## 1. Real Gasoline Prices, Oil Costs, and Elections: Basic Model

## Source details

**Canonical URL:** [1. Real Gasoline Prices, Oil Costs, and Elections: Basic Model](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp06260.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp06260.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp06260.pdf.json)

---

### I. Introduction — core hypothesis and empirical scope
- Core hypothesis: politicians set regulated prices to maximize electoral support by signaling pro-consumer behavior, producing lower real prices immediately preceding elections.
- Theoretical synthesis: combines regulatory capture perspective (Stigler (1971); Peltzman (1976)) with signaling-game political business cycle ideas (Paiva (1996); Rogoff and Sibert (1988); Rogoff (1990); Nordhaus (1975)).
- Empirical test: quarterly data from 32 industrial and developing countries over 1978–2004; evidence points to electoral price cycles in gasoline markets.

### II. Model of political price cycles in a regulated industry — setup and equilibrium
- Agents and preferences:
  - Two discrete regulator types: pro-consumer (α) and pro-industry (ᾱ), with α, ᾱ ∈ Α and Prob(α) = μ.
  - Social welfare in regulated market: Wt(α) = (1 − μ) V(p(α)) + μ π(p(α)).
  - Incumbent maximizes intertemporal objective: Φ = Σ_{i=0}^{1} β^{i} [ W_{t+i}(α) + σ R_{t+i}(·) ], with discount β < 1 and reelection weight σ.
- Technology and preferences (functional forms and assumptions preserved):
  - Firm profit: πt = pt qt − C(qt), with Cq > 0 and Cqq > 0.
  - Consumer utility: Ut = I t φ(qt) + ηt, with φ' > 0 and φ'' < 0.
  - ηt follows: ηt = ηt−1 + st with st iid, unimodal, twice continuously differentiable, mean zero (as stated: ηt follows 1− +=− t o t I t s s η η in source notation).
  - Demand: q = q(p).
- Timing (chronology reproduced):
  - At beginning of period t government observes αt, sets price pt.
  - After price set, both government and consumers observe popularity shock st, election happens.
  - Winner takes office for t+1 until t+2; new preference shock αt+2 occurs at start of t+2.
- Signaling equilibrium characterization:
  - Voters’ beliefs map observed price to inferred type: p̃t → α̃t.
  - Nonelectoral years: incumbents set welfare-maximizing price p*(α).
  - Electoral years: incumbents choose p to maximize [W + σ R].
  - Distortion defined: λα(p) = W(p*(α)) − W(p).
  - Separating equilibrium (propositions preserved):
    - Pro-industry type sets p*(ᾱ) with zero distortion.
    - Pro-consumer type sets p*(α) − positive distortion (i.e., lowers price) to signal type.
  - Comparative statics (as stated):
    - Greater probability of lower type winning relative to higher type increases distortion by lower type.
    - Greater weight σ on reelection relative to consumer welfare increases distortion.
  - Model implication: political price cycle occurs only when pro-consumer regulator is in office (probability 1−μ); continuum-of-types extension would generate cycles of varying intensity.

### III. Empirical evidence — data, definitions, and main econometric findings
- Data and variables:
  - Sample period: 1978–2004.
  - Number of countries: 32.
  - Maximum pool observations: 2,915 (reported regressions use 2,913 or 2,937 in various specs).
  - Quarterly series: domestic gasoline prices, consumer price indexes, exchange rates against U.S. dollar.
  - Additional series: average international oil prices (U.S. dollars) and dummy per country for parliamentary/congress elections (legislative elections only).
  - Sources: OECD Energy Prices and Taxes Database; IMF International Finance Statistics; IMF Commodity Price System Database; Election Results Archive.
- Key constructed variables (preserve exact formulations):
  - gasprofit = {[(1+ cpgas) ÷ (1+ cpoildomcur)] – 1} * 100, where cpgas = % change in gasoline prices, cpoildomcur = % change in international oil price in domestic currency.
  - lrgas = ln(domestic gasoline price) − ln(domestic CPI).
  - ldompoil = ln(real oil price in domestic currency).
  - lrer = ln(exchange rate) − ln(domestic CPI).
  - lpoilusd = ln(international oil prices in U.S. dollars) used in some specifications.
- Descriptive statistics (preserved):
  - Pooled sample (1978–2004): gasprofit increased at an average of 1.4 percent during “normal” quarters and declined by 0.4 percent in periods with electoral activity.
  - In 25 of 32 countries, average change in gasprofit during electoral quarters was lower than during “normal” quarters.
  - Average change in gasprofit during electoral quarters negative in 23 countries; negative during non-election quarters in only 7 countries.
  - Real gasoline prices: decline of 0.3 percent on average during “normal” quarters and about 0.7 percent during electoral campaign quarters.
  - In 15 countries the difference exceeded 2 percentage points; in 7 countries it exceeded 6 percentage points.
- Econometric specifications (exact forms preserved):
  - Basic panel specification (equation (1)):
    - clrgas_{n,t} = Σ_{i} γ_{i} cldompoil_{n,t-i} + δ Delec_{n,t} + α_{n} + τ_{t} + ε_{n,t}
    - Delec_{n,t} = 1 if election in current or next quarter, 0 otherwise.
  - Alternative specification separating international oil price and real exchange rate (equation (2)):
    - clrgas_{n,t} = Σ λ_{i} clpoilusd_{t-i} + Σ μ_{i} clrer_{n,t-i} + δ Delec_{n,t} + α_{n} + τ_{t} + ε_{n,t}
- Main estimation results (numeric findings preserved):
  - Contemporaneous coefficient on cldompoil_t ≈ 0.156 to 0.225 (statistically significant).
  - Electoral dummy δ ≈ −0.009 or −0.008 (t-ratios between about −1.93 and −2.24) in basic specs, implying reduction in real gasoline prices of about 0.8–0.9 percent during election periods.
  - Modified basic model (country-specific coefficients on oil cost lags): Delec coefficients range from −0.009** to −0.010** (t-ratios about −2.20 to −2.66), robust and significant.
  - Alternative specification separating international oil price and real exchange rate: Electoral dummy δ remains significant; point estimates suggest real gasoline price declines by at least 0.7 percent and up to 1 percent, on average, during quarters of election campaign across the 32 countries.
  - Preferred specification (period fixed effects) indicates suppliers/regulators pass onto domestic gasoline prices about 16 percent of contemporaneous changes in domestic oil costs.
  - Robustness: only first lag reached minimum significance; results stable across panel estimators (LS, GLS), fixed/random effects choices, and country-specific coefficient specifications.

### IV. Concluding remarks — theoretical and policy implications
- Theoretical contribution:
  - A signaling-game model where temporary information asymmetry about regulator type and electoral incentives generates political price cycles in regulated industries.
  - Pro-consumer types lower regulated prices in election periods as an equilibrium signaling strategy; pro-industry types do not mimic due to welfare cost constraints.
- Empirical contribution:
  - Quarterly panel evidence for 32 countries (1978–2004) supports the model: statistically and economically significant reductions in real gasoline prices during the two quarters of electoral campaign.
  - Point estimates indicate reductions of at least 0.8 percent and up to 1 percent during election quarters, versus average quarterly reduction of 0.3 percent in the full sample period.
  - Results robust across model specifications, country-specific coefficients, and alternative cost decompositions.
- Policy-relevant implications (preserved):
  - Political incentives can cause temporary distortions in regulated prices, especially for visible consumer goods like gasoline.
  - Magnitude of distortions linked to incumbent’s reelection weight σ and relative chances of types winning (probabilities R).
  - Expect heterogeneity: not all countries or periods exhibit political price cycles; where present effects can be substantial (differences > 2 percentage points in 15 countries; > 6 percentage points in 7 countries).

### Appendix 1 — list of countries included in the empirical work (sample composition)
- Number of countries: 32
- Countries:
  - Australia
  - Austria
  - Belgium
  - Brazil
  - Canada
  - Switzerland
  - Denmark
  - Finland
  - France
  - Germany
  - Greece
  - India
  - Indonesia
  - Ireland
  - Italy
  - Japan
  - Korea
  - Mexico
  - Netherlands
  - New Zealand
  - Norway
  - Poland
  - Portugal
  - Romania
  - Slovak Republic
  - South Africa
  - Spain
  - Sweden
  - Thailand
  - Turkey
  - United Kingdom
  - United States

*Source: _wp06260 - 1. Real Gasoline Prices, Oil Costs, and Elections: Basic Model — PDF chapter/section*

### 1. Real Gasoline Prices, Oil Costs, and Elections: Basic Model .............................................16

### 1. Real Gasoline Prices, Oil Costs, and Elections: Basic Model

### Overview of sections in this content unit
- "1. Real Gasoline Prices, Oil Costs, and Elections: Basic Model" — page 16
- "2. Real Gasoline Prices, Oil Costs, and Elections: Modified Basic Model" — page 17
- "3. Real Gasoline Prices, Oil Costs, and Elections: Alternative Model" — page 18

### Figures and appendices referenced within the unit
- Figure: "1. Chronological Order of Events" — page 8
- Appendix: "1. List of Countries Included in the Empirical Work" — page 21

### Structural notes (as presented)
- The content unit presents three model formulations in sequence: Basic Model, Modified Basic Model, and Alternative Model, with exact page locations provided above.
- A chronological figure is referenced earlier in the document (page 8), and an appendix lists the countries used in the empirical work (page 21).

*Source: _wp06260 - 1. Real Gasoline Prices, Oil Costs, and Elections: Basic Model — PDF chapter/section*

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

### _wp06260 - References..............................................................................................................

### I. INTRODUCTION
- Links macroeconomic policy decisions to electoral cycles via political business cycle theory (Nordhaus (1975); Rogoff and Sibert (1988); Rogoff (1990)).
- Combines industry-level regulatory capture perspective (Stigler (1971); Peltzman (1976)) with signaling-game political business cycle ideas (Paiva (1996)).
- Core hypothesis: politicians set regulated prices to maximize electoral support by signaling pro-consumer behavior, producing lower real prices immediately preceding elections.
- Empirical test: quarterly data from 32 industrial and developing countries over 1978–2004; evidence points to electoral price cycles in gasoline markets.

### II. BACKGROUND
- Nordhaus (1975) political business cycle: expansionary policy in election years exploiting a Phillips curve; assumes adaptive expectations and myopic voters.
- Rogoff and Sibert (1988), Rogoff (1990): political cycles as signaling equilibria with temporarily asymmetric information about government competency; rational voters infer type from policy.
- Stigler (1971), Peltzman (1976), Becker (1983): regulation as arbitration among interest groups; Peltzman models regulator choosing price to maximize political support (price between monopoly and competitive levels).
- Paiva (1996): introduces time dimension—regulator more pro-firm in nonelection periods, more pro-consumer before elections; relies on myopic consumers.
- Present paper: replaces myopic consumer assumption with a signaling-game framework where incumbents signal type via price; predicts lower regulated prices pre-election by pro-consumer types.

### III. MODEL OF POLITICAL PRICE CYCLES IN A REGULATED INDUSTRY
- General setup:
  - Two discrete regulator types: pro-consumer (α) and pro-industry (ᾱ), with ααα,∈ and αα< and Probμαα==)(.
  - Social welfare in regulated market is a weighted average of consumers’ utility and industry profits; weights follow a stochastic process with shocks at beginning of election periods.
  - Elections every other period. Incumbent observes type shock αt before setting price pt; voters learn type only after the election (information asymmetry).
  - Firms cannot affect election outcomes (assumption), consumers’ votes influenced by current price and an exogenous popularity shock st.
- Formal elements (preserve functional forms and assumptions):
  - Firm profit: πt = pt qt − C(qt), with Cq > 0 and Cqq > 0.
  - Consumer utility: Ut = I t φ(qt) + ηt, with 0' > φ and 0'' < φ; ηt follows 1− +=− t o t I t s s η η with st iid, unimodal, twice continuously differentiable, mean zero.
  - Demand: q = q(p).
  - Incumbent social welfare: Wt(α) = (1 − μ) V(p(α)) + μ π(p(α)) . (As stated in text: W expressed as weighted average; objective maximizes intertemporal function including reelection probability R, weight σ on being elected, and discount β < 1.)
  - Incumbent objective over term: maximize Φ = Σ_{i=0}^{1} β^{i} [ W_{t+i}(α) + σ R_{t+i}(·) ] (compactly as in text).
- Timing (chronology reproduced):
  - At beginning of period t government observes αt, sets price pt.
  - After price set, both government and consumers observe popularity shock st, election happens.
  - Winner takes office for t+1 until t+2; new preference shock αt+2 occurs at start of t+2.
- Signaling equilibrium characterization:
  - Voters’ beliefs map observed price to inferred type: p̃t → α̃t.
  - In nonelectoral years incumbents set welfare-maximizing price p*(α).
  - In electoral years incumbents solve one-period maximization: choose p to maximize [W + σ R].
  - Define distortion λα(p) = W(p*(α)) − W(p). In equilibrium separating strategies:
    - Pro-industry type sets p*(ᾱ) with zero distortion.
    - Pro-consumer type sets p*(α) − positive distortion (i.e., lowers price) to signal type.
  - Proposition (separating equilibrium): pro-industry sets p*(ᾱ) and distortion zero; pro-consumer sets p*(α) with positive distortion.
  - Theorem (closed-form separating sequential equilibrium):  
    - [][][] 1* αασλααRRpp −− = −  (as stated in text).
  - Comparative statics from model:
    - Greater probability of lower type winning relative to higher type increases distortion by lower type.
    - Greater weight σ on reelection relative to consumer welfare increases distortion.
  - Model implication: political price cycle occurs only when pro-consumer regulator is in office (probability 1−μ). Continuum-of-types extension would generate cycles of varying intensity; discrete-two-type model predicts cycles driven by pro-consumer types.

### IV. EMPIRICAL EVIDENCE
- Market choice and motivation:
  - Gasoline chosen because it is homogeneous, frequently purchased, visible in media, affects transportation costs, inflation, and consumer welfare.
  - Governments influence gasoline prices directly or indirectly via regulatory stocks, importation, taxes, subsidies, environment and safety standards.
- Data:
  - Quarterly series for domestic gasoline prices, consumer price indexes, and exchange rates against U.S. dollar for 32 industrial and developing countries, 1978–2004.
  - Maximum pool observations: 2,915 (due to some missing data).
  - Also included: average international oil prices (U.S. dollars) and dummy variable per country indicating quarters with parliamentary/congress elections (legislative elections only).
  - Sources: OECD Energy Prices and Taxes Database (domestic gasoline prices); IMF International Finance Statistics (consumer prices, exchange rates); IMF Commodity Price System Database (international crude oil prices); Election Results Archive (election dates).
- Preliminary statistics:
  - Gasprofit defined as: gasprofit = {[(1+ cpgas) ÷ (1+ cpoildomcur)] – 1} * 100, where cpgas = % change in gasoline prices, cpoildomcur = % change in international oil price in domestic currency.
  - Pooled sample (1978–2004): gasprofit increased at an average of 1.4 percent during “normal” quarters and declined by 0.4 percent in periods with electoral activity.
  - In 25 of 32 countries, average change in gasprofit during electoral quarters was lower than during “normal” quarters.
  - Average change in gasprofit during electoral quarters negative in 23 countries; negative during non-election quarters in only 7 countries.
  - Real gasoline prices: decline of 0.3 percent on average during “normal” quarters and about 0.7 percent during electoral campaign quarters.
  - In 15 countries the difference exceeded 2 percentage points; in 7 countries it exceeded 6 percentage points.
- Econometric modeling:
  - Variables generated per country:
    - lrgas = ln(domestic gasoline price) − ln(domestic CPI)
    - ldompoil = ln(real oil price in domestic currency)
    - lrer = ln(exchange rate) − ln(domestic CPI)
    - lpoilusd = ln(international oil prices in U.S. dollars) used in some specs
  - Unit root tests inconclusive in levels; first differences stationary; estimations conducted in first differences.
  - Basic panel specification (equation (1)):
    - clrgas_{n,t} = Σ_{i} γ_{i} cldompoil_{n,t-i} + δ Delec_{n,t} + α_{n} + τ_{t} + ε_{n,t}
    - Delec_{n,t} = 1 if election in current or next quarter, 0 otherwise.
  - Key results (Table 1 summary):
    - Contemporaneous coefficient on cldompoil_t ~ 0.156 to 0.225 (statistically significant).
    - Electoral dummy δ ≈ −0.009 or −0.008 (t-ratios between about −1.93 and −2.24), indicating reduction in real gasoline prices of about 0.8–0.9 percent during election periods.
    - Preferred specification (period fixed effects) passes tests; indicates suppliers/regulators pass onto domestic gasoline prices about 16 percent of contemporaneous changes in domestic oil costs.
  - Modified basic model (country-specific coefficients on oil cost lags) (Table 2):
    - Delec coefficients range from −0.009** to −0.010** (t-ratios about −2.20 to −2.66), robust and significant.
    - Adjusted R-squared improved to 0.36–0.40 in some specifications.
  - Alternative specification separating international oil price and real exchange rate (equation (2), Table 3):
    - clrgas_{n,t} = Σ λ_{i} clpoilusd_{t-i} + Σ μ_{i} clrer_{n,t-i} + δ Delec_{n,t} + α_{n} + τ_{t} + ε_{n,t}
    - Electoral dummy δ remains significant across various common and country-specific coefficient setups.
    - Point estimates suggest real gasoline price declines by at least 0.7 percent and up to 1 percent, on average, during quarters of election campaign across the 32 countries.
  - Robustness:
    - Only the first lag of explanatory variables reached minimum significance; results stable across panel estimators (LS, GLS), fixed/random effects choices, and country-specific coefficient specifications.
- Empirical summary statistics preserved:
  - Sample period: 1978–2004.
  - Number of countries: 32.
  - Maximum observations: 2,915; reported regressions use 2,913 or 2,937 in various specs.
  - Average gasprofit: 1.4 percent (normal quarters) vs. −0.4 percent (electoral quarters).
  - Average real gasoline price change: −0.3 percent (normal) vs. −0.7 percent (electoral).
  - Estimated electoral dummy δ ≈ −0.008 to −0.010 (implying −0.8 percent to −1.0 percent decline), significant at 5 percent (**) or 10 percent (*) levels.
  - Contemporaneous oil cost pass-through ≈ 0.156 (16 percent) in preferred specification.

### V. CONCLUDING REMARKS
- Theoretical contribution:
  - Presents a signaling-game model where temporary information asymmetry about regulator type and electoral incentives generate political price cycles in regulated industries.
  - Shows pro-consumer types lower regulated prices in election periods as an equilibrium signaling strategy; pro-industry types do not mimic due to welfare cost constraints.
- Empirical contribution:
  - Quarterly panel evidence for 32 countries (1978–2004) supports the model: statistically and economically significant reductions in real gasoline prices during the two quarters of electoral campaign.
  - Point estimates indicate reductions of at least 0.8 percent and up to 1 percent during election quarters, versus average quarterly reduction of 0.3 percent in the full sample period.
  - Robust across model specifications, country-specific coefficients, and alternative cost decompositions.
- Policy-relevant implications:
  - Political incentives can cause temporary distortions in regulated prices, especially for visible consumer goods like gasoline.
  - Magnitude of distortions linked to incumbent’s reelection weight σ and relative chances of types winning (probabilities R).
  - Expect heterogeneity: not all countries or periods exhibit political price cycles; where present effects can be substantial (differences > 2 percentage points in 15 countries; > 6 percentage points in 7 countries).

*Source: _wp06260 - References..............................................................................................................*

### Appendix 1. List of Countries Included in the Empirical Work

### Appendix 1. List of Countries Included in the Empirical Work

### Sample composition
- Number of countries: 32

- The 32 countries in our sample are the following:
  - Australia
  - Austria
  - Belgium
  - Brazil
  - Canada
  - Switzerland
  - Denmark
  - Finland
  - France
  - Germany
  - Greece
  - India
  - Indonesia
  - Ireland
  - Italy
  - Japan
  - Korea
  - Mexico
  - Netherlands
  - New Zealand
  - Norway
  - Poland
  - Portugal
  - Romania
  - Slovak Republic
  - South Africa
  - Spain
  - Sweden
  - Thailand
  - Turkey
  - United Kingdom
  - United States

*Source: Appendix 1. List of Countries Included in the Empirical Work*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp06260.pdf_