## _wp15101

## Source details

**Canonical URL:** [_wp15101](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15101.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15101.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15101.pdf.json)

---

### Major findings
- A 1 percentage point increase in energy subsidies to GDP leads, on average, to a reduction of public spending in education and health by 0.6 percentage point of GDP.
- The crowding-out effect is stronger:
  - in the presence of weak domestic institutions (delivery of public social spending is subject to important leakages, deadweight loss ζ);
  - when fiscal space is narrow;
  - among net oil importers.
- Resource wealth mitigates the crowding-out: a larger size of the fiscal “pie” limits the severity of the reduction in public social spending.

### Context and magnitudes cited in the study
- Pre-tax energy subsidies were estimated to amount to 0.7 percent of global GDP in 2011.
- Post-tax energy subsidies were evaluated at about 2.9 percent of global GDP in 2011, equivalent to 8.5 percent of total government revenues.
- In the MENA region in 2011, pre-tax energy subsidies absorbed around 22 percent of government revenues.
- Micro-based evidence: the top income quintile captures six times more in subsidies than the bottom, in absolute terms; the poorest 20 percent of households receive only about one-tenth of natural gas and electricity subsidies.

### Theoretical framework and mechanism (political game)
- Agents and preferences:
  - Two agent types: the rich (elite) with income y_r and the poor (middle-class) with income y_p (y_p < y_r).
  - Utility u_i(c_i, k) = ln(c_i) + v_i(k); private consumption c_i = c_{i,e}^{θ} c_{i,n}^{1−θ}, 0 < θ < 1.
- Timing and mechanism:
  1. Nature chooses bottlenecks (ζ), observed by all.
  2. The rich set energy subsidy rate (δ).
  3. The poor choose public social spending (k).
  4. Both groups choose consumption c_e and c_n.
- Key model implications:
  - Energy subsidies deliver immediate private benefits (favoring the rich); public social spending suffers from leakages (ζ).
  - Elite choose δ anticipating the poor’s response, potentially setting δ sub-optimally high and crowding out k, especially with weak institutions and limited fiscal space.
- Closed-form results (selected):
  - Poor’s optimal public spending: k* = (1−ζ) ( z + \bar{τ} y − δ / (1−δ) θ (1− \bar{τ}) y ) (Equation (7)).
  - ∂k*/∂δ = − (1−ζ) / (1−δ)^2 θ (1− \bar{τ}) y < 0 (Equation (8)).
  - Elite’s optimal subsidy rate: δ* = 1 − (1−ζ)(1− \bar{τ}) y v_r′(k*) (Equation (9)).
  - ∂δ*/∂ζ = (1− \bar{τ}) y v_r′(k*) > 0.

### Empirical strategy and identification
- Objective: identify causal relationship between energy subsidies and public social spending.
- Data and variables:
  - Energy subsidies from Clements and others (2013) using the "price gap" approach: s_{e i,t} = Σ_{e=1}^m (p_{e w,t} − p_{e i,t}) c_{e i,t} (Equation (10)); pre-tax subsidies censored at 0 when price gap negative.
  - Social spending: public expenditure in education + health (% of GDP) from World Development Indicators (WDI).
  - Stylized facts from panel of 109 low- and middle-income countries, 2000–11.
- Endogeneity concerns:
  - Simultaneity between subsidies and social spending; omitted variables correlated with both.
- Instruments:
  - Main instrument: energy subsidy intensity in neighbor countries (ESINC), ESINC_i = 1/n ∑_{j=1}^n s_j × d_{i,j}; weighting uses Cepii gravity database.
  - Second instrument: Constraints on the Executive (political constraints).
  - When interacting subsidies with moderators, instrument matrix expanded with interaction terms.
- Identification validation:
  - Relevance assessed by F-stat and Shea R^2; orthogonality tested by Sargan over-identification test.

### Estimation approach and key regression specification
- Cross-section specification using averages over 2000–11:
  - Social_i = α + (β_1 + β_2 F_i) Subsidies_i + β_3 F_i + X_i′ Γ + ε_i (Equation (11)).
  - "Social": education + health (% of GDP).
  - "Subsidies": total pre-tax energy subsidies (% of GDP).
  - X includes demographic controls, initial real per capita income, government size (excluding subsidies), trade openness, macro volatility, and measures of domestic institutions.
  - F_i includes fiscal space (public debt-to-GDP), natural resource dependency (oil rents-to-GDP), and institutional quality.

### Empirical results (IV estimates and diagnostics)
- Baseline linear IV second-stage coefficients for Energy subsidies-to-GDP (selected):
  - Column (1): -0.630** (standard error 0.282)
  - Column (2): -1.473* (standard error 0.764)
  - Column (3): -0.477 (standard error 0.303)
  - Column (4): 0.381 (standard error 0.488)
- Interaction effects (selected):
  - Energy subsidies * Oil rents-to-GDP: 0.055** (standard error 0.022) in Column (2).
  - Energy subsidies * Public debt-to-GDP: -0.022** (standard error 0.011) in Column (3).
  - Energy subsidies * 1[Corruption > 75th perc.]: -1.956* (standard error 1.068) in Column (4).
- Notable controls:
  - Government size: 0.182*** (0.028) in Column (1).
  - Energy subsidies intensity in neighbor countries (ESINC): 0.076*** (0.023) in Column (1).
  - Constraints on the Executive: -0.245*** (0.083) in Column (1).
- First-stage and diagnostics:
  - Joint significance of subsidies coefficients: P-value 0.031 (Column 1); 0.031 (Column 2); 0.089 (Column 3).
  - F-stat of first-stage: 15.545 (Column 1); 7.287 (Column 2); 9.100 (Column 3); 5.335 (Column 4).
  - Sargan OID test: P-value 0.043 (Column 1); 0.414 (Column 2); 0.507 (Column 3); 0.113 (Column 4).
  - Sample sizes: N 109 (Columns 1–3), N 108 (Column 4).

### Main quantitative interpretations
- Average crowding-out effect (baseline, Column 1): about 0.6. Interpretation: public spending in education and health were on average 0.6 percentage point of GDP lower in countries where energy subsidies were 1 percentage point of GDP higher.
- Fiscal space interaction:
  - Crowding-out rises to about 0.8 for countries with a debt-to-GDP ratio above the 75th percentile of the sample distribution, corresponding to a debt-to-GDP ratio of about 70.
- Institutional interaction:
  - Crowding-out is almost one-to-one in countries with government ineffectiveness above the 75th percentile.

### Policy implications and recommendations
- Energy subsidy reform can free fiscal space for priority public social spending (education and health), but effectiveness of reallocated resources depends on delivery efficiency (leakage ζ).
- Reforms are more likely to translate into improved public social outcomes when accompanied by improvements in governance, anti-corruption measures, and broader fiscal space.
- Address the political economy of subsidy retention: because subsidies provide visible, immediate, and often regressive benefits, reform strategies should consider distributional compensation and measures to increase credibility of reallocations to social spending.
- Non-resource-rich countries with narrow fiscal space should move expeditiously with subsidy reform to relax constraints on public social spending.
- Resource-rich economies face challenges sustaining high subsidies given mounting social spending pressures and volatile resource revenues; lower international oil prices erode fiscal buffers.
- The recent sharp drop in global oil prices represents an opportunity to durably reform energy subsidies.
- Depoliticizing domestic energy pricing—e.g., adopting an automatic pricing mechanism—is recommended as a transition toward fully-deregulated energy prices.

*Source — _wp15101 - 2.   Solving the Game (Backward Induction); 2.   Why Adopt a Cross-sectional Specification?; 3.   Identification Strategy; 4.   Estimation Results; IV.   Conclusion and Policy Implications (excerpts)._*

### 0.6 percentage point of GDP in countries where energy subsidies were 1 percentage point of GDP

### 0.6 percentage point of GDP in countries where energy subsidies were 1 percentage point of GDP

### Major findings
- A 1 percentage point increase in energy subsidies to GDP leads, on average, to a reduction of public spending in education and health by 0.6 percentage point of GDP.
- The crowding-out effect is stronger:
  - in the presence of weak domestic institutions (delivery of public social spending is subject to important leakages, deadweight loss ζ);
  - when fiscal space is narrow;
  - among net oil importers.
- Resource wealth mitigates the crowding-out: a larger size of the fiscal “pie” limits the severity of the reduction in public social spending.

### Context and magnitudes cited in the study
- Pre-tax energy subsidies were estimated to amount to 0.7 percent of global GDP in 2011.
- Post-tax energy subsidies were evaluated at about 2.9 percent of global GDP in 2011, equivalent to 8.5 percent of total government revenues.
- In the MENA region in 2011, pre-tax energy subsidies absorbed around 22 percent of government revenues.
- Micro-based evidence: the top income quintile captures six times more in subsidies than the bottom, in absolute terms; the poorest 20 percent of households receive only about one-tenth of natural gas and electricity subsidies.

### Theoretical framework (political game)
- Agents: two types — the rich (elite) with income y_r and the poor (middle-class) with income y_p (y_p < y_r).
- Preferences: utility u_i(c_i, k) = ln(c_i) + v_i(k), with private consumption c_i a Cobb-Douglas aggregate of energy and non-energy goods: c_i = c_{i,e}^{θ} c_{i,n}^{1−θ}, 0 < θ < 1.
- Timing:
  1. Nature chooses the level of bottlenecks (ζ), observed by all.
  2. The rich set the energy subsidy rate (δ).
  3. The poor choose public social spending (k).
  4. Both groups choose consumption of subsidized energy (c_e) and non-energy (c_n).
- Mechanism: energy subsidies provide a certain, immediate benefit to consumption (favoring the rich), while public social spending suffers from leakages (ζ) so the poor may prefer subsidies despite regressivity; elite choose δ anticipating these incentives, potentially setting subsidies sub-optimally high and crowding out social spending, especially under weak institutions and tight fiscal space.

### Empirical strategy and identification
- Cross-sectional analysis of a large sample of low- and middle-income countries.
- Main endogeneity concern: simultaneity in budget choices between energy subsidies and public social spending.
- Instrumental variable: level of subsidies in neighboring countries used to instrument a country’s subsidies.
- Political constraints on the executive are included to capture implementation ease of subsidy policies.

### Empirical results (high-level)
- Instrumental variable estimations indicate a statistically and economically meaningful crowding-out: 1 percentage point of GDP higher energy subsidies → 0.6 percentage point of GDP lower public spending in education and health.
- Non-linearities documented: the magnitude of the crowding-out varies with institutional quality and fiscal space and is larger for net oil importers.

### Policy implications (drawn in the chapter)
- Energy subsidy reform can free fiscal space for priority public social spending (education and health), but the effectiveness of reallocated resources depends on domestic institutions that govern delivery efficiency (leakage ζ).
- Reforms are more likely to translate into improved public social outcomes when accompanied by improvements in governance, anti-corruption measures, and broader fiscal space.
- The political economy of subsidy retention must be addressed: subsidies provide visible, immediate benefits (often regressive) that can be politically entrenched, so reform strategies should consider distributional compensation and measures to increase credibility of reallocations to social spending.

*Working Paper excerpt: JEL Classification Numbers: C21, C26, D71, H23, H41; Keywords: Energy subsidies; Public social spending; Causality; Author E-Mail Addresses: CEbeke@imf.org; CLonkeng@imf.org*

### 2.   Solving the Game (Backward Induction)

### 2.   Solving the Game (Backward Induction)

### Third stage of the game: each agent decides on the consumption of private goods
- For a given subsidy rate δ and a given provision of the public good k, each agent allocates a constant share of disposable income to energy and non-energy consumption under the Cobb-Douglas/log specification:
  - energy share: θ
  - non-energy share: 1−θ
- Consumption equations (agent i):
  - (1−δ)p_w c_{i,e} = θ y_i (1− \bar{τ}) (3)a
  - c_{i,n} = (1−θ) y_i (1− \bar{τ}) (3)b
- Notes and implications:
  - δ is the subsidy rate: the agent pays fraction 1−δ of the international energy price p_w; price of non-energy good normalized to unity.
  - Constant consumption shares imply energy consumption increases with the subsidy rate, for given disposable income and international energy price.
  - The rich benefit more from energy subsidies; the marginal response of energy consumption to δ satisfies:
    - ∂c_{r,e}/∂δ ÷ ∂c_{p,e}/∂δ = y_r / y_p

### Second-stage of the game: the poor chooses the amount of social spending
- Only a fraction 1−ζ of social expenditures contributes to social infrastructure; ζ captures leakages/inefficiencies (e.g., corruption, delivery losses, perceived return).
- Resource constraint (common pool):
  - δ p_w (λ c_{p,e} + (1−λ) c_{r,e}) + k / (1−ζ) ≤ z + \bar{τ} y (4)
  - where λ is fraction of rich in population and y = λ y_r + (1−λ) y_p
- Using Equation (3) the constraint becomes:
  - δ θ / (1−δ) (1− \bar{τ}) y + k / (1−ζ) ≤ z + \bar{τ} y (5)
- Poor’s maximization (u_p(c_p,k) = ln(c_p) + v_p(k)) yields:
  - ln(c_p) = α_p − θ ln(1−δ) (6)
    - α_p = ln((1− \bar{τ}) y_p / p_w) + θ ln(θ) + (1−θ) ln(1−θ)
  - Lagrangian first-order conditions imply μ_p = (1−ζ) v_p′(k) and the budget binds (μ_p > 0), giving optimal k*:
    - k* = (1−ζ) ( z + \bar{τ} y − δ / (1−δ) θ (1− \bar{τ}) y ) (7)
- Comparative statics:
  - ∂k*/∂δ = − (1−ζ) / (1−δ)^2 θ (1− \bar{τ}) y < 0 (8)
    - A higher subsidy rate reduces public social spending.
  - ∂k*/∂ζ < 0
    - Greater leakages (higher ζ) reduce chosen public social spending.
  - A larger resource endowment (z) or larger fiscal space mitigates the crowding-out effect of subsidies on k*.

### First-stage of the game: the elite sets the subsidy rate
- Rich’s utility, taking k*(δ,ζ) into account:
  - α_r − θ ln(1−δ) + v_r(k*(δ,ζ))
- First-order condition (interior solution) with respect to δ leads to:
  - θ / (1−δ) + (∂k*/∂δ) v_r′(k*) = 0
- Solving yields optimal subsidy rate δ*:
  - δ* = 1 − (1−ζ)(1− \bar{τ}) y v_r′(k*) (9)
- Comparative statics:
  - ∂δ*/∂ζ = (1− \bar{τ}) y v_r′(k*) > 0
    - Worse delivery (higher ζ) raises the elite’s chosen subsidy rate.
  - A higher marginal value of the public good for the rich (v_r′(k*)) leads the rich to choose a lower subsidy rate.
- Taxation insight:
  - If subsidies were contingent on higher income tax (or more progressive taxation), equilibrium would tend toward lower energy subsidies and higher priority social spending via Equations (9), (7), and (8).

### Mechanism and interpretation
- Transmission mechanism:
  - Weak domestic institutions (high ζ) reduce the perceived benefit of public social spending, inducing the rich to choose higher energy subsidies (δ*), which in turn crowd out public social spending k*, especially when fiscal space is limited.
  - The poor may support this equilibrium because subsidies provide a small certain private benefit, while public goods suffer from leakages.
- Model limitations and potential extensions:
  - Utility separability assumption between public and private goods could be relaxed to allow complements or substitutes.
  - Institutions ζ are treated as exogenous; an extension could endogenize ζ (e.g., set ζ = ζ(z)).
  - Heterogeneity beyond income could be introduced by allowing different consumption shares of energy/non-energy goods across groups.

### Empirical analysis (preview of approach and data)
- Goal: identify causal relationship between energy subsidies and public social spending.
- Data:
  - Energy subsidies from Clements and others (2013) using the "price gap" approach; subsidies computed as:
    - s_{e i,t} = Σ_{e=1}^m (p_{e w,t} − p_{e i,t}) c_{e i,t} (10)
    - Pre-tax subsidies are non-negative and censored at 0 when price gap negative.
  - Social spending: sum of public expenditure in education and health (% of GDP) from World Development Indicators (WDI).
- Stylized facts (panel 109 low- and middle-income countries, 2000–11):
  - Between 2002–2006 and 2007–2011:
    - Subsidies decreased in developing and emerging Asia; increased in Sub-Saharan Africa.
    - Social spending moved unevenly and in some regions inversely to subsidies, suggesting a potential trade-off; MENA region did not show a decline in social spending despite higher subsidies.
  - Country-level patterns:
    - Most countries experienced increases in subsidies; social spending changes were mixed.
    - Some resource-rich countries increased subsidies without cutting social spending substantially, indicating fiscal space conditions the crowding-out effect.
- Econometric model (cross-section specification, averages over 2000–11):
  - Social_i = α + (β_1 + β_2 F_i) Subsidies_i + β_3 F_i + X_i′ Γ + ε_i (11)
    - "Social": public spending in education + health (% of GDP)
    - "Subsidies": total pre-tax energy subsidies (% of GDP)
    - X includes demographic controls (dependency ratio, urbanization), initial real per capita income, government size (spending-to-GDP, excluding subsidies), trade openness, macro volatility (std dev of annual real GDP growth), and measures of domestic institutions.
    - F_i contains potential moderators: fiscal space (public debt-to-GDP), natural resource dependency (oil rents-to-GDP), and institutional quality.
  - Institutional quality captured by government effectiveness and corruption indicators from World Governance Indicators, rescaled to [0,1] via min-max transformation.

*Italic: Source — _wp15101 - 2.   Solving the Game (Backward Induction)_*

### 2.   Why Adopt a Cross-sectional Specification?

### 2.   Why Adopt a Cross-sectional Specification?

### Rationale for using cross-sectional averages instead of panel estimations
- Measurement error in the subsidy variable can lead to an attenuation bias in panel estimations with country-fixed effects; year-to-year variations in the subsidies variable often reflect benchmark-price changes or price-stickiness rather than regime shifts, risking identification of the impact of shocks to energy prices on social spending rather than the crowding-out effect of subsidies on social spending.
- The measure of energy subsidies captures implicit subsidies, not necessarily reflected in budgets. Yearly data may be misleading and produce stronger attenuation bias because year-to-year changes in implicit subsidies might be accommodated by SOE losses, arrears accumulation, or public debt rather than cuts in public social spending. Using within-country averages over the analysis period limits this bias as SOEs cannot run losses indefinitely and debt cannot be built up indefinitely without adjustment.
- The short time period (2000–11) and strong inertia in the dependent variable (public social spending) and slow-moving determinants (demographics, institutions, natural resource dependency) limit the information content of the time dimension, exposing panel estimations to the risk that fixed effects absorb almost all variation.

### Appeal of the between-country (cross-sectional) dimension
- Country-specific averages of energy subsidies-to-GDP ratios mostly reflect "subsidy regimes" or "pricing regimes"—the extent to which some countries tend to subsidize energy products more than others.
- Cross-sectional estimates allow assessment of whether countries more prone to subsidizing energy products also spend less on public social spending.

### Key excerpts on numeric features of the data
- Time period of short panel: 2000–11.
- Country sample sizes used in estimations: N 109 and N 108 (see Table 1).

---

### 3.   Identification Strategy: Addressing the Endogeneity of Subsidies

### Sources of endogeneity to address
- Simultaneity: subsidies and social spending may be jointly determined; OLS estimates of β1 and β2 would suffer from a simultaneity bias.
- Omitted variables: cross-sectional estimations cannot automatically control for unobservables jointly affecting subsidy regimes and public social spending.

### Instruments proposed
- Main instrument: average level of subsidies (percent of GDP) in neighbor countries (energy subsidy intensity in neighbor countries, ESINC).
  - Construction: ESINC_i = 1/n ∑_{j=1}^n s_j × d_{i,j}, where s_j is the subsidy-to-GDP ratio in country j and d_{i,j} = { 1 if countries i and j are neighbors; 0 otherwise }.
  - Weighting uses the worldwide gravity database assembled by the Cepii.
  - Identification intuition: spatial spillovers via political economy channels—citizens aware of neighbors’ policies over a 12-year horizon may pressure governments to adopt subsidies if neighbors do, particularly if the country is a net commodity exporter.
  - Empirical note: ESINC alone accounts for one quarter of total variation in energy subsidies across countries and up to half of the variation among net oil exporters (authors’ statement).
- Second instrument: degree of political constraints on the executive (Constraints on the Executive).
  - Intuition: stronger political constraints limit executive discretion to develop and implement energy subsidy policies.
  - Exclusion concern: political constraints might also correlate with other spending categories; authors limit this risk by controlling for overall government size (net of subsidies) and argue public social spending are more traditional budget items than energy subsidies.
- Instrumenting interactions: when specifications include interaction terms (e.g., subsidies × conditional variables), the instruments are expanded to include interaction terms of the two instruments with each conditional variable; the instrument matrix includes additive and interaction terms.

### Instrument relevance and validity testing
- Instrument relevance gauged by the F-stat and Shea R^2 from first-stage regressions.
- Orthogonality (exclusion) tested using the Sargan over-identification test.

---

### 4.   Estimation Results

### Estimation approach
- Two-stage least squares (IV) with robust standard errors.
- Second-stage estimates for public social spending-to-GDP (Panel A); first-stage estimates for energy subsidies-to-GDP (Panel B).
- All first-stage specifications include the full set of control variables as in the second-stage.

### Baseline linear IV estimates (selected coefficients and statistics from Table 1)
- Energy subsidies-to-GDP (second-stage coefficients):
  - Column (1): -0.630** (standard error 0.282)
  - Column (2): -1.473* (standard error 0.764)
  - Column (3): -0.477 (standard error 0.303)
  - Column (4): 0.381 (standard error 0.488)
- Interaction effects (selected):
  - Energy subsidies * Oil rents-to-GDP: 0.055** (standard error 0.022) in Column (2).
  - Energy subsidies * Public debt-to-GDP: -0.022** (standard error 0.011) in Column (3).
  - Energy subsidies * 1[Corruption > 75th perc.]: -1.956* (standard error 1.068) in Column (4).
- Controls with notable coefficients:
  - Government size: 0.182*** (0.028) in Column (1); positive and significant across columns.
  - Energy subsidies intensity in neighbor countries (ESINC): 0.076*** (0.023) in Column (1); significant also in other columns.
  - Constraints on the Executive: -0.245*** (0.083) in Column (1).
- First-stage and diagnostic statistics:
  - Joint significance of subsidies coefficients: P-value 0.031 in Column (1); 0.031 in Column (2); 0.089 in Column (3).
  - F-stat of first-stage: 15.545 (Column 1); 7.287 (Column 2); 9.100 (Column 3); 5.335 (Column 4).
  - Sargan OID test: P-value 0.043 (Column 1); 0.414 (Column 2); 0.507 (Column 3); 0.113 (Column 4).
  - Sample sizes: N 109 (Columns 1–3), N 108 (Column 4).

### Main estimated effects and interpretation
- Average crowding-out effect (baseline, Column 1): about 0.6 (less than unity). The authors interpret this as: public spending in education and health were on average 0.6 percentage point of GDP lower in countries where energy subsidies were 1 percentage point of GDP higher.
- Non-linearities and economic structure:
  - Resource-rich countries: marginal effect of energy subsidies on social spending tends to be significantly lower (interaction with oil rents-to-GDP positive and significant), implying resource-rich countries can to some extent sustain subsidies with less crowding-out.
  - Fiscal space (public debt-to-GDP): interaction estimated negative and significant; crowding-out effect stronger in more indebted countries. Example: crowding-out coefficient rises to about 0.8 for countries with a debt-to-GDP ratio above the 75th percentile of the sample distribution, corresponding to a debt-to-GDP ratio of about 70.
- Institutional quality:
  - Interaction terms between subsidies and governance fragility indicators (corruption, governance ineffectiveness) tend negative; for values above the 75th percentile of governance fragility the crowding-out is estimated to be almost one-to-one.
  - Authors’ summary: marginal effect of energy subsidies on public social spending is larger when governance is particularly fragile—the crowding-out is almost one-to-one in countries with government ineffectiveness above the 75th percentile.

---

### IV.   Conclusion and Policy Implications (key takeaways)
- Empirical evidence supports a negative causal relationship between energy subsidies and public social spending when using IVs (neighbor subsidies intensity and political constraints on the executive).
- Quantitative findings highlighted:
  - Public education and health spending on average 0.6 percentage point of GDP lower when energy subsidies are 1 percentage point of GDP higher.
  - Crowding-out rises to about 0.8 when debt-to-GDP reaches 70 percent (75th percentile benchmark).
  - Crowding-out is almost one-to-one when government ineffectiveness is above the 75th percentile.
- Policy implications:
  - Non-resource-rich countries with narrow fiscal space should move expeditiously with subsidy reform to relax constraints on public social spending.
  - Resource-rich economies will find it challenging to sustain high energy subsidies given mounting social spending pressures and volatile resource revenues; lower international oil prices erode fiscal buffers.
  - The recent sharp drop in global oil prices represents an opportunity to durably reform energy subsidies.
  - Depoliticizing domestic energy pricing—e.g., adopting an automatic pricing mechanism—is recommended as a transition toward fully-deregulated energy prices.

*Source: _wp15101 - 2.   Why Adopt a Cross-sectional Specification? (excerpt).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15101.pdf_