## wpiea2025073-print-pdf

## Source details

**Canonical URL:** [wpiea2025073-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025073-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025073-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025073-print-pdf.pdf.json)

---

### Context and motivation
- Ensuring sustainable debt levels is an important policy priority for Emerging Market and Developing Economies (EMDEs), particularly amid elevated debt vulnerabilities following the COVID-19 pandemic.
- Public debt surged during the COVID-19 pandemic as governments implemented expansionary policies to bolster demand.
- The heavy debt burden, combined with a tightening of monetary policy in major economies post-COVID-19, has increased concerns about debt affordability and high borrowing costs.
- Borrowing costs reflect investors’ perception of the borrower's credit risk, influenced by:
  - Country-specific (pull) factors: economic activity, fiscal position.
  - Global (push) factors: risk perception, demand shocks.

### Role and evolution of ESG in sovereign risk assessment
- Growing research on non-economic and non-financial determinants of sovereign borrowing costs includes Environment, Social, and Governance (ESG) factors.
- ESG considerations have gained popularity due to concerns about climate change and ethical investment drives, supported by initiatives such as:
  - Principles for Responsible Investment (PRI), established in the mid-2000s.
  - Sustainable Development Goals, formalized in 2015.
- Systematic inclusion of the "E" alongside "S" and "G" in sovereign risk assessment is a notable shift; environmental factors were previously considered implicitly in extreme cases (e.g., small countries or those prone to natural disasters).
- Major banks and rating agencies have adjusted investment frameworks and methodologies to incorporate ESG factors.

### Research objective and empirical approach
- Central research question: To what extent do ESG subcomponents influence sovereign borrowing costs in EMDEs?
- Empirical approach:
  - Panel regression analysis using a fixed effects model.
  - Sample: a panel of 79 EMDEs.
  - Main model includes traditional macroeconomic determinants of spreads plus measures of individual ESG components (novelty: focus on ESG subcomponents rather than aggregated indicators).
- An ESG index is constructed using principal component analysis on ESG subcomponents (approach similar to Capelle-Blancard (2019)).

### Main contributions and headline findings
- Threefold contribution:
  1. An improvement in ESG performance tends to lower sovereign spreads for EMDEs.
  2. Unpacking ESG shows that “S” and “G” are the most important drivers — contributing even more than global financial conditions.
  3. Construction of an indicator to summarize the overall impact of ESG factors.
- Policy-relevant headline: Improvements in social (GNI per capita) and governance (government effectiveness, regulatory quality, control of corruption) are highly effective at lowering sovereign borrowing costs, alongside conventional macroeconomic policies.

### Methodology and data (selected details)
- Empirical specification:
  - Spreadsi,t = α + β1 GDP Growthi,t + β2 Reservesi,t−1 + β3 Debti,t−1 + β4 VIXi,t + β5 FFRi,t + β6 ΔGHG emissions per capitai,t + β7 GNI per capitai,t−1 + β8 WGIi,t−1 + ηi + εi,t
- Dependent variable: J.P. Morgan Emerging Market Bond Index Global (EMBIG) spreads (in log terms), annual average of daily EMBIG spreads at market close.
- Panel: unbalanced, 79 countries, period 2001-2021.
- Key variable definitions:
  - Real GDP: annual growth in percent (WEO).
  - International reserves: total reserve assets as percent of GDP (lagged by one year).
  - Public debt: total public debt in percent of GDP (lagged by one year).
  - VIX: log terms (Cboe Volatility Index).
  - Federal Funds Rate: log terms.
  - GHG emissions per capita: gigagram CO2 equivalent, year-on-year difference (Environmental Performance Index).
  - GNI per capita: 2017 PPP$, log terms, lagged (UNDP HDI).
  - WGI variables: government effectiveness, regulatory quality, control of corruption (lagged, WGI units standard normal).

### Stylized facts and correlations
- Distribution of spreads skewed right; most observations below 1000 basis points (bps); notable outliers (e.g., Argentina average annual spread levels surpassed 5000 bps in 2001).
- EMBIG spreads correlated:
  - Positively with GHG emissions.
  - Negatively with GNI per capita, government effectiveness, regulatory quality, and control of corruption.
  - Strongest correlations observed for government effectiveness and regulatory quality.

### Baseline regression results (Table 3, Column (8))
- Sample and fit:
  - Observations: 961.
  - Number of countries: 79.
  - R-squared: 0.424.
  - Country fixed effects: Yes. Time fixed effects: No.
- Global factors:
  - VIX (log): coefficient 0.363*** (standard error 0.049). Interpretation: a one percent increase in the VIX index is predicted to increase spreads by 0.36 percent.
  - Federal Funds Rate (log): coefficient -0.093*** (standard error 0.015). Interpretation: a one percent increase in the Federal Funds Rate is predicted to reduce spreads by 0.09 percent.
- Macroeconomic domestic factors:
  - Real GDP Growth: coefficient -0.036*** (standard error 0.005). Interpretation: a one percentage point increase in real GDP growth is anticipated to reduce spreads by approximately 3.6 percent.
  - Debt to GDP (lagged): coefficient 0.008*** (standard error 0.003). Interpretation: deterioration in the debt-to-GDP ratio is associated with a 0.8 percent increase in spreads.
  - International Reserves to GDP (lagged): coefficient -0.012* (standard error 0.006).
- ESG indicators (all statistically significant in baseline):
  - Greenhouse Gas Emissions per Capita, yoy change: coefficient 0.017* (standard error 0.009).
  - GNI per Capita (log, lagged): coefficient -0.691*** (standard error 0.218).
  - WGI Government Effectiveness (lagged): coefficient -0.294* (standard error 0.153).
  - WGI Regulatory Quality (lagged): coefficient -0.293* (standard error 0.154).
  - WGI Control of Corruption (lagged): coefficient -0.223* (standard error 0.124).
- Summary interpretation:
  - Global risk sentiment and global liquidity conditions matter for EM spreads.
  - Macroeconomic fundamentals affect spreads consistent with prior literature.
  - ESG factors (E, S, G) are statistically significant; governance shows particularly strong relationships; social and environmental factors also matter.

### Robustness analysis — key findings
- Alternative specifications tested: inclusion of time fixed effects; removing outliers; winsorization; 3-year averages; additional explanatory variables; augmenting to test ESG importance over time.
- Selected robustness outcomes:
  - Including time fixed effects (Column (1), 961 observations) increases R-squared to 0.546. Federal Funds Rate coefficient becomes positive in this specification; ESG factors remain significant.
  - Removing outliers (Column (2), 910 observations) and winsorization (Column (3), 961 observations) do not alter main conclusions; ESG variables remain statistically significant.
  - 3-year averages (Column (4), 771 observations) increase coefficients and significance for GHG emissions per capita, government effectiveness, and regulatory quality.
- Time-split test (2001-2015 vs 2016-2021):
  - Greenhouse Gas Emissions per Capita, yoy change: coefficient 0.023*** (standard error 0.009); interaction with Time Dummy -0.035* (standard error 0.019). No evidence that ESG factors became uniformly more important in 2016-2021.
- Additional explanatory variables:
  - Inclusion of tree cover loss, ND-GAIN adaptive capacity, life expectancy, mean years of schooling, full set of WGI variables, and external arrears generally confirms baseline findings.
  - GHG emissions per capita can lose significance when additional environmental indicators are added; many supplementary environmental indicators also lack statistical significance, possibly due to overlapping information.
  - Social and governance variables remain robust negative determinants of spreads across specifications.
  - External arrears dummy (>1 percent of GDP) coefficient 0.127 (not statistically significant in Column (12)).

### Practical implications and ranked policy priorities
- Ranked policy priorities (Figure 3): reduction in spreads from a one standard deviation “improvement” in each independent variable (standardized effects).
  - Top three impacts on reducing spreads among baseline variables:
    1. GNI per capita
    2. Debt to GDP
    3. Regulatory quality
  - Social and governance factors dominate the top five policy levers.
- Environmental considerations:
  - GHG emissions per capita rank lowest among determinants in the ranking exercise, possibly reflecting nascent integration in investment decisions and measurement challenges.
  - However, statistical significance of ESG variables indicates ESG factors matter for sovereign spreads.
- Policy message:
  - EMDEs can lower perceived sovereign risk and borrowing costs by improving social conditions (higher GNI per capita, human capital) and governance (government effectiveness, regulatory quality, control of corruption), alongside sound fiscal and external management.
  - Investors are particularly sensitive to social and governance conditions; countries with poor performance in these categories pay a premium.

### A new ESG Index (PCA-based) and replacement regressions
- Construction:
  - Principal components analysis (PCA) used to combine five baseline ESG variables (GHG emissions per capita, GNI per capita, government effectiveness, regulatory quality, control of corruption) into a single ESG index.
- Baseline model replacing disaggregated ESG with ESG Index (Table 7):
  - Coefficient estimates (standard errors):
    - Real GDP Growth: -0.032*** (0.005)
    - International Reserves to GDP (lagged): -0.012* (0.006)
    - Debt to GDP (lagged): 0.007*** (0.002)
    - VIX (log): 0.374*** (0.045)
    - Federal Funds Rate (log): -0.091*** (0.016)
    - ESG Index (lagged): -0.351*** (0.060)
    - Constant: 4.539*** (0.215)
  - Sample and fit:
    - Observations: 936
    - Number of countries: 79
    - R-squared: 0.374
    - Country fixed effects: Yes. Time fixed effects: No.
- Interpretation:
  - ESG Index (lagged) coefficient -0.351*** indicates a negative and statistically significant relationship between sovereign spreads and the ESG Index.
  - The ESG index provides information beyond macro fundamentals and global financial conditions.

### Conclusion and policy implications
- Scope: Panel of 79 EMDEs over 2001-2021.
- Key substantive conclusions:
  - Macroeconomic country-specific factors (GDP growth, reserves, public debt to GDP) are statistically significant determinants of sovereign borrowing costs.
  - ESG indicators are statistically significant determinants of sovereign spreads:
    - Governance: strong impact via government effectiveness, regulatory quality, and control of corruption.
    - Social proxy: Gross National Income per Capita (log, lagged) — higher living standards relate to lower spreads.
    - Environmental: worsening environmental factors (e.g., GHG emissions) are associated with higher spreads.
  - Robustness: Findings hold under removal of outliers, tests for time fixed effects, and inclusion of additional ESG-related variables.
- Policy recommendations:
  - Alongside improving macroeconomic fundamentals (promoting growth, strengthening fiscal and external positions), authorities should intensify efforts to improve ESG performance.
  - Given sensitivity to ESG sub-components, policy makers should consider prioritizing selected reforms to reduce the premium for poor performance in those categories.
  - Improving ESG performance would not only improve creditworthiness but also provide essential buffers against external shocks.

### Annex highlights (selected statistics and variables)
- Annex I — Summary statistics (selected):
  - Spreads (in log terms): Observations 1017; Mean 5.735; Std. Dev. .789; Min 2.971; Max 8.664
  - Real GDP Growth: Observations 1680; Mean 4.134; Std. Dev. 5.164; Min -33.5; Max 81.787
  - International Reserves to GDP (lagged): Observations 1621; Mean 18.136; Std. Dev. 13.247; Min .231; Max 108.955
  - Debt to GDP (lagged): Observations 1653; Mean 46.613; Std. Dev. 29.901; Min 1.562; Max 344.317
  - VIX (in log terms): Observations 1680; Mean 2.936; Std. Dev. .309; Min 2.407; Max 3.486
  - Federal Funds Rate (in log terms): Observations 1680; Mean -.461; Std. Dev. 1.41; Min -2.079; Max 1.619
  - Greenhouse Gas Emissions per Capita, yoy change: Observations 1680; Mean -.195; Std. Dev. 1.643; Min -8.56; Max 16.9
  - Gross National Income per Capita (in log terms, lagged): Observations 1680; Mean 9.272; Std. Dev. .912; Min 6.411; Max 11.552
  - WGI Government Effectiveness (lagged): Observations 1600; Mean -.114; Std. Dev. .603; Min -1.963; Max 1.563
  - WGI Regulatory Quality (lagged): Observations 1600; Mean -.071; Std. Dev. .646; Min -2.243; Max 1.536
  - WGI Control of Corruption (lagged): Observations 1600; Mean -.288; Std. Dev. .632; Min -1.502; Max 1.718
- Annex II — Sample countries: Baseline model includes 79 countries.
- Annex III — Additional ESG variables described: Tree Cover Loss (lagged); ND-GAIN Vulnerability Adaptive Capacity (lagged); Life Expectancy (lagged); Mean Years of Schooling (lagged); additional WGI subcomponents; External Arrears dummy (>1 percent of GDP).
- Annex IV — ESG Index: figure presents correlation between ESG Index and spreads on the full sample (no outliers removed).

*International Monetary Fund — Do ESG Considerations Matter for Emerging Market Sovereign Spreads? Working Paper No. WP/2025/73*

### Introduction ...........................................................................................................

### wpiea2025073-print-pdf - Introduction ...........................................................................................................

### Context and motivation
- Ensuring sustainable debt levels is an important policy priority for Emerging Market and Developing Economies (EMDEs), particularly amid elevated debt vulnerabilities following the COVID-19 pandemic.
- Public debt surged during the COVID-19 pandemic as governments implemented expansionary policies to bolster demand.
- The heavy debt burden, combined with a tightening of monetary policy in major economies post-COVID-19, has increased concerns about debt affordability and high borrowing costs.
- Borrowing costs reflect investors’ perception of the borrower's credit risk, influenced by:
  - Country-specific (pull) factors: economic activity, fiscal position.
  - Global (push) factors: risk perception, demand shocks.

### Role and evolution of ESG in sovereign risk assessment
- There is a growing body of research on non-economic and non-financial determinants of sovereign borrowing costs, including Environment, Social, and Governance (ESG) factors.
- ESG considerations have gained popularity due to concerns about climate change and ethical investment drives, supported by initiatives such as:
  - Principles for Responsible Investment (PRI), established in the mid-2000s.2
  - Sustainable Development Goals, formalized in 2015.
- The systematic inclusion of the "E" alongside "S" and "G" in sovereign risk assessment is a notable shift; environmental factors were previously considered implicitly in extreme cases (e.g., small countries or those prone to natural disasters).
- Major banks and rating agencies have adjusted investment frameworks and methodologies to incorporate ESG factors.

### Research objective and approach
- Central research question: To what extent do ESG subcomponents influence sovereign borrowing costs in EMDEs?
- Empirical approach:
  - Panel regression analysis using a fixed effects model.
  - Sample: a panel of 79 EMDEs.
  - Main model includes traditional macroeconomic determinants of spreads plus measures of individual ESG components (novelty: focus on ESG subcomponents rather than aggregated indicators).
- An ESG index is constructed using principal component analysis on ESG subcomponents (approach similar to Capelle-Blancard (2019)).

### Main contributions and headline findings
- Threefold contribution:
  1. An improvement in ESG performance tends to lower sovereign spreads for EMDEs.
  2. Unpacking ESG shows that “S” and “G” are the most important drivers — contributing even more than global financial conditions.
  3. Construction of an indicator to summarize the overall impact of ESG factors.
- The paper provides policy insights and a ranked list of ESG areas for policymakers to target in order to reduce funding costs and enhance fiscal sustainability.

### Links to broader literature and themes
- The paper connects to three strands of literature:
  - Debt Sustainability: sovereign spreads affect liquidity and solvency risks and interest rate-growth differentials.
  - Climate Change: rising temperatures, more frequent extreme weather events, and biodiversity loss pose systemic risks.
  - Environmental, Social, and Governance (ESG) Factors: financial sector integration of ESG criteria and growing economics literature on ESG’s role in sovereign spreads (Capelle-Blancard et al., 2019).

### Paper structure (as organized in the source)
- Section 1: Impacts of domestic and global macroeconomic factors and ESG variables on sovereign spreads.
- Section 2: Related literature on determinants of sovereign spreads.
- Section 3: Methodology.
- Section 4: Results.

*Source: Introduction, wpiea2025073-print-pdf — INTERNATIONAL MONETARY FUND*

### Section 5 will provide our conclusions.

### wpiea2025073-print-pdf - Section 5 will provide our conclusions.

### I. What Drives Sovereign Spreads? The Critical Role of ESG Factors
- Sovereign bond spreads reflect market perceptions of a sovereign’s ability and willingness to meet debt obligations, shaped by global and domestic factors (macroeconomic and non-macroeconomic).
- Global factors:
  - Global demand or supply shocks can reduce external and fiscal revenues and thus resources available to repay debt.
  - Price shocks (e.g., decreases in export prices such as oil) reduce external revenues and fiscal capacity.
  - Currency depreciation in weak global demand contexts can increase inflation through higher import costs and reduce the fiscal base.
  - Shocks to global sentiment can reduce capital flows to emerging markets, lowering output and fiscal revenues.
  - U.S. monetary policy (Fed Funds rate) can attract capital to the U.S. and raise emerging market borrowing costs; relationship can be ambiguous and conditional on liquidity (examples from Eichengreen and Mody, 2000; Comelli, 2012).
- Country-specific factors:
  - Macroeconomic: economic activity, fiscal and external positions, and payment track record (defaults, arrears, restructurings) materially influence spreads.
  - Non-macroeconomic (ESG): environmental shocks (e.g., hurricanes), social disturbances, and governance quality have long-term effects on output, fiscal resources, and investor willingness to lend.
- Caveats:
  - Distinctions between macroeconomic and non-macroeconomic factors are not always clear (e.g., income per capita links to both human/social development and economic growth).
  - Variables may carry multiple information layers (e.g., public debt signals fiscal strength and past payment behavior).
- ESG indicators used in the analysis:
  - Environmental: Environmental Performance Index subcomponents; GHG emissions per capita (year-on-year change).
  - Social: UNDP Human Development Index component—Gross National Income per capita (2017 PPP$).
  - Governance: Worldwide Governance Indicators (WGI) subcomponents—government effectiveness, regulatory quality, control of corruption (WGI variables in units of a standard normal distribution, ranging between -2.5 and 2.5).

### II. Literature Review (summarized findings)
- Macroeconomic determinants:
  - Fiscal and external performance are key determinants of external borrowing costs (Akitoby and Stratmann, 2008; Comelli, 2012).
  - Sovereign rating upgrade to “investment grade” lowers spreads by 36 percent (Jaramillo and Tejada, 2011).
  - Net debt (adjusted for government financial assets) can be more informative than gross debt (Hadzi-Vaskov and Ricci, 2022).
  - Global financial developments can dominate country-specific fundamentals in certain episodes (Arora and Cerisola, 2002; Kodres et al., 2008).
- Political determinants:
  - Spreads increase in run-up to elections (Block and Vaaler, 2004).
  - Presidential systems and stronger governance associated with lower spreads (Eichler, 2014).
  - Political risk index: lower political risk reduces spreads, especially during financial stress (Baldacci, Gupta, and Mati, 2011).
- ESG-related literature:
  - Physical climate risk exposure raises spreads (Cevik and Jalles, 2022; Boitan and Marchewka-Bartkowiak, 2022).
  - Mixed evidence on environmental effects; governance often has larger impact than social or environmental components (Capelle-Blancard et al., 2019; Margaretic and Pouget, 2016).
  - ESG factors can impact sovereign credit ratings for advanced economies more than for emerging markets (Pineau, Le, and Estran, 2022).
  - Stronger ESG performance lowers CDS spreads and flattens term structure in some studies (Hübel, 2022).

### III. Methodology and Data
- Empirical specification:
  - Panel regression with country fixed effects estimating:
    - Spreadsi,t = α + β1 GDP Growthi,t + β2 Reservesi,t−1 + β3 Debti,t−1 + β4 VIXi,t + β5 FFRi,t + β6 ΔGHG emissions per capitai,t + β7 GNI per capitai,t−1 + β8 WGIi,t−1 + ηi + εi,t
  - Dependent variable: J.P. Morgan Emerging Market Bond Index Global (EMBIG) spreads (in log terms), annual average of daily EMBIG spreads at market close.
  - Panel: unbalanced, 79 countries, period 2001-2021.
- Variables and specification details (selected):
  - Real GDP: annual growth in percent (WEO).
  - International reserves: total reserve assets as percent of GDP (lagged by one year).
  - Public debt: total public debt in percent of GDP (lagged by one year).
  - VIX: log terms (Cboe Volatility Index).
  - Federal Funds Rate: log terms; sign ambiguous.
  - GHG emissions per capita: gigagram CO2 equivalent, year-on-year difference (Environmental Performance Index).
  - GNI per capita: 2017 PPP$, log terms, lagged (UNDP HDI).
  - WGI government effectiveness, regulatory quality, control of corruption: lagged by one year (WGI units standard normal).
- Stylized facts:
  - Distribution of spreads skewed right; most observations below 1000 basis points (bps); notable outliers (e.g., Argentina average annual spread levels surpassed 5000 bps in 2001).
  - Correlations (Figure 2 and Table 2): EMBIG spreads positively correlated with GHG emissions and negatively correlated with GNI per capita, government effectiveness, regulatory quality, and control of corruption. Correlation strongest for government effectiveness and regulatory quality.

### Results — Baseline Regression (Table 3, Column (8))
- Sample and fit:
  - Observations: 961.
  - Number of countries: 79.
  - R-squared: 0.424.
  - Country fixed effects: Yes. Time fixed effects: No.
- Global factors:
  - VIX (log): a one percent increase in the VIX index is predicted to increase spreads by 0.36 percent. Coefficient in Column (8): 0.363*** (standard error 0.049).
  - Federal Funds Rate (log): a one percent increase in the Federal Funds Rate is predicted to reduce spreads by 0.09 percent. Coefficient in Column (8): -0.093*** (standard error 0.015).
- Macroeconomic domestic factors:
  - Real GDP Growth: coefficient in Column (8): -0.036*** (standard error 0.005). Interpretation: a one percentage point increase in real GDP growth is anticipated to reduce spreads by approximately 3.6 percent.
  - Debt to GDP (lagged): coefficient in Column (8): 0.008*** (standard error 0.003). Interpretation: a deterioration in the debt-to-GDP ratio is found to increase spreads by 0.8 percent.
  - International Reserves to GDP (lagged): coefficient in Column (8): -0.012* (standard error 0.006).
- ESG indicators (all statistically significant in baseline):
  - Greenhouse Gas Emissions per Capita, yoy change: coefficient in Column (8): 0.017* (standard error 0.009). Rising GHG emissions per capita associated with higher spreads.
  - GNI per Capita (log, lagged): coefficient in Column (8): -0.691*** (standard error 0.218). Higher standards of living reduce perceived sovereign risk.
  - WGI Government Effectiveness (lagged): coefficient in Column (8): -0.294* (standard error 0.153). A one-point improvement in governance associated with spread reductions; reductions vary between 22 to 29 percent depending on indicator across specifications.
  - WGI Regulatory Quality (lagged): coefficient in Column (8): -0.293* (standard error 0.154).
  - WGI Control of Corruption (lagged): coefficient in Column (8): -0.223* (standard error 0.124).
- Summary interpretation:
  - Global risk sentiment and global liquidity conditions matter for EM spreads.
  - Macroeconomic fundamentals (growth, debt, reserves) affect spreads consistent with literature.
  - ESG factors (E, S, G) are statistically significant; governance shows particularly strong relationships; social (GNI per capita) and environmental (GHG emissions per capita) also matter.

### Robustness Analysis — Key Findings
- Alternative specifications tested: inclusion of time fixed effects; removing outliers; winsorization; 3-year averages; additional explanatory variables; augmenting to test ESG importance over time.
- Table 4 highlights:
  - Including time fixed effects (Column (1), 961 observations) increases R-squared to 0.546. Federal Funds Rate coefficient becomes positive in this specification (explained by controlling for global shocks); ESG factors remain significant.
  - Removing outliers (Column (2), 910 observations) and winsorization (Column (3), 961 observations) do not alter the main conclusions; ESG variables remain statistically significant.
  - 3-year averages (Column (4), 771 observations) increase coefficients and significance for GHG emissions per capita, government effectiveness, and regulatory quality—suggesting recent trends in environmental and governance performance matter to investors.
- Testing ESG importance over time (Table 5):
  - Time dummy splits periods 2001-2015 and 2016-2021 (Paris Agreement came into force in 2016).
  - Main coefficients remain aligned with baseline; interaction terms do not show evidence that ESG factors have become more important in 2016-2021. Example: Greenhouse Gas Emissions per Capita, yoy change coefficient 0.023*** (standard error 0.009); interaction with Time Dummy -0.035* (standard error 0.019).
- Additional explanatory variables (Table 6):
  - Including tree cover loss, ND-GAIN adaptive capacity, life expectancy, mean years of schooling, full set of WGI variables, and external arrears generally confirms baseline findings.
  - GHG emissions per capita can lose significance when additional environmental indicators are added; many supplementary environmental indicators also lack statistical significance, possibly due to overlapping information.
  - Social and governance variables remain robust negative determinants of spreads across specifications.
  - Including an external arrears dummy (>1 percent of GDP) shows positive coefficient (0.127) but not statistically significant in Column (12), possibly because arrears are captured in public debt.

### Practical Implications for Policymakers and Investors
- Ranked policy priorities (Figure 3): reduction in spreads from a one standard deviation “improvement” in each independent variable (standardized effects).
  - Among baseline variables, the top three impacts on reducing spreads are:
    1. GNI per capita
    2. Debt to GDP
    3. Regulatory quality
  - Social and governance factors dominate the top five policy levers, indicating that improvements in standards of living and institutional robustness are highly effective at lowering sovereign borrowing costs.
  - Fiscal consolidation and prudent debt management remain important, but ESG—particularly “S” and “G”—are crucial policy channels.
- Environmental considerations:
  - GHG emissions per capita rank lowest among determinants in the ranking exercise, possibly reflecting the nascent stage of environmental integration in investment decisions and measurement challenges.
  - Nevertheless, the statistical significance of ESG variables throughout the empirical work indicates ESG factors matter for sovereign spreads.
- Policy message:
  - EMDEs can lower perceived sovereign risk and borrowing costs by improving social conditions (higher GNI per capita, human capital) and governance (government effectiveness, regulatory quality, control of corruption), alongside sound fiscal and external management.
  - Investors are particularly sensitive to social and governance conditions; countries with poor performance in these categories pay a premium.

### A New ESG Index
- Construction:
  - Principal components analysis (PCA) used to combine the five baseline ESG variables (GHG emissions per capita, GNI per capita, government effectiveness, regulatory quality, control of corruption) into a single ESG index (methodology following Capelle-Blancard (2019) approach).
- Illustration:
  - Figure 4 presents the constructed ESG index and sovereign spreads (outliers excluded beyond two standard deviations in spreads). Note: a higher ESG index indicates better ESG performance.
- Uses:
  - The ESG index captures common variation across E, S, and G subcomponents and is used to model a country’s overall ESG performance in relation to sovereign spreads.

*International Monetary Fund — content from wpiea2025073-print-pdf (Section 1–4 and robustness analyses as provided).*

### Annex IV includes the same figure across the entire sample (i.e., not excluding outliers).

### Annex IV includes the same figure across the entire sample (i.e., not excluding outliers).

### Baseline model results: ESG index replacement (Table 7)
- Regression replaces disaggregated ESG components with an ESG Index (lagged).
- Coefficient estimates (standard errors in parentheses):
  - Real GDP Growth: -0.032*** (0.005)
  - International Reserves to GDP (lagged): -0.012* (0.006)
  - Debt to GDP (lagged): 0.007*** (0.002)
  - VIX (in log terms): 0.374*** (0.045)
  - Federal Funds Rate (in log terms): -0.091*** (0.016)
  - ESG Index (lagged): -0.351*** (0.060)
  - Constant: 4.539*** (0.215)
- Sample and fit statistics:
  - Observations: 936
  - Number of countries: 79
  - R-squared: 0.374
  - Inclusion of country fixed effects? Yes
  - Inclusion of time fixed effects? No
- Robust standard errors reported.
- Significance notation:
  - *** p<0.01, ** p<0.05, * p<0.1

### Main findings and interpretation
- Negative and statistically significant relationship between sovereign spreads and the ESG Index (lagged coefficient: -0.351***).
- Traditional macroeconomic determinants remain significant:
  - Real GDP Growth (negative effect): -0.032***.
  - International Reserves to GDP (lagged) (negative effect): -0.012*.
  - Debt to GDP (lagged) (positive effect): 0.007***.
- Global market uncertainty matters:
  - VIX (in log terms) associated with higher spreads: 0.374***.
  - Federal Funds Rate (in log terms) associated with lower spreads: -0.091***.
- Interpretation: The constructed ESG index provides information beyond macro fundamentals and global financial conditions and may be used by investors when assessing country risk.

### Conclusion and policy implications
- Scope: Panel of 79 EMDEs over 2001-2021.
- Key substantive conclusions:
  - Macroeconomic country-specific factors (GDP growth, reserves, public debt to GDP) are statistically significant determinants of sovereign borrowing costs.
  - ESG indicators are statistically significant determinants of sovereign spreads:
    - Governance: strong impact via government effectiveness, regulatory quality, and control of corruption.
    - Social proxy: Gross National Income per Capita (in log terms, lagged) and real income per capita in PPP capture social outcomes; higher living standards relate to lower spreads.
    - Environmental: worsening environmental factors (e.g., GHG emissions) are associated with higher spreads.
  - Robustness: Findings hold under removal of outliers, tests for time fixed effects, and inclusion of additional ESG-related variables.
- Policy recommendations:
  - In addition to improving macroeconomic fundamentals (promoting growth, strengthening fiscal and external positions), authorities should intensify efforts to improve ESG performance.
  - Given sensitivity to ESG sub-components, policy makers should consider prioritizing selected reforms to reduce the premium for poor performance in those categories.
  - Improving ESG performance would not only improve creditworthiness but also provide essential buffers against external shocks.

### Annex I — Summary statistics (selected)
- Spreads (in log terms): Observations 1017; Mean 5.735; Std. Dev. .789; Min 2.971; Max 8.664
- Real GDP Growth: Observations 1680; Mean 4.134; Std. Dev. 5.164; Min -33.5; Max 81.787
- International Reserves to GDP (lagged): Observations 1621; Mean 18.136; Std. Dev. 13.247; Min .231; Max 108.955
- Debt to GDP (lagged): Observations 1653; Mean 46.613; Std. Dev. 29.901; Min 1.562; Max 344.317
- VIX (in log terms): Observations 1680; Mean 2.936; Std. Dev. .309; Min 2.407; Max 3.486
- Federal Funds Rate (in log terms): Observations 1680; Mean -.461; Std. Dev. 1.41; Min -2.079; Max 1.619
- Greenhouse Gas Emissions per Capita, yoy change: Observations 1680; Mean -.195; Std. Dev. 1.643; Min -8.56; Max 16.9
- Gross National Income per Capita (in log terms, lagged): Observations 1680; Mean 9.272; Std. Dev. .912; Min 6.411; Max 11.552
- WGI Government Effectiveness (lagged): Observations 1600; Mean -.114; Std. Dev. .603; Min -1.963; Max 1.563
- WGI Regulatory Quality (lagged): Observations 1600; Mean -.071; Std. Dev. .646; Min -2.243; Max 1.536
- WGI Control of Corruption (lagged): Observations 1600; Mean -.288; Std. Dev. .632; Min -1.502; Max 1.718

### Annex II — Sample countries
- Baseline model includes 79 countries (list in source). Number of countries: 79

### Annex III — Additional ESG variables (descriptions and sources)
- Tree Cover Loss (lagged): Measure of environmental risk in the percentage of forest lost since 2000 — Environmental Performance Index
- ND-GAIN, Vulnerability Adaptive Capacity (lagged): Measure of environmental risk vis a vis a country’s capacity to adapt and react to adverse climate events — Notre Dame Global Adaptation Initiative
- Life Expectancy (lagged): Measure of health outcomes in number of years of life expectancy — UNDP HDI
- Mean Years of Schooling (lagged): Measure of human capital in number of years of schooling — UNDP HDI
- WGI Political Stability and Absence of Violence/Terrorism: Measure of governance risk — WGI
- WGI Voice and Accountability: Measure of governance risk — WGI
- WGI Rule of Law: Measure of governance risk — WGI
- External Arrears: Dummy variable to indicate the presence of arrears that are greater than one percent of GDP — World Bank International Debt Statistics

### Annex IV — ESG Index
- The figure presents the correlation between the ESG Index and spreads on the full sample (i.e., no outliers removed).

*International Monetary Fund — Do ESG Considerations Matter for Emerging Market Sovereign Spreads? Working Paper No. WP/2025/73*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025073-print-pdf.pdf_