## _wp1559 - 1. Descriptive Statistics

## Source details

**Canonical URL:** [_wp1559 - 1. Descriptive Statistics](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1559.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1559.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1559.pdf.json)

---

### Introduction
- Objective: analyze impacts of fiscal decentralization on the efficiency of public service delivery (health and education), focusing on efficiency (outcomes given similar inputs) rather than inputs or outcomes alone.
- Sample: large panel including developed, emerging, and developing economies.
- Key summary finding: fiscal decentralization can improve performance only under specific conditions (effective local autonomy, strong accountability, good governance, strong local capacity, sufficient expenditure decentralization, and accompanying revenue decentralization); absent these, decentralization can worsen efficiency.

### Literature and theoretical background
- Potential efficiency-enhancing mechanisms:
  - Preference matching and allocative efficiency (Hayek, 1945; Tiebout, 1956; Musgrave, 1969; Oates, 1972).
  - Stronger local accountability and “productive efficiency” (Ahmad, Brosio, and Tanzi, 2008; Cantarero and Pacual Sanchez, 2006).
  - Competition across local governments and “voting with one’s feet” (Seabright, 1996; Persson and Tabellini, 2000; Hindriks and Lockwood, 2005).
  - Reduced lobbying and rent-seeking distortions.
- Adverse channels and risks:
  - Losses from scale economies when services devolved to small jurisdictions.
  - Disruption of redistribution/equalization when large shares are decentralized (Ter-Minassian, 1997).
  - Weak local accountability leading to rent-seeking and misallocation (Davoodi and Zou, 1998; Woller and Phillips, 1998; Zhang and Zou, 1998; Rodriguez-Pose and others, 2009; Gonzalez Alegre, 2010; Grisorio and Prota, 2011).

### Empirical methodology (two-step strategy)
- Step 1: Estimate country-specific, time-varying efficiency coefficients for public service delivery (infant mortality rate and secondary school enrollment rate) using parametric stochastic frontier analysis (SFA) à la Battese and Coelli (1988); Jondrow and others (1988).
  - Dependent variables: infant mortality rate and secondary school enrollment rate.
  - Main input: public expenditure on health and education as a percent of GDP (lagged: 1itPE−).
  - Controls: structural variables including real GDP per capita, density, population size, average years of primary and secondary schooling (average years excluded when estimating education effects to avoid collinearity).
  - Error decomposition: εit = ωit ± ηit, where ωit is idiosyncratic disturbance and ηit is one-sided disturbance capturing inefficiency; time-varying country-specific efficiencies obtained following Battese and Coelli (1988) and Jondrow and others (1988).
- Step 2: Regress estimated efficiencies (η̂ it) on measures of fiscal decentralization and interaction terms to assess impacts and heterogeneities.
  - Baseline specification: η̂ it = α + δ fd1it− + φ GDP1it− + ... (lagged fiscal decentralization fd1it−).
  - Non-linearity: include (fd1it−)2 and compute derivatives to detect non-linear effects.
  - Political/institutional interactions: include additive term I1it− and interaction fd1it− × I1it−; parameters δ (decentralization effect), ρ (direct institutional effect), and τ (interaction effect) are estimated.
- Endogeneity and identification:
  - Explanatory variables (including fiscal decentralization) introduced with one-period lag.
  - Two-stage least squares (2SLS) for fiscal decentralization using three instruments:
    - Population size.
    - Existence/importance of natural resources.
    - Government fractionalization and legislative fractionalization (probability two deputies randomly picked are from different parties).
  - Instruments rationale: affect decentralization process; signs for resource and fractionalization instruments not signed a priori.

### Data
- Sample coverage: unbalanced panel of 64 countries, 1990–2012.
- Data sources: IMF’s Government Financial Statistics, World Bank’s World Development Indicators, Eurostat, OECD databases, among others.
- Fiscal decentralization measures:
  - Main index: share of subnational expenditure to general government expenditure (expenditure-side focus).
  - Revenue decentralization also considered: share of local government revenue to general government revenue.
  - Subnational shares obtained as residual (total general government minus central government share) due to data limitations.
  - Vertical fiscal imbalance indicator (share of local expenditure financed with own revenue) not available for full sample.
- Political and institutional variables included:
  - Corruption index (higher = more corrupt).
  - Autonomy dummy (value 1 when constitutionally autonomous regions exist).
  - Democracy score (higher = more democratic).
  - Political system index (binary: 1 parliamentary, 0 presidential).
  - Fractionalization: legislative and government fractionalization.

### Descriptive statistics (selected)
- Expenditure decentralization:
  - Average share implemented by subnational governments: about 30 percent.
  - Advanced economies: about 40 percent.
  - Emerging and developing countries: about 25 percent.
- Revenue decentralization:
  - Average share of subnational governments: about 27 percent.
  - Advanced economies: 37 percent.
  - Emerging and developing countries: 23 percent.
- Fractionalization:
  - Probability two deputies from legislature are from different parties (legislature): 65 percent.
  - Probability two members of governments are from different parties: 29 percent.
- Political system and corruption:
  - Corruption index: higher values indicate more corruption; corruption seems more pervasive in emerging and developing countries.
  - Political system index: advanced economies appear more parliamentary based than emerging and developing economies.
- Efficiency estimates (sample average): about 85 percent of the production frontier (country-specific, time-varying from SFA).

### Efficiency estimates and summary statistics
- Mean estimated efficiencies:
  - Health (Battese and Coelli (1988)): 0.82
  - Health (Jondrow et al. (1982)): 0.81
  - Health (Heterog.): 0.84
  - Education (Battese and Coelli (1988)): 0.88
  - Education (Jondrow et al. (1982)): 0.88
  - Education (Heterog.): 0.88
- Dispersion of estimated efficiencies:
  - Health standard deviations: 0.09, 0.10, 0.11 (columns 1–3).
  - Education standard deviations: 0.10, 0.10, 0.13 (columns 4–6).
  - Minimum efficiencies: Health 0.30, 0.29, 0.31; Education 0.33, 0.33, 0.27.
  - Maximum efficiencies: Health 0.94, 0.94, 0.98; Education 0.98, 0.98, 0.99.
- Interpretation: "An efficiency score of x percent implies that the country delivers x percent of the possible objective (reducing infant mortality rate or increasing school enrollment rate) as compared to a fully efficient country using similar input values (such as public expenditure)."

### Direct channel effects of fiscal decentralization (pooled and subsample results)
- Instruments and identification:
  - First-stage instruments significantly correlated with endogenous regressor in almost all cases (associated p-values < 0.05).
  - Kleibergen-Paap’s p values: the null that "the equations are underidentified" can be rejected at the 5 percent level.
- Pooled-sample result:
  - Pooling advanced economies, emerging markets, and developing economies: fiscal decentralization has no significant effect on the efficiency of public expenditure (columns 1 and 6 in reported tables).
- Subsample results:
  - Advanced economies:
    - Fiscal decentralization shows positive impacts on efficiency of public expenditure on health (column 2).
    - Quantified effect: "a 5 percent increase in fiscal decentralization would lead to 2.9 percentage points of efficiency gains in public service delivery."
    - Coefficient for education is statistically insignificant (column 7).
  - Emerging markets and developing economies:
    - Impacts of fiscal decentralization are negative for efficiency (columns 3 and 8).
- Robustness to time dummies:
  - Positive and negative effects for the two country groups robust to inclusion of time dummies, with slight reduction in magnitude (columns 4,5,9,10).

### Non-linear (U-shaped) relationship and indicative thresholds
- Non-linearity testing: fiscal decentralization and its squared term significantly affect efficiency for entire sample (columns 1 and 4 of Table 4).
  - Sign pattern: coefficient on fiscal decentralization negative; coefficient on fiscal decentralization squared positive → U-shaped relationship.
- Indicative thresholds derived from estimated parameters:
  - Health: 35.7 percent (computed as 100×fd* = 35.7 percent).
  - Education: 35.4 percent.
- Marginal effects around threshold (health):
  - When fiscal decentralization ratio < 35.7 percent: a 1 percent increase in fiscal decentralization ratio reduces efficiency by about 0.8 percentage point (column 2 of Table 4).
  - When fiscal decentralization ratio ≥ 35.7 percent: a 1 percent increase in fiscal decentralization ratio increases efficiency by about 0.2 percentage point (column 3 of Table 4).
- Country-group averages vs thresholds:
  - Average expenditure decentralization in advanced economies: about 40 percent (above the indicative threshold).
  - Average expenditure decentralization in emerging and developing countries: about 25 percent (below the indicative threshold).
- Interpretation: U-shape may reflect scale economies in production and delivery of public services; small-scale local provision may face large fixed costs that reduce service provision unless decentralization reaches sufficient scale.

### Political and institutional interactions (key coefficients and effects)
- Corruption:
  - Corruption negatively affects the impact of fiscal decentralization on efficiency.
  - Example marginal effect: a 5 percent increase in fiscal decentralization ratio associated on average with a 2.5 percent decrease in efficiency relative to mean when accounting for corruption (noted formula: (-0.488×0.05) ×100 = −2.4).
  - Interpretation: corruption may reflect stronger local interest groups, more discretion and fewer controls at local level, and room for leakage.
- Political system interactions:
  - FD × Parliamentary (t-1) interaction positive and statistically significant in some specifications → parliamentary system combined with fiscal decentralization may boost expenditure efficiency.
  - Parliamentary regimes argued to have stronger institutional frameworks to limit executive discretion.
- Autonomy of regions:
  - Existence of constitutionally autonomous regions has positive and statistically significant impacts on relationship between decentralization and expenditure efficiency.
  - Sufficient autonomy allows preference matching and allocation efficiency to operate.
- Subsample patterns:
  - Decentralization broadly improves efficiency in advanced economies and worsens or does not improve efficiency in emerging markets and developing countries.
  - For both groups and for health and education, corruption has negative impacts and regional autonomy has positive effects.

### Robustness checks and alternate specifications
- Outlier exclusion:
  - Results robust after excluding countries with decentralization ratios close to zero and those exceeding 90 percent (sample restricted to 0%<fd<90%).
  - Health findings: positive impact in advanced economies and negative in emerging/developing economies persist.
  - Education findings: similar thrust with slight coefficient differences.
- Alternative efficiency estimation:
  - Jondrow et al. (1982) approach and heterogeneity/heteroskedasticity adjustments used.
  - Under alternative estimates, corruption hinders impacts of fiscal decentralization with high statistical significance; parliamentary regimes and more democratic institutions show favorable contributions (sometimes weaker significance); autonomy of regions positive with high significance.
- Averaging over four-year periods:
  - Variables averaged over four-year period with fiscal decentralization lagged one period; results support baseline findings:
    - Decentralization improves efficiency in advanced economies.
    - Negative or non-significant impacts for emerging and developing economies.
    - Negative impact of corruption confirmed.
    - Favorable role of parliamentary regimes ascertained; autonomy positive but not always significant.
- Alternative institutional variables:
  - Bureaucracy, political stability, and checks and balances used; inferences broadly similar though statistical significance low in many cases.

### Revenue decentralization findings
- Revenue decentralization shows positive and statistically significant impacts on public service delivery across advanced economies and emerging/developing countries (Table 11).
- Robustness:
  - Confirmed under alternative efficiency estimates (Jondrow approach and heterogeneity-adjusted estimates; Table 12).
  - Remains robust after excluding outliers and restricting sample to revenue decentralization between 0 and 90 percent (Table 13).
  - For both health and education, and for both country groups, revenue decentralization positively affects efficiency.
- Institutional interactions:
  - Corruption decreases the positive impact of revenue decentralization on efficiency (reported FDR × Corruption (t-1): -0.170*** and -0.458*** in specified columns).
  - Checks and balances enhance contribution of revenue decentralization (FDR × Checks (t-1): 0.264** in a reported column).
- Interpretation: expenditure decentralization should be accompanied by sufficient revenue decentralization to ensure improved performance.

### Selected quantitative table highlights (A3.1 and A3.2 excerpts)
- Table A3.1 (selected FD (t-1) coefficients for estimated efficiencies):
  - FD (t-1) coefficients (columns (1) to (10)): -0.775***, -0.279***, -0.530**, -0.248, 0.367***, -1.474***, 0.931, -1.649***, -0.162, 0.547**
  - (FD (t-1))^2 coefficients shown in some columns: 1.069***, 0.441***, 0.868**, (blank), 2.719***, -1.069, 3.106***.
  - Real GDP pc (t-1) coefficients (selected): -0.0027, 0.0150***, 0.000, 0.006, 0.002, 0.045***, -0.006, 0.052***, 0.067***, 0.013.
  - Number of observations by column: 926, 303, 623, 528, 394, 569, 188, 381, 321, 246.
  - Countries by column: 58, 16, 42, 40, 30, 50, 14, 36, 32, 24.
- Table A3.2 (selected interaction coefficients):
  - FD (t-1) coefficients across columns: -0.065, -0.087***, -0.245***, -0.146***, -0.419, 0.003, -0.111, -0.373*.
  - FD × Corruption (t-1): -0.026 (t-statistic (-0.604)); -0.426*** (t-statistic (-2.830)) in another column.
  - FD × Autonomy (t-1): 0.530*** (t-statistic (5.022)); 1.561*** (t-statistic (3.259)).
  - Real GDP pc (t-1) coefficients across columns: 0.008, 0.0118*, -0.027***, 0.006, -0.038, 0.148***, 0.057**, 0.038**.
  - Number of observations by column: 861, 926, 925, 926, 529, 569, 568, 569.
  - Countries by column: 54, 58, 58, 58, 46, 50, 50, 50.
- Note on significance: (*), (**) and (***) denote statistical significance level of 10 percent, 5 percent and 1 percent respectively. Robust t-statistics reported in parentheses in source tables.

### Conclusions and policy implications
- Main conclusions:
  - Fiscal decentralization can improve public service delivery efficiency, but benefits depend on conditions.
  - Expenditure decentralization appears to have improved service delivery in advanced economies; impacts in emerging and developing economies are mixed or negative.
  - Expenditure decentralization needs to exceed an indicative threshold of about 35 percent to improve service delivery (35.7 percent for health; 35.4 percent for education).
  - Revenue decentralization has positive impacts across country groups; aligning responsibilities with adequate revenue decentralization is important.
- Policy implications and necessary conditions emphasized in source:
  - Accompany decentralization of responsibilities with sufficient decentralization of resources (revenue).
  - Ensure favorable institutional and political environment:
    - Effective autonomy of local governments to enable preference matching and allocative efficiency.
    - Strong accountability of local authorities to enable productive efficiency.
    - Anti-corruption measures to prevent misuse of decentralized resources.
    - Strengthen capacity at the local level.
  - Absent these conditions, fiscal decentralization can worsen public service delivery.
- Suggested extensions for future analysis:
  - Use alternative policy outcome indicators (e.g., life expectancy, school drop-out rates, PISA scores).
  - Investigate impact of decentralization on macroeconomic performance, such as fiscal outcomes and GDP growth.

### Annex I (countries and coverage) — selected rows preserved verbatim
- Argentina 1993–2004 GFS, WEO
- Australia 1990–2011 OECD database
- Austria 1990–2012 Eurostat
- BahrainT 1990–2004 GFS, WEO
- Belarus 2001–2010 GFS, WEO
- Belgium 1990–2012 Eurostat
- Bhutan 1990–2009 GFS, WEO
- Bolivia 1990–2007 GFS, WEO
- Brazil 1997–2012 GFS, WEO
- Bulgaria 1995–2012 Eurostat
- Canada 1990–2010 OECD database
- Chile 1990–2012 GFS, WEO
- Croatia 2002–2012 Eurostat
- Cyprus 1995–2012 Eurostat
- Czech Republic 1995–2012 Eurostat
- Denmark 1990–2012 Eurostat
- Egypt 2002–2012 GFS, WEO
- Estonia 1995–2012 Eurostat
- Finland 1990–2012 Eurostat
- France 1990–2012 Eurostat
- Georgia 1997–2012 GFS, WEO
- Germany 1990–2012 Eurostat
- Greece 1995–2012 Eurostat
- Hungary 1995–2012 Eurostat
- Iceland 1995–2012 Eurostat
- India 1990–2012 GFS, WEO
- Indonesia 1990–2004 GFS, WEO
- Iran 1990–2009 GFS, WEO
- Ireland 1990–2012 Eurostat
- Israel 1995–2012 OECD database
- Italy 1990–2012 Eurostat
- Japan 1990–2012 OECD database
- Korea 2000–2012 OECD database
- Latvia 1995–2012 Eurostat
- Lesotho 1990–2008 GFS, WEO
- Lithuania 1995–2012 Eurostat
- Luxembourg 1990–2012 Eurostat
- Maldives 1990–2011 GFS, WEO
- Malta 1995–2012 Eurostat
- Mauritius 2000–2011 GFS, WEO
- Mexico 1990–2012 GFS, WEO
- Mongolia 1992–2012 GFS, WEO
- Netherlands 1990–2012 Eurostat
- New Zealand 1990–2012 OECD database
- Norway 1990–2012 Eurostat
- Pakistan 1990–2007 GFS, WEO
- Peru 1995–2012 GFS, WEO
- Poland 1995–2012 Eurostat
- Portugal 1990–2012 Eurostat
- Romania 1995–2012 Eurostat
- Seychelles 1993–2012 GFS, WEO
- Singapore 1990–2012 GFS, WEO
- Slovak Republic 1995–2012 Eurostat
- Slovenia 1995–2012 Eurostat
- South Africa 1990–2012 GFS, WEO
- Spain 1995–2012 Eurostat
- Sweden 1993–2012 Eurostat
- Switzerland 1990–2012 Eurostat
- Tunisia 1990–2012 GFS, WEO
- Turkey 1990–2012 OECD database
- United Kingdom 1990–2012 Eurostat
- United States 1990–2012 OECD database
- Uruguay 1999–2012 GFS, WEO
- Venezuela 1990–2005 GFS, WEO
- Note: Expenditure and Revenue decentralization for European and OECD countries taken from Eurostat and OECD databases; for emerging economies and developing countries, data are from GFS and WEO.

*Source: _wp1559 - 1. Descriptive Statistics (PDF chapter/section) — content as provided.*

### 1. Descriptive Statistics ..............................................................................................

### _wp1559 - 1. Descriptive Statistics ..............................................................................................

### Introduction
- Objective: analyze impacts of fiscal decentralization on the efficiency of public service delivery (health and education), focusing on efficiency (outcomes given similar inputs) rather than inputs or outcomes alone.
- Sample: large panel including developed, emerging, and developing economies.
- Key finding summarized in the source: fiscal decentralization can improve performance only under specific conditions (effective local autonomy, strong accountability, good governance, strong local capacity, sufficient expenditure decentralization, and accompanying revenue decentralization); absent these, decentralization can worsen efficiency.

### Literature Review and Theoretical Background
- Mechanisms by which fiscal decentralization can improve efficiency:
  - Preference matching and allocative efficiency (Hayek, 1945; Tiebout, 1956; Musgrave, 1969; Oates, 1972).
  - Stronger local accountability and “productive efficiency” (Ahmad, Brosio, and Tanzi, 2008; Cantarero and Pacual Sanchez, 2006).
  - Competition across local governments and “voting with one’s feet” (Seabright, 1996; Persson and Tabellini, 2000; Hindriks and Lockwood, 2005).
  - Reduced lobbying and rent-seeking distortions.
- Risks and adverse channels:
  - Losses from scale economies when services are devolved to small jurisdictions.
  - Disruption of redistribution/equalization by the central government when large shares are decentralized (Ter-Minassian, 1997).
  - Weak local accountability leading to rent-seeking and misallocation (Davoodi and Zou, 1998; Woller and Phillips, 1998; Zhang and Zou, 1998; Rodriguez-Pose and others, 2009; Gonzalez Alegre, 2010; Grisorio and Prota, 2011).

### Empirical Analysis — Methodology
- Two-step empirical strategy:
  1. Estimate country-specific, time-varying efficiency coefficients for public service delivery (infant mortality rate and secondary school enrollment rate) using parametric stochastic frontier analysis (SFA) à la Battese and Coelli (1988); Jondrow and others (1988).
     - Dependent variables: infant mortality rate and secondary school enrollment rate.
     - Main input: public expenditure on health and education as a percent of GDP (lagged: 1itPE−).
     - Controls: a set of structural variables (real GDP per capita, density, population size, average years of primary and secondary schooling), with average years of schooling excluded when estimating education effects to avoid collinearity.
     - Error decomposition: εit = ωit ± ηit, where ωit is idiosyncratic disturbance and ηit is one-sided disturbance capturing inefficiency; time-varying country-specific efficiencies are obtained following Battese and Coelli (1988) and Jondrow and others (1988).
  2. Regress estimated efficiencies (η̂ it) on measures of fiscal decentralization and interaction terms to assess impacts and heterogeneities.
     - Baseline specification: η̂ it = α + δ fd1it− + φ GDP1it− + ... (lagged fiscal decentralization fd1it−).
     - Non-linearity: include (fd1it−)2 and compute derivatives to detect non-linear effects.
     - Political/institutional interactions: include additive term I1it− and interaction fd1it− × I1it−; parameters δ (decentralization effect), ρ (direct institutional effect), and τ (interaction effect) are estimated.
- Endogeneity and identification:
  - Explanatory variables (including fiscal decentralization) are introduced with one-period lag.
  - Two-stage least squares (2SLS) for fiscal decentralization using three instruments:
    - Population size.
    - Existence/importance of natural resources.
    - Government fractionalization and legislative fractionalization (probability two deputies randomly picked are from different parties).
  - Rationale: instruments affect decentralization process; signs for resource and fractionalization instruments are not signed a priori.

### Empirical Analysis — Data
- Sample coverage: unbalanced panel of 64 countries, 1990–2012.
- Data sources: IMF’s Government Financial Statistics, World Bank’s World Development Indicators, Eurostat, OCED databases, among others.
- Fiscal decentralization measurement:
  - Main index: share of subnational expenditure to general government expenditure (expenditure-side focus because of direct link to health and education outcomes).
  - Revenue decentralization also considered: share of local government revenue to general government revenue.
  - Note: subnational shares obtained as residual (total general government minus central government share) due to data limitations.
  - Vertical fiscal imbalance indicator (share of local expenditure financed with own revenue) not available for full sample.
- Political and institutional variables:
  - Corruption index (higher = more corrupt).
  - Autonomy dummy (value 1 when constitutionally autonomous regions exist).
  - Democracy score (higher = more democratic).
  - Political system index (binary: 1 parliamentary, 0 presidential).
  - Fractionalization: probability two deputies are from different parties in legislature (legislative fractionalization) and in government (government fractionalization).

### Descriptive Statistics (selected)
- Average share of public expenditure implemented by subnational governments: about 30 percent.
  - Advanced economies: about 40 percent.
  - Emerging economies and developing countries: about 25 percent.
- Revenue side: share of subnational governments about 27 percent.
  - Advanced economies: 37 percent.
  - Emerging economies and developing countries: 23 percent.
- Fractionalization:
  - Probability two deputies come from two different parties (legislature): 65 percent.
  - Probability two members of governments are from different parties: 29 percent.
- Political system and corruption:
  - Higher corruption index indicates more corruption; corruption seems more pervasive in emerging economies and developing countries.
  - Political system index: advanced economies appear more parliamentary based than emerging and developing economies.
- Efficiency estimates:
  - Average efficiency of countries in the sample is at about 85 percent of the production frontier.
  - Predicted efficiencies come from stochastic frontier analysis (country-specific, time-varying).

### Key Empirical Findings and Implications (as stated in the source)
- General conclusion: fiscal decentralization can serve as a policy tool to improve public service delivery efficiency, but only under specific conditions.
- Necessary conditions identified in the source:
  - Effective autonomy of local governments.
  - Strong accountability at various institutional levels.
  - Good governance.
  - Strong capacity at the local level.
  - A sufficient degree of expenditure decentralization.
  - Decentralization of expenditure needs to be accompanied by sufficient decentralization of revenue.
- Risk statement: Absent those conditions, fiscal decentralization can worsen the efficiency of public service delivery.

*Source: _wp1559 - 1. Descriptive Statistics (PDF chapter/section) — content as provided.*

### 82.2 percent on average for health and 87.8 percent for education (Table 2). An efficiency score

### _wp1559 - 82.2 percent on average for health and 87.8 percent for education (Table 2). An efficiency score

### Efficiency estimates and methodology
- Mean estimated efficiencies:
  - Health (Battese and Coelli (1988)): 0.82
  - Health (Jondrow et al. (1982)): 0.81
  - Health (Heterog.): 0.84
  - Education (Battese and Coelli (1988)): 0.88
  - Education (Jondrow et al. (1982)): 0.88
  - Education (Heterog.): 0.88
- Dispersion of estimated efficiencies:
  - Health standard deviations: 0.09 (column 1), 0.10 (column 2), 0.11 (column 3)
  - Education standard deviations: 0.10 (column 4), 0.10 (column 5), 0.13 (column 6)
  - Minimum efficiencies: Health 0.30, 0.29, 0.31; Education 0.33, 0.33, 0.27
  - Maximum efficiencies: Health 0.94, 0.94, 0.98; Education 0.98, 0.98, 0.99
- Interpretation: "An efficiency score of x percent implies that the country delivers x percent of the possible objective (reducing infant mortality rate or increasing school enrollment rate) as compared to a fully efficient country using similar input values (such as public expenditure)."
- Estimation approaches:
  - Benchmark estimates: Battese and Coelli (1988) (columns 1 and 4).
  - Alternative methodology: Jondrow et al. (1982) (columns 2 and 5).
  - Heterogeneity and heteroskedasticity adjustments (columns 3 and 6).
  - The estimates from the various approaches are reported as highly correlated.

### Descriptive statistics (Table 1)
- Variables, sample sizes, means, and dispersion measures (selected):
  - FD expenditure (%): Number of observations 108; Mean 29.6; Std. Dev. 39.0; Min 25.4; Max 21.3; further range 0.09–98.4 (as printed in table layout)
  - FD revenue (%): Number of observations 112; Mean 27.4; Std. Dev. 36.8; Min 23.5; Max 20.0; further range 0.07–73.6
  - Real GDP pc (in thousands): Number of observations 146; Mean 22.7; Std. Dev. 34.7; Min 17.6; Max 15.7; further range 1.39–97.4
  - Natural ress. (% GDP): Number of observations 146; Mean 4.5; Std. Dev. 1.9; Min 5.7; Max 8.1; further range 0.06–64.0
  - Government frac.: Number of observations 138; Mean 0.3; Std. Dev. 0.3; Min 0.3; Max 0.3; further range 0.01–1.0
  - Fractionalization: Number of observations 136; Mean 0.7; Std. Dev. 0.7; Min 0.6; Max 0.2; further range 0.01–1.0
  - Population size (in millions): Number of observations 147; Mean 48.6; Std. Dev. 43.3; Min 50.9; Max 138.7; further range 0.11–236.7
  - Corruption: Number of observations 128; Mean -2.7; Std. Dev. -3.5; Min -2.3; Max 1.3; further range -5.00–0.7
  - Parliamentary: Number of observations 143; Mean 0.6; Std. Dev. 0.9; Min 0.4; Max 0.5; further range 0.01–1.0
  - Democracy: Number of observations 142; Mean 30.1; Std. Dev. 51.0; Min 20.9; Max 26.4; further range 1.08–82.0
  - Autonomy: Number of observations 142; Mean 0.2; Std. Dev. 0.3; Min 0.2; Max 0.4; further range 0.01–1.0
- Source attribution for table: "Source: Authors' calculations."

### Direct channel effects of fiscal decentralization
- Instruments and identification:
  - Two-stage least squares first stage: instrument variables are significantly correlated with the endogenous regressor in almost all cases (associated p-values are < 0.05).
  - Kleibergen-Paap’s p values: the null hypothesis that "the equations are underidentified" can be rejected at the 5 percent level.
- Pooled-sample results:
  - Pooling advanced economies, emerging markets, and developing economies: fiscal decentralization has no significant effect on the efficiency of public expenditure (columns 1 and 6).
- Subsample results:
  - Advanced economies:
    - Fiscal decentralization shows positive impacts on the efficiency of public expenditure on health (column 2).
    - Quantified effect: "a 5 percent increase in fiscal decentralization would lead to 2.9 percentage points of efficiency gains in public service delivery."
    - Coefficient for education is statistically insignificant (column 7).
  - Emerging markets and developing economies:
    - Impacts of fiscal decentralization are negative for efficiency (columns 3 and 8).
  - Robustness:
    - Positive and negative effects for the two country groups are robust to the inclusion of time dummies, with a slight reduction in magnitude (columns 4,5,9, and 10).
    - This robustness suggests results are not driven by common shocks or time-trend evolution of efficiency scores.

### Non-linear (U-shaped) relationship between fiscal decentralization and efficiency
- Non-linearity testing:
  - Investigated through equation (4); results presented in Table 4.
  - For the entire sample, both the fiscal decentralization variable and its squared term significantly affect the efficiency of public services (columns 1 and 4).
  - Sign pattern:
    - Coefficient on fiscal decentralization: negative.
    - Coefficient on fiscal decentralization squared: positive.
  - Interpretation:
    - Relationship is U-shaped: a low level of fiscal decentralization appears harmful; it needs to exceed a threshold (text ends before the exact threshold value is printed in the supplied excerpt) for positive impacts to emerge.

*Source: Authors' calculations, excerpt from the supplied PDF content.*

### 35.7 percent for health and 35.4 percent for education to bring about improvements in the

### _wp1559 - 35.7 percent for health and 35.4 percent for education to bring about improvements in the efficiency of public services

### Non-linearity and Indicative Thresholds
- The relationship between fiscal decentralization and public expenditure efficiency is U-shaped.
- Indicative thresholds (derived from estimated parameters):
  - Health: 35.7 percent (computed as 100×fd* = 35.7 percent).
  - Education: 35.4 percent (derived similarly).
- Effects below and above the health threshold (based on Table 4 results):
  - When fiscal decentralization ratio is below 35.7 percent: a 1 percent increase in fiscal decentralization ratio reduces efficiency by about 0.8 percentage point (column 2 of Table 4).
  - When fiscal decentralization ratio ≥ 35.7 percent: a 1 percent increase in fiscal decentralization ratio increases efficiency by about 0.2 percentage point (column 3 of Table 4).
- For education, the coefficients of fiscal decentralization are not statistically significant when the sample is split by the indicative threshold.
- Country-group averages and threshold comparison:
  - Average expenditure decentralization in advanced economies: about 40 percent (above the indicative threshold of about 35 percent).
  - Average expenditure decentralization in emerging markets and developing countries: about 25 percent (below the indicative threshold of 35 percent).
- Interpretation: The non-linear relationship may reflect scale economies in production and delivery of public services; small-scale local provision may face large fixed costs that reduce service provision unless decentralization reaches sufficient scale. The sufficient level likely differs across countries.

### Political and Institutional Conditions
- Interactions of decentralization with political and institutional variables significantly affect public service delivery efficiency (Table 5 and related tables).
- Corruption:
  - Negatively affects the impact of fiscal decentralization on efficiency.
  - Example marginal effect: a 5 percent increase in the fiscal decentralization ratio is associated on average with a 2.5 percent decrease in the efficiency of public expenditure relative to the mean efficiency when taking into account the corruption variable (noted formula: (-0.488×0.05) ×100 = −2.4).
  - Corruption may reflect stronger local interest groups, more discretion and fewer controls at local level, and room for leakage of public resources.
- Political system interactions:
  - FD × Parliamentary (t-1) interaction is positive and statistically significant in some specifications, indicating that a parliamentary system combined with fiscal decentralization may boost expenditure efficiency.
  - Parliamentary regimes are argued to have stronger institutional frameworks to limit executive discretion.
- Autonomy of regions:
  - Existence of constitutionally autonomous regions has positive and statistically significant impacts on the relationship between decentralization and public expenditure efficiency.
  - Sufficient autonomy allows preference matching and allocation efficiency to operate.
- Subsample analysis (advanced economies vs emerging markets and developing countries):
  - Decentralization broadly improves efficiency in advanced economies and worsens (or does not improve) efficiency in emerging markets and developing countries.
  - For both subgroups and for health and education, corruption has negative impacts and regional autonomy has positive effects.

### Robustness Checks
- Outlier exclusion:
  - Baseline results are robust after excluding countries with decentralization ratios close to zero and those exceeding 90 percent (narrowed sample 0%<fd<90% in Table 7).
  - Health findings: positive impact of decentralization in advanced economies and negative in emerging markets and developing economies persist.
  - Education findings: thrust remains similar with slight differences in coefficient magnitudes.
- Alternative efficiency estimation methodologies:
  - Variant of stochastic frontier analysis (Jondrow et al., 1982) and a methodology accounting for sample heterogeneity and heteroskedasticity were used (Table 8).
  - Under both alternative estimates and for both health and education:
    - Corruption hinders the impacts of fiscal decentralization on public service efficiency with high statistical significance.
    - Parliamentary regimes and more democratic institutions show favorable contributions (sometimes with weaker statistical significance).
    - Autonomy of regions shows positive impacts with high statistical significance.
- Averaging to absorb short-term fluctuations:
  - Variables averaged over a four-year period with fiscal decentralization introduced with a one-period lag (Table 9).
  - Results support baseline findings:
    - Decentralization improves efficiency in advanced economies (columns 2 and 8).
    - Negative or non-significant impacts for emerging markets and developing countries.
    - Negative impact of corruption confirmed (columns 4 and 10).
    - Favorable role of parliamentary regimes ascertained (columns 5 and 11).
    - Autonomy of regions positive but not always statistically significant.
- Alternative political and institutional variables:
  - Bureaucracy, political stability, and checks and balances used as alternatives (Table 10).
  - Broadly similar inferences as baseline; coefficient signs mostly as expected though statistical significance is low in many cases.
- Overall: core findings are robust across exclusions, alternative efficiency computations, temporal averaging, and alternative institutional indicators.

### Revenue Decentralization
- Revenue decentralization shows positive and statistically significant impacts on public service delivery across advanced economies and emerging economies and developing countries (Table 11).
- Robustness of revenue decentralization findings:
  - Confirmed under alternative efficiency estimates (Jondrow approach and heterogeneity-adjusted estimates; Table 12).
  - Remains robust after excluding outliers and restricting sample to revenue decentralization between 0 and 90 percent (Table 13).
  - For both health and education, and for both country groups, revenue decentralization positively affects efficiency.
- Institutional interactions with revenue decentralization (Table 14):
  - Corruption decreases the positive impact of revenue decentralization on efficiency (FDR × Corruption (t-1): -0.170*** and -0.458*** in specified columns).
  - Regime variable (strength of democracy) shows a negative influence, but overall effect of revenue decentralization remains positive.
  - Checks and balances enhance the contribution of revenue decentralization (FDR × Checks (t-1): 0.264** in a reported column).
- Interpretation: Findings imply the need to accompany expenditure decentralization with sufficient revenue decentralization to ensure improved performance; revenue decentralization may facilitate effective local resource availability for service delivery.

### Conclusions and Policy Implications
- Key conclusions:
  - Fiscal decentralization can improve public service delivery efficiency, but benefits depend on conditions.
  - Expenditure decentralization appears to have improved service delivery in advanced economies; impacts in emerging and developing economies are mixed.
  - Expenditure decentralization needs to exceed an indicative threshold of about 35 percent to improve service delivery.
  - Revenue decentralization has positive impacts across country groups, suggesting the importance of aligning responsibilities with adequate revenue decentralization.
- Policy implications and necessary conditions:
  - Accompany decentralization of responsibilities with sufficient decentralization of resources (revenue).
  - Ensure a favorable institutional and political environment:
    - Effective autonomy of local governments to enable preference matching and allocative efficiency.
    - Strong accountability of local authorities to enable productive efficiency.
    - Anti-corruption measures to prevent misuse of decentralized resources.
    - Strengthen capacity at the local level.
  - Absent these conditions, fiscal decentralization can worsen public service delivery.
- Suggested extensions for future analysis:
  - Use alternative policy outcome indicators (e.g., life expectancy, school drop-out rates, PISA scores) which may exhibit larger variance across countries and time.
  - Investigate the impact of decentralization on macroeconomic performance, such as fiscal outcomes and GDP growth, given that improved public expenditure efficiency can be a transmission channel.

*Source: Authors’ calculations (content unit: _wp1559 - 35.7 percent for health and 35.4 percent for education to bring about improvements in the efficiency of public services).*

### References

### References

### Bibliographic entries
- Ahmad, E., Brosio, G. and Tanzi, V. (2008) “Local Service Provision in Selected OECD Countries: Do Decentralized Operations Work Better?”, IMF Working Paper 08/67.
- Arze del Granado, F.A., Martinez-Vazquez, J. and McNab, R. (2005) "Fiscal Decentralization and The Functional Composition of Public Expenditures", International Center for Public Policy Working Paper Series.
- Barankay, I. and Lockwood, B. (2007) "Decentralization and the Productive Efficiency of Government: Evidence from Swiss Cantons", Journal of Public Economics, 91(5-6), pp. 1197-1218.
- Bardhan, P. and Mukherjee, D. (2002) “Decentralization of Governance and Development”, Journal of Economic Perspectives, pp. 185-205.
- Battese G.E. and Coelli, T.J. (1988) “Prediction of Firm-level Technical Efficiencies: With a Generalized Frontier Production Function and Panel Data”, Journal of Econometrics, 38, pp. 387-399.
- Battese, G. E. and Coelli, T. J. (1995) “A Model for Technical Inefficiency Effects in a Stochastic Frontier Production Function for Panel Data”, Empirical Economics, 20, pp. 325–332.
- Bénassy-Quéré A., Gobalraja N. and Trannoy A. (2007) "Tax and Public Input Competition", Economic Policy, CEPR & CES & MSH, 22(4), pp. 385-430.
- Besley, T. and Smart, M. (2007) “Fiscal Restraints and Voter Welfare”, Journal of Public Economics, 91, pp. 755–773.
- Bordignon, M., Cerniglia, F., and Revelli, F. (2004) “Yardstick competition in intergovernmental relationships: theory and empirical predictions”, Economics Letters, 83, pp. 325–331.
- Cantarero, P.D. and Sanchez, M.P. (2006) “Decentralization and Health Care Outcomes: An Empirical Analysis within the European Union”, Working Paper, University of Cantabria, Spain.
- Cornwell, C., Schmidt, P. and Sickles, R. (1990) "Production Frontiers with Cross Sectional and Time Series Variation in Efficiency Levels", Journal of Econometrics, 46, pp. 185-200.
- Davoodi, H., Xie, D. and Zou, H. (1999) “Fiscal Decentralization and Economic Growth in the United States”, Journal of Urban Economics 45, pp. 228-239.
- Davoodi, H. and Zou, H. F. (1998) “Fiscal Decentralization and Economic Growth: A Cross-country Study”, Journal of Urban Economics, 43(2), pp. 244-257.
- Dziobek, C., Gutierrez-Mangas, C. and Kufa P. (2011) “Measuring Fiscal Decentralization – Exploring the IMF’s Databases”, IMF Working Paper 11/126.
- Ebeke, C. H. (2012) “The Power of Remittances on the International Prevalence of Child Labor”, Structural Change and Economic Dynamics, 23(4), pp. 452–462.
- Escolano, J., Eyraud, L., Moreno-Badia, M., Sarnes, J. and Tuladhar, A. (2012) “Fiscal Performance, Institutional Design and Decentralization in European Union Countries”, IMF Working Papers 12/45.
- Ezcurra, R. and Rodríguez-Pose, A. (2010) "Does Decentralization Matter for Regional Disparities? A Cross-country Analysis", Journal of Economic Geography, Oxford University Press, 10(5), pp. 619-644.
- Fredriksen, K. (2013) “Decentralization and Economic Growth-Part 3: Decentralization, Infrastructure Investment and Educational Performance”, OECD Working Papers on Fiscal Federalism, 16.
- Gauthier, B. and Wane, W. (2007) "Leakage of Public Resources in the Health Sector: an Empirical Investigation of Chad", Policy Research Working Paper Series, 4351.
- Gonzalez-Alegre, J. (2010) "Decentralization and the Composition of Public Expenditure in Spain", Regional Studies, 44(8), pp. 1067-1083.
- Greene, W.H. (2005) “Efficiency of Public Spending in Developing Countries: A Stochastic Frontier Approach,” World Bank.
- Grigoli, F. (2014) “A Hybrid Approach to Estimating the Efficiency of Public Spending on Education in Emerging and Developing Economies”, IMF Working Paper 14/19.
- Grigoli, F. and Kapsoli, J. (2013) “Waste Not, Want Not: The Efficiency of Health Expenditure in Emerging and Developing Economies”, IMF Working Papers 13/187.
- Grisorio, M.J. and Prota, F. (2011) “The Impact of Fiscal Decentralization on the Composition of Public Expenditure: Panel Data Evidence from Italy”, Societa italiana degli economisti, 52.
- Gupta, S. and Verhoeven, M. (2001) “The Efficiency of Government Expenditure. Experiences from Africa” Journal of Policy Modelling, 23, pp. 433–467.
- Gupta, S., Schwartz, G., Tareq, S., Allen, R., Adenauer, I., Fletcher, K. and Last D. (2007) “Fiscal Management of Scale-Up Aid” IMF Working Paper No. 07/222.
- Hayek, V. F. (1945) “The Use of Knowledge in Society”, American Economic Review, 35(4), pp. 519-530.
- Herrera, S. and Pang, G. (2005) “Efficiency of Public Spending in Developing Countries: An Efficiency Frontier Approach” World Bank Policy Research Working Paper, 3645.
- Hindriks, J. and Lockwood, B. (2005) “Decentralization and Electoral Accountability: Incentives, Separation and Voter Welfare”, CEPR Discussion Paper, 5125.
- Jayasuriya, R. and Wodon, Q. (2003) “Efficiency in Reaching the Millennium Development Goals” World Bank Working Paper, 9.
- Jiménez-Rubio, D. (2011) "The Impact of Fiscal Decentralization on Infant Mortality Rates: Evidence from OECD Countries", Social Science & Medicine, 73(9), pp. 1401-1407.
- Jondrow, J., Lovell, C.A.K., Materov, I.S. and Schmidt, P. (1982) "On the Estimation of Technical Inefficiency in the Stochastic Frontier Production Function Model", Journal of Econometrics, 19, pp. 233-238.
- Kappeler, A. and Valila, T. (2008) "Fiscal Federalism and the Composition of Public Investment in Europe”, European Journal of Political Economy, 24(3), pp. 562-570.
- Kavosi, Z., Keshtkaran, A., Samadi, A.H. and Vahedi, S. (2013) “The Effect of Fiscal Decentralization on Under-five Mortality in Iran: A Panel Data Analysis”, International Journal of Health Policy and Management, 1(4), pp. 301-306.
- Keen, M. and Marchand. M (1997) "Fiscal Competition and the Pattern of Public Spending" Journal of Public Economics, Elsevier, 66(1), pp. 33-53.
- Musgrave, R. (1969) “Theories of Fiscal Federalism”, Public Finance, 24(4), 521–32. __________ (1999) “An Essay on Fiscal Federalism” Journal of Economic Literature.
- Persson, T. and Tabellini, G. (2000) “Political Economics: Explaining Economic Policy”, MIT Press, Cambridge, MA.
- Phillips, K. and Woller, G.M. (1998) "Fiscal Decentralization and LDC Economic Growth: An empirical Investigation", Journal of Development Studies, 34(4), pp. 139-148.
- Redoano, M. (2003) “Does centralization affect the number and size of lobbies?”, CSGR Discussion Paper, University of Warwick, 604.
- Picazo, O. F., Robalino, D.A. and Voetberg, A. (2001) “Does Fiscal Decentralization Improve Health Outcomes? Evidence from a cross-country analysis”, World Bank Policy Research Working Paper, 2565.
- Seabright, Paul. (1996) “Accountability and Decentralization in Government: An Incomplete Contracts Model” European Economic Review, 40, pp. 61-89.
- Ter-Minassian, T. (1997) “Decentralization and Macroeconomic Management”, IMF Working Paper 97/155.
- Ter-Minassian, T. (1997a) “Fiscal Federalism in Theory and Practice”, (Washington: International Monetary Fund).
- Thornton, J. (2007) "Fiscal Decentralization and Economic Growth Reconsidered", Journal of Urban Economics, 61(1), pp. 64-70.
- Tiebout, C.M. (1956) “A Pure Theory of Local Expenditures”, Journal of Political Economy, 64, pp. 416-424.
- Treisman, D. (1999a) “After the Deluge: Regional Crises and Political Consolidation in Russia”, American Journal of Political Science.
- Treisman, D. ____ (2000a) “The Causes of Corruption: A Cross-National Study”, Journal of Public Economics.
- Zhang, T. and Zou, H. (1998) "Fiscal Decentralization, Public Spending, and Economic Growth in China", Journal of Public Economics, 67(2), pp. 221-240.

---

### Annex I. Countries, Coverage, and Sources

### Country coverage and data sources (selected rows preserved verbatim)
- Argentina 1993–2004 GFS, WEO
- Australia 1990–2011 OECD database
- Austria 1990–2012 Eurostat
- BahrainT 1990–2004 GFS, WEO
- Belarus 2001–2010 GFS, WEO
- Belgium 1990–2012 Eurostat
- Bhutan 1990–2009 GFS, WEO
- Bolivia 1990–2007 GFS, WEO
- Brazil 1997–2012 GFS, WEO
- Bulgaria 1995–2012 Eurostat
- Canada 1990–2010 OECD database
- Chile 1990–2012 GFS, WEO
- China (not listed explicitly in this table excerpt)
- Croatia 2002–2012 Eurostat
- Cyprus 1995–2012 Eurostat
- Czech Republic 1995–2012 Eurostat
- Denmark 1990–2012 Eurostat
- Egypt 2002–2012 GFS, WEO
- Estonia 1995–2012 Eurostat
- Finland 1990–2012 Eurostat
- France 1990–2012 Eurostat
- Georgia 1997–2012 GFS, WEO
- Germany 1990–2012 Eurostat
- Greece 1995–2012 Eurostat
- Hungary 1995–2012 Eurostat
- Iceland 1995–2012 Eurostat
- India 1990–2012 GFS, WEO
- Indonesia 1990–2004 GFS, WEO
- Iran 1990–2009 GFS, WEO
- Ireland 1990–2012 Eurostat
- Israel 1995–2012 OECD database
- Italy 1990–2012 Eurostat
- Japan 1990–2012 OECD database
- Korea 2000–2012 OECD database
- Latvia 1995–2012 Eurostat
- Lesotho 1990–2008 GFS, WEO
- Lithuania 1995–2012 Eurostat
- Luxembourg 1990–2012 Eurostat
- Maldives 1990–2011 GFS, WEO
- Malta 1995–2012 Eurostat
- Mauritius 2000–2011 GFS, WEO
- Mexico 1990–2012 GFS, WEO
- Mongolia 1992–2012 GFS, WEO
- Netherlands 1990–2012 Eurostat
- New Zealand 1990–2012 OECD database
- Norway 1990–2012 Eurostat
- Pakistan 1990–2007 GFS, WEO
- Peru 1995–2012 GFS, WEO
- Poland 1995–2012 Eurostat
- Portugal 1990–2012 Eurostat
- Romania 1995–2012 Eurostat
- Seychelles 1993–2012 GFS, WEO
- Singapore 1990–2012 GFS, WEO
- Slovak Republic 1995–2012 Eurostat
- Slovenia 1995–2012 Eurostat
- South Africa 1990–2012 GFS, WEO
- Spain 1995–2012 Eurostat
- Sweden 1993–2012 Eurostat
- Switzerland 1990–2012 Eurostat
- Tunisia 1990–2012 GFS, WEO
- Turkey 1990–2012 OECD database
- United Kingdom 1990–2012 Eurostat
- United States 1990–2012 OECD database
- Uruguay 1999–2012 GFS, WEO
- Venezuela 1990–2005 GFS, WEO

Note: Expenditure and Revenue descentralization for European and OECD countries are taken respectively from Eurostat and OECD databases. For emerging economies and developing countries, data are from GFS and WEO.

---

### Annex II. Variables, Definitions and Data Sources

### Fiscal variables
- Expenditure decentralization — Fiscal decentralization - Expenditures side
- Revenue decentralization — Fiscal decentralization - Revenue side

### Demographic and macro variables
- Imr — Mortality rate, infant (per 1,000 live births)
- Umr — Mortality rate, under-5 (per 1,000 live births)
- Primary education — Primary education, duration (years)
- Secondary education — Secondary education, duration (years)
- Average year of schooling — Average year of primary and secondary schooling
- Total population — Measures the size of the population
- Density — Population density (people per sq. km of land area)
- Real GDP pc — GDP per capita, PPP (constant 2011 international)
- Natural ressources (% GDP) — Natural resource rents

### Health and education variables
- Health expenditure — Health expenditure, public (% of GDP)
- Primary enrollment — Gross enrollment ratio, primary, both sexes (%)
- Secondary enrollment — Gross enrollment ratio, secondary, both sexes (%)
- Education exp. — Government expenditure on education as % of GDP (%)

### Political and institutional variables
- Polstab — Political stability measures the likelihood that the government will be destabilized by unconstitutional or violent means. The WGI, 2013 Update
- Government fractionalization — Probability that two deputies randomly picked from the government parties will be of different parties.
- Fractionalization — The probability that two deputies picked from the legislature will be of different parties.
- Parliamentary — Dummy variable that takes value 1 if the political system is parliamentary
- Democracy — Variable recording the strenght of the democracy
- Autonomy — Dummy variable taking value 1 with the existence of autonomous region
- Corruption — Assessment of corruption within the political system. ICRG database

### Data sources summary
- Eurostat, GFS, OECD and WEO
- World Bank, World Development Indicators 2014
- OECD and UNESCO databases
- DPI2012 Database of Political Institutions

---

### Annex III. Alternative Policy Outcome Variables (Life expectancy at birth and adjusted primary education net enrollment rate)

### Table A3.1: Fiscal Decentralization and Public Expenditure Efficiency — Health and Education (selected coefficients and statistics)
- Dependent variable: estimated efficiencies (Health / Education)
- FD (t-1): coefficients by column (1) to (10): -0.775***, -0.279***, -0.530**, -0.248, 0.367***, -1.474***, 0.931, -1.649***, -0.162, 0.547**
  - Corresponding t-statistics (in order): (-4.072), (-2.833), (-2.515), (-0.799), (4.924), (-3.182), (1.049), (-3.104), (-0.787), (2.115)
- (FD (t-1))^2: coefficients shown in some columns: 1.069***, 0.441***, 0.868**, (blank), 2.719***, -1.069, 3.106*** (with t-statistics (3.975), (3.134), (2.470), (3.613), (-0.976), (3.230))
- Real GDP pc (t-1): coefficients by column: -0.0027, 0.0150***, 0.000, 0.006, 0.002, 0.045***, -0.006, 0.052***, 0.067***, 0.013
  - Corresponding t-statistics: (-0.600), (3.588), (-0.019), (0.532), (0.520), (3.542), (-0.626), (3.149), (6.468), (0.972)
- Number of observations by column: 926, 303, 623, 528, 394, 569, 188, 381, 321, 246
- Countries by column: 58, 16, 42, 40, 30, 50, 14, 36, 32, 24
- Fisher (p-value) by column: 0.000, 0.000, 0.003, 0.119, 0.000, 0.000, 0.748, 0.000, 0.000, 0.066
- Hansen OID (p-value) by column: 0.014, 0.000, 0.002, 0.148, 0.253, 0.020, 0.198, 0.011, 0.001, 0.129
- KP-under by column: 0.000, 0.000, 0.000, 0.419, 0.000, 0.000, 0.061, 0.004, 0.048, 0.001
- Note: Table source: Authors’ calculations. Note on significance: (*), (**), and (***) denote statistical significance level of 10 percent, 5 percent and 1 percent respectively. Robust t-statistics in parentheses.

### Table A3.2: Fiscal Decentralization and Political/Institutional Environment — Health and Education (selected coefficients and statistics)
- Dependent variable: estimated efficiencies (Health / Education)
- FD (t-1): coefficients across columns: -0.065, -0.087***, -0.245***, -0.146***, -0.419, 0.003, -0.111, -0.373*
  - Corresponding t-statistics: (-1.261), (-2.969), (-3.783), (-3.492), (-1.378), (0.015), (-0.763), (-1.955)
- FD × Corruption (t-1): coefficient displayed: -0.026 (t-statistic (-0.604)); and in another column -0.426*** (t-statistic (-2.830))
- FD × Parliamentary (t-1): coefficient displayed: -0.064 (t-statistic (-0.355)); and in another column -3.643 (t-statistic (-1.585))
- FD × Regime (t-1): coefficient displayed: 0.008*** (t-statistic (3.961)); and in another column -0.002 (t-statistic (-0.493))
- FD × Autonomy (t-1): coefficients displayed: 0.530*** (t-statistic (5.022)); 1.561*** (t-statistic (3.259))
- Real GDP pc (t-1): coefficients across columns: 0.008, 0.0118*, -0.027***, 0.006, -0.038, 0.148***, 0.057**, 0.038**
  - Corresponding t-statistics: (1.449), (1.722), (-3.383), (1.642), (-1.066), (3.376), (2.407), (2.220)
- Number of observations by column: 861, 926, 925, 926, 529, 569, 568, 569
- Countries by column: 54, 58, 58, 58, 46, 50, 50, 50
- Fisher (p-value) by column: 0.002, 0.000, 0.000, 0.000, 0.000, 0.000, 0.000, 0.000
- Hansen OID (p-value) by column: 0.001, 0.002, 0.471, 0.113, 0.194, 0.206, 0.000, 0.024
- KP-under by column: 0.264, 0.122, 0.000, 0.000, 0.026, 0.445, 0.000, 0.001
- Note: Table source: Authors’ calculations. Note on significance: (*), (**) and (***) denote statistical significance level of 10%, 5% and 1% percent respectively. Robust t-statistics in parentheses.

---

*Source: _wp1559 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1559.pdf_