A. Closed Economy

We begin with the standard life cycle model. Suppose that at each period there are two coexisting generations, young and old, each with population of size N. The old do not supply any labor. Each young owns a unit of labor, which she supplies to the market inelastically. Therefore, the total labor supply would be equal to N in each period. There is only one type of good, which can be either consumed or invested.

At time t, denote the wage bill of a young individual by w_t, her consumption by c_yt, and her saving by a_t, all in units of the consumption good. In addition, each young individual receives a government transfer of g_yt, in units of consumption good.

Each individual who is young at time t becomes old at time t+1. This individual is referred to as a member of “generation t.” An old individual of generation t invests the entire savings of her young age in time t+1. Denote the capital stock due to this investment by k_t+1. An old person receives a total interest payment of r_t+1k_t+1, where r_t+1 is the real interest rate at time t+1. In addition, she receives a transfer of g_ot+1 from the government. She consumes c_ot+1 at the end of period t+1.

The lifetime budget constraint of a generation t individual would equate the present values of her lifetime consumption and income. Define g_t as the value of transfer to a generation t individual (g_t = g_yt + g_ot+1/(1 + r_t+1)). Her lifetime budget constraint will be:

Suppose a generation t individual enjoys lifetime utility of $u_t=c_{\mathit{\text{yt}}}^{\alpha }{c}_{\mathit{\text{ot}}+1}^{1-\alpha }$, where α is a

parameter that captures the marginal propensity to consume (mpc). A higher α implies a stronger taste for consumption when young, which leads to lower saving.

Each individual maximizes her utility subject to her lifetime budget constraint. The solution to this problem for a generation t individual would determine c_yt and c_ot+1 as functions of the parameters of the utility function, wage and interest rate, as follows:

and

The economy has a Cobb-Douglas production function: ${\mathit{\text{y}}}_t=A_t{K}_t^{\beta }{L}_t^{1-\beta }$, where Y, K, and L represent economy’s total output, capital and labor respectively, and A is the total factor productivity. At market equilibrium, interest rate and wage rate would be equal to marginal products of capital and labor, respectively:

As capital becomes more abundant (relative to labor), the reward to capital decreases and the reward to labor, the wage rate, increases.

All of the assets accumulated in each period will be invested in the following period:

Replace for w_t from equation [5] above, and write g_t in terms of its components to find the economy’s equation for evolution of capital (transition equation) as follows:

This equation suggests that the transition path of capital will be higher if more transfers are made to the young, and lower if more anticipated transfers are made to the old.⁴ This is intuitive, as transfers to savers encourage more (less) capital accumulation (consumption), whereas transfers to spenders encourage more (less) consumption (capital accumulation). In what follows, we study different possibilities for financing transfers and their impact for the capital accumulation in the economy:⁵

• Case 1 (Resource Curse): g_yt=0 and g_ot+1>0

In this case, natural resources are sold internationally to finance transfers to domestic consumers (the old) only. No transfers are made to the savers (the young). Figure 2 depicts the transition path without transfers and with transfers to the young. In this case, the transition path of the non resource sector of the economy with transfers is lower than without transfers, hence resources have caused a curse.

Transition Paths

Citation: IMF Working Papers 2009, 112; 10.5089/9781451872590.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Transition Paths

Citation: IMF Working Papers 2009, 112; 10.5089/9781451872590.001.A001

• Download Figure
• Download figure as PowerPoint slide

Transition Paths

Citation: IMF Working Papers 2009, 112; 10.5089/9781451872590.001.A001

• Download Figure
• Download figure as PowerPoint slide

• Case 2 (Resource Blessing): g_yt>0 and g_ot+1=0

This case is the opposite of Case 1. Resources are sold to finance transfers to savers (the young) only. No transfers are made to consumers (the old). Figure 2 above illustrates that the transition path with transfers to the young is higher than the transition path without such transfers. In other words, transfers to the young will be a blessing for the economy’s growth.

• Case 3 (tax the young and subsidize the old): g_yt<0 and g_ot+1>0

In this case, the savers (the young) are taxed to finance transfers to the consumers (the old). Similar to case 1, the non resource sector of the economy will have a lower transition path than in the absent of transfers. This holds whether the budget is balanced (g_yt = -g_ot) or not.

• Case 4 (subsidize the young and tax the old): g_yt>0 and g_ot+1<0

In this case, the consumers (the old) are taxed to make transfers to the savers (the young). This is similar to Case 2, and opposite to case 3; the non resource sector will have a higher transition path than in the absence of transfers.

Cases 1 and 2 show that the natural resource curse (or blessing) could be due to how natural resource revenues are spent potentially generating an income effect. This is different from other explanations for resource curse explained in the literature, such as the Dutch disease. Governments can “reverse the curse” by policies that motivate increased private capital accumulation and decreased current expenditures. It is clear from Cases 3 and 4 above that this policy recommendation is no different from advice on management of revenues from taxation.

B. Openness and Resource Curse

In this subsection, we show that the more open an economy is, the less it would be prone to have a resource curse through the income effect channel.⁶ This is shown by studying a small open economy framework with restrictions on the pass-through of foreign capital inflows.⁷ Openness is only studied on capital, as our single commodity model is not proper for studying commodity trade. While our main interest is the case of the small open economy, Appendix C provides the model’s extension to the larger open economy case, which can be of interest when considering regional integration between countries with similar size, but that are closed to the rest of the world.

Consider the possibility of foreign capital flows to an otherwise closed economy. The transition equation will be modified as below:

where, $k_{t+1}^f$ is the (per capita) foreign investment. It is a function of domestic interest rates, as follows: $k_{t+1}^f=f(r_{t+1}^{\mathit{\text{closed}}}-r_{t+1}^{\mathit{\text{world}}}),\frac{\partial f}{\partial r^{\mathit{\text{closed}}}}>0$. Further assume that capital cannot flow beyond $k_{t+1}^{\max ,f}=(1-\tau )\overline{k}$ where, τ ∈ [0,1] is the measure of the country’s restrictions on cross border capital mobility and $\overline{k}$ is the steady state level of capital in the closed economy. For the case of de jure fully closed economy τ = 1, and for the case of dejure fully open economy τ = 0.

Next, we will discuss the effect of an increase in current expenditures on capital accumulation. Consider an increase of Δg_ot+1 in transfers to the old. This leads to an increase in interest rates and immediate capital inflows. If $k_{t+1}^{\max ,f}-k_{t+1}^f>\frac{\alpha }{1+r_{t+1}}{\Delta }{g}_{\mathit{\text{ot}}+1}$, i.e. if the amount of foreign capital in the economy is much below the maximum allowed, foreign capital inflows neutralizes the effect of the current expenditure on domestic capital accumulation, leaving growth unchanged.⁸ Otherwise, foreign capital will flow into the country up to the maximum level of $k_{t+1}^{\max ,f}$, and the capital loss due to this expansion in capital expenditure would be: $\mathit{\text{Loss}}=\frac{\alpha }{1+r_{t+1}}{\Delta }{g}_{\mathit{\text{ot}}}-(k_{t+1}^{{max},f}-k_{t+1}^f)>0$

while Loss is positive, it is a decreasing function of $k_{t+1}^{\max ,f}$. We can thus conclude that increased openness (higher $k_{t+1}^{\max ,f}$) reduces the extent to which the income effect dampen capital accumulation.

C. Altruism and Resource Curse

This subsection shows that a higher degree of forward altruism dampens the income effect channel of the resource curse while backward altruism strengthens it. Two types of altruism are considered: the old individuals transfer part of their resources to the young (forward) and vice versa (backward). ⁹ Forward altruism strengthens the resource curse results derived in subsection A, but backward altruism weakens them; if resources are sold to finance transfers to the altruistic old, they will consume part of these resources, but transfer the rest to their children, who will save. This will weaken the resource curse through the income effect channel result found above. The opposite happens if resources are transferred to the altruistic young. Below we will formalize this intuitive argument.

Suppose the old individuals transfer fraction ϕ of their wealth to the young. Assume ϕ ∈ (–1,1), where a positive value represents positive transfers from the old to the young (forward altruism) and a negative value represents the reverse (backward altruism). The transition equation (for the closed economy case) will be as follows:

Ceteris paribus, a positive (negative) ϕ will bring the transition path of capital to a higher (lower) level. Whether resource curse (blessing) happens depends on whether $(1-\alpha )g_{\mathit{\text{yt}}}+[\varphi (1-\alpha )-\frac{(1-\varphi )\alpha }{1+r_{t+1}}]g_{\mathit{\text{ot}}}$ is negative (positive). For g_yt=0 and g_ot+1>0 resource curse (blessing) takes place if $[\varphi (1-\alpha )-\frac{(1-\varphi )\alpha }{1+r_{t+1}}]$ is negative (positive). We can thus conclude that a higher degree of forward altruism will tend to dampen the income effect. It is interesting to note that with altruistically enough parents (ϕ big enough), transfers to the old through natural resources can also be a blessing to the economy.

A. Empirical Methodology

We use a unique database, consisting of annual data for 25 oil-exporting countries over 1992-2005. Appendix A describes sources of data (Table 4) and their descriptive statistics (Table 5). Due to some missing observations in the early 1990s and the removal of outliers in the estimation, on average, nine observations per country are used in the regressions. The list of countries included in the empirical investigation is also mentioned in Appendix A (Table 6). The resource sector is the hydrocarbon sector, which includes oil/gas and its derivatives. The non-resource sector is the non-hydrocarbon sector. In the baseline empirical model, non-hydrocarbon GDP growth (NH-GDPGrowth_it) is the dependent variable, and the following are the explanatory variables (in natural logarithm, unless otherwise indicated):

• Lagged GDP per capita (y_it–1), which captures the convergence of income per capita to its steady state level.

• Government current spending as a percentage of non-hydrocarbon GDP (Spending_it), which proxies transfers to private agents and is anticipated at the prior time period (t–1), similar to the theoretical model. We have assumed an extension of our two-period model to a 55 period one, à la Auerbach and Kotlikoff (1987). All of the qualitative results derived for the two-period model apply to this extension as well.

• Vector of control variables (Control_it), which consists of the effects of other resource curse channels that are sketched in Figure 3 and are captured in the following variables: institutional quality index (Institution_it) measures the undermining of institutions channel and is proxied by political risk index from ICRG (2005); rate of change of real effective exchange rate (REER_it) proxies the Dutch disease channel. The arrows in Figure 3 indicate the direction of the causality. Thus, a two-sided arrow indicates the potential non-univocal of the causality between two elements.

Resource Curse Channels

Citation: IMF Working Papers 2009, 112; 10.5089/9781451872590.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Resource Curse Channels

Citation: IMF Working Papers 2009, 112; 10.5089/9781451872590.001.A001

• Download Figure
• Download figure as PowerPoint slide

Resource Curse Channels

Citation: IMF Working Papers 2009, 112; 10.5089/9781451872590.001.A001

• Download Figure
• Download figure as PowerPoint slide

We also test extensions of our theoretical model by including an additional vector of control variables (XControl_it). This vector includes restrictions on international goods and capital movements (Restrictions_it), and interactions of government current spending with other variables (see Appendix A for complete descriptions).

Our baseline empirical mode is as follows:

Δy_it=α + α₁y_it–1 + α₂Spending_it + α₃Control_it + μ_i + ε_it¹⁰

and our extension model is as follows:

Δy_it=α + α₁y_it–1 + α₁Spending_it + α₃Control_it + α₄XControl_it + μ_i + ε_it

μ_i captures the unobservable invariant effects for country i, which among other things can capture the degree of altruism in country i in the absence of an explanatory variable for altruism. ε_it captures the remaining unobservable effects of country i at year t, and α’s are the estimation coefficients.

B. Results

Table 2 shows the estimation results for our baseline specification, using two estimators that are fixed-effects ordinary least square (OLS) and generalized method of moments (GMM). Hausmann tests supported the choice of the fixed effects OLS model over the random effects model. The results using the fixed-effects OLS estimator with lagged dependent variables are presented in Column (1) in Table 2. A potential issue arising from using fixed-effects OLS estimator to estimate a lagged dependent model is a bias of order 0 (1/T), with T being the length of the time period of the sample. Column (2) presents the results of our preferred specification using Arrellano and Bond (1991) and Blundell and Bond (1998), Generalized Method of Moments (GMM) estimators, which corrects for the bias arising from the use of lagged dependent variables in the fixed effect model. ¹¹ In addition, we used instrumental variables techniques to correct for the endogeneity bias that may arise from the inclusion of some of the endogenous regressors in the specification, using up to three-year lagged variables as instruments. We also used the unit value of hydrocarbon export and time dummies as instruments for government current spending. We assume that movement in the unit value of hydrocarbon exports affects non-hydrocarbon growth only through its impact on fiscal policy. That assumption is supported by the some of the results of the estimation of a VAR model using a selected number of oil-exporting countries, as documented in Husain et al. (2008). The use of instrumental variables technique allows us to assert the causal impact of government current spending on non-hydrocarbon growth. Results presented in columns (1) and (2) in Table 2 are qualitatively similar, albeit the income effect is stronger when correcting for endogeneity.

Table 2.

Growth Regressions

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.* significant at 10%; ** significant at 5%; *** significant at 1%.

Table 2.

Growth Regressions

Dependent variable: NH-GDP Growth	OLS Fixed Effects (1)	GMM-SYS (2)
Initial NH-GDP per capita	-0.2434*** (0.0482)	-0.0827 (0.0536)
Spending	-0.0646** (0.0299)	-0.183** (0.0808)
Institution	0.0036** (0.0014)	0.0156*** (0.0037)
REER	0.0004 (0.0005)	0.0001 (0.0016)
R-squared	0.678
Hansen test		0.162
Serial correlation test (first order)		0.014
Serial correlation test (second order)		0.178
Number of countries	25	25
Observations	221	221

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.* significant at 10%; ** significant at 5%; *** significant at 1%.

View Table

Table 2.

Growth Regressions

Dependent variable: NH-GDP Growth	OLS Fixed Effects (1)	GMM-SYS (2)
Initial NH-GDP per capita	-0.2434*** (0.0482)	-0.0827 (0.0536)
Spending	-0.0646** (0.0299)	-0.183** (0.0808)
Institution	0.0036** (0.0014)	0.0156*** (0.0037)
REER	0.0004 (0.0005)	0.0001 (0.0016)
R-squared	0.678
Hansen test		0.162
Serial correlation test (first order)		0.014
Serial correlation test (second order)		0.178
Number of countries	25	25
Observations	221	221

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.* significant at 10%; ** significant at 5%; *** significant at 1%.

Testing whether more openness dampens the adverse effect of government current spending on non-hydrocarbon GDP growth

We further test the theoretical prediction that a higher degree of openness on international goods and capital movements dampens the reduction in investment in the private non-hydrocarbon sector generated by the expectation of government current spending. To do so, we augment the baseline specification with an interaction term between government current spending and an index of restriction to international capital and goods movements, as discussed previously. Restrictions_it is a simple average of four dummies of exchange arrangements and exchange restrictions taken from taken from IMF (2007b) namely the presence of multiple exchange rate practice, current account restrictions, capital account restrictions and surrenders of exports. A value of one for the Restrictions_it indicator implies that the country is closed to international movements of goods and capital and zero implies that the country is open to those movements. Other proxies, such as de facto openness to movements of goods and services, non-hydrocarbon openness were considered but not selected. Given the lack of diversification of most oil-exporting economies, that indicator is more likely to capture the lack of diversification than the degree of openness of government policies towards international trade in goods and service. In addition, indicators such as de facto financial integration are likely to capture large accumulation of foreign assets financed through large oil revenues rather than the degree of openness of government policies toward capital movements.

The results, presented in Table 3, support the theoretical predictions that fewer restrictions dampen the negative impact of government current spending on non-hydrocarbon sector development. We focus on column (4) that uses a specification excluding the variable Institution from the specification. Indeed, Restriction is likely to be collinear with Institution. The coefficients associated with Spending and its interaction with Restriction need now to be considered together to assess the adverse effect of current spending on the performance of the non-hydrocarbon sector. As shown in Appendix A, Restriction ranges from zero to one. Thus, the absence of restrictions on goods and capital movements in a given country will lead to a positive impact of current spending on non-hydrocarbon sector development as indicated by the coefficient associated with current expenditure in column (4) in Table 3. In the opposite case where there would be a high degree of restriction to goods and capital international movements in a given country, the implied coefficient associated with the impact of current spending equals 0.29 minus 0.70, which is negative.

Table 3.

Growth Regressions using Restrictions on Trade and Capital Flows

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.* significant at 10%; ** significant at 5%; *** significant at 1%.

Table 3.

Growth Regressions using Restrictions on Trade and Capital Flows

Dependent variable: NH-GDP Growth	OLS Fixed Effects (1)	GMM-SYS (2)	OLS Fixed Effects (3)	GMM-SYS (4)
Initial NH-GDP per capita	-0.2544*** (0.0474)	-0.1474*** (0.0535)	-0.2477*** (0.0483)	-0.2154*** (0.0944)
Spending	-0.0354 (0.0484)	-0.0285 (0.117)	-0.0488 (0.0476)	0.291** (0.141)
Institution	0.0035** (0.0014)	0.0121*** (0.0040)
REER	0.0003 (0.0005)	0.0005 (0.0014)	0.0005 (0.0005)	0.0021 (0.0020)
Restriction	0.154 (0.253)	0.889 (0.687)	0.217 (0.246)	2.254** (0.921)
Spending × Restriction	-0.0590 (0.0802)	-0.275 (0.210)	-0.0780 (0.0774)	-0.701** (0.308)
R-squared	0.68		0.67
Hansen test		0.729		0.510
Serial correlation test (first order)		0.039		0.056
Serial correlation test (second order)		0.147		0.077
Number of countries	25	25	25	25
Observations	221	221	221	221

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.* significant at 10%; ** significant at 5%; *** significant at 1%.

View Table

Table 3.

Growth Regressions using Restrictions on Trade and Capital Flows

Dependent variable: NH-GDP Growth	OLS Fixed Effects (1)	GMM-SYS (2)	OLS Fixed Effects (3)	GMM-SYS (4)
Initial NH-GDP per capita	-0.2544*** (0.0474)	-0.1474*** (0.0535)	-0.2477*** (0.0483)	-0.2154*** (0.0944)
Spending	-0.0354 (0.0484)	-0.0285 (0.117)	-0.0488 (0.0476)	0.291** (0.141)
Institution	0.0035** (0.0014)	0.0121*** (0.0040)
REER	0.0003 (0.0005)	0.0005 (0.0014)	0.0005 (0.0005)	0.0021 (0.0020)
Restriction	0.154 (0.253)	0.889 (0.687)	0.217 (0.246)	2.254** (0.921)
Spending × Restriction	-0.0590 (0.0802)	-0.275 (0.210)	-0.0780 (0.0774)	-0.701** (0.308)
R-squared	0.68		0.67
Hansen test		0.729		0.510
Serial correlation test (first order)		0.039		0.056
Serial correlation test (second order)		0.147		0.077
Number of countries	25	25	25	25
Observations	221	221	221	221

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.* significant at 10%; ** significant at 5%; *** significant at 1%.

Appendix B presents the empirical results with mixed support for the impact of altruism on the income effect channel of the resource curse.