## _wp1346

## Source details

**Canonical URL:** [_wp1346](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1346.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1346.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1346.pdf.json)

---

### Box 1 — Why Fiscal Decentralization May Inflate the Size of the Government: Principal channels
- Four principal channels by which decentralization may increase public sector size:
  - Decentralization does not generate unlimited efficiency gains:
    - Diseconomies of scale, negative inter-jurisdictional spillovers, and coordination issues can offset gains.
    - Macroeconomic stability and redistribution responsibilities are often argued to be best left at the center.
    - Resource-allocation guidance: central government should provide national public goods; subnational governments should deliver services with local benefits (examples: waste disposal, street maintenance, primary education).
  - Normative considerations seldom guide actual expenditure assignments:
    - Decentralization is largely driven by political motives, historical and cultural legacies.
    - Constitutional boundaries and the existing “political map” limit the scope for adjusting local entities’ size according to efficiency criteria.
  - Fiscal discipline is harder to enforce at the subnational level:
    - Soft budget constraint: bailouts from the center; subsidized loans from public banks or state-owned enterprises; supplier arrears; underfunded public sector pensions.
    - Common pool problem: local policymakers may not internalize the cost of spending when financed with intergovernmental transfers or shared revenue funded by other jurisdictions’ taxpayers.
    - Moral hazard and weak governance: devolved tax bases and spending responsibilities can outpace central monitoring; local bureaucracies may be lower quality; clientelism and corruption may be more prevalent.
  - Expenditure-control failures reflect structural local budget management difficulties:
    - Procyclicality of resources: subnational own revenue sources are narrow and volatile; transfers from the center are often procyclical.
    - Unclear spending assignments: concurrent responsibilities weaken accountability and encourage blame-shifting.
    - Weak PFM systems: lack of audit and control mechanisms; looser accounting standards (notably for arrears); absence of multi-year fiscal frameworks; ill-designed fiscal rules.
    - Flaws in transfer system design: matching grants and inadequate treatment of subnational heterogeneity can encourage overspending; transfers often allocated based on actual spending costs rather than “expenditure needs” and independent of service quality.

### Fragmentation, empirical evidence, and the role of Vertical Fiscal Imbalance (VFI)
- Government fragmentation:
  - Literature provides limited support that fragmented governments are smaller and spend less; effects differ between “general” and “special-purpose” subnational governments.
- Empirical cross-country findings:
  - Decentralization per se: no strong evidence that fiscal decentralization by itself increases government size; earlier literature finds no relation; more recent work shows opposite effects of spending and revenue decentralizations.
  - Vertical fiscal imbalance (VFI): consensus on a “conditional” effect—spending decentralization raises general government spending when financed by transfers or borrowing (i.e., associated with large VFI).
  - Expenditure decentralization without corresponding local tax powers does not generate tax competition or increase local accountability.
  - Evidence cited: negative effect of VFI and its interaction with spending decentralization found in Jin and Zou (2002), Rodden (2003), Eyraud and Lusinyan (2011); Fornasari et al. (2000) show subnational spending not funded by local taxes is additional to central spending.

### Stylized facts on expenditure decentralization in Europe (exact figures preserved)
- Fact 1 — Long-term trend and crisis interruption:
  - About 30 percent of public expenditure programs are carried out at the subnational level in the EU15.
  - Comparable averages: OECD 32 percent in 2010; EU27 27 percent in 2011.
  - Expenditure decentralization is more advanced in federal states (Austria, Belgium, Germany, Spain).
  - Ireland: ratio of subnational to general government expenditure fell sharply from about 40 percent in 2004 to 10 percent in 2010 (one central government item of €31bn in 2010 noted as an exceptionally large bank recapitalization; excluding that item the ratio would have been 14 percent).
  - Excluding Ireland, about 4 percent of total expenditure has been redirected from the central to subnational governments since 1995.
  - Disaggregated functions: decentralization occurred across-the-board except for health and environmental protection.
  - Financing of increased subnational spending (2001–11): only one fifth of the change was funded from own revenues; the rest came from transfers and borrowing roughly in equal parts, resulting in an average increase in VFI.
  - Since the Great Recession onset, the decentralization trend has been hampered.
- Fact 2 — Highly-decentralized countries have larger public sectors:
  - Positive correlation between expenditure decentralization and government size in the EU15 (holds for all years from 1995 onwards and for the whole EU sample).
- Fact 3 — Subnational contribution to crisis deterioration:
  - On average, subnational overall balances deteriorated by half percentage point of GDP between 2004-07 and 2008-11.
  - National government deficits increased by almost 4 percent of GDP over the same period.
  - In most countries subnational deficits remained below 1 percent of GDP during the crisis.
  - Using “own balance” metric: national own surplus declined by about 3 percent of GDP between 2004-07 and 2008-11; subnational deficit increased by 1 percent over the same period. Based on this metric, subnational governments accounted for one-fourth of the increase in the general government deficit during the crisis (excluding Spain, this share is one-fifth).
  - “Own-spending-to-GDP” indicator: 22 percent of the increase in general government spending between 2007 and 2011 occurred at the subnational level.
- Fact 4 — Crisis exacerbated but did not generate subnational spending pressures:
  - Accounting decomposition over 2001–11 shows subnational expenditure expanded significantly in the early 2000s as well as in 2008–09; excluding Ireland, about half of the subnational government increase during the last decade occurred prior to the crisis.
  - On average, the subnational deficit stayed relatively stable before the crisis due to an increase in the early 2000s later offset by a reduction in 2006.

### Financing patterns of subnational spending (1995–2011)
- Higher subnational spending in the early 2000s was primarily financed from own revenues in the first half of the 2000s, while transfers played a more important role in 2006-07.
- In 2008-09 the subnational spending increase was financed from central transfers (as part of the fiscal stimulus), and higher deficit roughly in equal parts.
- During 2010-11 (a consolidation period), subnational governments did not receive additional transfers; their deficit still increased moderately, offsetting the contraction in own revenues.
- Results suggest expenditure pressures built up prior to the crisis when subnational budgets benefited from buoyant own revenue sources.
- During the crisis, local governments experienced a steep decline in own revenues, reflecting the combined effects of the economic cycle and the asset cycle.
- Local authorities increased fiscal deficits, likely because reversing past expenditure increases was not feasible or politically acceptable.

### Accounting/resource decomposition (changes in shares relative to earlier period — exact figures)
- 2010-11 (relative to 08-09)
  - EU15: Spending -0.05; Transfers received 0.00; Own Revenue -0.19; Deficit 0.14
  - Excluding Ireland: Spending 0.06; Transfers received 0.06; Own Revenue -0.16; Deficit 0.16
- 2008-09 (relative to 06-07)
  - EU15: Spending 0.57; Transfers received 0.44; Own Revenue -0.24; Deficit 0.37
  - Excluding Ireland: Spending 0.60; Transfers received 0.46; Own Revenue -0.24; Deficit 0.38
- 2006-07 (relative to 04-05)
  - EU15: Spending 0.04; Transfers received 0.09; Own Revenue 0.12; Deficit -0.16
  - Excluding Ireland: Spending 0.25; Transfers received 0.32; Own Revenue 0.12; Deficit -0.19
- 2004-05 (relative to 01-03)
  - EU15: Spending -0.04; Transfers received -0.20; Own Revenue 0.18; Deficit -0.01
  - Excluding Ireland: Spending 0.24; Transfers received 0.04; Own Revenue 0.19; Deficit 0.01
- 2001-03 (relative to 95-00)
  - EU15: Spending 0.45; Transfers received -0.01; Own Revenue 0.24; Deficit 0.22
  - Excluding Ireland: Spending 0.37; Transfers received -0.14; Own Revenue 0.28; Deficit 0.23

### Evidence of pre-crisis expenditure pressures
- Real growth in subnational spending exceeded potential growth between 2001 and 2007 in most countries.
- Excluding Ireland, the gap between subnational spending growth and potential growth amounted to 1 percentage point per year.
- Real growth in national spending was on par with potential growth over the same 2001–2007 period.
- Since 2008, the difference between national and subnational spending growth has faded away.
- Note: increases in subnational spending could also reflect devolution of new responsibilities.

### Econometric approach — research questions and model
- Empirical questions:
  - Do different forms of decentralization have distinct impacts on general government fiscal performance?
  - Does financing of decentralization matter?
  - Is decentralization more beneficial when accompanied by subnational fiscal rules?
  - Has decentralization resulted in expenditure overlap?
  - How prevalent are soft budget constraints at the subnational level?
- Empirical approach:
  - Estimate a dynamic fiscal reaction function over 1995-2011 following specifications by Bohn (1998), Debrun et al. (2008), Escolano et al. (2012).
  - Estimated equation (notation preserved in source): , 11 ε η μ δ β α itt i it it itit d X Dec PBPB ++++++= −−
  - PB is the general government primary balance to GDP; Dec is overall spending decentralization (own subnational spending as a ratio of general government spending); X denotes control variables; ηi country fixed effects; dt time dummies; εit error term.
  - Lags of decentralization are not included because the variable changes slowly and analysis targets an equilibrium relationship.

### Model specification and estimation details (Section II)
- Preferred specification: dynamic panel model estimated using the Least Square Dummy Variable estimator proposed by Bruno (2005).
- Dependent variable (baseline): general government primary balance (percent of GDP); alternative: general government structural primary balance.
- Baseline controls: general government debt to GDP ratio (debt) and output gap (gap).
- Bruno (2005) estimator corrects bias from dynamic panel fixed-effects; does not correct for possible endogeneity of decentralization.
- Large number of other controls tested but not significant (age dependency, openness, legislative strength, timing of elections).
- Sample period: primarily 1995-2011; some results for pre-crisis 1995-2007.

### Main econometric findings (exact reported estimates and interpretations)
- Baseline result:
  - Spending decentralization generally improves fiscal outcomes but effect is not large.
  - Increasing spending decentralization by 10 percentage points is associated with a 1¼ percent of GDP improvement in the general government primary balance.
  - Spending decentralization in the EU15 has increased on average by 4 percentage points since 1995 — implying limited expected fiscal gains from past decentralization trends.
- Robustness and caveats:
  - Positive effect disappears when dependent variable is the structural primary balance.
  - Positive effect disappears if sample excludes Ireland.
  - Significant persistence in primary balance (high lag coefficient reported).
  - No evidence of procyclicality in specification with structural primary balance: output gap coefficient not significant there.

### Functional decomposition of decentralization effects (exact coefficients as reported)
- Spending decentralization split into functions (health; education; social protection (socp); economic affairs (ea); environmental protection; defence; general public services; public order and safety; housing and community assistance; recreation, culture and religion).
- Estimated effects:
  - Only decentralization of social protection (socp) and economic affairs (ea) have a positive effect on fiscal performance; other functions not statistically significant.
  - Point estimates reported:
    - Decentralization: socp = 0.55**.
    - Decentralization: ea = 0.94**.

### Fragmentation and financing interactions (exact reported values)
- Fragmentation interaction:
  - Proxy: average number of municipalities per million inhabitants.
  - Interaction estimate reported as -0.00 (Table 2, column 14) — fragmentation does not alter impact of spending decentralization.
- VFI (Vertical Fiscal Imbalance):
  - VFI defined as share of subnational spending financed through net borrowing or transfers.
  - Over full sample 1995-2011: VFI coefficient not significant (Table 2, column 15).
  - Pre-crisis sample 1995-2007: VFI has a negative effect — a 10 percent increase in VFI is associated with a 0.4 percent of GDP decline in the primary balance (Table 2, column 16).

### Expenditure overlap between national and subnational spending (identification and results)
- Identification: regress changes in national spending on changes in subnational spending and subnational own revenue plus net borrowing to identify transfer-financed increases.
- Interpretation of β:
  - β = 0: no overlap (subnational increase fully substituted by national own spending decline).
  - β = 1: full overlap (national spending increases by same amount as subnational spending).
  - 0 < β < 1: partial overlap.
- Results:
  - Estimated β = 0.5 (reported as 0.51* with t-statistic 1.95) — partial overlap: a 1 percent of GDP increase in subnational spending results in a 0.5 percent of GDP increase in national spending.
- Function-specific overlap:
  - Overlap found in social protection, environmental protection, housing and community service, recreation and culture.
  - Overlap magnitudes reported:
    - Social protection overlap 0.85 (0.85***; t-statistic 3.27).
    - Housing & community assistance overlap 0.90 (0.90***; t-statistic 17.32).
  - These two categories account for about one-fourth of subnational spending in the EU15 on average.
- Economic-category overlap:
  - Wages: no evidence of overlap (∆Subnational wage spending coefficient -0.50*).
  - Capital spending: overlap estimated at 0.88 (0.88***; t-statistic 5.64).
- Robustness: results broadly unchanged when reestimating over pre-crisis period and when excluding Ireland and Spain.

### Soft Budget Constraints (SBC) — empirical strategy and findings (exact interpretations)
- Concept: SBCs arise when subnational governments can unduly influence access to funding (expectation of central government bailouts), leading to riskier fiscal policies.
- Two-stage empirical strategy:
  - Stage 1: estimate subnational revenue growth model to derive expected revenue and compute revenue shocks; preferred specification includes nominal GDP growth (growth), change in general government balance (GGbal), share of population older than 65 (Pop65), unemployment rate (ur).
  - Stage 2: regress subnational spending in nominal terms on positive and negative revenue shocks and controls (output gap (gap), change in subnational balance (∆Subbal), parliamentary election dummy (elec), inflation (inf)).
- Hypotheses:
  - If SBCs present: response to negative shocks (γ1) is zero or small (no downward adjustment).
  - Without SBCs: γ1 positive (spending decreases in response to negative shocks), possibly partially (<1).
  - Asymmetry (γ1 ≠ γ2) indicates SBCs.
- Empirical findings:
  - Full period 1995-2011: evidence of SBCs — coefficient of negative revenue shock not statistically significant; subnational governments do not immediately cut spending when revenues decline unexpectedly.
  - Positive revenue shocks lead to spending increases, but not by the full amount (about one fourth).
  - Pre-crisis period 1995-2007: no evidence of SBCs; coefficients for positive and negative shocks not statistically different — suggests SBCs and bailout expectations increased during recent crisis.
  - Subnational spending found to be procyclical over period (positive coefficient of GDP growth).
- Role of fiscal rules:
  - Split sample by strength of general government and subnational fiscal rules (above/below median) yields counter-intuitive result: in countries with strong rules, spending does not adjust to negative fiscal shocks, while in countries with weaker rules it does — suggesting SBCs more prevalent where rules are stronger.
  - Possible explanation: fiscal rules may have been introduced where SBCs were already prevalent; rule strength indicator may not capture enforcement effectiveness.
- Caveats:
  - Aggregated subnational statistics mask heterogeneity across units.
  - Mandates and legal constraints may limit subnational ability to cut spending.
  - Reaction to shocks depends on source and persistence.
  - Fiscal rule indicators may not capture enforcement or effectiveness.

### Policy recommendations (preserved language and emphasis)
- Match subnational own resources with responsibilities:
  - Avoid solving adverse incentives by simply cutting grants (risks: arrears, excessive subnational borrowing, unfunded mandates).
  - Increase subnational own revenue where feasible; suggested options include property taxes or PIT surcharges, noting practical difficulties (tax base mobility, administrative costs, horizontal disparities).
  - Improve transfer systems (performance-based transfers; allocate based on expenditure needs rather than actual costs).
- Better define spending assignments:
  - Clarify responsibilities to limit overlap and enhance accountability.
  - Consider recentralization of functions where agency problems, negative externalities, or economies of scale losses are pronounced (noting trend toward recentralization of health in several countries).
  - Address economies of scale via municipal mergers or inter-municipal associations.
- Introduce expenditure rules:
  - Subnational expenditure rules may be promising given procyclicality of subnational spending.
  - Subnational expenditure rules largely non-existent across EU15; effectiveness requires monitoring and credible sanctions.
  - Rules alone are ineffective if unfunded mandates persist.
- Ensure sound local public financial management practices:
  - Capacity building for realistic budgeting, audit and control, fiscal risk disclosure, transparency and reporting.
  - Consider performance budgeting at subnational level.
- Strengthen local accountability:
  - Enhance external auditors, representative local assemblies, public interest bodies, and civil society oversight to improve cost-effective service delivery.

### Key quantitative highlights (values preserved exactly)
- Increase in spending decentralization by 10 percentage points → 1¼ percent of GDP improvement in general government primary balance.
- Average increase in spending decentralization in EU15 since 1995 → 4 percentage points.
- Estimated overlap coefficient β (overall) ≈ 0.5 (reported as 0.51*).
- Function-specific overlap estimates:
  - Social protection overlap ≈ 0.85 (0.85***).
  - Housing & community assistance overlap ≈ 0.9 (0.90***).
- Capital spending overlap ≈ 0.88 (0.88***).
- VFI pre-crisis effect: a 10 percent increase in VFI associated with a 0.4 percent of GDP decline in the primary balance.
- Subnational spending response to positive revenue shocks ≈ one fourth increase.
- Interaction term fragmentation x expenditure decentralization reported as -0.00.
- Bruno (2005) estimator referenced for bias correction in dynamic panel with fixed effects.

### Appendix I — Data sources and definitions (exact items preserved)
- Fiscal data source: Eurostat covering the period 1995-2011.
- Sample: unbalanced panel including the EU15 countries (Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden and United Kingdom).
- Average observations per country: about 17 observations per country on average.
- Subnational measures: aggregating regional and local subsectors (S1312 and S1313 in ESA95).
- “National” level: consolidated central government and social security funds (S1311 and S1314).
- Variable definitions and sources (selected):
  - Expenditure decentralization: Share of subnational own expenditure in total general government expenditure. Source: Eurostat.
  - VFI: Share of subnational own expenditure not financed with subnational own revenue. Source: Eurostat.
  - General government primary balance: general government revenue minus general government expenditure (excluding net interest). Source: IMF WEO database.
  - Output gap: percentage difference between actual GDP in constant prices and estimated potential GDP. Source: IMF WEO database.
  - Government fragmentation: Number of municipalities per million of inhabitants. Source: Dexia.
  - General Government Fiscal Rule: Fiscal Rule Strength Index. Source: European Commission (2012b).
  - Subnational Government Fiscal Rule: Composite Indicator from IMF Staff using EC (2012).
  - 65+ population: Population above 65 as a percentage of the total population. Source: World Bank.
  - GDP deflator: IMF WEO database.
- Note: All subnational measures exclude transfers received from other general government units when indicated.

### Appendix II — Simple model of Soft Budget Constraints (SBC)
- Behavioral priors:
  - Consumption smoothing behavior: anticipated vs unanticipated revenue changes should be asymmetric under permanent income/consumption smoothing.
  - Bailout expectations: if bailouts expected, negative revenue changes lead to little or no downward spending adjustment.
- Typology:
  - Type 1: prudent — no bailout expectations, smoothes consumption.
  - Type 2: imprudent — bailout expectations, does not smooth consumption.
  - Type 3: constrained — no bailout expectations, does not smooth consumption.
- Testable implications:
  - Response to negative revenue shocks discriminates type 2 (imprudent) from types 1/3.
  - Type 2 shows asymmetric responses to positive and negative shocks; types 1 and 3 react similarly (increase with positive shocks, decrease with negative shocks).
- Caveat: exceptional circumstances could justify ad hoc transfers without implying SBCs.

*Source: IMF Working Paper — Box 1: Why Fiscal Decentralization May Inflate the Size of the Government; Sections II, IV, Appendix I and Appendix II from _wp1346 (1995–2011 Eurostat/IMF staff analysis).*

### Box 1: Why Fiscal Decentralization May Inflate the Size of the Government

### Box 1: Why Fiscal Decentralization May Inflate the Size of the Government

### Four principal channels by which decentralization may increase public sector size
- Decentralization does not generate unlimited efficiency gains:
  - Diseconomies of scale, negative inter-jurisdictional spillovers, and coordination issues can offset gains.
  - Macroeconomic stability and redistribution responsibilities are often argued to be best left at the center.
  - Resource-allocation guidance: central government should provide national public goods; subnational governments should deliver services with local benefits (examples: waste disposal, street maintenance, primary education).
- Normative considerations seldom guide actual expenditure assignments:
  - Decentralization is largely driven by political motives, historical and cultural legacies.
  - Constitutional boundaries and the existing “political map” limit the scope for adjusting local entities’ size according to efficiency criteria.
- Fiscal discipline is harder to enforce at the subnational level:
  - Soft budget constraint: local authorities may overspend if not facing a fixed resource envelope—bailouts from the center; subsidized loans from public banks or state-owned enterprises; supplier arrears; underfunded public sector pensions.
  - Common pool problem: local policymakers may not internalize the cost of spending when financed with intergovernmental transfers or shared revenue funded by other jurisdictions’ taxpayers.
  - Moral hazard and weak governance: devolved tax bases and spending responsibilities can outpace central monitoring; local bureaucracies may be lower quality; clientelism and corruption may be more prevalent.
- Expenditure-control failures reflect structural local budget management difficulties:
  - Procyclicality of resources: subnational own revenue sources are narrow and volatile; transfers from the center are often procyclical, producing procyclical local spending when access to credit or escape clauses are limited.
  - Unclear spending assignments: concurrent responsibilities across levels (e.g., center sets health/education standards and finances; lower levels provide services) weaken accountability and encourage blame-shifting.
  - Weak public financial management (PFM) systems: lack of audit and control mechanisms; looser accounting standards (notably for arrears); absence of multi-year fiscal frameworks; ill-designed fiscal rules.
  - Flaws in transfer system design: matching grants and inadequate treatment of subnational heterogeneity can encourage overspending; transfers often allocated based on actual spending costs rather than “expenditure needs” and independent of service quality.

### Fragmentation, empirical evidence, and the role of vertical fiscal imbalances (VFI)
- Government fragmentation:
  - Literature on fragmentation and public spending (notably U.S. studies) provides limited support that fragmented governments are smaller and spend less; effects differ between “general” and “special-purpose” subnational governments.
- Empirical cross-country findings:
  - Decentralization per se: no strong evidence that fiscal decentralization by itself increases government size. Earlier literature finds no relation; more recent work shows opposite effects of spending and revenue decentralizations.
  - Vertical fiscal imbalance (VFI): consensus on a “conditional” effect—spending decentralization raises general government spending when financed by transfers or borrowing (i.e., associated with large VFI). Expenditure decentralization without corresponding local tax powers does not generate tax competition or increase local accountability.
  - Evidence cited: negative effect of VFI and its interaction with spending decentralization found in Jin and Zou (2002), Rodden (2003), Eyraud and Lusinyan (2011); Fornasari et al. (2000) show subnational spending not funded by local taxes is additional to central spending.

### Stylized facts on expenditure decentralization in Europe (key findings and exact figures)
- Fact 1 — Long-term trend and crisis interruption:
  - About 30 percent of public expenditure programs are carried out at the subnational level in the EU15.
  - Comparable averages: OECD 32 percent in 2010; EU27 27 percent in 2011.
  - Expenditure decentralization is more advanced in federal states (Austria, Belgium, Germany, Spain).
  - Ireland: ratio of subnational to general government expenditure fell sharply from about 40 percent in 2004 to 10 percent in 2010 (one central government item of €31bn in 2010 noted as an exceptionally large bank recapitalization; excluding that item the ratio would have been 14 percent).
  - Excluding Ireland, about 4 percent of total expenditure has been redirected from the central to subnational governments since 1995.
  - Disaggregated functions: decentralization occurred across-the-board except for health and environmental protection.
  - Financing of increased subnational spending (2001–11): only one fifth of the change was funded from own revenues; the rest came from transfers and borrowing roughly in equal parts, resulting in an average increase in VFI.
  - Since the Great Recession onset, the decentralization trend has been hampered; possible reasons: exhaustion of optimal decentralization for some functions, rise in counter-cyclical central expenditure (2008–09), and subnational share of consolidation efforts.
- Fact 2 — Highly-decentralized countries have larger public sectors:
  - Positive correlation between expenditure decentralization and government size in the EU15 (holds for all years from 1995 onwards and for the whole EU sample).
  - Explanations: expenditure overlaps and duplication; less prudent subnational fiscal policies leading to overspending; decentralization revealing voter preferences for higher taxes and services; caution advised due to reverse causality and omitted variables.
- Fact 3 — Subnational contribution to crisis deterioration:
  - On average, subnational overall balances deteriorated by half percentage point of GDP between 2004-07 and 2008-11.
  - National government deficits increased by almost 4 percent of GDP over the same period.
  - In most countries subnational deficits remained below 1 percent of GDP during the crisis.
  - Using “own balance” metric: national own surplus declined by about 3 percent of GDP between 2004-07 and 2008-11; subnational deficit increased by 1 percent over the same period. Based on this metric, subnational governments accounted for one-fourth of the increase in the general government deficit during the crisis (excluding Spain, this share is one-fifth).
  - “Own-spending-to-GDP” indicator: 22 percent of the increase in general government spending between 2007 and 2011 occurred at the subnational level.
- Fact 4 — Crisis exacerbated but did not generate subnational spending pressures:
  - Accounting decomposition over 2001–11 shows subnational expenditure expanded significantly in the early 2000s as well as in 2008–09; excluding Ireland, about half of the subnational government increase during the last decade occurred prior to the crisis.
  - On average, the subnational deficit stayed relatively stable before the crisis due to an increase in the early 2000s later offset by a reduction in 2006.

*Source: IMF Working Paper — Box 1: Why Fiscal Decentralization May Inflate the Size of the Government*

### 2007. Higher spending was primarily financed from own revenues in the first half of the

### _wp1346 - 2007. Higher spending was primarily financed from own revenues in the first half of the

### Financing patterns of subnational spending (1995–2011)
- Higher subnational spending in the early 2000s was primarily financed from own revenues in the first half of the 2000s, while transfers played a more important role in 2006-07.
- In 2008-09 (the first two years of the crisis), the subnational spending increase was financed from central transfers (as part of the fiscal stimulus), and higher deficit roughly in equal parts.
- During 2010-11 (a consolidation period), subnational governments did not receive additional transfers; their deficit still increased moderately, offsetting the contraction in own revenues.
- The results suggest expenditure pressures built up prior to the crisis when subnational budgets benefited from buoyant own revenue sources.
- During the crisis, local governments experienced a steep decline in own revenues, reflecting the combined effects of the economic cycle (automatic stabilizers) and the asset cycle (housing market collapse in Spain, for instance).
- Local authorities increased fiscal deficits, likely because reversing past expenditure increases was not feasible or politically acceptable.

### Accounting decomposition (notation and interpretation)
- The accounting decomposition reported is: ሻܲܦܩܲ ݃݊݅݀݊݁݌ݏ ݈ܽ݊݋݅ݐܾܽ݊ݑݏ∆ሺ
⁄
ሻܲܦܩܲ ݀݁ݒ݅݁ܿ݁ݎ ݏݎ݂݁ݏ݊ܽݎܐൌ∆ሺ
⁄
൅
ሻܲܦܩܲ ݁ݑ݊݁ݒ݁ݎ ݊ݓ݋∆ሺ 
⁄
ሻܲܦܩܲ ݐ݂݅ܿ݅݁݀ ݈ܽ݊݋݅ݐܾܽ݊ݑݏ൅∆ሺ
⁄
, with PDGP denoting potential GDP.
- Note: Using nominal GDP ratios would distort comparisons between pre-crisis and crisis periods because the share of subnational expenditure in nominal GDP soared during the fiscal stimulus period partly due to GDP’s cyclical decline.

### Resource decomposition by period (changes in shares relative to earlier period)
- 2010-11 (relative to 08-09)
  - EU15: Spending -0.05; Transfers received 0.00; Own Revenue -0.19; Deficit 0.14
  - Excluding Ireland: Spending 0.06; Transfers received 0.06; Own Revenue -0.16; Deficit 0.16
- 2008-09 (relative to 06-07)
  - EU15: Spending 0.57; Transfers received 0.44; Own Revenue -0.24; Deficit 0.37
  - Excluding Ireland: Spending 0.60; Transfers received 0.46; Own Revenue -0.24; Deficit 0.38
- 2006-07 (relative to 04-05)
  - EU15: Spending 0.04; Transfers received 0.09; Own Revenue 0.12; Deficit -0.16
  - Excluding Ireland: Spending 0.25; Transfers received 0.32; Own Revenue 0.12; Deficit -0.19
- 2004-05 (relative to 01-03)
  - EU15: Spending -0.04; Transfers received -0.20; Own Revenue 0.18; Deficit -0.01
  - Excluding Ireland: Spending 0.24; Transfers received 0.04; Own Revenue 0.19; Deficit 0.01
- 2001-03 (relative to 95-00)
  - EU15: Spending 0.45; Transfers received -0.01; Own Revenue 0.24; Deficit 0.22
  - Excluding Ireland: Spending 0.37; Transfers received -0.14; Own Revenue 0.28; Deficit 0.23

### Evidence of pre-crisis expenditure pressures (Figure 4 summary)
- Real growth in subnational spending exceeded potential growth between 2001 and 2007 in most countries.
- Excluding Ireland, the gap between subnational spending growth and potential growth amounted to 1 percentage point per year.
- Real growth in national spending was on par with potential growth over the same 2001–2007 period.
- Since 2008, the difference between national and subnational spending growth has faded away.
- Note: Conclusions are tentative because increases in subnational spending could also reflect devolution of new responsibilities.

### IV. Econometric evidence — research questions and approach
- Three empirical questions explored:
  - Have all forms of decentralization the same impact on the general government’s fiscal performance? Specifically:
    - Does the effect differ according to the type of spending function being decentralized?
    - Does the financing of decentralization matter?
    - Is decentralization more beneficial to fiscal outcomes when accompanied by subnational fiscal rules?
  - Has decentralization resulted in expenditure overlap (duplication and possible waste)?
  - How prevalent are soft budget constraints at the subnational level (moral hazard and bailout expectations)?
- Empirical approach:
  - Estimate a dynamic fiscal reaction function over the period 1995-2011, following specifications adopted by Bohn (1998), Debrun et al. (2008), and Escolano et al. (2012).
  - The estimated equation is:
    , 11 ε η μ δ β α itt i it it itit d X Dec PBPB ++++++= −−
  - Where:
    - indices i, t denote countries and years;
    - PB is the general government primary balance to GDP;
    - Dec is overall spending decentralization (own subnational spending as a ratio of general government spending);
    - X denotes a vector of control variables;
    - ηi represents country-specific fixed effects;
    - dt are time dummies;
    - εit is a time- and country-specific error term.
  - Because the decentralization variable does not change rapidly over time and the analysis targets an equilibrium relationship, lags of decentralization are not included.

*Source: IMF staff analysis and Eurostat data as presented in the source content.*

### Section II). A positive (negative) value for the estimated coefficient

### _wp1346 - Section II). A positive (negative) value for the estimated coefficient

### Model specification and estimation
- Preferred specification: dynamic panel model estimated using the Least Square Dummy Variable estimator proposed by Bruno (2005).  
- Dependent variable (baseline): general government primary balance (percent of GDP); alternative dependent variable: general government structural primary balance.  
- Control variables in baseline: general government debt to GDP ratio (debt) and output gap (gap).  
- Estimation notes:
  - Bruno (2005) estimator used to correct bias from dynamic panel fixed-effects; does not correct for possible endogeneity of the decentralization variable.
  - Large number of other controls tested but not significant (age dependency, openness, legislative strength, timing of elections).
  - Sample period: primarily 1995-2011; some results reported for pre-crisis 1995-2007.

### Main findings on decentralization and fiscal performance
- Baseline result:
  - Spending decentralization generally improves fiscal outcomes but effect is not large (Table 2, column 1).
  - Increasing spending decentralization by 10 percentage points is associated with a 1¼ percent of GDP improvement in the general government primary balance.
  - Spending decentralization in the EU15 has increased on average by 4 percentage points since 1995 (Section III) — implying limited expected fiscal gains from past decentralization trends.
- Robustness and caveats:
  - Positive effect disappears when dependent variable is the structural primary balance (Table 2, column 2).
  - Positive effect disappears if sample excludes Ireland (Table 2, column 3).
  - Significant persistence in primary balance (high lag coefficient reported in table notes).
  - No evidence of procyclicality in specification with structural primary balance: output gap coefficient not significant there.

### Functional decomposition of decentralization effects
- Spending decentralization split into 8 functions: health (h); education (educ); social protection (socp); economic affairs (ea); environmental protection (envip); defence (def); general public services (gps); public order and safety (poas); housing and community assistance (haca); recreation, culture and religion (rcar).
- Estimated effects:
  - Only decentralization of social protection (socp) and economic affairs (ea) have a positive effect on fiscal performance; other functions have no statistically significant effect (Table 2, columns 4–13).
  - Specific point estimates (Table 2 excerpts):
    - Decentralization: socp = 0.55** (z-statistic reported in table).
    - Decentralization: ea = 0.94**.
  - Non-significant or mixed results for health, education, environmental protection, defence, general public services, public order & safety, housing & community assistance, recreation/culture/religion.

### Fragmentation and financing interactions
- Fragmentation:
  - Interaction tested between expenditure decentralization and fragmentation (proxy: average number of municipalities per million inhabitants).
  - Expectation: interaction coefficient negative if fragmentation aggravates decentralization’s fiscal effect.
  - Result: fragmentation does not alter impact of spending decentralization (Table 2, column 14; interaction estimate reported as -0.00).
  - Caveat: proxy may not capture vertical and horizontal fragmentation fully; indicator is time-invariant.
- Vertical Fiscal Imbalance (VFI) and financing:
  - VFI defined as share of subnational spending financed through net borrowing or transfers.
  - Over full sample 1995-2011: VFI coefficient not significant (Table 2, column 15) — possible bias from crisis period reverse causality.
  - Pre-crisis sample 1995-2007: VFI has a negative effect on fiscal performance — a 10 percent increase in VFI is associated with a 0.4 percent of GDP decline in the primary balance (Table 2, column 16).

### Expenditure overlap between national and subnational spending
- Identification strategy:
  - Estimate correlation between changes in national spending (Nate) and changes in subnational spending (Sube) using equation (2); include subnational own revenue plus net borrowing (RB) to identify transfer-financed increases.
  - Interpretation of β:
    - β = 0: no overlap (subnational increase fully substituted by national own spending decline).
    - β = 1: full overlap (national spending increases by same amount as subnational spending).
    - 0 < β < 1: partial overlap.
- Results (Table 3, column 1):
  - Estimated β = 0.5 (coefficient reported as 0.51* with t-statistic 1.95 in table), interpreted as partial overlap: a 1 percent of GDP increase in subnational spending results in a 0.5 percent of GDP increase in national spending.
- Function-specific overlap (Table 3, columns 2–11):
  - Expenditure overlap found in four functions: social protection, environmental protection, housing and community service, recreation and culture.
  - Overlap magnitudes reported:
    - Social protection overlap ≈ 0.85 (reported 0.85***; t-statistic 3.27).
    - Housing & community assistance overlap ≈ 0.9 (reported 0.90***; t-statistic 17.32).
  - These two categories account for about one-fourth of subnational spending in the EU15 on average.
- Economic-category overlap (Table 3, columns 12–13):
  - Wages: no evidence of overlap (∆Subnational wage spending coefficient reported -0.50*).
  - Capital spending: sizable overlap estimated at 0.88*** (coefficient 0.88***; t-statistic 5.64).
- Robustness: results broadly unchanged when reestimating over pre-crisis period and when excluding Ireland and Spain.

### Soft Budget Constraints (SBC) at subnational level
- Concept: SBCs arise when subnational governments can unduly influence access to funding (expectation of central government bailouts), leading to riskier fiscal policies (Kornai, 1979; Rodden et al., 2003).
- Empirical strategy:
  - Two-stage approach:
    - Stage 1: estimate subnational revenue growth model to derive expected revenue and compute revenue shocks (equation (3)); preferred specification includes nominal GDP growth (growth), change in general government balance (GGbal), share of population older than 65 (Pop65), unemployment rate (ur).
    - Stage 2: regress subnational spending in nominal terms on positive and negative revenue shocks and controls (equation (4)); controls include output gap (gap), change in subnational balance (∆Subbal), parliamentary election dummy (elec), inflation (inf).
  - Distinguish expected vs unexpected revenue changes; focus on unexpected shocks for SBC identification.
- Hypotheses and interpretation:
  - If SBCs present: response to negative shocks (γ1) is zero or small (no downward adjustment).
  - In absence of SBCs: γ1 positive (spending decreases in response to negative shocks), possibly partially (<1).
  - Asymmetry test: γ1 ≠ γ2 indicates SBCs.
- Empirical findings (Table 4 and discussion):
  - Full period 1995-2011: evidence of SBCs — coefficient of negative revenue shock not statistically significant; subnational governments do not immediately cut spending when revenues decline unexpectedly.
  - Positive revenue shocks lead to spending increases, but not by the full amount (about one fourth reported in text).
  - Pre-crisis period 1995-2007: no evidence of SBCs; coefficients for positive and negative shocks not statistically different — suggests SBCs and bailout expectations increased during recent crisis.
  - Subnational spending found to be procyclical over period (positive coefficient of GDP growth).
- Role of fiscal rules:
  - Split sample by strength of general government and subnational fiscal rules (above/below median).
  - Counter-intuitive result: in countries with strong general government or subnational rules, spending does not adjust to negative fiscal shocks, while in countries with weaker rules it does — suggesting SBCs more prevalent where rules are stronger.
  - Possible explanation: fiscal rules may have been introduced where SBCs were already prevalent; rule strength indicator may not capture enforcement effectiveness.
- Caveats and limitations:
  - Aggregated subnational statistics mask heterogeneity across subnational units.
  - Standards and mandates may constrain subnational capacity to cut spending legitimately.
  - Reaction to shocks may depend on source and persistence of shocks; small or transitory shocks may not prompt adjustment even under hard constraints.
  - Fiscal rule indicators may not capture enforcement or effectiveness.
  - Data constraints prevent further sample splits (transfer dependency, history of bailouts, assignment of sensitive functions).

### Policy recommendations (from concluding remarks)
- Match subnational own resources with responsibilities:
  - Avoid solving adverse incentives by simply cutting grants (risks: arrears, excessive subnational borrowing, unfunded mandates).
  - Increase subnational own revenue where feasible; suggested options include property taxes or PIT surcharges, noting practical difficulties (tax base mobility, administrative costs, horizontal disparities).
  - Improve transfer systems (performance-based transfers; allocate based on expenditure needs rather than actual costs).
- Better define spending assignments:
  - Clarify responsibilities to limit overlap and enhance accountability.
  - Consider recentralization of functions where agency problems, negative externalities, or economies of scale losses are pronounced (noting trend toward recentralization of health in several countries).
  - Address economies of scale via municipal mergers or inter-municipal associations.
- Introduce expenditure rules:
  - Subnational expenditure rules may be promising given procyclicality of subnational spending.
  - Subnational expenditure rules largely non-existent across EU15; effectiveness requires monitoring and credible sanctions.
  - Rules alone are ineffective if unfunded mandates persist.
- Ensure sound local public financial management practices:
  - Capacity building for realistic budgeting, audit and control, fiscal risk disclosure, transparency and reporting.
  - Consider performance budgeting at subnational level.
- Strengthen local accountability:
  - Enhance external auditors, representative local assemblies, public interest bodies, and civil society oversight to improve cost-effective service delivery.

### Key quantitative highlights (preserved exactly as reported)
- Increase in spending decentralization by 10 percentage points → 1¼ percent of GDP improvement in general government primary balance.
- Average increase in spending decentralization in EU15 since 1995 → 4 percentage points.
- Estimated overlap coefficient β (overall) ≈ 0.5 (reported as 0.51* in Table 3).
- Function-specific overlap estimates:
  - Social protection overlap ≈ 0.85 (0.85***).
  - Housing & community assistance overlap ≈ 0.9 (0.90***).
- Capital spending overlap ≈ 0.88 (0.88***).
- VFI pre-crisis effect: a 10 percent increase in VFI associated with a 0.4 percent of GDP decline in the primary balance.
- Subnational spending response to positive revenue shocks ≈ one fourth increase (text description: “about one fourth”).
- Interaction term fragmentation x expenditure decentralization reported as -0.00 (Table 2, column 14).
- Bruno (2005) estimator referenced for bias correction in dynamic panel with fixed effects.

*Source: _wp1346 - Section II). A positive (negative) value for the estimated coefficient (IMF working paper content provided).*

### Appendix I. Data Sources and Definitions

### _wp1346 - Appendix I. Data Sources and Definitions

### Dataset coverage and structure
- Fiscal data source: Eurostat covering the period 1995-2011.
- Sample composition: unbalanced panel including the EU15 countries (Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden and United Kingdom).
- Average observations per country: about 17 observations per country on average.
- Subnational measures: calculated by aggregating the regional and local subsectors (S1312 and S1313 in ESA95).
- “National” level: refers to the consolidated central government and social security funds (S1311 and S1314).

### Variable definitions and data sources
- Expenditure decentralization
  - Definition: Share of subnational own expenditure in total general government expenditure.
  - Source: Eurostat
- Expenditure decentralization per function
  - Definition: Same ratio but for specific function (COFOG classification)
  - Source: Eurostat
- National (resp. subnational) Government Size
  - Definition: Ratio of national (resp. subnational) government expenditure to GDP
  - Source: Eurostat
- Government Debt
  - Definition: General government gross debt
  - Source: Eurostat
- Output gap
  - Definition: Percentage difference between actual GDP in constant prices and estimated potential GDP.
  - Source: IMF WEO database
- General Government Fiscal Rule
  - Definition: Fiscal Rule Strength Index
  - Source: European Commission (2012b)
- Subnational Government Fiscal Rule
  - Definition: Composite Indicator aggregating the subnational rule strength indexes using coverage data for weights
  - Source: IMF Staff using EC (2012)
- Government fragmentation
  - Definition: Number of municipalities per million of inhabitants
  - Source: Dexia
- National (resp. subnational) balance
  - Definition: Total revenue minus total expenditure of the national (resp. subnational) government
  - Source: Eurostat
- National (resp. subnational) own balance
  - Definition: Total revenue (excluding transfers received from other levels of government) minus total expenditure (minus transfers paid to other levels of government) of the national (resp. subnational) government
  - Source: Eurostat
- VFI (vertical fiscal imbalance)
  - Definition: Share of subnational own expenditure (i.e., excluding transfers paid to other general government units) not financed with subnational own revenue (i.e., excluding transfers received from other general government units). Subnational government is a consolidated state (when applicable) and local government. Transfers include both current and capital transfers.
  - Source: Eurostat
- General government primary balance
  - Definition: General government revenue minus general government expenditure (excluding net interest)
  - Source: IMF WEO database
- 65+ population
  - Definition: Population above 65 as a percentage of the total population.
  - Source: World Bank
- GDP deflator
  - Definition: GDP deflator
  - Source: IMF WEO database
- GDP
  - Definition: GDP, constant prices.
  - Source: Eurostat

### Notes
- Appendix pagination: page numbers indicated in source (30).
- All subnational measures exclude transfers received from other general government units when indicated by the variable definition (e.g., VFI, own balances).

---

### Appendix II. A Simple Model of Soft Budget Constraints

### Model setup and behavioral priors
- Purpose: Use the response of subnational governments to revenue changes to identify the existence of soft budget constraints (SBCs).
- Two key behavioral features assumed for subnational governments:
  - Consumption smoothing behavior:
    - (i) If spending follows the permanent income/consumption smoothing model, responses to anticipated versus unanticipated revenue changes should be asymmetric (Holtz-Eakin and Rosen, 1989; Holtz-Eakin et al., 1993). Anticipated changes should not affect spending (financial markets accommodate expected fluctuations); unanticipated revenue shocks affect spending if they revise permanent income.
    - (ii) If governments do not smooth consumption (credit-constrained, subject to budget balance rule, or myopic), they should respond similarly to unanticipated and anticipated revenue changes.
  - Bailout expectations:
    - If governments expect bailouts, responses to positive and negative revenue changes should be asymmetric. For negative revenue change, spending is not adjusted downward (or is adjusted less) compared with no-bailout expectation.

### Typology of subnational governments
- Type 1: prudent subnational government — does not have bailout expectations, and smoothes consumption.
- Type 2: imprudent government — has bailout expectations, and does not smooth consumption.
- Type 3: constrained type — does not have bailout expectations, but does not smooth consumption (e.g., no access to financial markets).
- Note: The fourth theoretical type (bailout expectations and expenditure smoothing) is not examined.

### Testable implications and identification strategy
- Objective: Discriminate between subnational governments with or without bailout expectations, i.e., between type 2 (imprudent) and types 1/3 (prudent and constrained).
- Main implications of the model:
  - First: Anticipated revenue changes are not informative to discriminate between type 2 and types 1/3. Both types 2 and 3 raise expenditure in response to positive anticipated revenue changes, while types 1 and 2 may have similar responses to negative anticipated revenue changes.
  - Second: The response to negative revenue shocks provides sufficient information to reveal “imprudent” governments. In the face of negative shocks, subnational governments without bailout expectations reduce spending (regardless of type 1 or 3), while expenditure of type 2 does not adjust (or does so by less).
  - Third: Type 2 has asymmetric responses to positive and negative shocks, while types 1 and 3 should react in a similar way (although in opposite directions: increase spending in response to positive shocks and decrease spending in response to negative shocks).
- Econometric focus: revenue shocks are used to identify whether European subnational governments have behaved, on average, as type 2, consistent with the existence of SBCs.

### Caveats and limitations
- Exceptional circumstances (e.g., natural disaster, terrorist attack) could lead prudent governments to legitimately expect ad hoc transfers without implying soft budget constraints.

### Appendix Table 1: Subnational Expenditure Response (qualitative summary)
- Type 1 (unexpected revenue change)
  - positive: +
  - negative: -
- Type 2 (unexpected revenue change)
  - positive: +++
  - negative: 0-
- Type 3 (unexpected revenue change)
  - positive: +
  - negative: -
- Expected revenue change responses (qualitative)
  - Type 1: positive 0; negative 0
  - Type 2: positive ++; negative 0
  - Type 3: positive ++; negative 0

---

*Source: _wp1346 - Appendix I. Data Sources and Definitions (PDF content as provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1346.pdf_