## _wp16141

## Source details

**Canonical URL:** [_wp16141](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16141.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16141.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16141.pdf.json)

---

### Introduction and research focus
- Purpose: analyse the role of net fiscal transfers in mitigating regional shocks on subnational public finances in federations, focusing on risk-sharing (idiosyncratic shocks) and stabilization (common shocks).
- Primary empirical focus: the extent of risk sharing and whether a central budget (or equivalent mechanisms) provides better insurance against macroeconomic shocks.
- Sample restricted to three large fiscal federations: Australia, Canada, and the U.S.
- Definition: "net fiscal transfers" = gross transfers from center to sub-national governments (to budgets or to residents) minus taxes paid by sub-national residents to the center. "Sub-national" = state/provincial governments.

### Theoretical background
- Pro-insurance arguments for a central budget:
  - Pooling smooths consumption against idiosyncratic shocks when state-specific exchange rates are absent, prices are sticky, input mobility limited, or private credit markets dysfunctional.
  - Scale economies: central governments can borrow on better terms (example: U.S. states pay a risk-premium over Treasury bond yields).
  - Fiscal spillovers create public-good stabilization rationale and can mitigate free-riding, especially when monetary policy constrained.
  - Ricardian equivalence: centralized non-liability-creating stabilization may be more effective than decentralized debt-financed stabilization (example: Bayoumi and Masson (1998) — in Canada non-liability creating stabilization two to three times more effective than decentralized stabilization financed through local public debt).
- Anti-insurance arguments:
  - Design difficulty: efficient tax-and-transfer design may create systematic redistribution from rich to poor regions; empirical patterns show richer regions pay higher federal taxes and receive lower transfers on average.
  - Moral hazard: implicit bailout guarantees can worsen subnational policy incentives and fiscal positions.
- Net theoretical conclusion: mixed; empirical question is how much central budgets cushion regional shocks in practice.

### Methodological contributions
- Reviews prior empirical approaches and proposes refinements:
  - Standard two-step approach (separate estimation of redistribution and risk-sharing) and its limitations.
  - New one-step PMG cointegration approach (Pooled Mean Group estimator of Pesaran et al., 1999) that combines level and first-difference regressions for efficiency and dynamic capture.
  - Alternative component-level approach to separate contributions of TRANS, TRANSDIR, TAX to risk-sharing and redistribution.
- PMG empirical specification:
  - Δ(YD_it / Y̅_t ) = φ_i [ (YD_{it−1} / Y̅_{t−1}) − α − β (Y_{it−1} / Y̅_{t−1}) ] + δ_i Δ(Y_{it} / Y̅_{t}) + μ_i + ε_{it}
  - Interpretation:
    - β measures redistribution (long-run response).
    - δ (and δ_i) measures risk-sharing (short-run response).
    - φ (φ_i) is speed of adjustment to long-run equilibrium.
    - μ_i: state fixed effects.

### Data and sample periods
- Federations: Australia, Canada, USA.
- Transfer components:
  - TRANS = gross transfers to state budgets.
  - TRANSDIR = gross transfers to individuals.
  - TAX = federal taxes.
  - T = TRANS + TRANSDIR − TAX (net fiscal transfers).
- Sample periods:
  - Australia and the U.S.: 1998 - 2010 (13 years).
  - Canada: 1992 – 2009 (18 years).
- Fiscal and macro sources include national statistical agencies, Department of Finance (Canada), BEA, IRS; Annex 2 lists series coverage (examples: Australia GDP 1990–2011; USA GDP 1997–2010; TRANS Australia 1999–2011; TRANSDIR USA 1998–2010; TAX USA 1998–2010).

### Descriptive statistics (selected)
- Federal tax revenues: range between 15-20 percent of state GDP.
- Gross fiscal transfers: range between 5-15 percent of state GDP.
- Median state-level net fiscal transfers (country samples):
  - U.S. (1998-2010): West Virginia — slightly less than 20 percent of state GDP; Minnesota — about 15 percent of state GDP.
  - Canada (1992-2009): Newfoundland and Labrador — 10 percent of state GDP; Ontario — -10 percent of state GDP.
  - Australia (1998-2010): Northern Territory — 15 percent of state GDP; Australian Capital Territory — -5 percent of state GDP.
- Post-crisis patterns: U.S. federal tax revenues declined markedly and outlays increased significantly in 2009-2010; Canada and Australia saw sharp tax declines after the crisis with spending somewhat less responsive (except Australia in 2009). Net transfers increased after the global financial crisis.

### Key empirical results — two-step (Bayoumi and Masson (1995) style)
- Redistribution magnitude (1 − β_redist) from cross-sectional regressions:
  - U.S.: 28 percent.
  - Canada: 31 percent.
  - Australia: 13 percent.
- Risk-sharing magnitude (1 − β_risk-sharing) from panel regressions (changes in demeaned income):
  - U.S.: around 5 percent.
  - Australia: around 5 percent.
  - Canada: 8 percent.
- Note: two-step risk-sharing estimates for 1998-2010 are much smaller than Bayoumi and Masson (1969-1986) estimates; possible reasons include cycle harmonization and improved cross-regional financial markets.

### Key empirical results — PMG one-step
- PMG long-run redistribution coefficients b (reported as b; redistribution magnitude = 1 − b):
  - USA: b = 0.87*** → Redistribution magnitude 13 percent.
  - Canada: b = 0.82*** / 0.80*** → Redistribution magnitude 18-20 percent.
  - Australia: b = 0.78*** / 0.76*** → Redistribution magnitude 22-24 percent.
- PMG short-run / risk-sharing (from short-run coefficients d mapped into (1 − d) interpretation):
  - Risk-sharing magnitudes reported (consolidated in conclusions): Australia 4 percent; Canada 4 percent; U.S. 11 percent.
- Speed of adjustment (φ averages):
  - USA: −0.70*** (about 70 percent of deviation corrected within one year).
  - Canada: −0.57*** / −0.56*** (56-57 percent per year).
  - Australia: −0.31*** / −0.38*** (state incomes adjust more slowly than U.S./Canada).
- Notes on PMG vs prior literature:
  - PMG redistribution estimates smaller than many prior-study averages (average 26 percent in Table 1) and than two-step results.
  - PMG risk-sharing estimates closer to Von Hagen (1992), Goodhart and Smith (1993), Obstfeld and Peri (1998), and substantially lower than Bayoumi and Masson (1995) 31 percent estimate.
  - Hausman test p-values reported across specifications: 0.39, 0.88, 0.34, 0.85, 0.11, 0.76.

### Component analysis (TRANS, TRANSDIR, TAX) — main findings (Table 4 summaries)
- Transfers to state budgets (TRANS):
  - Weak evidence that TRANS cushions regional shocks.
  - Significant but small redistribution contributions in Australia and Canada: 7-8 percent (β estimates).
  - Largely no significant association with regional cycles (γ) → limited role in risk-sharing.
- Gross transfers to individuals (TRANSDIR):
  - Strong evidence TRANSDIR contributes to both redistribution and risk-sharing.
  - Redistribution impact: 12-16 percent.
  - Risk-sharing impact: 7 percent.
  - Significant risk-sharing impact disappears when controlling for time fixed effects → these transfers serve important stabilization (response to common shocks).
- Federal taxes (TAX):
  - Significant impact on risk-sharing in the U.S.: 21 percent.
  - Significant impact on risk-sharing in Australia: 17 percent.
  - TAX does not significantly affect redistribution.
- Net fiscal transfers (T = TRANS + TRANSDIR − TAX):
  - Net transfers significantly affect both risk-sharing and redistribution in all three federations.
  - When time fixed effects are included, γ (output gap) turns insignificant in Australia and Canada and decreases in magnitude in the U.S., indicating an important stabilization role of net transfers.
  - Estimated magnitudes of stabilization (difference between γ in columns with and without time fixed effects):
    - U.S.: 16 percent.
    - Canada: 14 percent.
    - Australia: 23 percent.
  - Stabilization magnitudes exceed those of risk-sharing.
- Overall implication:
  - Limited cushioning from transfers to state budgets; significant net effects mainly arise from gross transfers to individuals and federal taxes (automatic stabilizers).
  - Federal responses to common shocks (stabilization) are stronger than responses to regional shocks (risk-sharing).
  - Greater synchronization of regional cycles in Australia and Canada may explain insignificant automatic stabilizer effects once common shocks are controlled for.

### Consolidated quantitative takeaways (Section VIII)
- Net fiscal transfers have larger impact on stabilization than on risk-sharing across the three federations.
- Risk-sharing (moderate impact):
  - Australia: 4 percent.
  - Canada: 4 percent.
  - U.S.: 11 percent.
- Redistribution (PMG estimates):
  - U.S.: 13 percent.
  - Canada: 18-20 percent.
  - Australia: 22-24 percent.
- Two-step comparisons:
  - Two-step redistribution: U.S. 28 percent; Canada 31 percent; Australia 13 percent.
  - Two-step risk-sharing: U.S. and Australia around 5 percent; Canada 8 percent.
- Interpretation:
  - Net fiscal transfers more responsive to common shocks than idiosyncratic shocks.
  - Redistribution occurs where explicit mandates exist (Canada, Australia) and as a byproduct of centralization in the U.S.
  - Moderate measured role of transfers in risk-pooling may reflect trade-offs or cycle synchronization reducing regional shock importance.

### Policy implications and design considerations
- Strengthening fiscal risk-sharing is more compelling where regional cycles are less synchronized and where centralization can promote private risk-sharing (financial integration, central deposit insurance).
- Options to increase fiscal risk-sharing:
  - Ad hoc rainy-day fund to transfer resources to states hit by shocks.
  - Centralization of cyclical spending and revenue functions (e.g., unemployment insurance and taxes).
  - Common provision of public services to avoid procyclical local fiscal behavior when local automatic stabilizers are constrained.
  - Establishment of a central budget to reduce idiosyncratic state fiscal policies and allow coordinated countercyclical response (Cottarelli, 2013).
- Design caveat:
  - To avoid persistent redistributive transfers, schemes should target transitory shocks only, but policymakers cannot reliably distinguish transitory from persistent shocks in real time; schemes that ignore heterogeneous risk profiles can lead to permanent transfers.

*Source: _wp16141 (References, Annex 1, Annex 2, Sections VII–VIII summaries as presented).*

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

### _wp16141 - References

### Introduction and research focus
- This chapter analyses the role of net fiscal transfers in mitigating the impact of regional shocks on subnational public finances in federations.
- Fiscal transfers have multiple roles: closing vertical imbalances, achieving redistribution goals, and insuring states or provinces against macroeconomic shocks.
- Insurance against macroeconomic shocks is subdivided into:
  - risk-sharing: insurance against idiosyncratic macroeconomic shocks hitting individual states; and
  - stabilization: insurance against common shocks hitting all states simultaneously.
- Primary empirical focus: the extent of risk sharing and whether a federal system with a central budget—or equivalent risk-sharing mechanisms—would achieve better insurance against macroeconomic shocks than a system without it.

### Scope and sample
- Empirical assessment is restricted to three large fiscal federations: Australia, Canada, and the U.S.; restriction driven by data availability.
- Terminology for this chapter (unless otherwise specified):
  - "net fiscal transfers" (or "fiscal transfers") refers to net fiscal flows between central and sub-national governments, defined as the difference between gross transfers from the center to sub-national governments (either to the sub-national budget or to residents) and taxes paid by sub-national government residents to the center.
  - "sub-national" refers to state/provincial governments; lower level governments such as local regions and municipalities are not considered.

### Empirical challenges and identification
- Disentangling redistribution, stabilization, and risk-sharing is complicated because fiscal transfers often serve these roles simultaneously.
- Ideal measurement of risk sharing would require a counterfactual analysis comparing insurance magnitudes before and after establishment of a central budget; this is not feasible because almost all existing federations have long-functioning centralized fiscal systems.
- Given this limitation, the chapter follows an alternative route:
  - review literature on centralized versus decentralized fiscal systems’ effectiveness in offsetting macroeconomic shocks;
  - empirically assess the importance of risk-sharing in Australia, Canada, and the U.S., and the mechanisms through which risk-sharing occurs.

### Methodological contributions
- The chapter offers methodological refinements for estimating risk-sharing and redistribution motives of fiscal transfers:
  - Starts with the standard two-step approach used in the literature (where redistribution and risk-sharing motives are estimated separately).
  - Proposes a new, more efficient one-step approach.
  - Proposes an alternative approach that allows separate assessment of the extent to which individual components of net fiscal transfers contribute to risk-sharing (and redistribution).

### Structure of the chapter and supplementary material
- Remainder of the chapter is organized as follows:
  - Section II: summarizes theoretical arguments on whether a central budget in federations could enhance insurance against macroeconomic shocks.
  - Section III: reviews existing empirical approaches for measuring the impact of fiscal transfers on redistribution and insurance, and summarizes evidence from previous literature.
  - Section IV: introduces an efficient one-step empirical methodology to estimate fiscal risk-sharing and redistribution.
  - Section V: describes the data and provides summary statistics.
  - Section VI: presents estimation results from two empirical methodologies: the standard two-step approach used in previous studies and the new more efficient methodology.
- Annexes and supporting materials listed in the chapter:
  - Annex 1. Empirical approaches used in existing literature to quantify the impact of fiscal transfers (page 21).
  - Annex 2. Data description and sources (page 24).
  - Tables and Figures with exact numbering and titles are provided (Tables 1–4 on pages 25–28; Figures 1–6 on pages 29–34).

*Source: _wp16141 - References*

### Section VII introduces a third empirical approach which allows us to assess separately the role of

### DOES A CENTRAL BUDGET ENHANCE INSURANCE AGAINST MACROECONOMIC SHOCKS? REVIEW OF MAIN THEORETICAL ARGUMENTS

### Theoretical arguments for central budget insurance
- Insurance (interregional risk-sharing) allows member states to smooth consumption paths against idiosyncratic shocks by pooling risks across members.
- Risk-sharing is especially valuable when there are no state-specific exchange rates, price stickiness, limited mobility of production inputs, or dysfunctional private credit markets.
- Scale economies: central governments can borrow on better terms than individual states, reducing financing costs for countercyclical policies (example: U.S. states pay a risk-premium over Treasury bond yields).
- Fiscal spillovers and coordination: stabilization can be a “public good” where measures in one state leak to others via interstate trade spillovers, creating a free-riding problem that may justify centralized policy—particularly when monetary policy is at the lower bound.
- Ricardian equivalence: with (partly) Ricardian households, decentralized state fiscal expansions may be offset by households increasing savings in anticipation of future taxes, weakening fiscal multipliers; centralized systems dilute these Ricardian responses (example: Bayoumi and Masson (1998) — in Canada a non-liability creating stabilization would be two to three times more effective than decentralized stabilization financed through local public debt).

### Theoretical arguments against central budget insurance
- Design issues: efficient tax-and-transfer design to achieve stabilization and risk-sharing is difficult and may produce systematic redistribution from rich to poor regions; empirical evidence suggests richer regions pay higher federal taxes and receive lower federal transfers on average.
- Moral hazard: implicit bailout guarantees can reduce incentives for good policies, worsening states’ fiscal positions and vulnerability to shocks; this applies to centralized transfers and instruments like Eurobonds.

### Net conclusion from theory
- Theoretical arguments are mixed; the central empirical question is how much central budgets actually cushion regional shocks in practice.

### Redistribution and risk-sharing in existing federations — empirical approaches & prior evidence
- Fiscal transfers have two distinct effects:
  - Redistribution: permanent transfers from richer to poorer regions (convergence of living standards).
  - Insurance (risk-sharing and stabilization): temporary transfers that dampen the impact of idiosyncratic and common shocks.
- Interpretation example: an insurance level of 10 percent means disposable income falls by $0.90 for a $1 temporary decline in output; redistribution of 20 percent means a permanent $1 lower output relative to the national average results in disposable income $0.80 below the national average (the remaining $0.20 covered by transfers).
- Three empirical approaches in the literature:
  - Sachs and Sala-i-Martin (1992) — combined (no distinction between redistribution and insurance): estimates for 9 U.S. regions (1970-80) show combined cushioning of net fiscal transfers in range 33-40 percent.
  - Von Hagen (1992) — two-step approach separating redistribution and risk-sharing: U.S. (48 states, 1981-86) finds insurance ~10 percent and large redistribution ~47 percent; Goodhart and Smith (1993) apply similar method to multiple countries (U.S., Canada, U.K.) finding U.S. risk-sharing ~11 percent and redistribution ~13 percent (smaller redistribution possibly due to different state coverage).
  - Bayoumi and Masson (1995) — two-step approach not separating federal taxes and gross transfers: finds redistribution of net transfers in Canada 39 percent vs U.S. 22 percent; risk-sharing of net transfers U.S. 31 percent vs Canada 17 percent. Variants and extensions (Melitz and Zumer (2002), Obstfeld and Peri (1998), Decressin (2002)) show wide cross-country and method variation.
- Empirical regularity: redistributive motive generally stronger than risk-sharing (example aggregate: U.S. average risk-sharing 14 percent vs redistribution 23 percent).
- Limitations of prior studies: relatively old datasets, arbitrary two-step procedures, and lack of separation between automatic stabilizers and discretionary transfers.

### New one-step estimation method (PMG one-step cointegration approach)
- Motivation: combine level and first-difference regressions in one step to improve efficiency and capture dynamic effects (short-run adjustment of disposable income to long-run equilibrium).
- Estimator: Pooled Mean Group (PMG) estimator of Pesaran et al. (1999) — panel-data error-correction model.
- Empirical specification (as given):
  - Δ(푌퐷_it / 푌̅̅̅̅_ t ) = φ_i [ (푌퐷_{it−1} / 푌̅̅̅̅_{t−1}) − α − β (푌_{it−1} / 푌̅_{t−1}) ] + δ_i Δ(푌_{it} / 푌̅_{t}) + μ_i + ε_{it}
- Interpretation of coefficients:
  - β measures the redistribution effect (long-run response of relative disposable income to a permanent change in relative GDP); maps to β_redist in two-step literature.
  - δ (and δ_i) measures risk-sharing (short-run response of changes in relative disposable income to transitory changes in relative GDP); maps to β_risk-sharing in two-step literature.
  - φ (φ_i) is the speed of adjustment of relative disposable income to long-run equilibrium — larger absolute φ implies faster adjustment facilitated by fiscal transfers.
  - μ_i are state-specific fixed effects capturing unobserved heterogeneity; ε_{it} is i.i.d. error.
- Advantages of PMG one-step over two-step:
  - More efficient by combining high- and low-frequency information.
  - Accounts for dynamics emphasized by Obstfeld and Peri (1998).
  - Allows state-specific short-run coefficients δ_i (heterogeneous risk-sharing across states) while pooling long-run coefficient β (consistent with broadly uniform redistribution formulas).
  - Can be tested against the Mean Group (MG) estimator via the Hausman test; if homogeneity rejected, MG can analyze state-varying redistribution and risk-sharing.

### Data used for empirical application
- Federations analyzed: Australia, Canada, and the U.S.
- Transfer data collected:
  - Total transfers from federal government to state budgets (earmarked transfers and block grants) based on pre-determined formulas not directly linked to regional cyclical fluctuations.
  - State-level gross federal transfers to individuals (retirement, medical and income maintenance benefits, education and training assistance, unemployment insurance compensation) and taxes (personal and corporate income taxes) that are largely driven by automatic stabilizers and more closely related to regional cycles.
- Sample periods:
  - Australia and the U.S.: 1998 - 2010 (13 years).
  - Canada: 1992 – 2009 (18 years).
- Note: length of time series comparable to previous studies (averaging 16 years).

*Source: _wp16141 - Section VII introduces a third empirical approach which allows us to assess separately the role of (PDF)*

### Annex 2 contains detailed information on data sources and sample periods.

### _wp16141 - Annex 2 contains detailed information on data sources and sample periods

### A. Descriptive statistics: magnitude and cyclical behavior of fiscal transfers
- Federal tax revenues range between 15-20 percent of state GDP.
- Gross fiscal transfers range between 5-15 percent of state GDP.
- U.S. case: federal tax revenues declined markedly and outlays increased significantly in 2009-2010.
- Canada and Australia: federal taxes saw a sharp decline in the aftermath of the crisis while spending was somewhat less responsive than in the U.S. (except in Australia in 2009).
- Net transfers from the federal government increased following the global financial crisis, reflecting automatic stabilizers and discretionary stimulus packages.

- Illustrative median state-level net fiscal transfers (1998-2010 and country-specific samples):
  - U.S. (Figure 4): West Virginia — median annual net fiscal transfer amounted to slightly less than 20 percent of state GDP; Minnesota — median net fiscal transfer amounted to about 15 percent of state GDP.
  - Canada (Figure 5): Newfoundland and Labrador — median of 10 percent of state GDP; Ontario — median of -10 percent of state GDP.
  - Australia (Figure 6): Northern Territory — median of 15 percent of state GDP; Australian Capital Territory — median of -5 percent of state GDP.

- Institutional note: redistribution in Canada and Australia is an explicit constitutional objective; in the U.S. redistribution occurs as a byproduct of revenue and spending centralization, including progressive features.

### B. Preliminary evidence on redistribution (box-plot results)
- Box plots of net fiscal transfers show clear evidence of redistribution in all three countries.
- Permanent net recipients and contributors identified in the U.S., Canada, and Australia (examples listed above).

### VI. Estimation results — methods overview
- Estimation based on equation (1) using the Pooled Mean Group (PMG) estimator; comparison with Bayoumi and Masson (1995) two-step approach.
- PMG constrains long-run slope coefficients to be equal across states; MG does not. Hausman test used to assess homogeneity assumption.
- State potential GDP estimated using linear and quadratic trends (equivalent of HP filter with smoothing parameter of 100).

### A. Bayoumi and Masson (1995) two-step approach — key estimates (Table 2)
- Redistribution magnitude (1-β_redist) from cross-sectional regressions:
  - U.S.: 28 percent (somewhat larger than the 22 percent reported in Bayoumi and Masson).
  - Canada: 31 percent (smaller than Bayoumi and Masson’s 39 percent).
  - Australia: 13 percent (smallest).
- Caveat: Canada and Australia results based on a small number of observations (equal to number of states), which could bias estimates.
- Risk-sharing estimates (1-β_risk-sharing) from panel regressions (changes in demeaned income), baseline and trend-augmented:
  - U.S.: around 5 percent.
  - Australia: around 5 percent.
  - Canada: 8 percent.
- Note: risk-sharing estimates for U.S. and Canada over 1998-2010 much smaller than Bayoumi and Masson (1969-1986); possible reasons include harmonization of regional cycles and improved cross-regional financial markets.

### B. PMG estimator: one-step approach — key estimates (Table 3)
- U.S. (Columns (1)-(2), baseline and trend-augmented):
  - Redistribution magnitude: 13 percent.
  - Risk-sharing magnitude: 11 percent.
  - Trend coefficient close to zero → no substantial increase in redistribution over time.
  - Speed of adjustment: about 70 percent of deviation from long-run equilibrium corrected within one year.
- Canada (Columns (3)-(4)):
  - Redistribution magnitude: 18-20 percent.
  - Risk-sharing: fiscal transfers cushion only 4 percent of the initial regional income shock.
  - Trend coefficient close to zero → no substantial increase in redistribution over time.
  - Speed of adjustment: 56-57 percent per year.
- Australia (Columns (5)-(6)):
  - Redistribution magnitude: 22-24 percent (largest among the three).
  - Risk-sharing effect: 4 percent.
  - Speed of adjustment smaller than in U.S. and Canada → state-level incomes adjust more slowly to long-run equilibrium.
- Comparison notes:
  - PMG redistribution estimates smaller than many prior studies’ averages (average 26 percent in Table 1) and two-step results.
  - PMG risk-sharing estimates closer to some prior studies (Von Hagen (1992), Goodhart and Smith (1993), Obstfeld and Peri (1998)) and differ substantially from Bayoumi and Masson (1995) 31 percent estimate.
  - Differences attributed to more recent data and methodological differences (one-step exploits cross-sectional and time variation).

### VII. Role of individual components of net fiscal transfers in cushioning shocks — empirical specification (equation (2))
- Dependent variables: ratios over state potential GDP of:
  - TRANS = gross transfers to state budgets,
  - TRANSDIR = gross transfers to individuals,
  - TAX = federal taxes,
  - T = TRANS + TRANSDIR - TAX (net fiscal transfers).
- Two explanatory variables:
  - Trend component of relative income gap (β): negative β (positive for TAX) supports redistribution motive.
  - State-specific output gap (γ): negative γ (positive for TAX) supports risk-sharing motive after controlling for stabilization via time fixed effects.
- Estimation strategy: specification (2) without time fixed effects (γ measures impact of risk-sharing and stabilization) and with time fixed effects (γ measures risk-sharing only).

### VII. Results on components (Table 4) — summary findings
- Transfers to state budgets (TRANS):
  - Weak evidence that TRANS cushions regional shocks.
  - Significant but small impact on redistribution in Australia and Canada: 7-8 percent (β).
  - Largely no significant association with regional cycles (γ) → limited role for risk-sharing.

- Gross transfers to individuals (TRANSDIR):
  - Strong evidence that TRANSDIR contributes to redistribution and risk-sharing.
  - Impact on redistribution: 12-16 percent.
  - Impact on risk-sharing: 7 percent.
  - Significant impact on risk-sharing disappears when controlling for time fixed effects → these transfers serve an important stabilization role (response to common shocks).

- Federal taxes (TAX):
  - Significant impact on risk-sharing in the U.S.: 21 percent.
  - Significant impact on risk-sharing in Australia: 17 percent.
  - TAX does not significantly affect redistribution.

- Net fiscal transfers (T = TRANS + TRANSDIR - TAX):
  - In net terms, fiscal transfers have significant impact on both risk-sharing and redistribution in all three federations.
  - When time fixed effects are included, coefficient γ (output gap) turns insignificant in Australia and Canada and decreases in magnitude in the U.S. This indicates the stabilization motive of net fiscal transfers is important.
  - Estimated magnitude of stabilization (difference between γ in columns (4) and (1)):
    - U.S.: 16 percent.
    - Canada: 14 percent.
    - Australia: 23 percent.
  - Stabilization magnitudes exceed those of risk-sharing.

- Overall implication from component analysis:
  - Transfers to state budgets play a limited role in cushioning regional shocks.
  - Significant net transfer effects mostly arise from gross transfers to individuals and federal taxes (automatic stabilizers).
  - Federal governments’ response to common shocks (stabilization) is stronger than response to regional shocks (risk-sharing).
  - Greater synchronization of regional cycles in Australia and Canada may explain insignificant impact of automatic stabilizers when common shocks are controlled for.

### VIII. Conclusions — consolidated quantitative takeaways and policy implications
- Comparative magnitudes (methodology-averaged findings across sections):
  - Net fiscal transfers have larger impact on stabilization than on risk-sharing.
  - Risk-sharing (moderate impact):
    - Australia: 4 percent.
    - Canada: 4 percent.
    - U.S.: 11 percent.
  - Redistribution:
    - U.S.: 13 percent (PMG estimate).
    - Canada: 18-20 percent (PMG).
    - Australia: 22-24 percent (PMG).
  - Two-step redistribution estimates (for comparison): U.S. 28 percent, Canada 31 percent, Australia 13 percent.
  - Two-step risk-sharing estimates: U.S. and Australia around 5 percent; Canada 8 percent.
- Interpretation:
  - Net fiscal transfers are more responsive to common shocks than to idiosyncratic shocks in the three federations studied.
  - Redistribution is observed in federations with explicit mandates (Canada, Australia) and also in the U.S., implying redistribution can arise as a consequence of centralization.
  - The moderate measured role of net fiscal transfers in risk-pooling may reflect trade-offs with other considerations, or fiscal centralization may have synchronized business cycles and reduced regional shock importance.
- Policy implications and options for currency unions (e.g., euro area):
  - Case for enhancing fiscal risk-sharing stronger where cycles are less synchronized and centralization can promote private risk-sharing (financial integration, central deposit insurance).
  - Options to increase fiscal risk-sharing:
    - Ad hoc rainy-day fund to transfer resources to states hit by shocks.
    - Centralization of cyclical spending and revenue functions (e.g., unemployment insurance and taxes).
    - Common provision of public services to prevent procyclical local fiscal behavior when local automatic stabilizers cannot operate fully (due to local fiscal rules or financing constraints).
    - Establishment of a central budget to reduce idiosyncratic state fiscal policies and allow coordinated countercyclical response (Cottarelli, 2013).
- Design caveat:
  - To avoid persistent redistributive transfers, transfer systems should target transitory shocks only, but policymakers cannot reliably distinguish transitory from persistent shocks in real time; schemes that ignore heterogenous risk profiles can lead to permanent transfers.

*Source: _wp16141 - Annex 2 contains detailed information on data sources and sample periods.*

### REFERENCES

### _wp16141 - REFERENCES

### Key references cited
- Andersson, L., 2004, “Regional Risk-Sharing Provided by the Fiscal System: Empirical Evidence from Sweden,” Regional Studies, 38 (3): pp. 269-80.
- Bayoumi, T., and Masson, P., 1995, “Fiscal Flows in the United States and Canada: Lessons from Monetary Union in Europe,” European Economic Review, 39: pp. 253–74.
- Bayoumi, T. and Masson, P., 1998, “Liability-Creating versus Non-Liability-Creating Fiscal Stabilization Policies: Ricardian Equivalence, Fiscal Stabilization, and EMU”, Economic Journal, 108: pp. 1026-45.
- Beetsma, R. and Debrun, X., 2004, “The Interaction Between Monetary and Fiscal Policies in a Monetary Union: A Review of Recent Literature”, In Monetary Policy, Fiscal Policies, and Labor Markets: Macroeconomic Policymaking in the EMU, ed. Roel Beetsma, (Cambridge University Press, Cambridge, UK).
- Bornhorst, F., Dobrescu, G., Fedelino, A., Gottschalk, J., and Nakata, T., 2011, “When and How to Adjust Beyond the Business Cycle? A Guide to Structural Fiscal Balances,” Technical Guidance Note, Fiscal Affairs Department (Washington: International Monetary Fund).
- Cottarelli, C., 2013, “European Fiscal Union: A Vision for the Long Run,” Swiss Journal of Economics and Statistics, 149 (2): pp. 167-174.
- Decressin, J., 2002, “Regional Income Redistribution and Risk-Sharing: How Does Italy Compare in Europe?” Journal of Public Economics, 86: pp. 287-306.
- European Commission, 1977, “Report of the Study Group on the Role of Public Finances in European Integration” (MacDougall Report), (Brussels: European Commission).
- Friedman, M., 1957, A Theory of the Consumption Function, Princeton, N.J. (Princeton University Press).
- Goodhart, C., and Smith, S., 1993, “Stabilization,” In The Economics of Community Public Finance, European Economy Reports and Studies, 5: 417–55 (Brussels: European Commission).
- Hamada, K., 1985, The Political Economy of International Monetary Interdependence (MIT Press, Cambridge, MA).
- Koutsogeorgopoulou, V., 2007, “Fiscal Relations Across Levels of Government in Australia,” OECD Economics Department Working Paper No. 541 (Paris: OECD).
- Masson, P. and Taylor, M., 1993, “Fiscal Policy within Common Currency Areas,” Journal of Common Market Studies, 31 (1): pp. 29-44.
- Melitz, J., and Zumer, F., 2002, “Regional Redistribution and Stabilization by the Centre in Canada, France, the U.K. and the U.S.: A Reassessment and New Tests,” Journal of Public Economics, 86 (2): pp. 263–84.
- Obstfeld, M., and Peri, G., 1998, “Regional Non-Adjustment and Fiscal Policy,” Economic Policy, 26: pp. 205–69.
- Pesaran, H., Shin, Y., and Smith, R., 1999, “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels,” Journal of the American Statistical Association, 94: pp. 621-34.
- Sachs, J. and Sala-i-Martin, X., 1992, “Fiscal Federalism and Optimum Currency Areas: Evidence for Europe from the United States”, In Establishing a Central Bank: Issues in Europe and Lessons from the U.S., ed. Matthew Canzoneri, Vittorio Grilli, and Paul Masson, (Cambridge: Cambridge University Press).
- Von Hagen, J., 1992, “Fiscal Arrangements in a Monetary Union: Evidence from the US”, In Fiscal Policy, Taxes, and the Financial System in an Increasingly Integrated Europe, ed. Don Fair and Christian de Boissieux, (Kluwer, London).

### Annex 1 — Empirical approaches used to quantify the impact of fiscal transfers
- Overview:
  - Three main empirical approaches reviewed: Sachs and Sala-i-Martin (1992); Von Hagen (1992); Bayoumi and Masson (1995).
  - Purpose: quantify the impact of fiscal transfers on redistribution and risk-sharing across regions.

- Sachs and Sala-i-Martin (1992) approach
  - Disposable income identity: ∆YD = ∆Y − ∆TX + ∆TR (variables in real per capita terms).
  - Define elasticities:
    - 훽_TX = (∆TX / ∆Y) (Y / TX)
    - 훽_TR = (∆TR / ∆Y) (Y / TR)
  - Rewritten as: ∆YD = ∆Y * λ (A.1)
    - where λ = (1 − 훽_TX (TX / Y) + 훽_TR (TR / Y)) measures effect of net fiscal transfers on regional disposable income.
  - Regressions to estimate elasticities (A.2) and (A.3):
    - log(TX_it / TX̅_t) = α_TX + β_TX log(Y_it / Y̅_t) + γ_TX trend + ε_it
    - log(TR_it / TR̅_t) = α_TR + β_TR log(Y_it / Y̅_t) + γ_TR trend + u_it
  - Application: 9 U.S. regions, period 1970-80.
  - Limitation: assesses overall impact but does not disentangle redistribution vs. insurance (risk-sharing).

- Von Hagen (1992): two-step approach
  - Regressions (A.4) and (A.5):
    - ∆log(TX_it) = α_i + θ_TX ∆log(Y_it) + D_Oil + γ_t + ε_it
    - ∆log(TR_it) = α_i + θ_TR ∆log(Y_it) + D_Oil + γ_t + ε_it
    - D_Oil: dummies for Alaska, Oklahoma, Wyoming.
    - α_i: regional fixed effects; γ_t: time fixed effects.
  - Interpretation: θ_TX and θ_TR measure risk-sharing effects of federal taxes and transfers.
  - Application: 48 U.S. states, period 1981-86.
  - Von Hagen also conducts cross-sectional regression on regional averages to assess redistribution.
  - Use of time fixed effects controls for aggregate shocks.

- Bayoumi and Masson (1995): two-step approach
  - Combine taxes and transfers; two-step method separates redistribution and risk-sharing:
    - Cross-section for redistribution (A.6): (YD̅_i / YD̿) = α + β_redist (Y̅_i / Y̿) + ε_i
      - Slope coefficient yields (1 − β_redist) as redistribution measure.
    - Panel for risk-sharing (A.7): ∆(YD_it / YD̅_t) = α_i + β_risk-sharing ∆(Y_it / Y̅_t) + u_it
      - Slope coefficient yields (1 − β_risk-sharing) as risk-sharing measure.
  - Demeaning by national averages controls for common shocks (within transformation), as alternative to time fixed effects.
  - Application: U.S. states and Canadian provinces, period 1969-1986.

- Comparison and pitfalls
  - Sachs and Sala-i-Martin: evaluates overall impact but cannot disentangle redistribution vs. risk-sharing.
  - Von Hagen and Bayoumi & Masson: attempt to separate effects via cross-sectional averages (redistribution) and panel transformations or time effects (risk-sharing).
  - Limitation of two-step approaches: averaging reduces degrees of freedom and can be inefficient, especially with small numbers of states (e.g., Canada).
  - Authors propose also reporting a one-step approach that estimates redistribution and risk-sharing simultaneously to address efficiency concerns.

### Annex 2 — Data description and sources (selection of variables and sample periods)
- Macroeconomic variables
  - GDP:
    - Australia – Australia Bureau of Statistics, from 1990 to 2011.
    - Canada – Department of Finance, from 1990 to 2010.
    - USA – US Census Bureau, from 1997 to 2010.
  - CPI (country-level):
    - Australia – Australia Bureau of Statistics, from 1990 to 2011.
    - Canada – Department of Finance, from 1990 to 2011.
    - USA – US Census Bureau, from 1990 to 2010.
  - Population:
    - Australia – Australia Bureau of Statistics, from 1990 to 2011.
    - Canada – Department of Finance, from 1990 to 2010.
    - USA – US Census Bureau, from 1990 to 2010.

- Fiscal variables
  - Gross transfers from central government to state budgets:
    - Australia – Australia Bureau of Statistics, from 1999 to 2011.
    - Canada – Department of Finance, from 1990 to 2011.
    - USA – US Census Bureau, from 1990 to 2010.
  - Gross transfers from central government to individuals:
    - Australia – Australian Bureau of Statistics, Australian National Accounts table 5220.0, from 1998 to 2011.
    - Canada – Statistics Canada, Provincial and Territorial Economic Accounts, from 1990 to 2009.
    - USA – Bureau of Economic Analysis (BEA), from 1998 to 2010.
  - Taxes paid to central government:
    - Australia – Australian Bureau of Statistics, Australian National Accounts table 5220.0, from 1998 to 2011.
    - Canada – Statistics Canada, Provincial and Territorial Economic Accounts, from 1990 to 2009.
    - USA – Internal Revenue Service (IRS), 1998 to 2010.

### Key empirical estimation results reported
- Table 1 (summary from literature)
  - Examples of reported redistribution and risk-sharing estimates from prior studies (selected entries as reported in table):
    - von Hagen (1992), USA, 1981-86: Redistribution 47; Risk-sharing 10.
    - Goodhart and Smith (1993), Canada, 1966-88: Redistribution 15; Risk-sharing 17.
    - Bayoumi and Masson (1995), Canada, 1969-86: Redistribution 39; Risk-sharing 31.
    - Bayoumi and Masson (1995), USA, 1969-86: Redistribution 22; Risk-sharing 17.
    - Decressin (2002), Italy, 1970-95: Redistribution 25--35; Risk-sharing 10--15.
    - Obstfeld and Peri (1998), Canada, 1971-95: Redistribution 53; Risk-sharing 13.
    - Federal fiscal flows (aggregate reported): Redistribution 33--40; Risk-sharing 13; additional range 7--22.

- Table 2. Estimation Results Based on Bayoumi and Masson (1995) Two-step Approach
  - Panel A. Redistribution (cross-sectional OLS)
    - USA (column (1)): b_redist = 0.72*** (0.030); Constant = 0.28*** (0.030); # of obs. 48; R-squared 0.92.
    - Canada (column (2)): b_redist = 0.69*** (0.050); Constant = 0.31*** (0.051); # of obs. 10; R-squared 0.87.
    - Australia (column (3)): b_redist = 0.87*** (0.154); Constant = 0.13 (0.135); # of obs. 8; R-squared 0.82.
  - Panel B. Risk-sharing (fixed-effects OLS)
    - USA columns (1)-(2): b_risk-sharing = 0.95*** (0.107); g (trend) = 0.00 (0.001 or 0.000); Constant = 0.00 (various s.e.); # of obs. 624; # of states 48; R-squared 0.364.
    - Canada columns (3)-(4): b_risk-sharing = 0.92*** (0.020)/(0.021); # of obs. 180; # of states 10; R-squared 0.816.
    - Australia columns (5)-(6): b_risk-sharing = 0.95*** (0.034)/(0.037); # of obs. 104; # of states 8; R-squared 0.947.

- Table 3. Estimation Results Based on the Pooled Mean Group (PMG) Estimator (Pesaran et al., 1999)
  - Notes: Estimations include province-specific fixed effects. Sample periods: Australia (1998-2010), Canada (1992-2009), USA (1998-2010).
  - Long-run coefficients b (redistribution):
    - USA columns (1)-(2): 0.87*** [0.01]
    - Canada columns (3)-(4): 0.82*** / 0.80*** [0.02][0.02]
    - Australia columns (5)-(6): 0.78*** / 0.76*** [0.03][0.02]
  - g (trend):
    - Reported as 0.00***, 0.00, -0.00*** in various specifications (standard errors reported as [0.00]).
  - Speed of adjustment (average of state-specific estimates):
    - USA: -0.70*** [-0.04] / -0.70*** [-0.04]
    - Canada: -0.57*** [-0.08] / -0.56*** [-0.08]
    - Australia: -0.31*** [-0.06] / -0.38*** [-0.09]
  - Short-run coefficients d (risk-sharing):
    - USA: 0.89*** [0.04] / 0.89*** [0.04]
    - Canada: 0.96*** [0.03] / 0.96*** [0.02]
    - Australia: 0.96*** [0.03] / 0.96*** [0.03]
  - # of obs. and # of states:
    - USA: # of obs. 624; # of states 48.
    - Canada: # of obs. 180; # of states 10.
    - Australia: # of obs. 104; # of states 8.
  - Hausman test (p-value) reported for specifications: 0.39, 0.88, 0.34, 0.85, 0.11, 0.76 (respectively for columns shown).

- Table 4. Alternative specification to assess stabilization and risk-sharing (fixed effects OLS)
  - Dependent variable: ratio of respective measure of fiscal transfers relative to state potential output. Ypot denotes potential output; GAP denotes output gap (quadratic trend).
  - Selected coefficient estimates (with t-statistics in parentheses) across panels (samples: Australia, USA, Canada):
    - Relative real per capita income (b) coefficients (examples):
      - USA, (1): -0.01 (-0.63)
      - USA, (2): -0.12** (-2.17)
      - USA, (4): -0.13** (-2.11)
      - Canada, (2) [with time fixed effects]: -0.12* (-1.97)
      - Australia, (1): -0.07*** (-3.75)
    - State output gap (g) coefficients (examples):
      - USA, (2): -0.07*** (-6.21)
      - USA, (3): 0.21** (2.51)
      - USA, (4): -0.28*** (-3.13)
      - Canada, (3): 0.15*** (3.11)
      - Australia, (3): 0.17*** (3.66)
    - Constants and fit:
      - Example constants: 4.50*** (3.11), 24.75*** (4.46), 15.07*** (3.08), 14.17** (2.22) in USA panel examples.
      - # of obs. frequently 672 (USA pooled), 180 (Canada), 104 (Australia); # of states 48, 10, 8 respectively.
      - R-squared values vary by specification, e.g., USA panel R-squared 0.269, 0.458, 0.016, 0.352 across columns; Canada and Australia panels show higher R-squared in some specifications (e.g., Canada 0.537, 0.571, 0.046, 0.497; Australia 0.001, 0.819, 0.108, 0.104).

### Figures and graphical summaries (as reported)
- Time-series plots reported (Figures 1–3) show transfers to states, transfers to households, tax revenues, and net transfers as averages percent of state GDP across years:
  - Australia: annual series from 1999 to 2011.
  - Canada: annual series from 1990 to 2009.
  - USA: annual series from 1998 to 2010.
- Distributional boxplots (Figures 4–6) show redistribution through net fiscal transfers:
  - Figure 4: USA (1998-2010) — distribution across 48 states (plot shows median, 25th/75th percentiles, min/max).
  - Figure 5: Canada (1992-2009) — distribution across provinces (Ontario, Alberta, British Columbia, Quebec, Saskatchewan, Manitoba, Nova Scotia, New Brunswick, Prince Edward Island, Newfoundland and Labrador).
  - Figure 6: Australia (1998-2010) — distribution across states/territories (Australian Capital Territory, New South Wales, Victoria, Western Australia, Queensland, South Australia, Tasmania, Northern Territory).

*Italic: Source document _wp16141 - REFERENCES (selected references, Annex 1 and Annex 2 content, tables and figures as presented).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16141.pdf_