in sovereign debt stress.
  - ‘Twin-crisis’ risks from a large sovereign-bank nexus, risks to the exchange rate peg, and financial sector exposure to regional sovereigns.
  - Banks’ sovereign debt exposure has risk weights (RWs) of zero and sovereign debt represents roughly 30 percent of banks’ total assets.
  - Maldives is a highly dollarized economy and financial system; abandoning the peg, or re-pegging at a depreciated rate, would exert significant pressure on banks and the economy.
- Policy response:
  - Reduce domestic financing needs and adjust the RW on banking sector sovereign debt exposure.
  - Reduce excess liquidity in the system and implement measures to reduce the level of dollarization in the economy.

### Intensification of military conflicts, commodity price shock, and supply chain disruptions
- Likelihood: Medium
- Impact of Risk: High
- Key risk drivers and transmission:
  - Emergence or escalation of regional conflict(s) and resulting economic sanctions that exacerbate disruptions in trade, remittances, tourism, FDI, and financial flows.
  - Decreased domestic growth, tightened financial conditions, and impaired confidence.
  - Slower growth and damage to business confidence adversely impact investment and consumption.
  - Global financial tightening threatens FX stability and external debt sustainability.
  - As risk of sovereign distress increases, banks face spillover risks through the sovereign-bank nexus.
- Policy response:
  - Extend support measures: (i) asset purchase program to support liquidity in bond markets and compress risk premia; (ii) provide liquidity support to banks.
  - Implement fiscal consolidation to reduce financial needs and adjust banks’ RWs on sovereign debt exposures to mitigate sovereign-bank nexus risks.

### Abrupt global slowdown or domestic economic shock (tourism shock)
- Likelihood: Medium
- Impact of Risk: High
- Key risk drivers and transmission:
  - A global downturn or a domestic security incident leading to a shock to tourism would reverberate through trade and financial channels.
  - Medium impact from key source countries could put pressure on FX reserves and the budget deficit.
- Policy response:
  - Provide targeted and temporary fiscal support to the most vulnerable and hardest-hit workers and businesses.
  - If necessary, keep financial conditions accommodative to support banks.

### Extreme climate events
- Likelihood: Medium
- Impact of Risk: High
- Key risk drivers and transmission:
  - More severe than expected damage to infrastructure and loss of lives and livelihoods, amplifying supply chain disruptions and inflationary pressures, causing water and food shortages, and reducing medium-term growth.
  - Disruption in tourism and other economic activity, damage to properties from coastal flooding, higher food and energy inflation weakening debtor capacity to repay obligations.
- Policy response:
  - Establish contingent financing plans with development partners.
  - With clear criteria of exceptional circumstances, provide targeted and temporary assistance to affected groups and sectors.
  - Prioritize public investment in disaster-resistant infrastructure and sustainable growth.
  - Identify financial sector exposures to physical risks from climate change.

*Source: IMF staff (Appendix III).*

### Banking Sector: Solvency Risk (Stress Testing Framework)

### Institutional perimeter and data
- Institutions included: All 8 domestic banks
- Market share: 100 percent of banking system assets
- Data and baseline date: Supervisory data as of December 2022

### Channels of risk propagation and methodology
- Methodology: IMF stress testing framework
- Satellite models for macro-financial linkages:
  - Model-based forecast of main macroeconomic and financial variables.
  - Satellite panel data models for bank-level NPL ratios.
  - Bottom-up stress test: Banks used own models or expert judgment to estimate impact on NPL ratio, income and CAR.
- Stress test horizon: 3 years (end-2022 to end-2025).

### Scenario design (Tail shocks)
- Macro scenarios: baseline, moderate, severe (variables include tourist arrivals, bed-nights, real GDP growth, exports, imports, lending rate, consumer inflation).
- Baseline scenario: Latest baseline scenario of the APD country team.
- Moderate scenario: Latest adverse scenario of the APD country team built on an economic downturn in advanced countries important to Maldives.
- Severe scenario: Bespoke adverse scenario involving significant economic deterioration, depreciation of the MVR against the U.S. dollar (i.e., abandoning the peg), leading to a twin-coupled period of extreme sovereign stress and debt restructuring.

### Sensitivity analysis and calibrated shocks
- Credit Risk:
  - Derived from projection of NPL ratio based on satellite models with regressors including inflation and import growth, and assumptions about credit growth.
  - Areas with increased NPLs: local currency loan portfolio (mostly household and SME loans), foreign currency loan portfolio (mostly corporate loans), default of largest 5 borrowers.
- FX-induced Credit Risk (severe scenario currency devaluation):
  - Analyze effect on loans to unhedged borrowers in FX (data permitting).
- Sovereign Risk (severe scenario domestic debt exchange):
  - Assumes restructuring of domestic debt where T-bills and any other outstanding domestic government bonds (excl. MMA) are converted into a new bond on financial terms comparable to that given to the MMA in the most recent securitization of MMA claims: 20-year maturity at an interest rate of 2.5 percent, equating to a 24.9 percent value loss applied as a haircut to the face value.

### Risks and buffers, behavioral adjustments
- Risks/factors assessed:
  - Total credit losses, pre-provision profit, repricing gap, shocks to credit quality of sectoral and large exposures, domestic debt exchange (DDE) losses, losses from maturity and currency mismatches.
- Behavioral adjustments:
  - Assumptions for pre-provision income growth and credit growth in scenarios, income taxes paid, dividend payouts, maturity mismatch related repricing gap, NOP upon devaluation.

### Regulatory and market-based parameters
- Calibration of risk parameters:
  - Projections of bank-by-bank NPL ratios from satellite panel regressions combined with projections for credit growth, pre-provision income, income taxes, dividends, and regulatory information to project banks’ solvency positions.
- Regulatory/Accounting Standards:
  - Minimum CAR hurdle rate: 12 percent

*Source: IMF staff (Banking sector solvency risk).*

### Banking Sector: Liquidity Risk

### Institutional perimeter and data
- Institutions included: All 8 domestic banks
- Market share: 100 percent of total banking sector assets and liabilities
- Data and baseline date: Supervisory data as of December 2022

### Methodology and channels
- Methodology: Cash-flow-based using maturity buckets
- Assumptions: Withdrawals of five largest deposits of each bank used in propagation

### Risks and buffers
- Risks: Funding liquidity shock
- Buffers: Counterbalancing capacity

### Tail shocks and calibration
- Size of the shock: Assumptions for run-off rates on funding sources and roll-off rates on assets to estimate the funding gap based on historical volatility.

### Regulatory standards and reporting
- Hurdle metrics: funding gap used for pass/fail result
- Output presentation: Funding gap by bank, aggregated; number of banks found illiquid and their aggregate share in system assets

*Source: IMF staff (Banking sector liquidity risk).*

### Banking Sector: Market Risk

### Tail shocks and sensitivity analysis
- Interest Rate Risk (Income Effect):
  - Parallel shifts in interest rates of interest-bearing assets and liabilities using maturity buckets supplied for the liquidity stress test.
- Currency Risk:
  - Severe scenario (currency devaluation): analyzed effect on Net Open Position (direct currency risk effect).

*Source: IMF staff (Banking sector market risk).*

### Banking Sector: Climate Risk Analysis

### Institutional perimeter and data
- Institutions included: All 8 domestic banks
- Data sources:
  - Islands shapefiles from the MLSA
  - Population by island, and industrial population by island from the MBS
  - Tourism sector data: number of beds for hotel and resorts, and number of guesthouses by administrative Atolls from the MoT
  - Projection of coastal flood levels considering: Storm surge estimated from (Biffis & Wang, 2022); Sea level rise data from IPCC AR 6
  - Elevation data from ASTER-DEM NASA
  - Supervisory data as of December 2022

### Methodology
- Micro approach on sensitivity of banks' portfolios to coastal flood tail events.
- Hazards: coastal floods modeled from storm surge tail distribution plus sea level rise from future climate scenarios SSP2-4.5, SSP3-7.0, and SSP5-8.5.
- Exposure: disaggregated economic activity by industry using spatial explicit proxies as a proxy of capital stock geographic distribution.
- Vulnerability: Huizinga et al. (2017) floods damage function, relating flood depth to percentage damage to physical assets, differentiated by industry.
- Country losses: computed by applying industry damage rates by administrative Atolls weighted by corresponding capital stock distribution, considering geographical interactions between administrative Atolls with a Monte Carlo simulation assuming administrative Atolls as independent.
- Bank impact: computed by applying damages by affected industries to immovable asset-related portfolios and calculating additional provisions assuming no insurance coverage.

*Source: IMF staff (Climate risk analysis).*

### FSAP Macro Variables: Baseline, Moderate, and Severe Scenarios (In percent; arrivals and exports growth as provided)

- Real GDP growth (Est. 2022; Proj. 2023 2024 2025)
  - Baseline: 12.3 7.2 5.7 6.5
  - Moderate: 12.3 -14.3 13.4 10.4
  - Severe: 12.3 -24.4 8.8 10.5

- Inflation rate (eop)
  - Baseline: 3.3 5.1 2.5 2.1
  - Moderate: 3.3 5.8 4.2 2.4
  - Severe: 3.3 9.5 5 2.6

- Bank lending rate
  - Baseline: 10.9 10.9 10.9 10.9
  - Moderate: 10.9 12.5 11.2 10.9
  - Severe: 10.9 16.3 13.2 10.9

- Exchange rate (MVR per USD)
  - Baseline: 15.4 15.4 15.4 15.4
  - Moderate: 15.4 15.4 15.4 15.4
  - Severe: 15.4 18.5 18.5 18.5

- Credit growth
  - Baseline: 11.7 9.8 8.7 8.9
  - Moderate: 11.7 6.6 13.9 10.7
  - Severe: 11.7 2.2 10.4 9.5

- Fiscal balance (percent of GDP)
  - Baseline: -11.1 -9.1 -8.5 -5.7
  - Moderate: -11.1 -18.2 -13.9 -9.6
  - Severe: -11.1 -22.1 -19.5 -15.1

- Total Arrival (thousand)
  - Baseline: 1675.26 1765.199 1899.109 2140.794
  - Moderate: 1675.26 753.752 1214.506 1602.94
  - Severe: 1675.26 676.116 893.058 1221.722

- Exports Growth (percent)
  - Baseline: 35.952 4.122 6.295 6.945
  - Moderate: 35.952 1.86 6.095 6.708
  - Severe: 35.952 -0.403 5.909 6.486

*Source: IMF staff estimates (Appendix V).*

### Bottom-Up Stress Test: Instructions and Assumptions (Summary)
- Banks projected key variables for a three-year period through end-2025 under baseline, moderate, and severe scenarios:
  - Loans, other assets (total assets); deposits and other liabilities (total liabilities)
  - NPLs and NPL ratio
  - Loan loss provisions on additional NPLs
  - Pre-provision income (net interest income, net fee income, net other income)
  - Pre-tax income (pre-provision income and provisioning)
  - Income tax paid, post-tax income and effective tax rate applied
  - Dividend payout and payout ratio
  - Change in retained earnings relative to retained earnings at end-2022
  - Risk-weighted assets
  - Regulatory capital (Tier 1, total capital), capital adequacy ratio, leverage ratio (MMA definition)
- Banks required to explain methodology in a short note.
- Assumptions:
  - No deleveraging permitted (e.g., asset sales to generate additional revenue).
  - Provisions on additional NPLs must increase with scenario severity.
  - Pre-provision income should be lowest under the severe scenario (moderate lower than baseline).
  - Banks to apply realistic income tax payment under each scenario.
  - Dividend payouts should be lower under moderate and severe scenarios than baseline.

*Source: IMF staff (Appendix VI).*

### Climate Data, Monte-Carlo, and Loss Distributions (Highlights)
- Climate data treatment and Monte-Carlo simulation processes were used for climate risk analysis (Appendices VII–VIII).
- Loss distributions and impact of coastal floods on capital stock computed with 10,000 simulations (Appendices IX–X).
- Percentile and Average Annual Loss statistics were produced for historical and future scenarios (SSP2-4.5, SSP3-7.0, SSP5-8.5) across multiple metrics including percentage damage to capital stock and damages as percentage of GDP.
- Appendix XI reports historical impact of previous events: Flood (1987), Storm (1991), Tsunami (2004) with damages as percentage of GDP:
  - Flood (1987): 4.25
  - Storm (1991): 12.28
  - Tsunami (2004): 38.32

*Source: IMF staff (Appendices VII–XI).* 

*Appendices XII lists Administrative Atolls Codes and Names and References (IMF staff and external sources).*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1mdvea2024001.pdf_