Natural resource dependence and economic performance in the 19702000 period

Review of Development Economics, 15(1), 124–138, 2011DOI:10.1111/j.1467-9361.2010.00597.x Natural Resource Dependence and Economic
Performance in the 1970–2000 Periodrode_597124.138
George Mavrotas, Syed Mansoob Murshed, and Sebastian Torres* Abstract
We look at the type of natural resource dependence and growth in developing countries. Certain naturalresources called point-source, such as oil and minerals, exhibit concentrated and capturable revenue patterns,while revenue flows from resources such as agriculture are more diffused. Developing countries that exportthe former type of products are regarded prone to growth failure due to institutional failure. We present anexplicit model of growth collapse with micro-foundations in rent-seeking contests with increasing returns.
Our econometric analysis is among the few in this literature with a panel data dimension. Point-source-typenatural resource dependence does retard institutional development in both governance and democracy,which hampers growth. The resource curse, however, is more general and not simply confined to mineralexporters.
1. Introduction
The purpose of this paper is to explore the link between certain types of naturalresource dependence and poor recent economic performance in certain developingcountries—the resource curse. The term natural resources can include agricultural andmineral assets, as well as resources associated with water and forests. Among these,certain types of natural resources such as oil and minerals have a tendency to lead toproduction and revenue patterns that are concentrated, while revenue flows from othertypes of resources such as agriculture are more diffused throughout the economy.
Following the classification in Auty (1997), countries rich in the former category may becalled point-sourced economies, while nations abundant in the latter type may bereferred to as diffuse. Sometimes, agricultural commodities such as coffee/cocoa arealso considered point-sourced, if they are produced or marketed in a manner akin tothe concentrated conditions prevalent with minerals. Isham et al. (2005) find thatpoint-sourced economies identified as exporters of oil, mineral, and plantation-basedcrops have lower growth rates compared to other diffuse and manufactured exportersbetween 1975 and 1997 because of the poorer governance engendered by a fuel,mineral, or plantation-dependent economy. The challenge is to extend the pure cross-sectional econometric analysis so that it has a time dimension, and this is what we do inthis paper.
* Mavrotas (corresponding author): Chief Economist, Global Development Network, 2nd Floor, West Wing,ISID Complex, Plot No. 4, Vasant Kunj Institutional Area, New Delhi 110 070, India. Tel: (+91)-11-26139494/2613 6885; Fax: (+91)-11-2613 6893; E-mail: [email protected]. Murshed: Institute of Social Studies(ISS), PO Box 29776, 2502 LT The Hague, Netherlands; Birmingham Business School, University of Bir-mingham, Edgbaston, Birmingham B15 2TT, UK. E-mail: [email protected]. Torres: Departamento deEconomía, Universidad Católica del Uruguay, Montevideo, Uruguay. E-mail: [email protected]. MansoobMurshed is grateful to the International Institute for Environment and Development (IIED) in London forfinancial support. The authors thank the Editor and an anonymous referee for their insightful comments andsuggestions; Nicky Pouw and Admasu Shiferaw for excellent research assistance; as well as Jörg Mayer forcomprehensive comments on previous drafts. Any remaining errors or omissions remain the responsibility ofthe authors.
RESOURCE DEPENDENCE AND ECONOMIC PERFORMANCE With regard to the channels between a rich point-sourced resource dependence and growth failure two separate avenues can be identified. The first is to do with “Dutchdisease” and the fact that a temporary but substantial increase in the price of minerals(or remittances, as in Vargas-Silva, 2009) can cause resource misallocation via themechanism of relative prices; see the survey in Murshed (2004). There is a switch frominternationally traded goods to nontraded goods. Collier and Goderis (2007), forexample, find that commodity booms have a positive short-term effect on output, butadverse long-term effects.
The second mechanism is to do with the political economy of resource rents.
Increases in the availability of resource rents following a boom in their world prices canincrease the appetite for resource rents amongst certain individuals or groups withinsociety. This is known as a voracity effect (Lane and Tornell, 1996), or the rentier effect(Ross, 2001), and it can instigate serious diversion from normal productive activities.
Competition over lootable natural resource rents can sometimes descend in to outrightcivil war (Murshed, 2002). A wealth of mineral resources or plantation-based produc-tion can spawn extractive and nondevelopmental institutions that eventually becomeentrenched (Sokoloff and Engerman, 2000).
Acemoglu et al. (2005) argue that good economic institutions (property rights and the rule of law) are required for growth in the longer term. They assert that these aremore likely to materialize in the context of settings that constrain the arbitrary exerciseof political power, where there is a broad-based interest in property rights and if thereare fewer rents that can be appropriated by a small political elite. Collier and Hoeffler(2009) unpack democracy into (i) electoral competition and (ii) checks and balances,and examine their interaction with natural resource rents in determining GDP growth.
The blend between resource rents and strong electoral competition is growth reducing;while the mix of resource rents and strong checks and balances yield growth-enhancingoutcomes. Granted, good institutions can be the key to future economic progress. Thisraises the question as to what actually determines good institutions. The Acemogluet al. (2005) thesis is that it is whether or not Europeans settled a colony. Our, alter-native, explanation is that a nation’s endowments or economic dependence are thatwhich is important in determining future institutional quality.
The rest of the paper is organized as follows: section 2 constructs a theoretical model of macroeconomic growth collapse embedded within micro-foundations of rent-seeking behavior and contests, linked to the institutional environment. The majortheoretical innovation is the incorporation of increasing returns to scale in rent seeking.
Section 3 is concerned with empirical findings on the link between natural resourcedependence and growth, where the type of resource dependence exerts a positive ornegative impact on growth via institutional quality (the Fraser index of governance) orthe quality of democracy (the Polity index). The second innovation in our paper lies inthe fact that our empirical investigation is based on panel data models, unlike the purelycross-sectional regressions of most of this literature. The third innovation in our paperis that we employ a variety of econometric techniques to establish the robustness of ourresults. Finally, section 4 concludes.
2. A Dynamic Model of Growth Collapse Combined with Rent Seeking
We begin with a competitive game of rent seeking in the spirit of Tullock (1967).
Several agents compete for rents and the competition to capture this entails a cost, bethat bribery or lobbying expenditure. Let P represent the winner-take-all prize ofresource rents that each agent is attempting to seize. Each agent’s probability of success George Mavrotas, Syed Mansoob Murshed, and Sebastian Torres will depend on their own rent-seeking expenditure relative to others. The expectedutility (E) of an agent (i) in a symmetrical setting can take the form: where p is the probability of winning based upon the contest success function,and c represents lobbying costs or expenditures. The contest success function isgiven by: In this example above there are only two agents, i = 1, 2. The crucial parameter s represents the “efficiency or productivity” of lobbying expenditure or bribery; if s > 1,there are increasing returns to scale in such expenditure. If that is so, under weakinstitutions of governance, lobbying expenditure is even more productive as far as rentseekers are concerned. In many ways, s can be characterized to be negatively related togood governance and institutional quality, with s > 1 being a sign of very poor institu-tional environment. Our analysis has some similarities to the grabber-friendlyinstitutions described in Mehlum et al. (2006), but we incorporate increasing returns toscale to that activity, and the possibility of an attrition game. It is not only the totalavailable prize (P) that determines rent seeking, but also the institutional environmentparameterized by s in our model.
Substituting (2) into (1) and maximizing with respect to ci we find that: Equation (3) gives us the Cournot–Nash equilibrium level of lobbying spending byeach agent. The substitution of (3) into (1) yields the following expected utility: The above expression becomes negative if s > 2. If this is so, it will lead to an even moresocially wasteful war of attrition game, where the object is to make one’s opponentsexit the rent-seeking game because an opponent’s very presence yields negativeexpected utility. Lobbying or rent-seeking expenditure is wasteful and detracts fromthe capital stock. Total lobbying expenditures may cause a decline in the capital stock,as investment in capital declines.At this juncture we introduce two definitions which weintend to utilize in the macro model of growth collapse: We now turn to the macro model, akin to the Ramsey growth model (see Murshed, 2004, for details). All variables are given in per capita values. The equilibrium level ofthe capital stock in the steady state implicitly defines growth rates. An “infinitely”-livedindividual maximizes utility at each period (t): where utility (U or u) depends on consumption, C; exp is the exponential operator; andthe real interest rate is i. Maximization is subject to two budget constraints at t: RESOURCE DEPENDENCE AND ECONOMIC PERFORMANCE D(t) = C(t) + I(t) + iD(t) − f (k(t)), k(t) = I(t).
We ignore the rate of depreciation. Output subject to constant returns, Y is given by: The stock constraint (8) tells us that the rate of capital accumulation at time t, is equalto investment (I ) at time t. Equation (9) is the production function for Y (output)written in per capita fashion, k is the capital/labor ratio. Equation (7) is the flowconstraint in an open economy. It informs us that the rate of accumulation of interna-tional debt is given by the excess of consumption (C), investment (I ), and debt servicing(iD, where i is the interest rate and D is debt stock), over output ( f(k)).
The current value Hamiltonian (H) is: ( (t)) − μ(t) C(t) + I(t)⎨1+ g + iD(t) − f (k Here pk is the shadow price of capital. The function inside g(·) gives us the cost ofinstalling capital per unit of existing capital, g′ > 0. The two costate variables are: m(t)and m(t)pk.
Maximization yields the following, among other, first-order conditions: δH(t) = u′ C ( (t) ) −μ(t) = 0, Equation (11) tells us that the optimizing agent will equate the marginal utility ofconsumption, u′(C(t)) to the shadow price of consumption, m (the Ramsey rule). Theconsumption and investment decisions are separable. Equation (12) informs us that theratio of investment to the existing stock of capital is equated to the shadow price ofcapital ( pk). We may write the ratio of investment to capital as a function of the shadowprice of capital ( pk).This allows us a steady-state differential equation in k, with g(·) = 0: ) = k(t)ϕ (p ( )); ϕ′ > , indicating that investment is an increasing function of the shadow price of capital (pk).
The picture is incomplete unless we postulate an equation determining pk: Equations (13) and (14) can be utilized to describe the dynamics and steady-state equilibrium of the system. A resource boom is incorporated by an additive and multi-plicative effect to the production function in (9): − z0) f (k(t)) + z1.
Here z0 represents the diversion of a part of the capital stock from ordinary productionto rent-seeking activities, and z1 is the revenue component. They are described by George Mavrotas, Syed Mansoob Murshed, and Sebastian Torres equation (5). The additive component, z1 has no effect on the marginal product ofcapital, and therefore no effect on investment and the capital stock, but only onconsumption. Note that the resource rents (z1) are exogenous in the sense that they arelike a pure transfer or manna from heaven. The costs to the economy (the multiplica-tive term z0) are, however, an endogenous outcome of rent-seeking activities describedabove in equations (1) to (4). We postulate that rent seeking will reduce the effectivemarginal product of capital, due to the diversion of productive investment away fromnormal activities towards rent seeking.
Following Murshed (2004), the events related to investment and capital accumula- tion are depicted in Figure 1 in k- and pk-space. In Figure 1, the k = 0 schedule is ahorizontal line; the optimal capital stock is related to marginal productivity and not pk.
The p = 0 line is negatively sloped as a rise in pk increases the rate of investment, which in turn raises the capital stock (k). However, with a fixed marginal productivity ofcapital, the capital stock is given at its optimal level, k*, such that pk = 1 in the steadystate; hence k will decline.
When there is a negative effect on the capital stock, the economy jumps from the initial equilibrium at E1 to the new saddle-path (SS2) at point F. The p = will then shift to the left, and the final equilibrium is at E2. There is an initial, but notsteady-state, fall in the shadow price of capital. This makes the rate of investmentnegative between F and E2, which, in turn, causes the capital stock to decline, prompt-ing negative growth. The economy comes to a rest with a lower steady-state capitalstock and growth rate at E2. The growth collapse occurs between F and E2. Netoutput declines in the new steady state due to the combined effect of the diversionof output to rent seeking and a lower capital stock. Rent-seeking expenditure per seis greater the higher the prize, P in (1) to (4), or z1 in (15). This implies alarge amount of resource rents. But, more importantly, total rent-seeking expenditurefor any level of natural resource rents will be greater when the polity is morepredatory, oligarchic, and poor institutions abound. This means that z0 is large in (5)and s > 1.
RESOURCE DEPENDENCE AND ECONOMIC PERFORMANCE 3. Empirical Analysis
We investigate the link between natural resources, institutional development, andeconomic growth in 56 developing countries (excluding industrialized nations andformer socialist economies) over the period 1970–2000. For each country the first twomajor export items were identified based on UNCTAD (2002) and is utilized as anindicator of resource dependence in each nation. The export commodities were thengrouped as originating from one of the following four sources: (i) point-source naturalresources; (ii) diffuse-source natural resources; (iii) coffee/cocoa; or (iv) manufacturing.
Our definition of “point” includes both proper point-source commodities such asminerals, as well as coffee and cocoa. This is because various authors including Ishamet al. (2005) have argued in favor of aggregating coffee/cocoa economies along withpoint-source goods due to similar conditions of production and distribution. The dataspan the 1970–2000 period over which the structure of exports is allowed to vary, unlikein previous studies where export structure is held fixed. A case can be made for theGDP share of natural resources, as in Sachs and Warner (1999); however, a nation’sprincipal exports are both indicative of its competitive and comparative advantage,which has to be related to its endowments over time. Brunnschweiler and Bulte (2008)differentiate between resource dependence and resource abundance. Our measure iscloser to a definition of resource dependence.
Many datasets on democracy exist. Perhaps the best known is the Polity dataset that gives a democracy score of between 0 (lowest) and 10 (highest), compiled by Gurr andJaggers (1996). A similar autocracy dataset gives an autocracy score of between -10 and0, with -10 being the worst score. The Polity 2 score is a combination of both autocracyand democracy, and a reflection, both, of a country’s democratic and nondemocraticcredentials. Table 1 (descriptive statistics) indicates that point-sourced economies haveworse democracy scores and lower growth rates than diffuse economies, and diffuseeconomies perform more poorly on these two counts compared to manufacturingexporters.
We use a second data source for institutions, the Fraser chain-linked index: see Gwartney and Lawson (2005) for details. This dataset pertains more to governance,and is perhaps closer to what Acemoglu et al. (2005) have in mind when they referto “economic” institutions. It covers the following: (a) the size of government expen-diture and tax structure, (b) the legal structure and property rights, (c) access tosound money, (d) the freedom to trade internationally (the trade policy stance), and(e) the regulatory structure. The maximum score in this index is 10. Besides beingcomprehensive in its coverage of economic governance, the greatest advantage of thisdataset is that these data are available going back to 1970, in five-yearly intervals.
This allows its application to panel data estimates. Human capital is measured fol-lowing the Barro and Lee dataset as an estimate of educational attainment (averageschooling years) for the population over age 15. For macroeconomic variables likeper capita GDP, investment, and real effective exchange rates, the main source wasthe World Bank (2002).
Consider the following model that consists of equations for George Mavrotas, Syed Mansoob Murshed, and Sebastian Torres Note: Variable sample means and standard deviations (in parentheses).
The econometric model is motivated by the theory outlined in section 2. Our econometric methodology addresses both the indirect effect of resource dependenceon growth, as well as the potential endogeneity problem that runs between institu-tions and growth, because not only do institutions promote growth, but increasedprosperity also affects institutional quality. One issue relates to the fact that theimpact of resource dependence on economic prosperity is indirect, working throughthe institutional quality channel. Ultimately the political process determines institu-tional quality, and the Polity 2 index could be a proxy for the political conditions thatreproduce well-functioning economic institutions. Alternatively, we could have directmeasures of governance affected by the resource dependence type, such as with theFraser index. We try both, because good governance is also associated with certaintypes of autocracies. The second point is related to the Lipset (1960) hypothesis,which argues that democratic institutions automatically emerge when countriesbecome more affluent, because the body politic becomes more educated and con-scious of their rights. Thus, there is a problem of reverse causality betweeninstitutions and growth. We therefore need to estimate institutions separately in (16)as a function of resource-dependence type, and use predicted values for institutionsin the growth equation, (17).
The first equation, (16), is a reduced form of the rent-seeking effect where Institutionsit represent either the polity or governance score for country i at time t as a proxy for RESOURCE DEPENDENCE AND ECONOMIC PERFORMANCE rent-seeking activities. This is in line with equations (3) and (5) above, where rent-seeking expenditure is a function of total available rents (P, z0, z1) and the institutionalenvironment (s).This in turn depends on the type of resource dependence, where we canexpect a priori that point-sourced economies fare worst.Thus, we are hypothesizing thatinstitutions are determined by resource dependence and not geography, culture, orcolonial heritage. Our proxies for a country’s resource dependence are its two principalexports. Pointit and Diffuseit first reflect actual percentages of a nation’s principal exports;we also later utilize dummy variables coded 1, if a country’s major export in a particularyear is point-source and 0 otherwise in (16); similarly for the variable diffuse. The latterprocedure is likely to magnify any negative effect of point-source or diffuse exportdependency on institutions. The point and diffuse variables also act as instruments thatcorrect for any endogeneity between growth and institutions. The model has two errorcomponents: ui which is unobserved country specific and time invariant, and eit which isnormally distributed with zero mean and constant variance.
Equation (17) is the reduced form of the growth collapse theoretical equation (15) with proxies for the reduction in the marginal product of capital via already estimatedinstitutions in the first equation and other standard features of growth regressions thatinclude human and physical capital (measured by gross fixed capital formation) as wellas the effects of resource booms via terms of trade and real exchange rate movements.
Zi is a vector of these growth regressors, also utilized in (16), including lagged invest-ment as a ratio to GDP, initial stocks of human capital (this is justified by the fact thathuman capital levels take longer gestation periods before yielding returns), indicatorsof macroeconomic instability, and a control for initial levels of per capita GDP. Mac-roeconomic instability is proxied by standard deviations of the terms of trade and realexchange rate on a three-year moving average basis. These are important, as thesemacroeconomic indicators point to Dutch disease effects; that is the effect of relativeprices and exchange rates following resource booms on growth, in contrast to politicaleconomy effects. By including them we are controlling for the other (nonpoliticaleconomy) channel through which resource booms could adversely affect growth. Ourmodel, therefore, has more control variables than is usual in this literature. Equation(17) also has two error terms; a country-specific term (hi) and a normally distributeddisturbance (vit).
Unlike previous studies (Isham et al., 2005, for example) our estimates are over the 1970–2000 period for a five-year average panel of 56 developing countries wherechanges in the pattern of exports and comparative advantage are therefore taken intoaccount. In order to test the robustness of our empirical specifications, and in additionto using two new variables to measure Institutions, we depart from previous studies byusing not one, but several different econometric techniques, some of which take care ofunobserved country fixed effects.
First, in order to test the direction of causality in the institutions–growth correlation, we follow the Acemoglu et al. (2005) strategy and estimate both an instrumentalvariables pooled regression (Table 2) and an instrumental variables panel regression(Table 3), where our institutional measure is instrumented with our point and diffuseindices. The Sargan and Basmann tests for overidentifying restrictions suggest that ourinstruments are valid (Table 2).
Several alternatives were then available. The random effects model has a conceptual appeal in large datasets since it provides a more general specification for the sources oferrors and uses up fewer degrees of freedom. However, a crucial assumption in therandom effects model is that the two error components (time and nontime invariant)are homoskedastic and serially uncorrelated. To address this issue and obtain efficient George Mavrotas, Syed Mansoob Murshed, and Sebastian Torres Table 2. Instrumental Variables Pooled Regressions of GDP Per Capita Growth Rate Using Pointand Diffuse as Instruments Panel A: Instrumental Variables Two-Stage Least Squares Regression
Institutions
Panel B: First-Stage Regression for Institutions
Point
Notes: Panel A reports the two-stage least squares estimates instrumenting for the Fraser chain-link indexand for Polity 2 using point and diffuse; Panel B reports the corresponding first stage.
***, **, * Estimate is significant at the 1%, 5%, 10% level, respectively. Standard errors are in parentheses.
p-Values reported for the Sargan and Basmann overidentification tests.
estimates we deployed the feasible generalized least squares (FGLS) method andproceed first by correcting for autocorrelation and then for heteroskedasticity in Tables4 and 5, respectively. Finally, we also utilized the instrumental variables generalizedmethod of moments (GMM) estimator to take into account the possible autocorrela-tion between the dependent variable and our regressors (Table 6). We first use actualpoint and diffuse exports as a share of the first two exports for our definition of point RESOURCE DEPENDENCE AND ECONOMIC PERFORMANCE Table 3. Instrumental Variables Panel Regressions of GDP Per Capita Growth Rate Using Pointand Diffuse as Instruments Panel A: Instrumental Variables Two-Stage Least Squares Regression
Institutions
Panel B: First-Stage Regression for Institutions
Point
Notes: Panel A reports the two-stage least squares estimates instrumenting for the Fraser chain-link indexand for Polity 2 using point and diffuse; Panel B reports the corresponding first stage.
***, **, * Estimate is significant at the 1%, 5%, 10% level, respectively. Standard errors are in parentheses.
and diffuse in (16) for the regressions reported in the first two columns of Tables 2–6,followed by regressions using dummy variables for point (0, 1) and diffuse (0, 1) in thelast two columns of Tables 2–6.
First, in all but one of the 20 cases of different estimation pro- cedures, two types of data definitions of point and diffuse export dependence, andtwo definitions of institutions, reported in Tables 2–6, the impact of the institutionalvariable upon growth is positive and significant. The exception is for Polity 2 (or George Mavrotas, Syed Mansoob Murshed, and Sebastian Torres Table 4. Feasible Generalized Least Squares Panel Data Regressions of GDP Per Capita Growth,Using Point and Diffuse to Predict Institutions Panel A: Dependent Variable—GDP Per Capita Growth Rate (Equation (17))
Predicted institutions
Panel B: Dependent Variable—Institutions (Equation (16))
Point
Notes: Panel A reports the two-stage least squares estimates instrumenting for the Fraser chain-link indexand for Polity 2 using point and diffuse; Panel B reports the corresponding first stage.
***, **, * Estimate is significant at the 1%, 5%, 10% level, respectively. Standard errors are in parentheses.
democracy) in the instrumental variables panel regression reported in Table 3.
Second, governance seems to be more important for growth, as it is alwayssignificant, with a larger coefficient compared to democracy. Third, both types ofresource dependence, point and diffuse, exert a negative influence on growth viainstitutions when compared to manufactured goods exporting economies. Resourcedependence of any kind appears to be bad for institutional development in devel-oping countries.
RESOURCE DEPENDENCE AND ECONOMIC PERFORMANCE Table 5. Feasible Generalized Least Squares with Heteroskedastic Error Structure Panel DataRegressions of GDP Per Capita Growth, Using Point and Diffuse to Predict Institutions Panel A: Dependent Variable—GDP Per Capita Growth Rate (Equation (17))
Predicted institutions
Panel B: Dependent Variable—Institutions (Equation (16))
Point
Notes: Panel A reports the two-stage least squares estimates instrumenting for the Fraser chain-link indexand for Polity 2 using point and diffuse; Panel B reports the corresponding first stage.
***, **, * Estimate is significant at the 1%, 5%, 10% level, respectively. Standard errors are in parentheses.
Fourth, and most importantly, it is not always the case that a point sourced export structure is worse than diffuse resource dependence as far as institutionaldevelopment is concerned (Tables 3 and 5). It depends upon which institution we arelooking at. Here we must again reiterate the differences between governance, as mea-sured by the Fraser index, and democracy captured by Polity 2. The first is a pictureof how well a country is run, and the latter an indication of the right to chooseleaders and executive constraints. Often democracies, especially weak and embryonic George Mavrotas, Syed Mansoob Murshed, and Sebastian Torres Table 6. Instrumental Variables Generalized Method of Moments Estimation of GDP Per CapitaGrowth Rate Using Point and Diffuse as Instruments Panel A: Instrumental Variables Two-Stage Least Squares Regression
Institutions
Panel B: First-Stage Regression for Institutions
Point
Notes: Panel A reports the two-stage least squares estimates instrumenting for the Fraser chain-link indexand for Polity 2 using point and diffuse; Panel B reports the corresponding first stage.
***, **, * Estimate is significant at the 1%, 5%, 10% level, respectively. Standard errors are in parentheses.
p-Values reported for the test of excluded instruments and for the underidentification tests (Anderson andCragg–Donald).
democracies, are badly governed. The point is that a weak democracy (those withPolity scores consistently below 8) may also be poorly governed. Our finding isthat good governance is more important for growth when compared to democracy.
This is explained by the fact that many of the democracies in our sample are rather RESOURCE DEPENDENCE AND ECONOMIC PERFORMANCE imperfect and new democracies that may be poorly governed. Note that bothtypes of resource dependency retard both categories of institutional development.
With regard to governance, point resources are worse than diffuse, except very mar-ginally in two cases (Tables 3 and 5). As far as democracy is concerned, a diffuseexport pattern appears to retard democratic development by more in only twocases (Table 3).
Finally, the magnitudes of the macroeconomic control variables are either small or insignificant, suggesting that the political economy channels dominate traditionalDutch disease effects that manifest themselves in terms of trade and exchange ratevolatility. Investment and the level of initial human capital are significant in some cases,but in other instances are insignificant with the wrong signs, particularly when associ-ated with Polity 2. This serves to underscore the explanatory power of institutions, as faras growth is concerned.
4. Conclusions
Our theoretical model encapsulates macroeconomic growth collapse as a consequenceof resource dissipation resulting from a wasteful rent-seeking contest, where the objectof the contest is the capture of resource rents. The novelty here is that the institutionalenvironment can make rent-seeking outlays subject to increasing returns to scale, andeven more profitable. Our econometric analysis also contains major innovations. First,it has theoretical foundations. Second, it is one of the few econometric investigationsinto the resource curse that includes analysis over time, as it is a panel data estimation.
Our proxies for institutional quality are the degree of democracy and the quality ofgovernance, thus we have attempted to capture both the effects of how a country isruled as well as its polity; governance seems to have a greater impact on development.
Finally, our results are more robust than in other cases in the literature, as a variety ofeconometric models have been used to test the same specification. Our general resultssuggest that both point-source and diffuse-type natural resource dependence retard thedevelopment of democracy and good governance, which in turn hampers economicgrowth. So there is a more widespread resource curse, valid for both types of depen-dence. Point-sourced economies have a worse impact on governance, which is moreimportant for growth.
The resource curse in point-sourced economies depends on the quality of institu- tions, and the quality of institutions may be endogenously affected by resourcedependence. Historically, however, rich mineral resource endowments did notprevent economic growth in Australia, Canada, and the USA a century ago, aspointed out by Findlay and Lundahl (1994). Brunnschweiler and Bulte (2008) chal-lenge the notion of resource curse in their cross-sectional analysis. They distinguishbetween resource dependence and resource abundance, finding that resource depen-dence has no significant effect on growth. By contrast, they find that resourceabundance has significantly positive effects on growth. A resource-abundantnation may not be very resource dependent, if it has a diversified production struc-ture. Indeed, resource dependence may be a reflection of the failure to grow anddevelop good economic and political institutions, along with the associated poverty,inequality, and poor human development outcomes. Finally, institutional improve-ment is like administering an antibiotic to a patient curing him of a serious infection;changes in policy are more akin to the ingestion of aspirin that works like a tempo-rary palliative.
George Mavrotas, Syed Mansoob Murshed, and Sebastian Torres References
Acemoglu, D., S. Johnson, and J. A. Robinson, “Institutions as the Fundamental Cause of Long-Run Growth,” in P. Aghion and S. Durlauf (eds), Handbook of Economic Growth,Vol. 1, Amsterdam: Elsevier (2005):385–472.
Auty, R. M., “Natural Resources, the State and Development Strategy,” Journal of International Development 9 (1997):651–63.
Brunnschweiler, C. N. and E. H. Bulte, “The Resource Curse Revisited and Revised: A Tale of Paradoxes and Red Herrings,” Journal of Environmental Economics and Management 55(2008):248–64.
Collier, P. and B. Goderis, “Commodity Prices, Growth, and the Natural Resource Curse: Rec- onciling a Conundrum,” CSAE working paper 2007-15 (2007).
Collier, P. and A. Hoeffler, “Testing the Neo-Con Agenda: Democracy in Resource-Rich Soci- eties,” European Economic Review 53 (2009):293–308.
Findlay, R. and M. Lundahl, “Natural Resources, ‘Vent-for-Surplus’ and the Staples Theory,” in G. Meir (ed.), From Classical Economics to Development Economics, New York: St. Martin’sPress (1994):68–93.
Gurr, T. R. and K. Jaggers, “POLITY III: Regime Change and Political Authority, 1800–1994,” computer file, Study #6695, Ann Arbor, MI: Inter-University Consortium for Political andSocial Research [distributor] (1996).
Gwartney, J. and R. Lawson, Economic Freedom of the World: 2005 Annual Report, Vancouver: Fraser Institute. Available at www.freetheworld.com (2005).
Isham, J., L. Pritchett, M. Woolcock, and G. Busby, “The Varieties of Resource Experience: Natural Resource Export Structures and the Political Economy of Economic Growth,” WorldBank Economic Review 19 (2005):141–74.
Lane, P. and A. Tornell, “Power, Growth and the Voracity Effect,” Journal of Economic Growth Lipset, S., Political Man: The Social Bases of Politics, New York: Doubleday (1960).
Mehlum, H., K. Moene, and R. Torvik, “Institutions and the Resource Curse,” Economic Journal Murshed, S. M., “Civil War, Conflict and Underdevelopment,” Journal of Peace Research 39 ———, “When Does Natural Resource Abundance Lead to a Resource Curse,” IIED-EEP working paper 04-01. Available at www.iied.org (2004).
Ross, M., “Does Oil Hinder Democracy?” World Politics 53 (2001):325–61.
Sachs, J. and A. Warner, “The Big Push, Natural Resource Booms and Growth,” Journal of Development Economics 59 (1999):43–76.
Sokoloff, K. L. and S. L. Engerman, “Institutions, Factor Endowments, and Paths of Development in the New World,” Journal of Economic Perspectives 14 (2000):217–32.
Tullock, G., “The Welfare Costs of Tariffs, Monopolies and Theft,” Western Economic Journal 5 UNCTAD, Handbook of Statistics, online database, Geneva: UNCTAD (2002).
Vargas-Silva, C., “The Tale of Three Amigos: Remittances, Exchange Rates, and Money Demand in Mexico,” Review of Development Economics 13 (2009):1–14.
World Bank, World Development Indicators 2002, CD-ROM database, Washington, DC: World

Source: http://deptoeconomia.ucu.edu.uy/attachments/143_Mavrotas-et-al%20-2011.pdf