## wp18189

## Source details

**Canonical URL:** [wp18189](https://www.imf.org/-/media/files/publications/wp/2018/wp18189.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18189.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18189.pdf.json)

---

### Scope & objectives
- Explores the economic impact of fiscal transparence in emerging market (EM) and developing countries.
- Focuses on two impacts:
  - Effect on countries’ borrowing costs (risk premia / EMBI spreads).
  - Effect on foreign investors’ holdings of an EM’s sovereign debt relative to other EMs’, after controlling for the price effect.

### Measurement and data
- Fiscal transparency conceptualization (Kopits and Craig (1998)) — three dimensions:
  - Transparency of the budget process (availability of reliable information on fiscal policy intentions and forecasts).
  - Transparency of fiscal data (availability of detailed data and information on government operations).
  - Transparency of accountability (measures to ensure officials are held accountable).
- Indicators used:
  - Open Budget Index (OBI) — captures budget process transparency. OBS survey years: 2006, 2008, 2010, 2012, 2015, and 2017 (2017 OBS used as 2016 data).
  - Government Finance Statistics Reporting Index (GFSRI) — constructed by the authors; captures level of detail in fiscal data reported to the IMF.
  - Voice and Accountability index (WGIVA) — World Bank WGI indicator capturing accountability.
- Rationale: separate indicators to avoid losing information and introducing weighting choices.

### Sample, dependent variables, and controls
- Sample: 33 countries (2005–16). List in Appendix C.
- Dependent variables:
  - EMBI Global spread (natural logarithm used in regressions).
  - S1: normalized share of global foreign holdings allocated to country j.
  - S2: share of a country’s general government debt held by foreign investors (Fj/Gj).
- Control variables:
  - Pull (country-specific): Credit Rating (converted to numbers), NLB/GDP, Debt/GDP, Macro (first principal component of GDP growth, inflation, REER, current account ratio), Political Risk Rating (PRR), Financial Risk Rating (FRR).
  - Push (global): Fedfunds, VIX, Crisis dummy (1 if year in 2009-2013).
- Estimation: panel regressions; Breusch-Pagan LM rejects OLS in favor of random effects; Hausman supports random effects. Many variables (EMBI, external debt ratio, political risk, financial risk, Fed Funds rate and VIX) entered in log; push factors lagged one year.

### Stylized facts and descriptive statistics (selected)
- OBI coverage: increased from 17 countries in 2006 to 33 in 2016.
- OBI: average steadily increased but never exceeds 50 (max 100); standard deviation fluctuates between 18 and 22; range between least and most transparent often over 75 points.
- GFSRI: average about 54 in 2012 to close to 80 in 2016 (a 26-point increase in four years).
- WGIVA: averages are negative and stable.
- EMBI spreads:
  - Pre-crisis average: 171 basis points.
  - During crisis: above 550 basis points.
  - Between 2010 and 2016, spreads increased by more than 125 basis points with acceleration in last two years.
- Foreign investor allocations:
  - S1: on average between 3 and 4 percent allocated to each EM in sample.
  - S2: represents on average 36 to 40 percent of these countries’ total general government debt.
- Sample summary 2006–2016:
  - OBI: Mean 47.55 Std. dev 20.24
  - GFSRI: Mean 61.33 Std. dev 36.39
  - WGIVA: Mean -0.15 Std. dev 0.73
  - EMBI: Mean 415.64 Std. dev 398.26
  - S1: Mean 0.03 Std. dev 0.02
  - S2: Mean 0.37 Std. dev 0.19

### Main estimation results — effect on borrowing costs (EMBI)
- Key control effects (consistent signs; examples from Table 1):
  - Credit Rating (-1): negative effect on EMBI (example coefficient -0.0398**).
  - Macro (-1): positive and significant (example coefficient 0.0055***).
  - Political Risk (-1): large negative effect (example coefficient -2.1717***).
  - Financial Risk (-1): negative effect (example coefficient -1.0376**).
  - Crisis (2009-2013): negative coefficient (search-for-yield), example -0.1957***.
  - VIX: positive and significant, example 0.7269***.
  - Fedfunds: negative and significant in regressions, example -10.6074***.
- Fiscal transparency effects on EMBI:
  - OBI (budget process transparency) is the only transparency measure with a statistically significant effect in the baseline.
    - OBI coefficient (baseline): -0.0059** (Table 1, column (1)).
    - Interpretation: a one-point improvement in OBI decreases the spread by 0.59 percent.
    - Example: most transparent country in sample in 2016 (OBI = 89) would have an EMBI global spread 23 percent lower than a country at the median OBI (OBI = 50), all else equal.
  - Heterogeneity:
    - Effect of OBI on spreads matters primarily for high-OBI countries (OBI >= median). Example coefficient OBI High: -0.0058**.
  - GFSRI and WGIVA: coefficients statistically insignificant for EMBI in baseline and heterogeneity specifications.

### Main estimation results — effect on foreign holdings (S1 and S2)
- Key control effects (examples):
  - Debt/GDP (-1): positive and significant (example 0.4178***).
  - Financial Risk (-1): positive and significant (example 0.8939**).
  - Fedfunds: significant and positive in S1 regressions (example 8.7974*** in S1 column (1)); not significant for S2.
- Fiscal transparency effects on foreign holdings:
  - GFSRI (fiscal data transparency) has a statistically significant positive effect.
    - GFSRI coefficient: 0.0035** on S1 (Table 2, column (1)).
    - Interpretation: a one-point improvement in GFSRI leads to 0.35 percent increase in S1 and 0.34 percent increase in S2.
    - Example: the most transparent country in 2016 (GFSRI = 100) would have S1 (resp. S2) 35 percent (resp. 34 percent) higher than the country with GFSRI = 0, all else equal.
    - Increasing a low-GFSRI country to the sample average (GFSRI = 80.69) would raise its foreign holdings by about 28 percent (text calculation).
  - Heterogeneity:
    - Both high and low fiscal-data-transparency countries benefit from improving GFSRI; low-transparency countries benefit slightly more (difference ~3 basis points for S1, ~5 basis points for S2).
  - OBI and WGIVA: coefficients not statistically significant for S1 and S2 in baseline; heterogeneity estimates generally insignificant.

### Robustness checks and alternative specifications
- Random effects supported by Breusch-Pagan LM and Hausman tests; push factors lagged one year to address simultaneity.
- Robustness exercises (qualitative summary):
  - Excluding fiscal space measures — results unchanged.
  - Using squares and interactions of transparency measures — inconclusive.
  - One-year lags of transparency measures — similar results.
  - Constructing GFSRI2 (dropping balance sheet items): effect on spreads unchanged; effect on foreign holdings smaller; with GFSRI2, only high-transparency countries benefit on foreign holdings.
  - Expanding institutional coverage (aggregate data-transparency index): sample shrinks; one-point increase in aggregate index leads to 0.6–0.7 percent increase in foreign holdings.
  - Including exchange rate regime dummy, Chinn-Ito index, Financial Markets Depth index — improve fit but coefficients often not significant; financial risk rating captures many effects.
  - Controlling for World Bank statistical capacity indicator — results robust; indicator not significant.
  - Alternative global liquidity proxies (3-month LIBOR, Wu and Xia Shadow Fed rate) — results robust in sign and qualitatively.
  - Results robust to exclusion of Credit Rating for foreign holdings regressions.
- Selected robustness numerical examples (GFSRI2 results, Table 6 and Table 7):
  - OBI: -0.0058** (standard error 0.0027) in sovereign spreads regression (Observations: 311; Overall R-Square: 0.5582).
  - GFSRI (S1, GFSRI2): 0.0025* (0.0013) (Observations: 195; Overall R-Square: 0.388).

### Interpretation and policy-relevant findings
- Distinct channels by transparency dimension:
  - Budget process transparency (OBI) lowers sovereign borrowing costs (EMBI spreads), primarily for already high-OBI countries.
    - Mechanism: forward-looking budget information helps investors anticipate fiscal issues and price risk premia.
  - Fiscal data transparency (GFSRI) increases foreign holdings of EM sovereign debt (S1 and S2), with benefits for both low- and high-transparency countries.
    - Mechanism: detailed, cross-country-comparable fiscal data (including balance sheet items) facilitates investors’ forward- and backward-looking analyses and portfolio allocation decisions.
- Magnitudes (baseline interpretations preserved):
  - OBI: one-point increase → 0.59 percent decrease in EMBI spread (baseline).
  - GFSRI: one-point increase → 0.35 percent increase in S1; 0.34 percent increase in S2.
  - Example contrasts: OBI = 89 vs OBI = 50 → 23 percent lower EMBI spread; GFSRI = 100 vs GFSRI = 0 → 35 percent higher S1.
- Overall message:
  - Improving fiscal transparency yields measurable financial benefits for emerging market economies.
  - Different transparency dimensions matter for different investor outcomes: budget process transparency for borrowing costs; fiscal data transparency for foreign demand and holdings.

*_Source: wp18189 - Section IV reports the regression results and the last section concludes with policy._*

### References ________________________________________________________________25

### wp18189 - References ________________________________________________________________25

### Scope & objectives
- Explores the economic impact of fiscal transparence in emerging market (EM) and developing countries.
- Focuses on two impacts:
  - Effect on countries’ borrowing costs (risk premia / EMBI spreads).
  - Effect on foreign investors’ holdings of an EM’s sovereign debt relative to other EMs’, after controlling for the price effect.

### Measurement and data
- Fiscal transparency conceptualization follows Kopits and Craig (1998): three dimensions:
  - Transparency of the budget process (availability of reliable information on fiscal policy intentions and forecasts).
  - Transparency of fiscal data (availability of detailed data and information on government operations).
  - Transparency of accountability (measures to ensure officials are held accountable).
- Indicators used (each capturing one dimension):
  - Open Budget Index (OBI) from the International Budget Partnership (IBP) — captures budget process transparency. OBS survey years noted: 2006, 2008, 2010, 2012, 2015, and 2017.
  - Government Finance Statistics Reporting Index (GFSRI) — constructed by the authors; captures the level of detail in fiscal data reported to the IMF for publication in the Government Finance Statistics Yearbook (GFSY).
  - Voice and Accountability index from the World Bank’s Worldwide Governance Indicators (WGI) — captures accountability dimension.
- Rationale for separate indicators: investors may weight these dimensions differently; aggregating could lose information and introduce weighting choices.

### Key empirical findings
- Fiscal transparency has two measurable economic benefits:
  - Positively affects EM countries’ borrowing costs (reduces EMBI spreads).
  - Increases foreign investors’ demand for sovereign debt.
- Different dimensions drive different benefits:
  - High transparency of the budget process contributes most to lower EMBI spreads.
  - Availability of detailed and cross-country comparable fiscal data (GFSRI) leads to higher foreign holdings of EM sovereign debt.
- Heterogeneity of effects:
  - Effect of fiscal data transparency (GFSRI) applies to all countries in the sample.
  - Effect of budget process transparency (OBI) is mainly felt by countries that have already achieved a sufficient degree of budget process transparency.
- Supporting numeric finding from related literature:
  - Choi and Hashimoto (2017) find that subscription to the IMF’s Data Standards Initiatives (SDDS and GDDS) leads to a 15 percent reduction in EMBI spread one year following such reforms.

### Theoretical considerations
- Two reasons why changes in spreads and holdings may capture different channels:
  - Stiglitz and Weiss (1981) argument: imperfect credit markets imply interest rates may not clear demand and supply for funds; tracking volumes (foreign holdings) provides additional insights.
  - Portfolio allocation choices and investor constraints (risk mitigation, preferences) may cause non-linear responses of holdings to interest rates.

### Contribution relative to existing literature
- Twofold contribution claimed:
  - Investigates the three dimensions of fiscal transparency separately using different indicators (OBI, GFSRI, WGI-Voice and Accountability).
  - Explores both price impact (EMBI spread) and quantity impact (foreign holdings of EM sovereign debt).

### Appendix and tables (inventory from source)
- Appendix sections and page references:
  - A. Construction of the GFSRI — page 27
  - B. Summary Statistics — page 29
  - C. List of Countries Used in the Analyses — page 30
  - D. Regression Results Using Less Stringent Measure of Fiscal Data Transparency — page 31
  - E. Numerical Conversion of Sovereign Credit Ratings — page 33
- Tables listed with titles and page references:
  - Table 1. Effect of Fiscal Transparency on Sovereign Spreads — page 18
  - Table 2. Effect of Fiscal Transparency on Foreign Investors’ Holdings of EM Debt — page 20
  - Table 3. Scores for the GFS Reporting Index (GFSRI) — page 28
  - Table 4. Descriptive Statistics — page 29
  - Table 5. Pairwise Correlations — page 30
  - Table 6. Effect of Fiscal Transparency on Sovereign Spreads — page 31
  - Table 7. Effect of Fiscal Transparency on Foreign Investors' Holdings of EM Debt — page 32

### Glossary (selected acronyms preserved as in source)
- AREAER IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions
- EMBI Emerging Market Bond Index
- EMs Emerging and developing economies
- Fedfunds Federal funds rate
- FRR Financial Risk Rating
- GDDS General Data Dissemination Standards
- GFSM 2014 Government Finance Statistics Manual, 2014
- GFSRI Government Finance Statistics Reporting Index
- GFSY Government Finance Statistics Yearbook
- IFS International Financial Statistics
- ICRG International Country Risk Guide
- IBP International Budget Partnership
- OBI Open Budget Index
- PRR Political Risk Rating
- PRS Political Risk Services
- QPSD Quarterly Public-Sector Debt
- ROSC Report on Observation of Standards and Codes
- SDDS Special Data Dissemination Standards
- VIX CBOE Volatility Index
- WGI World Bank’s Worldwide Governance Indicators
- WGIVA The Voice and Accountability Index

*Source: wp18189 - References ________________________________________________________________25*

### Section IV reports the regression results and the last section concludes with policy

### wp18189 - Section IV reports the regression results and the last section concludes with policy

### Data, measures, and sample
- Three fiscal transparency measures used:
  - Open Budget Index (OBI): 0–100 scale; assesses publication of eight budget documents and budget process openness.
  - Government Finance Statistics Reporting Index (GFSRI): 0–100 normalized score based on data submitted to IMF Statistics Department following GFSM2014.
  - Voice and Accountability Index (WGIVA): World Bank governance indicator ranging between -2.5 and 2.5.
- Dependent variables:
  - EMBI Global spread (natural logarithm of EMBI used in regressions).
  - S1: normalized share of global foreign holdings allocated to country j (formula (1) in text).
  - S2: share of a country’s general government debt held by foreign investors (Fj/Gj) (formula (2) in text).
- Sample and timing:
  - 33 (mostly emerging and developing) economies (list in Appendix C).
  - Analysis period: 2005–16.
  - OBI annual data available for 2006, 2008, 2010, 2012, 2015, and 2017; 2017 OBS used as 2016 data in analyses.
- Control variables:
  - Pull (country-specific): Credit Rating (Bloomberg, converted to numbers), NLB/GDP, Debt/GDP, Macro (first principal component of GDP growth, inflation, REER, current account ratio), Political Risk Rating (PRR), Financial Risk Rating (FRR).
  - Push (global): Fedfunds, VIX, Crisis dummy (1 if year in 2009-2013).
- Estimation framework:
  - Baseline model derived from Edwards (1986) asset pricing formulation.
  - General specification: Y_it = β0 + β1 X_it + β2 Z_t + α_i + ε_it (equation (6)).
  - Tests: Breusch-Pagan LM rejects OLS in favor of random effects; Hausman suggests random effects preferred.
  - Log transformations: EMBI, external debt ratio, political risk, financial risk, Fed Funds rate and VIX entered in log; push factors enter with a one-year lag.

### Stylized facts and descriptive statistics
- OBI coverage increased from 17 countries in 2006 to 33 in 2016.
- Average OBI steadily increased over time but never exceeds 50 (max 100) — budget process transparency remains broadly moderate.
- OBI standard deviation fluctuates between 18 and 22; range between least and most transparent often over 75 points.
- GFSRI:
  - Average increased slowly 2006–2012, then accelerated.
  - From about 54 in 2012 to close to 80 in 2016 (a 26-point increase in four years).
- WGIVA averages are negative and stable over time.
- EMBI spreads:
  - Pre-crisis average: 171 basis points.
  - During crisis: above 550 basis points.
  - Between 2010 and 2016, spreads increased by more than 125 basis points with acceleration in last two years.
- Foreign investors allocations:
  - S1: on average between 3 and 4 percent allocated to each EM in sample.
  - S2: represents on average 36 to 40 percent of these countries’ total general government debt.
- Pairwise correlations:
  - Correlation OBI–GFSRI not significantly different from zero.
  - Highest correlation OBI–WGIVA = 0.6.
  - High OBI and WGIVA associated with low EMBI and high foreign holdings; correlations are significant but low.
  - GFSRI correlations with dependent variables not statistically different from zero in unconditional correlations.

### Main estimation results — effect on borrowing costs (EMBI)
- Baseline regression period: 2005–16.
- Key significant control effects (consistent signs):
  - Credit Rating (-1): negative effect on EMBI (e.g., coefficient -0.0398** in Table 1 column (1)).
  - Macro (-1): positive and significant (Macro deterioration raises EMBI), coefficient 0.0055***.
  - Political Risk (-1): large negative effect (better political risk lowers EMBI), coefficient -2.1717***.
  - Financial Risk (-1): negative effect, -1.0376**.
  - Crisis (2009-2013): negative coefficient (search-for-yield) e.g., -0.1957***.
  - VIX: positive and significant, e.g., 0.7269***.
  - Fedfunds: negative and significant in these regressions (counterintuitive but documented), e.g., -10.6074***.
- Fiscal transparency effects on EMBI:
  - OBI (budget process transparency) is the only transparency measure with a statistically significant effect in the baseline.
    - OBI coefficient (baseline): -0.0059** (Table 1, column (1)).
    - Interpreted in text: a one-point improvement in OBI decreases the spread by 0.59 percent.
    - Example: most transparent country in sample in 2016 (OBI = 89) would have an EMBI global spread 23 percent lower than a country at the median OBI (OBI = 50), all else equal.
  - Heterogeneity by OBI level:
    - Effect of OBI on spreads matters primarily for high-OBI countries (OBI >= median).
    - OBI High coefficient e.g., -0.0058** (Table 1, column (2)).
    - OBI Low coefficients generally insignificant except some sensitivity in alternative specifications.
  - GFSRI and WGIVA: coefficients statistically insignificant for EMBI in baseline and heterogeneity specifications.

### Main estimation results — effect on foreign holdings (S1 and S2)
- Dependent variables: S1 (share of global foreign holdings allocated to country) and S2 (share of country debt held by foreigners).
- Key significant control effects:
  - Debt/GDP (-1): positive and significant (e.g., 0.4178***).
  - Financial Risk (-1): positive and significant (e.g., 0.8939**).
  - Fedfunds: significant and positive in S1 regressions (counterintuitive direction), e.g., 8.7974*** in S1 (column (1)); not significant for S2.
- Fiscal transparency effects on foreign holdings:
  - GFSRI (fiscal data transparency) is the transparency measure with a statistically significant positive effect.
    - GFSRI coefficient: 0.0035** on S1 (Table 2, column (1)).
    - Interpreted in text: a one-point improvement in GFSRI leads to 0.35 percent increase in S1 and 0.34 percent increase in S2.
    - Example: the most transparent country in 2016 (GFSRI = 100) would have S1 (resp. S2) 35 percent (resp. 34 percent) higher than the country with GFSRI = 0, all else equal.
    - Increasing a low-GFSRI country to the sample average (GFSRI = 80.69) would raise its foreign holdings by about 28 percent (text calculation).
  - Heterogeneity by GFSRI level:
    - Both high and low fiscal-data-transparency countries benefit from improving GFSRI.
    - Low-transparency countries benefit slightly more (difference ~3 basis points for S1, ~5 basis points for S2).
  - OBI and WGIVA: coefficients not statistically significant for S1 and S2 in baseline; heterogeneity estimates generally insignificant.

### Robustness checks
- Estimation choices and diagnostics:
  - Random effects specification supported by Breusch-Pagan LM and Hausman tests.
  - Push factors entered with one-year lag to address simultaneity.
- Robustness exercises performed (results reported qualitatively):
  - Excluding fiscal space measures (Debt and NLB) — results unchanged.
  - Using squares and interactions of transparency measures — inconclusive.
  - Using one-year lags of transparency measures — similar results.
  - Constructing GFSRI2 (dropping balance sheet items): effect on spreads unchanged; effect on foreign holdings smaller; with GFSRI2, only high-transparency countries benefit on foreign holdings.
  - Expanding institutional coverage (Budgetary Central, Central, General government average): sample shrinks; one-point increase in aggregate data-transparency index leads to 0.6–0.7 percent increase in foreign holdings.
  - Including exchange rate regime dummy (de facto floating), Chinn-Ito index, Financial Markets Depth index — improve model fit but coefficients often not significant; financial risk rating correlated with these variables and captures their effects.
  - Controlling for World Bank statistical capacity indicator — results robust; indicator not significant.
  - Alternative global liquidity proxies (3-month LIBOR, Wu and Xia Shadow Fed rate) — results robust in sign and qualitatively.
  - Results robust to exclusion of Credit Rating for foreign holdings regressions.

### Interpretation and policy-relevant findings
- Distinct channels by transparency dimension:
  - Budget process transparency (OBI) lowers sovereign borrowing costs (EMBI spreads), primarily for already high-OBI countries.
    - Mechanism: forward-looking budget information helps investors anticipate fiscal issues and price risk premia.
  - Fiscal data transparency (GFSRI) increases foreign holdings of EM sovereign debt (both S1 and S2), with benefits for both low- and high-transparency countries.
    - Mechanism: detailed, cross-country-comparable fiscal data (including balance sheet items) facilitates investors’ forward- and backward-looking analyses and portfolio allocation decisions.
- Magnitudes:
  - OBI: one-point increase → 0.59 percent decrease in EMBI spread (baseline).
  - GFSRI: one-point increase → 0.35 percent increase in S1; 0.34 percent increase in S2.
  - Example contrasts: OBI = 89 vs OBI = 50 → 23 percent lower EMBI spread; GFSRI = 100 vs GFSRI = 0 → 35 percent higher S1.
- Overall message:
  - Improving fiscal transparency yields measurable financial benefits for emerging market economies.
  - Different transparency dimensions matter for different investor outcomes: budget process transparency for borrowing costs; fiscal data transparency for foreign demand and holdings.

*Source: IMF Working Paper (content unit: wp18189 - Section IV reports the regression results and the last section concludes with policy).*

### References

### wp18189 - References

### A. Construction of the GFSRI
- Most aggregated data: net transactions in assets and liabilities (푋푋) with three subcomponents: net/gross investment in nonfinancial assets (퐴퐴), net acquisition of financial assets (퐵퐵), and net incurrence of liabilities (퐶퐶).
- Disaggregated components example: 퐴퐴 broken down into a1: fixed assets, a2: inventories, a3: valuables, a4: non-produced assets; 퐵퐵 and 퐶퐶 broken down by residency: 푏푏1: domestic debtor, 푏푏2: external debtor, 푐푐1: domestic creditor, 푐푐2: external creditor.
- Scoring rules:
  - Each disaggregated item (푎푎1–푎푎4, 푏푏1, 푏푏2, 푐푐1, 푐푐2): score = 1 if data reported, 0 if not.
  - Subcomponent score (example 퐴퐴): score = 2 if data reported by country; otherwise score = 1 if 퐴퐴 can be computed from subcomponents (퐴퐴 = 푎푎1 + 푎푎2 + 푎푎3 + 푎푎4), else 0.
  - Main aggregate (푋푋): score = 3 if published by country authorities; else score = 2 if recoverable from all subcomponents (푋푋 = 퐴퐴 + 퐵퐵 − 퐶퐶); else 0.
- Rationale: distinction between published aggregates and aggregates recoverable only from subcomponents preserves users’ ability to check internal consistency; published aggregates score higher.
- Same logic applied to the other two blocks: Government’s current operations and Balance sheet.
- Final index construction: sum scores across the three blocks for each country, then normalize to range between 0 and 100.
- Interpretation: GFSRI score close to 100 means high fiscal data transparency; low scores denote high fiscal opacity.

### B. Summary statistics (selected)
- Table 4. Descriptive Statistics — Fiscal transparency measures (OBI, GFSRI, WGIVA) by year:
  - 2006:
    - OBI: N 17 Mean 48.24 Std. dev 21.50 Min 3 Max 86
    - GFSRI: N 24 Mean 50.07 Std. dev 38.24 Min 0 Max 98
    - WGIVA: N 24 Mean -0.13 Std. dev 0.79 Min -1.75 Max 1.08
  - 2008:
    - OBI: N 28 Mean 46.61 Std. dev 18.17 Min 10 Max 87
    - GFSRI: N 32 Mean 52.54 Std. dev 37.95 Min 0 Max 100
    - WGIVA: N 32 Mean -0.17 Std. dev 0.75 Min -1.70 Max 1.01
  - 2010:
    - OBI: N 30 Mean 48.17 Std. dev 20.24 Min 0 Max 92
    - GFSRI: N 32 Mean 56.79 Std. dev 35.42 Min 0 Max 100
    - WGIVA: N 32 Mean -0.18 Std. dev 0.75 Min -1.68 Max 1.10
  - 2012:
    - OBI: N 30 Mean 47.60 Std. dev 19.67 Min 4 Max 90
    - GFSRI: N 31 Mean 53.63 Std. dev 38.07 Min 0 Max 100
    - WGIVA: N 31 Mean -0.15 Std. dev 0.70 Min -1.64 Max 1.08
  - 2015:
    - OBI: N 33 Mean 46.84 Std. dev 21.38 Min 2 Max 86
    - GFSRI: N 32 Mean 71.92 Std. dev 32.08 Min 0 Max 100
    - WGIVA: N 32 Mean -0.13 Std. dev 0.72 Min -1.66 Max 1.04
  - 2016:
    - OBI: N 33 Mean 48.09 Std. dev 22.13 Min 0 Max 89
    - GFSRI: N 32 Mean 79.98 Std. dev 28.78 Min 0 Max 100
    - WGIVA: N 32 Mean -0.14 Std. dev 0.73 Min -1.62 Max 1.00
- Table 4. Dependent variables (EMBI, S1, S2) by year:
  - 2006:
    - EMBI: N 24 Mean 170.92 Std. dev 108.23 Min 59.30 Max 538.07
    - S1: N 18 Mean 0.04 Std. dev 0.02 Min 0.01 Max 0.09
    - S2: N 18 Mean 0.37 Std. dev 0.20 Min 0.05 Max 0.79
  - 2008:
    - EMBI: N 33 Mean 564.28 Std. dev 402.33 Min 162.25 Max 1864.00
    - S1: N 18 Mean 0.04 Std. dev 0.02 Min 0.00 Max 0.08
    - S2: N 18 Mean 0.34 Std. dev 0.18 Min 0.03 Max 0.67
  - 2010:
    - EMBI: N 33 Mean 330.77 Std. dev 230.70 Min 77.35 Max 1081.19
    - S1: N 19 Mean 0.03 Std. dev 0.02 Min 0.00 Max 0.06
    - S2: N 19 Mean 0.36 Std. dev 0.19 Min 0.02 Max 0.67
  - 2012:
    - EMBI: N 31 Mean 416.17 Std. dev 271.31 Min 147.66 Max 1062.02
    - S1: N 20 Mean 0.03 Std. dev 0.02 Min 0.00 Max 0.05
    - S2: N 20 Mean 0.40 Std. dev 0.20 Min 0.01 Max 0.69
  - 2015:
    - EMBI: N 33 Mean 491.22 Std. dev 578.44 Min 96.23 Max 2776.13
    - S1: N 20 Mean 0.03 Std. dev 0.02 Min 0.00 Max 0.05
    - S2: N 20 Mean 0.38 Std. dev 0.21 Min 0.02 Max 0.68
  - 2016:
    - EMBI: N 33 Mean 459.30 Std. dev 465.95 Min 106.54 Max 2743.53
    - S1: N 20 Mean 0.03 Std. dev 0.02 Min 0.00 Max 0.05
    - S2: N 20 Mean 0.39 Std. dev 0.20 Min 0.02 Max 0.69
- Table 4. Summary statistics for 2006–2016 (entire sample):
  - OBI: Mean 47.55 Std. dev 20.24
  - GFSRI: Mean 61.33 Std. dev 36.39
  - WGIVA: Mean -0.15 Std. dev 0.73
  - EMBI: Mean 415.64 Std. dev 398.26
  - S1: Mean 0.03 Std. dev 0.02
  - S2: Mean 0.37 Std. dev 0.19

### C. Pairwise correlations (selected, Table 5)
- Correlations (asterisks indicate significance at a 10 percent level):
  - OBI with GFSRI: -0.013
  - OBI with WGIVA: 0.589*
  - OBI with EMBI: -0.242*
  - GFSRI with WGIVA: -0.314*
  - GFSRI with EMBI: 0.055
  - WGIVA with EMBI: -0.138*
  - S1 correlations: with OBI 0.117*, with WGIVA 0.235*, with S2 0.923*
  - S2 correlations: with OBI 0.146*, with S1 0.923*, with EMBI 0.113*
  - Rating correlations: with GFSRI -0.284*, with EMBI -0.396*
  - FFR correlations: with EMBI -0.196*, with S1 0.245*
  - VIX correlations: with EMBI 0.220*, with FFR -0.255*
- Source: Authors calculations.

### D. List of countries used in the analyses (Section C)
- Argentina, Brazil, Bulgaria, Chile, China, Colombia, Croatia, Dominican Republic, Ecuador, Egypt, El Salvador, Ghana, Hungary, Indonesia, Iraq, Kazakhstan, Lebanon, Malaysia, Mexico, Pakistan, Peru, Philippines, Poland, Russia, Serbia, South Africa, Sri Lanka, Trinidad and Tobago, Tunisia, Turkey, Ukraine, Venezuela, Vietnam.

### E. Regression results using less stringent measure of fiscal data transparency (GFSRI2)
- Table 6. Effect of Fiscal Transparency on Sovereign Spreads (with GFSRI2) — selected coefficient estimates and statistics:
  - Column (1):
    - OBI: -0.0058** (standard error 0.0027)
    - Credit Rating (-1): -0.0424** (0.0171)
    - Macro (-1): 0.0055*** (0.0017)
    - Political Risk (-1): -2.1453*** (0.5069)
    - Financial Risk (-1): -1.0289** (0.4573)
    - Crisis (1 if year in 2009-2013): -0.1997*** (0.0376)
    - Fedfunds: -10.7430*** (1.8665)
    - VIX: 0.7264*** (0.0859)
    - Observations: 311
    - Overall R-Square: 0.5582
    - Wald Chi-Square: 318.93***
  - Column (2):
    - OBI High (OBI >= median): -0.0057** (0.0027)
    - Credit Rating (-1): -0.0431** (0.0179)
    - Macro (-1): 0.0056*** (0.0018)
    - Political Risk (-1): -2.1336*** (0.4626)
    - Financial Risk (-1): -1.0254** (0.4595)
    - Fedfunds: -10.8380*** (1.9404)
    - VIX: 0.7281*** (0.0898)
    - Observations: 311
    - Overall R-Square: 0.5623
    - Wald Chi-Square: 403.28***
  - Column (3):
    - OBI High: -0.0061*** (0.0030)
    - OBI Low: -0.0059 (0.0036)
    - Macro (-1): 0.0057*** (0.0017)
    - Financial Risk (-1): -1.2030*** (0.4552)
    - Fedfunds: -10.8150*** (1.9613)
    - VIX: 0.7082*** (0.0872)
    - Observations: 313
    - Overall R-Square: 0.4688
    - Wald Chi-Square: 419.58***
  - Column (4):
    - OBI High: -0.0065* (0.0034)
    - OBI Low: -0.0075* (0.0042)
    - Macro (-1): 0.0054*** (0.0018)
    - Financial Risk (-1): -1.3296*** (0.4883)
    - Fedfunds: -13.9260*** (2.1097)
    - VIX: 0.6274*** (0.0846)
    - Observations: 311
    - Overall R-Square: 0.4387
    - Wald Chi-Square: 432.64***
- Notes: Regressions similar to those in Table 1 but using GFSRI2; numbers in parentheses are robust standard errors; significance: * p<10%, ** p<5%, *** p<1%.

- Table 7. Effect of Fiscal Transparency on Foreign Investors' Holdings of EM Debt (with GFSRI2) — selected coefficient estimates and statistics:
  - Dependent variables: S1 & S2
  - Column (1) (S1):
    - OBI: 0.0028 (0.0036)
    - GFSRI: 0.0025* (0.0013)
    - Debt/GDP (-1): 0.4276*** (0.1539)
    - Financial Risk (-1): 0.8954** (0.3904)
    - Fedfunds: 8.7191*** (1.6907)
    - VIX: 0.1297* (0.0745)
    - Constant: -8.5870*** (2.9181)
    - Observations: 195
    - Overall R-Square: 0.388
    - Wald Chi-Square: 223.02***
  - Column (2) (S1, alternative specification):
    - GFSRI High: 0.0023** (0.0011)
    - Debt/GDP (-1): 0.4140*** (0.1342)
    - Financial Risk (-1): 0.9220** (0.3711)
    - Fedfunds: 8.0329*** (1.7267)
    - Observations: 195
    - Overall R-Square: 0.4686
    - Wald Chi-Square: 354.80***
  - Column (3) (S2):
    - OBI: 0.0028 (0.0035)
    - GFSRI: 0.0025** (0.0013)
    - Debt/GDP (-1): 0.4401*** (0.1537)
    - Financial Risk (-1): 0.9445** (0.3840)
    - Observations: 195
    - Overall R-Square: 0.3842
    - Wald Chi-Square: 66.03***
  - Column (4) (S2, alternative specification):
    - GFSRI High: 0.0024** (0.0017)
    - Debt/GDP (-1): 0.4305*** (0.1348)
    - Financial Risk (-1): 0.9809*** (0.3678)
    - Fedfunds: -1.9899 (1.8049)
    - Observations: 195
    - Overall R-Square: 0.4300
    - Wald Chi-Square: 225.96***
- Notes: Regressions similar to those in Table 2 but using GFSRI2; numbers in parentheses are robust standard errors; significance: * p<10%, ** p<5%, *** p<1%.

### F. Numerical conversion of sovereign credit ratings (selected mapping, Section E)
- Fitch — Numerical Ratings:
  - AAA = 23
  - AA+ = 22
  - AA = 21
  - AA- = 20
  - A+ = 19
  - A = 18
  - A- = 17
  - BBB+ = 16
  - BBB = 15
  - BBB- = 14
  - BB+ = 13
  - BB = 12
  - BB- = 11
  - B+ = 10
  - B = 9
  - B- = 8
  - CCC+ = 7
  - CCC = 6
  - CCC- = 5
  - CC = 4
  - C = 3
  - DDD = 2
  - SD = 1
  - DD = 1
  - D = 0
  - RD = 0
  - WD, WR, NR indicate "withdrawn" (the country was not rated).
- Moody's — Numerical Ratings:
  - Aaa = 20
  - Aa1 = 19
  - Aa2 = 18
  - Aa3 = 17
  - A1 = 16
  - A2 = 15
  - A3 = 14
  - Baa1 = 13
  - Baa2 = 12
  - Baa3 = 11
  - Ba1 = 10
  - Ba2 = 9
  - Ba3 = 8
  - B1 = 7
  - B2 = 6
  - B3 = 5
  - Caa1 = 4
  - Caa2 = 3
  - Caa3 = 2
  - Ca = 1
  - C = 0
- S&P — Numerical Ratings:
  - AAA = 22
  - AA+ = 21
  - AA = 20
  - AA- = 19
  - A+ = 18
  - A = 17
  - A- = 16
  - BBB+ = 15
  - BBB = 14
  - BBB- = 13
  - BB+ = 12
  - BB = 11
  - BB- = 10
  - B+ = 9
  - B = 8
  - B- = 7
  - CCC+ = 6
  - CCC = 5
  - CCC- = 4
  - CC = 3
  - C = 2
  - D = 0

*Source: Authors.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18189.pdf_