## _wp0927 - 1. Means and Standard Deviations

## Source details

**Canonical URL:** [_wp0927 - 1. Means and Standard Deviations](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0927.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0927.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0927.pdf.json)

---

### I. INTRODUCTION — purpose and research question
- Research question: whether changes in political institutions (checks and balances, elections, democratic accountability) have a direct effect on financial markets, measured as a reduction in the sovereign risk premium (spreads).
- Hypothesized channels:
  - Better institutions → better fiscal policy → reduced default risk → lower spreads.
  - Better institutions → greater trust and exchange → improved financial market outcomes and growth.
- Scope: emerging market economies that have issued external debt since the early 1990s; dependent variable: secondary-market sovereign spread over the risk-free world interest rate.

### II. ANALYTICAL FRAMEWORK — model and estimation
- Equilibrium/Lending and spread formulation:
  - Equations (1)–(4) derive the log-spread relation from a logistic probability of default specification (equations preserved in the source).
  - Final estimating equation with country and time effects: Log s_it = α_i + λ_t + ∑_k β_k Z_kit + η_it (equation (5)).
- Political institutions tested (Pol_it):
  - political system (presidential vs. parliamentary);
  - political constraints (Henisz, 2000);
  - Freedom House indexes of political rights and civil liberties;
  - Kaufmann voice and accountability index;
  - Polity democracy index;
  - ICRG democratic accountability index;
  - ICRG political risk index;
  - election-year dummy.
- Fiscal policy controls included explicitly:
  - total revenue–to–GNI ratio (Rev_it);
  - current government spending–to–GNI ratio (Curex_it);
  - public investment–to–GNI ratio (Ginv_it).
- Additional control vector Q_it includes:
  - debv_it = total debt outstanding to gross national income (GNI);
  - foreign exchange reserves to GDP;
  - inflation rate;
  - output gap;
  - default history dummy;
  - regional spread index (regions: Africa, Asia, Europe, Latin America).

### III. EMPIRICAL STRATEGY FOR FISCAL VARIABLES — endogeneity and instruments
- Endogeneity concerns for fiscal variables: interest payments, automatic stabilizers, timing of discretionary adjustments.
- Derivation of discretionary fiscal variables via estimating F_it (government revenue, current spending, or investment as percent of GNI) on output gap, inflation, and inflation squared (equation (7)); residual ξ_it interpreted as discretionary fiscal component.
- First-stage IV instruments (for equation (7) residuals): lagged output gap, lagged inflation, lagged inflation squared.
- Additional instruments to address remaining endogeneity for discretionary fiscal variables:
  - dummy for political system (presidential vs. parliamentary);
  - dummy for electoral system (proportional vs. majoritarian);
  - electoral indicator;
  - right-wing government indicator;
  - Henisz political constraints indicator;
  - dependency ratio;
  - urbanization rate;
  - log of real GDP per capita;
  - openness (total trade as percent of GDP).
- Second stage: estimate baseline specification (equation (6)) with country and year fixed effects using fitted values for the three fiscal variables.

### IV. DATA DESCRIPTION
- Dependent variable: stripped spread from the Emerging Markets Bond Index Global (EMBI Global), measured against comparable U.S. government bond.
  - EMBI Global inclusion criteria (as stated): face value > US$500 million; maturity > 2½ years; verifiable daily prices and cash flows; market-capitalization-weighted index; currently covers 32 countries in the index description.
- Data sources for right-hand-side variables:
  - World Bank’s World Development Indicators;
  - IMF’s World Economic Outlook database and International Financial Statistics;
  - default history: classification by Manasse and Roubini (2005) and Gelos, Sahay, and Sandleris (2004);
  - political/institutional variables: Persson and Tabellini (2001); ICRG; World Bank’s Database of Political Institutions (DPI).
- Descriptive statistics (Table 1 exact figures):
  - Log annual mean spread: Obs 169; Mean 6.272; Std. Dev. 0.815; Min 4.050; Max 8.664.
  - Total debt to GNI: Obs 169; Mean 0.569; Std. Dev. 0.254; Min 0.163; Max 1.57.
  - Reserves to GNI: Obs 169; Mean 0.127; Std. Dev. 0.076; Min 0.008; Max 0.375.
  - Log inflation: Obs 169; Mean 2.356; Std. Dev. 1.174; Min 2.38E-08; Max 7.991.
  - Output gap: Obs 169; Mean -0.283; Std. Dev. 2.667; Min -11.674; Max 6.356.
  - Country has defaulted in past: Obs 169; Mean 0.243; Std. Dev. 0.43; Min 0; Max 1.
  - Total government revenues: Obs 169; Mean 0.270; Std. Dev. 0.079; Min 0.149; Max 0.511.
  - Government investment: Obs 169; Mean 0.046; Std. Dev. 0.03; Min 0.007; Max 0.149.
  - Current government expenditures: Obs 169; Mean 0.259; Std. Dev. 0.095; Min 0.077; Max 0.491.
  - Kaufmann voice index: Obs 75; Mean 0.010; Std. Dev. 0.546; Min -1.31; Max 1.12.
  - Political constraints—Henisz: Obs 169; Mean 0.582; Std. Dev. 0.215; Min 0; Max 0.858.
  - ICRG political risk index: Obs 167; Mean 65.77; Std. Dev. 7.784; Min 45.833; Max 79.42.
  - ICRG democratic accountability index: Obs 167; Mean 3.952; Std. Dev. 1.102; Min 1; Max 6.
  - Election year dummy: Obs 169; Mean 0.260; Std. Dev. 0.440; Min 0; Max 1.
  - Presidential system dummy: Obs 169; Mean 0.852; Std. Dev. 0.356; Min 0; Max 1.
  - Polity index: Obs 150; Mean 15.54; Std. Dev. 4.549; Min 3; Max 20.
  - Freedom House political rights: Obs 159; Mean 3.006; Std. Dev. 1.503; Min 1; Max 6.
  - Freedom House civil rights: Obs 159; Mean 3.434; Std. Dev. 1.134; Min 1; Max 6.

### V. EMPIRICAL RESULTS — main findings (Tables 2–5 summarized)
- Controls and macro fundamentals:
  - Debt to GNI: positive and significant. Example: Table 3, column 1 coefficient 0.987 (standard error 0.331) — interpreted as a 1 percentage point increase in debt–to–GNI leads to about a 1 percent increase in spreads.
  - Reserves to GDP: negative and significant. Example: Table 3, column 1 coefficient -4.462 (standard error 1.114) — a 1 percentage point increase in reserves–to–GDP causes spreads to fall by about 4.5 percent.
  - Log inflation: positive and significant in many specifications.
  - Output gap: negative (spreads lower in boom periods; higher in recessions).
  - Default history: positive and significant.
- Political institutions:
  - Democracy and accountability lower spreads across multiple measures:
    - Kaufmann voice & accountability index: negative and significant (e.g., Table 3 coefficient -0.514, standard error 0.225).
    - ICRG democratic accountability index: negative and significant (e.g., Table 3 coefficient -0.123, standard error 0.040).
    - Polity index: negative sign; significance varies (Table 3 coefficient -0.037, standard error 0.020; significance noted at 10% in some specifications).
    - Freedom House indexes: higher values imply less democracy; positive coefficients on these indices imply that more democracy (lower Freedom House scores) lowers spreads (e.g., one point reduction in political rights reduces spreads by 8.5 percent; one point reduction in civil rights reduces spreads by 19 percent as reported).
  - ICRG political risk index: negative and significant (lower political risk reduces spreads).
  - Election-year dummy: markets demand higher spreads in election years; evidence significant in some specifications (fixed effects).
  - Presidential system dummy: positive coefficient in some specifications (e.g., Table 3 dummy for presidential system 0.413, standard error 0.143), indicating higher spreads for presidential systems in those specifications.
- Fiscal composition and discretionary policy:
  - Government revenue to GDP: negative and statistically significant — higher revenue reduces spreads (Table 4 and Table 5 results).
  - Current expenditure to GDP: positive and significant — higher current spending raises spreads.
  - Public investment to GDP: coefficient often not statistically significant; effect ambiguous.
  - Interaction on financing composition: example interpretation (Table 5, column 1): a 1 percentage point increase in current spending financed by debt raises spreads by about 11 percent; financed by taxes it reduces spreads by about 10 percent (note: source states this number is obtained by adding the coefficients on debt to GNI and current expenditure to GDP).
- Robustness checks and additional tests:
  - Results robust to alternative measures of solvency and liquidity.
  - Currency crisis measures not statistically significant in explaining spreads.
  - Regional contagion index tested (regions: Europe, Latin America, Africa, Asia); contagion variable not statistically significant, while political institutional results hold.
  - Nonlinear threshold tests at 40 percent and 50 percent debt–to–GDP: interaction variable not statistically significant.
  - Tests for interaction between political variables and debt variable: no supporting evidence found.

### VI. CONCLUSIONS and policy implications
- Core conclusion: political institutions matter for sovereign borrowing costs; financial markets reward democracy and accountability by charging lower interest rate spreads, beyond the effect of institutions on fiscal and macroeconomic outcomes.
- Policy implications for lowering international borrowing costs:
  - Strengthen checks and balances and democratic accountability.
  - Ensure appropriate macroeconomic and structural policies to attain lower debt–to–GDP ratios and comfortable reserves.
  - When fiscal adjustment is required, prefer revenue-driven adjustment and cuts to current spending rather than cuts to public investment, as markets pay attention to the composition of fiscal consolidation.
- Suggestions for future research: investigate specific institutions (central bank independence, public expenditure management, bankruptcy law, labor market institutions) with larger cross-country and longer time-series data.

### Appendix Table A1: List of Countries Included
- Argentina  10
- Brazil  10
- Bulgaria  10
- Chile  5
- Colombia  7
- Côte d'Ivoire  6
- Croatia  8
- Ecuador  9
- Egypt  3
- El Salvador  2
- Malaysia  8
- Mexico  10
- Morocco  7
- Pakistan  1
- Panama  8
- Peru  7
- Philippines  7
- Russia  7
- South Africa  10
- Thailand  7
- Tunisia  2
- Turkey  8
- Ukraine  4
- Uruguay  3
- Venezuela  10

*Content derived from the supplied IMF working paper text.*

### 1. Means and Standard Deviations........................................................................................

### _wp0927 - 1. Means and Standard Deviations.......................................................................................

### Means and Standard Deviations
- Section title appears on page 9.

### Effects of Institutions on Spreads—OLS Estimates (Baseline)
- Section title appears on page 10.

### Effects of Institutions on Spreads—Fixed Effects Estimates (Baseline)
- Section title appears on page 11.

### Effects of Institutions on Spreads in the Presence of Fiscal Variables—OLS Estimates
- Section title appears on page 14.

### Effects of Institutions on Spreads in the Presence of Fiscal Variables—Fixed Effects Estimates
- Section title appears on page 15.

### Appendix Table A1: List of Countries Included
- Appendix table appears on page 18.

*Document: _wp0927 - 1. Means and Standard Deviations........................................................................................*

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

### _wp0927 - References................................................................................................................................19

### I. INTRODUCTION — purpose and research question
- Research question: whether changes in political institutions (checks and balances, elections, democratic accountability) have a direct effect on financial markets, measured as a reduction in the sovereign risk premium (spreads).
- Hypothesized channels:
  - Better institutions → better fiscal policy → reduced default risk → lower spreads.
  - Better institutions → greater trust and exchange → improved financial market outcomes and growth.
- Scope: emerging market economies that have issued external debt since the early 1990s; dependent variable: secondary-market sovereign spread over the risk-free world interest rate.

### II. ANALYTICAL FRAMEWORK — model and estimation
- Equilibrium/Lending and spread formulation:
  - Equations (1)–(4) derive the log-spread relation from a logistic probability of default specification (equations preserved in the source).
  - Final estimating equation with country and time effects: Log s_it = α_i + λ_t + ∑_k β_k Z_kit + η_it (equation (5)).
- Political institutions tested (Pol_it):
  - political system (presidential vs. parliamentary);
  - political constraints (Henisz, 2000);
  - Freedom House indexes of political rights and civil liberties;
  - Kaufmann voice and accountability index;
  - Polity democracy index;
  - ICRG democratic accountability index;
  - ICRG political risk index;
  - election-year dummy.
- Fiscal policy controls included explicitly:
  - total revenue–to–GNI ratio (Rev_it);
  - current government spending–to–GNI ratio (Curex_it);
  - public investment–to–GNI ratio (Ginv_it).
- Additional control vector Q_it includes:
  - debv_it = total debt outstanding to gross national income (GNI);
  - foreign exchange reserves to GDP;
  - inflation rate;
  - output gap;
  - default history dummy;
  - regional spread index (regions: Africa, Asia, Europe, Latin America).

### III. EMPIRICAL STRATEGY FOR FISCAL VARIABLES — endogeneity and instruments
- Endogeneity concerns for fiscal variables: interest payments, automatic stabilizers, timing of discretionary adjustments.
- Derivation of discretionary fiscal variables via estimating F_it (government revenue, current spending, or investment as percent of GNI) on output gap, inflation, and inflation squared (equation (7)); residual ξ_it interpreted as discretionary fiscal component.
- First-stage IV instruments (for equation (7) residuals): lagged output gap, lagged inflation, lagged inflation squared.
- Additional instruments to address remaining endogeneity for discretionary fiscal variables:
  - dummy for political system (presidential vs. parliamentary);
  - dummy for electoral system (proportional vs. majoritarian);
  - electoral indicator;
  - right-wing government indicator;
  - Henisz political constraints indicator;
  - dependency ratio;
  - urbanization rate;
  - log of real GDP per capita;
  - openness (total trade as percent of GDP).
- Second stage: estimate baseline specification (equation (6)) with country and year fixed effects using fitted values for the three fiscal variables.

### IV. DATA DESCRIPTION
- Dependent variable: stripped spread from the Emerging Markets Bond Index Global (EMBI Global), measured against comparable U.S. government bond.
  - EMBI Global inclusion criteria (as stated): face value > US$500 million; maturity > 2½ years; verifiable daily prices and cash flows; market-capitalization-weighted index; currently covers 32 countries in the index description.
- Data sources for right-hand-side variables:
  - World Bank’s World Development Indicators;
  - IMF’s World Economic Outlook database and International Financial Statistics;
  - default history: classification by Manasse and Roubini (2005) and Gelos, Sahay, and Sandleris (2004);
  - political/institutional variables: Persson and Tabellini (2001); ICRG; World Bank’s Database of Political Institutions (DPI).
- Descriptive statistics (Table 1 exact figures retained):
  - Log annual mean spread: Obs 169; Mean 6.272; Std. Dev. 0.815; Min 4.050; Max 8.664.
  - Total debt to GNI: Obs 169; Mean 0.569; Std. Dev. 0.254; Min 0.163; Max 1.57.
  - Reserves to GNI: Obs 169; Mean 0.127; Std. Dev. 0.076; Min 0.008; Max 0.375.
  - Log inflation: Obs 169; Mean 2.356; Std. Dev. 1.174; Min 2.38E-08; Max 7.991.
  - Output gap: Obs 169; Mean -0.283; Std. Dev. 2.667; Min -11.674; Max 6.356.
  - Country has defaulted in past: Obs 169; Mean 0.243; Std. Dev. 0.43; Min 0; Max 1.
  - Total government revenues: Obs 169; Mean 0.270; Std. Dev. 0.079; Min 0.149; Max 0.511.
  - Government investment: Obs 169; Mean 0.046; Std. Dev. 0.03; Min 0.007; Max 0.149.
  - Current government expenditures: Obs 169; Mean 0.259; Std. Dev. 0.095; Min 0.077; Max 0.491.
  - Kaufmann voice index: Obs 75; Mean 0.010; Std. Dev. 0.546; Min -1.31; Max 1.12.
  - Political constraints—Henisz: Obs 169; Mean 0.582; Std. Dev. 0.215; Min 0; Max 0.858.
  - ICRG political risk index: Obs 167; Mean 65.77; Std. Dev. 7.784; Min 45.833; Max 79.42.
  - ICRG democratic accountability index: Obs 167; Mean 3.952; Std. Dev. 1.102; Min 1; Max 6.
  - Election year dummy: Obs 169; Mean 0.260; Std. Dev. 0.440; Min 0; Max 1.
  - Presidential system dummy: Obs 169; Mean 0.852; Std. Dev. 0.356; Min 0; Max 1.
  - Polity index: Obs 150; Mean 15.54; Std. Dev. 4.549; Min 3; Max 20.
  - Freedom House political rights: Obs 159; Mean 3.006; Std. Dev. 1.503; Min 1; Max 6.
  - Freedom House civil rights: Obs 159; Mean 3.434; Std. Dev. 1.134; Min 1; Max 6.

### V. EMPIRICAL RESULTS — main findings (Tables 2–5 summarized)
- Controls and macro fundamentals:
  - Debt to GNI: positive and significant. Example: Table 3, column 1 coefficient 0.987 (standard error 0.331) — interpreted as a 1 percentage point increase in debt–to–GNI leads to about a 1 percent increase in spreads.
  - Reserves to GDP: negative and significant. Example: Table 3, column 1 coefficient -4.462 (standard error 1.114) — a 1 percentage point increase in reserves–to–GDP causes spreads to fall by about 4.5 percent.
  - Log inflation: positive and significant in many specifications.
  - Output gap: negative (spreads lower in boom periods; higher in recessions).
  - Default history: positive and significant.
- Political institutions:
  - Democracy and accountability lower spreads across multiple measures:
    - Kaufmann voice & accountability index: negative and significant (e.g., Table 3 coefficient -0.514, standard error 0.225).
    - ICRG democratic accountability index: negative and significant (e.g., Table 3 coefficient -0.123, standard error 0.040).
    - Polity index: negative sign; significance varies (Table 3 coefficient -0.037, standard error 0.020; significance noted at 10% in some specifications).
    - Freedom House indexes: higher values imply less democracy; positive coefficients on these indices imply that more democracy (lower Freedom House scores) lowers spreads (e.g., one point reduction in political rights reduces spreads by 8.5 percent; one point reduction in civil rights reduces spreads by 19 percent as reported).
  - ICRG political risk index: negative and significant (lower political risk reduces spreads).
  - Election-year dummy: markets demand higher spreads in election years; evidence significant in some specifications (fixed effects).
  - Presidential system dummy: positive coefficient in some specifications (e.g., Table 3 dummy for presidential system 0.413, standard error 0.143), indicating higher spreads for presidential systems in those specifications.
- Fiscal composition and discretionary policy:
  - Government revenue to GDP: negative and statistically significant — higher revenue reduces spreads (Table 4 and Table 5 results).
  - Current expenditure to GDP: positive and significant — higher current spending raises spreads.
  - Public investment to GDP: coefficient often not statistically significant; effect ambiguous.
  - Interaction on financing composition: example interpretation (Table 5, column 1): a 1 percentage point increase in current spending financed by debt raises spreads by about 11 percent; financed by taxes it reduces spreads by about 10 percent (note: source states this number is obtained by adding the coefficients on debt to GNI and current expenditure to GDP).
- Robustness checks and additional tests:
  - Results robust to alternative measures of solvency and liquidity.
  - Currency crisis measures not statistically significant in explaining spreads.
  - Regional contagion index tested (regions: Europe, Latin America, Africa, Asia); contagion variable not statistically significant, while political institutional results hold.
  - Nonlinear threshold tests at 40 percent and 50 percent debt–to–GDP: interaction variable not statistically significant.
  - Tests for interaction between political variables and debt variable: no supporting evidence found.

### VI. CONCLUSIONS and policy implications
- Core conclusion: political institutions matter for sovereign borrowing costs; financial markets reward democracy and accountability by charging lower interest rate spreads, beyond the effect of institutions on fiscal and macroeconomic outcomes.
- Policy implications for lowering international borrowing costs:
  - Strengthen checks and balances and democratic accountability.
  - Ensure appropriate macroeconomic and structural policies to attain lower debt–to–GDP ratios and comfortable reserves.
  - When fiscal adjustment is required, prefer revenue-driven adjustment and cuts to current spending rather than cuts to public investment, as markets pay attention to the composition of fiscal consolidation.
- Suggestions for future research: investigate specific institutions (central bank independence, public expenditure management, bankruptcy law, labor market institutions) with larger cross-country and longer time-series data.

*Italicized source attribution: Content derived from the supplied IMF working paper text.*

### Appendix Table A1: List of Countries Included

### Appendix Table A1: List of Countries Included

### Countries and number of observations
- Argentina  10
- Brazil  10
- Bulgaria  10
- Chile  5
- Colombia  7
- Côte d'Ivoire  6
- Croatia  8
- Ecuador  9
- Egypt  3
- El Salvador  2
- Malaysia  8
- Mexico  10
- Morocco  7
- Pakistan  1
- Panama  8
- Peru  7
- Philippines  7
- Russia  7
- South Africa  10
- Thailand  7
- Tunisia  2
- Turkey  8
- Ukraine  4
- Uruguay  3
- Venezuela                                 10

*Source: Appendix Table A1 from the provided PDF content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0927.pdf_