## _wp11126

## Source details

**Canonical URL:** [_wp11126](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11126.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11126.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11126.pdf.json)

---

### I. Introduction
- Purpose: present and discuss statistics available to support empirical analysis of fiscal decentralization (fiscal federalism) using IMF Government Finance Statistics Yearbook (GFSY) data.
- Definitions and theoretical context:
  - Fiscal decentralization = study of the structure and functioning of multi-tiered governments.
  - First-generation literature: Samuelson (1954, 1955), Arrow (1970), Musgrave (1959).
  - Second-generation literature emphasizes information problems, moral hazard, free riding, hard vs. soft budget constraints, and reliance on local own revenues (Weingast 1995, McKinnon 1997, Rodden 2003).
- Institutional framework and measurement:
  - GFSM2001 institutional definition: GL3 = general government; GL2 = consolidated central government; GL1 = budgetary central government.
  - Decentralization measured as GL2/GL3 ratio for selected fiscal indicators.
- Database scope and sample:
  - GFSY data considered from 1990 to 2008.
  - IMF membership: currently 187, approximately 150 in 1990.
  - Sample construction: eliminate countries never reporting GL3; leaves approximately 80 countries.
  - Dimensions: 19 years × ~80 countries × 4 fiscal indicators = 5,760 potential observations.
  - Actual observations:
    - Revenue = 730
    - Tax Effort = 762
    - Expenditures = 687
    - Compensation of Employees = 736
    - Total = 2,915
- Organization of paper:
  - Section II: decentralization for four fiscal indicators (revenue, tax effort, expenditure, compensation of employees).
  - Section III: global trends.
  - Section IV: findings by WEO country groupings, constitutional form, and accounting basis.
  - Section V: summary, conclusions, and statistical recommendations.

### II. Global trends in fiscal decentralization — measurement and indicators
- Four GL2/GL3 indicators used:
  - Revenue (GL2/GL3 ratio of government revenue).
  - Tax Effort (sum of tax revenue and compulsory social security contributions; GFSM2001, paragraph 5.11).
  - Expenditure.
  - Compensation of Employees.
- Data processing:
  - Consolidated for intra-flows to reflect autonomous execution by level of government.
  - Code descriptions and aggregation methodology described in a forthcoming technical note (Working with the Government Finance Statistics Yearbook Database: A User’s Guide (forthcoming 2011)).
- Example divergence (South Africa, 2008):
  - GL2 collected about 95 percent of GL3 taxes (centralized by tax effort).
  - GL2 accounts for about 34 percent of compensation of employees (decentralized by employment).

### II.A. Revenue (GL2/GL3 ratios)
- Cross-section (2008):
  - Most countries fall in the range of 80-100 percent GL2/GL3 revenue ratio.
  - A small number of countries with ratios well below 50 percent; a larger number at 100 percent.
- Countries among the most decentralized (central government collects lower share):
  - Argentina, Canada, China, Denmark, India, South Africa, Switzerland, United States.
- Top five (most decentralized by GL2/GL3 revenue ratio) central government shares:
  - central government in the five most decentralized countries collects 47 to 62 percent of general government’s revenue.
- Reported ranges for top-five lowest GL2/GL3 ratios across indicators:
  - Revenue: 47 to 62 percent.
  - Tax Effort: 50 to 64 percent.
  - Expenditure: 35 to 50 percent.
  - Compensation of Employees: 12 to 23 percent.
- Time changes (1990–2008 caveats):
  - Many countries in “no change” range; ratios relatively stable.
  - Largest revenue GL2/GL3 ratio decreases: Bolivia, Italy, San Marino, Spain, Switzerland.
  - Largest revenue GL2/GL3 ratio increases: Bulgaria, Georgia, Kazakhstan, Mongolia, Romania.
- Observed patterns:
  - Small-area countries often show high centralization (Maldives, San Marino, Seychelles, Singapore, St. Kitts and Nevis).
  - Large-area countries often show greater decentralization (Canada, P.R. of China (mainland), India, United States), with exceptions (Switzerland decentralized despite small area; Russia not among most decentralized).
  - No country observed with GL2/GL3 ratio equal to zero; perfect centralization (ratio equal to one) is relatively common.
- Sample table excerpts (preserved verbatim format notes in source):
  - Canada: 47 (2007)
  - China, P.R.: Mainland: 48 (2007)
  - Switzerland: 49 (2007)
  - United States: 54 (2008)
  - Argentina: 62 (2004)
  - GL2/GL3 = 100 examples include Kuwait, Maldives, Seychelles, Singapore, St. Kitts and Nevis (entries preserved verbatim in source).

### II.B. Tax Effort (GL2/GL3 ratio of tax effort)
- Cross-section and trends:
  - Tax effort has the lowest dispersion across indicators in both cross-section and over time.
  - Cross-sectional range for tax effort reported as 52 percent (similar to revenue range of 53 percent).
  - Many countries in “no change” zone; small number with large changes.
- Examples and extremes:
  - Most decentralized by tax effort: China, P.R.: Mainland (48), Canada (51), Switzerland (58), India (62), United States (64).
  - GL2/GL3 = 100 for tax effort: Malta, San Marino, Seychelles, Singapore, St. Kitts and Nevis (entries preserved verbatim).
- Countries with largest tax effort ratio changes (Table 4 summary):
  - Largest decreases: Spain, Switzerland, Bolivia, Italy, Slovak Republic.
  - Largest increases: Romania, Mongolia, Bulgaria, Kazakhstan, Georgia.

### II.C. Expenditure (GL2/GL3 ratio)
- Cross-section:
  - GL2/GL3 ratio range for expenditure reported as 65 percent.
  - Central government average responsible for expenditures less than for revenue.
- Examples:
  - Most decentralized (lowest GL2/GL3 expenditure ratios): Canada (35), Denmark (36), Switzerland (43), India (48), South Africa (48).
  - GL2/GL3 = 100 examples: Kuwait, Maldives, Seychelles, Singapore, St. Kitts and Nevis.
- Trends (countries with largest changes, Table 6):
  - Largest decreases: Spain (-17), Peru (-14), Bolivia (-12), Slovak Republic (-12), Thailand (-11).
  - Largest increases: Moldova (+7), Colombia (+12), Ireland (+12), Georgia (+15), Mongolia (+24).

### II.D. Compensation of Employees (GL2/GL3 ratio)
- Dispersion and volatility:
  - Compensation of employees has the widest range: 88 percent cross-sectional range.
  - Over time changes up to 63 percent in some cases.
- Cross-section extremes:
  - Lowest central shares (most decentralized by compensation): Switzerland (12), Canada (18), Spain (22), Sweden (23), Germany (23).
  - GL2/GL3 = 100 examples: Maldives, San Marino, Seychelles, Singapore, St. Kitts and Nevis.
- Trend extremes (Table 8):
  - Largest decreases: Slovak Republic (-29), Macedonia, FYR (-24), Spain (-24), Thailand (-11), Russian Federation (-8).
  - Largest increases: Armenia (+24), Ireland (+29), Bulgaria (+43), Georgia (+45), Mongolia (+63).

### III. Stability over the 20-year period (summary)
- Overall finding: financial decentralization appears relatively stable over the 20-year period analyzed (1990–2008), with only small moves toward decentralization during 1998–2008 for the full sample.
- Exceptions: transition economies showed significant changes (Armenia, Bulgaria, Georgia, Kazakhstan, Macedonia, Romania, Russian Federation, Slovak Republic).
- Magnitude of changes (absolute value over sample):
  - Every fiscal aggregate except compensation of employees ranges from 7 to 28 percent (absolute value).
  - Compensation of employees range is 8 to 63 percent.
- Implications for IMF operational work:
  - Annual forecasts for general government (GL3) updated with partial outcomes often available only at GL2 level.
  - If GL2/GL3 ratios are known and stable, they can supplement missing local government data in models; analysts must monitor policy changes and business-cycle turning points that can magnify forecasting errors.
- Note on interpretation:
  - Changes in GL2/GL3 ratios may reflect true decentralization or nominal changes in taxation policy that do not represent transfers to lower levels.

### IV. Data on fiscal decentralization by country groups (summary statistics)
- Groupings analyzed:
  - Full sample (“All”).
  - Geographic regions (WEO definitions).
  - Fuel/primary material exporters (WEO definitions).
  - Constitutional power arrangement: unitary vs. federal.
  - Accounting basis of recording: accrual vs. cash (GFSY metadata).
- Key cross-country means and dispersion (Table 9 summaries and selected exact values):
  - Central government responsible for revenue collection: overall average 88 percent (GL2/GL3 revenue).
  - Central government responsible for expenditures: overall average 79 percent (GL2/GL3 expenditure).
  - Dispersion (standard deviation) across indicators:
    - Tax effort: standard deviation 10 percent.
    - Revenue: standard deviation 11 percent.
    - Expenditures: standard deviation 16 percent.
    - Compensation of employees: standard deviation 26 percent.
- Selected peer-group exact values (Fiscal Year 2008, GL2/GL3 ratios):
  - All (revenue): Average 88; Std. Dev. 11; Countries 63; 10-year Change 22.
  - All (expenditure): Average 79; Std. Dev. 16; Countries 62; 10-year Change 19.
  - Advanced Economies (revenue): Average 82; Std. Dev. 12; Countries 24; 10-year Change 15; 10-year Change -1 (expenditure).
  - Developing and Emerging Economies (revenue): Average 91; Std. Dev. 7; Countries 39; 10-year Change 7.
  - Unitary Government (revenue): Average 90; Std. Dev. 8; Countries 55; 10-year Change 18.
  - Federal Government (revenue): Average 72; Std. Dev. 10; Countries 8; 10-year Change 4; 10-year Change -3 (expenditure).
  - Cash basis system (revenue): Average 92; Std. Dev. 8; Countries 30.
  - Accrual basis system (revenue): Average 84; Std. Dev. 11; Countries 36.
- Statistically significant contrasts (Difference-in-Means t-statistics reported):
  - Advanced vs Developing and Emerging (tax effort): Difference-in-Means statistic = (-2.75); interpretation: central government execution of tax effort is 85 percent (Std. Dev. 11 percent) in advanced economies compared to 92 percent (Std. Dev. 8 percent) in developing and emerging economies; result significant at 95 percent level of confidence.
  - Federal vs Unitary (revenue and compensation): t-Statistics reported (revenue) (4.59) and (expenditure) (4.23) indicating large statistical differences with federal constitutions tending to be more decentralized.
  - Cash vs Accrual (revenue and compensation): t-Statistic (revenue) (3.18); (compensation) (2.92); reported differences significant at over a 95 percent confidence level.
- Accrual versus cash recording contrasts:
  - Revenue GL2 to GL3 ratio: cash basis countries = 92 percent; accrual basis countries = 84 percent.
  - Compensation of employees GL2 to GL3 ratio: cash basis countries = 80 percent; accrual basis countries = 59 percent.
- Geography and size:
  - Large geographic area economies tend to be more decentralized; exceptions include Switzerland.
  - Most centralized countries listed (small area examples): Malta, San Marino, Seychelles, Singapore, St. Kitts and Nevis.

### V. Key findings, questions, and implications for statistical work
- Key empirical findings:
  - GL2/GL3 ratios are generally high for revenue and tax effort (centralized tax collection) and somewhat lower for expenditure and compensation of employees.
  - Tax effort is the most stable and least dispersed indicator across countries and over time.
  - Compensation of employees is the most variable indicator across countries and over time.
  - Many countries show little change in centralization over 1990–2008; notable exceptions are transition economies.
  - Perfect centralization (GL2/GL3 = 1) is common; perfect decentralization (GL2/GL3 = 0) is not observed.
- Open questions for further research:
  - Are there economies of scale in tax collection that justify centralization?
  - To what extent are observed patterns driven by governance or political choices?
  - Explore differences between GL2/GL3 ratios of revenues and expenditures and the possibility of a composite indicator of decentralization.
  - Further scrutiny of highly centralized economies (GL2/GL3 equal or close to one) to identify other indicators of decentralization not captured by current data.
- Statistical and operational recommendations:
  - Use knowledge of relative stability of GL2/GL3 ratios to supplement missing local government data in forecasting models, while monitoring for policy shifts or nominal tax changes that could invalidate ratio stability.
  - Develop strategies to improve communication among government levels for functions where execution has been devolved (e.g., compensation of employees).
  - Recognize data limitations: differing time coverage across countries, missing observations, and pre-1990 methodological breaks in GFSY (GFSM 1986 to GFSM2001).
  - Timely fiscal data on general government (GL3) are important; statistical dissemination standards recommend annual GL3 data, with increasing demand for quarterly GL3 data about three months after the reference period.

*Source: _wp11126 (chapter content and tables/figures as presented).*

### Bibliography ...........................................................................................................

### _wp11126 - Bibliography ............................................................................................................................29

### Tables
- 1. GL2/GL3 Ratios Revenue....................................................................................................11
- 2. Countries with the Largest Changes in GL2/GL3 Ratios: Revenue ....................................11
- 3. GL2/GL3 Ratios: Tax Effort ................................................................................................13
- 4. Countries with the Largest Changes in GL2/GL3 Ratios: Tax Effort .................................13
- 5. GL2/GL3 Ratios: Expenditure .............................................................................................15
- 6. Countries with the Largest Changes in GL2/GL3 Ratios: Expenditure ..............................15
- 7. GL2/GL3 Ratios: Compensation of Employees ..................................................................17
- 8. Countries with the Largest Changes in GL2/GL3 Ratios: Compensation of Employees ....17
- 9. Summary Statistics on Fiscal Decentralization by Country Groups, Fiscal Year 2008 ......23

### Figures
- 1. Institutional Structure of the General Government ................................................................6
- 2. South Africa GL2/GL3 Ratios for Selected Fiscal Indicators ...............................................8
- 3. Distribution of GL2/GL3 Ratios: Revenue ..........................................................................12
- 4. Changes in GL2/GL3 over Time: Revenue .........................................................................12
- 5. Distribution of GL2/GL3 Ratios: Tax Effort .......................................................................14
- 6. Changes in GL2/GL3 over Time: Tax Effort ......................................................................14
- 7. Distribution of GL2/GL3 Ratios: Expenditure ....................................................................16
- 8. Changes in GL2/GL3 over Time: Expenditure ....................................................................16
- 9. Distribution of GL2/GL3 Ratios: Compensation of Employees ..........................................18
- 10. Changes in GL2/GL3 over Time: Compensation of Employees .......................................18
- 11. The Share of Central Government Is Broadly Stable Overtime: Four Country Examples 20
- 12. Cross-Country Average of GL2/GL3 Ratios for Four Fiscal Indicators ...........................26
- 13. GL2/GL3 Ratios, Standard Deviation of Four Fiscal Indicators, 2008 .............................26

### Boxes
- (No boxes listed in the provided content.)

*Source: _wp11126 - Bibliography (page listing of tables and figures).*

### 1. Classifying Governments in Terms of Constitution Power Arrangement ...........................21

### 1. Classifying Governments in Terms of Constitution Power Arrangement

### I. Introduction
- Purpose: present and discuss statistics available to support empirical analysis of fiscal decentralization (fiscal federalism) using IMF Government Finance Statistics Yearbook (GFSY) data.
- Definitions and theoretical context:
  - Fiscal decentralization = study of the structure and functioning of multi-tiered governments.
  - First-generation literature: Samuelson (1954, 1955), Arrow (1970), Musgrave (1959) — central government for macro stabilization, income redistribution, national public goods; local governments for local public goods.
  - Tax assignment problem: local taxation should focus on property taxes and user fees due to mobile tax bases.
  - Second-generation literature: information problems, moral hazard, free riding; hard vs. soft budget constraints; importance of reliance on local own revenues (Weingast 1995, McKinnon 1997, Rodden 2003).
- Institutional framework:
  - Uses GFSM2001 institutional definition of government: GL3 = general government; GL2 = consolidated central government; GL1 = budgetary central government; includes state and local governments, social security, extra-budgetary units.
  - Decentralization measured as GL2/GL3 ratio for selected fiscal indicators.
- Database scope and sample:
  - GFSY data considered from 1990 to 2008 (database has a break in the series in 1990).
  - IMF membership: currently 187, approximately 150 in 1990.
  - Sample construction: eliminate countries never reporting GL3; leaves approximately 80 countries.
  - Dimensions: 19 years × ~80 countries × 4 fiscal indicators = 5,760 potential observations.
  - Actual observations: Revenue = 730; Tax Effort = 762; Expenditures = 687; Compensation of Employees = 736; total = 2,915 observations.
- Organization of paper:
  - Section II: decentralization for four fiscal indicators (revenue, tax effort, expenditure, compensation of employees) with cross-section extremes and largest changes.
  - Section III: global trends.
  - Section IV: findings by IMF WEO country groupings, constitutional form (unitary vs. federal), and accounting basis (accrual vs. cash).
  - Section V: summary, conclusions, and statistical recommendations.

### II. Global trends in fiscal decentralization — measurement and indicators
- Four GL2/GL3 indicators used:
  - Revenue (GL2/GL3 ratio of government revenue).
  - Tax Effort (sum of tax revenue and compulsory social security contributions; GFSM2001, paragraph 5.11).
  - Expenditure.
  - Compensation of Employees.
- Data processing:
  - Consolidated for intra-flows to reflect autonomous execution by level of government.
  - Code descriptions and aggregation methodology described in a forthcoming technical note (Working with the Government Finance Statistics Yearbook Database: A User’s Guide (forthcoming 2011)).
- Example illustrating indicator divergence:
  - South Africa, 2008: GL2 collected about 95 percent of GL3 taxes (centralized by tax effort) while GL2 accounts for about 34 percent of compensation of employees (decentralized by employment).

### A. Fiscal decentralization measured in terms of Revenue (GL2/GL3 ratios)
- Cross-section distribution (2008):
  - Most countries fall into the range of 80-100 percent GL2/GL3 revenue ratio.
  - Small number of countries with ratios well below 50 percent; larger number at 100 percent.
- Countries identified among the most decentralized (central government collects lower share):
  - Argentina, Canada, China, Denmark, India, South Africa, Switzerland, United States.
- Top five (most decentralized by GL2/GL3 revenue ratio) central government shares:
  - central government in the five most decentralized countries collects 47 to 62 percent of general government’s revenue.
- Ranges reported for top-five lowest GL2/GL3 ratios across indicators:
  - Revenue: 47 to 62 percent.
  - Tax Effort: 50 to 64 percent.
  - Expenditure: 35 to 50 percent.
  - Compensation of Employees: 12 to 23 percent.
- Time changes (1990–2008 sample caveats):
  - Largest changes show many countries in “no change” range — ratios relatively stable over period.
  - Examples of countries with largest revenue GL2/GL3 ratio decreases (local governments obtained greater revenue role): Bolivia, Italy, San Marino, Spain, Switzerland.
  - Examples with largest increases (more centralized): Bulgaria, Georgia, Kazakhstan, Mongolia, Romania.
- Observed patterns and hypotheses:
  - Small-area countries often show high centralization (Maldives, San Marino, Seychelles, Singapore, St. Kitts and Nevis).
  - Large-area countries often show greater decentralization (Canada, P.R. of China (mainland), India, United States), with exceptions (Switzerland decentralized despite small area; Russia not among most decentralized).
  - No country observed with GL2/GL3 ratio equal to zero (perfect decentralization); perfect centralization (ratio equal to one) is relatively common.
- Data notes and sample excerpts:
  - Table 1 (GL2/GL3 ratios for revenue — highest and lowest) highlights:
    - Canada: 47 (2007) observations/time period 8-0 (as shown in source).
    - China, P.R.: Mainland: 48 (2007) observations/time period 3-11.
    - Switzerland: 49 (2007) observations/time period 10-9.
    - United States: 54 (2008) observations/time period 10-5.
    - Argentina: 62 (2004) observations/time period 3-3.
    - Countries with GL2/GL3 = 100: Kuwait (1002009 80?), Maldives (1002008 190?), Seychelles (1002008 70?), Singapore (1002008 80?), St. Kitts and Nevis (1002006 20?) — note: entries preserved verbatim as in source table format.

### B. Tax Effort (GL2/GL3 ratio of tax effort)
- Cross-section and trend features:
  - Tax effort has the lowest dispersion across indicators in both cross-section and over time.
  - Cross-sectional range for tax effort reported as 52 percent (similar to revenue range of 53 percent).
  - Stability: many countries in “no change” zone; small number with large changes.
- Examples from tables:
  - Countries among most decentralized by tax effort: China, P.R.: Mainland (48), Canada (51), Switzerland (58), India (62), United States (64).
  - Countries with GL2/GL3 = 100 for tax effort include Malta, San Marino, Seychelles, Singapore, St. Kitts and Nevis (entries preserved verbatim).
  - Table 4 lists countries with largest tax effort ratio changes: Spain, Switzerland, Bolivia, Italy, Slovak Republic (largest decreases); Romania, Mongolia, Bulgaria, Kazakhstan, Georgia (largest increases) — preserved as in source.

### C. Expenditure (GL2/GL3 ratio)
- Cross-section:
  - GL2/GL3 ratio range for expenditure reported as 65 percent.
  - Central government average responsible for expenditures less than for revenue (see Table 9 summary in Section IV).
- Examples:
  - Most decentralized expenditure ratios (lowest GL2/GL3): Canada (35), Denmark (36), Switzerland (43), India (48), South Africa (48).
  - Most centralized (GL2/GL3 = 100) include Kuwait, Maldives, Seychelles, Singapore, St. Kitts and Nevis (entries preserved).
- Trends:
  - Figure 8 and Table 6 show countries with largest changes in expenditure GL2/GL3 ratio: Spain (-17), Peru (-14), Bolivia (-12), Slovak Republic (-12), Thailand (-11) for decreases; Moldova (+7), Colombia (+12), Ireland (+12), Georgia (+15), Mongolia (+24) for increases.

### D. Compensation of Employees (GL2/GL3 ratio)
- Dispersion and volatility:
  - Compensation of employees has the widest range: 88 percent cross-sectional range; over time changes up to 63 percent in some cases.
  - Figure 10 indicates greater volatility: five countries with increase > 20 percent (Armenia, Bulgaria, Georgia, Ireland, Mongolia); three countries show changes > 40 percent.
- Cross-section extremes:
  - Lowest central shares (most decentralized by compensation): Switzerland (12), Canada (18), Spain (22), Sweden (23), Germany (23).
  - Highest central shares (GL2/GL3 = 100): Maldives, San Marino, Seychelles, Singapore, St. Kitts and Nevis.
- Trend extremes (Table 8):
  - Largest decreases: Slovak Republic (change -29), Macedonia, FYR (-24), Spain (-24), Thailand (-11), Russian Federation (-8).
  - Largest increases: Armenia (+24), Ireland (+29), Bulgaria (+43), Georgia (+45), Mongolia (+63).

### III. Stability over the 20-year period (summary)
- Overall finding: financial decentralization appears relatively stable over the 20-year period analyzed (1990–2008), with only small moves toward decentralization during 1998–2008 for the full sample.
- Exceptions: countries undergoing transition to market economies showed significant changes (examples: Armenia, Bulgaria, Georgia, Kazakhstan, Macedonia, Romania, Russian Federation, Slovak Republic).
- Magnitude of changes (absolute value over sample):
  - Every fiscal aggregate except compensation of employees ranges from 7 to 28 percent (absolute value).
  - Compensation of employees range is 8 to 63 percent.
- Implications for IMF operational work:
  - Annual forecasts for general government (GL3) are updated with partial outcomes often available only at GL2 level.
  - If GL2/GL3 ratios are known and stable, they can supplement missing local government data in models; analysts must remain aware of policy changes and business-cycle turning points that can magnify forecasting errors.
- Note on interpretation:
  - Changes in GL2/GL3 ratios may reflect true decentralization or nominal changes in taxation policy (e.g., change in tax rates collected by GL2) that do not represent transfers to lower levels.

### IV. Data on fiscal decentralization by country groups (summary statistics)
- Groupings analyzed:
  - Full sample (“All”).
  - Geographic regions (WEO definitions).
  - Fuel/primary material exporters (WEO definitions).
  - Constitutional power arrangement: unitary vs. federal (political science literature definitions).
  - Accounting basis of recording: accrual vs. cash (GFSY metadata).
- Key cross-country means and dispersion (Table 9 summarization as described in text):
  - Central government responsible for revenue collection: overall average 88 percent (GL2/GL3 revenue).
  - Central government responsible for expenditures: overall average 79 percent (GL2/GL3 expenditure).
  - Dispersion (standard deviation) across indicators:
    - Tax effort: standard deviation 10 percent.
    - Revenue: standard deviation 11 percent.
    - Expenditures: standard deviation 16 percent.
    - Compensation of employees: standard deviation 26 percent.
- Interpretation in light of fiscal federalism literature:
  - Findings are consistent with view that decentralization of taxation desirable for limited revenues (property taxes and user fees); central governments retain control on collection of taxes and mandatory social contributions.

### V. Key findings, questions, and implications for statistical work
- Key empirical findings:
  - GL2/GL3 ratios are generally high for revenue and tax effort (centralized tax collection) and somewhat lower for expenditure and compensation of employees.
  - Tax effort is the most stable and least dispersed indicator across countries and over time.
  - Compensation of employees is the most variable indicator across countries and over time.
  - Many countries show little change in centralization over 1990–2008; notable exceptions are transition economies.
  - Perfect centralization (GL2/GL3 = 1) is common; perfect decentralization (GL2/GL3 = 0) is not observed.
- Open questions (not analyzed in the paper):
  - Are there economies of scale in tax collection that justify centralization?
  - To what extent are observed patterns driven by governance or political choices?
- Statistical and operational recommendations:
  - Use knowledge of relative stability of GL2/GL3 ratios to supplement missing local government data in forecasting models, while monitoring for policy shifts or nominal tax changes that could invalidate ratio stability.
  - Recognize data limitations: differing time coverage across countries, missing observations, and pre-1990 methodological breaks in GFSY (GFSM 1986 to GFSM2001).
- Data caveats reiterated:
  - Time period (number of years) varies across countries; changes reported differ in meaningful comparability because of varying sample lengths.
  - Prior to 1990 GFSY statistics were in GFSM 1986 format; from 1990 back-filled and current statistics compiled under GFSM2001. A project is in train to link pre-1990 and post-1990 data and to fill or explain missing observations.

*Source: IMF Government Finance Statistics Yearbook database and manuscript figures and tables as presented in the chapter.*

### Box 1. Classifying Governments in Terms of Constitution Power Arrangement

### Box 1. Classifying Governments in Terms of Constitution Power Arrangement

### Definitions of constitutional power arrangements
- Unitary government
  - Based on a constitution which sets supreme authority to the central government.
  - Central government can delegate power through decentralization to local governing institutional units, serving as an administrative arm of the central government to provide uniform and equal access to public services.
  - Unitary government can be single (e.g. Monaco, Singapore, Sweden) or multitiered (the central government, state government, and local government).
- Federal government
  - Based on a constitution that defines how power is shared between the government institutional units (the central government, state government, and local government).
  - The constitution allocates duties, rights, and privileges to each level of government.
  - Multiordered government, with somewhat independent governments that share decision-making responsibilities for the supply of public services.

### Empirical patterns from Table 9 and peer-group analysis (Fiscal Year 2008, GL2/GL3 ratios)
- Data and measures
  - GL2/GL3 refers to the share of central government (GL2) in relation to general government (GL3).
  - Four fiscal aggregates analyzed include: revenue (tax effort), expenditures, compensation of employees, and overall fiscal indicators.
- Key cross-country means and contrasts (selected exact values reported)
  - All (revenue): Average 88; Std. Dev. 11; Countries 63; 10-year Change 22.
  - All (expenditure): Average 79; Std. Dev. 16; Countries 62; 10-year Change 19.
  - Advanced Economies (revenue): Average 82; Std. Dev. 12; Countries 24; 10-year Change 15; 10-year Change -1 (expenditure).
  - Developing and Emerging Economies (revenue): Average 91; Std. Dev. 7; Countries 39; 10-year Change 7.
  - Unitary Government (revenue): Average 90; Std. Dev. 8; Countries 55; 10-year Change 18.
  - Federal Government (revenue): Average 72; Std. Dev. 10; Countries 8; 10-year Change 4; 10-year Change -3 (expenditure).
  - Cash basis system (revenue): Average 92; Std. Dev. 8; Countries 30.
  - Accrual basis system (revenue): Average 84; Std. Dev. 11; Countries 36.
- Statistically significant contrasts (Difference-in-Means t-statistics reported)
  - Advanced vs Developing and Emerging (tax effort): Difference-in-Means statistic = (-2.75) — interpretation: the central government execution of tax effort is 85 percent (Std. Dev. 11 percent) in advanced economies compared to 92 percent (Std. Dev. 8 percent) in developing and emerging economies; result significant at 95 percent level of confidence.
  - Federal vs Unitary (revenue and compensation): t-Statistics reported (revenue) (4.59) and (expenditure) (4.23) indicating large statistical differences with federal constitutions tending to be more decentralized.
  - Cash vs Accrual (revenue and compensation): t-Statistic (revenue) (3.18); (compensation) (2.92); reported differences significant at over a 95 percent confidence level.
- Empirical conclusions drawn from peer-group analysis
  - Advanced economies tend to be more decentralized than developing and emerging economies across the four fiscal aggregates.
  - Economies with a federal constitution tend to be more decentralized than economies with a unitary constitution; the constitutional arrangement is not perfectly determinative (example: Sweden is unitary but decentralized).
  - Governments tend to decentralize expenditure execution more than revenue collection; revenues (mostly taxes) are more centralized due to investment costs and economies of scale in tax collection and processing.

### Accrual versus cash recording and decentralization
- GFSM 2001 recommendation and practice
  - GFSM 2001 recommends preparing government finance statistics on an accrual basis in addition to cash flow statements.
  - Many countries implement accruals-based accounting for key areas (e.g., grants-in-kind, accounts payable including arrears).
- Empirical contrasts from GFSY metadata
  - Revenue GL2 to GL3 ratio: cash basis countries = 92 percent; accrual basis countries = 84 percent.
  - Compensation of employees GL2 to GL3 ratio: cash basis countries = 80 percent; accrual basis countries = 59 percent.
  - Differences in sample means between accrual and cash are statistically significant (t-statistics reported earlier), with significance at over a 95 percent confidence level.
- Interpretation and causality
  - The observed association does not prove causality between decentralization and sophisticated recording formats.
  - Possible causal scenarios:
    - Developments in accounting systems may allow or support devolvement of activities.
    - Economic development may enable modernization of public administration and accrual systems, and concurrently greater autonomy for local governments.
  - For compensation of employees and expenditures, there is a significant mean difference between cash and noncash recording bases.

### Geography, country size, and decentralization
- Pattern: Economies with large geographic areas tend to be more decentralized than economies with a small geographic area.
- Exceptions and examples:
  - Switzerland appears an exception (small area effect countered by mountainous topography).
  - Most centralized countries listed: Malta, San Marino, Seychelles, Singapore, and St. Kitts and Nevis — all with small geographic areas.

### Main conclusions and implications for statistics and policy
- Data coverage and approach
  - Study uses data on fiscal decentralization for about 80 countries from the Government Finance Statistics Yearbook (GFSY) over up to 20 years starting in 1990.
  - Analyses focus on two main levels of government (GL2 and GL3) per GFSM 2001 definitions.
- Stability and estimation implications
  - Except for countries undergoing systemic reforms, levels of decentralization are relatively stable over time.
  - Stability supports the development of estimates for lower-level government data to supplement more timely GL1 or GL2 data.
- Policy and statistical recommendations
  - Timely fiscal data on general government (GL3) are important; statistical dissemination standards recommend annual GL3 data, with increasing demand for quarterly GL3 data about three months after the reference period.
  - Data compilers can develop strategies to improve communication among government levels for functions where execution has been devolved (e.g., compensation of employees).
- Suggested further research directions
  - Explore differences between GL2/GL3 ratios of revenues and expenditures and the possibility of a composite indicator of decentralization.
  - Further scrutiny of highly centralized economies (GL2/GL3 equal or close to one) to determine other indicators of decentralization not captured by current data.
  - Address significant gaps in the GFSY database to improve coverage and analysis.

*Italic: Source — _wp11126 - Box 1. Classifying Governments in Terms of Constitution Power Arrangement*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11126.pdf_