Resource rent and economic growth: A Panel Threshold Analysis
Abstract
The effects of Institution’s quality on the nexus between resource rent and economic growth is the most debated and addressed issue within the contemporary economics literature. Apart from the ban or boom effect of resource rent for economic growth, this study uses Rule-of-law as a threshold variable and revisits the nonlinear relation between the natural resource and economic growth under the resource curse hypothesis. The study use Panel data of 14 resources abundant Sub-Saharan Africa (SSA) countries over the period of 1998-2016. A panel threshold regression (PTM) model is employed to identify the threshold level of rule-of-law above which the effect of resource rent on economic growth changes. The robustness effect of the resource rent with in the thresholds regimes are estimated using system GMM. The study finds the existence of two statistically significant thresholds that indicates significant positive relationship between resource rent and economic growth when the resource rent is above the threshold level of -1.28 and -1.37 of rule-of-law. Below the threshold value, the resource curse (RC) start to operate as the resource rent starts to impede economic growth due to inferior values of rule-of-law. Hence, the lesson drawn from the study is that for a nation desiring to benefit fully from its natural resources should not neglect the important role of good institutions. With good institutions, the RC puzzle can also be challenged. Therefore, policymakers of the resource-abundant SSA has to give due attention to the institution’s quality in general and to the rule-of-law in specific to curtail the impact of RC.
Keywords: Resource rent, economic growth, Panel Threshold Regression model, Rule-of-law and Resource reach SSA countries
- Introduction
For decades, the role of natural resources in economic growth has become the topic of debate in the economics literature. Arguments suggest economies that specialize in natural resource extraction and export experience comparatively low rates of economic growth (Sachs and Warner, 1995; Sachs and Warner, 2001; Haggard and Tiede, 2011; Badeeb et al., 2017), though intuitively, resource abundance increase wealth and purchasing power (Holden, 2013). Such findings brought dichotomy in the literature of the resource and economic growth nexus. This puzzle is called a natural resource curse (NRC) paradox; a paradox in which countries with an abundance of natural resources experience stagnant or even negative economic growth (Roy et al., 2013; Badeeb et al., 2017).
Several reasons are provided to support the curse associated with the natural resource (Sachs and Warner, 1995; Sachs and Warner, 2001; Saha and Ben Ali, 2017). First, natural resource extraction processes consume economic resources that otherwise could be allocated to industries that are thought to promote long-term economic growth such as manufacturing or services. Second, global prices for natural resources can be volatile and thus destabilize exporting economies. Third, the concentrated nature of many natural resource supplies facilitates rent capture by non-inclusive governments whose institutions support the retention of political power rather than facilitating economic growth. The resource curse also occurs as a country begins to focus all of its energies on a single industry, such as mining, and neglect other major productive sectors (Gylfason, 2001, 2006; James, 2015). As a result, the nation becomes over-dependent on the price of commodities, and the overall gross domestic product becomes extremely volatile. Additionally, government corruption often results when proper resource rights, rule of law and an income distribution framework is not established in the society, resulting in unfair regulation of the industry (Boyce and Herbert Emery, 2011; Roy et al., 2013).
Contrary to this, studies which challenge the resource curse are reemerging with empirical evidence on the natural resource as a means for economic growth and prosperity (North and Thomas, 1973; Brunnschweiler, 2008; Brunnschweiler and Bulte, 2008; Alexeev and Conrad, 2011). Prior to the twentieth century, natural resources, usually comprising primary commodities played a pivotal role in world trade. Notable countries such as Australia, United States of America and Canada greatly benefited from primary commodity exports in the early stages of their economic development. Lately, Ecuador has also experienced a significant increase in income after the resource boom and Norway has made good use of their natural wealth and turned it into economic prosperity for all next-generation (North and Thomas, 1973; Auty and Mikesell, 1998; Papyrakis and Gerlagh, 2003; Papyrakis, 2016).
The other spontaneous reason to doubt the existence of the resource curse and the obstacle to make meaningful comparison between the developed and less developed countries is the apparent persistence of institutional quality between countries and regions (Alexeev and Conrad, 2011; Bennett et al., 2017). That is, the qualities of institutions of sub-Saharan countries and other less developed countries are inferior to the developed regions. Unfortunately, resource-rich countries in Sub-Saharan Africa tend to have weaker institutions than non-resource reach Sub-Saharan African countries (Gueye and Lee, 2015). Economies with abundant natural resources and better institutional quality and governance such as strong democratic accountability, high rule of law and order, lower corruption, and higher integration among government institutions are evident to have better economic growth and higher human welfare (Damania and Bulte, 2003 and Mehlum et al., 2006).
Both cross-country and dynamic panel regressions confirm the critical role of institutional quality (measured by either institutional design or institutional performance) in turning natural resources into an economic boon or ban (Bakwena et al., 2009). Therefore, one might expect, that if the curse of natural resources does exist the clues may be found and pronounced in a state’s baseline levels of institutional development, economic structure, and political status (Ayelazuno, 2014; Arvanitis and Weigert, 2017) of the resource-abundant countries (Gylfason and Zoega, 2002; Bennett et al., 2017). Study also identifies improving the rule-of-law can cure a remarkably wide array of ailments in developing and post-communist countries, from corruption and surging crime to lagging foreign investment and growth. Despite this, policy makers and promoters are surprisingly short of understanding and knowledge about it. The twin rationale for improving the rule-of-law is that it promotes economic development and democratization (Carothers, 2003). Thus, the importance of rule of law is a key to enable resource-rich nations to have control over their destiny by shaping policies that are likely to be most effective in promoting economic growth (Bakwena et al., 2009; Karabegović, 2009). Contrary to these, when economies of a country or group of countries are characterized by weak rule of law, then the resource is not growth-enhancing rather than a source of NRC (Antonakakis et al., 2017).
An important question in development and economic growth studies is how natural resources richness affects long-term economic growth. The jury is yet open with approximately 40% of empirical papers finding a negative effect, 40% finding no effect and 20% finding a positive effect (Havranek et al., 2016). Our study leaps forward from the ban and boom effects of resource rent and attempts to suggest a threshold level of the rule of law at which the natural resource rent affect economic growth differently on the resource-abundant SSA countries.
The study designs are especially effective in explaining the effects regarding the endogenity effects of explaining variables. Hence, to find out the threshold levels the panel threshold regression econometric techniques due to (Hansen, 1999) and (Gonzalo and Pitarakis, 2002) methods are used. The question such as is there any thresholds in the relationship between resource abundance and economic growth of SSA countries? Is natural resource is a source of a curse at all possible threshold level of INSTQ, and is there any threshold level of which resource abundance to act differently from the RC hypothesis? are answered.
In this way, substantial contributions are expected to be made to the literature about the examination of economic growth and natural resource effects in light of the resource curse of the resource abundant SSA economies. The study also contributes a threshold level and different regimes through examining the non-linear relationship between economic growth and natural resource rent using rule-of-law (INSTQ) as a threshold variable.
The remainder of the paper is organized as follows. In Section I, we briefly discuss previous empirical studies on resource curse and on economic growth. In Section III we describe our empirical specification, our procedures for estimation and inference, and the dataset that we use. In Section IV, we present our results. Finally, in Section V we present the main conclusions of our study and discuss future extensions of this line of research.
- Literature review
The relation between the resource curse and the abundant resource of the sub-Saharan African countries may continue to invite mixed hope and threat for economic growth of the countries. As evidences show some resource-endowed countries of the world are victories from the resource curse contest (Gylfason, 2004; Stijns, 2005) while others though maintained economic growth for short period of time, fail to improve the living standards and bring meaningful economic leap (Ouoba, 2016).
Managing a nation’s extractive natural resource endowments can advance national development if done meaningfully. Unfortunately, across Africa, the apparent mismanagement of such resources, poor growth rates, social tensions, and civil strife in resource-rich countries have thrown up a great deal of literature on what is now known as resource curse. Particularly, rule-of-law forms the basis of the socio-economic development, especially where the presence of the factors such as corruption and freedom of expression, an inferior institutional structure has a significant impact on economic growth (Ozpolat et al., 2016). There are empirical studies that state institutional efficiency boosts economic growth in developed countries, whereas it doesn’t have an impact or has a negative impact on economic growth of developing countries (Haggard and Tiede, 2011; Ozpolat et al., 2016).
Isham et al. (2003) and Frankel (2010) either argue the reason for inconsistency in the empirical findings by previous researchers could be due to the different type of resources; a point or diffuse, and different economic backgrounds in the level of human capital, level of debt overhang, and export diversification. Brunnschweiler (2008), postulated that the inconsistencies in the empirical finding to originate from the inappropriateness of resource abundance measurement to proxy natural resources in the empirical estimation. The other reason to doubt Sachs and Warner’s (S&W) findings of the resource curse is that their study drew on data for 1970–1989 when the global incidence of resource curse effects peaked (Auty, 2017).
To answer the inconsistency in the RC analysis since the turn of the new millennium, the global economic growth has faced a shift in patterns and dynamics of economic growth. The inclusion of recent statistics increases the weighting of the post-1997 years of economic recovery relative to those impacted by the 1973–1985 growth collapses (Auty, 2017). In view of the fact that earlier studies fail to account for such divergent growth experiences despite similar resource type and abundance, more plausible explanations have recently been proposed in the literature of e.g. (Mehlum et al. , 2006b; Arezki and van der Ploeg, 2007; Boschini et al., 2007; Humphreys et al., 2007).
Many of the empirical studies that have been made on the impact of rule-of-law and institutions on resource and economic growth brought different interesting findings that compromise the RC hypothesis. Such analysis and findings on the issues of RC hypothesis show how the issues are far from conclusive. For instance, Auty (2017) identified the importance of Institutions in nullifying the curse through avoiding of rent-seeking behavior, reducing corruption and improve rule-of-law (Ishan et al. 2005 and Robinson et al. 2006), lowering the risk of violent civil conflict (Collier and Hoeffler, 2005), and accelerating efficient resource allocation (Atkinson and Hamilton, 2003). The rule-of-law comprises but not limited to general lawlessness such as guerrilla warfare, revolution, frequent constitutional change, government instability or inability to enforce the law and an absence of government accountability (Deacon, 1994).
To see the bigger picture of the Institution’s impact, exemplary comparisons between Norway and Nigeria; the two oil-endowed countries, one with very low institution’s quality and the other with better or good institution’s quality may indicate us the magnitude it may play on economic growth. The economic prosperity of Norway and its emancipation from the RC indicates the proper utilization of the oil revenue highly correlates with the quality of the institutions of the country. Norway has transformed its economy from one of the poorest countries in Europe during the early 1900s into the country of the highest quality of life today (UNDP, 2009). On the other hand, Nigeria is notorious for its mismanagement of resource proceeds as well as general lack of rule-of-law and rampant corrupt tendencies, hence little economic growth was witnessed due to the oil resource (Gylfason and Zoega, 2002; Sala-i-Martin and Subramanian, 2013).
Boschini et al., (2007) indicated the impact of natural resources on economic growth to be non-monotonic in institutional quality. Mehrara (2009) also examined the issue of the existence of the threshold effects in the relationship between oil revenues and output growth in oil-exporting countries, applying panel regressions. The empirical results strongly suggest the existence of a threshold beyond which oil revenues growth exerts a negative effect on output. The results indicate that the threshold of the growth rate of oil revenues above which oil revenues significantly slows growth is around 18–19% for oil-exporting countries. A study also made on NRC hypothesis in the case of oil exporting countries using Panel smooth transitioning regression (PSTR) model identified a threshold of oil dependence where the relationship between economic growth and its determinants could move smoothly from a regime to another. The study offered strong evidence that oil revenues have non-linear impacts on economic growth and that resource curse only exists under the condition of high oil dependence (Seghir and Damette, 2013).
In contrast to the linear estimation, without any allowance for threshold effects would misleadingly imply that an increase in the resource revenues increases the economic growth rate. Failure to account for nonlinearities conceal the resource curse in these countries particularly during extreme resource booms as suggested in previous studies (Mehrara, 2009). For all these reasons, studying the threshold influences of rule-of-law on the resource rent nexus economic growth in resource abundant SSA countries may give better insights for the policymakers and practitioners of the region on the ways to curb the RC form the countries.
- Data, Variables and Empirical Strategy
There are two categories of explanatory variables in the panel data. The threshold variables, i.e., the rule-of-law hereafter referred as Institution’s quality index (INSTQ) are the variable that used to investigate whether or not there is an asymmetric threshold effect of Institution’s quality on the impacts of resource rent on economic growth. The study also included control variables commonly used in the analysis of the economic growth, namely; resource rent, gross primary school enrolment rate, gross fixed capital formation, debt, a term of trade, exchange rate, life expectancy and real per capita GDP growth.
The study used balanced panel data from the World Development Indicator (WDI) and IMF database for 14 SSA countries.[1] Table (annex B) shows the variables that are used in our growth regression with their data source. The descriptive statistics, namely mean, standard deviation, minimum and maximum are also included in table 1.The study used INSTQ as a threshold index. The index runs from -2.5 to 2.5[2]. According to the table, the average value of the INSTQ is -0.916; while the minimum of is -2.13 and the average resource rent is 22.2 and the while the maximum rent is 89.17 for the countries. Furthermore, table (1) reveals asymmetric distributions of the variables. The dataset covers the period from 1998 to 2016.
The INSTQ thresholds enable us to identify the asymmetrical kinds of relationships and the points of the threshold effects between resource abundance and economic growth. The types of impacts resource may have on economic growth are also find within the threshold levels. A country below the threshold level of INSTQ may experience retarded economic growth and there by suffer from the RC (the negative effect) and conversely, above the threshold level country may experience economic growth due to resource rent (the negative effect).
Table 1
Summary statistics
| Variable | Observation | Mean | Std.Dev. | Min | Max |
| GDP Growth | 266 | 4.904549 | 4.522089 | -8.94 | 121.03 |
| INSTQ | 266 | -0.91693 | 0.509513 | -2.13 | 0.154609 |
| LEX | 266 | 54.37166 | 5.11223 | 43.636 | 66.1658 |
| GFDC | 266 | 23.1172 | 21.80144 | -8.11858 | 218.993 |
| DBT | 266 | 59.61893 | .50.01031 | 0 | 284.168 |
| TOT | 266 | 100.4441 | 45.65923 | 11.1647 | 263.257 |
| SCL | 266 | 93.67978 | 23.28864 | 51.0578 | 204.686 |
| EXC | 266 | 629.2342 | 1240.56 | 0.231166 | 7596.91 |
| RR | 266 | 22.1923 | 16.27612 | 2.995603 | 89.16611 |
Panel Threshold Model
To determine the existence of threshold effects between the variables endogenously is different from the traditional approach in which the threshold level is determined exogenously. If the threshold level is chosen arbitrarily or is not determined within an empirical model, it is not possible to derive confidence intervals for the chosen threshold (Hansen, 1999, 2000). Hence, to achieve this economic estimation is made based endogenous sample splitting. Bootstrap methods, which can be used to construct appropriate confidence intervals were used to assess the statistical significance of the threshold effect. We found the methodology instrumental to test the threshold effect of INSTQ on resource abundance and economic growth relationship.
The problem of the likelihood of endogeneity, where explanatory variables might be correlated with the error term is solved by using lagged values of the dependent variables. The administration of dynamic panel data method eliminates the relationship between the lagged values of the dependent variable and error terms, thus improves the reliability of the predictions made and strengthens the consistency of the obtained predictors (Wang et al., 2014; Aydın et al., 2016). Therefore, one lagged value of annual real GDP growth is used as the explanatory variable (Roodman, 2009).
After establishing the threshold values using Hansen’s (1999) panel fixed effect threshold model (hereafter PTM), intercepts coefficient estimation (threshold effect) is made by using system GMM predictors, a predictor which suits the estimation of endogenous variables rule-of-law (hereafter INSTQ) (Arellano and Bover, 1995).
The observed data are from a balanced panel
yit, qit, xit 1≤i≤n, 1 ≤t ≤T. The subscript
iand t describe the individual and the time indexes respectively. The dependent variable
yitis scalar, the threshold variable
qit,is scalar, and the repressor’s
xitis a k vector. The specification of the model is as follow:
Yit= μit+ β1’xitI qit≥γ+ β2’xitI qit<γ+ εit 1
Re-writing equation 1 is;
Yit=μi+ β1’xit+εit qit ≤ γ μi+ β2’xit+εit qit > γ
Another squeezed way of writing equation 1 is also;
xitγ=xitI(qit ≤ γ)xitI(qit > γ)
And
β=β1′ β2”so that
1equals
yit= μ+ β’xit γ+ εit 2
Each regime is characterized by different regression slopes
β1and
β2. To identify them it is required that
β1and
β2,
qit, and
xitare assumed to be time-invariant. The error
εitis assumed to be independent and identically distributed (iid) with mean zero and finite variance
σ2.
The panel specification of the model studies can be summarized in the form of the following:
yit= βiXit+μt+ θi+ εit
3
Where
yitis the real GDP growth rates of the countries,
Xitis a vector of explanatory variables,
μt and θiare the country and time fixed effects, respectively,
εitis a serially uncorrelated measurement error, and the subscripts, t and
irefers to the country and period, respectively.
Chan (1993) and Hansen (1999) recommended estimating
γby LS minimizing the concentrated SSE. Thus estimator of
γbecomes
γ̂=argminS1(γ)
4
Since it is undesirable for a threshold
γ̂to be selected when it sorts too few observations in regimes, we can exclude it by restricting the minimization in (3) to values
γsuch that a minimal percentage of observation lies in each regime. Once
γ̂is obtained the slope estimate become
β̂= β̂ (γ̂), the residual vector is
ε̂*= ε̂*(γ̂)and the residual variance is
σ̂2= 1n(T-1)ε̂*’ε̂*= 1n T-1S1γ̂.
It is important to establish the threshold and test its significance statistically. The null hypothesis of no threshold effects against the alternative hypothesis of threshold effects are given as follow:
H0= β1= β1
H1= β1≠ β1
Under the null hypothesis, the effects of threshold
γare not known, therefore the classical tests such as the Lagrange Multiplier (LM) test does not have the standard distribution. Hence, a bootstrap procedure suggested by Hansen (1999) is used to determine and simulate the asymptotic distribution of the likelihood ratio test. The p-values constructed from the bootstrap procedure are asymptotically valid.
Then after, individual fixed effect removed from the model by removing the individual mean, taking the averages over time of (2) getting
yit*= β*xit*γ+ εit *
After fixed effect transformation, the equation is transformed as:
yit*= βit’xit*(γ)+εit* 5
Under the null hypothesis of no threshold effect, the model is represented by:
yit= β1’xit+εit 6
Alternatively, as in (5) under the fixed effects transformation, the regression parameter
βcan be estimated by OLS, yielding the restricted estimate
β1residuals
ε⃛, and the sum of squared errors
S0= εit*’ε̂it*. Thus, the likelihood ratio test of
H0against the alternative of a threshold is based on
F1= S0-S1(γ̂1)σ̂2 7
Where
S0and
S1are the residual sum of squared errors obtained from equation (1) without and with threshold effects respectively, and
σ̂2is the residual variance of the panel threshold estimation.
Hansen (1999) recommended the implementation of bootstrap to get asymptotically valid tests keeping the regressors and threshold variable fixed in repeated bootstrap samples.
Procedurally, first, we estimate the model by grouping the regression residuals
ε̂i*by an individual, i.e. for each
i, we construct
ε̂i*=(ε̂i1*, ε̂i2*……..ε̂iT*)treating the sample
ε̂i*=(ε̂1*, ε̂2*……..ε̂N*)as the empirical distribution for the bootstrap. Then we draw with replacement a sample of size N from the empirical distribution and use these errors
ε̂*(b), where the superscript indicates
bothbootstrap replication, to create a bootstrap sample under the null hypothesis, i.e.
ŷit*(b)= β̂’xit*+ εit*(b). Using the bootstrap sample
ŷit*(b), we estimate the model under the null, i.e.
ŷit*(b)= β’xit*+ uitand under the alternative of a threshold, i.e.
ŷit*(b)= β’xit*(γ)+ vit. Therefore, we compute the bootstrap value of the likelihood ratio statistic
F10as in (7), repeating this procedure a large number (say 500) of times. We finally calculate the bootstrap p-value as the percentage of draws for which the simulated statistic exceeds the actual value (Hansen, 1999; 2000). If there is a threshold effect (
β1≠ β2), then
γ̂is inconsistent with
γ0. This requires the computation of confidence interval around the estimated threshold values.
Under normality the likelihood ratio test statistics,
LR1=(S1γ-S1γ̂)/σ̂2. When endogenous sample splitting procedure is used LR does not have standard
x2distribution (Chang, 2010). Hence, the best way to form confidence intervals for
γis to form the “no-rejection region” using the likelihood ratio statistics for a test of
γ. A test of
H0: γ= γ0at the asymptotic level
∝if
LR1(γ0)exceeds
c(∝)reject large values
LR1.
F1
in equation 7 tests the hypothesis of no threshold against one threshold, and if
F1rejects the null hypothesis of no threshold, a further step based on the model equation (8) is used to test one and two thresholds. Same procedure followed for the approximation of likelihood ration of one versus two thresholds.
After obtaining the threshold value, the resource rent effect within the threshold level is tested using system GMM model proposed by Arellano and Bover, (1995). The model further uses one lagged value of annual real GDP growth rate as the explanatory variable. This makes the model dynamic. To conduct the estimation of the effect of resource rent using GMM, transformation of PTM in the first step into linear panel data model through multiplying the threshold value (INSTQ) with resource rent variable is required. Following the specification (5) and by adding a lagged value of the dependent variable as one of the regressors we specify the model as follow:
Yit= μit+yit-1+ β1’xit Iqit≤ γ1+ β2’xit Iγ1< qit≤ γ2+ β3’xit I qit> γ2+ φit 8
Where the resource rent containing the thresholds are ordered so that
γ1<γ2.
γ1 and γ2are they the divide the equation into three regimes with coefficients β1; β2, and β3, respectively. The same strategy can be adopted for the model (3) with different regimes.
Before conducting the GMM estimation, we must ascertain whether the orthogonal condition holds. The condition is:
Ey1,t-s1*ei,t*-ei,t-1*=0 s=2,….,t-1, t=3,….,T 9
Ey1,t-sR*ei,t*-ei,t-1*=0 s=2,….,t-1, t=3,….,T 10
It is obvious that as t becomes larger the number of orthogonal conditions increases. Rewriting Eq.10
EWi*’ei,t*-ei,t-1*=0 i=1,….,N 11
Let GMM estimator matrix be
θ̂ρ̂’, β̂’and
θ ̂=XR* ‘W* AN W* ‘WR*-1 XR* ‘W* ANW*Y’ 12
In Eq.(12),
AN=1N ∑i=1NWi* ‘HWi*-1,where H is a (T-2)*(T-2) matrix
- Empirical Result
Based on our primitive assumption, we expect nonlinear relationship impacts of INSTQ on the relationship between resource rent and economic growth. Within the specified threshold level, the impacts of resource rent are expected to be positive and significant, unlike the resource curse hypothesis. Specifically, the model takes the form:
The notation for the multi-threshold regimes of the study’s model (3 & 6) is fixed as follows:
yit=μi+β1yit-1+β2Dbtit-1+ β3Lexit-1+ β4Totit-1+ β5 RRit-1+ β6cfit-1+ β7Sclit-1+β8Excit-1+β9Insit-1 Iqit≤ γ1+ δ2Insit Iγ1< qit≤ γ2+ δitinsitI qit> γ2+ φit 13
Where:
μi= is an individual effect
yit=Real GDP growth annual percentage change at a time t,
yit-1=Real GDP growth annual percentage change at a time t-1,
Dbtit-1=Repayable Debt owed at a time t-1
Lexit-1=number of a newborn infant would live at a time t-1
Totit-1=percentage of the ratio of export price to import price at a time t-1
RRit-1=total resource rent % of GDP at a time t-1
Cfit-1=Fixed assets of the economy at a time t-1
Instit-1=rule of law at a time t-1
Excit-1= exchange rate at a time t-1
Sclit-1=total enrollment in primary education at a time t-1
φit-1= εit is iid across countries and years
Institutions and resource curse
Before estimating the threshold regression, we test the non-linearity relationship and the existence of a threshold effect between resource rent, INSTQ, and economic growth. Then reject the null hypothesis of no threshold effects of the model using the p-value of the bootstrapping as proposed by Hansen (1999). To obtain the approximations of the F statistics and then calculate the p-values, the bootstrap procedure is repeated 500 times for each of the panel threshold tests. Then after, the coefficient’s impacts within the threshold level were tested using system GMM model.
Table2 shows the results of tests for single, double and triple thresholds with the bootstrap P-values. We find that the P-value for the single threshold to be significant with a value of 0.013 and the test statistics for a double threshold is also found to be strongly significant with P-value of 0.006. However, the test statistics for a triple-threshold found to be insignificant with a value of 0.924. Hence, we may ascertain the presence of two thresholds in the regression estimation and also we no longer consider the triple-threshold estimation for the study.
Table 2
Test for threshold effect
| 1 | Test for a single threshold
F1 P-value 10%, 5%, 1% critical values |
18.5
0.013 11.4, 14.3, 20.1 |
| 2 | Test for double threshold
F1 P-value 10%, 5%, 1% critical values |
44.9
0.006 15.4, 23.48, 33.45 |
| 3 | Test for Triple threshold
F1 P-value 10%, 5%, 1% critical values |
53.2
0.924 7.6, 13.7, 34.3 |
The points of the thresholds estimate and their asymptotic value at 95% confidence intervals are reported in table 2. The estimates of the threshold value of INSTQ, which divides the SSA countries into three categories/regimes, are -1.375 and -1.283, a country with ‘very low INSTQ’, ‘good INSTQ’ and ‘others’ with very small value difference of INSTQ’s confidence intervals.
As mentioned before, after finding the first likelihood ratio function
LR (γ), the second and the third threshold values are computed consecutively. The first threshold estimate is the point where the
LR (γ) equals -1.375. There is also consecutive major fall in the likelihood ratio after the first threshold at -1.283. Thus, the first likelihood threshold invites for the estimation of the second threshold in the regression. The tight asymptotic confidence intervals of the threshold estimations show the high certainty of the divisions and the predictions.
Table 3
Test for threshold estimates
| Thresholds | Estimation | 95% Confi. Interval |
| LR (
γ) |
-1.375 | -1.386, -1.365 |
| LR (
γ) |
-1.283 | -1.288, -1.270 |
The rise and fall of the graph may indicate the existence of thresholds in the regression, though needs to be confirmed by empirical tests. Figures (Annex C, D & E) show the threshold estimate from the plots of the likelihood ration function. The plots may provide more inference about the threshold values and the LR tests. The point of estimates are the values of
γat which the likelihood ratio hits zero axis. The 95% confidence interval for
γ2and
γ1can be found from the
LR2r(γ)and
LR1r(γ)by the value of
γfor which the likelihood ratio lies beneath the dotted line (table 3).
Table 4 shows the categories based on the threshold values. The classification includes resource abundant countries with institutional quality lower than -1.367. The percentage of values in this category ranges from 14% to 50% of the samples over 19 years. The second and medium level group is considered as medium level INSTQ countries ranging from 0% to 14%. Countries with better INSTQ refers to countries with INSTQ threshold value greater than -1.283. The percentage of these countries in this groups ranges from 50% to 79%
Table 4
Percentage of observations in Three Regimes, By Year
| year | Dit-1 ≤ -1.366982 | -1.3367<Dit-1≤-1.2382 | -1.2382<Dit-1 | ||
| 1998 | 21 | 14 | 64 | ||
| 1999 | 36 | 0 | 64 | ||
| 2000 | 29 | 7 | 64 | ||
| 2001 | 36 | 0 | 64 | ||
| 2002 | 21 | 0 | 79 | ||
| 2003 | 36 | 0 | 64 | ||
| 2004 | 21 | 0 | 79 | ||
| 2005 | 14 | 14 | 71 | ||
| 2006 | 29 | 0 | 71 | ||
| 2007 | 36 | 0 | 64 | ||
| 2008 | 21 | 7 | 71 | ||
| 2009 | 29 | 0 | 71 | ||
| 2010 | 50 | 0 | 50 | ||
| 2011 | 21 | 0 | 79 | ||
| 2012 | 14 | 14 | 71 | ||
| 2013 | 14 | 7 | 79 | ||
| 2014 | 21 | 7 | 71 | ||
| 2015 | 29 | 7 | 64 | ||
The primary interest of the study is those under the threshold values of resource rent. The results of the three regimes indicate significant relationship between resource rent and economic growth. Regime two and three of the conventional OLS and standard error white regression results of table 5, indicates positive correlation between economic growth and resource rent when the quality of rule-of-law is “better and good” , but when the INSTQ is below the threshold level “inferior” the correlation turns to be negative and significant, supporting the RC hypothesis. For robustness analysis of the coefficient of estimates within the threshold values, we run the Arellano-bond system GMM method as it has an advantage in estimating variables with endogenous effect. The Arellano-Bond test for AR (1) in first differences and Arellano-Bond test for AR (2) in first differences test reported (annex F) both show the appropriateness of the test. The results of the regression shows similar finding of positive and significant correlation between resource rent and economic growth when the INSTQ is “better and good” the result are given in Table 5.
The most important finding of the study is that, economic institutions such as the rule of law, determine whether natural resources are a curse or a blessing for economic growth. Nations with economic institutions of higher quality are more capable of managing their resource revenue and can easily turn it into positive input for economic growth. Rule of law increases efficiency by eliminating barriers to entrepreneurial activity and establishing a rule of law that is crucial for economic activity (Karabegović, 2009).
The coefficients of interest are those on threshold values of resource rent (W1,W2, and W3)[3]. Referring to table 4, the second and third regime’s effect (threshold value of INSTQ multiplied by the resource rent) on economic growth is positive and statistically significant. While the first regime shows statistically insignificant results. Implying that in regime one (W1), the resource curse captures countries with inferior INSTQ and economic growth retarders due to the negative impacts of resource rent. On the other hand, countries with better and average INSTQ experience the positive fruits of resource rent. Countries within these regimes are supported by better rule of law than the other regime and the effect of RC diminishes.
The bottom-line of the finding is, if countries are able to improve their INSTQ within the limits or above the threshold values, they can survive from the RC effect and the coefficients support the argument of INSTQ has a major role in turning resource rents to boom or ban for economic growth. Overall, the result indicates that impact of INSTQ is not similar across the distribution and proves the non-linear relationship between resource rent and economic growth.
Table 5
Regression estimates
| Regressor | Coefficient estimate | OLS SE | White SE | GMM |
| yit-1 | 0.231 | 0.0211 | 0.0192 | 1.030*** |
| Lexit-1 | 0.0009 | 0.039408 | 0.028666 | 0.079** |
| Gdcfit-1 | 0.000272 | 0.005924 | 0.004143 | -000 |
| Dbtit-1 | 0.004963 | 0.004093 | 0.003274 | -0.004 |
| Sclit-1 | 0.004122 | 0.011298 | 0.008131 | -0.00 |
| Totit-1 | -0.00158 | 0.003635 | 0.001898 | 0.001 |
| Excit-1 | 0.043629505 | 0.039877215 | 0.037316316 | -0.00 |
| Insit-1 | 0.683555 | 1.914003 | 1.775699 | -7.187*** |
| Lexit-1 Insit-1 | -0.043629505 | 0.039877 | 0.037316 | 0.108*** |
| Gdcfit-1Insit-1 | 0.009833 | 0.00851 | 0.01118 | 0.009 |
| Dbtit-1Insit-1 | 0.016681 | 0.009821 | 0.008595 | 0.009*** |
| Sclit-1Insit-1 | 0.006538 | 0.003587 | 0.004127 | 0.004 |
| Totit-1Insit-1 | -0.004409704 | 0.003305 | 0.002712 | -0.00 |
| Excit-1Insit-1 | -0.000005352 | 0.008739535 | 0.009155970 | 0.003 |
| Insit-1
≤ -1.37 |
-0.005436276 | 0.00874 | 0.009156 | 0.008 |
| -1.37<Insit-1≤-1.24 | 0.071291 | 0.01121 | 0.019757 | 0.033*** |
| -1.24<Insit-1 | 0.003694 | 0.00775 | 0.005839 | 0.017** |
- Conclusion
In this empirical study, we have examined the resource curse effects of resource abundant SSA countries by considering a rule-of-law as a threshold value. We used panel data of 14 countries, covering the period from 1998 to 2016. One of the contributions of the paper is the adoption of the regression model based on the idea of a dynamic panel threshold effect to capture the rich dynamic change in the relationship between economic growth and resource rent, using rule of law as threshold variable. We found that the relationship between natural resource rent and economic growth is not linear and due to the threshold effects, the entire data of the countries are categorized into three regimes.
Furthermore, considering the effect of rule-of-law on the economic growth of the countries, the second and third regime coefficients indicate statistically positive significant result with 95% confidence interval, whereas, the first regime shows statistically significant negative correlation between resource rent and economic growth, indicating the resource curse dominance in the regime. Overall, the result indicates that impact of rule-of-law is not similar across the distribution and proves that their relationship is non-linear. Hence, within the threshold level, the impact of the INSTQ diminishes leading the resource rent to impact economic growth positively. Furthermore, the results indicate where the resource curse is limited and dominate the economic growth of SSA resource abundant countries. Hence, Countries with better and average rule-of-law quality in the resource reach countries experience the positive fruits of resource rent due to the better quality of their institution (rule-of-law). Above the threshold regime, the estimated coefficient indicates that the effect of RC does not operate. Henceforth, if countries able to improve their INSTQ within the limits of thresholds or above the threshold values, they can survive from the RC effect and the hypothesis of resource curse does not hold within and above the estimated values of the thresholds.
In summary, a nation desiring to benefit fully from its natural resources should not neglect the important role of good institutions for a sustainable economic growth, and with good institutions, the NCR puzzle can also be challenged. Therefore, policymakers of the resource-abundant SSA has to give due attention to the institution’s quality in general and to the rule of law in specific to curtail the impact of RC.
Annex
Annex A
List of resource reach countries
| Resource-Rich Countries |
| Angola |
| Cote d’Ivoire |
| Cameroon |
| Congo, Dem. Rep. |
| Gabon |
| Ghana |
| Guinea |
| Equatorial Guinea |
| Mali |
| Mozambique |
| Mauritania |
| Nigeria |
| Chad |
| Zambia |
Annex B: list of research variables
| Variables | Abbreviations | Descriptions | Sources |
| Real GDP Growth per capita | Yit | Real annual GDP growth | IMF |
| Institutional Quality | INSTQ | Rule of Law of resource abundant countries | WDI |
| Gross Capital Formation (investment) as % GDP | GCF | Additions to the fixed assets of the economy, plus net changes in the level of inventories. | WDI |
| Natural resource Rents | RR | Natural resources rents are the sum of oil rents, natural gas rents, coal rents (hard and soft) and mineral rents. | WDI |
| Primary education | SCL | The total enrollment in primary education, regardless of age, expressed as a percentage of the population of official primary education age. | WDI |
| Exchange rate | EXC | Determined in the legally sanctioned exchange market. It is calculated as an average annual (local currency units relative to the U.S. dollar). | WDI |
| Term of trade | TOT | The percentage of the ratio of export prices to import prices | WDI |
| Life expectancy | LEX | number of years a newborn infant would live | WDI |
| External debt | DBT | Debt owed to non-residents repayable in currency, goods, or services. | WDI |
Annex C
Confidence Interval Construction
Annex D
| Arellano-Bond test for AR(1) in first differences: z = -1.35 Pr > z = 0.177 |
| Arellano-Bond test for AR(2) in first differences: z = 0.48 Pr > z = 0.633 |
References
Alexeev, M., Conrad, R., 2011. The natural resource curse and economic transition. Economic Systems 35, 445-461.
Antonakakis, N., Cunado, J., Filis, G., de Gracia, F.P., 2017. Oil dependence, quality of political institutions and economic growth: A panel VAR approach. Resources Policy 53, 147-163.
Arellano, M., Bover, O., 1995. Another look at the instrumental variable estimation of error-components models. Journal of Econometrics 68, 29-51.
Arvanitis, Y., Weigert, M., 2017. Turning resource curse into development dividends in Guinea-Bissau. Resources Policy 53, 226-237.
Auty, R.M., 2017. Natural resources and small island economies: Mauritius and Trinidad and Tobago. The Journal of Development Studies 53, 264-277.
Auty, R.M., Mikesell, R.F., 1998. Sustainable development in mineral economies. Oxford University Press.
Aydın, C., Esen, Ö., Bayrak, M., 2016. Inflation and Economic Growth: A Dynamic Panel Threshold Analysis for Turkish Republics in Transition Process. Procedia-Social and Behavioral Sciences 229, 196-205.
Ayelazuno, J., 2014. Oil wealth and the well-being of the subaltern classes in Sub-Saharan Africa: A critical analysis of the resource curse in Ghana. Resources Policy 40, 66-73.
Badeeb, R.A., Lean, H.H., Clark, J., 2017. The evolution of the natural resource curse thesis: A critical literature survey. Resources Policy 51, 123-134.
Bakwena, M., Bodman, P., Le, T., Tang, K.K., 2009. Avoiding the resource curse: The role of institutions. MRG discussion paper (Brisbane: University of Queensland, 2009).
Bennett, D.L., Faria, H.J., Gwartney, J.D., Morales, D.R., 2017. Economic Institutions and Comparative Economic Development: A Post-Colonial Perspective. World Development 96, 503-519.
Boyce, J.R., Herbert Emery, J.C., 2011. Is a negative correlation between resource abundance and growth sufficient evidence that there is a “resource curse”? Resources Policy 36, 1-13.
Brunnschweiler, C.N., 2008. Cursing the blessings? Natural resource abundance, institutions, and economic growth. World development 36, 399-419.
Brunnschweiler, C.N., Bulte, E.H., 2008. The resource curse revisited and revised: A tale of paradoxes and red herrings. Journal of environmental economics and management 55, 248-264.
Carothers, T., 2003. Promoting the rule of law abroad: the problem of knowledge.
Chang, C.-L., Khamkaew, T., McAleer, M., Tansuchat, R., 2010. A panel threshold model of tourism specialization and economic development. International Journal of Intelligent Technologies and Applied Statistics 3, 159-186.
Deacon, R.T., 1994. Deforestation and the rule of law in a cross-section of countries. Land Economics, 414-430.
Gonzalo, J., Pitarakis, J.-Y., 2002. Estimation and model selection based inference in single and multiple threshold models. Journal of Econometrics 110, 319-352.
Gueye, C.A., Lee, M., 2015. Natural resource wealth: making resource windfalls work for Sub-Saharan African countries. International Growth Centre Blog.
Gylfason, T., 2001. Natural resources, education, and economic development. European economic review 45, 847-859.
Gylfason, T., 2004. Natural Resources and Economic Growth: From Dependence to Diversification. In: organized by European Institute for International Economic Relations (EIIW) (Ed.), Sustainable Economic Liberalization and Integration Policy Options for Eastern Europe and Russia,, University of Wuppertal, Germany, and held in Brussels.
Gylfason, T., 2006. Natural resources and economic growth: From dependence to diversification. Economic liberalization and integration policy, 201-231.
Gylfason, T., Zoega, G., 2002. Inequality and economic growth: Do natural resources matter?
Haggard, S., Tiede, L., 2011. The Rule of Law and Economic Growth: Where are We? World Development 39, 673-685.
Hansen, B.E., 1999. Threshold effects in non-dynamic panels: Estimation, testing, and inference. Journal of Econometrics 93, 345-368.
Hansen, B.E., 2000. Sample splitting and threshold estimation. Econometrica 68, 575-603.
Havranek, T., Horvath, R., Zeynalov, A., 2016. Natural resources and economic growth: A meta-analysis. World Development 88, 134-151.
Holden, S., 2013. Avoiding the resource curse the case Norway. Energy Policy 63, 870-876.
James, A., 2015. The resource curse: A statistical mirage? Journal of Development Economics 114, 55-63.
Karabegović, A., 2009. Institutions, economic growth, and the” curse” of natural resources. Fraser Institute.
Mehrara, M., 2009. Reconsidering the resource curse in oil-exporting countries. Energy Policy 37, 1165-1169.
North, D.C., Thomas, R.P., 1973. The rise of the western world: A new economic history. Cambridge University Press.
Ouoba, Y., 2016. Natural resources: Funds and economic performance of resource-rich countries. Resources Pol. 50, 108-116.
Ozpolat, A., Guven, G.G., Ozsoy, F.N., Bahar, A., 2016. Does Rule of Law Affect Economic Growth Positively? Research in World Economy 7, 107.
Papyrakis, E., 2016. The Resource Curse – What Have We Learned from Two Decades of Intensive Research: Introduction to the Special Issue. The Journal of Development Studies 53, 175-185.
Papyrakis, E., Gerlagh, R., 2003. Natural resources: a blessing or a curse?
Roodman, D., 2009. A note on the theme of too many instruments. Oxford Bulletin of Economics and Statistics 71, 135-158.
Roy, B.C., Sarkar, S., Mandal, N.R., 2013. Natural Resource Abundance and Economic Performance—A Literature Review. Current Urban Studies 01, 148-155.
Sachs, J.D., Warner, A.M., 1995. Natural resource abundance and economic growth. National Bureau of Economic Research.
Sachs, J.D., Warner, A.M., 2001. The curse of natural resources. European Economic Review 45, 827-838.
Saha, S., Ben Ali, M.S., 2017. Corruption and Economic Development: New Evidence from the Middle Eastern and North African Countries. Economic Analysis and Policy 54, 83-95.
Sala-i-Martin, X., Subramanian, A., 2013. Addressing the natural resource curse: An illustration from Nigeria. Journal of African Economies 22, 570-615.
Seghir, M., Damette, O., 2013. Natural resource curse: a nonlinear approach in a panel of oil exporting countries.
Stijns, J.-P.C., 2005. Natural resource abundance and economic growth revisited. resource policy 30, 107-130.
Wang, Y., Cheng, L., Wang, H., Li, L., 2014. Institutional quality, financial development, and OFDI. Pacific Science Review 16, 127-132.
[1] The countries are presented in appendix table A
[2] http://www.govindicators.org” for more information on governance indicators available for countries of the world
during the 1998–2016 periods.
