## A SAVINGS FUND FOR NATURAL DISASTERS: AN APPLICATION TO DOMINICA

## Source details

**Canonical URL:** [A SAVINGS FUND FOR NATURAL DISASTERS: AN APPLICATION TO DOMINICA](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16245.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16245.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16245.pdf.json)

---

### Introduction and motivation
- Dominica faces recurrent natural disasters (ND) that produce immediate social protection and rehabilitation expenditures and contribute to longer-term reconstruction costs and higher public debt.
- Public debt is reported as "over 85 percent of GDP", constraining borrowing capacity after NDs.
- Proposal: create a savings fund (SF) financed by Economic Citizenship Program (ECP) revenues to:
  - Smooth the macroeconomic impact of volatile and potentially unstable ECP inflows.
  - Provide self-insurance for immediate reconstruction and rehabilitation needs.
  - Reduce the need for post-disaster debt issuance and support fiscal sustainability.

### Rationale for a savings fund versus market insurance
- Market-based insurance and regional schemes are insufficient and costly:
  - Private-sector underinsurance is pervasive; actuarial assessment is complicated by low-probability/high-damage events and heterogeneous ND types (hurricanes, tropical-storm rainfall, earthquakes).
  - CAT bonds characteristics cited: coupons of "Libor plus a spread in the range of 3-20 percent" and "maturities of less than 3 years"; CAT bond triggers can be imperfectly correlated with actual losses.
- Governments act as de-facto ultimate insurers; external or domestic financing after NDs is often slow or constrained.
- A dedicated SF:
  - Ensures rapid availability of resources for immediate expenditure needs, rehabilitation, and reconstruction.
  - Imposes recurrent saving discipline, reducing probability of ad hoc post-disaster borrowing.
  - Does not necessarily increase long-run public debt relative to no-SF scenarios because saved resources are expected to be used at some point for recurrent NDs.

### Empirical methodology and simulation design
- VAR specification and estimation:
  - Endogenous variables: cyclical components of GDP; government revenues excluding grants; grants; current primary expenditures; capital expenditures.
  - Cyclical components computed as ratio to estimated trend; GDP cyclical component estimated with Hodrick-Prescott filter on "1990-2015" annual data.
  - All variables are real (GDP deflator) and in logarithms. Shock identification uses Choleski decomposition in the presented ordering.
- Controls: U.S. real effective exchange rate; oil price; cyclical component of U.S. output; dummy for the September 2001 shock.
- Monte-Carlo simulation:
  - "1000 simulations".
  - Simulation horizon: "2016-2030".
  - Shocks drawn from a normally-distributed probability density function estimated from model residuals.
  - Simulations produce probabilistic paths for the five endogenous variables each projected year, expressed in percent of GDP after assuming deterministic trends that grow at the same constant rate from the sample end point.
  - Interest expenditures projected by multiplying prior-year debt stock by an implicit interest rate path; implicit interest rate treated as a calibration parameter. Debt dynamics use the debt accumulation identity expanded to include SF-government financing flows.

### ND identification and SF trigger logic within simulations
- ND identification algorithm:
  - Compute year-on-year changes (sum) of: (i) non-grant revenue (with negative sign), (ii) grant revenues, (iii) current primary expenditure, and (iv) capital expenditure.
  - The largest X percent fiscal deteriorations (highest tail of this sum's distribution) are labeled as NDs; the remaining 1-X percent interpreted as other smaller shocks.
  - Typical ND pattern: significant declines in non-grant revenue and significant increases in grants, current primary expenditure, and capital expenditure.
- Triggering SF flows:
  - Identified NDs trigger financing flows between the SF and the government budget in the simulations to produce probabilistic public debt projections that account for SF dynamics.

### Modeling SF inflows/outflows, base spending and re-prioritization
- Inflows:
  - In years with no ND, the budget generates an additional overall balance surplus as a percent of GDP that is deposited in the SF.
  - Budget contributions to the SF modeled as a fixed parameter as a percent of the previous year’s GDP.
  - Annual saving calibrated to achieve financial sustainability with a sufficiently low probability of depletion and stability of SF stock in expected terms.
- Outflows in a ND year: sum of four components:
  - + Gap of non-grant revenues below trend.
  - - Gap of grant revenues above trend.
  - + Gap of current primary expenditure above trend (includes additional fixed amount as percent of GDP for social support and rehabilitation).
  - + Gap of capital expenditure above trend (includes additional fixed amount as percent of GDP for reconstruction).
- Contributions to the budget from the SF continue until each indicator returns to a level below the pre-ND year value.
- Expenditure re-prioritization assumption:
  - SF finances only increases of fiscal needs above trends; large estimated fiscal costs are offset by reallocation and re-prioritization of pre-ND allocations.

### Parametric calibration (selected exact values)
- Historical sample for VAR estimation: "1990-2015".
- Monte-Carlo runs: "1000 simulations".
- Simulation horizon: "2016-2030".
- Public debt: "over 85 percent of GDP".
- CAT bond characteristics: coupons of "Libor plus a spread in the range of 3-20 percent"; "maturities of less than 3 years".
- Table 1 selected parameter values:
  - GDP potential growth, percent: 1.7 (applied through 2016-2022 and 2022-35 as 1.7).
  - Implicit interest rate on public debt, percent: 2.5 (applied through 2016-2022 and 2022-35 as 2.5).
  - Fiscal consolidation measures, Primary Balance, percent of GDP: -0.4 (2016), 0.1 (2017), 1.6 (2018), 1.6 (2019), 1.6 (2020), 0.5 (2021), 0.0 (2022-35).
    - Non-grant revenue: 1.0 (2016), 1.0 (2017), 1.0 (2018), 1.0 (2019), 1.0 (2020), 0.0 (2021), 0.0 (2022-35).
    - Grant revenue: 0.0 across 2016-2022 and 2022-35.
    - Current primary expenditure: -0.1 across 2016-2020, 0.0 (2021), 0.0 (2022-35).
    - Capital expenditure: 1.5 (2016), 1.0 (2017), -0.5 (2018), -0.5 (2019), -0.5 (2020), -0.5 (2021), 0.0 (2022-35).
  - Cumulative fiscal consolidation, percent of GDP: -0.4 (2016), -0.3 (2017), 1.3 (2018), 2.9 (2019), 4.5 (2020), 5.0 (2021), 5.5 (2022-35).
- Government Fund parameters:
  - Initial Fund stock / GDP: 10.00 (percent).
  - Inflows into the Fund from the budget, percent of GDP:
    - Non-grant revenue t-1: 1.500
    - Grant revenue t-1: 0.000
  - Outflows of the Fund to the budget parameters:
    - Storm probability threshold: 0.200
    - Base capital expenditure level / GDP: 0.064
      - Capital expenditure trend / GDP trend: 0.084
      - Year average capital expenditure on reconstruction / GDP: 0.020
    - Base current primary expenditure / GDP: 0.190
      - Current primary expenditure trend / GDP trend: 0.230
      - Year average current expenditure social support and rehabilitation / GDP: 0.040

### Calibration choices, sustainability targets and sensitivity
- ND probability threshold set at 0.2 (ND every 5 years on average).
  - Example: Probability[x(t) < X] = 0.2 implies that with 1,000 simulations per year on average 200 out of 1,000 simulations in each year would be identified as a ND.
- Initial size of Fund stock set at 10 percent of GDP; chosen to obtain a probability of depletion within the next ten years of 0.08.
- Budget saving flows into the SF in years without a ND set at 1.5 percent of previous-year’s GDP.
- Depletion and alternative calibrations:
  - If ND probability = 0.25 (ND every 4 years on average), budget savings of 2 percent of GDP per year would be needed for SF sustainability.
  - If SF starts with assets of 2 percent of GDP and ND probability = 0.2, SF would be depleted during 2016-2030 with probability of more than 30 percent.
  - To reduce probability of depletion to less than 10 percent, a SF of at least 8 percent of GDP is required.
- Key sensitivity threshold: with annual ND probability = 0.2 and specified budget financing rules, annual budget savings of 1.5 percent of GDP are needed to achieve SF financial sustainability over time.

### Illustration and simulated outcomes
- Example random simulation (2016-2030):
  - Three NDs occur.
  - SF stock increases up to 2020 as 1.5 percent of GDP saved every year to more than 15 percent of GDP.
  - A ND in 2021 leads to SF disbursements to the budget of about 5 percent of GDP.
- Under calibrated parameters, SF would be financially sustainable with a low probability of depletion; results show significant dispersion depending on shock realizations.

### Impact on public debt and risk bands
- With this calibration, public debt would decline to near 60 percent of GDP by 2030 in expected terms, but with significant dispersion depending on realization of shocks.
- Result aligns with regional commitments to reach a public debt ratio of 60 percent of GDP or lower by 2030.
- Even with fiscal consolidation, significant probability remains that the target will not be met under more extreme ND frequency or severity scenarios; probabilistic fan charts illustrate these risk bands.

### Policy implications and recommendations
- SF design and governance:
  - Include unambiguous budget contribution and disbursement rules with triggers based on verifiable criteria.
  - Clearly-stated objective.
  - Strict information disclosure requirements to ensure transparency.
- Funding and fiscal alignment:
  - Start-up costs and subsequent saving flows could be financed with ECP resources under clear savings rules established in legislation.
  - Benefits: avoid issuing public debt for start-up cost; discipline to save for recurrent NDs; reduce scope for allocating ECP revenues to recurrent spending given fiscal consolidation needs.
  - Other funding sources: donor partner contributions for budget support and international loans for investment projects financed with Fund resources.
- Broader fiscal policy:
  - Fiscal consolidation calibrated in line with macroeconomic framework: cumulative fiscal consolidation measures of 5.5 percent of GDP, largely from revenue measures introduced smoothly through 2016-2021.
  - Capital expenditures calibrated to map expected increase related to tropical storm Erika in 2015 and the subsequent unwinding towards 2021.
- Institutional safeguards:
  - Strong institutional design and oversight are critical to avoid political capture and ensure SF resources are used for intended reconstruction and rehabilitation purposes.

---

### Credit unions, stress testing, and disaster context (selected complementary findings)
- Credit union sector overview and size:
  - Total assets increased from US$172 million (35 percent of GDP) to US$233 million (45 percent of GDP) during 2009–14.
  - Average annual growth rate of credit union assets: 7 percent.
  - Credit unions extended proportionally more loans: loans-to-GDP ratio increased by about 7 percentage points to 31 percent during 2009–14.
  - Credit union loans account for almost 32 percent of total private sector credit.
- Financial soundness and risks (end-2014):
  - Non-performing loans (NPLs) for the sector as a whole: 6.4 percent.
  - Aggregate CAR: 8.7 percent (below minimum regulatory requirement of 10 percent).
  - Provision coverage: 60 percent.
  - Liquidity: liquid assets represent 10 percent of credit unions’ assets (compared with 40 percent for commercial banks).
- Stress test scenarios and outcomes:
  - Baseline CAR: 8.7 percent.
  - Scenario 1 (full provisioning of expected losses): CAR decreases to 5.9 percent.
  - Scenario 2 (NPLs increased by one standard deviation, about 50 percent higher, and full provisioning): CAR decreases to 3.3 percent.
  - Interest rate risk scenarios: CAR decreases slightly from 8.7 percent to 8.5 percent with a 2 ppt. increase in interest rates.
- Systemic and monetary aggregates (selected levels, end period, in millions of Eastern Caribbean dollars):
  - Net foreign assets 2014: 611.9
  - Net domestic assets 2014: 1,072.3
  - Private sector credit 2014: 1,210.8
    - credit from banks 2014: 773.2
    - credit from credit unions 2014: 437.6
  - Broad Money 2014: 1,642.4
  - Credit union currency and deposits (net) 2014: 395.8
- Economic impact of tropical storms (selected estimates):
  - Average annualized actual cost of a tropical cyclone: 7.4 percent of GDP (1970-2015).
  - Estimated cost: 10.8 percent of GDP (1970-2015, as estimated using Acevedo and World Bank data).
  - More than half of the tropical cyclones affecting Dominica caused damages of over 50 percent of the country’s national output in the given year.
- Policy recommendations related to credit unions and disaster resilience:
  - Strengthen regulation and supervision of credit unions; step up information exchange and collaboration between the ECCB, SRUs, and local regulators.
  - Special supervision for large credit unions (largest CU accounts for about 71 percent of CU assets).
  - Boost capitalization, consider consolidation and mergers to enhance stability and efficiency.
  - Improve collection and compilation of data from credit unions.
  - Disaster preparedness measures: early warning systems, better building standards and zoning, contingency funds, improved market insurance and regional insurance pools, and prudent fiscal policies to improve resilience.

*Source: IMF staff analysis and simulations (sections 15–30).*

### References ____________________________________________________________________________ 15

### A SAVINGS FUND FOR NATURAL DISASTERS: AN APPLICATION TO DOMINICA

### Introduction and motivation
- Dominica faces recurrent natural disasters (ND) that produce immediate social protection and rehabilitation expenditures and contribute to longer-term reconstruction costs and higher public debt.
- Public debt is reported as "over 85 percent of GDP", constraining borrowing capacity after NDs.
- The paper proposes creating a savings fund (SF) financed by Economic Citizenship Program (ECP) revenues to:
  - Smooth the macroeconomic impact of volatile and potentially unstable ECP inflows.
  - Provide self-insurance for immediate reconstruction and rehabilitation needs.
  - Reduce the need for post-disaster debt issuance and support fiscal sustainability.

### Rationale for a savings fund vs market insurance
- Market-based insurance and regional schemes are insufficient and costly:
  - Private-sector underinsurance is pervasive; actuarial assessment is complicated by low-probability/high-damage events and heterogeneous ND types (hurricanes, tropical-storm rainfall, earthquakes).
  - CAT bonds are noted as typically risky, paying coupons of "Libor plus a spread in the range of 3-20 percent", and having "maturities of less than 3 years"; CAT bond triggers can be imperfectly correlated with actual losses.
- Governments act as de-facto ultimate insurers; access to rapid external or domestic financing after NDs is often slow or constrained.
- A dedicated SF:
  - Ensures rapid availability of resources for immediate expenditure needs, rehabilitation, and reconstruction.
  - Imposes recurrent saving discipline, reducing the probability of ad hoc post-disaster borrowing.
  - Does not necessarily increase long-run public debt relative to no-SF scenarios because saved resources are expected to be used at some point for recurrent NDs.

### Comparative regional experience and institutional considerations
- Regional examples using ECP or resource revenues for funds are cited:
  - Sugar Industry Diversification Fund (St. Kitts and Nevis), National Transformation Fund (Grenada), Trinidad and Tobago SWF, and various funds in Turks and Caicos.
- Experiences have been mixed, with political influence and capture undermining outcomes—underscoring the need for strong institutional design and oversight.

### Methodology for calibrating SF size
- Empirical framework:
  - A Vector Auto-regression Model (VAR) is estimated with endogenous variables: cyclical components of GDP; government revenues excluding grants; grants; current primary expenditures; and capital expenditures.
  - Cyclical components computed as ratio to estimated trend; GDP cyclical component estimated with the Hodrick-Prescott filter on "1990-2015" annual data. All variables are real (GDP deflator) and in logarithms. Shock identification uses Choleski decomposition in the presented ordering.
- Controls and identification:
  - Control variables: the U.S. real effective exchange rate; oil price; cyclical component of U.S. output; a dummy for the September 2001 shock.
  - The controls aim to remove major alternative sources of shocks so residuals predominantly reflect ND-related variability.
- Simulation design:
  - Monte-Carlo experiment: "1000 simulations" are run over 2016-2030.
  - Shocks are drawn from a normally-distributed probability density function estimated from the model residuals.
  - Simulations produce probabilistic paths (probability density functions) for the five endogenous variables each projected year, expressed in percent of GDP after assuming deterministic trends that grow at the same constant rate from the sample end point.
  - Interest expenditures are projected by multiplying prior-year debt stock by an implicit interest rate path (ratio of interest expenditures to public debt stock), treated as a calibration parameter. Debt dynamics use the debt accumulation identity expanded to include SF-government financing flows.

### ND identification and SF trigger logic within simulations
- Natural disasters in a simulation are identified algorithmically:
  - Compute year-on-year changes (sum) of: (i) non-grant revenue (with negative sign), (ii) grant revenues, (iii) current primary expenditure, and (iv) capital expenditure.
  - The algorithm treats the largest X percent fiscal deteriorations (highest tail of this sum's distribution) as NDs; the remaining 1-X percent are interpreted as other smaller shocks.
  - If a realization features significant declines in non-grant revenue and significant increases in grants, current primary expenditure, and capital expenditure (a typical ND pattern), it is labeled a ND in that simulation.
- Identified NDs then trigger the financing flows between the SF and the government budget in the simulations to produce probabilistic public debt projections that account for SF dynamics.

### Key quantitative elements preserved from the source
- Monte-Carlo runs: "1000 simulations".
- Simulation horizon: "2016-2030".
- Historical sample for VAR estimation: "1990-2015".
- Public debt reported as "over 85 percent of GDP".
- CAT bond characteristics cited: coupons of "Libor plus a spread in the range of 3-20 percent" and "maturities of less than 3 years".

### Policy implications and conclusions (as presented)
- Using ECP revenues to seed and fund an ND-targeted SF can:
  - Reduce the need for emergency borrowing after disasters, supporting fiscal sustainability.
  - Limit the use of volatile ECP flows for recurrent spending that undermines macroeconomic management.
- Strong institutional design and oversight are critical to avoid political capture and ensure that SF resources are used for intended reconstruction and rehabilitation purposes.

*Prepared by Alejandro Guerson; content as presented in the source document.*

### 15.      The calibration of the probability threshold is important, as it determines the annual

### Calibration and Results for a Saving Fund for Natural Disasters (sections 15–30)

### Calibration of the probability threshold and ND frequency
- The probability threshold determines the annual frequency of NDs in the simulations.
- Example calibration: if recent episodes indicate a ND occurs every 5 years, then set Probability[x(t) < X] = 0.2.
- With Probability[x(t) < X] = 0.2 and 1,000 simulations per year through 2016-2035, on average 200 out of the 1,000 simulations in each year would be identified as a ND.
- The distribution of ND intensity (size) is captured by simulated fluctuations of government revenues and expenditures: severe negative impacts on revenues and expenditures correspond to large NDs.

### Modeling of the Saving Fund (SF) inflows and outflows
- In years with no ND, the budget generates an additional overall balance surplus as a percent of GDP that is deposited in the SF.
- Budget contributions to the SF are modeled as a fixed parameter as a percent of the previous year’s GDP.
- Annual saving amount is calibrated to achieve financial sustainability of the Fund with a sufficiently low probability of depletion and stability of the SF stock in expected terms.
- In the event of a ND, a financing inflow to the budget from the SF takes place, computed as the sum of four components:
  - + Gap of non-grant revenues below trend (decline in tax and non-tax revenues after NDs).
  - - Gap of grant revenues above trend (increased donor support after NDs reduces need for SF financing).
  - + Gap of current primary expenditure above trend (higher expenditures in social support and rehabilitation; additional fixed amount as percent of GDP captures below-trend reprioritization of spending).
  - + Gap of capital expenditure above trend (higher public investment after NDs; additional fixed amount as percent of GDP captures below-trend reprioritization of spending).
- Contributions to the budget from the SF continue until the year in which each indicator returns to a level below the value in the year prior to the ND.

### Expenditure re-prioritization assumption
- The SF finances only the increase of fiscal needs above trends; large estimated fiscal costs are offset by reallocation and re-prioritization of pre-ND allocations.
- Example: destruction of public infrastructure equal to 50 percent of GDP might suggest a 5 percent of GDP increase in public investment over ten years, but in practice reconstruction uses reallocated resources so reconstruction expenditures do not require equivalent increases in public investment.
- This justifies the additional savings relative to the estimated trends allowed for current primary and capital expenditures.

### Initial stock (start-up cost) and implications
- The initial stock affects the probability of depletion over the horizon: too low increases probability of depletion; too high increases opportunity cost relative to debt repayment or public investment returns.
- The proposal assumes the start-up cost is funded with existing ECP assets and has not been added to the debt stock at the beginning of the projection horizon (end-2015).

### Simulation strategy and public debt projections
- Simulated series of revenues and primary expenditures allow calculation of primary balances and public debt dynamics using the debt accumulation identity.
- In years with no ND, the budget contributes specified savings to the Fund instead of reducing debt by that amount.
- If a ND occurs, the Fund finances the additional fiscal needs as specified in SF disbursement rules instead of issuing public debt.

### Parametric calibration (selected values preserved from Table 1)
- GDP potential growth, percent: 1.7 (applied through 2016-2022 and 2022-35 as 1.7).
- Implicit interest rate on public debt, percent: 2.5 (applied through 2016-2022 and 2022-35 as 2.5).
- Fiscal consolidation measures, Primary Balance, percent of GDP: -0.4 (2016), 0.1 (2017), 1.6 (2018), 1.6 (2019), 1.6 (2020), 0.5 (2021), 0.0 (2022-35).
  - Non-grant revenue: 1.0 (2016), 1.0 (2017), 1.0 (2018), 1.0 (2019), 1.0 (2020), 0.0 (2021), 0.0 (2022-35).
  - Grant revenue: 0.0 across 2016-2022 and 2022-35.
  - Current primary expenditure: -0.1 across 2016-2020, 0.0 (2021), 0.0 (2022-35).
  - Capital expenditure: 1.5 (2016), 1.0 (2017), -0.5 (2018), -0.5 (2019), -0.5 (2020), -0.5 (2021), 0.0 (2022-35).
- Cumulative fiscal consolidation, percent of GDP: -0.4 (2016), -0.3 (2017), 1.3 (2018), 2.9 (2019), 4.5 (2020), 5.0 (2021), 5.5 (2022-35).

- Government Fund parameters:
  - Initial Fund stock / GDP: 10.00 (percent).
  - Inflows into the Fund from the budget, percent of GDP:
    - Non-grant revenue t-1: 1.500
    - Grant revenue t-1: 0.000
  - Outflows of the Fund to the budget parameters:
    - Storm probability threshold: 0.200
    - Base capital expenditure level / GDP: 0.064
      - Capital expenditure trend / GDP trend: 0.084
      - Year average capital expenditure on reconstruction / GDP: 0.020
    - Base current primary expenditure / GDP: 0.190
      - Current primary expenditure trend / GDP trend: 0.230
      - Year average current expenditure social support and rehabilitation / GDP: 0.040

### Calibration choices and sustainability targets
- ND probability threshold set at 0.2, broadly consistent with historical frequency of ND occurring every 5 years on average.
- Initial size of Fund stock set at 10 percent of GDP; this was chosen to obtain a probability of depletion within the next ten years of 0.08.
- Budget saving flows into the SF in years without a ND were set at 1.5 percent of previous-year’s GDP.
- If in a given simulation the Fund is depleted it is assumed the deficit is covered with debt issuance.

### SF financing to the budget after ND and “base” spending levels
- “Base” capital expenditure set at 6.4 percent of GDP by specifying a gap from the estimated capital expenditure trend in 2015 of 2 percent of GDP.
- “Base” current primary expenditure set at 19 percent of GDP, obtained after specifying 4 percent of trend current primary expenditures associated with ND.
- SF disburses financing equal to the gap between simulated amounts of current primary expenditures and capital expenditures and the calibrated base levels (net of simulated increase in grants).
- Financing flows continue after a ND as long as simulated spending is higher than pre-ND levels.

### Fiscal consolidation alignment
- Fiscal consolidation calibrated in line with macroeconomic framework: cumulative fiscal consolidation measures of 5.5 percent of GDP, largely from revenue measures introduced smoothly through 2016-2021.
- Capital expenditures calibrated to map expected increase related to tropical storm Erika in 2015 and the subsequent unwinding towards 2021.

### Simulation results and sensitivity
- Under proposed parameter calibrations, the SF would be financially sustainable with a low probability of depletion.
- Illustration: in one random simulation (2016-2030) three NDs occur; SF stock increases up to 2020 as 1.5 percent of GDP saved every year to more than 15 percent of GDP; a ND in 2021 leads to SF disbursements to the budget of about 5 percent of GDP.
- Key thresholds from sensitivity analysis:
  - With annual ND probability = 0.2 and specified budget financing rules, annual budget savings of 1.5 percent of GDP are needed to achieve SF financial sustainability over time (no gradual depletion and no unnecessary perpetual accumulation).
  - If ND probability is set at 0.25 (ND every 4 years on average), budget savings of 2 percent of GDP per year would be needed for SF sustainability.
  - If SF starts with assets of 2 percent of GDP and annual ND probability = 0.2, the SF would be depleted during 2016-2030 with probability of more than 30 percent.
  - To reduce the probability of depletion to less than 10 percent (a more prudent level), a SF of at least 8 percent of GDP is required.

### Impact on public debt and risk bands
- With this calibration, public debt would decline to near 60 percent of GDP by 2030 in expected terms, but with significant dispersion depending on realization of shocks.
- Result is broadly in line with regional commitments to reach a public debt ratio of 60 percent of GDP or lower by 2030.
- Even with the fiscal consolidation effort, there remains significant probability that the target will not be met under more extreme ND frequency or severity scenarios; probabilistic fan charts illustrate these risk bands.

### Conclusions and policy recommendations
- A Saving Fund for ND can support immediate needs after NDs and provide reconstruction financing within fiscally sustainable bounds.
- Probabilistic simulations indicate:
  - A saving Fund stock of about 10 percent of GDP and annual budget savings of 1.5 percent of GDP in years with no ND are needed to have sufficient savings commensurate to expected fiscal costs and observed ND frequency.
  - A SF of this size would finance increases in current primary and capital expenditures after a ND with a low probability of depletion (except in the most extreme events).
- Supporting the SF with a strong institutional setup is critical:
  - Include unambiguous budget contribution and disbursement rules with triggers based on verifiable criteria.
  - Clearly-stated objective.
  - Strict information disclosure requirements to ensure transparency.
- Start-up costs and subsequent saving flows could be financed with ECP resources:
  - ECP revenues could fund the start-up and subsequent annual savings with clear savings rules established in legislation.
  - Benefits: avoid issuing public debt for start-up cost; discipline to save for recurrent NDs; reduce scope for allocating ECP revenues to recurrent spending given fiscal consolidation needs.
- Other possible funding sources include donor partner contributions for budget support and international loans for investment projects financed with Fund resources.

*Source: IMF staff analysis and simulations (sections 15–30).*

### 1.      Credit Unions play a prominent role in the Dominican financial system. They provide an

### _cr16245 - 1.      Credit Unions play a prominent role in the Dominican financial system. They provide an

### Overview
- Credit unions provide important financial intermediation, particularly for middle and lower income groups and other key segments of the population that might otherwise find it difficult to access credit through the commercial banking system.
- Objectives differ from commercial banks: commercial banks emphasize profit maximization, while credit unions prioritize serving members with a greater focus on thrift and less on risk taking.
- Credit unions provide services similar to banks and are therefore subject to the same macro-economic shocks and stresses.
- Some of the larger credit unions are comparable in size to the indigenous bank in Dominica.
- There is a pressing need to strengthen the sector through more timely data disclosure and improved regulation and supervision.

### Size and Growth of the Sector (2009–14)
- Total assets of the credit union sector increased from US$172 million (35 percent of GDP) to US$233 million (45 percent of GDP) during 2009–14.
- Average annual growth rate of credit union assets: 7 percent.
- Commercial banks’ assets increased to US$ 462 million (88 percent of GDP) during the same period, with an average annual growth of about 5.5 percent.
- Credit unions extended proportionally more loans: loans-to-GDP ratio increased by about 7 percentage points to 31 percent during 2009–14.

### Membership and Market Penetration
- Credit union membership in Dominica is very high compared to other ECCU countries.
- Among the ECCU countries, Montserrat and Dominica had the highest ratios of credit union members to total population compared to the ECCU average of 61 percent, as of end-2014.
- Higher penetration observed in rural communities and for small and micro businesses with limited access to commercial banks.

### Concentration and Institution Size
- Sector is heavily concentrated: out of 10 credit unions, the largest credit union accounts for about 71 percent of the credit union’s total assets.
- A large majority of credit unions are small; some are likely too small to be viable.
- As of end-2014, total loans to members amounted to 31 percent of GDP.

### Role in Private Sector Credit and Deposits
- Credit union loans account for almost one-third of total private sector credit in Dominica: credit union loans amount to almost 32 percent of the total private sector credit.
- During 2009–14, credit union lending to the private sector increased steadily while commercial bank lending to the private sector declined.
- Sectoral lending composition (NCCU, 2014): mortgage 59%, personal 23%, vehicle 5%, land 7%, business 4%, other loans 2%.
- Credit unions make substantial deposits in commercial banks:
  - During 2009–14, credit union deposits accounted for about 32 percent of commercial bank deposits and about 41 percent of commercial bank’s liquid deposits.
  - Credit unions account for about one-third of the total bank deposits.

### Financial Performance and Soundness (end-2014)
- Non-performing loans (NPLs) for the sector as a whole: 6.4 percent.
- Capital adequacy ratios (aggregate CAR) stood around 8.7 percent as of end-2014—below the minimum regulatory requirement of 10 percent.
- Total equity grew by 6.6 percent over the year (2014).
- Deposit liabilities were the main funding source: at end-2014, deposits represented 90 percent of the sector’s balance sheet.
- Ratio of interest income to total income: approximately 90 percent.
- Asset quality: gross NPLs to total loans had dropped for the entire CU sector, but significant variation exists across individual credit unions.
- Provision coverage: provision for NPLs stood at 60 percent and was deemed insufficient.
- Liquidity: liquid assets represent 10 percent of credit unions’ assets (compared with 40 percent for commercial banks).

### Key Risks Identified
- Credit risk: rapid credit growth increases vulnerability; credit risk is the major source of solvency risk.
- Concentration risk: heavy concentration in a few large CUs, notably the National credit union which accounts for about 71 percent of total CU assets.
- Contagion risk: large CU deposits at commercial banks create potential for deposit withdrawals and financial contagion.
- Liquidity risk: sector vulnerable to bank-run type shocks given limited liquid assets and high interlinkages with commercial banks.
- Interest rate risk: potential US interest rate hikes could increase CU funding costs due to EC$ peg to US$, affecting net interest margin.
- Operational/regulatory risk: FSU lacks power to issue financial penalties; Dominica Co-Operative Societies League lacks legal enforcement powers—response to undercapitalized CUs has been confined to advisories and moral suasion.

### Stress Testing: Method and Results (using end-2014 balance sheet data)
- Approach: bottom-up solvency and liquidity tests covering all CUs in Dominica; CU capital needs assessed against regulatory requirement of 10 percent of risk-weighted assets; liquidity and interest rate scenarios simulated.
- Purpose: understand potential short- and medium-term vulnerabilities, not to estimate recapitalization needs for specific CUs.
- Credit risk stress scenarios:
  - Baseline CAR: 8.7 percent.
  - Scenario 1 (full provisioning of expected losses): CAR decreases to 5.9 percent.
  - Scenario 2 (NPLs increased by one standard deviation, about 50 percent higher, and full provisioning): CAR decreases to 3.3 percent.
- Liquidity risk tests:
  - Simulate loss of confidence resulting in higher than-expected retail deposit withdrawals; examine liquid assets to total assets and liquid assets to short-term liabilities under a run.
  - Example scenario: draw down of deposits by 5% — relevant charts show declines in liquidity ratios (figures in source).
- Interest rate risk stress tests:
  - Scenario: interest rate increases linked to US rate hikes; assumed impact on CU net interest margin.
  - Results: CAR decreases slightly from 8.7 percent in baseline to 8.5 percent with a 2 ppt. increase in interest rates; to 8.6 percent with alternative scenario shown in charts.
  - Conclusion: interest rate risk does not appear to be a large factor affecting profitability or CAR ratios of CUs.

### Consolidated Monetary Accounts and Systemic Implications
- Monetary survey in Dominica excludes credit unions; consolidated monetary account constructed using CU balance sheet information (credit union deposits at commercial banks excluded from Bank’s assets to avoid double counting).
- Even though bank lending to the private sector declined since 2012, credit unions’ financial intermediation activity increased steadily and compensated for declining bank lending, especially to the non-tourism sector.
- Increased overall risk to financial stability given weaker regulation and supervision of credit unions and the lack of a lender of last resort arrangement.
- Given large CU deposits at commercial banks, stronger monitoring and management of macro-financial spillover risks is needed.

### Policy Recommendations and Conclusions
- Regulation and supervision need to be stepped up by strengthening information exchange and collaboration between the ECCB, SRUs, and local regulators.
- Strengthen institutional framework of credit unions, particularly supervision to a level equivalent to commercial banks.
- Special attention and supervision required for big credit unions, notably the National credit union (about 71 percent of CU assets) due to potential spillover risks.
- Credit union loan portfolios should be supervised closely given steady CU lending while banks are deleveraging.
- Ongoing efforts to boost capital are welcome and should be sustained until every credit union is in compliance with capital requirements.
- Consolidation and merger activity can enhance capitalization, stability, and efficiency (example: consolidation plans among three CUs and merger talks among another three cooperatives cited).
- Strengthen collection and compilation of data from credit unions to improve accuracy and effectiveness of regulation and supervision.

*Prepared by Saji Thomas; content based on Dominican credit union sector analysis as presented in the source text.*

### 20.      Stress testing indicates that credit unions are vulnerable to credit and liquidity risks.

### 20. Stress testing indicates that credit unions are vulnerable to credit and liquidity risks.

### Stress test findings
- The stress tests were aimed at understanding the potential short and medium-term vulnerabilities in the system, rather than estimating recapitalization needs (and potential liquidity shortages) for specific credit unions.
- The stress tests indicate that the system is vulnerable to credit and liquidity risks.
- Since this sector has an important role in the financial system of the country the sector needs to be more carefully supervised.

### Dominica: Consolidated Accounts of the Monetary Sector (selected levels, end of period, in millions of Eastern Caribbean dollars)
- Net foreign assets: 2009: 557.3; 2010: 536.2; 2011: 456.7; 2012: 545.0; 2013: 527.9; 2014: 611.9
- Net domestic assets: 2009: 715.6; 2010: 832.8; 2011: 932.9; 2012: 974.1; 2013: 1,039.4; 2014: 1,072.3
- Public sector credit, net: 2009: -140.6; 2010: -139.1; 2011: -96.6; 2012: -133.2; 2013: 4.7; 2014: 21.4
- Private sector credit: 2009: 957.9; 2010: 1,056.8; 2011: 1,121.6; 2012: 1,183.8; 2013: 1,201.9; 2014: 1,210.8
  - credit from banks: 2009: 654.3; 2010: 716.2; 2011: 763.4; 2012: 796.5; 2013: 790.9; 2014: 773.2
  - credit from credit unions: 2009: 303.6; 2010: 340.7; 2011: 358.3; 2012: 387.3; 2013: 411.0; 2014: 437.6
- Other items (net): 2009: -101.7; 2010: -85.0; 2011: -92.1; 2012: -76.5; 2013: -167.2; 2014: -159.9
- Cross deposits: 2009: -30.9; 2010: -32.8; 2011: -34.0; 2012: -35.6; 2013: -38.1; 2014: -41.8
- Broad Money: 2009: 1,242.1; 2010: 1,336.1; 2011: 1,355.6; 2012: 1,483.6; 2013: 1,529.2; 2014: 1,642.4
- Money: 2009: 198.4; 2010: 205.3; 2011: 187.2; 2012: 221.3; 2013: 210.6; 2014: 232.4
- Quasi-money: 2009: 771.0; 2010: 822.9; 2011: 844.2; 2012: 910.5; 2013: 945.7; 2014: 1,014.2
- Credit union currency and deposits (net): 2009: 272.7; 2010: 307.9; 2011: 324.2; 2012: 351.8; 2013: 372.9; 2014: 395.8

### (12-month percentage change) contributions and changes (selected)
- Net foreign assets (contribution): -3.8; -14.8; 19.3; -3.1; 15.9
- Net domestic assets, of which (growth): 16.4; 12.0; 4.4; 6.7; 3.2
- Public sector credit, net (growth): -1.1; -30.6; 37.9; -103.5; 361.0
- Private sector credit (growth): 10.3; 6.1; 5.5; 1.5; 0.7
  - credit from banks (growth): 9.5; 6.6; 4.3; -0.7; -2.2
  - credit from credit unions (growth): 12.2; 5.2; 8.1; 6.1; 6.5
- Broad money (growth): 7.6; 1.5; 9.4; 3.1; 7.4
- NFA contribution: -1.7; -5.9; 6.5; -1.2; 5.5
- NDA contribution: 9.4; 7.5; 3.0; 4.4; 2.2
- Money (growth): 3.5; -8.8; 18.2; -4.8; 10.4
- Credit union currency and deposits (net, growth): 12.9; 5.3; 8.5; 6.0; 6.1

- Note: "2/ Credit union data are from the survey and by Fund staff estimates" and "1/ This table includes estimations on Credit Unions."

### Economic impact of tropical storms (key findings)
- The Caribbean region is one of the most disaster-prone regions in the world; 15 Caribbean islands are among the top 25 countries affected by tropical cyclone disasters.
- The probability of a hurricane hitting seven Caribbean islands, including Dominica, in any given year is above 10 percent.
- Using Acevedo’s work and the World Bank (2015) Rapid Damage and Impact Assessment report on Tropical Storm Erika in 2015:
  - The average annualized actual cost of a tropical cyclone is 7.4 percent of GDP during the period of 1970-2015.
  - The estimated cost is 10.8 percent of GDP (period 1970-2015, as estimated using Acevedo and World Bank data).
- More than half of the tropical cyclones affecting Dominica caused damages of over 50 percent of the country’s national output in the given year (based on the estimated cost series).

### Climate change scenarios and damages
- Acevedo’s estimates of damage increases if all disasters over the last 65 years occurred in warmer temperatures:
  - Low temperature scenario (3°C higher than normal): damages would have been between 4 percent of GDP larger.
  - High temperature scenario (5.6°C higher than normal): damages would have been between 8 percent of GDP larger.
- Charted scenario averages (1950-2014) shown by temperature scenario: Low (3°C), Mean (4.3°C), High (5.6°C).

### Short-term and long-term transmission channels of natural disaster impacts
- Short-term channels:
  - Declines in national production and export receipts due to damage to infrastructure, agriculture, tourism, utilities and service sectors.
  - Imports increase because of reconstruction needs and disruptions to domestic supplies.
  - Deterioration in external trade is partially offset by foreign grants, remittances and insurance payments.
  - Negative impacts on growth and the balance of payments; adverse fiscal effects as the tax base contracts and expenditures for rehabilitation and reconstruction increase.
  - Exchange rate depreciation pressures and higher inflationary pressure; potential spillovers to other countries through regional input/output networks and financial linkages.
- Long-term channels:
  - More severe or longer-lasting versions of short-term channels.
  - Greater risks to public debt sustainability, worsening external sector, high inflation, increased risk of capital flight, and potential banking and/or balance of payments crises.
  - The adequacy of policy responses is critical to mitigate short-term and long-term impacts.

### Historical impacts on Dominica (selected qualitative outcomes)
- Past storms have significantly lowered real GDP, worsened current account balances, and put pressure on the fiscal sector.
- Typical post-disaster patterns illustrated for selected storms (David and Frederick 1979; Hugo 1989; Marilyn and Luis 1995; Dean 2007; Ophelia 2011; Erika 2015):
  - Real GDP falls in large disaster events (indexes shown with year prior to disaster = 100).
  - Current account balance widens reflecting rebuilding needs.
  - Government revenues may decline but are sometimes offset by an increase in grants.
  - Government expenditures jump as repairs boost public capital expenditures, which may boost external debt.
  - Inflation pressures increase.

### Policy recommendations and preparedness measures
- Disaster adaptation and mitigating measures should include:
  - Implementation of improved early warning systems.
  - Enforcement of better building standards and zoning laws that protect or reduce the size of losses.
- Precautionary policy measures to provide financing buffers:
  - Contingency funds.
  - Improved market insurance and regional insurance pools to facilitate internalization of shocks and fast-track disaster relief and reconstruction.
- Fiscal policy:
  - More prudent fiscal policies over the medium term to improve Dominica’s fiscal position, debt sustainability, and financing options.
  - Prudent fiscal positions provide flexibility to absorb adverse shocks through timely countercyclical fiscal policy.
- International support:
  - Precautionary measures could be supported by the donor community and IFI’s to assist the country in promoting self-protection.
- Climate change strategy actions:
  - Expedite ratification of the Paris Climate Agreement and strengthen commitments to slow global warming and reduce emissions by 2030.
  - Prepare regional systems to enable countries’ access to climate finance for adaptation and resilience to climate shocks.
  - Increase the share of renewable energy and improve energy efficiency to help achieve targets.

*Source: IMF staff report chapter titled "20. Stress testing indicates that credit unions are vulnerable to credit and liquidity risks."*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16245.pdf_