## 3.1 Empirical strategy

## Source details

**Canonical URL:** [3.1 Empirical strategy](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024230-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024230-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024230-print-pdf.pdf.json)

---

### Empirical strategy and identification
- Framework:
  - Staggered Difference-in-Differences (DiD) following Callaway and Sant’Anna (2021) to estimate Average Treatment Effect on the Treated (ATT) with multiple treatment timings.
  - Event-study formulation (Equation 3 in source): approval_ct = Σ_{−L}^{−2} γ_l D_lct + Σ_{0}^{L} γ_l D_lct + X_{c,t∈[−L,−2]} α + ε_ct.
  - Main outcome: log of presidential approval ratings (robustness checks with raw levels and demeaned outcomes).
- Identification assumptions (per Callaway and Sant’Anna (2021)):
  1. Treatment turns on and stays on (staggered adoption; observation period restricted Q2 2007 — Q3 2012).
  2. No treatment anticipation (plausible in Mexico due to prior unfulfilled communications in 2008–09; plausible in Bolivia as decision surprised economic actors).
  3. Generalized propensity score bounded away from one.
- Control group (“not-yet treated”): Argentina, Brazil, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Guatemala, Honduras, Nicaragua, Panama, Paraguay, Peru, United States, Uruguay (United States included as important control for Mexico).
- Parallel trends assessment:
  - Bolivia: parallel trends plausible in macro data (two borderline pre-treatment quarters 2008 Q1 and 2009 Q2 noted).
  - Mexico: parallel trends less clear in macro data; visible pre-treatment differences after Q4 2009 in log approval series; rolling-mean transformations strengthen parallel trends but are not used in main specification due to drawbacks.

### Data and sample
- Balanced quarterly panel for 16 Latin American countries and the United States covering Q3 1998 to Q3 2018; main analysis restricted to Q2 2007 to Q3 2012.
- Presidential approval ratings: Executive Approval Project (EAP), quarterly (1980–2018); primary series approval (percentage of positive ratings, 0–100%).
- Socio-economic and governance variables: International Country Risk Guide (ICRG) rating of PRS Group.
- Treatment timing:
  - Mexico: policy announced December 2009 → treatment starts Q4 2009.
  - Bolivia: policy announced December 2010 → treatment starts Q4 2010.
- Descriptive statistics (pooled before log transform, sample period 2007–2012 unless stated):
  - Sample mean approval: 51%.
  - Minimum approval: 13%.
  - Maximum approval: 84%.
  - Bolivia (Table 2): Presidential approval mean 46.64; Variance 66.82; Min 32.36; Max 60.84. Term duration mean 8.50; Variance 17.02; Min 1.00; Max 16.00.
  - Mexico (Table 2): Presidential approval mean 52.72; Variance 18.97; Min 47.19; Max 60.41. Term duration mean 13.50; Variance 42.17; Min 3.00; Max 24.00.
  - Controls (Table 2): Presidential approval mean 51.15; Variance 231.86; Min 13.23; Max 84.06. Term duration mean 8.87; Variance 26.33; Min 1.00; Max 20.00.
  - Table 3 pre/post means:
    - Bolivia pres. approval: pre-treatment 51.22 → post-treatment 36.83.
    - Controls (Bolivia): 51.10 → 51.27.
    - Mexico: 56.37 → 49.06.
    - Controls (Mexico): 49.09 → 53.21.

### 3.3 Results

### Macro-level empirical findings (staggered DiD and SCM)
- Evidence of political cost after announcement of fossil-fuel subsidy (FFS) removal: declines in approval for up to five quarters.
- Time-varying point estimates (DiD, log approval):
  - Up to −0.27 for Bolivia.
  - Up to −0.19 for Mexico.
- Duration and average ATT:
  - Bolivia: negative effect persists up to five quarters; average ATT of −0.19 over five quarters (textual average result: 19% decrease in presidential approval).
  - Mexico: effect lasts only a couple of quarters; average ATT of −0.15 over two quarters.
- Synthetic Control Method (SCM) results (alternative method, larger magnitude):
  - ATT = 7% for Mexico.
  - ATT = 21% for Bolivia.
- Statistical precision and parallel trends:
  - Coefficients less precisely estimated for Mexico (larger standard errors).
  - Bolivia: plausibility of parallel trends supported (with two borderline pre-treatment quarters noted).
  - Mexico: visible pre-treatment increase in approval in Q2 2009 complicates causal identification; rolling-mean transformations strengthen parallel trends but are not used in main specification.

### Micro-level empirical findings (Mexico; heterogeneity and trust)
- Motivation: subsidy distribution in Mexico is regressive — higher-income households (third and fourth income quartiles) receive a larger share of the subsidy.
- Data: Americas Barometer (LAPOP) survey rounds 2008–2012 for Mexico and Colombia (control); ENIGH merged with subsidy data.
- Micro DiD specification: log-log OLS with Treat × Post interaction; controls include perceived economic situation, corruption perceptions, role of government in improving safety, trust in the president, ideology.
- Pooled micro effect:
  - Approval ratings decrease by approximately 2.3% due to subsidy removal (pooled DiD ≈ −0.023).
- Income-quartile heterogeneous effects (summary from Table 4 and robustness tables):
  - Third and fourth income quartiles (3Q and 4Q): negative and statistically significant Treat×Post coefficients (greater loss in approval among richer households).
  - First and second quartiles (1Q and 2Q): generally less responsive; coefficients often positive.
- Trust in the president moderates effects:
  - Higher trust increases approval across income groups.
  - Three-way interaction (Treat×Post×Trust in president): positive for richest quartiles → higher trust attenuates or can reverse the negative effect of subsidy removal on approval among high-income groups.
- Robustness and placebo tests:
  - Results robust across alternative specifications (raw outcome, demeaned outcome).
  - Placebo tests generally detect no statistical evidence of treatment effects; Nicaragua noted as an exception for parallel trends plausibility but with no robust negative treatment effect.

### Placebo tests and robustness (macro and SCM)
- Placebo DiD: simulating treatment across all countries finds absence of parallel trends for most countries except potentially the United States; for the United States no statistical evidence of a treatment effect is found.
- SCM details:
  - Pre-treatment period for SCM: 2001 Q1 to 2012 Q2.
  - Synthetic Mexico donor weights (highest): Honduras 0.35; Ecuador 0.33; ElSalvador 0.22; Panama 0.09; Argentina 0.01.
  - Synthetic Bolivia donor weights (highest): Ecuador 0.40; Paraguay 0.20; Argentina 0.19; Panama 0.15; United States 0.05; Colombia 0.02.
  - Placebo-in-time (artificial treatment at −10 quarters, 2007 Q2) yields no negative treatment effect.
- Outcome-choice robustness:
  - Alternative outcomes (raw approval, rolling mean k=3, demeaned outcomes) produce qualitatively similar patterns though point estimates and precision vary.

### Theoretical framework and interpretation

### Probabilistic voting model: structure and key equations
- Electorate partitioned into income groups j ∈ J with population shares σ_j.
- Voter utility: W_i = c_i + μ_j H(g), with μ_j group-specific trust in government; H′(g) > 0, H″(g) < 0.
- Government budget: y τ; FFS share α; public good g = (1−α) y τ; average subsidy per voter s = α y τ; group-specific subsidy s_j = a_j σ_j s.
- Voter welfare: W_j(α) = (1−τ) y_j + (a_j/σ_j)(α y τ) + μ_j H((1−α) y τ).
- Voter preferred subsidy α_j defined implicitly via H′ and group parameters.
- Equilibrium α° (Equation (11)) solved from first-order condition of parties maximizing probability of winning given probabilistic voting.
- Key implicit definitions and derivatives presented (Equations (13)–(16) in source).

### Model predictions (Result 1)
- (i) Removing FFS results in a loss of political support.
- (ii) If FFS are regressive and μ_j equal across groups, loss of support is driven by high-income groups (low-income groups may increase support).
- (iii) If low-income groups have sufficiently low trust in government (low μ_j), they can also reduce support after subsidy removal even if subsidies are regressive.
- Interpretation: low trust raises preferred subsidy levels and can cause removal to reduce approval among groups that do not directly benefit from subsidies because they distrust alternative redistribution channels.

### Case narratives and numeric details

### Mexico (chronology and fiscal context)
- Net subsidies persisted; in 2008 represented 1.8% of Mexico’s GDP.
- Subsidy phase-out announced December 2009; gradual monthly price hikes of less than US-$0.1/month started then; full phase-out by 2014.
- Energy reform in 2016; market liberalization phased in starting 2017; new price-smoothing decree in 2019.
- Macro evidence: pre-trends complicate causal identification; micro-level evidence finds average 2% decrease in presidential approval driven by richest income groups.

### Bolivia (chronology and fiscal context)
- Gasoline nominal prices reported at 3.74 bolivianos per liter since 2005 (OLADE, 2023); prices marginally increased since 2000 level of 3.1 bolivianos per liter.
- Subsidy phase-out announced December 2010; Decree (748) set low-octane gasoline from 3.74 up to 6.47 Bolivianos/liter and high-octane from 4.79 to 7.41 Bolivianos; diesel increased from 3.72 to 6.80 Bolivianos/liter.
- Subsidies estimated just above 1 percent of GDP in 2006 and rose to 3 percent in GDP in 2013 (Di Bella et al., 2015).
- Macro evidence: removal clearly identified with average 19% decrease in presidential approval; effects stronger and more persistent than Mexico.

### Regional magnitudes and context
- Fossil fuel subsidies estimated as high as $1 trillion (IEA, 2023b) to $1.3 trillion (Black et al., 2023) in 2022.
- Gasoline represents 46% of road transport fuel demand in the region (IEA, 2023a).
- Latin America spent more than $360 billion in 2022 on fossil-fuel subsidies, equivalent to more than 5% of GDP (Black et al., 2023).
  - Explicit subsidies in LAC region: $56.7 billion.
  - Implicit subsidies: $304.5 billion (Black et al., 2023).

### Policy implications and recommendations
- Recognize moderate but tangible political costs:
  - Subsidy removal is politically costly but costs can be moderate and short-lived (case-dependent: Bolivia vs Mexico).
- Leverage trust-building measures:
  - Higher trust in the president can mitigate or reverse negative approval effects; policy designs should incorporate transparency, communication, and credible commitment on the use of saved fiscal resources.
- Targeted redistribution and clear revenue use:
  - Given regressivity of FFS in many settings, emphasize redirecting savings to progressive and visible public goods or targeted transfers; specifying optimal use of fiscal savings increases public acceptance.
- Sequence and design of reforms matter:
  - Gradual, predictable price adjustments (Mexico’s monthly hikes of less than US-$0.1/month starting Dec. 2009) may yield different political dynamics compared to abrupt large one-off increases (Bolivia’s Dec. 2010 decree raising prices by roughly 60–80 percent for some grades).
- Use frequent approval and trust metrics:
  - Presidential approval ratings provide a timely measure of political repercussions and can guide reform pacing and compensatory measures.

### Limitations noted
- External validity limited by case selection (two cases: Mexico and Bolivia) and data availability.
- Results apply to FFS phase-outs examined; cannot be generalized without caution to other carbon-pricing instruments (carbon taxes, fuel excise taxes) that may have different coverage and behavioral effects.
- Analysis abstracts from the role of how government uses revenue saved from subsidy removal (compensatory or spending choices), which can be decisive for acceptance.

*Italic: Source — IMF Working Paper excerpt: “The Political Economy of Fossil-Fuel Subsidy Removal” (content unit: wpiea2024230-print-pdf).*

### 3.1  Empirical strategy  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12

### 3.1  Empirical strategy

### Chapter and section map
- 3.1  Empirical strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12
- 3.2  Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17
- 3.3  Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19

### Subsidies, voters and trust (Chapter 4)
- 4  Subsidies, voters and trust23
- 4.1  A simple probabilistic voting model of redistributive fossil fuel subsidies . . . . .23
  - 4.1.1  Voter utility and preferred subsidy levels . . . . . . . . . . . . . . . . . . . . . .23
  - 4.1.2  The political environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .24
- 4.2  Empirical evidence: Fuel expenditure, trust, and presidential approval across income groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26
  - 4.2.1  Data and methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26
  - 4.2.2  DiD results from micro data: Heterogeneous effects by income level . . . .28

### Conclusions and appendix
- 5  Conclusions33
- 6  Appendix34
  - 6.1  Data and descriptives  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .34
  - 6.2  Placebo tests: Empirical evidence at the macro level . . . . . . . . . . . . . . .35
  - 6.3  Robustness: Empirical evidence at the macro level  . . . . . . . . . . . . . . . . .35
  - 6.4  Synthetic control estimation as robustness . . . . . . . . . . . . . . . . . . . . . . .36
  - 6.5  Derivation of Equation (11) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41
  - 6.6  Proof of Result 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41
  - 6.7  Robustness: Empirical evidence at the micro level  . . . . . . . . . . . . . . . .42

*https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024230-print-pdf.pdf*

### 6.8  Placebo tests: Empirical evidence at the micro level  . . . . . . . . . . . . . . . . . .44

### 6.8  Placebo tests: Empirical evidence at the micro level

### Introduction and research question
- Removing fossil fuel subsidies (FFS) improves public finances, reduces greenhouse gas emissions, and decreases local air pollutants, yet subsidies remain politically stable and entrenched.
- Key empirical question: How significant are the political costs of subsidy removal for elected officials, and which income groups disapprove most — wealthy elites or middle and low-income voters?
- Scope: Empirical analysis (macro and micro) and theoretical modeling focused on gasoline subsidy removal in Mexico and Bolivia in the early 2010s; micro analysis limited to Mexico due to data availability.
- Contextual facts preserved from source:
  - Fossil fuel subsidies estimated as high as $1 trillion (IEA, 2023b) to $1.3 trillion (Black et al., 2023) in 2022.
  - Gasoline represents 46% of road transport fuel demand in the region (IEA, 2023a).

### Main empirical findings (macro and micro)
- Aggregate political-cost findings:
  - Bolivia: removal is clearly identified in macro data with an average 19% decrease in presidential approval.
  - Mexico: macro-level identification is complicated by pre-trends; micro-level evidence indicates an average 2% decrease in presidential approval driven by the richest income groups.
- Duration and intensity:
  - Effect in Bolivia is stronger and more long-lasting than in Mexico.
  - In Mexico, pre-trends in macro data make causal identification ambiguous; micro-level DiD supports theoretical predictions that the richest income groups drive the approval decline.
- Role of trust:
  - Trust in the president moderates the public response; higher trust can mitigate or even reverse the negative impact of subsidy removal on approval.
- Regional magnitudes and fiscal context:
  - Latin America spent more than $360 billion in 2022 on fossil-fuel subsidies, equivalent to more than 5% of GDP (Black et al., 2023).
  - Explicit subsidies in LAC region: $56.7 billion; implicit subsidies: $304.5 billion (Black et al., 2023).
- Case-specific numeric details retained:
  - Mexico: net subsidies persisted and in 2008 represented 1.8% of Mexico’s GDP. The subsidy phase-out was announced in Dec. 2009; gradual monthly price hikes of less than US-$0.1/month started then; full phase-out by 2014. Energy reform in 2016; market liberalization phased in starting 2017; new price-smoothing decree in 2019.
  - Bolivia: gasoline nominal prices reported at 3.74 bolivianos per liter since 2005 (OLADE, 2023); prices marginally increased since 2000 level of 3.1 bolivianos per liter. The subsidy phase-out was announced in Dec. 2010; Decree (748) set low-octane gasoline from 3.74 up to to 6.47 Bolivianos/liter and high-octane from 4.79 to 7.41 Bolivianos; diesel increased from 3.72 to 6.80 Bolivianos/liter. Subsidies estimated just above 1 percent of GDP in 2006 and rose to 3 percent in GDP in 2013 (Di Bella et al., 2015).
  - Regional examples and timing:
    - Mexico’s Ministry of Finance set fuel prices for several decades until reforms changed the market.
    - Mexico liberalized gasoline market in 2017; earlier smoothing via excise tax variation created subsidy dynamics in 2006-2014.

### Empirical strategy and identification
- Methods:
  - Staggered Difference-in-Differences (DiD) framework following Callaway and Sant’Anna (2021) to estimate Average Treatment Effect on the Treated (ATT) in presence of multiple treatment timings.
  - Event-study formulation as in Equation 3 of the source:
    - approval_ct = Σ_{−L}^{−2} γ_l D_lct + Σ_{0}^{L} γ_l D_lct + X_{c,t∈[−L,−2]} α + ε_ct
  - Main outcome: log of presidential approval ratings (also robustness checks with raw levels and demeaned outcomes referenced in the source).
- Identification assumptions explicitly used (per Callaway and Sant’Anna (2021)):
  1. Treatment turns on and stays on (staggered treatment adoption focused within the presidential term and adjacent same-person terms; observation period restricted from second quarter of 2007 until third quarter of 2012).
  2. No treatment anticipation (plausible in Mexico due to prior unfulfilled communications in 2008-09; plausible in Bolivia as decision surprised economic actors).
  3. Generalized propensity score bounded away from one (regular inference condition).
- Control group composition for staggered DiD:
  - “Not-yet treated” countries include Argentina, Brazil, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Guatemala, Honduras, Nicaragua, Panama, Paraguay, Peru, United States and Uruguay (United States included as an important control for Mexico given economic integration).
- Parallel trends assessment:
  - Bolivia: parallel trends plausible in macro data.
  - Mexico: parallel trends less clear in macro data; pre-trends visible after Q4 2009 in the log approval series. Rolling-mean transformations strengthen parallel trends but have drawbacks (mixing periods, undermining DiD assumptions, contradicting non-anticipation) and are not used in main specification.

### Theoretical framework and micro-level tests (Mexico)
- Theoretical model:
  - Probabilistic voting model à la Lindbeck and Weibull (1987) is used to analyze how trust and income-group preferences shape political equilibria and subsidy lock-in.
  - Model prediction summary:
    - If FFS are regressive and trust in government is equal across income groups, the overall loss in support from subsidy removal is driven by high-income groups decreasing support, while low-income groups increase support.
    - If trust is sufficiently low in low-income groups, these groups may also reduce support after subsidy removal, sustaining political lock-in of subsidies.
- Micro-level empirical tests (Mexico) using survey data and DiD:
  - Hypothesis 1: Decline in approval associated with FFS phase-out is predominantly due to diminished support from high-income groups when FFS are regressive.
  - Hypothesis 2: The adverse effect is exacerbated by lack of trust in the president.
  - Results: Micro-level DiD corroborates theoretical predictions — the of 2% decrease in presidential approval in Mexico is driven by the richest income groups; trust in the president plays a crucial moderating role.

### Policy implications and recommendations (derived from analysis)
- Recognize moderate but tangible political costs:
  - Subsidy removal is politically costly but costs can be moderate and short-lived (case-dependent: Bolivia vs Mexico).
- Leverage trust-building measures to reduce political backlash:
  - Higher trust in the president can mitigate or reverse negative approval effects; policy designs should incorporate transparency, communication, and credible commitment on the use of saved fiscal resources.
- Targeted redistribution and clear revenue use:
  - Given the regressivity of FFS in many settings, emphasize how savings will be redirected to progressive and visible public goods or targeted transfers; prior literature and the source note that specifying optimal use of fiscal savings increases public acceptance.
- Sequence and design of reforms matter:
  - Gradual, predictable price adjustments (as in Mexico’s monthly hikes of less than US-$0.1/month starting Dec. 2009) may yield different political dynamics compared to abrupt large one-off increases (as in Bolivia’s Dec. 2010 decree raising prices by roughly 60-80 percent for some grades).
- Use frequent approval and trust metrics:
  - Presidential approval ratings provide a timely measure of political repercussions and are less susceptible to manipulation than protest indicators; monitoring these metrics can guide reform pacing and compensatory measures.

*Italic: Source — IMF Working Paper excerpt: “The Political Economy of Fossil-Fuel Subsidy Removal”, pages covering 6.8 Placebo tests and preceding sections as provided in the source PDF.*

### 2003. Similarly, the spike in approval ratings in 2010 coincides with the start of Evo Morales’ second

### wpiea2024230-print-pdf - 2003. Similarly, the spike in approval ratings in 2010 coincides with the start of Evo Morales’ second

### Data and sample
- Balanced quarterly panel for 16 Latin American countries and the United States covering Q3 1998 to Q3 2018; analysis restricted to Q2 2007 to Q3 2012 to fit presidential terms.
- Presidential approval ratings from the Executive Approval Project (EAP) (quarterly, 1980–2018); primary series used: approval (percentage of positive ratings, 0-100%).
- Socio-economic and governance country variables from the International Country Risk Guide (ICRG) rating of the PRS Group (monthly indicators, 1984–2023).
- Treatment definition:
  - Mexico: policy announced December 2009 → treatment starts in Q4 2009.
  - Bolivia: policy change announced December 2010 → treatment starts in Q4 2010.
- Descriptive statistics (pooled data before log transformation, sample period 2007–2012 unless stated):
  - Sample mean approval: 51%.
  - Minimum approval: 13%.
  - Maximum approval: 84%.
  - Bolivia (Table 2): Presidential approval mean 46.64; Variance 66.82; Min 32.36; Max 60.84. Term duration mean 8.50; Variance 17.02; Min 1.00; Max 16.00.
  - Mexico (Table 2): Presidential approval mean 52.72; Variance 18.97; Min 47.19; Max 60.41. Term duration mean 13.50; Variance 42.17; Min 3.00; Max 24.00.
  - Controls (Table 2): Presidential approval mean 51.15; Variance 231.86; Min 13.23; Max 84.06. Term duration mean 8.87; Variance 26.33; Min 1.00; Max 20.00.
  - Table 3 means comparison (pre-/post-treatment):
    - Bolivia pres. approval (pre-treatment) 51.22; (post-treatment) 36.83.
    - Controls (Bolivia) 51.10 → 51.27.
    - Mexico 56.37 → 49.06.
    - Controls (Mexico) 49.09 → 53.21.

### Main empirical results (macro-level, staggered DiD)
- Estimation: staggered DiD (Callaway and Sant’Anna, 2021) on log presidential approval; robustness checks with raw approval model, rolling means (k=3), Synthetic Control Method (SCM).
- Key findings:
  - Evidence of political cost after announcement of fossil-fuel subsidy (FFS) removal: declines in approval for up to five quarters.
  - Point estimates (time-varying effects):
    - Up to -0.27 for Bolivia.
    - Up to -0.19 for Mexico.
  - Duration and average ATT:
    - Bolivia: negative effect persists up to five quarters; average ATT of -0.19 over five quarters.
    - Mexico: effect lasts only a couple of quarters; average ATT of -0.15 over two quarters.
  - Magnitude summary:
    - Bolivia experiences a 19% decrease in presidential approval compared to the control group, all else equal (textual average result).
  - SCM results (alternative method, larger magnitude):
    - ATT = 7% for Mexico.
    - ATT = 21% for Bolivia.
- Statistical precision and parallel trends:
  - Coefficients less precisely estimated for Mexico (larger standard errors).
  - Parallel trends plausibility:
    - Bolivia: plausibility of parallel trends confirmed (with two borderline pre-treatment quarters 2008 Q1 and 2009 Q2 noted).
    - Mexico: parallel trends less convincing due to visible pre-treatment differences (e.g., significant approval increase in Q2 2009).
- Placebo tests:
  - Placebo simulations in non-treated countries find no statistical evidence of treatment effects; for most countries pre-treatment parallel trends are absent except potentially the United States (no treatment effect found).

### Heterogeneous effects and micro-level evidence (Mexico)
- Motivating observation: subsidy's distribution is regressive in Mexico — higher-income households (third and fourth income quartiles) receive a larger share of the subsidy.
- Data used for heterogeneity analysis:
  - Americas Barometer (LAPOP) survey rounds 2008–2012 for Mexico and Colombia (control).
  - Mexican National Survey of Household Income and Expenditure (ENIGH) merged with subsidy data.
- Micro DiD specification: log-log OLS with Treat × Post interaction; controls include perceived economic situation, corruption perceptions, role of government in improving safety, trust in the president, ideology.
- Main micro findings:
  - Pooled effect (micro data): approval ratings decrease by approximately 2.3% due to the subsidy removal (reported pooled DiD ~ -0.023).
  - Income-quartile heterogeneous effects (Table 4 summary):
    - Third and fourth income quartiles (3Q and 4Q): negative and statistically significant Treat×Post coefficients (greater loss in approval among richer households).
    - First and second quartiles (1Q and 2Q): generally less responsive; coefficients often positive.
  - Trust in the president moderates effects:
    - Higher trust increases approval across income groups.
    - Three-way interaction (Treat×Post×Trust in president): positive for richest quartiles → higher trust attenuates (or can reverse) the negative effect of subsidy removal on approval among high-income groups.
- Robustness and placebo tests:
  - Results robust across alternative specifications (raw outcome, demeaned outcome).
  - Placebo tests generally detect no statistical evidence of treatment effects; Nicaragua noted as an exception for parallel trends plausibility but with no robust negative treatment effect.

### Theoretical framework: probabilistic voting model (intuition and predictions)
- Model features:
  - Electorate partitioned into income groups j ∈ J with population shares σ_j.
  - Voter utility: W_i = c_i + μ_j H(g) ; μ_j interpretable as group-specific trust in government; H′(g) > 0, H″(g) < 0.
  - Government budget: y τ, split between FFS (share α) and other public goods g = (1−α)yτ; average subsidy per voter s = α y τ; group-specific subsidy s_j = a_j σ_j s.
  - Voter welfare: W_j(α) = (1−τ) y_j + (a_j/σ_j)(α y τ) + μ_j H((1−α) y τ).
  - Voter preferred subsidy level:
    - α_j = 1 − ( H′−1( (a_j μ_j)/(σ_j y τ) ) ).
  - Probabilistic voting with parties A and B; equilibrium subsidy α° determined by weighted preferences and ideological distributions (equation (11)).
- Model implications (Result 1):
  - (i) Removing FFS results in a loss of political support.
  - (ii) If FFS are regressive and μ_j equal across groups, loss of support is driven by high-income groups (low-income groups may increase support).
  - (iii) If low-income groups have suﬃciently low trust in government (low μ_j), they can also reduce support after subsidy removal even if subsidies are regressive.
- Interpretation: low trust raises preferred subsidy levels and can cause removal to reduce approval among groups that do not directly benefit from subsidies (because they distrust alternative redistribution channels).

### Policy implications and conclusions
- Empirical conclusion: removal of regressive FFS is politically costly but effects are moderate and often short-lived; costs are more pronounced and persistent in Bolivia than in Mexico in the studied episodes.
- Mechanisms:
  - Regressivity concentrates subsidy benefits among higher-income groups; these groups drive political backlash when subsidies are removed.
  - Low trust in government amplifies opposition to subsidy removal and can lead low-income groups to oppose removals even when they are net beneficiaries of alternative fiscal policies.
- Policy recommendations derived from findings:
  - Account for distributional incidence when designing subsidy phase-outs: identify which income groups receive the largest share of subsidies.
  - Build and strengthen trust in government and institutions to increase acceptance of subsidy removal; higher trust moderates or can reverse negative popularity effects.
  - Consider targeted compensation or credible redistribution mechanisms that are perceived as effective by low-trust groups to avoid political backlash.
- Limitations noted in the analysis:
  - External validity limited by case selection (two prominent cases: Mexico and Bolivia) and data availability.
  - Results apply to FFS phase-outs examined; cannot be generalized without caution to other carbon-pricing instruments (carbon taxes, fuel excise taxes) that may have different coverage and behavioral effects.
  - The analysis abstracts from the role of how government uses revenue saved from subsidy removal (compensatory or spending choices), which can be decisive for acceptance.

*Italic: Source — IMF Working Paper: "The Political Economy of Fossil-Fuel Subsidy Removal" (content unit: wpiea2024230-print-pdf).*

### 12. The higher the index rating, the lower the risk.

### 12. The higher the index rating, the lower the risk.

### Rating construction and component scales
- The rating is obtained as the sum of three sub-components: unemployment, consumer confidence and poverty.
- Each of these components ranges from 0-4, where 4 equates to a very low risk and 0 to a very high risk.

### ICRG indicators (quarterly)
- Internal conflict (risk rating measure assessing political violence and its actual or potential impact on governance)
  - Rating ranges from 0-12.
  - A rating of 12 equates to no armed or civil opposition, and where the government does not indulge in arbitrary violence against its own people.
  - A rating of 0 equates to on-going civil war.
  - The rating is obtained as the sum of three sub-components: 1) Civil War/Coup Threat, 2) Terrorism/Political violence and 3) Civil Disorder. Each sub-component ranges from 0-4, where 4 equates to a very low risk and 0 to a very high risk.
- Law and order (risk rating)
  - Rating ranges from 0-6 points.
  - Composed of two sub-elements: law (strength and impartiality of the legal system) and order (popular observance of the law).

### LAPOP/Americas Barometer variables (bi-annual)
- Presidential approval
  - Presidential approval of a representative sample of citizens above voting age.
  - Respondents asked: "Generally speaking regarding the current government, ¿Would you say that the job the president [Name] is?" Options: 1) Very good, 2) Good, 3) Neither good, nor bad (regular), 4) Bad, 5) Very bad.
  - The variable has been recoded so that the higher the number, the higher the approval.
- Trust in the president
  - Question: "To what extent do you trust in the President?"
  - Answers range from 1-7, where 1=no trust, 7=high trust.
- Ideology
  - Political ideology self-placement on a scale of 1-10, where 1 represents the political "left" and 10 represents the political "right".
- Government improves safety
  - Scale from 1-7, where 1 represents "nothing" and 7 represents "very much".
  - Question: "To which extend would you agree with the statement that the current administration improves citizen’s safety?"
- Economic situation
  - Question: "How would you judge the economic situation of the country" Options: 1) Very good, 2) Good, 3) Neither good, nor bad (regular), 4) Bad, 5) Very bad.
  - Variable recoded so that the higher the number, the better the economic situation.
- Corruption
  - Question: "Taking into consideration your own experience or what you have heard, the corruption of the public servants in the country is" Options: 1) Very widespread, 2) Somewhat widespread, 3) Little widespread, 4) Not at all widespread.
  - Variable recoded so that the higher the number, the higher (more generalized) the corruption.

### Placebo tests and macro-level robustness
- Placebo test: artificially assign treatment by simulating treatment across all countries in the sample; each country replaces Bolivia in staggered treatment assignment with Mexico.
  - For most countries, except the United States, the absence of parallel trends is notable.
  - For the United States, the assumption of parallel trends could be reasonable yet no statistical evidence of a treatment effect is found.
- Figure 9 (description)
  - Y-axis: DiD estimated effect of removal on presidential approval ratings.
  - X-axis: time (quarters).
  - Dots: DiD estimates per quarter; values below zero denote a negative effect of subsidy removal on presidential approval.
  - Bars: 95% confidence intervals; bars touching the zero y-axis denote insignificant effects.

### Robustness: outcome choice and parallel trends
- Figure 10: DiD results for Bolivia and Mexico using (a) Raw approval model and (b) Rolling mean approval model.
  - Treatment in Bolivia and Mexico denoted by vertical red dotted line in Q4 2010 and Q4 2009 respectively.
  - Point estimates and confidence intervals represented by dots and bars; bars denote 95% confidence intervals.
- Figure 11: Presidential approval: Parallel trends plots for Mexico and Bolivia in raw and rolling mean specifications.
  - Controls represent the average approval ratings for the control countries.

### Synthetic control method (SCM) as robustness
- SCM computes ATT as the post-treatment average difference between approval ratings of Mexico (or Bolivia) and Synthetic Mexico (or Synthetic Bolivia).
- Donor pool: same set of control countries as in the main regression in section 3.3, with Bolivia removed from Mexico's donor pool and viceversa.
- Pre-treatment period used for SCM: 2001 Q1 to 2012 Q2.
- SCM results are consistent with the main results but of larger magnitude.
- SCM path plot assumes treatment starts at 0 when price hikes are announced in Mexico (Dec 2009) and Bolivia (Dec 2010).
- Synthetic Mexico constructed as a weighted average of five countries; highest donor weights:
  - Honduras 0.35
  - Ecuador 0.33
  - ElSalvador 0.22
  - Panama 0.09
  - Argentina 0.01
- Synthetic Bolivia donor weights (highest):
  - Ecuador 0.40
  - Paraguay 0.20
  - Argentina 0.19
  - Panama 0.15
  - United States 0.05
  - Colombia 0.02
- Placebo tests in SCM
  - Placebo in-time: artificial treatment at -10 quarters before treatment (i.e., 2007 Q2) results in no negative treatment effect.
  - Placebos in time assume treatment starts at -10 (placebo). Placebos in space compare outcome gap of treated vs. artificially (placebo) treated countries from the donor pool using the actual treatment date.

### SCM covariate predictor means and weights (selected entries)
- Table 6: Outcome predictor means and weights: Mexico
  - Socio-econ. conditions: Mexico 1.98 | Synthetic Mexico 1.21 | Donor pool 1.52 | Covariate weights 0.00
  - Corruption: 0.79 | 0.90 | 0.94 | 0.07
  - Law and order: 0.87 | 0.81 | 0.99 | 0.16
  - Bureacracy quality: 1.10 | 0.70 | 0.71 | 0.01
  - Duration: 12.33 | 7.84 | 9.10 | 0.00
  - Duration squared: 202.73 | 88.18 | 113.38 | 0.00
  - Internal conflict: 2.19 | 2.20 | 2.18 | 0.00
  - Special predictor -3: Mexico 4.07 | Synthetic Mexico 4.03 | Donor pool 3.64 | Covariate weights 0.34
  - Special predictor -2: 3.86 | 3.90 | 3.59 | 0.12
  - Special predictor -1: 3.99 | 3.83 | 3.74 | 0.06
  - Special predictor -10: 4.10 | 4.03 | 3.86 | 0.06
  - Special predictor -1: 4.01 | 4.03 | 3.96 | 0.17
- Table 7: Outcome predictor means and weights: Bolivia
  - Socio-econ. conditions: Bolivia 1.64 | Synthetic Bolivia 1.51 | Donor pool 1.53 | Covariate weights 0.06
  - Corruption: 0.80 | 0.81 | 0.94 | 0.35
  - Law and order: 1.08 | 1.02 | 0.99 | 0.08
  - Bureacracy quality: 0.69 | 0.68 | 0.71 | 0.16
  - Duration: 5.50 | 7.27 | 9.01 | 0.01
  - Duration squared: 47.14 | 77.95 | 111.88 | 0.01
  - Internal conflict: 2.11 | 2.13 | 2.19 | 0.12
  - Special predictor -16: 4.00 | 3.96 | 3.87 | 0.09
  - Special predictor -22: 3.89 | 3.91 | 3.72 | 0.12
  - Special predictor -26: 4.13 | 3.74 | 3.62 | 0.01

### Analytical derivations and proofs (selected equations)
- Definition and implicit equation for ˆδ:
  - G(αA, αB, δ) := ∑_{j∈J} σ_j B_j( W_j(αA) − W_j(αB) − δ ).
  - G_{αA} = ∑_{j∈J} σ_j b_j(ξ_j) W_{jαA}(αA) and G_δ = −∑_{j∈J} σ_j b_j(ξ_j) ≤ 0.
  - There exists ˆδ(αA, αB) implicitly defined by G(αA, αB, ˆδ) = 1/2. (Equation (13))
  - Probability of winning: Pr[ˆδ ≥ δ] = ψ_{ˆδ} + 1/2. (Equation (14))
  - Implicit differentiation: dˆδ/dαA = − G_{αA} / G_δ = ( ∑_{j∈J} σ_j b_j(ξ_j) W_{jαA}(αA) ) / ( ∑_{j∈J} σ_j b_j(ξ_j) ). (Equation (15))
  - First-order condition yields equation (16) and solving for α yields Equation (11).
- Proof of Result 1: key implications
  - (i) Equilibrium level of FFS results from maximization of expected probability of winning; all deviations reduce expected political support.
  - Vote share for party A from group j: π_{jA} = σ_j B_j( ξ_j(αA, αB) ) = σ_j B_j( W_j(αA) − W_j(αB) − δ ).
  - ∂π_{jA}/∂αA = σ_j b_j(ξ_j) W_{jαA}(αA).
  - W_j(αA) has global maximum at α_j (income group j’s bliss point), so σ_j b_j(ξ_j) W_{jαA}(αA) ⋛ 0 ⇐⇒ αA ⋚ α_j.
  - ∂α_j/∂μ_j < 0 implies if μ_j is sufficiently low, α_j > α^◦_A and ∂π_{jA}/∂αA > 0.

### Micro-level empirical robustness and placebo results (selected table entries)
- Table 10: DiD Robustness: Outcome variable in raw levels (columns denote income quartiles 1Q through 4Q and Pooled)
  - Treat×Post: 0.044 (1Q), 0.154 (2Q), -0.059 (3Q), -0.176 (4Q), -0.027 (Pooled)
    - Standard errors (parentheses): (0.002), (0.001), (0.002), (0.012), (0.005)
  - Trust in president: 0.171, 0.146, 0.165, 0.238, 0.174
    - Std. errors: (0.004), (0.01), (0.007), (0.032), (0.017)
  - Treat: -0.269, -0.305, -0.177, -0.127, -0.195
    - Std. errors: (0.017), (0.001), (0.004), (0.016), (0.005)
  - Post: -0.081, -0.112, -0.051, 0.062, -0.026
    - Std. errors: (0.003), (0.001), (0.002), (0.018), (0.003)
  - Ideology: 0, 0.013, 0.025, 0.041, 0.023
    - Std. errors: (0.006), (0.004), (0.018), (0.004), (0.006)
  - Gov. improves safety: 0.054, 0.089, 0.104, 0.125, 0.095
    - Std. errors: (0.005), (0.011), (0), (0.016), (0.008)
  - Econ. sit. country: 0.145, 0.068, 0.144, 0.142, 0.133
    - Std. errors: (0.072), (0.007), (0.013), (0.013), (0.005)
  - Corruption: -0.042, -0.033, -0.019, 0.011, -0.005
    - Std. errors: (0.019), (0.019), (0.031), (0.009), (0.016)
  - Intercept: 2.297, 2.433, 1.99, 1.294, 1.912
    - Std. errors: (0.241), (0.056), (0.009), (0.185), (0.053)
  - Obs.: 5609 (1Q), 14 (2Q?) — table shows "Obs.560914    1114    603    3603" (as printed)
- Table 11: DiD Robustness: Outcome variable demeaned
  - Treat×Post: 0.104, 0.028, -0.059, -0.05, -0.006
    - Std. errors: (0.004), (0.005), (0.01), (0.001), (0.006)
  - Trust in president: 0.334, 0.2, 0.279, 0.372, 0.28
    - Std. errors: (0.051), (0.001), (0.057), (0.065), (0.049)
  - Treat: 0.058, 0.014, 0.092, 0.141, 0.073
    - Std. errors: (0.017), (0.013), (0.016), (0.04), (0.014)
  - Post: -0.137, -0.029, -0.027, -0.024, -0.039
    - Std. errors: (0.006), (0.006), (0.008), (0.03), (0.007)
  - Ideology: 0.043, 0.05, 0.037, 0.12, 0.066
    - Std. errors: (0.015), (0), (0.042), (0.038), (0.003)
  - Govt. improves safety: 0.116, 0.165, 0.17, 0.23, 0.179
    - Std. errors: (0.028), (0.013), (0.005), (0.03), (0.005)
  - Econ. sit. country: 0.199, 0.06, 0.166, 0.25, 0.154
    - Std. errors: (0.031), (0.051), (0.035), (0.13), (0.011)
  - Corruption: -0.115, -0.069, 0.015, 0.065, -0.011
    - Std. errors: (0.069), (0.005), (0.075), (0.088), (0.034)
  - Intercept: 0.127, 0.369, 0.071, -0.48, 0.028
    - Std. errors: (0.046), (0.065), (0.137), (0.461), (0.141)
  - Obs.: "560913    1112    600    3594" (as printed)
- Table 12: DiD results Placebo Nicaragua (selected entries)
  - Treat×Post (various columns): 0.206, 0.194, 0.086, -0.044, 0.056, 0.089, 0.028, -0.260, 0.072, 0.027
    - Std. errors: (0.028), (0.003), (0.04), (0.042), (0.023), (0.007), (0.008), (0.014), (0.022), (0.016)
  - Treat: -0.24, -0.306, -0.156, -0.12, -0.102, -0.117, -0.080, 0.226, -0.116, -0.08
    - Std. errors: (0.034), (0.015), (0.035), (0.066), (0.01), (0.03), (0.031), (0.034), (0.008), (0.035)
  - Post: -0.028, -0.056, -0.004, 0.127, 0, 0.004, 0.007, 0.307, -0.003, 0.067
    - Std. errors: (0.001), (0.014), (0.008), (0.023), (0.004), (0.001), (0.002), (0.012), (0.003), (0.006)
  - Trust in president: 0.183, 0.137, 0.159, 0.193, 0.212, 0.214, 0.231, 0.419, 0.20, 0.233
    - Std. errors: (0.004), (0.03), (0.017), (0.015), (0.014), (0.003), (0.051), (0.006), (0.025), (0.011)
  - Treat×Trust in president: 0.035, -0.013, 0.02, -0.211, -0.016 with std. errors (0.002), (0.034), (0.016), (0.034), (0.022) (as printed)
  - Post×Trust in president: 0.015, -0.086, -0.002, -0.197, -0.046 with std. errors (0.006), (0.01), (0.003), (0.004), (0.001) (as printed)
  - Treat×Post×Trust in president: 0.020, 0.08, -0.036, 0.191, 0.021 with std. errors (0.009), (0.008), (0.012), (0.004), (0.004) (as printed)
  - Ideology (various): -0.032, -0.03, -0.039, -0.035, -0.033, -0.035, -0.044, -0.053, -0.033, -0.033
    - Std. errors: (0.015), (0.016), (0.036), (0.039), (0.039), (0.042), (0.053), (0.051), (0.039), (0.041)
  - Gov. improves safety (various): 0.130, 0.129, 0.154, 0.154, 0.128, 0.128, 0.196, 0.202, 0.150, 0.151
    - Std. errors: (0.038), (0.038), (0.052), (0.051), (0.015), (0.015), (0.035), (0.03), (0.034), (0.033)
  - Econ sit. country (various): 0.140, 0.139, 0.120, 0.119, 0.142, 0.143, 0.197, 0.197, 0.157, 0.158
    - Std. errors: (0.093), (0.093), (0.082), (0.083), (0.036), (0.037), (0.005), (0.004), (0.046), (0.046)
  - Corruption (various): -0.023, -0.021, -0.022, -0.023, -0.012, -0.014, 0.010, 0.007, -0.009, -0.011
    - Std. errors: (0.003), (0.004), (0.008), (0.008), (0.003), (0.005), (0.024), (0.024), (0.011), (0.012)
  - Intercept (various): 0.774, 0.847, 0.796, 0.738, 0.714, 0.714, 0.514, 0.238, 0.678, 0.628
    - Std. errors: (0.108), (0.064), (0.063), (0.059), (0.03), (0.02), (0.143), (0.066), (0.02), (0.008)
  - Obs.: "97097015721572237223721485148569006900" (as printed)

_Italic: Source — IMF Working Paper chapter: The Political Economy of Fossil-Fuel Subsidy Removal (excerpts provided)._

### References

### References

### Methodology and Econometric Methods
- Abadie, A., Diamond, A., and Hainmueller, J.  Comparative Politics and the Synthetic Control Method.American Journal of Political Science., 59(2):495–510, 2015.
- Abadie, A. Using synthetic controls: Feasibility, data requirements, and methodological aspects. Journal of Economic Literature, 59(2):391–425, 2021.
- Abadie, A. and Gardeazabal, J.  The Economic Costs of Conflict: A Case Study of the Basque Country.American Economic Review., 93(1):113–32, 2003.
- Abadie, A. and Gardeazabal, J. Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California’s Tobacco Control Program.Journal of the American Statistical Association., 105:494–505, 2010.
- Callaway, B. and Sant’Anna, P. H. Difference-in-differences with multiple time periods.Journal of Econometrics, 225(2):200–230, 2021.
- Sun, L. and Abraham, S. Estimating dynamic treatment effects in event studies with heterogeneous treatment effects.Journal of Econometrics, 225(2):175–199, 2021.

### Political Economy, Political Behavior, and Electoral Accountability
- Alesina, A. and Tabellini, G. A positive theory of fiscal deficits and government debt.The review of economic studies, 57(3):403–414, 1990.
- Becker, G. S. A theory of competition among pressure groups for political influence.The quarterly journal of economics, 98(3):371–400, 1983.
- Berlemann, M. and Enkelmann, S. The economic determinants of us presidential approval: A survey. European Journal of Political Economy, 36:41–54, 2014.
- Berlemann, M., Enkelmann, S., and Kuhlenkasper, T. Unraveling the relationship between presidential approval and the economy: A multidimensional semiparametric approach.Journal of Applied Econometrics, 30(3):468–486, 2015.
- Boix, C.Democracy and redistribution.Cambridge University Press., 2003.
- Clarke, H. D. and Stewart, M. C. Prospections, retrospections, and rationality: The ”bankers” model of presidential approval reconsidered.American Journal of Political Science., 38(4):1104–1123, 1994.
- Commander, S. A guide to the political economy of reforming.IZA Policy Paper No. 52., 2012.
- Fox, G. and Phillips, E. N. Interrelationship between presidential approval, presidential votes and macroeconomic performance, 1948–2000.Journal of Macroeconomics, 25(3):411–424, 2003.
- Harring, N., Jönsson, E., Matti, S., Mundaca, G., and Jagers, S. C.  Cross-national analysis of attitudes towards fossil fuel subsidy removal.Nature Climate Change, 13(3):244–249, 2023.
- Hodler, R., Luechinger, S., and Stutzer, A. The effects of voting costs on the democratic process and public finances.American Economic Journal: Economic Policy, 7(1):141–71, 2015.
- Jung, J. W. and Oh, J. Determinants of presidential approval ratings: Cross-country analyses with reference to Latin America.International Area Studies Review, 23(3):251–267, 2020.
- Kojima, M. Petroleum product pricing and complementary policies: experience of 65 developing countries since 2009.World Bank Policy Research Working Paper, 6396, 2013.
- Kone, S. L. and Winters, R. F. Taxes and Voting: Electoral Retribution in the American States. The Journal of Politics, 55(1):22–40, 1993.
- Lindbeck, A. and Weibull, J. W.  Balanced-budget Redistribution as the Outcome of Political Competition.Public Choice, 52(3):273–297, 1987.
- List, J. A. and Sturm, D. M. How elections matter: Theory and evidence from environmental policy. The Quarterly Journal of Economics, 121(4):1249–1281, 2006.
- Lowry, R. C., Alt, J. E., and Ferree, K. E. Fiscal policy outcomes and electoral accountability in american states.American Political Science Review, 92(4):759–774, 1998. doi: 10.2307/2586302.
- Marzilli Ericson, K. M. and Fuster, A. The endowment effect.Annu. Rev. Econ., 6(1):555–579, 2014.
- Niemi, R. G., Stanley, H. W., and Vogel, R. J. State economies and state taxes: Do voters hold governors accountable?American Journal of Political Science, 39(4):936–957, 1995. ISSN 00925853, 15405907. URLhttp://www.jstor.org/stable/2111664.
- Nurdianto, D. A. and Resosudarmo, B. P. The economy-wide impact of a uniform carbon tax in asean.Journal of Southeast Asian Economies, pages 1–22, 2016.
- Oates, W. E. and Portney, P. R. The political economy of environmental policy. InHandbook of environmental economics, volume 1, pages 325–354. Elsevier, 2003.
- Persson, T. and Tabellini, G. E.Political economics: explaining economic policy. MIT press, 2002.
- Stimson, J. A. Public support for american presidents: A cyclical model.Public Opinion Quarterly, 40(1):1–21, 1976.
- Tribin, A. Chasing votes with the public budget.European Journal of Political Economy, 63:101875, 2020.
- Van de Graaf, T. and Blondeel, M. Fossil fuel subsidy reform.The politics of fossil fuel subsidies and their reform, page 83, 2018.
- Victor, D. G. The politics of fossil-fuel subsidies., 2009. URLhttp://dx.doi.org/10.2139/ssrn.1520984.

### Energy Subsidies, Reform Experience, and Distributional Effects
- Alleyne, T., Clements, B., Coady, D., Fabrizio, S., Gupta, S., Sdralevich, C., Shang, B., and Villafuerte, M. Reforming energy subsidies: lessons from experience. In Clements, B., Coady, D., Fabrizio, S., Gupta, S., Alleyne, T., and Sdralevich, C., editors,Energy Subsidy Reform. International Monetary Fund, 2013.
- Agnolucci, P., Fischer, C., Heine, D., Montes de Oca, M., Patroni, K., Pryor, J., and Hallegate, S. Measuring total carbon pricing.World Bank Policy Research Working Papers, 2023.
- Brenner, M., Riddle, M., and Boyce, J. K. A Chinese sky trust?: Distributional impacts of carbon charges and revenue recycling in China.Energy Policy, 35(3):1771–1784, 2007.
- Davis, L. W.  The economic cost of global fuel subsidies.American Economic Review, 104(5):581–585, 2014.
- Del Granado, F. J. A., Coady, D., and Gillingham, R. The unequal benefits of fuel subsidies: A review of evidence for developing countries.World Development, 40(11):2234–2248, 2012.
- Di Bella, M. G., Norton, M. L., Ntamatungiro, M. J., Ogawa, M. S., Samake, I., and Santoro, M.Energy subsidies in Latin America and the Caribbean: Stocktaking and policy challenges. International Monetary Fund, 2015.
- Dorband, I. I., Jakob, M., Kalkuhl, M., and Steckel, J. C. Poverty and distributional effects of carbon pricing in low-and middle-income countries–a global comparative analysis.World Development, 115:246–257, 2019.
- Jakob, M., Chen, C., Fuss, S., Marxen, A., and Edenhofer, O. Development incentives for fossil fuel subsidy reform.Nature Climate Change, 5(8):709–712, 2015.
- Kojima, M. Petroleum product pricing and complementary policies: experience of 65 developing countries since 2009.World Bank Policy Research Working Paper, 6396, 2013.
- Kyle, J. Local corruption and popular support for fuel subsidy reform in indonesia.Comparative Political Studies, 51(11):71–88, 2018.
- Munoz-Pina, C., Montes de Oca, M., and Rivera-Planter, M. From Negative to Positive Carbon Pricing in Mexico.Economics of Energy and Environmental Policy., 11(2):113–32, 2022.
- Parry, I., Black, S., and Vernon, N.Still not getting energy prices right: A global and country update of fossil fuel subsidies. IMF Working Paper, 2021.
- Renner, S. Poverty and distributional effects of a carbon tax in mexico.Energy Policy, 112:98–110, 2018.
- Sdralevich, M. C. A., Sab, M., Zouhar, M. Y., and Albertin, M. G.Subsidy reform in the Middle East and North Africa: Recent progress and challenges ahead.  International Monetary Fund, 2014.   URL\url{https://www.imf.org/-/media/Websites/IMF/imported-publications/external/pubs/ft/dp/2014/_1403mcdsum.ashx}.
- Sterner, T.Fuel Taxes and the Poor: The Distributional Effects of Gasoline Taxation and Their Implications for Climate Policy. RFF Press with Environment for Development initiative., 2012.
- Sterner, T. Fuel taxes: An important instrument for climate policy.Energy policy, 35(6):3194–3202, 2007.
- Strand, J. Political economy aspects of fuel subsidies: a conceptual framework.World Bank Policy Research Working Paper, 6392, 2013.
- Van Heerden, J., Gerlagh, R., Blignaut, J., Horridge, M., Hess, S., Mabugu, R., and Mabugu, M. Searching for triple dividends in South Africa: Fighting CO2 pollution and poverty while promoting growth.The Energy Journal, 27(2), 2006.
- Vagliasindi, M.  Implementing energy subsidy reforms. overview of the key issues.World Bank, Policy Research Working Paper(6122), 2013.
- WEF.Lessons  Drawn  From  Reforms  of  Energy  Subsidies.World  Economic Forum, 2013.URL\url{https://www3.weforum.org/docs/GAC13/WEF_GAC13_LessonsReformsEnergySubsidies_Report.pdf}.

### Carbon Pricing, Climate Mitigation, and Global Energy Outlooks
- Alesina, A. and Tabellini, G. A positive theory of fiscal deficits and government debt.The review of economic studies, 57(3):403–414, 1990.
- Carattini, S., Kallbekken, S., and Orlov, A. How to win public support for a global carbon tax. Nature, 565(7739):289–291, 2019.
- Dorband, I. I., Jakob, M., Kalkuhl, M., and Steckel, J. C. Poverty and distributional effects of carbon pricing in low-and middle-income countries–a global comparative analysis.World Development, 115:246–257, 2019.
- Munoz-Pina, C., Montes de Oca, M., and Rivera-Planter, M. From Negative to Positive Carbon Pricing in Mexico.Economics of Energy and Environmental Policy., 11(2):113–32, 2022.
- IEA.World Energy Outlook (WEO). International Energy Agency, 2015.
- IEA.   Latin america energy outlook.2023,https://iea.blob.core.windows.net/assets/1055131a-8dc4-488b-9e9e-7eb4f72bf7ad/LatinAmericaEnergyOutlook.pdf, 2023a.
- IEA.   Fossil fuels consumption subsidies 2022, iea.2023,https://www.iea.org/reports/fossil-fuels-consumption-subsidies-2022, 2023b.
- Overland, I. Subsidies for fossil fuels and climate change: a comparative perspective.International Journal of Environmental Studies., 67(10):303–317, 2010.
- Renner, S. Poverty and distributional effects of a carbon tax in mexico.Energy Policy, 112:98–110, 2018.
- Nurdianto, D. A. and Resosudarmo, B. P. The economy-wide impact of a uniform carbon tax in asean.Journal of Southeast Asian Economies, pages 1–22, 2016.

### Country Case Studies, Regional Analyses, and Institutional Reports
- Black, M. S., Kirabaeva, K., Parry, I., Raissi, M. M., and Zhunussova, K.A Comprehensive Climate Mitigation Strategy for Mexico. International Monetary Fund, 2021.
- Black, S., Liu, A. A., Parry, I. W., and Vernon, N. IMF Fossil Fuel Subsidies Data: 2023 Update. IMF Working Paper, 2023/169, 2023.
- Di Bella, M. G., Norton, M. L., Ntamatungiro, M. J., Ogawa, M. S., Samake, I., and Santoro, M.Energy subsidies in Latin America and the Caribbean: Stocktaking and policy challenges. International Monetary Fund, 2015.
- Hallegate, S., Godinho, C., Rentschler, J., Avner, P., Dorband, I., Knudsen, C., Lemke, J., and Mealy, P. Within Reach: Navigating the Political Economy of Decarbonization.Climate Change and Development Series, Washington DC, 2024.
- ILOADE. Sistema de Información Energética de Latinoamérica y el Caribe, 2023. URL\url{https://sielac.olade.org}.
- IMF.Mexico:2010    Article    IV    Consultation-Press    Release    and Staff    Report.International   Monetary   Fund.   Country   Report.   No. 10/71.,https://www.imf.org/en/Publications/CR/Issues/2016/12/31/Mexico-2010-Article-IV-Consultation-Staff-Report-and-Public-Information-Notice-on-the-23711, 2010.
- IMF.Bolivia:2011    Article    IV    Consultation-Press    Release    and Staff    Report.International   Monetary   Fund.   Country   Report.   No. 11/124.,https://www.imf.org/en/Publications/CR/Issues/2016/12/31/Bolivia-2011-Article-IV-Consultation-Staff-Report-Supplement-and-Staff-Statement-Public-24896, 2011a.
- IMF. Mexico: 2011 Article IV Consultation-Press Release and Staff Report.International Monetary Fund. Country Report. No. 11/250.,https://www.imf.org/en/Publications/CR/Issues/2016/12/31/Mexico-Staff-Report-for-the-2011-Article-IV-Consultation-25146, 2011b.
- IMF. Bolivia: 2011 article iv consultation.International Monetary Fund,https://www.imf.org/external/pubs/ft/scr/2011/cr11124.pdf, 2011c.
- IMF.Mexico:2022    Article    IV    Consultation-Press    Release    and Staff    Report.International   Monetary   Fund.   Country   Report.   No. 22/334.,https://www.imf.org/en/Publications/CR/Issues/2022/11/04/Mexico-2022-Article-IV-Consultation-Press-Release-and-Staff-Report-525448, 2022.
- IPSOS.  Aprobacion Presidencial Bolivia. Estudio Multicliente de Opinion Publica.Aprobacion Presidencial 2006-2019. IPSOS, page 10, 2020.
- INEGI. Encuesta nacional de ingreso y gasto de los hogares (enigh).I4CE,https://www.i4ce.org/download/global-carbon-account-2019/, 2022.
- INCHAUSTE, G. and Victor, D. G.The political economy of energy subsidy reform. Washington, DC: World Bank, 2017.
- World Bank. Program document for a proposed first development policy loan.Report No. 50152-DO, hhttp://web.worldbank.org/archive/website01506/WEB/IMAGES/501520PG.PDF, 2009.
- Sdralevich, M. C. A., Sab, M., Zouhar, M. Y., and Albertin, M. G.Subsidy reform in the Middle East and North Africa: Recent progress and challenges ahead.  International Monetary Fund, 2014.   URL\url{https://www.imf.org/-/media/Websites/IMF/imported-publications/external/pubs/ft/dp/2014/_1403mcdsum.ashx}.

### Public Opinion, Protests, Media Accounts, and Social Unrest
- BBC News Mundo.  El precio social y político del ”gasolinazo” en Bolivia.Veronica Smink, BBC Mundo, Cono Sur,https://www.bbc.com/mundo/noticias/2010/12/101227_bolivia_combustible_repercusion_smink, 2010.
- El Economista. Hacienda hace oficial el ”gasolinazo”.Carolina Ruiz / El Economista,https://www.eleconomista.com.mx/economia/Hacienda-hace-oficial-el-gasolinazo--20091230-0035.html, 2009.
- El Pais.   Violent protests force Bolivian U-turn on removing fuel subsidies.El Pais. Jan 5, 2011,https://english.elpais.com/elpais/2011/01/05/inenglish/1294208446_850210.html, 2011.
- La   Republica.Vicepresidente   de   bolivia   se   casa   por   rito   aymara. September8,2012,https://www.larepublica.ec/blog/2012/09/08/vicepresidente-de-bolivia-se-casa-por-rito-aymara/, 2012.
- Douenne, T. and Fabre, A. Yellow vests, pessimistic beliefs, and carbon tax aversion.American Economic Journal: Economic Policy, 14(1):81–110, 2022.
- McCulloch, N., Natalini, D., Hossain, N., and Justino, P. An exploration of the association between fuel subsidies and fuel riots.World Development, 157:105935, 2022.
- Drabo, A., Eklou, K., Imam, P. A., and Kpodar, K. Social unrests and fuel prices: The role of macroeconomic, social and institutional factors.IMF Working Paper, 2023.
- Douenne, T. and Fabre, A. Yellow vests, pessimistic beliefs, and carbon tax aversion.American Economic Journal: Economic Policy, 14(1):81–110, 2022.
- Cheon, A., Urpelainen, J., and Lackner, M. Why do governments subsidize gasoline consumption? An empirical analysis of global gasoline prices, 2002 – 2009.Energy Policy, 56(C):382–390, 2013.
- Fairbrother, M., Sevä, I. J., and Kulin, J. Political trust and the relationship between climate change beliefs and support for fossil fuel taxes: Evidence from a survey of 23 European countries.Global Environmental Change, 59:102003, 2019.
- Finnegan, J. J. Changing prices in a changing climate: Electoral competition and fossil fuel taxation. Comparative Political Studies, 56(8):1257–1290, 2023.
- Romero, V.  Impacto de los temas de seguridad pública en aprobación presidencial.Política y Gobierno, volumen temático., pages 139–160., 2012.
- Romero, V., Magaloni, B., and Díaz-Cayeros, A. Presidential approval and public security in mexico’s war on crime.Latin American Politics and Society, 58(2):100–123, 2016.
- Kyle, J. Local corruption and popular support for fuel subsidy reform in indonesia.Comparative Political Studies, 51(11):71–88, 2018.

### Behavioral Economics, Endowments, and Perception
- Marzilli Ericson, K. M. and Fuster, A. The endowment effect.Annu. Rev. Econ., 6(1):555–579, 2014.
- Douenne, T. and Fabre, A. Yellow vests, pessimistic beliefs, and carbon tax aversion.American Economic Journal: Economic Policy, 14(1):81–110, 2022.
- Fairbrother, M., Sevä, I. J., and Kulin, J. Political trust and the relationship between climate change beliefs and support for fossil fuel taxes: Evidence from a survey of 23 European countries.Global Environmental Change, 59:102003, 2019.

*The Political Economy of Fossil Fuel Subsidy Removal: Evidence from Bolivia and Mexico — References (Working Paper No. WP/24/230).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024230-print-pdf.pdf_
