## Executive Summary

## Source details

**Canonical URL:** [Executive Summary](https://www.imf.org/-/media/files/publications/tar/2025/english/tarea2025036-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/tar/2025/english/tarea2025036-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/tar/2025/english/tarea2025036-print-pdf.pdf.json)

---

### Missions and scope
- Two CARTAC missions to the Turks and Caicos Islands Financial Services Commission (TCIFSC):
  - First mission: January 29 – February 2, 2024 — reviewed the TCIFSC Financial Stability Report (FSR), discussed financial stability indicators and essential elements of climate risk, and covered credit risk modeling as a precondition for new stress testing (ST).
  - Second mission: March 11–15, 2024 — focused on constructing a new multi-factor and multi-period bank solvency ST tool based on macroeconomic scenarios.
- Financial system context:
  - Financial system size: 275% of GDP.
  - Commercial banks share: 65% of total financial system assets.

### Assessment of the TCIFSC Financial Stability Report (FSR)
- Overall assessment:
  - The TCIFSC’s FSR has been published since 2015 and is assessed as relatively good.
  - The report provides overview, motivation, macro-financial narrative, discussion on the real sector, coverage of the entire financial system (banks, investment businesses, domestic insurance companies, trust companies, money service businesses), results of banking ST, and key initiatives.
- Suggested improvements to content and structure:
  - Streamline the FSR to organize text coherently around the central risk story.
  - Focus on key risk drivers and their transmission channels.
  - Provide fewer nominal figures and use relative indicators instead; move some statistics to annexes.
  - Add mission/theme statements for paragraphs to guide readers; consider a foreword from the Managing Director to spell out key messages.
  - Integrate chapters on systemic risk and banking sector resilience into the financial sector development chapter.
  - Make the financial sector developments chapter less descriptive and connect it to financial stability risks.
  - Shorten the macro-financial environment chapter to focus more directly on the TCI economy.
  - Add discussion on emerging risks, such as climate and cyber risks.
- Regulatory and analytical enhancements:
  - Discuss the relevant regulatory framework for different financial system segments within the FSR to make indicators self-explanatory (critical for insurance sector solvency indicators where market valuation frameworks matter).
  - Enhance background financial stability analysis by adding: average probabilities of defaults (PDs), loss given defaults (LGDs), collaterals (e.g., loan-to-value (LTV) ratios), and for non-banks include duration to assess market risk.
  - Reflect on IFRS17 regulatory framework details (discounting of liabilities, types of reinsurance contracts).
  - Consider hiring one person proficient in actuarial science.

### Credit risk modeling
- Work completed:
  - Several credit risk models estimated to project the impact of alternative macroeconomic scenarios on banks’ non-performing loans (NPLs).
  - Introduced Bayesian Model Averaging (BMA) to address uncertainty in modeling sectoral NPLs.
  - Delivered an R script with initial estimates for four sectoral credit risk models:
    - Loans to households
    - Loans for the construction and land development sector
    - Loans to the tourism sector
    - Other corporate loans
- Model variables and coordination:
  - Initial model explanatory variables include tourist arrivals, inflation, unemployment, and real GDP.
  - Discussion initiated with the Ministry of Finance to generate baseline and adverse scenarios based on an existing macroeconomic model including six variables: tourist arrivals, inflation, real GDP, imports, real estate sector GDP, and government revenues.
  - Note: unemployment is not part of the Ministry of Finance macroeconomic model but can be forecasted directly by the TCIFSC.
  - Recommendation to check alternative model specifications using variables generated by the Ministry of Finance and to investigate historical/current NPL write-off practices to improve projections.

### Solvency stress testing framework
- New framework developed:
  - Multi-factor and multi-period solvency ST tool tailored to TCI banks, respecting existing accounting, tax, and regulatory rules.
  - Framework links common explicit macroeconomic scenarios (one baseline and two adverse) to NPL satellite models to project NPLs and credit losses.
  - Produces scenario-specific macroeconomically-consistent projections of institutions’ key balance sheet, profit & loss, and capital adequacy items over up to three years.
- Data, training, and operationalization:
  - Reviewed available data and regulatory framework; constructed tool and provided staff training.
  - TCIFSC staff need to become familiar with the ST framework to update it with new macro-financial and institution-specific data, calibrate scenarios, and adjust assumptions/parameters.
  - Strengthen cooperation with the Ministry of Finance on scenarios; develop new satellite models; regularly publish and share ST results with bank supervisors to support risk-based supervisory practices.

### Illustrative stress tests and results
- Inputs used:
  - Illustrative stress tests run using December 2023 data.
  - Baseline scenario: official TCI Ministry of Finance macroeconomic projection as of March 2024.
  - Adverse scenarios: calibrated by expert judgment to capture different severity levels of economic recessions.
- High-level results:
  - Results suggest banks are generally resilient to economic stress given:
    - Relatively high initial capital adequacy.
    - Good pre-provision profitability.

### Processes, communication, and data management
- Enhancements recommended:
  - Production plan for FSR should be discussed before the production cycle and approved by the Managing Director.
  - External communication of the FSR should be regular (press releases, social media, interviews, etc.).
  - Consider establishing a centralized database for all existing and newly collected data sources, including appropriate analytics tools, to support financial stability monitoring and a fully-fledged solvency ST system.

### Recommendations (numbered, with priority and timeframe)
- Definitions: Near term: < 12 months; Medium term: 12 to 24 months.
- Financial Stability Report
  1. The FSR should be streamlined to organize the text more coherently around the central risk story.
     - Priority: High
     - Timeframe: Near-term
  2. Financial stability indicators need to be communicated together with the key aspects of the existing regulatory framework.
     - Priority: High
     - Timeframe: Near-term
  3. New emerging risks, such as cyber and climate risks, should be consistently covered in the FSR.
     - Priority: High
     - Timeframe: Medium-term
  4. The TCIFSC background financial stability analysis could be enhanced.
     - Priority: High
     - Timeframe: Medium-term
- Processes related to financial stability work, communication, and data management
  5. A production plan for FSR should be discussed before the production cycle starts and approved by the Managing Director.
     - Priority: High
     - Timeframe: Near-term
  6. The external communication of the FSR should be regular (press releases, social media, interviews, etc.).
     - Priority: High
     - Timeframe: Medium-term
  7. A centralized database for all existing and newly collected data sources, including appropriate tools for analytics, should be considered to support financial stability monitoring and a fully-fledged solvency stress testing system.
     - Priority: High
     - Timeframe: Medium-term
- Credit risk modeling
  8. The team needs to become familiar with the methodology and corresponding R script provided.
     - Priority: High
     - Timeframe: Near-term
  9. The team should further check for alternative specifications to improve the initial estimates of sectoral credit risk models.
     - Priority: Medium
     - Timeframe: Medium-term
  10. The team should test the models for the macroeconomic scenarios generated by the Ministry of Finance.
      - Priority: High
      - Timeframe: Near-term
  11. Historical and current practices for NPL write-offs should be investigated.
      - Priority: Medium
      - Timeframe: Medium-term
- Stress testing
  12. The team needs to become familiar with the new stress testing framework.
      - Priority: High
      - Timeframe: Near-term
  13. Cooperation between the TCI FSC and TCI MoF in the area of macroeconomic scenarios needs to be developed.
      - Priority: High
      - Timeframe: Medium-term
  14. Conduct regular (at least twice a year) top-down stress test exercises with the new framework, ideally in spring and autumn.
      - Priority: High
      - Timeframe: Near-term
  15. Publish the aggregate stress test results of the spring exercise in the annual FSR and include a description of the new stress test framework as a “research note” in the FSR when the results are published for the first time.
      - Priority: High
      - Timeframe: Near-term
  16. Share and discuss the stress test results with the Bank Supervision team and use them in the conduct of supervision and supervisory reviews.
      - Priority: Medium
      - Timeframe: Medium term
  17. Consider developing additional models for banks’ key items such as loan growth, deposit growth, interest rate, or non-interest income.
      - Priority: Medium
      - Timeframe: Medium term

*Source: Executive Summary, tarea2025036-print-pdf*

---

### Tourism, real estate, and construction: outlook and targets

- Tourism sector expected to maintain positive trend following the decline initiated by COVID-19, conditional on effective implementation of the National Tourism Strategy and Implementation Plan for the TCI (published March 2023).
- Plan quantitative targets:
  - Increase total tourist arrivals from 1.6 million in 2019 to 3.2 million in 2032.
  - Grow tourism revenue from $825 million in 2019 to $1.7 billion in 2032 while stemming the decline in visitor spend.
  - Expand total registered accommodation rooms from 4,008 in 2019 to 6,143 by 2032.
- Connectivity and market diversification objectives:
  - Improve air connectivity from Providenciales to other tourism destinations within the island.
  - Expand marketing in current major markets: USA, Canada, UK.
  - Gain market share in new potential markets in Europe (Germany, Italy, France, Spain, Benelux), Latin America (Brazil, Colombia, Chile, Argentina), and the Caribbean region.
- Real estate and construction: continued significant contributions anticipated, supported by recent completions of various resort and development projects and a robust pipeline of ongoing initiatives.

### Downside risks and resilience
- Key downside risks:
  - Weather-related shocks (high vulnerability due to location in the Atlantic hurricane belt).
  - Recession fears in the US.
  - Ongoing geopolitical tensions (e.g., Europe, Middle East) that could affect commodity prices and keep inflation elevated.
- Fiscal/financial buffers and mitigation:
  - National Wealth Fund stood at $51 million as of 31 December 2023.
  - Government participation in the Caribbean Catastrophe Risk Insurance Facility (CCRIF) to help mitigate hurricane-related costs.
  - Past major storms (e.g., Hurricane Irma in 2017) had impacts mitigated by strong financial reserves.

### Financial sector structure and key statistics (end-2023)
- Financial sector size: 275% of GDP at the end of 2023.
- Distribution of financial sector assets:
  - Banks: 65% of total financial assets; 6 banks.
  - Investment business: 31% (5 investment dealers, 8 mutual fund administrators).
  - Insurance companies: 3%; 18 companies.
- Market concentration and ownership:
  - Banking sector controlled by three Canadian banks (97% of total banking assets) and one British-Belizean bank.
  - Three Canadian banks operate as large retail Systematically Important Banks (SIBs) for the TCI economy.
  - Two other banks are niche players specializing in offshore wealth management.
- Dominant risk types:
  - For largest banks: credit and liquidity are main sources of risk; interest rate risk could also be significant.
- Distribution visual/statistics reproduced:
  - 66.0% — Banks
  - 31.0% — Investment business
  - 3.0% — Insurance companies

### Financial stability risks and transmission channels
- Primary financial stability risk: credit risk linked to a potential drop in tourist arrivals, with implications for banks’ credit portfolios and increased NPLs.
- Additional risks:
  - Non-life insurers vulnerable to climate change risk and rising inflation, possibly facing higher-than-expected claims.
  - External environment, especially US economic conditions, is a major determinant given TCI’s exposure to US tourists.
- Importance of measuring and communicating credit risk and interest rate risk for proper assessment of the financial system.

### FSR best practices and recommendations
- Objectives and definition:
  - FSRs should explicitly state objectives and provide a clear, dynamic definition of financial stability, including operational benchmarks and monitored indicators.
- Executive summary:
  - Should be brief, easy to read, and present main conclusions and evolution of risks since the previous issue; analytical chapters can contain advanced technical material in boxes/annexes.
- Macro-financial integration:
  - FSRs should integrate macroeconomic and financial analysis in both directions, identify transmission channels, and quantify shock reverberations where possible.
- Transparency of assumptions and methodologies:
  - Stress tests and other analyses should clearly justify assumptions (severity, speed of impact, hurdle rates, dividend policies, etc.), with limitations described.
- Data and methodology disclosure:
  - Sources and cut-off dates should be clearly indicated; data used should be made publicly available on the publishing authority’s website; methodologies for projections, sensitivity analyses, and stress tests should be explained, with technical details in boxes/annexes.
  - Reports should consider distributional indicators and, where appropriate, extreme or individual values (with confidentiality safeguards).
- Structure and content:
  - Follow a logical, integrated structure centered on key risks; include executive summary, chapters on external and domestic developments, financial infrastructure and regulatory framework, table of acronyms, glossary, methodological annex, and statistical annex.
- Publication timetable and frequency:
  - Publish relatively shortly after data cut-off (preferably within three months).
  - Frequency should change over the medium term to two times a year to align with best international practices.
  - Publication date should be announced in advance and indicated on the website.

### Organizational and communication recommendations for the TCIFSC
- Production planning and governance:
  - Policy unit should prepare a detailed FSR production plan with regular structural meetings and clear responsibilities; senior management should approve the plan.
  - Final discussion of the FSR in the TCIFSC Board needs to be scheduled before final publication and included in the production plan.
- Internal and external review:
  - Current internal commenters include Directors of Bank & Trusts, Insurance, and ALM Supervision Departments, the Managing Director, and externally the Head of the Statistics Department, Ministry of Finance.
- Communication strategy:
  - Complement production plan with a communication strategy positioning the FSR as the key communication tool for financial stability.
  - Consider setting up a communication team to plan press conferences, media interviews, YouTube videos, press releases (as a rule sent to media and key financial sector representatives), and seminars with market participants/analysts.
  - Include Foreword from the Managing Director in the FSR and use quotes in press materials.
  - Ensure all charts and tables’ underlying data are provided on the website alongside the report.

### Content, structure, and analytical enhancements for the FSR
- Streamlining and risk-focused narrative:
  - Organize text coherently around a central risk story and key risk drivers; reduce reliance on nominal values in the text, use relative indicators or move nominals to a statistical annex.
  - Integrate chapters on systemic risk and banking sector resilience earlier and link them to financial sector developments.
  - Shorten regional macro-financial sections and focus more directly on impacts to the TCI economy.
  - Add discussion on emerging risks such as climate and cyber risks.
  - Consider adding a foreword from the Managing Director.
- Regulatory framework and sector-specific information:
  - Add descriptions of relevant elements of the regulatory framework for different financial segments (critical for insurance solvency interpretation, IFRS17 implications, reinsurance contract types).
- Enhance background analysis:
  - Strengthen credit risk analysis by including average probabilities of default, loss given default, and collateral measures (e.g., loan-to-value ratios).
  - Enhance nonbank analysis by adding critical information such as duration and IFRS17-related clarifications on discounting of liabilities and reinsurance arrangements.

### Strengthening the financial stability analytical toolkit
- Develop econometric models to project key variables: credit volumes to the private sector, nonperforming loans, banks’ interest and non-interest income and expenses.
- Recommended methodology:
  - Bayesian Model Averaging (BMA) presented as a suitable approach to tackle model uncertainty, limited time series in TCI, and to explore links between macro-financial variables and banks’ performance indicators.
  - BMA’s flexibility aids in handling uncertainty in predictions and parameter estimates, variable selection, and dynamic relationships, making it effective for limited-data environments.

*IMF | CARTAC Turks and Caicos Islands | excerpt from chapter detailing short- to medium-term outlook, financial sector structure, risks, and FSR recommendations*

---

### Credit risk modeling work and methodology

### Purpose and model suite
- Purpose:
  - Develop credit risk models to project risk under different macroeconomic scenarios as a precondition for multi-period, multi-factor macro stress testing of the TCI banking sector.
- Sectoral models developed:
  - Four sectoral credit risk models: loans to households (personal loans), loans for the construction and land development sector, loans to the tourism sector, and other corporate loans.
- Dependent variable and transformation:
  - Nonperforming loan (NPL) ratios for the four sectors were employed as the dependent variable.
  - NPL ratios were transformed using a logit transformation to make them suitable for use in linear models.
- Estimation method and model suite:
  - The four credit risk models were estimated using the BMA method.
  - In each sector, between 3-4 models were estimated; the final sectoral projection is a weighted average of the individual model projections.
  - The mission provided an R script with an initial estimate for the four sectoral credit risk models; final set of models for each sector is available in Appendix 1.
- Explanatory variables and data frequency:
  - Considered explanatory variables: tourist arrivals, inflation, real GDP, and unemployment.
  - Data availability: NPL ratios and tourist arrivals are available quarterly; real GDP, inflation, and unemployment are currently available only annually, so quarterly proxies were constructed via interpolation.

### Recommendations for further work
- TCIFSC actions:
  - Further test and improve the initial estimates of sectoral credit risk models.
  - Team members must become familiar with the R script provided.
- Coordination with Ministry of Finance:
  - Further discussion with the Ministry of Finance to generate baseline and adverse scenarios for stress testing based on the existing macroeconomic model, including six variables: tourist arrivals, inflation, real GDP, imports, real estate sector GDP, and government revenues.
  - Note: The Ministry of Finance's macroeconomic model projects tourist arrivals based on US GDP; this then serves to estimate domestic GDP.
- Model extensions and data issues:
  - Unemployment is not part of the Ministry's macroeconomic model but could be forecasted directly by the TCIFSC.
  - Alternative sectoral credit risk model specifications with variables generated by the Ministry of Finance that are not included in the initial estimates should be further checked.
  - Historical and current NPL write-off practices should be investigated to improve projections.

---

### Stress testing (ST) tool: structure and key features

### General framework and scope
- New Excel-based macro ST tool explicitly projects key balance sheet items (performing and non-performing loans across the main loan segments, loan loss provisions, total and risk-weighted assets, and capital) and P&L items (net interest income, other income and expenses, and credit losses) over a horizon of up to three years.
- Tool works with four key loan segments: tourism & entertainment, personal loans, construction, and others.
- Tool can be run on any end-quarter data and generate evolution charts and contribution charts; includes a detailed step-by-step guide and a detailed user manual.

### Accounting, tax, and regulatory alignment
- Projections handle both traditional coefficient-based provisioning by asset quality classes and possible IFRS 9 provisioning.
- Definition of regulatory capital (Tier 1 and Tier 2) fully respected.
- Tool acknowledges the 0% corporate income tax rate for banks.

### Scenario setup and required macro variables
- Scenarios:
  - Baseline scenario: reflect most recent official macroeconomic outlook prepared by the TCI Ministry of Finance.
  - Two adverse scenarios: “moderate” and “severe”, calibrated to capture different intensities of an economic recession.
- Key macroeconomic variables to always be projected for all three scenarios:
  - Real GDP growth, Inflation, Unemployment rate, Tourist arrivals, Income from tourism, US 3M T-bill rate.
- Future macro projections could include a credit forecast to the private sector to calibrate loan portfolio growth and enable dynamic balance sheet sizing.

### How sectoral credit models feed the ST tool and conservatism adjustments
- NPL projection mechanics:
  - Scenario-specific macro projections are used to predict NPLs with the sectoral credit risk satellite models.
  - BMA weighting schemes were adjusted by expert judgment to yield more conservative NPL projections for larger recessions:
    - The weight applied on the most “reactive” NPL model (generating the most significant NPL increases) was set at the highest level for the severe scenario and the second highest level for the moderate scenario (at the expense of the remaining models) compared to the baseline.
- Application of projected NPL changes across institutions:
  - The absolute change in the projected aggregate NPL ratio is applied to each loan segment's starting NPL ratio level.
  - Using absolute rather than relative changes ensures institutions with a 0% starting level of the NPL ratio in some loan segments are hit by new NPLs in adverse conditions.
  - Tool allows additional expert-based adjustments:
    - explicit segment-specific NPL ratio add-ons
    - explicit bank-specific NPL ratio add-ons
- Special-mention-to-performing loans ratio:
  - Tool projects special-mention-to-performing loans ratio for each institution and segment.
  - Projection is linked to the NPL projection using an elasticity specified as a parameter set by expert judgment (changeable).
  - Loan growth projection (calibrated per scenario and applied equally across segments and institutions for simplicity) jointly with credit risk projections determines paths of good, special mention and non-performing loan exposures.
  - Three explicit loan quality classes facilitate future transition to IFRS 9 Stage 1-3 classification.

### Provisioning, write-offs, and other balance sheet assumptions
- Loan loss provisioning logic:
  - Projections based on existing provisioning rules and bank practices.
  - NPL provisioning levels are institution- and segment-specific: starting NPL provisioning rate plus an assumed increase of around 10-20 ppts (cumulative) in both adverse scenarios (typically no change for the baseline scenario).
  - Shock can be set separately for each scenario and year; reflects worsened NPL quality and migration to worse sub-classes or deterioration in collateral value.
  - Same logic applied to special mention (Stage 2) and good (Stage 1) loans—starting observed levels used for baseline, parameter-driven increases in adverse scenarios.
  - Provisions created for good loans are considered general provisions and qualify as Tier 2 regulatory capital up to 1.25% of credit risk-weighted assets.
- NPL write-offs and final provisions:
  - Final loan loss provisions impacting P&L derived from projected stocks, considering scenario- and year-specific NPL write-offs calibrated by expert judgment using qualitative information about bank write-off practices.
  - Even if model projects unchanged NPL stocks, new provisions may still be needed to account for NPL inflows required to keep NPLs stable amid write-offs.
- Other assets, deposits, and funding:
  - Projections for additional financial assets (cash, claims on financial institutions, debt securities, other assets) and deposits can be specified.
  - Calibration of growth rates can follow loan growth assumptions or expert judgment.
  - Evolution of asset items impacts total and risk-weighted assets and net interest income; deposit projection co-determines interest expenses.
  - Calibration must ensure assets equal liabilities (may require iterations).

### Market, FX, concentration, and income assumptions
- Market and FX risk:
  - Market risk from revaluation of securities can be switched on as a sensitivity; almost all securities held are foreign government bonds (typically US treasuries) normally held at amortized cost until maturity.
  - No domestic securities issued; no sovereign risk included.
  - USD as TCI currency and virtual absence of other currencies remove exchange rate risk from balance sheets; FX risk not considered for stress.
- Concentration risk:
  - Exposure to one or more of the largest five borrowers can be stressed with LGD calibrated between 0% and 100%; typically applied as a sensitivity within an adverse scenario.
- Pre-provision income (PPI):
  - PPI projected as interest and non-interest income minus interest and non-interest expenses.
  - Interest income = time-varying average implicit interest rate modeled along scenario-specific development of USD 3M T-bill rates (with pre-defined passthrough) × interest-bearing assets.
  - Interest-bearing assets typically decline in adverse scenarios as performing loans migrate to NPLs.
  - Interest expenses projected using assumed path of deposit funding and deposit interest rates (initial observed level, but can include funding cost shocks).
  - Non-interest income and expenses projected as institution-specific starting points adjusted by haircuts set by expert judgment; can later be linked to macro scenarios via satellite models.

### Capital, assets, RWAs, and reporting
- Capital and dividend treatment:
  - Capital projected as a function of net income; negative net income reduces capital.
  - Positive net income is subject to distribution decision; only retained part after dividend payouts increases capital.
  - Dividend payout assumptions are institution-specific.
- Total assets and RWAs:
  - Total assets = time-varying net loans + other financial assets.
  - Credit RWAs projected as function of net exposures and risk weights for key asset classes; market and operational RWAs kept constant.
  - Total assets and RWAs driven by credit growth, NPL evolution, and provisioning assumptions; typically decline in adverse scenarios.
  - Decline in assets/RWAs can somewhat help capital ratios but is small relative to the impact of economic recessions causing accounting losses.
- Reporting outputs:
  - Key results reported based on capital ratios: overall capital adequacy ratio, simple capital ratio, and Tier 1 ratio.
  - Tool shows contributors to changes in capital adequacy: net interest income (+), other income/expenses (+/-), credit losses (-), dividend payouts (-), change in denominator (risk-weighted assets) (+/-).
  - Reports institutions below a specified hurdle rate for total capital adequacy ratio and their share in sector assets, plus capital injections (expressed as a % of GDP) needed to bring all institutions to at least the minimum capital adequacy ratio of 11%.

### Overview table components (selected items)
- Institutional perimeter: All institutions (6 banks, additional banks can be added).
- Key macro variables for scenario calibration: Real GDP growth, y-o-y %; Inflation, y-o-y %; Unemployment rate, %; Tourist arrivals, ths.; Income from tourism, mil. USD; US 3-month T-bill rate, %.
- Credit risk satellite models: Estimated by BMA as a suite of simple models; dependent variable = logistic transformation of NPL ratio; explanatory variables include autoregressive term, GDP growth, inflation, tourist arrivals, constant.
- Additional credit assumptions that need to be set: Private sector credit growth, y-o-y%; Assumed (annual) NPL write-off rate, %; Change in NPL provisioning from initial level, ppts; Change in special mention loan provisioning from the initial level, ppts; Change in good loan provisioning (general provisions) from the initial level, ppts; Factor of special-mention-to-performing loans absolute increase compared to NPL ratio increase; Bank-specific NPL add-ons; Sectoral NPL add-ons.
- Assumptions for balance sheet items: Annual growth, % for cash, claims on financial institutions, government securities, other securities, other assets (incl. fixed assets), deposits and other funding.
- Concentration risk needs specification of which of 5 largest borrowers fail, in which year, in which scenario, and LGD in %.
- Assumptions for PPI: Passthrough coefficient from 3M USD T-bill rates to US government securities, other securities, claims on financial institutions, loans and advances; Funding shock (change in deposit rates from initial level); Scenario-specific changes in non-interest income and non-interest expense applied on institution-specific initial annual volumes, %; Dividend payout rate; Minimum total regulatory capital ratio.

---

### Illustrative stress test setup and headline scenario parameters

- Data and coverage:
  - Illustrative ST prepared using December 2023 data; all six banks covered.
- Scenario calibration notes:
  - Baseline scenario captures the official MoF forecast as of March 2023.
  - Severe scenario calibrated to reflect a 3-standard deviation shock to GDP growth, assuming a real GDP decline by around 20% at the bottom of the cycle.
  - Moderate scenario captures a milder recession, with GDP growth values set as averages of the baseline and the Severe scenario values.
  - Both adverse scenarios reflect possible deterioration in tourism revenues and a resulting recession similar to the 2009-2011 post-Global Financial Crisis period (drop in GDP by 20%) or the recent 2020-2021 COVID-19 recession (drop in GDP by 25%).
- Loan growth assumptions:
  - Loan growth assumed at 10%, 5%, and 0% annually for the Baseline, Moderate, and Severe scenarios, respectively.

### Illustrative model projections (summary)
- NPL model projections:
  - For loans to households and other corporate sectors: baseline assumes a slight decline in NPLs amid relatively good GDP growth projection and declining unemployment.
  - For the construction and tourism sectors: credit risk models suggest a mild increase even in the baseline.
  - For both adverse scenarios: increase in NPL ratios to high levels experienced in post-GFC times around 2013.
  - No sector-specific shocks or failures of the largest borrowers were assumed in the illustrative ST.

*Source: Mission team calculations; Turks and Caicos Islands Financial Services Commission materials included in the mission report.*

---

### NPL provisioning shocks, projected loan losses, and P&L assumptions

### NPL provisioning shocks and related parameters
- Shock to NPL provisioning rate assumed:
  - Moderate scenario: 15 ppts (cumulatively, with the first-year shock of 10 ppts and the second-year increase of 5 ppts).
  - Severe scenario: 30 ppts (cumulatively, with the first-year shock of 20 ppts and the second-year increase of 10 ppts).
- Average starting NPL provisioning: 47%.
- Resulting provisioning level in Severe scenario: would increase to about 87%.
- Special mention loans provisioning shock: 1 ppts (Moderate) and 5 ppts (Severe).
- General provisions/Stage 1 loans shock: 0.5% (Moderate) and 1% (Severe).
- NPL write-off rate: 10% for all scenarios.

### Projected loan losses and credit loss rates
- Baseline scenario:
  - Credit loss rate (defined as new provisioning booked in the P&L over the initial stock of net loans) hovers around 0.5% annually.
- Severe scenario:
  - Stock of loan loss provisions would increase about four times over the 3-year horizon.
  - Provisions would reach almost 13% of gross loans.
  - Large new provisioning of around 6% per year on average in the first two years of the horizon.

### Income, expenses, and other P&L assumptions
- Pre-provision income serves as first line of defense but is lower in adverse scenarios.
- Net interest income would drop considerably in the Severe scenario due to:
  - Declining interest-bearing assets (driven by lower-performing loans).
  - Lower average loan interest rates due to a large decline in US interest rates.
- Funding shock: none assumed.
- Other asset items, deposits, and funding: assumed to remain at the initial level.
- Non-interest income assumptions:
  - Remain at last year’s baseline level in Baseline scenario.
  - Drop by 15% in the Moderate scenario.
  - Drop by 30% in the Severe scenario.
- Non-interest expenses: assumed to grow somewhat (by 35 annually) in all scenarios.
- No revaluation of securities assumed.

### Bank-level profitability, capital, and resilience
- Profitability outcomes:
  - Four larger banks become loss-making only in the Severe scenario.
  - Two smaller banks experience negative profitability already in the Moderate scenario.
  - In total, all banks would, at one point, experience negative profits.
- Tax and payout assumptions:
  - Assuming no corporate tax.
  - 20% dividend payout rate for all banks.
- Impact on capital:
  - Projected net income would impact the final capital stock.
  - Risk-weighted assets (RWAs) would change following the implied trajectory of net loans and (unchanged) other asset items.
- Capital adequacy outcomes:
  - Aggregate CAR would grow in both the Baseline and the Moderate scenario due to continuing aggregate profitability driven by the largest banks.
  - Only in the Severe scenario would the sector’s CAR drop (faster drop in capital than in RWAs), but it would still remain comfortably above the limit.
  - All banks would remain above the regulatory minimum for capital adequacy of 11% in all scenarios.

### Stress test framework use and calibration recommendations
- Further analysis and fine-tuning of scenario calibration needed before regular use of the ST tool.
- Scenarios should be prepared consistently and underpinned by a good narrative jointly developed by the TCIFSC and TCI Ministry of Finance.
- Calibration of parameters that might need additional analysis and data:
  - Credit growth.
  - NPL write-offs.
  - Passthrough coefficient for interest rates.

### Operational and data recommendations (linked to stress testing)
- ST team at TCI FSC should become familiar with the new framework to make regular updates; staff must be able to:
  - Update the tool with new macro-financial and institution-specific data regularly.
  - Calibrate scenarios.
  - Adjust additional assumptions and parameters.
- Frequency of stress tests:
  - Use the new framework for regular internal stress tests, ideally twice a year — once in the spring and once in the autumn.
  - Aggregate ST results of the spring exercise should be published in the annual FSR.
  - A “research note” describing the new framework and all key assumptions should be provided in the FSR once the results are published for the first time.
- Use of ST results:
  - Share and discuss ST results with the Bank Supervision team at the TCIFSC.
  - Utilize results in supervisory reviews and overall supervision of individual banks to foster synergies between financial stability and supervision teams.

*IMF | CARTAC Turks and Caicos Islands*

### Executive Summary

### Executive Summary

### Missions and scope
- Two CARTAC missions to the Turks and Caicos Islands Financial Services Commission (TCIFSC):
  - First mission: January 29 – February 2, 2024 — reviewed the TCIFSC Financial Stability Report (FSR), discussed financial stability indicators and essential elements of climate risk, and covered credit risk modeling as a precondition for new stress testing (ST).
  - Second mission: March 11–15, 2024 — focused on constructing a new multi-factor and multi-period bank solvency ST tool based on macroeconomic scenarios.
- Financial system context:
  - Financial system size: 275% of GDP.
  - Commercial banks share: 65% of total financial system assets.

### Assessment of the TCIFSC Financial Stability Report (FSR)
- Overall assessment:
  - The TCIFSC’s FSR has been published since 2015 and is assessed as relatively good.
  - The report provides overview, motivation, macro-financial narrative, discussion on the real sector, coverage of the entire financial system (banks, investment businesses, domestic insurance companies, trust companies, money service businesses), results of banking ST, and key initiatives.
- Suggested improvements to content and structure:
  - Streamline the FSR to organize text coherently around the central risk story.
  - Focus on key risk drivers and their transmission channels.
  - Provide fewer nominal figures and use relative indicators instead; move some statistics to annexes.
  - Add mission/theme statements for paragraphs to guide readers; consider a foreword from the Managing Director to spell out key messages.
  - Integrate chapters on systemic risk and banking sector resilience into the financial sector development chapter.
  - Make the financial sector developments chapter less descriptive and connect it to financial stability risks.
  - Shorten the macro-financial environment chapter to focus more directly on the TCI economy.
  - Add discussion on emerging risks, such as climate and cyber risks.
- Regulatory and analytical enhancements:
  - Discuss the relevant regulatory framework for different financial system segments within the FSR to make indicators self-explanatory (critical for insurance sector solvency indicators where market valuation frameworks matter).
  - Enhance background financial stability analysis by adding: average probabilities of defaults (PDs), loss given defaults (LGDs), collaterals (e.g., loan-to-value (LTV) ratios), and for non-banks include duration to assess market risk.
  - Reflect on IFRS17 regulatory framework details (discounting of liabilities, types of reinsurance contracts).
  - Consider hiring one person proficient in actuarial science.

### Credit risk modeling
- Work completed:
  - Several credit risk models estimated to project the impact of alternative macroeconomic scenarios on banks’ non-performing loans (NPLs).
  - Introduced Bayesian Model Averaging (BMA) to address uncertainty in modeling sectoral NPLs.
  - Delivered an R script with initial estimates for four sectoral credit risk models:
    - Loans to households
    - Loans for the construction and land development sector
    - Loans to the tourism sector
    - Other corporate loans
- Model variables and coordination:
  - Initial model explanatory variables include tourist arrivals, inflation, unemployment, and real GDP.
  - Discussion initiated with the Ministry of Finance to generate baseline and adverse scenarios based on an existing macroeconomic model including six variables: tourist arrivals, inflation, real GDP, imports, real estate sector GDP, and government revenues.
  - Note: unemployment is not part of the Ministry of Finance macroeconomic model but can be forecasted directly by the TCIFSC.
  - Recommendation to check alternative model specifications using variables generated by the Ministry of Finance and to investigate historical/current NPL write-off practices to improve projections.

### Solvency stress testing framework
- New framework developed:
  - Multi-factor and multi-period solvency ST tool tailored to TCI banks, respecting existing accounting, tax, and regulatory rules.
  - Framework links common explicit macroeconomic scenarios (one baseline and two adverse) to NPL satellite models to project NPLs and credit losses.
  - Produces scenario-specific macroeconomically-consistent projections of institutions’ key balance sheet, profit & loss, and capital adequacy items over up to three years.
- Data, training, and operationalization:
  - Reviewed available data and regulatory framework; constructed tool and provided staff training.
  - TCIFSC staff need to become familiar with the ST framework to update it with new macro-financial and institution-specific data, calibrate scenarios, and adjust assumptions/parameters.
  - Strengthen cooperation with the Ministry of Finance on scenarios; develop new satellite models; regularly publish and share ST results with bank supervisors to support risk-based supervisory practices.

### Illustrative stress tests and results
- Inputs used:
  - Illustrative stress tests run using December 2023 data.
  - Baseline scenario: official TCI Ministry of Finance macroeconomic projection as of March 2024.
  - Adverse scenarios: calibrated by expert judgment to capture different severity levels of economic recessions.
- High-level results:
  - Results suggest banks are generally resilient to economic stress given:
    - Relatively high initial capital adequacy.
    - Good pre-provision profitability.

### Processes, communication, and data management
- Enhancements recommended:
  - Production plan for FSR should be discussed before the production cycle and approved by the Managing Director.
  - External communication of the FSR should be regular (press releases, social media, interviews, etc.).
  - Consider establishing a centralized database for all existing and newly collected data sources, including appropriate analytics tools, to support financial stability monitoring and a fully-fledged solvency ST system.

### Recommendations (numbered, with priority and timeframe)
- Definitions: Near term: < 12 months; Medium term: 12 to 24 months.

- Financial Stability Report
  1. The FSR should be streamlined to organize the text more coherently around the central risk story.
     - Priority: High
     - Timeframe: Near-term
  2. Financial stability indicators need to be communicated together with the key aspects of the existing regulatory framework.
     - Priority: High
     - Timeframe: Near-term
  3. New emerging risks, such as cyber and climate risks, should be consistently covered in the FSR.
     - Priority: High
     - Timeframe: Medium-term
  4. The TCIFSC background financial stability analysis could be enhanced.
     - Priority: High
     - Timeframe: Medium-term

- Processes related to financial stability work, communication, and data management
  5. A production plan for FSR should be discussed before the production cycle starts and approved by the Managing Director.
     - Priority: High
     - Timeframe: Near-term
  6. The external communication of the FSR should be regular (press releases, social media, interviews, etc.).
     - Priority: High
     - Timeframe: Medium-term
  7. A centralized database for all existing and newly collected data sources, including appropriate tools for analytics, should be considered to support financial stability monitoring and a fully-fledged solvency stress testing system.
     - Priority: High
     - Timeframe: Medium-term

- Credit risk modeling
  8. The team needs to become familiar with the methodology and corresponding R script provided.
     - Priority: High
     - Timeframe: Near-term
  9. The team should further check for alternative specifications to improve the initial estimates of sectoral credit risk models.
     - Priority: Medium
     - Timeframe: Medium-term
  10. The team should test the models for the macroeconomic scenarios generated by the Ministry of Finance.
      - Priority: High
      - Timeframe: Near-term
  11. Historical and current practices for NPL write-offs should be investigated.
      - Priority: Medium
      - Timeframe: Medium-term

- Stress testing
  12. The team needs to become familiar with the new stress testing framework.
      - Priority: High
      - Timeframe: Near-term
  13. Cooperation between the TCI FSC and TCI MoF in the area of macroeconomic scenarios needs to be developed.
      - Priority: High
      - Timeframe: Medium-term
  14. Conduct regular (at least twice a year) top-down stress test exercises with the new framework, ideally in spring and autumn.
      - Priority: High
      - Timeframe: Near-term
  15. Publish the aggregate stress test results of the spring exercise in the annual FSR and include a description of the new stress test framework as a “research note” in the FSR when the results are published for the first time.
      - Priority: High
      - Timeframe: Near-term
  16. Share and discuss the stress test results with the Bank Supervision team and use them in the conduct of supervision and supervisory reviews.
      - Priority: Medium
      - Timeframe: Medium term
  17. Consider developing additional models for banks’ key items such as loan growth, deposit growth, interest rate, or non-interest income.
      - Priority: Medium
      - Timeframe: Medium term

*Source: Executive Summary, tarea2025036-print-pdf*

### 5. In the short to medium term, economic growth is expected to be buoyed by momentum in the tourism,

### 5. In the short to medium term, economic growth is expected to be buoyed by momentum in the tourism,

### Tourism, real estate, and construction: outlook and targets
- Tourism sector expected to maintain positive trend following the decline initiated by COVID-19, conditional on effective implementation of the National Tourism Strategy and Implementation Plan for the TCI (published March 2023).
- Plan quantitative targets:
  - Increase total tourist arrivals from 1.6 million in 2019 to 3.2 million in 2032.
  - Grow tourism revenue from $825 million in 2019 to $1.7 billion in 2032 while stemming the decline in visitor spend.
  - Expand total registered accommodation rooms from 4,008 in 2019 to 6,143 by 2032.
- Connectivity and market diversification objectives:
  - Improve air connectivity from Providenciales to other tourism destinations within the island.
  - Expand marketing in current major markets: USA, Canada, UK.
  - Gain market share in new potential markets in Europe (Germany, Italy, France, Spain, Benelux), Latin America (Brazil, Colombia, Chile, Argentina), and the Caribbean region.
- Real estate and construction: continued significant contributions anticipated, supported by recent completions of various resort and development projects and a robust pipeline of ongoing initiatives.

### Downside risks and resilience
- Key downside risks:
  - Weather-related shocks (high vulnerability due to location in the Atlantic hurricane belt).
  - Recession fears in the US.
  - Ongoing geopolitical tensions (e.g., Europe, Middle East) that could affect commodity prices and keep inflation elevated.
- Fiscal/financial buffers and mitigation:
  - National Wealth Fund stood at $51 million as of 31 December 2023.
  - Government participation in the Caribbean Catastrophe Risk Insurance Facility (CCRIF) to help mitigate hurricane-related costs.
  - Past major storms (e.g., Hurricane Irma in 2017) had impacts mitigated by strong financial reserves.

### Financial sector structure and key statistics (end-2023)
- Financial sector size: 275% of GDP at the end of 2023.
- Distribution of financial sector assets:
  - Banks: 65% of total financial assets; 6 banks.
  - Investment business: 31% (5 investment dealers, 8 mutual fund administrators).
  - Insurance companies: 3%; 18 companies.
- Market concentration and ownership:
  - Banking sector controlled by three Canadian banks (97% of total banking assets) and one British-Belizean bank.
  - Three Canadian banks operate as large retail Systematically Important Banks (SIBs) for the TCI economy.
  - Two other banks are niche players specializing in offshore wealth management.
- Dominant risk types:
  - For largest banks: credit and liquidity are main sources of risk; interest rate risk could also be significant.
- Distribution visual/statistics reproduced:
  - 66.0% — Banks
  - 31.0% — Investment business
  - 3.0% — Insurance companies

### Financial stability risks and transmission channels
- Primary financial stability risk: credit risk linked to a potential drop in tourist arrivals, with implications for banks’ credit portfolios and increased NPLs.
- Additional risks:
  - Non-life insurers vulnerable to climate change risk and rising inflation, possibly facing higher-than-expected claims.
  - External environment, especially US economic conditions, is a major determinant given TCI’s exposure to US tourists.
- Importance of measuring and communicating credit risk and interest rate risk for proper assessment of the financial system.

### Financial Stability Report (FSR) best practices and recommendations
- Objectives and definition:
  - FSRs should explicitly state objectives and provide a clear, dynamic definition of financial stability, including operational benchmarks and monitored indicators.
- Executive summary:
  - Should be brief, easy to read, and present main conclusions and evolution of risks since the previous issue; analytical chapters can contain advanced technical material in boxes/annexes.
- Macro-financial integration:
  - FSRs should integrate macroeconomic and financial analysis in both directions, identify transmission channels, and quantify shock reverberations where possible.
- Transparency of assumptions and methodologies:
  - Stress tests and other analyses should clearly justify assumptions (severity, speed of impact, hurdle rates, dividend policies, etc.), with limitations described.
- Data and methodology disclosure:
  - Sources and cut-off dates should be clearly indicated; data used should be made publicly available on the publishing authority’s website; methodologies for projections, sensitivity analyses, and stress tests should be explained, with technical details in boxes/annexes.
  - Reports should consider distributional indicators and, where appropriate, extreme or individual values (with confidentiality safeguards).
- Structure and content:
  - Follow a logical, integrated structure centered on key risks; include executive summary, chapters on external and domestic developments, financial infrastructure and regulatory framework, table of acronyms, glossary, methodological annex, and statistical annex.
- Publication timetable and frequency:
  - Publish relatively shortly after data cut-off (preferably within three months).
  - Frequency should change over the medium term to two times a year to align with best international practices.
  - Publication date should be announced in advance and indicated on the website.

### Organizational and communication recommendations for the TCIFSC
- Production planning and governance:
  - Policy unit should prepare a detailed FSR production plan with regular structural meetings and clear responsibilities; senior management should approve the plan.
  - Final discussion of the FSR in the TCIFSC Board needs to be scheduled before final publication and included in the production plan.
- Internal and external review:
  - Current internal commenters include Directors of Bank & Trusts, Insurance, and ALM Supervision Departments, the Managing Director, and externally the Head of the Statistics Department, Ministry of Finance.
- Communication strategy:
  - Complement production plan with a communication strategy positioning the FSR as the key communication tool for financial stability.
  - Consider setting up a communication team to plan press conferences, media interviews, YouTube videos, press releases (as a rule sent to media and key financial sector representatives), and seminars with market participants/analysts.
  - Include Foreword from the Managing Director in the FSR and use quotes in press materials.
  - Ensure all charts and tables’ underlying data are provided on the website alongside the report.

### Content, structure, and analytical enhancements for the FSR
- Streamlining and risk-focused narrative:
  - Organize text coherently around a central risk story and key risk drivers; reduce reliance on nominal values in the text, use relative indicators or move nominals to a statistical annex.
  - Integrate chapters on systemic risk and banking sector resilience earlier and link them to financial sector developments.
  - Shorten regional macro-financial sections and focus more directly on impacts to the TCI economy.
  - Add discussion on emerging risks such as climate and cyber risks.
  - Consider adding a foreword from the Managing Director.
- Regulatory framework and sector-specific information:
  - Add descriptions of relevant elements of the regulatory framework for different financial segments (critical for insurance solvency interpretation, IFRS17 implications, reinsurance contract types).
- Enhance background analysis:
  - Strengthen credit risk analysis by including average probabilities of default, loss given default, and collateral measures (e.g., loan-to-value ratios).
  - Enhance nonbank analysis by adding critical information such as duration and IFRS17-related clarifications on discounting of liabilities and reinsurance arrangements.

### Strengthening the financial stability analytical toolkit
- Develop econometric models to project key variables: credit volumes to the private sector, nonperforming loans, banks’ interest and non-interest income and expenses.
- Recommended methodology:
  - Bayesian Model Averaging (BMA) presented as a suitable approach to tackle model uncertainty, limited time series in TCI, and to explore links between macro-financial variables and banks’ performance indicators.
  - BMA’s flexibility aids in handling uncertainty in predictions and parameter estimates, variable selection, and dynamic relationships, making it effective for limited-data environments.

*IMF | CARTAC Turks and Caicos Islands | excerpt from chapter detailing short- to medium-term outlook, financial sector structure, risks, and FSR recommendations*

### 28. Macro credit risk modeling is a crucial financial stability analytical toolkit component. This is because

### 28. Macro credit risk modeling is a crucial financial stability analytical toolkit component. This is because

### Credit risk modeling work and methodology
- Purpose
  - Develop credit risk models to project risk under different macroeconomic scenarios as a precondition for multi-period, multi-factor macro stress testing of the TCI banking sector.
- Sectoral models developed
  - Four sectoral credit risk models: loans to households (personal loans), loans for the construction and land development sector, loans to the tourism sector, and other corporate loans.
- Dependent variable and transformation
  - Nonperforming loan (NPL) ratios for the four sectors were employed as the dependent variable.
  - NPL ratios were transformed using a logit transformation to make them suitable for use in linear models.
- Estimation method and model suite
  - The four credit risk models were estimated using the BMA method.
  - In each sector, between 3-4 models were estimated; the final sectoral projection is a weighted average of the individual model projections.
  - The mission provided an R script with an initial estimate for the four sectoral credit risk models; final set of models for each sector is available in Appendix 1.
- Explanatory variables and data frequency
  - Considered explanatory variables: tourist arrivals, inflation, real GDP, and unemployment.
  - Data availability: NPL ratios and tourist arrivals are available quarterly; real GDP, inflation, and unemployment are currently available only annually, so quarterly proxies were constructed via interpolation.

### Recommendations for further work
- TCIFSC actions
  - Further test and improve the initial estimates of sectoral credit risk models.
  - Team members must become familiar with the R script provided.
- Coordination with Ministry of Finance
  - Further discussion with the Ministry of Finance to generate baseline and adverse scenarios for stress testing based on the existing macroeconomic model, including six variables: tourist arrivals, inflation, real GDP, imports, real estate sector GDP, and government revenues.
  - Note: The Ministry of Finance's macroeconomic model projects tourist arrivals based on US GDP; this then serves to estimate domestic GDP.
- Model extensions and data issues
  - Unemployment is not part of the Ministry's macroeconomic model but could be forecasted directly by the TCIFSC.
  - Alternative sectoral credit risk model specifications with variables generated by the Ministry of Finance that are not included in the initial estimates should be further checked.
  - Historical and current NPL write-off practices should be investigated to improve projections.

### Stress testing (ST) tool: structure and key features
- General framework
  - New Excel-based macro ST tool explicitly projects key balance sheet items (performing and non-performing loans across the main loan segments, loan loss provisions, total and risk-weighted assets, and capital) and P&L items (net interest income, other income and expenses, and credit losses) over a horizon of up to three years.
  - Tool works with four key loan segments: tourism & entertainment, personal loans, construction, and others.
  - Tool can be run on any end-quarter data and generate evolution charts and contribution charts; includes a detailed step-by-step guide and a detailed user manual.
- Accounting, tax, and regulatory alignment
  - Projections handle both traditional coefficient-based provisioning by asset quality classes and possible IFRS 9 provisioning.
  - Definition of regulatory capital (Tier 1 and Tier 2) fully respected.
  - Tool acknowledges the 0% corporate income tax rate for banks.
- Scenario setup and required macro variables
  - Baseline scenario: reflect most recent official macroeconomic outlook prepared by the TCI Ministry of Finance.
  - Two adverse scenarios: “moderate” and “severe”, calibrated to capture different intensities of an economic recession.
  - Key macroeconomic variables to always be projected for all three scenarios: real GDP growth, inflation, unemployment rate, tourist arrivals, income from tourism, and US 3M T-bill rate.
  - Future macro projections could include a credit forecast to the private sector to calibrate loan portfolio growth and enable dynamic balance sheet sizing.

### How sectoral credit models feed the ST tool and conservatism adjustments
- NPL projection mechanics
  - Scenario-specific macro projections are used to predict NPLs with the sectoral credit risk satellite models.
  - BMA weighting schemes were adjusted by expert judgment to yield more conservative NPL projections for larger recessions:
    - The weight applied on the most “reactive” NPL model (generating the most significant NPL increases) was set at the highest level for the severe scenario and the second highest level for the moderate scenario (at the expense of the remaining models) compared to the baseline.
- Application of projected NPL changes across institutions
  - The absolute change in the projected aggregate NPL ratio is applied to each loan segment's starting NPL ratio level.
  - Using absolute rather than relative changes ensures institutions with a 0% starting level of the NPL ratio in some loan segments are hit by new NPLs in adverse conditions.
  - Tool allows additional expert-based adjustments:
    - explicit segment-specific NPL ratio add-ons
    - explicit bank-specific NPL ratio add-ons
- Special-mention-to-performing loans ratio
  - Tool projects special-mention-to-performing loans ratio for each institution and segment.
  - Projection is linked to the NPL projection using an elasticity specified as a parameter set by expert judgment (changeable).
  - Loan growth projection (calibrated per scenario and applied equally across segments and institutions for simplicity) jointly with credit risk projections determines paths of good, special mention and non-performing loan exposures.
  - Three explicit loan quality classes facilitate future transition to IFRS 9 Stage 1-3 classification.

### Provisioning, write-offs, and other balance sheet assumptions
- Loan loss provisioning logic
  - Projections based on existing provisioning rules and bank practices.
  - NPL provisioning levels are institution- and segment-specific: starting NPL provisioning rate plus an assumed increase of around 10-20 ppts (cumulative) in both adverse scenarios (typically no change for the baseline scenario).
  - Shock can be set separately for each scenario and year; reflects worsened NPL quality and migration to worse sub-classes or deterioration in collateral value.
  - Same logic applied to special mention (Stage 2) and good (Stage 1) loans—starting observed levels used for baseline, parameter-driven increases in adverse scenarios.
  - Provisions created for good loans are considered general provisions and qualify as Tier 2 regulatory capital up to 1.25% of credit risk-weighted assets.
- NPL write-offs and final provisions
  - Final loan loss provisions impacting P&L derived from projected stocks, considering scenario- and year-specific NPL write-offs calibrated by expert judgment using qualitative information about bank write-off practices.
  - Even if model projects unchanged NPL stocks, new provisions may still be needed to account for NPL inflows required to keep NPLs stable amid write-offs.
- Other assets, deposits, and funding
  - Projections for additional financial assets (cash, claims on financial institutions, debt securities, other assets) and deposits can be specified.
  - Calibration of growth rates can follow loan growth assumptions or expert judgment.
  - Evolution of asset items impacts total and risk-weighted assets and net interest income; deposit projection co-determines interest expenses.
  - Calibration must ensure assets equal liabilities (may require iterations).

### Market, FX, concentration, and income assumptions
- Market and FX risk
  - Market risk from revaluation of securities can be switched on as a sensitivity; almost all securities held are foreign government bonds (typically US treasuries) normally held at amortized cost until maturity.
  - No domestic securities issued; no sovereign risk included.
  - USD as TCI currency and virtual absence of other currencies remove exchange rate risk from balance sheets; FX risk not considered for stress.
- Concentration risk
  - Exposure to one or more of the largest five borrowers can be stressed with LGD calibrated between 0% and 100%; typically applied as a sensitivity within an adverse scenario.
- Pre-provision income (PPI)
  - PPI projected as interest and non-interest income minus interest and non-interest expenses.
  - Interest income = time-varying average implicit interest rate modeled along scenario-specific development of USD 3M T-bill rates (with pre-defined passthrough) × interest-bearing assets.
  - Interest-bearing assets typically decline in adverse scenarios as performing loans migrate to NPLs.
  - Interest expenses projected using assumed path of deposit funding and deposit interest rates (initial observed level, but can include funding cost shocks).
  - Non-interest income and expenses projected as institution-specific starting points adjusted by haircuts set by expert judgment; can later be linked to macro scenarios via satellite models.

### Capital, assets, RWAs, and reporting
- Capital and dividend treatment
  - Capital projected as a function of net income; negative net income reduces capital.
  - Positive net income is subject to distribution decision; only retained part after dividend payouts increases capital.
  - Dividend payout assumptions are institution-specific.
- Total assets and RWAs
  - Total assets = time-varying net loans + other financial assets.
  - Credit RWAs projected as function of net exposures and risk weights for key asset classes; market and operational RWAs kept constant.
  - Total assets and RWAs driven by credit growth, NPL evolution, and provisioning assumptions; typically decline in adverse scenarios.
  - Decline in assets/RWAs can somewhat help capital ratios but is small relative to the impact of economic recessions causing accounting losses.
- Reporting outputs
  - Key results reported based on capital ratios: overall capital adequacy ratio, simple capital ratio, and Tier 1 ratio.
  - Tool shows contributors to changes in capital adequacy: net interest income (+), other income/expenses (+/-), credit losses (-), dividend payouts (-), change in denominator (risk-weighted assets) (+/-).
  - Reports institutions below a specified hurdle rate for total capital adequacy ratio and their share in sector assets, plus capital injections (expressed as a % of GDP) needed to bring all institutions to at least the minimum capital adequacy ratio of 11%.

### Overview table components (selected items)
- Institutional perimeter: All institutions (6 banks, additional banks can be added).
- Key macro variables for scenario calibration: Real GDP growth, y-o-y %; Inflation, y-o-y %; Unemployment rate, %; Tourist arrivals, ths.; Income from tourism, mil. USD; US 3-month T-bill rate, %.
- Credit risk satellite models: Estimated by BMA as a suite of simple models; dependent variable = logistic transformation of NPL ratio; explanatory variables include autoregressive term, GDP growth, inflation, tourist arrivals, constant.
- Additional credit assumptions that need to be set: Private sector credit growth, y-o-y%; Assumed (annual) NPL write-off rate, %; Change in NPL provisioning from initial level, ppts; Change in special mention loan provisioning from the initial level, ppts; Change in good loan provisioning (general provisions) from the initial level, ppts; Factor of special-mention-to-performing loans absolute increase compared to NPL ratio increase; Bank-specific NPL add-ons; Sectoral NPL add-ons.
- Assumptions for balance sheet items: Annual growth, % for cash, claims on financial institutions, government securities, other securities, other assets (incl. fixed assets), deposits and other funding.
- Concentration risk needs specification of which of 5 largest borrowers fail, in which year, in which scenario, and LGD in %.
- Assumptions for PPI: Passthrough coefficient from 3M USD T-bill rates to US government securities, other securities, claims on financial institutions, loans and advances; Funding shock (change in deposit rates from initial level); Scenario-specific changes in non-interest income and non-interest expense applied on institution-specific initial annual volumes, %; Dividend payout rate; Minimum total regulatory capital ratio.

### Illustrative stress test setup and headline scenario parameters
- Data and coverage
  - Illustrative ST prepared using December 2023 data; all six banks covered.
- Scenario calibration notes
  - Baseline scenario captures the official MoF forecast as of March 2023.
  - Severe scenario calibrated to reflect a 3-standard deviation shock to GDP growth, assuming a real GDP decline by around 20% at the bottom of the cycle.
  - Moderate scenario captures a milder recession, with GDP growth values set as averages of the baseline and the Severe scenario values.
  - Both adverse scenarios reflect possible deterioration in tourism revenues and a resulting recession similar to the 2009-2011 post-Global Financial Crisis period (drop in GDP by 20%) or the recent 2020-2021 COVID-19 recession (drop in GDP by 25%).
- Loan growth assumptions
  - Loan growth assumed at 10%, 5%, and 0% annually for the Baseline, Moderate, and Severe scenarios, respectively.

### Illustrative model projections (summary)
- NPL model projections
  - For loans to households and other corporate sectors: baseline assumes a slight decline in NPLs amid relatively good GDP growth projection and declining unemployment.
  - For the construction and tourism sectors: credit risk models suggest a mild increase even in the baseline.
  - For both adverse scenarios: increase in NPL ratios to high levels experienced in post-GFC times around 2013.
  - No sector-specific shocks or failures of the largest borrowers were assumed in the illustrative ST.

*Source: Mission team calculations; Turks and Caicos Islands Financial Services Commission materials included in the mission report.*

### 49. Additional  assumptions  that  significantly  impact  the  projected  loan  losses  are  related  to  NPL

### 49. Additional  assumptions  that  significantly  impact  the  projected  loan  losses  are  related  to  NPL provisioning

### NPL provisioning shocks and related parameters
- Shock to NPL provisioning rate assumed:
  - Moderate scenario: 15 ppts (cumulatively, with the first-year shock of 10 ppts and the second-year increase of 5 ppts).
  - Severe scenario: 30 ppts (cumulatively, with the first-year shock of 20 ppts and the second-year increase of 10 ppts).
- Average starting NPL provisioning: 47%.
- Resulting provisioning level in Severe scenario: would increase to about 87%.
- Special mention loans provisioning shock: 1 ppts (Moderate) and 5 ppts (Severe).
- General provisions/Stage 1 loans shock: 0.5% (Moderate) and 1% (Severe).
- NPL write-off rate: 10% for all scenarios.

### Projected loan losses and credit loss rates
- Baseline scenario:
  - Credit loss rate (defined as new provisioning booked in the P&L over the initial stock of net loans) hovers around 0.5% annually.
- Severe scenario:
  - Stock of loan loss provisions would increase about four times over the 3-year horizon.
  - Provisions would reach almost 13% of gross loans.
  - Large new provisioning of around 6% per year on average in the first two years of the horizon.

### Income, expenses, and other P&L assumptions
- Pre-provision income serves as first line of defense but is lower in adverse scenarios.
- Net interest income would drop considerably in the Severe scenario due to:
  - Declining interest-bearing assets (driven by lower-performing loans).
  - Lower average loan interest rates due to a large decline in US interest rates.
- Funding shock: none assumed.
- Other asset items, deposits, and funding: assumed to remain at the initial level.
- Non-interest income assumptions:
  - Remain at last year’s baseline level in Baseline scenario.
  - Drop by 15% in the Moderate scenario.
  - Drop by 30% in the Severe scenario.
- Non-interest expenses: assumed to grow somewhat (by 35 annually) in all scenarios.
- No revaluation of securities assumed.

### Bank-level profitability, capital, and resilience
- Profitability outcomes:
  - Four larger banks become loss-making only in the Severe scenario.
  - Two smaller banks experience negative profitability already in the Moderate scenario.
  - In total, all banks would, at one point, experience negative profits.
- Tax and payout assumptions:
  - Assuming no corporate tax.
  - 20% dividend payout rate for all banks.
- Impact on capital:
  - Projected net income would impact the final capital stock.
  - Risk-weighted assets (RWAs) would change following the implied trajectory of net loans and (unchanged) other asset items.
- Capital adequacy outcomes:
  - Aggregate CAR would grow in both the Baseline and the Moderate scenario due to continuing aggregate profitability driven by the largest banks.
  - Only in the Severe scenario would the sector’s CAR drop (faster drop in capital than in RWAs), but it would still remain comfortably above the limit.
  - All banks would remain above the regulatory minimum for capital adequacy of 11% in all scenarios.

### Stress test framework use and calibration recommendations
- Further analysis and fine-tuning of scenario calibration needed before regular use of the ST tool.
- Scenarios should be prepared consistently and underpinned by a good narrative jointly developed by the TCIFSC and TCI Ministry of Finance.
- Calibration of parameters that might need additional analysis and data:
  - Credit growth.
  - NPL write-offs.
  - Passthrough coefficient for interest rates.

### Operational and data recommendations (linked to stress testing)
- ST team at TCI FSC should become familiar with the new framework to make regular updates; staff must be able to:
  - Update the tool with new macro-financial and institution-specific data regularly.
  - Calibrate scenarios.
  - Adjust additional assumptions and parameters.
- Frequency of stress tests:
  - Use the new framework for regular internal stress tests, ideally twice a year — once in the spring and once in the autumn.
  - Aggregate ST results of the spring exercise should be published in the annual FSR.
  - A “research note” describing the new framework and all key assumptions should be provided in the FSR once the results are published for the first time.
- Use of ST results:
  - Share and discuss ST results with the Bank Supervision team at the TCIFSC.
  - Utilize results in supervisory reviews and overall supervision of individual banks to foster synergies between financial stability and supervision teams.

*IMF | CARTAC Turks and Caicos Islands*

---


_Source: https://www.imf.org/-/media/files/publications/tar/2025/english/tarea2025036-print-pdf.pdf_
