## 1. Composition of Government Expenditures in Oil Exporting

## Source details

**Canonical URL:** [1. Composition of Government Expenditures in Oil Exporting](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09112.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09112.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09112.pdf.json)

---

### Introduction and central hypothesis
- Proposes a new explanation for the “resource curse”: private agents anticipate future government current spending financed through natural resource revenues, which reduces private saving and investment and hampers non-hydrocarbon private-sector growth.
- Empirical strategy: estimate non-hydrocarbon sector growth regressions using a panel of 25 oil-exporting countries over 1992–2005.
- Places the proposed income-effect channel in context of existing literature:
  - Dutch disease explanations (Wijnbergen (1984), Corden (1984)).
  - Institutional/rent-seeking explanations (Tornell and Lane (1999); Robinson et al. (2006); Sala-i-Martin and Subramanian (2004)).
  - Other channels: revenue volatility and excessive borrowing (Hausmann and Rigobon (2004)); non-productive spending categories (Tanzi and Schuknecht (2000)).

### Empirical evidence and key statistics
- Aggregate finding:
  - Current government spending in oil-exporting countries increased over recent years, "amounting to 36 percent of non-hydrocarbon GDP over 1999-2005."
- Sample window and scope:
  - Panel of 25 oil-exporting countries over 1992–2005.
- Figure 1 summary:
  - Negative relationship between non-hydrocarbon GDP growth per capita and current expenditure as a share of non-hydrocarbon GDP, average over 1992–2005; pattern shown for: Algeria, Bahrain, Cameroon, Chad, Congo, Ecuador, Equatorial Guinea, Gabon, Indonesia, Iran, Kazakhstan, Kuwait, Mexico, Nigeria, Norway, Oman, Russia, Saudi Arabia, Sudan, Syria, Trinidad and Tobago, UAE, Venezuela, Vietnam, Yemen.
- Data sources explicitly cited: Ossowski and al. (2008) and authors' own calculations.
- Table 1 (numbers preserved verbatim as shown in source):
  - 32.8
  - 53.0
  - 37.6
  - 13.7
  - 2.8
  - 15.1
  - 74.5
  - 53.7
  - 13.4
  - 3.2
  - 10.8
  - 51.6
  - 34.4
  - 33.6
  - 14.0
  - 4.9
  - 8.2
  - 38.9
  - 20.0
  - 13.6
  - 3.9
  - 10.9
  - 53.9
  - 2005
  - 47.242.047.2
  - 1999-2005                           19992003
  - 36.036.2

### Theoretical framework (simple lifecycle model) — mechanism and implications
- Model setup (overlapping generations):
  - Young supply labor inelastically; old do not supply labor.
  - Young: wage wt, consumption cyt, saving at, government transfer gyt.
  - Old: interest rt+1 kt+1 on saved capital, transfer got+1, consumption cot+1.
  - Define gt = gyt + got+1/(1 + rt+1).
  - Lifetime budget: cyt + cot+1/(1+rt+1) = wt + gt  [equation [1]].
  - Utility: U = (c_yt^(1-α) + c_ot+1^(1-α))/(1-α).
  - First-order conditions:
    - c_yt = α(wt + gt)  [equation [2]]
    - c_ot+1 = (1-α)(1 + rt+1)wt + (1-α)gt  [equation [3]]
  - Production: Cobb-Douglas with parameters A, K, L, β [equations [4] and [5]].
  - Capital accumulation/transition equations derived [equations [6] and [7]] show how transfers affect kt+1.
- Mechanism:
  - Anticipated transfers to consumers (non-savers) raise expected lifetime income of the young while they are savers, inducing higher consumption and lower saving → lower private capital accumulation → lower non-hydrocarbon GDP growth (income-effect channel).
- Comparative static:
  - Transition path of capital is higher if more transfers are made to the young and lower if more anticipated transfers are made to the old.

### Policy-relevant cases and comparative outcomes (model cases)
- Case 1 (Resource Curse): gyt = 0 and got+1 > 0
  - Natural resource revenues finance transfers to domestic consumers (the old) only.
  - Transition path of the non-resource sector with transfers is lower than without transfers → resources cause a curse.
- Case 2 (Resource Blessing): gyt > 0 and got+1 = 0
  - Resources finance transfers to savers (the young) only.
  - Transition path with transfers to the young is higher than without transfers → resources cause a blessing.
- Case 3 (tax the young and subsidize the old): gyt < 0 and got+1 > 0
  - Savers (the young) are taxed to finance transfers to consumers (the old).
  - Non-resource sector transition path is lower than in absence of transfers (holds whether budget is balanced gyt = -got or not).
- Case 4 (subsidize the young and tax the old): gyt > 0 and got+1 < 0
  - Consumers (the old) are taxed to make transfers to savers (the young).
  - Non-resource sector transition path is higher than in absence of transfers.
- Policy implication emphasized:
  - Governments can “reverse the curse” by policies that motivate increased private capital accumulation and decreased current expenditures; channel resource revenues toward savers (capital accumulation) rather than current consumption.

### Extensions and comparative statics on openness
- Open-economy extension (Appendix C):
  - Transition equation for capital accumulation modified to include per capita foreign investment ft: Equation [8] incorporates ft.
  - ft = f(rclosed,t, rworld,t+1) with 0 > ∂f/∂rclosed.
  - Capital mobility constraint: capital cannot flow beyond kmax,t+1 f, where kmax,t+1 = (1−τ)k, τ ∈ [0,1]; τ = 1 fully closed, τ = 0 fully open.
  - Effect of an increase in current expenditures (Δgot+1):
    - If 1 1 1 max, 1 1    Δ  > − ot t f t f t g r kk α , i.e., foreign capital is well below the maximum, foreign capital inflows neutralize the effect of current expenditure on domestic capital accumulation, leaving growth unchanged.
    - Otherwise, foreign capital flows in up to kmax,t+1 f and the capital loss due to the expansion in current expenditure is (expression presented verbatim in source).
  - Comparative-static conclusion:
    - Loss is positive and is a decreasing function of kmax,t+1 f.
    - Increased openness (higher kmax,t+1 f, i.e., lower τ) reduces the extent to which the income effect dampens capital accumulation.
- Implication stated:
  - The income-channel resource curse is stronger for countries with lower degrees of openness.

### Altruism and the resource curse (theoretical)
- Two types of altruism:
  - Forward altruism: old → young transfers (positive φ).
  - Backward altruism: young → old transfers (negative φ).
- Parameterization: φ ∈ (−1,1).
- Modified transition equation (closed economy) includes φ (Equation [9]).
- Theoretical implications:
  - Positive φ raises the transition path of capital; negative φ lowers it.
  - For gyt = 0 and got+1 > 0, resource curse (blessing) occurs if 1(1 1)1 )] t r [( φα φα + − − − + is negative (positive) — expression presented verbatim in source.
  - Higher forward altruism tends to dampen the income effect; sufficiently large forward altruism can make transfers a blessing.

### Empirical investigation: data and methodology
- Data:
  - Annual data for 25 oil-exporting countries over 1992–2005.
  - On average, nine observations per country used due to missing observations and removal of outliers.
  - Resource sector: hydrocarbon sector (oil/gas and derivatives). Non-resource sector: non-hydrocarbon sector.
- Baseline empirical model:
  - Dependent variable: non-hydrocarbon GDP growth (NH-GDPGrowthit).
  - Explanatory variables:
    - Lagged GDP per capita (yit-1).
    - Government current spending as a percentage of non-hydrocarbon GDP (Spendingit), anticipated at t−1.
    - Control vector: institutional quality index (Institutionit) proxied by political risk index from ICRG (2005); REERit (rate of change of real effective exchange rate).
  - Extensions include Restrictionsit and interactions (XControlit).
- Estimation approach:
  - Fixed-effects OLS and GMM (Arellano-Bond / Blundell-Bond) estimators.
  - Instrumental variables with up to three-year lags; unit value of hydrocarbon export and time dummies used as instruments for government current spending.
  - Tests: Hausman supported fixed effects over random effects; first- and second-order serial correlation tests and Hansen test used for specification validity.
- Baseline reduced form:
  - Δyt = α0 + α1 yit-1 + α2 Spendingit + α3 Controlit + μi + εit

### Empirical results — baseline specification (Table 2)
- Preferred specification: Column (2) GMM-SYS.
- Key coefficient estimates (preserved exactly from source):
  - Coefficient on lagged non-hydrocarbon GDP: estimated at -0.08 (GMM-SYS reported as -0.0827 (0.0536) in Table 2).
  - Coefficient on current expenditure (Spending): estimated at -0.18 (GMM-SYS reported as -0.183** (0.0808) in Table 2).
    - Interpretation in source: an increase in current spending of 1 percent of non-hydrocarbon GDP reduces non-hydrocarbon sector growth by about 0.2 percent.
  - Coefficient on institutional quality index (Institution): estimated at 0.016 (GMM-SYS reported as 0.0156*** (0.0037)).
  - Coefficient on REER: estimated at 0.0001 (GMM-SYS reported as 0.0001 (0.0016)).
- Exact Table 2 values (as presented):
  - Initial NH-GDP per capita:
    - OLS Fixed Effects (1) = -0.2434*** (0.0482)
    - GMM-SYS (2) = -0.0827 (0.0536)
  - Spending:
    - (1) = -0.0646** (0.0299)
    - (2) = -0.183** (0.0808)
  - Institution:
    - (1) = 0.0036** (0.0014)
    - (2) = 0.0156*** (0.0037)
  - REER:
    - (1) = 0.0004 (0.0005)
    - (2) = 0.0001 (0.0016)
  - R-squared (reported for OLS): 0.678.
  - Hansen test (GMM-SYS): 0.162.
  - Serial correlation tests: first order = 0.014; second order = 0.178.
  - Number of countries = 25; Observations = 221.
  - Note on significance: * significant at 10%; ** significant at 5%; *** significant at 1%.
- Robustness:
  - Using non-wage government primary spending as Spending proxy yields a negative coefficient.
  - Including government capital spending does not change main results; coefficient on government capital spending not statistically significant.
  - Results robust to three-year averages.

### Empirical results — openness (Restrictions) interaction (Table 3)
- Restrictionsit: index (simple average of four IMF-derived dummies) measuring exchange arrangements and exchange restrictions; ranges from 0 (open) to 1 (closed).
  - Components: multiple exchange rate practice, current account restrictions, capital account restrictions, surrenders of exports.
- Theoretical prediction tested:
  - Higher openness (lower Restrictions) dampens the negative impact of government current spending on non-hydrocarbon sector development.
- Key empirical finding (focus on column (4) in Table 3):
  - In the absence of restrictions (Restriction = 0), the coefficient on Spending is positive: Spending = 0.291** (0.141) in column (4).
  - For a highly restricted country (Restriction = 1), the implied coefficient on Spending equals 0.29 − 0.70 = −0.41.
- Exact Table 3 values (selected as presented):
  - Initial NH-GDP per capita:
    - (1) OLS = -0.2544*** (0.0474)
    - (2) Fixed Effects GMM-SYS = -0.1474*** (0.0535)
    - (3) OLS = -0.2477*** (0.0483)
    - (4) GMM-SYS = -0.2154*** (0.0944)
  - Spending:
    - (1) = -0.0354 (0.0484)
    - (2) = -0.0285 (0.117)
    - (3) = -0.0488 (0.0476)
    - (4) = 0.291** (0.141)
  - Institution:
    - (1) = 0.0035** (0.0014)
    - (2) = 0.0121*** (0.0040)
  - REER:
    - (1) = 0.0003 (0.0005)
    - (2) = 0.0005 (0.0014)
    - (3) = 0.0005 (0.0005)
    - (4) = 0.0021 (0.0020)
  - Restriction:
    - (1) = 0.154 (0.253)
    - (2) = 0.889 (0.687)
    - (3) = 0.2172 (0.246)
    - (4) = 0.254** (0.921)
  - Spending × Restriction:
    - (1) = -0.0590 (0.0802)
    - (2) = -0.275 (0.210)
    - (3) = -0.0780 (0.0774)
    - (4) = -0.701** (0.308)
  - Model statistics:
    - R-squared (reported for OLS columns) ~ 0.68 (varies by column).
    - Hansen test values reported: 0.729 and 0.510 in some columns.
    - Serial correlation tests (first order) ~ 0.039 and 0.056; (second order) ~ 0.147 and 0.077.
    - Number of countries = 25; Observations = 221 per specification.
- Interpretation:
  - Interaction results support the theoretical prediction that fewer restrictions (greater openness) dampen the negative impact of government current spending on non-hydrocarbon growth; in sufficiently open countries the net effect of current spending can be positive.

### Appendix B — altruism empirical tests (summary)
- Purpose: test whether higher forward altruism dampens the adverse effect of government current spending on non-hydrocarbon GDP growth using regional dummies as proxies for altruism.
- Main empirical findings (Table 7):
  - Column (1) OLS Fixed Effects indicates Western Hemisphere, followed by Middle East, are most severely affected by current spending impacts even after controlling for institutional quality.
    - Column (1): interaction between government current spending and Middle East dummy = 0.197* (0.117); government current spending = -0.260** (0.109).
  - Column (2) GMM-SYS does not indicate significant interactive effects.
- Detailed Table 7 reported coefficients (as presented):
  - Initial NH-GDP per capita:
    - Column (1): -0.2518*** (0.0495)
    - Column (2): -0.1699** (0.0786)
  - Spending:
    - Column (1): -0.260** (0.109)
    - Column (2): -0.0578 (0.105)
  - Spending × Africa Dummy:
    - Column (1): 0.188 (0.115)
    - Column (2): -0.0697 (0.0436)
  - Spending × Asia Dummy:
    - Column (1): 0.274** (0.107)
    - Column (2): 0.0002 (0.0243)
  - Spending × Middle East Dummy:
    - Column (1): 0.197* (0.117)
    - Column (2): 0.0228 (0.0345)
  - Spending × Europe Dummy:
    - Column (1): 0.444 (0.278)
    - Column (2): -0.0699 (0.126)
  - Institution:
    - Column (1): 0.0035** (0.0016)
    - Column (2): 0.0157*** (0.0053)
  - REER:
    - Column (1): 0.0002 (0.0005)
    - Column (2): 0.0011 (0.0013)
  - Other statistics:
    - R-squared 0.68
    - Hansen test 0.949
    - Serial correlation test (first order) 0.019
    - Serial correlation test (second order) 0.191
    - Number of countries 2525 (as presented)
    - Observations 221221 (as presented)
- Authors' caution:
  - Results are mixed; GMM-SYS corrects for endogeneity and does not confirm significant interactive effects.
  - Caution due to small number of countries in regional groups and limitations of regional dummies as proxies for altruism.

### Conclusion: income-effect channel of the resource curse and policy implications
- Core conclusion:
  - The paper offers an alternative explanation for the “resource curse” based on the income effect of high government current expenditures financed by natural resource revenues that reduce private saving and investment in the non-resource sector.
- Empirical support:
  - Evidence from panel regressions for 25 oil-exporting countries over 1992–2005 supports main theoretical predictions.
- Policy implications to dampen this channel:
  - limit current transfers;
  - further liberalize international goods and capital movements;
  - introduce policies that promote domestic private investment.
- Suggested model extensions:
  - Include human capital accumulation to link lower investment in human capital to high transfers.
  - Allow for labor supply endogeneity to explain lower labor supply in resource rich countries resulting from expectations of future transfers.

*Source: _wp09112 - 1. Composition of Government Expenditures in Oil Exporting (IMF working paper chapter/PDF content provided).*

### 1. Composition of Government Expenditures in Oil Exporting ................................................4

### 1. Composition of Government Expenditures in Oil Exporting

### Introduction and central hypothesis
- Proposes a new explanation for the “resource curse”: private agents anticipate future government current spending financed through natural resource revenues, which reduces private saving and investment and hampers non-hydrocarbon private-sector growth.
- Empirical strategy: estimate non-hydrocarbon sector growth regressions using a panel of 25 oil-exporting countries over 1992–2005.
- Places the proposed income-effect channel in context of existing literature:
  - Dutch disease explanations (Wijnbergen (1984), Corden (1984))—real exchange rate appreciation harming non-resource sectors.
  - Institutional/rent-seeking explanations (Tornell and Lane (1999); Robinson et al. (2006); Sala-i-Martin and Subramanian (2004)).
  - Other channels: revenue volatility and excessive borrowing (Hausmann and Rigobon (2004)); non-productive spending categories (Tanzi and Schuknecht (2000)).

### Empirical evidence and key statistics
- Reported aggregate finding: current government spending in oil-exporting countries increased over recent years, "amounting to 36 percent of non-hydrocarbon GDP over 1999-2005."
- Paper notes that many oil-exporting countries also undertook large public investment plans, but this paper does not analyze the impact of public investment on private-sector growth.
- Table 1 (Composition of Government Expenditures in Oil Exporting) — numbers shown in source (preserved verbatim):
  - 32.8
  - 53.0
  - 37.6
  - 13.7
  - 2.8
  - 15.1
  - 74.5
  - 53.7
  - 13.4
  - 3.2
  - 10.8
  - 51.6
  - 34.4
  - 33.6
  - 14.0
  - 4.9
  - 8.2
  - 38.9
  - 20.0
  - 13.6
  - 3.9
  - 10.9
  - 53.9
  - 2005
  - 47.242.047.2
  - 1999-2005                           19992003
  - 36.036.2
- Figure 1 (summary from source): there is a negative relationship between non-hydrocarbon GDP growth per capita and current expenditure as a share of non-hydrocarbon GDP, on average over 1992–2005; this pattern is shown for the following countries in the sample: Algeria, Bahrain, Cameroon, Chad, Congo, Ecuador, Equatorial Guinea, Gabon, Indonesia, Iran, Kazakhstan, Kuwait, Mexico, Nigeria, Norway, Oman, Russia, Saudi Arabia, Sudan, Syria, Trinidad and Tobago, UAE, Venezuela, Vietnam, Yemen.
- Data sources explicitly cited in text/figure: Ossowski and al. (2008) and authors' own calculations.
- Sample window and scope explicitly stated: panel of 25 oil-exporting countries over 1992–2005.

### Theoretical framework (simple lifecycle model) — mechanism and implications
- Model setup:
  - Overlapping generations (young and old) with population size N each period; young supply labor inelastically; old do not supply labor.
  - Single consumption/investment good; young receive wage wt, consumption cyt, saving at, and a government transfer gyt.
  - Old receive interest rt+1 kt+1 on saved capital and a transfer got+1; consumption cot+1.
  - Define gt = gyt + got+1/(1 + rt+1).
  - Lifetime budget: cyt + cot+1/(1+rt+1) = wt + gt  [equation labeled [1] in source].
  - Utility specification: U = (c_yt^(1-α) + c_ot+1^(1-α))/(1-α) (presented in source as the lifetime utility where α captures the marginal propensity to consume when young).
  - First-order conditions yield:
    - c_yt = α(wt + gt)  [equation [2] in source]
    - c_ot+1 = (1-α)(1 + rt+1)wt + (1-α)gt  [equation [3] in source]
  - Production: Cobb-Douglas with parameters A, K, L, β (expressions in the source give marginal products and relations for r_t and w_t) [equations [4] and [5] in source].
  - Capital accumulation/transition equation derived in source (equations [6] and [7] in source) show how transfers affect next-period capital kt+1.
- Mechanism: anticipated transfers to consumers (non-savers) raise expected lifetime income of the young while they are savers, inducing higher consumption and lower saving → lower private capital accumulation → lower non-hydrocarbon GDP growth (income-effect channel of the resource curse).
- Comparative static from model: transition path of capital is higher if more transfers are made to the young and lower if more anticipated transfers are made to the old.

### Policy-relevant cases and comparative outcomes (as presented)
- Case 1 (Resource Curse): gyt = 0 and got+1 > 0
  - Natural resource revenues finance transfers to domestic consumers (the old) only.
  - Transition path of the non-resource sector with transfers is lower than without transfers → resources cause a curse.
- Case 2 (Resource Blessing): gyt > 0 and got+1 = 0
  - Resources finance transfers to savers (the young) only.
  - Transition path with transfers to the young is higher than without transfers → resources cause a blessing.
- Case 3 (tax the young and subsidize the old): gyt < 0 and got+1 > 0
  - Savers (the young) are taxed to finance transfers to consumers (the old).
  - Non-resource sector transition path is lower than in absence of transfers (holds whether budget is balanced gyt = -got or not).
- Case 4 (subsidize the young and tax the old): gyt > 0 and got+1 < 0
  - Consumers (the old) are taxed to make transfers to savers (the young).
  - Non-resource sector transition path is higher than in absence of transfers.
- Policy implication emphasized in model section: governments can “reverse the curse” by policies that motivate increased private capital accumulation and decreased current expenditures; revenue management that channels resource revenues toward savers (capital accumulation) rather than current consumption can transform a curse into a blessing.

### Extensions and comparative statics on openness
- The paper extends the model to show that the income-channel resource curse is stronger for countries with lower degrees of openness and lower degrees of forward altruism.
- Openness is studied with respect to capital flows (small open economy framework with restrictions on pass-through of foreign capital inflows); the more open an economy is, the less it is prone to the income-effect resource curse (as stated in the source).

*Source: _wp09112 - 1. Composition of Government Expenditures in Oil Exporting (IMF working paper chapter).*

### Appendix C provides the model’s extension to the larger open economy case, which can be

### _wp09112 - Appendix C provides the model’s extension to the larger open economy case, which can be

### Open-economy extension of the transition equation
- The transition equation for capital accumulation is modified to include per capita foreign investment ft:
  - Equation [8] (as presented) incorporates ft into the dynamic of capital accumulation.
- ft is a function of domestic interest rates and world rates, with ft increasing when domestic rates rise relative to world rates:
  - ft = f(rclosed,t, rworld,t+1) with 0 > ∂f/∂rclosed.
- Capital mobility constraint:
  - Capital cannot flow beyond kmax,t+1 f, where kmax,t+1 = (1−τ)k, with τ ∈ [0,1] measuring de jure restrictions on cross-border capital mobility and k the steady-state capital in the closed economy.
  - τ = 1 denotes de jure fully closed economy; τ = 0 denotes de jure fully open economy.
- Effect of an increase in current expenditures (Δg ot+1):
  - If 1
    1
    1
    max,
    1
    1
    
    
    
    Δ
    
    >
    −
    ot
    t
    f
    t
    f
    t
    g
    r
    kk
    α
    , i.e., foreign capital is well below the maximum, foreign capital inflows neutralize the effect of current expenditure on domestic capital accumulation, leaving growth unchanged.
  - Otherwise, foreign capital flows in up to kmax,t+1 f and the capital loss due to the expansion in current expenditure is:
    - f
      t
      max,
      1
      k
      0)(
      1
      1
      
      
      t
      r
      1
      max,
      1
      
      
      f
      t
      f
      tot
      kkgLoss
      
- Comparative-static conclusion:
  - Loss is positive and is a decreasing function of kmax,t+1 f.
  - Increased openness (higher kmax,t+1 f, i.e., lower τ) reduces the extent to which the income effect dampens capital accumulation.

### Altruism and the resource curse (theoretical)
- Two types of altruism considered:
  - Forward altruism: old → young transfers (positive φ).
  - Backward altruism: young → old transfers (negative φ).
- Parameterization:
  - φ ∈ (−1,1), positive values represent forward altruism; negative values represent backward altruism.
- Modified transition equation for closed economy with altruism:
  - Equation [9] (as presented) modifies the transition equation to include φ:
    - The transition equation includes terms with αφ and αφαβ etc., as shown in the source.
- Theoretical implications:
  - A positive φ (forward altruism) brings the transition path of capital to a higher level; a negative φ (backward altruism) brings it to a lower level.
  - Whether resource curse (or blessing) occurs depends on the sign of a derived expression involving φ, α, gt, rt, and gyt:
    - For gyt = 0 and got+1 > 0, resource curse (blessing) occurs if 1(1 1)1 )] t r [( φα φα + − − − + is negative (positive) — (expression presented verbatim in source).
  - A higher degree of forward altruism tends to dampen the income effect of the resource curse; sufficiently large forward altruism can make transfers via natural resources a blessing.

### Empirical investigation: data and methodology
- Data:
  - Annual data for 25 oil-exporting countries over 1992–2005.
  - On average, nine observations per country are used in regressions due to missing observations and removal of outliers.
  - Resource sector: hydrocarbon sector (oil/gas and derivatives).
  - Non-resource sector: non-hydrocarbon sector.
- Baseline empirical model:
  - Dependent variable: non-hydrocarbon GDP growth (NH-GDPGrowthit).
  - Explanatory variables (in natural logarithm unless indicated):
    - Lagged GDP per capita (yit-1).
    - Government current spending as a percentage of non-hydrocarbon GDP (Spendingit), anticipated at t−1.
    - Vector of control variables (Controlit): institutional quality index (Institutionit) proxied by political risk index from ICRG (2005); REERit (rate of change of real effective exchange rate).
  - Extension includes XControlit (Restrictionsit and interactions).
- Estimation approach:
  - Fixed-effects OLS and GMM (Arellano-Bond / Blundell-Bond) estimators used.
  - Instrumental variables with up to three-year lags; unit value of hydrocarbon export and time dummies used as instruments for government current spending.
  - Tests: Hausman supported fixed effects over random effects; first- and second-order serial correlation tests and Hansen test used for specification validity.
- Baseline reduced form estimated as:
  - Δyt = α0 + α1 yit-1 + α2 Spendingit + α3 Controlit + μi + εit
  - Extension model includes XControlit and additional α coefficients.

### Empirical results — baseline specification (Table 2)
- Column (2) (GMM-SYS preferred specification) summarized results:
  - Coefficient on lagged non-hydrocarbon GDP: estimated at -0.08.
  - Coefficient on current expenditure (Spending): estimated at -0.18.
    - Interpretation: an increase in current spending of 1 percent of non-hydrocarbon GDP reduces non-hydrocarbon sector growth by about 0.2 percent.
  - Coefficient on institutional quality index (Institution): estimated at 0.016.
  - Coefficient on REER: estimated at 0.0001 (positive but not statistically significant).
- Additional robustness findings:
  - Using non-wage government primary spending as Spending proxy yields a negative coefficient.
  - Including government capital spending does not change main results; coefficient on government capital spending not statistically significant.
  - Results robust to three-year averages instead of yearly frequency.
- Exact table values (Table 2):
  - Initial NH-GDP per capita: OLS Fixed Effects (1) = -0.2434*** (0.0482); GMM-SYS (2) = -0.0827 (0.0536).
  - Spending: (1) = -0.0646** (0.0299); (2) = -0.183** (0.0808).
  - Institution: (1) = 0.0036** (0.0014); (2) = 0.0156*** (0.0037).
  - REER: (1) = 0.0004 (0.0005); (2) = 0.0001 (0.0016).
  - R-squared (reported for OLS): 0.678.
  - Hansen test (GMM-SYS): 0.162.
  - Serial correlation tests: first order = 0.014; second order = 0.178.
  - Number of countries = 25; Observations = 221.
  - Note on significance: * significant at 10%; ** significant at 5%; *** significant at 1%.

### Empirical results — openness (Restrictions) interaction (Table 3)
- Restrictionsit: an index (simple average of four IMF-derived dummies) measuring exchange arrangements and exchange restrictions; ranges from 0 (open) to 1 (closed).
  - Components: multiple exchange rate practice, current account restrictions, capital account restrictions, surrenders of exports.
- Main theoretical prediction tested:
  - Higher openness (lower Restrictions) dampens the negative impact of government current spending on non-hydrocarbon sector development.
- Key empirical finding (focus on column (4) in Table 3):
  - In the absence of restrictions (Restriction = 0), the coefficient on Spending is positive as indicated by the coefficient in column (4): Spending = 0.291** (0.141).
  - For a highly restricted country (Restriction = 1), the implied coefficient on Spending equals 0.29 − 0.70 = −0.41 (i.e., negative).
- Exact table values (selected from Table 3):
  - Initial NH-GDP per capita:
    - (1) OLS = -0.2544*** (0.0474)
    - (2) Fixed Effects GMM-SYS = -0.1474*** (0.0535)
    - (3) OLS = -0.2477*** (0.0483)
    - (4) GMM-SYS = -0.2154*** (0.0944)
  - Spending:
    - (1) = -0.0354 (0.0484)
    - (2) = -0.0285 (0.117)
    - (3) = -0.0488 (0.0476)
    - (4) = 0.291** (0.141)
  - Institution:
    - (1) = 0.0035** (0.0014)
    - (2) = 0.0121*** (0.0040)
  - REER:
    - (1) = 0.0003 (0.0005)
    - (2) = 0.0005 (0.0014)
    - (3) = 0.0005 (0.0005)
    - (4) = 0.0021 (0.0020)
  - Restriction:
    - (1) = 0.154 (0.253)
    - (2) = 0.889 (0.687)
    - (3) = 0.2172 (0.246)
    - (4) = 0.254** (0.921)
  - Spending × Restriction:
    - (1) = -0.0590 (0.0802)
    - (2) = -0.275 (0.210)
    - (3) = -0.0780 (0.0774)
    - (4) = -0.701** (0.308)
  - Model statistics:
    - R-squared (reported for OLS columns) ~ 0.68 (varies by column).
    - Hansen test values reported: 0.729 and 0.510 in some columns.
    - Serial correlation tests (first order) ~ 0.039 and 0.056; (second order) ~ 0.147 and 0.077.
    - Number of countries = 25; Observations = 221 per specification.
- Interpretation:
  - Interaction results support the theoretical prediction that fewer restrictions (greater openness) dampen the negative impact of government current spending on non-hydrocarbon growth; in sufficiently open countries the net effect of current spending can be positive.

*Source: _wp09112 - Appendix C provides the model’s extension to the larger open economy case, which can be (PDF chapter/section).*

### Appendix B presents the empirical results with mixed support for the impact of altruism on

### _wp09112 - Appendix B presents the empirical results with mixed support for the impact of altruism on

### Conclusion: income-effect channel of the resource curse
- The paper offers an alternative explanation for the “resource curse” based on the income effect resulting from the high level of government current expenditures in resource rich economies.
- Using a simple life cycle framework, the paper shows that private investment in the non-resource sector is negatively affected by current transfers financed through natural resource revenues because expectation of transfers dampens savings within the economy.
- Higher degrees of openness and forward altruism reduce this adverse effect.
- Empirical support for the main theoretical predictions is found by estimating non-hydrocarbon sector growth regressions using panel data for 25 oil-exporting countries over the period 1992 to 2005.
- Policy implications to dampen this channel of the resource curse:
  - limit current transfers;
  - further liberalize international goods and capital movements;
  - introduce policies that promote domestic private investment.

### Model implications and possible extensions
- The paper studies the resource curse in the standard context of capital accumulation and growth but notes the model is rich enough to study welfare implications of this channel of the resource curse.
- Suggested model extensions:
  - Include human capital accumulation to potentially link lower investment in human capital to the provision of high transfers in resource-rich countries.
  - Allow for labor supply endogeneity to explain lower labor supply in resource rich countries resulting from the expectation of future transfers.
- These extensions would enable a richer income effect explanation of the resource curse.

### Appendix B: testing whether higher forward altruism dampens the adverse effect of government current spending on non-hydrocarbon GDP growth
- Purpose: test the theoretical prediction that a higher degree of forward altruism dampens the adverse impact on saving resulting from current expenditure on non-hydrocarbon sector growth.
- Empirical approach:
  - Augment the core specification with an interaction term between current spending and regional dummies to proxy altruism using regional dummies.
  - Note: survey data could be used to proxy the degree of forward altruism but such data is not available for many sample countries.
- Main empirical findings (Table 7):
  - Column (1) (OLS Fixed Effects) indicates Western Hemisphere, followed by Middle East, are the most severely affected by the impact of current spending on non-hydrocarbon sector growth even after controlling for institutional quality.
    - The individual effect of government spending on non-hydrocarbon GDP growth by default corresponds to the Western Hemisphere.
    - In column (1), the coefficient associated with the interaction between government current spending and Middle East dummy is 0.197* (standard error (0.117)), which is lower than the absolute value of the negative coefficient associated with government current spending, -0.260** (standard error (0.109)).
  - Column (2) (GMM-SYS) corrects for potential endogeneity bias and does not indicate any significant interactive effect of government spending with regional dummies.
  - Authors caution these results due to:
    - small number of countries in regional groups (data availability);
    - questionable adequacy of using dummies to capture the degree of altruism.
- Detailed reported regression coefficients and statistics (as presented):
  - Initial NH-GDP per capita:
    - Column (1): -0.2518*** (0.0495)
    - Column (2): -0.1699** (0.0786)
  - Spending:
    - Column (1): -0.260** (0.109)
    - Column (2): -0.0578 (0.105)
  - Spending × Africa Dummy:
    - Column (1): 0.188 (0.115)
    - Column (2): -0.0697 (0.0436)
  - Spending × Asia Dummy:
    - Column (1): 0.274** (0.107)
    - Column (2): 0.0002 (0.0243)
  - Spending × Middle East Dummy:
    - Column (1): 0.197* (0.117)
    - Column (2): 0.0228 (0.0345)
  - Spending × Europe Dummy:
    - Column (1): 0.444 (0.278)
    - Column (2): -0.0699 (0.126)
  - Institution:
    - Column (1): 0.0035** (0.0016)
    - Column (2): 0.0157*** (0.0053)
  - REER:
    - Column (1): 0.0002 (0.0005)
    - Column (2): 0.0011 (0.0013)
  - Other reported statistics (as shown in Table 7):
    - R-squared 0.68
    - Hansen test 0.949
    - Serial correlation test (first order) 0.019
    - Serial correlation test (second order) 0.191
    - Number of countries 2525
    - Observations 221221
- Notes from Table 7:
  - Country dummies are included in the various specifications but estimates are not shown.
  - Observations with excessively high leverage have been dropped from the sample.
  - Robust standard errors in parentheses.
  - Significance codes: * significant at 10%; ** significant at 5%; *** significant at 1%.

### Interpretation and caution
- OLS-fixed effects results suggest regional variation in how current spending affects non-hydrocarbon growth, consistent with the hypothesis that higher forward altruism (proxied by regional dummies) can dampen adverse effects.
- System GMM results do not confirm significant interactive effects, indicating mixed empirical support for the impact of altruism on the income effect channel of the resource curse.
- The paper emphasizes cautious interpretation due to data limitations and the limitations of using regional dummies as proxies for forward altruism.

*Source: _wp09112 - Appendix B presents the empirical results with mixed support for the impact of altruism on (IMF working paper PDF content provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09112.pdf_