## _wp08253

## Source details

**Canonical URL:** [_wp08253](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08253.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08253.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08253.pdf.json)

---

### I. Introduction — core question and headline findings
- Research question: Do world oil price changes exert an independent influence on economic activity in oil-exporting countries (e.g., via confidence effects or monetary/financial channels), or is the impact transmitted only through fiscal policy changes?
- Method: Estimate impulse responses to oil price shocks using panel VARs of a three-variable system—oil prices, fiscal stance, and output—and compare with an alternate VAR where fiscal policy is exogenous (unresponsive to oil price shocks).
- Core results:
  - Apart from their effect on fiscal policy, oil prices do not independently influence underlying non-oil output, especially in countries where the oil sector is a large share of the economy.
  - Fiscal policy is the mechanism by which oil price shocks are transmitted to the non-oil economy.
  - Fiscal policy tends to be procyclical because it drives the output cycle; fiscal policy changes do not amplify an already-existing, independent relationship between oil prices and non-oil output.

### II. Data, sample, and cyclicality
- Sample composition and coverage:
  - Countries: Algeria, Iran, Kuwait, Libya, Nigeria, Norway, Oman, Saudi Arabia, United Arab Emirates (UAE), Yemen.
  - The 10 economies accounted for about 40 percent of the world’s oil output in 2007 and almost 50 percent of world oil exports; in proven oil reserves the sample’s share is above 50 percent.
  - Data sources: International Financial Statistics, IMF country staff reports, and country desks’ databases.
- Heterogeneity and grouping:
  - High dependence group: Algeria, Kuwait, Libya, Nigeria, Oman, Saudi Arabia, UAE.
  - Low/medium dependence group: Iran, Norway, Yemen.
  - GCC subset of high-oil: Kuwait, Oman, Saudi Arabia, UAE (public spending to non-oil GDP highest in GCC).
- Non-oil output cycle properties (selected entries from Table 2 as reported):
  - Iran (1980-2007): Cycle Duration 6.1; Coefficient of Variation 2.5
  - Norway (1980-2007): Cycle Duration 11.3; Coefficient of Variation 1.5
  - Kuwait (1983-2007): Cycle Duration 4.0; Coefficient of Variation 29.5
  - Oman (1990-2007): Cycle Duration 5.0; Coefficient of Variation 23.2
  - Observation: Economies with greater dependence on oil show higher output volatility and shorter cycles; GCC countries generally have the shortest average cyclical duration (except Saudi Arabia).
- Fiscal volatility (selected entries from Table 3 as reported):
  - Iran (1980-2007): Fiscal Impulse 0.04; Non-oil Revenue to Non-oil GDP 0.02; Public Spending to Non-oil GDP 0.07
  - Kuwait (1986-2007): Fiscal Impulse 0.30; Non-oil Revenue to Non-oil GDP 0.14; Public Spending to Non-oil GDP 0.41
  - Libya (1990-2007): Fiscal Impulse 0.13; Non-oil Revenue to Non-oil GDP 0.02; Public Spending to Non-oil GDP 0.35
  - With the exception of Algeria, high-oil group countries have greater volatility in their fiscal impulse than low/medium-oil group countries.
- Summary point: Oil-exporting economies show greater volatility in both economic cycles and fiscal stance; disentangling whether volatility is driven by oil prices, fiscal policy, or both motivates the VAR analysis.

### III. Correlations and panel regressions — preliminary evidence
- Country-level correlations with oil price changes (selected entries from Table 4; significance: * at 10 percent, ** at 5 percent):
  - Iran: Non-oil GDP 0.44 ** | Fiscal Impulse 0.25 | Public spending to non-oil GDP 0.19
  - Norway: Non-oil GDP -0.17 | Fiscal Impulse -0.32 | Public spending to non-oil GDP -0.13
  - Yemen: Non-oil GDP 0.05 | Fiscal Impulse 0.57 ** | Public spending to non-oil GDP 0.60 **
  - Saudi Arabia: Non-oil GDP 0.36 * | Fiscal Impulse 0.60 ** | Public spending to non-oil GDP 0.60 **
  - Libya: Non-oil GDP 0.56 ** | Fiscal Impulse 0.46 * | Public spending to non-oil GDP 0.31
  - Oman: Non-oil GDP 0.51 ** | Fiscal Impulse 0.44 * | Public spending to non-oil GDP 0.58 **
- Panel fixed effects regression findings:
  - Simple panel FE: Δy_NO = α + γ Δp_O + u_t shows high explanatory power, but γ weakens when fiscal variables are included.
  - General specification: Δy_NO = α + β Δ(G/Y_NO) + γ Δp_O + ε_t yields:
    - Magnitude and significance of γ weaken substantially (especially for high-oil and GCC groups).
    - F-test of γ=0 cannot be rejected for high-oil and GCC groups; rejected only at 10 percent for full sample.
    - F-tests strongly reject β=0 for the full sample, high-oil, and GCC groups—highlighting importance of fiscal stance.
- Representative coefficient estimates (Table 5; dependent variable = first difference of log real non-oil GDP; standard errors in parentheses):
  - ∆p_O coefficients: 0.037 ** (0.017); 0.030 * (0.016); 0.056 *** (0.021); 0.030 (0.019); 0.057 ** (0.025); 0.016 (0.027)
  - ∆(G/Y_NO) coefficients: 0.059 ** (0.030); 0.059 ** (0.030); 0.096 *** (0.033); 0.098 *** (0.033); 0.098 *** (0.038); 0.100 *** (0.037)
  - R-sqr ranges: 0.230 to 0.443; Adj. R-sqr ranges: 0.181 to 0.394
  - F-stat (∆p_O): 3.52 * and 2.38 (other values reported)
  - F-stat (∆(G/Y_NO)): 4.02 **, 8.64 ***, 6.87 ***
- Fiscal impulse alternative (Table 6; lagged imp coefficients and ∆p_O coefficients reported):
  - ∆p_O coefficients: 0.037 ** (0.017); 0.030 * (0.016); 0.056 *** (0.021); 0.030 (0.019); 0.057 ** (0.025); 0.025 (0.022)
  - imp (lagged) coefficients: 0.082 *** (0.031); 0.079 *** (0.031); 0.115 *** (0.033); 0.114 *** (0.033); 0.114 *** (0.036); 0.116 *** (0.036)
  - R-sqr ranges: 0.230 to 0.513; Adj. R-sqr ranges: 0.181 to 0.471
  - F-stat (imp): 6.97 ***, 11.89 ***, 10.04 ***
- Interim interpretation: Evidence supports that oil prices do not exert a direct influence on non-oil activity apart from their effect on the fiscal stance, particularly in high-oil-share economies.

### IV. Panel VAR specification and key coefficients
- VAR specification (equation (5)):
  - Reduced form panel VAR: z_nt = C + C_1 z_{n,t-1} + f + u_{nt}, where z_nt = [Δy_NO, Δ(G/Y_NO), Δp_O]'.
  - One lag included due to annual data and relatively short time series.
- Selected VAR coefficient estimates (Table 7; coefficient then standard error):
  - Full Sample — Δy_NO equation:
    - Lag Δy_NO: 0.239 (0.070)
    - Lag Δ(G/Y_NO): 0.083 (0.029)
    - Lag Δp_O: 0.024 (0.015)
    - c: 0.052 (0.009)
    - R-squared 0.356; Adj. R-squared 0.302
  - Full Sample — Δ(G/Y_NO) equation:
    - Lag Δy_NO: -0.398 (0.229)
    - Lag Δ(G/Y_NO): -0.032 (0.095)
    - Lag Δp_O: 0.159 (0.049)
    - c: -0.010 (0.029)
    - R-squared 0.137; Adj. R-squared 0.065
  - High Oil Share — Δy_NO equation:
    - Lag Δy_NO: 0.178 (0.082)
    - Lag Δ(G/Y_NO): 0.108 (0.033)
    - Lag Δp_O: 0.023 (0.020)
    - c: 0.083 (0.014)
    - R-squared 0.410; Adj. R-squared 0.354
  - High Oil Share — Δ(G/Y_NO) equation:
    - Lag Δy_NO: -0.430 (0.289)
    - Lag Δ(G/Y_NO): -0.052 (0.115)
    - Lag Δp_O: 0.225 (0.070)
    - c: -0.027 (0.048)
    - R-squared 0.165; Adj. R-squared 0.085
  - GCC — Δy_NO equation:
    - Lag Δy_NO: 0.085 (0.107)
    - Lag Δ(G/Y_NO): 0.097 (0.038)
    - Lag Δp_O: 0.018 (0.027)
    - c: 0.088 (0.015)
    - R-squared 0.451; Adj. R-squared 0.387
  - GCC — Δ(G/Y_NO) equation:
    - Lag Δy_NO and associated entries reported as in source (table formatting issues noted in source).
- VAR conclusions:
  - Lagged changes in the public spending ratio and in oil prices are significant determinants of the non-oil output cycle in the VAR, though oil prices are only marginally significant for the high-oil group and not significant for the GCC group.
  - Oil prices significantly affect the fiscal stance across full sample and sub-samples, with the estimated effect larger for the high-oil group and larger still for the GCC group.

### V. Impulse response analysis — transmission and counterfactuals
- Main impulse response findings:
  - A one standard deviation shock to the oil price results in a significant increase in non-oil output and public spending for the full sample (Chart 3).
  - Shocks to public spending have a positive effect on non-oil output.
  - Ordering in the VAR: non-oil output first, government spending second, oil prices last.
  - For the high-oil group and the GCC group:
    - Responsiveness of non-oil output to oil price shocks weakens as the sample is reduced to the high-oil group.
    - For the GCC group, the effect of oil price shocks on non-oil output is not significantly different from zero (Chart 4).
  - Oil prices significantly affect public spending across samples, and spending shocks significantly impact the output cycle where public spending is relatively important.
- Fiscal channel vs. direct effect (fiscal channel shut down counterfactual):
  - When fiscal stance is not changed in response to oil price shocks, the response of non-oil output to a one standard deviation oil price shock is not different from zero at the 5 percent significance level for the full sample (Chart 5).
  - For the high-oil and GCC groups, the impact is insignificantly different from zero and the estimated mean magnitude is lower when fiscal policy is held fixed.
- Interpretation:
  - Oil price changes do not have a significant independent effect—directly or via confidence and/or monetary and financial effects—on the underlying non-oil economic cycle in oil-exporting countries.
  - Movements in oil prices tend to be associated with changes in the fiscal stance, which in turn affect the economic cycle.
  - Intuition: oil revenue principally accrues to the government in these countries; if additional oil income is saved (accumulated abroad) the domestic non-oil economy is unaffected, whereas increased government spending affects output, especially where public spending is large relative to the economy.

### VI. Robustness checks and extensions
- Robustness: main results robust to:
  - Small changes in the sample period.
  - Addition of one more lag in the VAR.
  - Replacing government spending ratio with the fiscal impulse.
  - Adding real oil output changes to the VAR.
  - Using an alternative measure of the cyclical component of non-oil output (HP-filtered non-oil output).
- Fiscal impulse specification:
  - Replacing spending ratio with lagged fiscal impulse yields similar results: with endogenous fiscal policy, oil price shocks have marginal significance for the full sample and high-oil group; with fiscal policy fixed, oil price shocks do not significantly affect output in any group (Chart 6).
  - Lagged output changes are not a significant determinant of the fiscal impulse in the VAR—so oil prices affecting fiscal stance via output is not a major concern.
- Sample extension 1980–89 for Iran, Norway, Saudi Arabia:
  - Yields a significant, albeit small, direct effect of oil price shocks on output for the full sample even with fiscal channel closed—possibly due to different transmission in the 1980s or weaker data quality.
  - High-oil and GCC group results remain unaffected.
- Adding oil output as exogenous variable:
  - Explanatory power of VAR equations improved somewhat.
  - Oil price shocks continued to have an insignificant—or marginally significant for the full sample—effect on non-oil output when fiscal policy was held fixed.
- HP-filtered non-oil output:
  - Main result strengthened: oil price changes did not have a significant impact on the output cycle for full sample, high-oil, and GCC groups even when fiscal policy responded.
  - Caveat: HP filter sensitivity to endpoint problems in short time series may affect accuracy.

### VII. Policy-relevant conclusions
- Empirical evidence argues against the view that fiscal policy reactions to oil price shocks amplify the underlying business cycle in oil-exporting countries—especially where the non-oil sector is relatively small.
- In countries where public spending occupies a large share of the economy, spending increases will by definition increase the cyclical component of output; whether such spending coincides with oil price changes does not change the degree of cyclical pressure.
- Caution for policy:
  - Fiscal expansion in response to positive oil price shocks is not necessarily optimal; assessing optimality requires weighing costs from increased cyclical pressure against benefits of greater spending, which depend on magnitude and persistence of the price shock.
  - The analysis implies that cyclical-stability costs associated with fiscal responses will not be larger during periods of high oil prices.

*Content based solely on _wp08253 - References..............................................................................................................*

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

### _wp08253 - References..............................................................................................................

### I. INTRODUCTION
- Macroeconomic performance in oil-exporting countries depends largely on developments in world oil prices.
- Pooled data from 10 oil-exporting countries indicate a strong and statistically significant correlation—about 0.4—between real non-oil GDP growth and oil price changes over the past decade and a half.
- The correlation between spending and oil prices is around 0.6 during 1991–2007.
- Key research question: Do world oil price changes exert an independent influence on economic activity in oil-exporting countries (e.g., via confidence effects or monetary/financial channels), or is the impact transmitted only through fiscal policy changes?
- Approach: Estimate impulse responses to oil price shocks using panel VARs of a three-variable system—oil prices, fiscal stance, and output—and compare with an alternate VAR where fiscal policy is exogenous (unresponsive to oil price shocks).
- Core result summary:
  - Apart from their effect on fiscal policy, oil prices do not independently influence underlying non-oil output, especially in countries where the oil sector is a large share of the economy.
  - Fiscal policy is the mechanism by which oil price shocks are transmitted to the non-oil economy.
  - Fiscal policy tends to be procyclical because it drives the output cycle; fiscal policy changes do not amplify an already-existing, independent relationship between oil prices and non-oil output.
- Paper organization: Section II (data and empirical facts), Section III (relations among oil prices, output, fiscal cycles), Section IV (panel VAR and impulse responses), Section V (sensitivity checks and extensions), Section VI (conclusion).

### II. DATA AND CYCLICAL PROPERTIES
- Sample and data sources:
  - Sample comprises 10 oil-exporting countries: Algeria, Iran, Kuwait, Libya, Nigeria, Norway, Oman, Saudi Arabia, United Arab Emirates (UAE), and Yemen.
  - The 10 economies accounted for about 40 percent of the world’s oil output in 2007 and almost 50 percent of world oil exports; in proven oil reserves the sample’s share is above 50 percent.
  - Data drawn from International Financial Statistics, IMF country staff reports, and country desks’ databases.
- Heterogeneity in oil importance (Table 1 summary, 1990-2007):
  - Share of nominal oil GDP in total GDP ranges (examples from table):
    - Iran 0.26 (Mean) St.Dev. 0.05
    - Norway 0.17 Mean St.Dev. 0.08
    - Kuwait 0.48 Mean St.Dev. 0.06
  - Share of oil revenue in total fiscal revenue and ratio of public spending to non-oil GDP reported for each country in Table 1.
- Grouping by oil dependence:
  - High dependence: Algeria, Kuwait, Libya, Nigeria, Oman, Saudi Arabia, UAE.
  - Low/medium dependence: Iran, Norway, Yemen.
  - High-oil group further partitioned into GCC and non-GCC due to larger government size (public spending to non-oil GDP highest in GCC: Kuwait, Oman, Saudi Arabia, UAE).
- Output cycle properties (Table 2):
  - Non-oil output defined as total output excluding oil (and gas) related activities; includes some oil-related activities (e.g., petrochemicals).
  - Cycle duration (average; years) and coefficient of variation (real non-oil output) by country (sample periods vary):
    - Iran (1980-2007): Cycle Duration 6.1; Coefficient of Variation 2.5
    - Norway (1980-2007): Cycle Duration 11.3; Coefficient of Variation 1.5
    - Kuwait (1983-2007): Cycle Duration 4.0; Coefficient of Variation 29.5
    - Oman (1990-2007): Cycle Duration 5.0; Coefficient of Variation 23.2
  - Economies with greater dependence on oil show higher output volatility and shorter cycles; GCC countries generally have the shortest average cyclical duration (except Saudi Arabia).
- Fiscal volatility (Table 3 standard deviations):
  - Standard deviation of Fiscal Impulse, Non-oil Revenue to Non-oil GDP, Public Spending to Non-oil GDP reported per country (sample periods vary). Examples:
    - Iran (1980-2007): Fiscal Impulse 0.04; Non-oil Revenue to Non-oil GDP 0.02; Public Spending to Non-oil GDP 0.07
    - Kuwait (1986-2007): Fiscal Impulse 0.30; Non-oil Revenue to Non-oil GDP 0.14; Public Spending to Non-oil GDP 0.41
    - Libya (1990-2007): Fiscal Impulse 0.13; Non-oil Revenue to Non-oil GDP 0.02; Public Spending to Non-oil GDP 0.35
  - With the exception of Algeria, high-oil group countries have greater volatility in their fiscal impulse than low/medium-oil group countries.
- Summary: Oil-exporting economies show greater volatility in both economic cycles and fiscal stance; determining whether volatility is driven by oil prices, fiscal policy, or both is the focus of subsequent analysis.

### III. OIL PRICES, OUTPUT CYCLES, AND FISCAL POLICY
- Correlations with oil price changes (Table 4):
  - Positive correlation between growth rates of real non-oil output and oil prices for virtually all countries; statistically significant for about one-half of them.
  - Fiscal stance (public spending ratio and fiscal impulse) also strongly related to oil prices; correlations generally higher in high-oil group.
  - Selected country correlations with oil price changes (Table 4 format: Non-oil GDP | Fiscal Impulse | Public spending to non-oil GDP):
    - Iran: 0.44 ** | 0.25 | 0.19
    - Norway: -0.17 | -0.32 | -0.13
    - Yemen: 0.05 | 0.57** | 0.60**
    - Saudi Arabia: 0.36 * | 0.60** | 0.60**
    - Libya: 0.56** | 0.46* | 0.31
    - Oman: 0.51** | 0.44* | 0.58**
  - Notes: Some correlations use lagged changes in oil prices (Nigeria, Saudi Arabia, Libya, Oman). Significance: * at 10 percent, ** at 5 percent.
- Preliminary panel regression evidence:
  - Simple panel fixed effects specification: Δy_NO = α + γ Δp_O + u_t (equation (1)) shows high explanatory power but the coefficient on oil prices weakens when controlling for fiscal variables.
  - More general specification (equation (2)): Δy_NO = α + β Δ(G/Y_NO) + γ Δp_O + ε_t shows:
    - Magnitude and significance of γ (oil price coefficient) weaken substantially, especially for high-oil and GCC groups.
    - F-test of γ=0 cannot be rejected for high-oil and GCC groups; rejected only at 10 percent for full sample.
    - F-tests strongly reject β=0 for full sample, high-oil, and GCC groups—indicating the importance of fiscal stance.
- Representative coefficient estimates and model statistics (from Table 5; dependent variable = first difference of log real non-oil GDP; standard errors in italics):
  - ∆p_O coefficients (various columns): 0.037 ** (std err 0.017), 0.030 * (0.016), 0.056 *** (0.021), 0.030 (0.019), 0.057 ** (0.025), 0.016 (0.027)
  - ∆(G/Y_NO) coefficients: 0.059 ** (0.030), 0.059 ** (0.030), 0.096 *** (0.033), 0.098 *** (0.033), 0.098 *** (0.038), 0.100 *** (0.037)
  - Constant c across specifications: 0.055 *** (0.007) through 0.096 *** (0.012)
  - R-sqr ranges: 0.230 to 0.443; Adj. R-sqr ranges: 0.181 to 0.394
  - F-stat (∆p_O): 3.52 * and 2.38 (other values reported)
  - F-stat (∆(G/Y_NO)): 4.02 **, 8.64 ***, 6.87 ***
- Alternate fiscal stance measure: cyclically-adjusted balance (cab) and fiscal impulse (imp):
  - cab defined as cab_t = (R_NO_t / Y_NO_t) - (G_NO_t / Y_NO_t) adjusted for potential output Y*_NO (equation (3)); potential output obtained by HP filter.
  - Fiscal impulse imp_t = cab_t-1 - cab_t-2 (equation (4)); positive imp corresponds to a decline in cab.
  - Replacing government spending with lagged fiscal impulse in panel FE regressions yields similar results (Table 6).
- Representative estimates with fiscal impulse (Table 6):
  - ∆p_O coefficients: 0.037 ** (0.017), 0.030 * (0.016), 0.056 *** (0.021), 0.030 (0.019), 0.057 ** (0.025), 0.025 (0.022)
  - imp (lagged) coefficients: 0.082 *** (0.031), 0.079 *** (0.031), 0.115 *** (0.033), 0.114 *** (0.033), 0.114 *** (0.036), 0.116 *** (0.036)
  - R-sqr ranges: 0.230 to 0.513; Adj. R-sqr ranges: 0.181 to 0.471
  - F-stat (∆p_O): 3.78 * and 2.51 (other values reported)
  - F-stat (imp): 6.97 ***, 11.89 ***, 10.04 ***

- Interim interpretation: Evidence supports view that oil prices do not exert a direct influence on non-oil activity apart from their effect on the fiscal stance, particularly in high-oil-share economies. The fiscal stance appears to be the transmission channel for oil price shocks.

### IV. PANEL VAR FRAMEWORK
- Specification:
  - Reduced form panel VAR: z_nt = C + C_1 z_{n,t-1} + f + u_{nt} (equation (5)), where z_nt is a 3×1 vector: [Δy_NO, Δ(G/Y_NO), Δp_O]'.
  - C is 3×N matrix of fixed effects; f is N×1 vector of country dummies; one lag included due to annual data and relatively short time series (most countries 1990–2007).
- VAR results (Table 7; coefficient estimate then standard error in each cell):
  - For Full Sample:
    - Δy_NO equation:
      - Lag Δy_NO coefficient: 0.239 (std err 0.070)
      - Lag Δ(G/Y_NO): 0.083 (0.029)
      - Lag Δp_O: 0.024 (0.015)
      - c: 0.052 (0.009)
      - R-squared 0.356; Adj. R-squared 0.302
    - Δ(G/Y_NO) equation:
      - Lag Δy_NO: -0.398 (0.229)
      - Lag Δ(G/Y_NO): -0.032 (0.095)
      - Lag Δp_O: 0.159 (0.049)
      - c: -0.010 (0.029)
      - R-squared 0.137; Adj. R-squared 0.065
  - For High Oil Share:
    - Δy_NO equation:
      - Lag Δy_NO: 0.178 (0.082)
      - Lag Δ(G/Y_NO): 0.108 (0.033)
      - Lag Δp_O: 0.023 (0.020)
      - c: 0.083 (0.014)
      - R-squared 0.410; Adj. R-squared 0.354
    - Δ(G/Y_NO) equation:
      - Lag Δy_NO: -0.430 (0.289)
      - Lag Δ(G/Y_NO): -0.052 (0.115)
      - Lag Δp_O: 0.225 (0.070)
      - c: -0.027 (0.048)
      - R-squared 0.165; Adj. R-squared 0.085
  - For GCC:
    - Δy_NO equation:
      - Lag Δy_NO: 0.085 (0.107)
      - Lag Δ(G/Y_NO): 0.097 (0.038)
      - Lag Δp_O: 0.018 (0.027)
      - c: 0.088 (0.015)
      - R-squared 0.451; Adj. R-squared 0.387
    - Δ(G/Y_NO) equation:
      - Lag Δy_NO: -0.690 (-0.430) [note: table shows negative standard error formatting issue; entries reported as provided]
      - Lag Δ(G/Y_NO): -0.082 (0.153)
      - Lag Δp_O: 0.293 (-0.106) [note: table shows coefficient 0.293 and standard error -0.106 as provided]
      - c: -0.012 (-0.059)
      - R-squared 0.189; Adj. R-squared 0.094
- Key VAR conclusions:
  - Lagged changes in the public spending ratio and in oil prices are significant determinants of the non-oil output cycle in the VAR, though oil prices are only marginally significant for the high-oil group and not significant for the GCC group (Table 7).
  - Oil prices significantly affect the fiscal stance across full sample and sub-samples, with the estimated effect larger for the high-oil group and larger still for the GCC group.
- Analytical next steps (as described in paper structure): Use impulse responses from the panel VAR to assess transmission of oil price shocks through fiscal policy, compare with counterfactuals where fiscal policy is exogenous.

*Italic: Content based solely on _wp08253 - References..............................................................................................................*

### Appendix B summarizes the unit root tests for the variables used in the VAR.

### _wp08253 - Appendix B summarizes the unit root tests for the variables used in the VAR.

### Impulse response findings (main VAR results)
- Impulse responses show a one standard deviation shock to the oil price results in a significant increase in non-oil output and public spending for the full sample (Chart 3).
- Shocks to public spending have a positive effect on non-oil output.
- Ordering in the VAR: non-oil output first, government spending second, oil prices last.
- For the high-oil group and the GCC group:
  - The responsiveness of non-oil output to oil price shocks weakens as the sample is reduced to the high-oil group.
  - For the GCC group, the effect of oil price shocks on non-oil output is not significantly different from zero (Chart 4).
- Oil prices significantly affect public spending across samples, and spending shocks significantly impact the output cycle where public spending is relatively important.

### Fiscal channel vs. direct effect of oil price shocks
- Modified impulse response (fiscal channel shut down by assuming fiscal spending exogenous) isolates the direct effect of oil price shocks on output.
- Results:
  - When fiscal stance is not changed in response to oil price shocks, the response of non-oil output to a one standard deviation oil price shock is not different from zero at the 5 percent significance level for the full sample (Chart 5).
  - For the high-oil and GCC groups, the impact is insignificantly different from zero and the estimated mean magnitude is lower when fiscal policy is held fixed.
- Interpretation:
  - Oil price changes do not have a significant independent effect—directly or via confidence and/or monetary and financial effects—on the underlying non-oil economic cycle in oil-exporting countries.
  - Movements in oil prices tend to be associated with changes in the fiscal stance, which in turn affect the economic cycle.
  - Intuition: oil revenue principally accrues to the government in these countries; if additional oil income is saved (accumulated abroad) the domestic non-oil economy is unaffected, whereas increased government spending affects output, especially where public spending is large relative to the economy.

### Sensitivity checks and extensions (robustness)
- Main results were robust to:
  - Small changes in the sample period.
  - Addition of one more lag in the VAR.
  - Replacing government spending ratio with the fiscal impulse.
  - Adding real oil output changes to the VAR.
  - Using an alternative measure of the cyclical component of non-oil output (HP-filtered non-oil output).
- Fiscal impulse specification:
  - Fiscal impulse captures changes in the spending ratio and non-oil taxation (via non-oil revenue ratio) and broadens transmission channels.
  - Results similar to government spending ratio specification: with endogenous fiscal policy, oil price shocks have marginal significance for the full sample and high-oil group; with fiscal policy fixed, oil price shocks do not significantly affect output in any group (Chart 6).
  - Lagged output changes are not a significant determinant of the fiscal impulse in the VAR—so oil prices affecting fiscal stance via output is not a major concern.
- Sample extension to include 1980–89 for Iran, Norway, and Saudi Arabia:
  - Yields a significant, albeit small, direct effect of oil price shocks on output for the full sample even with fiscal channel closed — possibly due to different transmission in the 1980s or weaker data quality.
  - High-oil and GCC group results remain unaffected.
- Adding oil output as an exogenous variable:
  - Explanatory power of VAR equations improved somewhat.
  - Oil price shocks continued to have an insignificant—or marginally significant for the full sample—effect on non-oil output when fiscal policy was held fixed.
- HP-filtered non-oil output:
  - Main result strengthened: oil price changes did not have a significant impact on the output cycle for full sample, high-oil, and GCC groups even when fiscal policy responded.
  - Caveat: HP filter sensitivity to endpoint problems in short time series may affect accuracy.

### Policy-relevant conclusions
- Strong empirical support against the view that fiscal policy reactions to oil price shocks amplify the underlying business cycle in oil-exporting countries—especially where the non-oil sector is relatively small.
- In countries where public spending occupies a large share of the economy, spending increases will by definition increase the cyclical component of output; whether such spending coincides with oil price changes does not change the degree of cyclical pressure.
- Caution: fiscal expansion in response to positive oil price shocks is not necessarily optimal; assessing optimality requires weighing costs from increased cyclical pressure against benefits of greater spending, which depend on magnitude and persistence of the price shock.
- The analysis implies that cyclical-stability costs associated with fiscal responses will not be larger during periods of high oil prices.

### Appendix A — Data sources and variable definitions (as provided)
- p_O: World oil price, defined as a natural log of real annual average petroleum spot price. Source: IMF, International Financial Statistics (IFS).
- y_NO: Real non-oil output, defined as a natural log of real non-oil GDP. Source: Country desk data.
- G/Y_NO: Fiscal spending, defined as a ratio of central government expenditure to nominal non-oil GDP. Sources: IMF IFS, and country desk data.
  - Government expenditure for Norway is based on general government data.
  - For UAE the government expenditure contains consolidated accounts of the federal government, and the emirates Abu Dhabi, Dubai, and Sharjah.
- R_NO/Y_NO: Non-oil revenue, defined as a ratio of nominal non-oil revenue to nominal non-oil GDP. Sources: IMF IFS, and country desk data.
  - Non-oil revenue for UAE is based on consolidated accounts of the federal government, and the emirates Abu Dhabi, Dubai, and Sharjah.
- Note: The definition and quality of data related to the non-oil sector varies across countries in the sample, and may include some oil-related activities.

### Appendix B — Unit root tests (Augmented Dickey-Fuller results; Table B1)
- General findings:
  - ADF tests for unit roots were performed for y_NO and G/Y_NO for each country, and for oil prices.
  - For all the time series, except Yemen (y_NO) and Iran (G/Y_NO), the presence of a unit root could not be rejected.
  - Oil prices follow a random walk: test statistics equalled –1.77 (p_O) and –4.92*** (Δp_O) for levels and first differences respectively.
  - Johansen’s log likelihood tests found one cointegrating equation only for Norway, Saudi Arabia, and UAE.
- ADF tests on first differences (Δy_NO, ΔG/Y_NO, and fiscal impulse) with a constant:
  - All first differences were stationary except for:
    - Oman and UAE (Δy_NO)
    - Norway and Libya (ΔG/Y_NO)
  - Where unit root could not be rejected in differenced data, small sample size likely caused imprecision.
  - Additional checks:
    - Error terms from country-specific, equation-by-equation estimations were stationary in all cases and white noise in most cases.
    - Panel robustness check excluding countries that failed ADF on differenced data produced qualitatively and quantitatively similar results.

- Table B1 (ADF Unit Root Test) — rows reproduced exactly as in source:
  - Algeria 1990 2007 -0.431 -0.97 -3.54 *1 -6.27 *** -3.54 **1
  - Iran 1980 2007 -1.941 -3.27 ** -6.94 ***2 -2.65 * -4.79 ***
  - Kuwait 1985 2007 -1.10 -1.492 -4.70 *** -4.32 ***1 -4.11 ***1
  - Libya 1990 2007 -1.431 1.65 -9.71 ***2 -2.25 -3.31 *1
  - Nigeria 1981 2007 -0.64 -2.14 -5.19 *** -3.63 ** -3.51 **1
  - Norway 1980 2007 -2.021 -2.502 -3.13 *1 -2.46 -4.31 **2
  - Oman 1990 2007 -0.92 -1.431 -2.78 -4.37 *** -4.24 **
  - Saudi Arabia 1980 2007 -1.481 -2.18 -4.52 *** -4.76 *** -5.20 ***
  - UAE 1990 2007 -2.41 -2.121 -0.62 -7.83 *** -5.70 ***
  - Yemen 1990 2007 -4.97 ***1 -0.71 -7.91 ***1 -4.54 *** -5.30 ***1

- Notes from Table B1:
  - DA stands for degree of augmentation.
  - * denotes significance at 10 percent, ** at 5 percent, *** at 1 percent.
  - Table header variables: y_NO, G/Y_NO, Δ(G/Y_NO), Δy_NO, imp, Time period (as presented in the source).
- Additional footnotes quoted verbatim:
  - "1 The presence of a unit root could not be rejected for the log level of oil prices (p_O), but was rejected for first differences (Δp_O). The test statistics equalled –1.77 (p_O) and –4.92*** (Δp_O), respectively."
  - "2 The results are based on both trace and maximum eigenvalue statistics with 5 percent significance level."

*Source: _wp08253 - Appendix B summarizes the unit root tests for the variables used in the VAR.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08253.pdf_