## wp1925

## Source details

**Canonical URL:** [wp1925](https://www.imf.org/-/media/files/publications/wp/2019/wp1925.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wp1925.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wp1925.pdf.json)

---

### I. Introduction and research questions
- Sample and period:
  - 77 developing countries over the period 2000 – 2014.
- Research objectives:
  - Assess whether an increase in domestic fuel prices hampers a country’s external competitiveness as reflected in export performance.
  - Evaluate whether energy dependency or access to other energy sources condition the effect.
  - Examine whether these effects differ by temporal horizon (short versus medium-long run).
- Key methodological approach:
  - Use of an original fuel price database compiled by Kpodar and Abdallah (2017).
  - Fixed-effect estimations and local projection approach to distinguish short-run from medium-long-run dynamics.

### II. Main empirical findings
- Baseline effects (fixed-effect estimations):
  - Change in gasoline pump price (t-1) coefficients reported: -0.623, -0.616, -0.624, -0.523, -0.612, -0.475, -0.636, -0.537.
  - Change in gasoline pump price (t-2) coefficients reported: -0.372, -0.333, -0.341, -0.265, -0.343, -0.323, -0.501, -0.563.
  - Change in diesel pump price (t-1) coefficients reported: -0.558, -0.553, -0.563, -0.440, -0.320, -0.364, -0.544, -0.576.
  - Change in diesel pump price (t-2) coefficients reported: -0.343, -0.308, -0.313, -0.255, -0.373, -0.277, -0.440, -0.482.
  - Interpretation example: “a one percentage point change in gasoline price would reduce export growth by a 0.6 percentage point” (column 3).
- Heterogeneity by energy dependency and access to electricity:
  - Interaction coefficients (Change in gasoline pump price * Log energy dependency (t-1)) include -0.484, -0.537, -0.525, -0.294 (negative and often significant).
  - Interaction coefficients (Change in diesel pump price * Log energy dependency (t-1)) include -0.276, -0.331, -0.335, -0.189 (negative and often significant).
  - Interaction coefficients (Change in diesel pump price * Log access to electricity (t-1)) include 0.514, 0.586, 0.492, 0.524 (positive and in several specifications significant), indicating mitigation by higher access to electricity.
  - For gasoline interactions with access to electricity reported examples: 0.230, 0.336 (positive but not significant in reported specifications).
- Temporal dynamics (local projections):
  - Fuel price increases are negatively correlated with export growth with magnitudes comparable to fixed-effect estimates.
  - The adverse effect is concentrated in the short run and mostly within the first two years after the shock; the impact becomes weaker and non-significant thereafter.
  - IRFs for large shocks are very close to IRFs for all shocks, indicating no disproportionately large contractions for large shocks.
- Robustness:
  - Results robust to additional controls (private investment ratio, secondary education enrollment rate, rule of law index, Trade Freedom index, share of export tax revenue, infrastructure quality, flexibility of wage determination).
  - Robust to exclusion of outliers (observations beyond two standard deviations and exclusion reducing sample by about 50 percent).
  - Robust when controlling for share of nonrenewable electricity production.

### III. Theoretical channels of transmission
- Direct channel:
  - Increase in energy input costs that is instantaneous unless hedged; exposure depends on energy required per unit of output.
- Indirect channel:
  - Higher energy prices raise production costs of intermediate inputs via supply-chain feed-through (e.g., higher transportation costs).
- Firms’ response mechanisms (Rentschler et al., 2017):
  - Absorption: firms reduce margins to absorb additional production cost.
  - Substitution: switch to other energy sources, contingent on reliable access and potentially requiring technology changes.
  - Resource efficiency: reduce overall energy consumption via updated production processes/technology.
  - Pass-on: shift costs to end-users depending on demand price elasticity.
- Constraints in developing countries:
  - Limited ability to pass through costs because exporters are often price-takers.
  - Substitution and efficiency improvements are structural and time-consuming; short-term response often limited to absorption.
  - SMEs may lack margins and face bankruptcy or market exit if unable to absorb costs.

### IV. Monetary and fiscal interactions
- Monetary:
  - Higher energy prices increase consumer prices; if inflation expectations are not well-anchored, wage indexation can trigger wage-price spirals, further raising production costs and harming competitiveness.
- Fiscal:
  - Reducing fuel subsidies can free fiscal space for growth-enhancing public spending (infrastructure, human capital) which could offset negative competitiveness effects with a lag.
  - If fiscal savings are used to expand current expenditure, inflationary pressures could worsen competitiveness.

### V. Medium- and long-run considerations
- Potential benefits where prices were below cost-recovery:
  - Reallocation away from less productive, energy-intensive activities, improving total factor productivity and competitiveness.
  - Incentives for technological change and investments in energy-efficient technology.
  - Realization of these benefits depends on availability and affordability of modern technology, financing, and implementation support.
- Empirical synthesis:
  - Mixed evidence in literature: some studies find negative short-run impacts; others find no effect or small positive effects via induced efficiency improvements over the medium/long run.
  - Examples preserved from literature: simulation of a 12 percent increase in fuel prices in Chile; a 30 percent increase in energy prices across Europe leading to up to a 0.5 percent decline in exports in Sato and Dechezleprêtre (2015); Rentschler and Kornejew (2017) evidence for Indonesia of small adverse long-run effects with larger marginal impacts for diesel and LPG.

### VI. Empirical strategy, data, and measurement
- Baseline model (as specified in the source):
  - ∆ln(Export)_{c,t} = α_0 + Σ_{j=1}^n δ_j ∆ln(Fuel)_{c,t−j} + A X_{c,t−1} + u_c + ε_{c,t}
- Dependent variable:
  - ln(Export)_{c,t} = logarithm of real non-fuel export of country c in year t (first difference used for growth).
- Variable of interest:
  - ln(Fuel) = logarithm of domestic fuel prices (retail prices per liter of gasoline and diesel converted to US dollars; yearly averages computed from monthly retail fuel price database).
- Controls (X):
  - change in international fuel prices, real effective exchange rate (REER), product diversification index, net FDI/GDP, inflation rate (CPI), external demand (average real GDP growth of the rest of the world).
- Fixed effects:
  - α_c to control for time-invariant unobserved heterogeneities.
- Lagging:
  - All explanatory variables except external demand are lagged to reduce feedback bias.
- Stationarity:
  - Phillips-Perron and Dickey-Fuller tests failed to reject stationarity for all included variables (results not shown).

### VII. Key coefficient and diagnostic highlights (selected reported values)
- Gasoline baseline (Appendix Table 3, selected FE and Sys-GMM columns):
  - Change in gasoline pump price (t-1): (1) -0.799 [0.198]***; (4) -0.648 [0.202]***; (8) -0.556 [0.151]***.
  - Change in gasoline pump price (t-2): (1) -0.458 [0.153]***; (4) -0.306 [0.154]*; (8) -0.480 [0.155]***.
  - Change in gasoline pump price (t-3): examples include 0.088 [0.140] and 0.247 [0.120]**.
  - Log REER (t-1): (1) -0.225 [0.119]*; (4) -0.316 [0.111]***.
  - Net oil export as a percent of GDP: approx. -0.018 [0.010]* in FE columns.
  - Sample size: Number of observations: 349; Number of countries: 64.
  - R-squared examples: 0.13, 0.14, 0.15, 0.19, 0.15, 0.08, 0.25, 0.39.
  - Sys-GMM diagnostics (where reported): Hansen test p-values: 0.86; AR(2) p-values: 0.58.
- Diesel baseline (Appendix Table 4, selected FE and Sys-GMM columns):
  - Change in diesel pump price (t-1): (1) -0.700 [0.130]***; (4) -0.537 [0.151]***; (8) -0.567 [0.116]***.
  - Change in diesel pump price (t-2): (1) -0.418 [0.188]**; (4) -0.292 [0.174]*; (8) -0.417 [0.132]***.
  - Change in diesel pump price (t-3): examples include 0.114 [0.108] and 0.267 [0.094]***.
  - Log REER (t-1): (1) -0.303 [0.096]***; (4) -0.381 [0.100]***.
  - Net oil export as a percent of GDP: approx. -0.018 [0.010]* in FE columns.
  - Sample size: Number of observations: 350; Number of countries: 62.
  - R-squared examples: 0.13, 0.15, 0.15, 0.18, 0.15, 0.10, 0.25, 0.37.
  - Sys-GMM diagnostics (where reported): Hansen test p-values: 0.86; AR(2) p-values: 0.65.

### VIII. Summary statistics (Appendix Table 2 — Observations = 486 unless otherwise noted)
- ∆ln(Export): Observations: 486; Mean: 0.0765; Standard deviation: 0.186; Min: -1.516; Max: 0.992.
- ln(Export): Observations: 486; Mean: 22.78; Standard deviation: 1.806; Min: 18.27; Max: 28.42.
- ∆ln(Diesel Price): Observations: 486; Mean: 0.0300; Standard deviation: 0.0741; Min: -0.327; Max: 0.211.
- ∆ln(Gasoline Price): Observations: 486; Mean: 0.0295; Standard deviation: 0.0688; Min: -0.389; Max: 0.240.
- Log Energy dependency: Observations: 407; Mean: 6.558; Standard deviation: 0.681; Min: 4.731; Max: 8.474.
- Access to electricity: Observations: 473; Mean: 4.370; Standard deviation: 0.329; Min: 2.416; Max: 4.615.
- Log REER: Observations: 486; Mean: 4.578; Standard deviation: 0.110; Min: 4.047; Max: 4.933.
- Log Diversification index: Observations: 486; Mean: 0.524; Standard deviation: 0.0749; Min: 0.323; Max: 0.653.
- Log FDI: Observations: 486; Mean: 0.989; Standard deviation: 1.166; Min: -3.911; Max: 4.009.
- Inflation: Observations: 486; Mean: 4.668; Standard deviation: 0.0441; Min: 4.581; Max: 4.892.
- Change in international gasoline prices: Observations: 486; Mean: 0.101; Standard deviation: 0.204; Min: -0.399; Max: 0.300.
- Change in international diesel prices: Observations: 486; Mean: 0.105; Standard deviation: 0.268; Min: -0.563; Max: 0.376.
- World growth: Observations: 486; Mean: 4.172; Standard deviation: 1.583; Min: 0.373; Max: 6.331.

### IX. Policy implications (drawn from empirical findings and channels)
- Anticipate a short-run trade-off between subsidy removal (or higher fuel prices) and export performance, especially in:
  - Energy-dependent economies.
  - Economies with limited access to alternative energy (electricity).
- Recommended complementary policies to mitigate short-run adverse effects:
  - Improve access to reliable electricity and alternative energy sources to facilitate substitution.
  - Support investments in energy efficiency and modern production technology, including financing and technical assistance.
  - Use fiscal savings from subsidy reduction to finance growth-enhancing public investment (infrastructure, human capital) rather than inflationary current spending.
  - Strengthen monetary policy credibility and frameworks to anchor inflation expectations and avoid second-round wage-price dynamics.
- Additional considerations:
  - Consider gradual price increases to ease short-term impacts and allow firms time to adjust, even though empirical non-linearity tests do not decisively favor gradualism over shock therapy.
  - Monitor inflationary effects when raising fuel prices, since inflation can amplify adverse export outcomes.
  - Recognize potential medium- and long-run gains from removing below-cost fuel pricing through better resource allocation and productivity improvements, contingent on appropriate complementary reforms and investments.

*Italic source: Excerpt from wp1925 (PDF chapter/section) content provided.*

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

### wp1925 - References

### I. Introduction and research questions
- Sample and period:
  - 77 developing countries over the period 2000 – 2014.
- Research objectives:
  - Assess whether an increase in domestic fuel prices hampers a country’s external competitiveness as reflected in export performance.
  - Evaluate whether energy dependency or access to other energy sources condition the effect.
  - Examine whether these effects differ by temporal horizon (short versus medium-long run).
- Motivation and context:
  - Many developing countries maintain tight control on domestic fuel prices, resulting in prolonged freezes or low pass-through of international oil prices and large fuel subsidies that carry fiscal, equity, and environmental costs.
  - Policymakers fear higher domestic fuel prices could reduce competitiveness for export-oriented sectors; empirical evidence at the macro level for developing countries is relatively limited.
- Key methodological approach:
  - Use of an original fuel price database compiled by Kpodar and Abdallah (2017).
  - Fixed-effect estimations and local projection approach to distinguish short-run from medium-long-run dynamics.

### II. Main empirical findings
- Baseline effects:
  - Domestic gasoline or diesel price increases have a negative effect on real non-fuel export growth (fixed-effect estimations).
- Heterogeneity by energy dependency:
  - Negative impact is mild or non-significant for countries with low energy dependency ratios.
  - Negative impact is significant and sizeable for countries with high energy dependency ratios.
- Role of access to electricity:
  - The adverse effect of fuel price increases on exports declines with better access to electricity.
- Temporal dynamics (local projections):
  - The adverse effect is concentrated in the short run and mostly within the first two years after the shock.
  - The impact becomes weaker and non-significant thereafter (medium-long run).

### III. Theoretical channels of transmission
- Direct channel:
  - Increase in energy input costs that is instantaneous unless hedged.
  - Degree of exposure depends on energy required per unit of output.
- Indirect channel:
  - Higher energy prices raise production costs of intermediate inputs via supply-chain feed-through (e.g., higher transportation costs).
- Firms’ response mechanisms (Rentschler et al., 2017):
  - Absorption: firms reduce margins to absorb additional production cost.
  - Substitution: switch to other energy sources, contingent on reliable access and potentially requiring technology changes.
  - Resource efficiency: reduce overall energy consumption via updated production processes/technology.
  - Pass-on: shift costs to end-users depending on demand price elasticity.
- Constraints for export sectors in developing countries:
  - Limited ability to pass through costs because exporters are often price-takers.
  - Substitution and efficiency improvements are structural and time-consuming; short-term response often limited to absorption.
  - SMEs in export sectors may lack margins and face bankruptcy or market exit if unable to absorb costs.

### IV. Monetary and fiscal interactions
- Monetary credibility and second-round effects:
  - Higher energy prices increase consumer prices; if inflation expectations are not well-anchored, wage indexation can trigger wage-price spirals, further raising production costs and harming competitiveness.
- Fiscal policy role:
  - Reducing fuel subsidies can free fiscal space for growth-enhancing public spending (infrastructure, human capital) which could offset negative competitiveness effects with a lag.
  - If fiscal savings are used to expand current expenditure, inflationary pressures could worsen competitiveness.

### V. Medium- and long-run considerations
- Potential benefits of higher energy prices where prices were below cost-recovery:
  - Reallocation away from less productive, energy-intensive activities, improving total factor productivity and competitiveness.
  - Incentives for technological change and investments in energy-efficient technology.
  - Realization of these benefits depends on availability and affordability of modern technology, financing, and implementation support.

### VI. Literature synthesis (selected quantitative findings preserved)
- Mixed empirical evidence:
  - Some studies find negative impacts of energy price increases on competitiveness; others find no effect or small positive effects depending on firms’ adaptation capacity.
  - Adverse impacts tend to be pronounced in the short-term and may phase out or reverse in the long run.
- Examples cited:
  - O‘Ryan et al. (2003): simulation of a 12 percent increase in fuel prices (to reach import-parity prices) leads to short-run output loss in Chile (static CGE model).
  - Sato and Dechezleprêtre (2015): energy prices explain only 0.01% of variation in trade flows; a 30 percent increase in energy prices across Europe would lead to only up to a 0.5 percent decline in exports.
  - Gonseth et al. (2015): diesel tax changes in 11 manufacturing sectors across six European countries over 1990-2003 find that increases in fuel taxes lower total factor productivity and net trade; adaptive capacity (measured by relative level of labor compensation) mitigates adverse effects.
  - Costantini and Mazzanti (2012), Rammera et al. (2017), Arlinghaus (2015): in some contexts energy taxes are neutral or positively associated with exports via induced efficiency improvements.

### VII. Policy implications (drawn from empirical findings and channels)
- Anticipate a short-run trade-off between subsidy removal (or higher fuel prices) and export performance, especially in energy-dependent economies and those with limited access to alternative energy.
- Complementary policies to mitigate short-run adverse effects:
  - Improve access to reliable electricity and alternative energy sources to facilitate substitution.
  - Support investments in energy efficiency and modern production technology, including financing and technical assistance.
  - Use fiscal savings from subsidy reduction to finance growth-enhancing public investment (infrastructure, human capital) rather than inflationary current spending.
  - Strengthen monetary policy credibility and frameworks to anchor inflation expectations and avoid second-round wage-price dynamics.
- Recognize potential medium- and long-run gains from removing below-cost fuel pricing through better resource allocation and productivity improvements, contingent on appropriate complementary reforms and investments.

_Italic source: Excerpt from wp1925 (PDF chapter/section) content provided._

### 0.3 percent, and a decline in exports by 0.8 percent. But the distributional impact is progressive,

### wp1925 - 0.3 percent, and a decline in exports by 0.8 percent. But the distributional impact is progressive,

### Literature evidence on energy prices and competitiveness
- Rentschler and Kornejew (2017): energy price increases have a small and statistically significant adverse long-run effect on competitiveness for Indonesian firms; marginal impacts larger for diesel and LPG than for electricity and kerosene; firm adaptation (energy substitution, higher energy efficiency, or higher consumer prices) mitigates but does not eliminate adverse effects.
- Clement et al. (2007) (CGE for Indonesia): reduction in fuel subsidies → small contraction in output (higher production costs, falling incomes) but reduction in fiscal deficit → boost to private investment that ultimately offsets initial decline in output.
- Jiang and Tan (2013): subsidy reforms in China would significantly affect price levels, especially the producer price index.
- Lin and Li (2012) (multi-world-region general equilibrium): subsidy removal in China → loss of competitiveness through the trade channel; recommend export subsidy and capital tax reduction to dampen negative effects.
- Solaymani and Kari (2012) (Malaysia): elimination of fuel subsidies → contraction of manufacturing output and exports; total output and investment rise as subsidy removal boosts oil production, offsetting non-oil output decline.
- Hope and Singh (1995) (six countries, 1980s reforms): industrial-level impact depends on energy intensity, substitutability of energy, and ability to pass costs to consumers; manufacturing output growth typically higher post-reform (except Turkey); energy price increases reduced distortions in input prices and contributed to economic growth.
- Hang and Tu (2007) (China, 1985–2004): energy price deregulation → lower energy intensity and higher energy efficiency, favorable for medium- and long-run competitiveness.
- Gaps noted: evidence from developing countries remains sparse; many studies are country-specific and use firm-level data, limiting control for macroeconomic variables and economy-wide competitiveness assessment.

### Empirical strategy (objective and scope)
- Objective: investigate linkages between fuel price changes and external competitiveness (measured by export growth) using a large panel dataset to address gaps in literature.
- Sample: panel of 77 developing countries during 2000-2014.
- Baseline approach: explain export growth by change in fuel prices, controlling for standard competitiveness determinants; include up to three lags of change in domestic fuel prices to capture pass-through timing.

### Data and model specification
- Baseline model (as specified in the source):
  ∆ln⁡(퐸푥푝표푟푡)
  푐,푡
  =훼
  0
  +훿
  푗
  ∑
  ∆ln⁡(퐹푢푒푙)
  푐,푡−푗
  푛
  푗=1
  +퐴푋
  푐,푡−1
  +푢
  푐
  +휀
  푐,푡
- Dependent variable: ln(Export)_{c,t} = logarithm of real non-fuel export of country c in year t (first difference used for growth).
- Variable of interest: ln(Fuel) = logarithm of domestic fuel prices (retail prices per liter of gasoline and diesel converted to US dollars; yearly averages computed from monthly retail fuel price database).
- Controls (X): change in international fuel prices, real effective exchange rate (REER), product diversification index, ratio of foreign direct investment to GDP (net FDI/GDP), inflation rate (CPI), and external demand measured by average real GDP growth of the rest of the world.
- Fixed effects: 훼_c to control for time-invariant unobserved heterogeneities.
- Lagging: all explanatory variables except external demand are lagged to reduce feedback bias.
- Stationarity: Phillips-Perron and Dickey-Fuller tests failed to reject stationarity for all included variables (results not shown).

### Measurement choices and expected signs
- Export growth: uses non-fuel real export growth (first difference of logarithm) from World Development Indicators (WDI).
- Domestic fuel prices: yearly averages of retail prices per liter for gasoline and diesel (converted to USD) from Kpodar and Abdallah (2017) database.
- Expected sign for fuel price change coefficient: negative and significant if fuel price rises harm competitiveness; non-significant or positive if no harm or enhancement of competitiveness.
- REER: appreciation expected to be negatively correlated with export growth (data from World Economic Outlook—WEO).
- Product diversity index (UNCTAD): expected positive correlation with export growth.
- Net FDI ratio to GDP (WDI): higher FDI expected to expand export base.
- CPI inflation (WDI): included to isolate domestic price shocks and test transmission channels (e.g., pass-through of fuel cost to domestic prices).
- International fuel prices data source: GIZ (Gesellschaft für Internationale Zusammenarbeit).

### Stylized facts from the sample
- Distribution of fuel price changes (gasoline and diesel): right-skewed, indicating positive price changes are more frequent.
- Prevalence of price controls: about 18 percent of sample observations are zero fuel price change, revealing prevalence of fuel price controls in developing countries.
- Fuel price declines: scarce and mostly coincide with the slump in international oil prices in the second half of 2008.
- Correlation patterns:
  - Overall sample: negative slope between fuel price changes and export growth, suggesting fuel price hikes may reduce competitiveness.
  - Sub-samples by energy dependency (measured by energy use in kilogram of oil equivalent per capita):
    - Countries with energy dependency below sample average: negative correlation dies out.
    - Countries with energy dependency above sample average: negative correlation persists.
  - Patterns consistent for both gasoline and diesel.

*Source: wp1925 (https://www.imf.org/-/media/files/publications/wp/2019/wp1925.pdf)*

### Appendix Table 2 provides the summary statistics on the dependent and explanatory variables.

### wp1925 - Appendix Table 2 provides the summary statistics on the dependent and explanatory variables.

### C. Effect of domestic fuel price changes on external competitiveness (fixed-effect estimations)
- Sample and setup:
  - Sample: 77 developing countries over the period 2000-2014 (tables report 76–77 countries and observations around 485–486 depending on specification).
  - Dependent variable: non-fuel real export growth.
  - Estimators: country fixed effects (FE) and System GMM (Sys-GMM) for dynamic specifications. Clustered standard errors at country level.
- Key baseline coefficient estimates (Table 1 — gasoline; Table 2 — diesel):
  - Change in gasoline pump price (t-1): -0.623, -0.616, -0.624, -0.523, -0.612, -0.475, -0.636, -0.537 (standard errors in brackets; many significant at *** p < 0.01).
    - Interpretation example: column 3—“a one percentage point change in gasoline price would reduce export growth by a 0.6 percentage point.”
  - Change in gasoline pump price (t-2): -0.372, -0.333, -0.341, -0.265, -0.343, -0.323, -0.501, -0.563.
  - Change in diesel pump price (t-1): -0.558, -0.553, -0.563, -0.440, -0.320, -0.364, -0.544, -0.576.
  - Change in diesel pump price (t-2): -0.343, -0.308, -0.313, -0.255, -0.373, -0.277, -0.440, -0.482.
  - Lagged effects: longer lags (t-2, t-3) are generally smaller and less significant, implying the negative impact phases out over time.
- Controls and additional findings:
  - Log REER (t-1) coefficients are negative and significant in most regressions (examples: -0.242, -0.257, -0.236, -0.360, -0.239, -0.307; diesel: -0.318, -0.338, -0.319, -0.409, -0.356, -0.373), indicating REER appreciation inhibits export growth.
  - Log Diversification index (t-1) is generally positive and significant in many specifications (examples: 1.689, 1.667, 1.502, 1.673).
  - Log FDI (t-1) shows small positive coefficients (examples around 0.002–0.004) and is not strongly associated with export performance.
  - Inflation (t-1) can be negatively associated with export growth (e.g., -1.132 in column 4), suggesting fuel price increases accompanied by inflationary pressures can worsen export outcomes.
  - International fuel price coefficients are negative; statistically significant for diesel in some specifications (e.g., column 5). Dropping domestic fuel prices increases size and significance of international price coefficients (column 6), consistent with international prices affecting the domestic economy primarily through pump prices.

### Non-linearity and magnitude tests
- Tests for non-linearity (Table 3):
  - Introduced change in fuel price squared and interaction with a large-shock dummy (large shocks defined as deviations from country-specific average by one standard deviation).
  - Results do not provide evidence of a nonlinear effect: marginal impact does not depend on shock size.
  - Coefficients (examples): Change in gasoline pump price (t-1) -0.438, -0.372; Change in diesel pump price (t-1) -0.317, -0.436.
  - Conclusion: no support for either gradualism or shock therapy purely from differential effect sizes of small vs. large shocks.

### D. Conditioning factors: energy dependency and access to electricity
- Energy dependency (Table 4):
  - Energy dependency measured as energy use per capita (kilogram of oil equivalent divided by total population).
  - Interaction: Change in gasoline pump price * Log energy dependency (t-1) coefficients negative and often significant (examples: -0.484, -0.537, -0.525, -0.294).
  - For diesel: Change in diesel pump price * Log energy dependency (t-1) coefficients also negative and significant (examples: -0.276, -0.331, -0.335, -0.189).
  - Interpretation: countries with higher energy dependency experience larger negative effects of fuel price increases on export growth.
  - Log Energy dependency (t-1) main effect examples: -0.289, -0.288, -0.123, 0.409.
- Access to electricity (Table 5):
  - Access to electricity proxied by percentage of urban population connected to the grid.
  - Interaction: Change in diesel pump price * Log access to electricity (t-1) positive and in several specifications significant (examples: 0.514, 0.586, 0.492, 0.524), indicating higher access mitigates the negative impact of diesel price increases.
  - For gasoline interactions are positive but not significant in the reported specifications (examples: 0.230, 0.336).
  - Log access to electricity (t-1) main effect examples: -0.133, -0.128, -0.039, 0.095.
  - Interpretation: better access to electricity reduces competitiveness loss from fuel price increases.
- Figures and marginal effects:
  - Figure 4: marginal impact of gasoline/diesel price change on non-fuel exports becomes larger and significant as energy dependency increases.
  - Figure 5: countries with high access to electricity tend to suffer less from export competitiveness loss following fuel price increases.

### E. Persistence and local projection results (short-term vs long-term)
- Methodology: local projection approach (Jordà, 2005) with correction by Teulings and Zubanov (2014), estimating impulse responses at horizons h = 0,1,2,3.
- Key results (Figure 6 and related discussion):
  - Fuel price increases are negatively correlated with export growth with magnitudes comparable to fixed-effect estimates.
  - The impact is short-lived: becomes statistically non-significant within two years after the shock.
  - IRFs for large shocks are very close to IRFs for all shocks, supporting the finding that large fuel price shocks do not produce disproportionately large contractions in exports.
- Conditional IRFs (Figures 7 and 8):
  - For countries with energy use per capita below the sample average, fuel price increases do not have a statistically significant effect.
  - For countries above the average energy dependency, there is a negative and statistically significant effect on non-fuel real exports that dies out within two years after the shock.
  - Countries with high access to electricity show little meaningful impact from fuel price changes; countries with low access show a negative effect.
- Overall persistence conclusion: contractions in exports following fuel price increases appear temporary rather than permanent.

### Robustness tests
- Robustness checks included adding controls: private investment ratio, secondary education enrollment rate, rule of law index, Trade Freedom index, share of export tax revenue in total government tax revenue, infrastructure quality (WEF index), and flexibility of wage determination (WEF index). Most of these additional variables are not statistically significant and baseline findings remain unaltered.
- Robustness to outliers: excluding observations deviating by more than two standard deviations from sample average, or excluding observations beyond half a standard deviation (reducing sample by about 50 percent), does not alter main results.
- Additional robustness: results robust when controlling for share of nonrenewable electricity production.

### IV. Conclusion and policy implications
- Summary findings:
  - Fuel price increases tend to lead to a non-negligible decline in real non-fuel export growth, concentrated in countries with high energy dependency and low access to electricity.
  - Where fuel price hikes reduce export growth, the effect is short-run (within the first two years) with no permanent effect on external competitiveness.
  - Results robust across fixed-effect estimators and the local projection approach.
- Policy recommendations:
  - Countries increasing domestic fuel prices should not be overly concerned about long-term impacts on export competitiveness given the short-lived nature of adverse effects.
  - Mitigating measures and structural reforms that reduce energy inefficiencies and enhance access to alternative energy sources (e.g., grid electricity) can minimize adverse consequences for exports.
  - Consideration should be given to gradual price increases to ease short-term impacts and allow firms time to adjust, even though empirical results do not decisively favor gradualism over shock therapy on the non-linearity test.
  - Monitor inflationary effects when raising fuel prices, since inflation can amplify adverse export outcomes.

*Source: wp1925 - Appendix Table 2 provides the summary statistics on the dependent and explanatory variables.*

### REFERENCES

### REFERENCES

### Bibliographic sources
- A curated list of empirical and policy literature on energy pricing, subsidies, competitiveness, and econometric methods, including but not limited to:
  - Aldy, J. E., & Pizer, W. A. (2015). “The Competitiveness Impacts of Climate Change Mitigation Policies,” Journal of the Association of Environmental and Resource Economists, 2(4), 565-595.
  - Arlinghaus, J. (2015). “Impacts of Carbon Prices on Indicators of Competitiveness: A Review of Empirical Findings”, OECD Environment Working Papers, 87, OECD Publishing, Paris.
  - Clements, B., Coady, D., Fabrizio, S., Gupta, S., Alleyne, T., & Sdralevich, C., eds. (2013). Energy Subsidy Reform: Lessons and Implications. International Monetary Fund, Washington, DC.
  - Kpodar, K. and Abdallah, C., 2017. Dynamic fuel price pass-through: Evidence from a new global retail fuel price database. Energy Economics, 66, pp.303-312.
  - IMF (2013a). Case Studies on Energy Subsidy Reform: Lessons and Implications. International Monetary Fund, Washington, DC.
  - Additional empirical studies on gasoline/diesel demand elasticities, fuel subsidy reform case studies, impulse response estimation methods, and international fuel price data sources (GIZ).

### Appendix Table 1 — Variable definitions and sources
- Export growth: First difference of the natural log of real non-fuel exports — World Development Indicators (WDI)
- Diesel price growth: First difference of the natural log of real domestic retail Diesel price in USD per liter — Kpodar and Abdallah (2017)
- Gasoline price growth: First difference of the natural log of real domestic retail Gasoline price in USD per liter — Kpodar and Abdallah (2017)
- Energy dependency: Energy use (kilogram of oil equivalent) per capita — World Development Indicators (WDI)
- Access to electricity: Percentage of the urban population with access to electricity — World Development Indicators (WDI)
- Real effective exchange rate: Real effective exchange rate (base 100=2010) — World Economic Outlook (WEO)
- Product diversity index: Product diversity index — UNCTAD
- FDI: Net Foreign Direct Investment inflows as share of GDP — World Development Indicators (WDI)
- Inflation: Inflation rate based on Consumer Price Index (Annual %) — World Development Indicators (WDI)
- World growth: Real GDP growth of the rest of the world — Authors' calculation
- International Diesel price growth: First difference of the natural log of international retail Diesel price in USD per liter — GIZ (International Fuel Prices 2012/2013, 8th edition)
- International Gasoline price growth: First difference of the natural log of international retail Gasoline price in USD per liter — GIZ (International Fuel Prices 2012/2013, 8th edition)

### Appendix Table 2 — Summary statistics (Observations = 486 unless otherwise noted)
- ∆ln(Export)
  - Observations: 486
  - Mean: 0.0765
  - Standard deviation: 0.186
  - Min: -1.516
  - Max: 0.992
- ln(Export)
  - Observations: 486
  - Mean: 22.78
  - Standard deviation: 1.806
  - Min: 18.27
  - Max: 28.42
- ∆ln(Diesel Price)
  - Observations: 486
  - Mean: 0.0300
  - Standard deviation: 0.0741
  - Min: -0.327
  - Max: 0.211
- ∆ln(Gasoline Price)
  - Observations: 486
  - Mean: 0.0295
  - Standard deviation: 0.0688
  - Min: -0.389
  - Max: 0.240
- Log Energy dependency
  - Observations: 407
  - Mean: 6.558
  - Standard deviation: 0.681
  - Min: 4.731
  - Max: 8.474
- Access to electricity
  - Observations: 473
  - Mean: 4.370
  - Standard deviation: 0.329
  - Min: 2.416
  - Max: 4.615
- Log REER
  - Observations: 486
  - Mean: 4.578
  - Standard deviation: 0.110
  - Min: 4.047
  - Max: 4.933
- Log Diversification index
  - Observations: 486
  - Mean: 0.524
  - Standard deviation: 0.0749
  - Min: 0.323
  - Max: 0.653
- Log FDI
  - Observations: 486
  - Mean: 0.989
  - Standard deviation: 1.166
  - Min: -3.911
  - Max: 4.009
- Inflation
  - Observations: 486
  - Mean: 4.668
  - Standard deviation: 0.0441
  - Min: 4.581
  - Max: 4.892
- Change in international gasoline prices
  - Observations: 486
  - Mean: 0.101
  - Standard deviation: 0.204
  - Min: -0.399
  - Max: 0.300
- Change in international diesel prices
  - Observations: 486
  - Mean: 0.105
  - Standard deviation: 0.268
  - Min: -0.563
  - Max: 0.376
- World growth
  - Observations: 486
  - Mean: 4.172
  - Standard deviation: 1.583
  - Min: 0.373
  - Max: 6.331

### Appendix Table 3 — Baseline with Gasoline Price: key estimated coefficients and model diagnostics
- Dependent variable: non-fuel real export growth. All estimates include country-fixed effects. Clustered standard errors at country level in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.
- Change in gasoline pump price (t-1): coefficients across columns (FE and Sys-GMM)
  - (1) -0.799 [0.198]***
  - (2) -0.790 [0.203]***
  - (3) -0.795 [0.211]***
  - (4) -0.648 [0.202]***
  - (5) -0.813 [0.431]*
  - (6) -0.607 [0.195]***
  - (7) -0.738 [0.175]***
  - (8) -0.556 [0.151]***
- Change in gasoline pump price (t-2):
  - (1) -0.458 [0.153]***
  - (2) -0.423 [0.156]***
  - (3) -0.421 [0.155]***
  - (4) -0.306 [0.154]*
  - (5) -0.417 [0.147]***
  - (6) -0.380 [0.132]***
  - (7) -0.649 [0.159]***
  - (8) -0.480 [0.155]***
- Change in gasoline pump price (t-3):
  - Coefficients range from -0.158 to 0.247; selected values:
    - (1) 0.088 [0.140]
    - (6) 0.247 [0.120]**
    - (7) -0.158 [0.171]
- Log REER (t-1):
  - (1) -0.225 [0.119]*
  - (4) -0.316 [0.111]***
  - (8) 0.199 [0.121]
- Log Diversification index (t-1):
  - (1) 1.967 [0.986]*
  - (5) 2.373 [0.981]**
  - (8) -2.592 [1.873]
- Net oil export as a percent of GDP:
  - Coefficients approximately -0.018 to -0.016 across FE columns, e.g., (1) -0.018 [0.010]*.
- Log non-oil real exports (t-1) (included in some specifications):
  - Coefficients: -0.325 [0.104]*** and 0.906 [0.055]*** in columns where included.
- Sample size and model fit:
  - Number of observations: 349
  - Number of countries: 64
  - R-squared: ranges reported (e.g., 0.13, 0.14, 0.15, 0.19, 0.15, 0.08, 0.25, 0.39)
- Sys-GMM diagnostics (column where reported):
  - Hansen test p-values: 0.86
  - AR(2) test (p-values): 0.58
- Constants (selected):
  - (1) 1.185 [0.548]**
  - (4) 7.277 [1.097]***
  - (9) 4.586 [3.457]

### Appendix Table 4 — Baseline with Diesel Price: key estimated coefficients and model diagnostics
- Dependent variable: non-fuel real export growth. All estimates include country-fixed effects. Clustered standard errors at country level in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.
- Change in diesel pump price (t-1): coefficients across columns (FE and Sys-GMM)
  - (1) -0.700 [0.130]***
  - (2) -0.686 [0.139]***
  - (3) -0.693 [0.147]***
  - (4) -0.537 [0.151]***
  - (5) -0.446 [0.201]**
  - (6) -0.463 [0.127]***
  - (7) -0.625 [0.109]***
  - (8) -0.567 [0.116]***
- Change in diesel pump price (t-2):
  - (1) -0.418 [0.188]**
  - (2) -0.383 [0.189]**
  - (3) -0.380 [0.185]**
  - (4) -0.292 [0.174]*
  - (5) -0.454 [0.178]**
  - (6) -0.324 [0.164]*
  - (7) -0.539 [0.168]***
  - (8) -0.417 [0.132]***
- Change in diesel pump price (t-3):
  - Coefficients range from -0.046 to 0.267; selected values:
    - (1) 0.114 [0.108]
    - (6) 0.267 [0.094]***
    - (8) 0.175 [0.096]*
- Log REER (t-1):
  - (1) -0.303 [0.096]***
  - (4) -0.381 [0.100]***
  - (8) 0.178 [0.121]
- Log Diversification index (t-1):
  - (1) 1.972 [0.984]**
  - (5) 2.304 [0.969]**
  - (8) -2.161 [1.802]
- Net oil export as a percent of GDP:
  - Coefficients approximately -0.018 to -0.015 across FE columns, e.g., (1) -0.018 [0.010]*.
- Log non-oil real exports (t-1) (included in some specifications):
  - Coefficients: -0.312 [0.100]*** and 0.937 [0.046]*** in columns where included.
- Sample size and model fit:
  - Number of observations: 350
  - Number of countries: 62
  - R-squared: ranges reported (e.g., 0.13, 0.15, 0.15, 0.18, 0.15, 0.10, 0.25, 0.37)
- Sys-GMM diagnostics (column where reported):
  - Hansen test p-values: 0.86
  - AR(2) test (p-values): 0.65
- Constants (selected):
  - (1) 1.538 [0.439]***
  - (4) 7.277 [1.448]***
  - (8) 7.874 [3.021]**

*Source: wp1925 - REFERENCES (IMF working paper appendices and references).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wp1925.pdf_