## wpiea2021137-print-pdf

## Source details

**Canonical URL:** [wpiea2021137-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021137-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021137-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021137-print-pdf.pdf.json)

---

### I. Introduction — scope and research questions
- Focus: frontier low-income economies issuing sovereign debt in international capital markets (panel of 18 frontier economies, 1990–2019).
- Core questions:
  - Are frontier economies’ increased market access and declining spreads driven by domestic macroeconomic fundamentals (pull factors) or external conditions (push factors: global market volatility, US growth/interest rates)?
  - Is bond issuance driven more by rising financing needs (supply-side) or by investors’ appetite (demand-side)?

### Main empirical findings — summary
- Both country-specific macroeconomic factors and external variables matter; the relative importance varies by market-access measure.
- Market-access measures and dominant drivers:
  - Outstanding PPG external bond stock (stock to GDP): relates to both investor demand and financing needs.
  - Probability of issuance (binary, year-by-year): depends more critically on contemporaneous investor demand.
  - EMBI spreads: respond to both domestic fundamentals and external factors affecting investor risk appetite.
- Stylized trends:
  - Outstanding PPG external bond stock for the 18 frontier economies:
    - USD 4.5 billion at end-2009.
    - USD 55.9 billion by end-2019.
  - 1990–2008: 11 international bond issuances by five frontier economies.
  - 2009–2019: 16 out of 18 frontier economies issued internationally, with 52 bonds.
  - EMBI spreads: average of 1,000-2,000 basis points in the 1990s to around 500 basis points after 2009.
  - Post-2009 issuance rise coincided with decreasing VIX (2009–2019) and low yields in advanced economies.

### Data, variables, and sample
- Sample: annual panel for 18 frontier economies, 1990–2019; 11 out of 18 are African.
- Frontier economies list:
  - PRGT and IDA eligible (15): Cameroon, Cote d'Ivoire, Ethiopia, Ghana, Guinea, Honduras, Kenya, Lao PDR, Mozambique, Moldova, Papua New Guinea, Rwanda, Senegal, Tajikistan and Zambia
  - IDA only (3): Mongolia, Nigeria and Pakistan
- Three market-access proxies:
  - PPG Bond Stock — Outstanding PPG bond stock as percent of GDP — WBG IDS
  - Probability of Issuance — Binary (1 if PPG bonds issued that year, 0 otherwise) — WBG IDS
  - EMBIG spreads — EMBIG Index — Bloomberg (available for 12 countries)
- Key pull factors (country-specific): GDP per capita (USD Thousands), GDP growth (Percent), Fiscal Balance (Percent of GDP), Current Account (Percent of GDP), Reserves (Months of Imports), Government Debt (Percent of GDP), External Debt (Percent of GDP), Terms of Trade (US Dollars), Inflation (Percent), Private Credit to GDP (Percent), Rule of Law (percentile 0–100), IMF Support dummies.
- Key push factors (external): US policy interest rate (Percent), US Bond 10yr yield (Percent), US growth (Percent), CBOE VIX Index (Index).
- Data sources: WEO, WBG IDS/WDI/WGI, IMF Finance Department Database, Bloomberg.

### Empirical approach
- Three model frameworks:
  1. Regressions of PPG Bond Stock (% of GDP) on pull and push variables (pooled OLS; country FE; time FE; country+time FE).
  2. Probability of issuance models: RE probit/logit, conditional logit, and LPM with country and time FE (Hausman tests informed choices).
  3. EMBIG spreads regressions with same identification strategies as (1).
- Identification choices:
  - Main analysis focuses on 2009–2019.
  - Use of lagged explanatory variables to reduce endogeneity where appropriate.
  - GMM/Arellano-Bond used as robustness check for dynamic panel concerns (short T); produced similar results.

### Selected quantitative results — determinants of market access
- PPG Bond Stock (selected associations):
  - Positive associations: Log(GDPpc); GDP growth; Private Credit to GDP; Rule of Law.
  - Negative associations: Inflation; VIX; EMBI (when included).
  - Larger bond stock also associated with lower reserves and higher government debt to GDP (indicative of financing needs/supply-side).
  - US growth generally insignificant for bond stock; US 10-year yield negative and significant in some robustness specs.
- Probability of issuance (nonlinear and LPM results):
  - Positive associations: GDP per capita, Rule of Law, Current Account (in LPM).
  - Negative associations: VIX, EMBI spreads, Reserves (higher reserves → lower likelihood), Private Credit to GDP (suggesting domestic financial development reduces immediate external issuance need).
  - Reported magnitudes (selected):
    - A 10 percent increase in GDP per capita associated with a 2 percentage point higher likelihood of issuing a bond (column 3).
    - An increase in reserves by 1 month of imports associated with a 12 percentage point decrease in likelihood of issuing (column 2).
    - A two-point increase in the VIX index associated with a decrease of 5 percentage points in likelihood of issuing (column 1).
  - Caveat: correlations are non-causal; grants/concessional loans and political choices affect issuance.
- EMBI spreads:
  - Narrower spreads associated with: higher GDP per capita, higher GDP growth, higher current account, higher reserves, lower public debt, lower inflation.
  - IMF program presence associated with narrower spreads in some specifications.
  - Push factors: higher VIX and higher US growth associated with higher EMBI spreads in pooled specs.

### Robustness and additional checks
- GMM/Arellano-Bond for PPG bond stock: similar significance and magnitudes.
- Replacing US growth with US 10-year yield: US 10-year yield significant only for outstanding bond stock (negative sign).
- Differentiating IMF facilities (ECF and SCF):
  - SCF and ECF financing related to lower probability of issuance.
  - SCF related to lower outstanding bond stock.
  - ECF associated with lower EMBI spreads.
- Interaction tests:
  - VIX x Rule of Law examined to test whether stronger institutions mitigate global volatility effects.
  - Appendix 4 suggests interactions matter in some regressions (e.g., EMBIG: VIX x Rule of Law positive and significant; Rule of Law negative and significant in some specs).
- Commodity price effects (BCOM index; Appendix 5):
  - BCOM inclusion materially improved prediction for 2014 (commodity price shock year).
  - BCOM: inverse correlation with PPG bond stock; positive correlation with probability of issuance in nonlinear model.

### Predictive performance — prospective market access exercise
- In-sample prediction (2009–2019) on 12 countries that issued at least twice during 1990–2019:
  - Two best-performing models: pooled LPM and Probit.
  - Average percentage of correct predictions between 2009–19: about 75 percent.
  - For 2013–19: average correct predictions about 67.8 percent.
  - Models struggle in years with large unmodeled shocks (e.g., 2014); inclusion of BCOM improved 2014 predictions.
  - Conclusion: Probit and pooled LPM can inform assessments of prospective market access but require caution due to sample/time limitations.

### Key policy-relevant implications
- Both pull (domestic fundamentals) and push (global conditions) matter for frontier economies’ market access; policy should address both channels.
- Policy priorities for frontier economies seeking sustainable market access:
  - Strengthen domestic fundamentals: raise GDP per capita, bolster rule of law, maintain prudent inflation, manage public debt.
  - Reserve management and domestic financial development:
    - Higher reserves reduce immediate external issuance needs.
    - Deeper private credit markets correlate with larger cumulative external bond stocks but can reduce immediate issuance needs.
  - Monitor global conditions: falling global volatility and low advanced-economy yields increase investor appetite; higher US yields and higher VIX reduce market access.
  - Use probability models (Probit, pooled LPM) as inputs for prospective market-access assessments, and incorporate country-specific shocks (commodity prices, political events) for better accuracy.
- Risks and tradeoffs:
  - External bond issuance can expose governments to exchange rate and other risks; careful debt management is required.
  - COVID-19 highlighted vulnerabilities in debt dynamics and sustainability for frontier economies.

### Selected summary statistics (Appendix 1, Table 9 excerpts)
- PPG Bond Disbursements (USD Billions): N 198; Mean 0.3; St. Dev. 0.7; Min 0; Pctl(25) 0; Pctl(75) 0; Max 5
- PPG Bond Stock (USD Billions): N 198; Mean 1.0; St. Dev. 1.7; Min 0; Pctl(25) 0; Pctl(75) 1.4; Max 11
- GDP Growth: N 198; Mean 5.6; St. Dev. 3.2; Min -6.0; Pctl(25) 3.8; Pctl(75) 7.4; Max 17.3
- GDP per capita (USD Thousands): N 198; Mean 4.2; St. Dev. 2.4; Min 0.9; Pctl(25) 2.5; Pctl(75) 5.3; Max 14.2
- Government Debt to GDP: N 187; Mean 41.4; St. Dev. 18.7; Min 8.6; Pctl(25) 28.4; Pctl(75) 52.4; Max 119.9
- Current Account to GDP: N 198; Mean -7.2; St. Dev. 10.4; Min -41.8; Pctl(25) -10.1; Pctl(75) -2.7; Max 27.2
- Reserves (Months of Imports): N 198; Mean 12.2; St. Dev. 6.7; Min 3.6; Pctl(25) 7.9; Pctl(75) 14.8; Max 34.4
- Rule of Law percentile rank: N 198; Mean 28.9; St. Dev. 15.6; Min 2.8; Pctl(25) 15.4; Pctl(75) 42.2; Max 60.6
- Private Credit to GDP: N 169; Mean 23.5; St. Dev. 13.1; Min 2.7; Pctl(25) 15.1; Pctl(75) 27.5; Max 63.9
- EMBIG Index: N 105; Mean 558.3; St. Dev. 276.4; Min 0.0; Pctl(25) 418.1; Pctl(75) 613.2; Max 1,702.7
- Inflation rate: N 198; Mean 6.8; St. Dev. 5.1; Min -2.2; Pctl(25) 3.2; Pctl(75) 9.3; Max 33.2
- US Growth: N 198; Mean 1.9; St. Dev. 1.5; Min -2.5; Pctl(25) 1.6; Pctl(75) 2.6; Max 2.9
- VIX Index: N 198; Mean 18.2; St. Dev. 5.5; Min 11.1; Pctl(25) 14.2; Pctl(75) 22.5; Max 31.6
- US 10 Years Bond Yield: N 198; Mean 2.5; St. Dev. 0.6; Min 1.8; Pctl(25) 2.1; Pctl(75) 2.9; Max 3.9
- US policy interest rate: N 198; Mean 0.6; St. Dev. 0.8; Min 0.1; Pctl(25) 0.1; Pctl(75) 1.4; Max 2.4

*Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021137-print-pdf.pdf*

### References .............................................................................................................

### References

### I. INTRODUCTION
- Recent years saw a new group of low-income economies increasingly issue sovereign debt in international capital markets; these countries are referred to as "frontier economies."
- Key observations:
  - The amount of issuance by individual frontier economies has been increasing, and the number of these countries accessing international sovereign markets has spiked, especially in the last ten years.
  - Continuous decline in frontier economies’ EMBI spreads suggests reduced perceived risk or increased investor appetite.
- Research questions:
  - Are domestic macroeconomic fundamentals or external factors (global market volatility, US growth and interest rates) the main drivers of frontier economies’ increased market access and declining spreads in the post-Global Financial Crisis decade?
  - Is bond issuance by frontier economies driven more by rising financing needs (supply-side) or by investors’ appetite (demand-side)?
- Main results (summary):
  - Both country-specific macroeconomic factors and external variables are relevant, but results vary with the market access measure used.
  - Market access measured as outstanding external bond debt (stock to GDP) relates to both investor demand and financing needs.
  - Market access measured as the probability of bond issuance in a given year depends more critically on investor demand at that time.
  - Sovereign spreads respond to both domestic and external factors that impact investors’ risk appetite.
- Policy relevance:
  - Findings inform assessments of prospective market access, useful for frontier economy authorities and multilateral institutions (e.g., IMF) when determining LIC access to international capital markets for LIC lending operations.

### Box 1. The IMF Low-Income Country Facilities Framework
- Purpose:
  - Framework established in 2009 to provide concessional financing to poorer and more vulnerable member countries via PRGT and associated facilities.
- Key operational features:
  - Loans via these facilities are provided at zero interest rates until at least end-June 2021 through the PRGT.
  - Limits on the size of PRGT loans exist to maintain trust viability.
  - Eligibility for PRGT depends on level of income per capita, access to international financial markets, and existence/severity of short-term vulnerabilities.
  - Reviews of eligibility occur every two years.
- Blending and graduation:
  - Prior to graduation, LICs with income per capita above a certain threshold and substantial market access may tap concessional resources only in a blend with non-concessional resources.
  - PRGT-eligible members meeting income or market access criteria for blending should tap Fund resources in a blend of concessional PRGT resources with non-concessional GRA resources in a ratio of 1:2.
- Assessment of market access:
  - Past market access is determined by whether a member has borrowed from international financial markets in two/three of the last five years in a cumulative amount of at least 25/50 percent of quota (depending on debt distress rating).
  - Prospective market access assessment for countries at high risk of debt distress relies mainly on forward-looking debt vulnerabilities and judgement.

### II. LITERATURE REVIEW
- Two general categories of drivers for international market access:
  - Pull factors: country-specific characteristics (macroeconomic fundamentals, institutions, liquidity/solvency positions).
  - Push factors: exogenous external variables (global market volatility, US policy rate) affecting investor risk appetite.
- Findings from the literature:
  - Most studies find both domestic fundamentals and external factors influence emerging markets’ ability to tap international capital markets.
  - Comelli et al (2012) and Fratzscher (2012): the impact of domestic fundamentals can be diminished during periods of high global market volatility; push factors were main drivers during the GFC, but country-specific determinants became more important post-GFC.
  - Presbitero et al (2015), Min et al (2003): stronger fiscal/external balances, sound institutions, and solid liquidity and solvency positions increase likelihood of market access.
  - Gelos et al (2011): probability of market access is not influenced by frequency of defaults; a quickly resolved default does not necessarily reduce issuance probability.
- Frontier economies and LICs:
  - Mecagni et al (2014): increase in international bond issuance by African frontier economies—sound fiscal environment is paramount to attract investor demand.
  - Haque et al (2017): strong growth performance and favorable sovereign rating increase LICs’ chances to access markets; warns of risk of excessive borrowing when investors chase yield.
  - Berger et al (2011): frontier markets’ exchange-traded funds have low integration with the world and offer diversification benefits.
- Role of IFI support:
  - Mixed evidence on catalytic effects of IMF/World Bank support:
    - Gelos et al (2011): no catalytic effect from IMF-supported programs.
    - Presbitero et al (2015): IMF support associated with higher likelihood of issuing an international bond.
    - Corsetti et al (2006): theoretical model shows IFI support can help prevent destructive bank runs in LICs; effects stronger with larger support.

### III. DATA
- Sample:
  - Panel dataset: annual data for 18 frontier economies from 1990 to 2019.
  - Frontier economies defined by two criteria:
    - Income criterion: PRGT and/or IDA eligible, excluding economies eligible only due to small-economy status.
    - Market access criterion: issued public and publicly guaranteed (PPG) external bonds at least once between 1990 and 2019.
  - Final list:
    - PRGT and IDA eligible (15 countries): Cameroon, Cote d'Ivoire, Ethiopia, Ghana, Guinea, Honduras, Kenya, Lao PDR, Mozambique, Moldova, Papua New Guinea, Rwanda, Senegal, Tajikistan and Zambia
    - IDA only (3 countries): Mongolia, Nigeria and Pakistan
  - Most sample countries: 11 out of 18 are African countries.
- Three proxies for market access:
  - Outstanding PPG bond stock to GDP ratio (proxy for cumulative volume of issuance).
  - Probability of issuance (binary: 1 if issued PPG bonds in a given year, 0 otherwise).
  - EMBI spreads (investor risk assessment and cost of issuance).
  - Note: EMBI spreads available for 12 countries in sample with more developed bond markets.
- Data sources:
  - Bond stocks and issuance: World Bank Group’s International Debt Statistics (IDS).
  - Country-specific pull factors: IMF World Economic Outlook (WEO, October 2020) and other databases.
  - Institutions (Rule of Law): World Bank’s World Governance Indicators (WGI) percentile rank (0 lowest, 100 highest).
  - IMF support dummies: IMF Finance Department Database.
  - EMBIG spreads: Bloomberg (JP Morgan’s Emerging Market Bond Index Global).
  - Push factors (external): US GDP growth rate, US policy rate, 10-year Treasury bond yields, VIX index (CBOE VIX Index) from Bloomberg and WEO.
- Table 2. Measures of Market Access (variable — description — source):
  - PPG Bond Stock — Outstanding PPG bond stock as percent of GDP — WBG International Debt Statistics (IDS)
  - Probability of Issuance — Binary variable equal to 1 if issued PPG bonds in a given year, 0 otherwise — WBG International Debt Statistics (IDS)
  - EMBIG spreads — EMBIG Index — Bloomberg
- Table 3. Push and Pull Factors — Pull factors (country-specific) include:
  - GDP per capita — Thousand US Dollars — IMF WEO
  - GDP growth — Percent — IMF WEO
  - Fiscal Balance — Percent of GDP — IMF WEO
  - Current Account — Percent of GDP — IMF WEO
  - Reserve Assets — Months of Imports — IMF WEO
  - Government Debt — Percent of GDP — IMF WEO
  - External Debt — Percent of GDP — IMF WEO
  - Terms of Trade — US Dollars — IMF WEO
  - Inflation — Percent — IMF WEO
  - Private Credit Lending — Percent of GDP — WBG WDI
  - Institutions (Rule of Law) — Percentile rank (0–100) — WBG WGI
  - IMF Support — Dummy (1 if PRGT or GRA in place) — IMF Finance Department Database
  - IMF PRGT-ECF Support — Dummy — IMF Finance Department Database
  - IMF PRGT-SCF Support — Dummy — IMF Finance Department Database
- Push factors (global external):
  - US policy interest rate — Percent — Bloomberg
  - US Bond 10yr yield — Percent — Bloomberg
  - US growth — Percent — IMF WEO
  - CBOE VIX Index — Index — Bloomberg
- Data challenges acknowledged:
  - Missing observations for spreads and interest rates for some frontier economies that began accessing markets recently.
  - Baseline identification uses US growth in place of US policy rate due to historically low and stable US interest rates in the past decade.

*wpiea2021137-print-pdf - References*

### 2009. Second, since we are working with a relatively short timeframe, this could lead to

### wpiea2021137-print-pdf - 2009. Second, since we are working with a relatively short timeframe, this could lead to 

### IV. Stylized Facts
- Four key trends:
  - Frontier economies started issuing bonds after 1990, with issuance really taking off after 2009.
  - After the GFC, the proportion of bond issuance became more evenly distributed among the sample of frontier economies.
  - The increase in bond issuance after 2009 coincided with a decreasing trend in global financial market volatility (CBOE VIX).
  - Frontier economies’ sovereign spreads narrowed during the sample period.
- Outstanding PPG external bond stock (proxy for cumulative PPG external bond issuance) for the 18 frontier economies in the sample:
  - USD 4.5 billion at end-2009.
  - USD 55.9 billion by end-2019.
- Frequency and country participation (1990-2019):
  - 1990-2008: 11 international bond issuances by five frontier economies (Pakistan, Moldova, Honduras, Guinea, Mongolia).
  - 2009-2019: 16 out of 18 frontier economies issued internationally, with 52 bonds (Moldova and Guinea exceptions).
  - A median country: no issuances per year during 1990-2008; issued in two to three years during 2009-2019.
  - Five countries issued in at least half of the years 2009-2019: Laos, Senegal, Nigeria, Ghana, Cote d’Ivoire.
  - African economies accounted for about two-thirds of issuance in 2009-2019.
  - Note: Many low issuance countries in 1990-2008 were participating in HIPC; by 2009 all but two HIPC countries in sample had reached completion point (Guinea and Cote d’Ivoire reached HIPC completion point in 2012).
- Correlations with global market conditions:
  - The post-2009 rising bond stock corresponds with a decreasing VIX slope (2009–2019).
  - EMBI spreads among pre-selected frontier economies: average of 1,000-2,000 basis points in the 1990s to around 500 basis points after 2009.
  - The 2009-19 period had low yields in advanced economies and extensive quantitative easing, prompting a search for yield into emerging and frontier economies.

### V. Empirical Analysis — Models and Approach
- Three model frameworks:
  1. Regression of outstanding PPG external bond stock (PPG Bond Stock as % of GDP) on pull and push variables with four identifications: pooled OLS, country fixed effects, time fixed effects, and country+time fixed effects. Equation (1).
  2. Probability of issuance model: binary Issuancei,t (1 if issued in year t) estimated via random effects probit, random effects logit, conditional logit, and a linear probability model (LPM) with country and time fixed effects due to Hausman test results. Equation (2).
  3. EMBIG spreads as dependent variable (EMBIGi,t) regressed on pull and push variables with the same four identifications as the first model. Equation (3).
- Empirical design choices:
  - Analyses focus on 2009-2019.
  - Use of lagged explanatory variables to reduce endogeneity, except for fast-moving variables (VIX, EMBI) and IMF Support dummy.
  - Dynamic panel with T < N (18 countries over 2009-19) creates a short T problem and potential Nickell bias; GMM (Arellano-Bond) used as robustness check for PPG bond stock regressions with similar results.

### Main Results — Determinants of Market Access
- General patterns across dependent variables (quantity: bond stock; timing: probability of issuance; price: EMBI spreads):
  - Pull variables (domestic fundamentals) and push variables (global conditions) both matter, but impact differs by market-access measure.
- Model 1 — Outstanding PPG bond stock (select findings):
  - Positive associations with: GDP per capita (Log(GDPpc)), GDP growth, private sector credit to GDP, rule of law.
  - Negative association with inflation.
  - Larger bond stock associated with lower reserves and higher public debt to GDP (indicative of financing needs/supply side).
  - Push variables: larger bond stock associated with lower VIX and lower EMBI spreads.
  - US growth insignificant across specifications.
- Model 2 — Probability of issuance (nonlinear and LPM results):
  - Nonlinear specifications (RE probit/logit, conditional logit):
    - Probability of issuance positively associated with GDP per capita and rule of law.
    - Inversely associated with global market volatility (VIX) and EMBI spreads.
    - Government debt shows an inverse relationship with probability of issuance in nonlinear specs (contrasting with bond stock results).
  - LPM with fixed effects:
    - Positive association: GDP per capita, rule of law, current account.
    - Negative association: VIX, reserves (higher reserves associated with lower likelihood of issuance).
    - Private credit to GDP shows inverse relationship with probability of issuance (suggesting domestic financial development reduces immediate need to issue externally).
  - Selected economic magnitudes reported:
    - A 10 percent increase in GDP per capita associated with a 2 percentage point higher likelihood of issuing a bond (column 3).
    - An increase in reserves by 1 month of imports associated with a 12 percentage point decrease in likelihood of issuing (column 2).
    - A two-point increase in the VIX index (about the 75th percentile of changes in last 20 years) associated with a decrease of 5 percentage points in likelihood of issuing (column 1).
  - Caveat: Non-causal correlations; non-issuance can reflect lack of need or lack of demand; availability of grants/concessional loans and political choices may affect issuance.
- Model 3 — EMBI spreads (select findings):
  - Narrower EMBI spreads associated with: higher GDP per capita, higher GDP growth, higher current account, higher reserves, lower public debt, lower inflation.
  - IMF program presence associated with narrower spreads.
  - Push factors: higher VIX and higher US growth associated with higher EMBI spreads.
  - Rule of law insignificant in EMBI regressions; private credit to GDP shows a positive (odd) association with spreads but not highly significant.

### Regression Evidence — Selected Coefficients and Table Highlights
- Table 4 (PPG bond stock regressions) selected significance patterns:
  - Log(GDPpc) shows positive and sometimes significant coefficients (e.g., 1.076** in column 1).
  - Government Debt to GDP often positive and significant (e.g., 0.054*** in column 2).
  - Inflation negative and significant in several specs (e.g., -0.095*** column 1).
  - Private Credit to GDP positive and significant across many columns (e.g., 0.051*** column 1).
  - VIX Index shows negative/significant sign in pooled and some specs (e.g., -0.041* column 1).
  - EMBI negative and significant when included (e.g., -0.001** column 2).
- Table 5 (Non-linear marginal effects for probability of issuance):
  - Log(GDPpc) positive and marginally significant in several specs (e.g., 0.241* column 1).
  - Rule of Law positive and significant in some models (e.g., 0.00582** column 1).
  - VIX negative and significant in some models (e.g., -0.0268* column 1).
  - EMBI negative and marginally significant (e.g., -0.000403* column 1).
- Table 6 (LPM probability of issuance):
  - Log(GDPpc) positive and sometimes significant (e.g., 0.248** column 3).
  - Current Account to GDP positive and significant in some specs (e.g., 0.0371*** column 2).
  - Reserves negative and significant in some specs (e.g., -0.121* column 2).
  - Private Credit to GDP negative and significant in some specs (e.g., -0.0337* column 2).
  - VIX negative and significant (e.g., -0.0259*** column 1).
- Table 7 (EMBIG regressions):
  - Log(GDPpc) large negative and significant in country fixed effects/time FE specs (e.g., -1,079.748*** column 2; -997.252*** column 4).
  - GDP growth negative and significant in pooled and pooled with time FE (e.g., -24.912*** column 1).
  - Current Account to GDP negative and significant (e.g., -8.472** column 1).
  - Reserves negative and significant (e.g., -41.622* column 1; -82.952*** column 4).
  - Government Debt to GDP positive and significant (e.g., 8.978*** column 2).
  - IMF Support associated with lower spreads in some specs (e.g., -91.284* column 2; -101.355** column 4).
  - VIX positive and significant in pooled spec (e.g., 15.149** column 1).
  - US growth positive and significant in pooled specs (e.g., 56.109*** column 1; 95.738*** column 2).

### B. Prospective Market Access — Predictive Power of Probability Models
- In-sample prediction exercise (2009-2019) on 12 countries that issued at least twice during 1990-2019 with at least one issuance in 2009-19 (Cameroon, Côte d’Ivoire, Ghana, Honduras, Kenya, Lao PDR, Mongolia, Nigeria, Pakistan, Rwanda, Senegal, Zambia).
- Methodology for correctness:
  - Fitted probability ≥ 0.5 and actual issuance in that year → correct.
  - Fitted probability < 0.5 and no issuance → correct.
  - % Correct Predictions = Amount of Correct Predictions / Amount of Total Predictions.
- Performance:
  - Two best-performing models: pooled LPM and Probit.
  - Average percentage of correct predictions between 2009-19: about 75 percent.
  - For 2013-19 (90 percent of issuances occurred in this sub-period): average correct predictions falls to about 67.8 percent.
  - Models struggle in years with large unmodeled shocks (e.g., 2014 commodity price reversal); inclusion of Bloomberg Commodities Index materially improves 2014 predictions.
- Conclusion: Probit and pooled LPM provide useful inputs for assessing prospective market access, though sample size and timeframe limit definitive conclusions.

### C. Robustness Analysis and Additional Checks
- GMM/Arellano-Bond:
  - Robustness check for PPG bond stock regressions using GMM/Arellano-Bond produced similar significance and magnitude as reported.
- US interest rate proxy:
  - Replacing US growth with US 10-year yields: US 10-year yield significant only for Model 1 (outstanding bond stock) and exhibits a negative sign — higher US yields associated with lower frontier bond stock.
- Differentiated IMF concessional facilities:
  - Replacing IMF Support dummy with ECF and SCF dummies:
    - Both SCF and ECF financing related to lower probability of issuance (IMF financing may preclude issuance need).
    - SCF related to lower outstanding bond stock.
    - ECF associated with lower EMBI spreads (consistent with main results).
- Interaction tests:
  - Interaction between VIX and rule of law explored to test whether stronger fundamentals preserve issuance likelihood during global turmoil.

### Key Policy-Relevant Implications (from analysis)
- Both pull (domestic fundamentals) and push (global conditions) factors matter for frontier economies’ market access; policy should address both:
  - Strengthen domestic fundamentals: raise GDP per capita, bolster rule of law, maintain prudent inflation, and manage public debt to improve access and lower spreads.
  - Manage reserves and domestic financial development recognizing tradeoffs: higher reserves reduce urgency to issue externally; deeper private credit markets can both attract larger cumulative external bond stocks and reduce immediate issuance needs.
  - Monitor global conditions: falling global volatility and low advanced economy yields increase foreign investor appetite; higher US yields and higher VIX can reduce market access.
- Use of probability models (Probit and pooled LPM) can inform assessments of prospective market access, but should incorporate country-specific shocks (commodity prices, political events) for better predictive accuracy.

*Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021137-print-pdf.pdf*

### Appendix 4 show that this is not the case; i.e., good institutions matter less for market access

### wpiea2021137-print-pdf - Appendix 4 show that this is not the case; i.e., good institutions matter less for market access

### Commodity prices
- Several sample countries are rich in natural resources, with commodities accounting for a significant share of their exports.
- Including the Bloomberg Commodities Index (BCOM) as an independent variable materially improved the percentage of correct predictions for the probability of issuance in 2014, a year with a major negative reversal in commodity prices.
- Appendix 5 robustness check: main results are largely robust to inclusion of the BCOM Index (in level terms).
- Correlations:
  - BCOM Index shows a significant and inverse correlation with market access as measured by the PPG bond stock.
  - BCOM Index shows a positive correlation with the probability of issuance in the nonlinear model.
- Interpretation: BCOM appears to affect PPG bond stock via financing needs or supply side, and probability of issuance via investor demand.

### Key conclusions on market access drivers
- Both domestic macroeconomic fundamentals and external conditions have played a role in frontier economies’ increased access to capital markets; effects vary by market-access measure and causal channel.
- Channels and variable roles:
  - Outstanding external bond stock (proxy for total volume of issuance over time) relates to both investors’ demand and frontier economies’ financing needs.
  - Probability of issuance and EMBI spreads (timing and cost of issuance) respond more to investors’ demand.
  - Reserves (level of reserves) operate more through impact on external financing needs (bond supply side).
  - GDP per capita, GDP growth, current account, and rule of law primarily affect investors’ demand for frontier economies’ bonds.
  - Private credit-to-GDP (domestic financial market development):
    - Positively correlates with market access over time (higher capacity to issue internationally, lower repayment risks).
    - Decreases the need to issue externally at a given point in time (greater domestic financing options).
- External factors:
  - Higher global market volatility (lower global risk appetite) is associated with:
    - Lower bond stocks
    - Lower probability of issuance
    - Higher EMBI spreads
  - US interest rates (proxied through US growth in main regressions):
    - Only indirect evidence for a ‘chase for yield’ via a positive association with EMBI spreads.
    - Robustness analysis replacing US growth with US 10-year yields provides firmer evidence: higher bond yields in advanced economies tend to reduce investors’ interest in riskier frontier markets.
- Crisis-period effect:
  - Robustness analysis suggests importance of domestic macroeconomic fundamentals recedes at times of global financial turmoil as investors’ risk appetite falls and there is a flight to safety.
- Predictive performance:
  - Probability of issuance model tested over 2009-2019 for a subset of countries.
  - Probit and pooled LPM specifications show an average percentage of correct predictions over this period of about 77 percent.
  - Caveat: longer time period and larger sample needed for definitive assessment; model can be used as an additional input in judging frontier economies’ prospective market access (relevant for IMF concessional lending framework).

### Policy and research implications
- External bond issuance benefits frontier economies but entails exchange rate and other risks that governments must carefully manage.
- The COVID-19 shock exposed vulnerabilities in debt dynamics and sustainability in frontier economies and emerging markets.
- Further research directions:
  - Extend the timeframe as more data becomes available and assess COVID-19 impact on market access.
  - Test predictive power with out-of-sample analysis when feasible.
  - Apply machine learning techniques to infer causality rather than correlation (examples: causal forest algorithm frameworks cited).

### Appendix 1 — Selected summary statistics (from Table 9)
- PPG Bond Disbursements (USD Billions): N 198; Mean 0.3; St. Dev. 0.7; Min 0; Pctl(25) 0; Pctl(75) 0; Max 5
- PPG Bond Stock (USD Billions): N 198; Mean 1.0; St. Dev. 1.7; Min 0; Pctl(25) 0; Pctl(75) 1.4; Max 11
- GDP Growth: N 198; Mean 5.6; St. Dev. 3.2; Min -6.0; Pctl(25) 3.8; Pctl(75) 7.4; Max 17.3
- GDP per capita (USD Thousands): N 198; Mean 4.2; St. Dev. 2.4; Min 0.9; Pctl(25) 2.5; Pctl(75) 5.3; Max 14.2
- Government Debt to GDP: N 187; Mean 41.4; St. Dev. 18.7; Min 8.6; Pctl(25) 28.4; Pctl(75) 52.4; Max 119.9
- Current Account to GDP: N 198; Mean -7.2; St. Dev. 10.4; Min -41.8; Pctl(25) -10.1; Pctl(75) -2.7; Max 27.2
- Reserve in Months of Imports (% GDP): N 198; Mean 12.2; St. Dev. 6.7; Min 3.6; Pctl(25) 7.9; Pctl(75) 14.8; Max 34.4
- Rule of Law percentile rank: N 198; Mean 28.9; St. Dev. 15.6; Min 2.8; Pctl(25) 15.4; Pctl(75) 42.2; Max 60.6
- Private Credit to GDP: N 169; Mean 23.5; St. Dev. 13.1; Min 2.7; Pctl(25) 15.1; Pctl(75) 27.5; Max 63.9
- EMBIG Index: N 105; Mean 558.3; St. Dev. 276.4; Min 0.0; Pctl(25) 418.1; Pctl(75) 613.2; Max 1,702.7
- Inflation rate: N 198; Mean 6.8; St. Dev. 5.1; Min -2.2; Pctl(25) 3.2; Pctl(75) 9.3; Max 33.2
- US Growth: N 198; Mean 1.9; St. Dev. 1.5; Min -2.5; Pctl(25) 1.6; Pctl(75) 2.6; Max 2.9
- VIX Index: N 198; Mean 18.2; St. Dev. 5.5; Min 11.1; Pctl(25) 14.2; Pctl(75) 22.5; Max 31.6
- US 10 Years Bond Yield: N 198; Mean 2.5; St. Dev. 0.6; Min 1.8; Pctl(25) 2.1; Pctl(75) 2.9; Max 3.9
- US policy interest rate: N 198; Mean 0.6; St. Dev. 0.8; Min 0.1; Pctl(25) 0.1; Pctl(75) 1.4; Max 2.4

### Appendix 2 — Notable regression findings (selected coefficients and significance)
- PPG Bond Stock regressions (Table 8A) — selected marginal/statistical results:
  - Log(GDPpc) (t-1): coefficient examples include 1.096**, 0.613, 1.096**, 0.295, 0.380, 1.260, 0.318, -0.373 (standard errors shown in table).
  - Government Debt to GDP (t-1): coefficients include 0.028*, 0.054***, 0.013, 0.051***, 0.034***, 0.041***, 0.028**, 0.039**.
  - Inflation (t-1): coefficients include -0.099***, -0.008, -0.101**, -0.021, -0.050*, -0.010, -0.047*, -0.018.
  - Private Credit to GDP (t-1): coefficients include 0.050**, 0.029, 0.051***, 0.030, 0.043***, 0.067***, 0.044***, 0.071***.
  - Rule of Law (t-1): coefficients include 0.033**, 0.120***, 0.036**, 0.136***, 0.027***, 0.046***, 0.027***, 0.050**.
  - US Bond 10yr yield (t-1): coefficients include -0.267, -0.233, -0.316**, -0.240* (in columns where included).
  - EMBI: coefficients include -0.0003, -0.001***, 0.0003, -0.001* (in columns where included).
  - Observations shown: 103 and 183 in various specifications; R2 values range from 0.620 to 0.256; Adjusted R2 values range from 0.569 to 0.072.
- Probability of Issuance regressions — Non-Linear Marginal Effects (Table 8B) — selected estimates:
  - Log(GDPpc) (t-1): 0.267**, 0.269**, -0.000700, 0.261**, 0.254**, 0.372 (with standard errors in table).
  - Rule of Law (t-1): 0.00522* , 0.00527* , 0.000240 , 0.00394 , 0.00398 , 0.00240.
  - US 10yr Bond Yield (t-1): -0.243, -0.241, -0.000673, -0.130, -0.139, -0.0440.
  - EMBI: -0.000494* , -0.000509* , -0.00000349 (where included).
  - Observations: 99, 98, 169, 152 across specifications.
- Probability of Issuance regressions — LPM (Table 8C) — selected estimates:
  - Log(GDPpc) (t-1): 0.202, -0.180, 0.245*, 0.857, 0.205*, 0.293, 0.216*, 0.481.
  - Current Account to GDP (t-1): 0.000749, 0.0402**, 0.000702, 0.0429**, 0.00484, 0.00763, 0.00474, 0.00731.
  - Private Credit to GDP (t-1): -0.00504, -0.0388*, -0.00566, -0.0529*, -0.00352, -0.0104, -0.00396, -0.0116.
  - Rule of Law (t-1): 0.00364*, 0.0335*, 0.00407*, 0.0442*, 0.00369, 0.00864, 0.00421, 0.0114.
  - Observations: 99 and 169 across specifications; models vary by pooling, Country FE, Time FE.

*Source: wpiea2021137-print-pdf - Appendix 4 show that this is not the case; i.e., good institutions matter less for market access*

### Appendix 3: IMF ECF/SCF Support Arrangement

### Appendix 3: IMF ECF/SCF Support Arrangement

### Summary of content
- Contains regression tables for 2009 to 2018 analyzing:
  - PPG Bond Stock (Tables 9A, 10A, 11A)
  - Probability of PPG Bond Issuance (Tables 9B, 9C, 10B, 11B, 11C)
  - EMBIG Index (Tables 9D, 10C, 11D)
  - Interaction of VIX Index with Rule of Law (Appendix 4: Tables 10A–10C)
  - Bloomberg Commodities Index (BCOM) related regressions (Appendix 5: Tables 11A–11D)

### Key regression findings — PPG Bond Stock (Table 9A; Appendix variants)
- Log(GDPpc) (t-1):
  - Coefficients reported across specifications: 0.662; 0.157; 0.698; 0.328; 0.244; 1.460; 0.178; -0.274
- GDP growth (t-1):
  - Coefficients: 0.025; 0.007; 0.018; -0.001; 0.050; 0.064*; 0.050; 0.060*
- Current Account to GDP (t-1):
  - Coefficients: 0.0001; -0.035*; -0.011; -0.039*; 0.016; -0.007; 0.012; -0.006
- Reserves Assets/Months of Imports (t-1):
  - Coefficients: 0.028; -0.096**; 0.035; -0.039; -0.103; -0.031; -0.087; -0.017
- Government Debt to GDP (t-1):
  - Coefficients: 0.024; 0.056***; 0.010; 0.055***; 0.032***; 0.044***; 0.027**; 0.041**
- Inflation (t-1):
  - Coefficients: -0.075**; -0.005; -0.083**; -0.021; -0.044**; -0.001; -0.047*; -0.017
- Private Credit to GDP (t-1):
  - Coefficients: 0.066***; 0.045; 0.066***; 0.045; 0.051***; 0.069***; 0.051***; 0.072***
- IMF PRGT-ECF Support:
  - Coefficients: 0.225; 0.035; 0.159; 0.033; 0.083; -0.189; 0.050; -0.227
- IMF PRGT-SCF Support:
  - Coefficients: -0.963***; -0.928***; -1.023***; -0.850***; -0.625*; -0.793***; -0.607*; -0.642***
- Rule of Law (t-1):
  - Coefficients: 0.032**; 0.128***; 0.036**; 0.137***; 0.027***; 0.052***; 0.027***; 0.051***
- VIX Index:
  - Coefficients (where reported): -0.055**; 0.012; -0.038**; -0.014
- US Bond 10yr yield (t-1):
  - Coefficients (where reported): 0.101; -0.001; 0.049*; -0.034
- EMBI:
  - Coefficients (where reported): -0.0002; -0.001**; 0.0004; -0.001**

- Model diagnostics (selected):
  - Observations: 103 (first block) and 183 (second block) depending on specification
  - R2 reported across specs: 0.622; 0.591; 0.597; 0.321; 0.533; 0.571; 0.432; 0.272
  - Adjusted R2: 0.567; 0.452; 0.493; -0.019; 0.500; 0.489; 0.362; 0.086

### Probability of PPG Bond Issuance — Non-linear marginal effects and LPM (Tables 9B, 9C)
- Non-linear marginal effects (Table 9B) — selected coefficients:
  - Log(GDPpc) (t-1): 0.165; 0.172; -0.00445; 0.214*; 0.209*; 0.548
  - Government Debt to GDP (t-1): -0.00640*; -0.00649*; -0.0000531; -0.00353; -0.00351; -0.00625
  - IMF PRGT-ECF Support: -0.0528; -0.0569; -0.0274; -0.152*; -0.158*; -0.186*
  - IMF PRGT-SCF Support: -0.315; -0.284; -0.0552; -0.300*; -0.282*; -0.278*
  - Rule of Law (t-1): 0.00573*; 0.00590*; 0.000427; 0.00465*; 0.00477*; 0.00510
  - VIX Index: -0.0264**; -0.0264**; -0.000258; -0.0195; -0.0187*; -0.0136*
  - EMBI: -0.000400; -0.000423*; -0.00000594 (where reported)
  - Observations: 103; 103; 102; 183; 183; 165

- LPM results (Table 9C) — selected coefficients:
  - Log(GDPpc) (t-1): 0.134; -0.124; 0.204; 1.004; 0.170*; 0.411; 0.175*; 0.458
  - Current Account to GDP (t-1): 0.00180; 0.0331**; 0.000593; 0.0328**; 0.00390; 0.00508; 0.00339; 0.00427
  - Reserves Assets/Months of Imports (t-1): -0.0377; -0.117*; -0.0239; -0.0678; -0.0692*; -0.0874*; -0.0707*; -0.0785*
  - IMF PRGT-ECF Support: -0.0501; -0.220; -0.0682; -0.209; -0.135*; -0.159; -0.157*; -0.146
  - IMF PRGT-SCF Support: -0.220; -0.367; -0.202; -0.315; -0.247*; -0.341**; -0.283**; -0.334*
  - Rule of Law (t-1): 0.00401; 0.0297**; 0.00392; 0.0364*; 0.00378; 0.00867; 0.00482; 0.0102
  - VIX Index: -0.0267***; -0.0265*; (reported as significant in some specs)
  - Observations: 103 and 183 depending on specification

### EMBIG Index regressions (Table 9D and Appendix variants)
- Table 9D (2009–2018) — selected coefficients:
  - Log(GDPpc) (t-1): -55.843; -1,052.293***; -53.175; -1,083.911***
  - GDP growth (t-1): -24.123***; -9.856; -21.683**; -8.186
  - Current Account to GDP (t-1): -8.496**; -4.563; -7.448**; -6.101
  - Reserves Assets/Months of Imports (t-1): -39.997*; -61.748***; -39.491*; -83.339***
  - Government Debt to GDP (t-1): 4.122**; 8.335***; 5.542***; 7.811***
  - Inflation (t-1): 11.719*; 19.039***; 11.668**; 20.613***
  - Private Credit to GDP (t-1): -2.880; 12.976*; -2.482; 16.776*
  - IMF PRGT-ECF Support: 32.234; -108.741; 43.466; -122.070***
  - IMF PRGT-SCF Support: 54.126; 27.344; 31.853; -27.784
  - Rule of Law (t-1): -1.314; 0.352; -1.893; -2.627
  - VIX Index: 14.603**; 3.231 (where reported)
  - US growth (t-1): 58.674***; 96.088*** (where reported)
  - Observations: 99 in the reported specifications
  - R2 reported: 0.524; 0.487; 0.551; 0.482
  - Adjusted R2: 0.458; 0.330; 0.436; 0.242

### VIX x Rule of Law interaction results (Appendix 4: Tables 10A–10C)
- PPG Bond Stock (Table 10A) — selected coefficients:
  - VIX Index x Rule of Law (t-1): -0.003; 0.0004; -0.004***; 0.0002; -0.002**; -0.001; -0.002**; -0.001
  - Rule of Law (t-1): 0.081*; 0.116**; 0.102***; 0.133***; 0.070***; 0.064**; 0.070***; 0.069**
  - Reserves Assets/Months of Imports (t-1): -0.035; -0.111**; -0.050; -0.057; -0.123; -0.031; -0.115; -0.024
  - IMF Support (aggregated label): -0.344; -0.146; -0.276; -0.064; -0.241; -0.129; -0.287; -0.166

- Probability of Issuance — LPM (Table 10B) — selected coefficients:
  - VIX Index x Rule of Law (t-1): 0.000235; 0.0000260; 0.000312; 0.000253; 0.000393; 0.000203; 0.000448; 0.000296
  - VIX Index: -0.0317*; -0.0267; (significance varies by spec)
  - Rule of Law (t-1): reported as 0.0299 and other specifications; some cells report 0 or (.) depending on FE inclusion
  - Constant terms reported across specs (examples): 1.195**; 1.867; 6.546**; -1.320; 0.801**; 0.751; 4.295; 0.0757

- EMBIG regressions with interaction (Table 10C) — selected coefficients:
  - VIX Index x Rule of Law (t-1): 0.659**; 0.742**; 0.683**; 0.781***
  - Rule of Law (t-1): -12.806**; -11.876**; -13.691**; -12.717***
  - IMF Support (aggregated label): -49.215; -116.247***; -52.535; -123.872***
  - Observations: 99
  - R2: 0.550; 0.546; 0.580; 0.551
  - Adjusted R2: 0.487; 0.407; 0.472; 0.343

### Bloomberg Commodities Index (BCOM) related regressions (Appendix 5: Tables 11A–11D)
- PPG Bond Stock (Table 11A) — selected coefficients:
  - BCOM Index: -0.098***; -0.004; -0.075***; -0.046
  - Log(GDPpc) (t-1): 1.074**; 0.480; 1.096**; 0.295; 0.351; 0.751; 0.318; -0.373
  - Reserves Assets/Months of Imports (t-1): -0.057; -0.111**; -0.032; -0.060; -0.131; -0.039; -0.126; -0.027
  - IMF Support (aggregated label): -0.464; -0.131; -0.394; -0.053; -0.339; -0.221; -0.330; -0.208
  - Rule of Law (t-1): 0.033**; 0.121***; 0.036**; 0.136***; 0.026***; 0.043**; 0.027***; 0.050**
  - EMBI: -0.0002; -0.001**; 0.0003; -0.001*

- Probability of Issuance — Non-linear marginal effects (Table 11B) — selected coefficients:
  - Log(GDPpc) (t-1): 0.244*; 0.249*; 0.000585; 0.246**; 0.241**; 0.0581
  - Private Credit to GDP (t-1): -0.00471; -0.00523; -0.0000401; -0.00385*; -0.00382*; -0.000927
  - VIX Index: -0.0367**; -0.0363**; -0.0000291; -0.0208; -0.0188*; -0.000996
  - US growth: 0.0963; 0.0923*; 0.0000315; 0.0257; 0.0219; 0.000883
  - BCOM Index (marginal effect): 0.0303*; 0.0296*; 0.0000511; 0.00999; 0.00752; 0.000890
  - EMBI: -0.000442*; -0.000464*; -0.000000325 (where reported)

- Probability of Issuance — LPM (Table 11C) — selected coefficients:
  - Log(GDPpc) (t-1): 0.192; 0.652; 0.248**; 1.201; 0.196*; 0.657; 0.192*; 0.473
  - Current Account to GDP (t-1): 0.00305; 0.0369***; 0.000836; 0.0380**; 0.00439; 0.00677; 0.00413; 0.00633
  - VIX Index: -0.0292***; -0.0209; -0.0136**; -0.0166*
  - BCOM Index (LPM): 0.0220; 0.0412; 0.00294; 0.0115
  - Observations: 103 and 183 depending on specification

- EMBIG regressions (Table 11D) — selected coefficients:
  - Log(GDPpc) (t-1): -58.827; -1,108.834***; -47.778; -997.252***
  - GDP growth (t-1): -24.606**; -10.912; -22.430**; -9.278
  - Reserves Assets/Months of Imports (t-1): -39.282*; -64.476***; -43.008**; -82.952***
  - Government Debt to GDP (t-1): 4.839**; 8.802***; 5.543***; 8.615***
  - Inflation (t-1): 11.139; 18.706**; 10.742*; 19.667***
  - IMF Support: -24.213; -91.933*; -31.105; -101.355**
  - US growth (t-1): 63.958***; 96.118***
  - BCOM Index: 8.379; -1.827 (where reported)
  - Observations: 99
  - R2: 0.529; 0.495; 0.550; 0.495
  - Adjusted R2: 0.464; 0.339; 0.441; 0.273

### Notation on statistical significance and estimation
- Standard errors are robust and clustered by country where noted.
- Significance notation used in tables:
  - * p<0.1
  - ** p<0.05
  - *** p<0.01
- Model variations include:
  - Pooling vs. Country Fixed Effects (Country FE) vs. Time Fixed Effects (Time FE)
  - RE Logit, RE Probit, Conditional Logit (Country FE), and LPM specifications for probability regressions
- Observations and R2 / Adjusted R2 are reported per specification and vary across tables and panels exactly as shown in the tables above.

*Appendix 3: IMF ECF/SCF Support Arrangement — Tables 9A–11D, as presented in the source PDF.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021137-print-pdf.pdf_