Resource abundance: Blessing or curse? Comparative analyses of point and diffuse resources

Abstract

The objective of this study is to assess the short and long run effects of renewable and non-renewable resource rents on economic growth in Cameroon. Taking crude oil rents and forest resource rents as proxies for non-renewable and renewable resources respectively for the period 1977–2018, we employed the autoregressive and dynamic autoregressive distributive lag (ARDL/DynARDL) modelling frameworks to achieve the stated objective. Results from the ARDL model indicate that, in the short run, both the renewable and non-renewable resources have a positive and significant effect on economic growth but the point resource is more significant than the diffused. A clear disparity in results is however noticed in the long run. While the point resources show that natural resources are a curse to long run growth, the diffuse resources reveal that natural resources are a blessing to long run growth. From the DynARDL simulation, a negative shock of the point resources leads to a fall in economic growth whereas diffuse resource indicates an increase. This shows that point resources are more prone to the resource-curse thesis and diffuse resources to resource-bless thesis. Contingent on these findings, the Cameroon government should ensure a proper allocation of natural resource revenues especially point resource rents to growth-inducing investment or social overhead capital such as open new markets, transport infrastructures, and power sectors, so as to enhance growth and development.

1. Introduction

An important question to address in the field of development studies is how richness in natural resources affects the economic growth of countries. The valorisation of the contribution of renewable resources against non-renewable natural resources in the sustainable development drive of countries is consistent with the global Sustainable Development Goals [1] and the resolutions of the recent conference of the parties (COP271) reaffirming the global resolve to foster green growth. Africa remains one of the few continents blessed with different natural resources like minerals, fisheries and forest [2]. The abundance of natural resources in a region or country can serve as a pivot of economic development since it can provide the needed energy for development ([3,4]), boost the capital structure of the economy, and provide employment which constitutes major drivers of economic growth and development. However, the natural resource and economic growth nexus is one of the most controversial issues in empirical research.

Indeed, there exist no consensus answers, with about 40 % of empirical studies confirming no significant effect, 40 % finding a negative effect and 20 % finding a positive effect ([5,6]). These conflicting results within the literature are likely due to differences in resource wealth indicators used by different authors. For instance Ref. [7], used oil income per capita while [8,9] used share of natural resources capital to total capital. Equally, natural resources rent as a share of GDP has been used by authors like [4,[10], [11], [12], [13], [14], [15]].

Besides the observed differences in natural resource indicators, some authors have sought to differentiate the various types of natural resources. For example [16], distinguish between point resources2 and diffused resources.3 Likewise [17], added coffee/cocoa to the diffused resources. However, diffused resources, such as coffee and cocoa have been shown to have a negative effect on institutions and consequently on economic growth, whereas diffused resources indicate no significant effect [17]. Moreover [18], show that high dependence on both point and diffused resources are responsible for the resource curse. This growth-unfriendliness of natural resources types has been corroborated in a recent study for Africa ([19]).

Other reasons for the disparities in results relate to estimation techniques. For instance, several authors have used panel data techniques like panel ordinary least squares ([20]) and instrumental variables (2SLS) ([21]), Pooled Mean Group (PMG) ([22]), panel VAR techniques ([9]), as well as nonlinear panel approaches ([23,24]). Time series estimation techniques have equally been applied by other authors like the vector autoregressive model (VAR) ([25]), the vector error correction model (VECM) ([26]), the autoregressive distribution lag model (ARDL) ([27,28]), ([29]) among others. In this study we adopt the ARDL technique and examine short and long run effects of natural resources (point and diffuse) on economic growth. We equally apply the new dynamic ARDL technique proposed by Ref. [30] to do simulations and examine variations of growth to negative shocks of the point and diffuse resources.

Although Africa is generally blessed with both renewable (diffuse) and non-renewable (point) natural resources [2], the role played by these resources in enhancing the livelihood of its population remain questionable [19]. Besides, Cameroon stands as one of the countries in Africa that have been blessed with both point and diffuse resources and the country is commonly referred to as Africa in miniature ([31]). Just like most developing countries wherein, there are limited competitive economic sectors like manufacturing and services, Cameroon greatly depends on the proceeds from the use of its natural resources to enhance economic and social development. Given that these resources are different, their effects on the development of the economy are not obviously the same. In this paper, we argue that there exist disparities in the effects of point and diffuse natural resource rents on economic growth. Cameroon is chosen among African countries based on the abundance of both point and diffuse resources, the availability of data, and due to the fact that very little research has been done in this domain in Cameroon. To the best of our knowledge, only two studies have been conducted on that issue by Refs. [29,31] which used total natural resource rents as independent variables on which economic growth depend. Their data were analysed using the GMM and the Structural Vector-Auto Regressive techniques respectively. This study differs with theirs in that, it distinguishes point and diffuse resources for the period 1977–2018. This is essential given that the resolutions of the recent COP27, in line with the global Sustainable Development Goals [1] emphasize the need to valorise the contribution of renewable resources against non-renewable natural resources, thereby making an appraisal of their distinct contribution vital. In this light, this study provides a more specific policy orientation with respect to the type of natural resource considered. As a result, this paper assesses the short and long run effects of point and diffuse resources on economic growth in Cameroon from 1977 to 2018, and equally investigates the response of growth to negative shocks of point and diffuse resources, that which have been given little or no consideration within extant literature.

To attain this objective, time series data collected from the World Development Indicators (32) was analysed with the helped of the ARDL and dynamic ARDL techniques to empirically investigate this relation. After ensuring that the estimated model passes all diagnostic tests, the result indicates that both non-renewable (point) and renewable (diffuse) resources have a positive and significant short run effect on economic growth but diffuse resources is more significant than point resources. Meanwhile in the long run, point resources show a negative and significant effect whereas diffuse resources continue to show a positive effect. The impulse response graph from the Dynamic ARDL simulation indicates a long run fall in economic growth as a result of a negative shock of point resources and a long run increase in economic growth as a result of a negative shock of diffuse resources. This indicates that the Cameroonian economy can enhance growth by controlling the negative effect of point resource shocks and equally continue to enhance the benefit obtained from diffuse resources. This study is different from other studies in that, it distinguishes the effect of point and diffuse resources and provides evidence of disparities on their effect on growth. Equally, it provides short and long run dynamics, accounts for shocks within the resource market and uses a more recent and robust econometric technique for data analysis. Moreover, it is one of the few studies that have considered shock simulation in the natural resource – economic growth literature. Based on the identified gap, this study sought to examine the hypothesis that point and diffuse resources does not have a significant effect on economic growth in Cameroon.

The remainder of the paper is organized as follows: section 2 reviews the relevant literature while section 3 outlines the methodological strategy. The results are presented and discussed in section 4 while section 5 provides a conclusion and policy implications.

2. Literature review

Important discussions on the role of natural resources date back to the views of Adam Smith and David Ricardo who considered these resources as the basis for economic development. Meanwhile [32], highlighted that natural resource wealth can propel the transition of countries from underdevelopment to industrial take-off [33], stated that the use of natural resources in current production activities imposes a burden on the earth's capacity and depletes the current stock of natural resources. This shows that for development to be sustainable, the current stock of total natural capital should be at least maintained at the current level. Since most natural resources are exhaustible [34,35]. argued that the sustainable development of natural resource rich countries can be guaranteed if they invest the gains from these resources into long term accumulation of other forms of capital (physical, human, institutional and social) as against the financing of current consumption.

Another strand of literature on the role of natural resources on economic development has been based on the resource curse and blessing thesis ([6,19,[36], [37], [38], [39], [40]]). Resource curse literature sought to explain why resource rich countries have not been able to use their natural resources to develop or enhance growth within their economies compared to non-resources rich countries. They argue that resources tend to retard development rather than promoting it. This development retarding phenomenon can be through the volatile nature of natural income, rent seeking behaviour of political elites, civil conflict and the Dutch disease [2].

Meanwhile, the rent seeking and corruption literature argue that natural resource wealth has the potential of increasing rent seeking and corruption within an economy by political elites who seek to make away with the resource rents [41], posit that technological and institutional development can be intentionally blocked by political elites since such development can wipe out their possibilities to loot from public income leading to poor economic development [42]. argue that rent seeking and other unproductive behaviours may be stimulated by lack of good quality institutions in resource rich countries [43]. highlighted that dictators in resource rich countries use resource rents to buy off political competitors which can consequently increase political instability.

In highlighting the wealth of a nation [44], stated that the wealth of a nation consists of produced capital, intangible capital and natural capital. Natural capital consists of non-renewable resources like oil, natural gas, coal and mineral resources, forest, cropland and protected areas. The wealth of a nation only increases if new resources are discovered and their extraction produces rents. Inefficient extraction of natural resources does not add value but rather reduces the wealth of a nation. Hence, the government should endeavour to reinvest rents from exhaustible resources into other productive assets, like human capital and infrastructural development to increase the wealth of a nation and ensure sustainable growth and development.

Recently [20], empirically showed that while natural resource abundance has a significant positive impact on economic growth, it has a negative and insignificant impact for human development [45]. employed the cross section augmented autoregressive distributive lag model along with the common correlation effect mean group to investigate the resource curse thesis and validates the natural resource blessing hypothesis for the BRIC economies. Further [46], concluded within a panel of Gulf Cooperation Council (GCC) countries that oil abundance in the GCC economies has a short and long run growth enhancing effect. However [15], empirically showed the existence of a resource curse hypothesis and further showed evidence of a mitigating effect from human capital development. On a similar page [47], demonstrated that resource abundance is a blessing to the environmental sustainability for countries of the GCC. Conversely, within a panel of G-7 economies [48], empirically settled on a resource curse hypothesis. Equally [19], recently confirmed the validity of resource curse hypothesis is in the context 39 developing African economies.

Furthermore, while applying the two stage least square technique (2SLS) from 1992 to 2016 [21], investigated the institutional and economic indicators that are more negatively affected by natural resource rents in Africa. He found that corruption, problem of rule of law, bad regulation, inefficiency of public administration, political instability, lack of voice and accountability are the main institutional problems causing GDP volatility [42]. found that natural resource abundance increases growth where good institutions exist, even though the independence effect of resources abundance was negative. While examining the relationship between agriculture, mining and quarrying resource endowment and regional economic growth over the period 2007 to 2016 for 260 European Union regions [49], authenticate the natural resource curse thesis particularly related to agriculture. Using a panel of 24 African economies from 1995 to 2017 [50], observed that natural resource rent has a negative effect on economic growth in Africa. Further, within a panel framework [51], concluded among others that natural resources enhance growth globally.

While investigating the link between natural resource rents and economic growth in the top resource-abundant countries from 1970 to 2013 using the PMG estimation technique [22], confirm the resource blessing hypothesis in the long-run, although an insignificant effect is found in the short run. The finding also suggests that economic growth exerts a positive impact on resource rent. Meanwhile [52] show that some natural resources especially oil and minerals exert a positive and nonlinear impact on growth in Nigeria through their deleterious impact on institutional quality. Employing the ARDL technique [53], show that oil revenue promotes economic growth in the short run and reduces it in the long run. Using Chinas provincial data and employing the system GMM [3], investigate the resource curse hypotheses and conclude among others that, the resources curse hypothesis exist in China but with the strength depending on the control variables employed. In a similar study [54], using data from the Chinese economy empirically highlight financial risk and fiscal decentralization aid to avoid the resource curse hypothesis. [55], highlighted that, though natural rent degrades the environment, it has a positive effect on economic growth for Africa economies.

[25] examine the link between oil exports revenue and long run economic growth in Algeria from 1979 to 2013. Their results indicate a positive association between oil revenue and long run economic growth but a negative impact between the volatility of oil revenues and economic growth. In the same light [56], show that the presence of natural resource rents alters democratic environments, which always result in institutional decay, slow growth, corruption and conflict. Similarly [57], confirm the validity of the resource curse hypothesis at the provincial level in China. Withal [24], using precious metal as an indicator of resource wealth applied the Smooth Transitional Autoregressive Distributive lag (STARDL) model from 1963 to 2017, in seven top precious metal producing countries (Australia, Canada, Mexico, Philippine, Peru, South Africa and USA). Their results show a regime specific short and long run relationship between economic growth and each precious metal type under consideration. However [23], found evidence of non-linear relationship between oil income and economic performance.

Moreover [26]. in their study for Pakistan and India employ the vector error correcting model (VECM) and find that total natural resource rent have a positive effect on GDP per capita both in both countries. Adopting a similar approach for the Kern County [27], show that there is causality between oil, agricultural abundance and education [8]. however illustrate that both natural resource abundance and dependence negatively impact on health care expenditure. In the same vein [58], note in their empirical analyses in China that there exist a resources curse and this is generally dependent on different quantiles considered [59]. empirically demonstrated among others that, natural resources curb environmental degradation for BRICS economies [60]. in a cross-sectional study of 97 developing countries show that a high share of primary products (mineral and oil) exports in GDP result in low economic growth. Whence [61], find that natural resource abundance and dependence positively and negatively impacts on economic growth respectively. With a particular focus on for sub-Saharan African (SSA) [62], recently argued that the environment-unfriendly nature of crude oil price shocks indirectly impacts negatively living standards and are therefore detrimental to the sustainable development drive of these economies. These views for SSA are corroborated by Ref. [6]. Likewise, it is argued that the recurrent oil price variations further dampen the efforts towards energy transition ([[63], [64], [65]]). Similar results are obtained by ([29,66]) in their study for Cameroon. Consequently, the negative nexus between natural resource rents and economic growth confirm the validity of the resource curse hypothesis.

Based on the aforementioned empirical literature, it can be observed that most empirical studies have focused on panel analyses with few on country specific cases. Further, the few existing country specific studies have mostly focused on the reality of more advanced economies with little work on developing countries like Cameroon. Again, simultaneously considering point and diffuse resources have been limited within literature and appraising shock simulation of natural resource exploitation on growth has generally been lacking to the best of our knowledge. As such, this study sough to fill these identified gaps within extant literature.

3. Material and methods

After reviewing some salient extant literature, this research investigates the short and long run effect of renewable and non-renewable natural resources on economic growth in Cameroon. The Cameroon economy is chosen based on the fact that limited work has been done in this domain for Cameroon, with limited comparative analyses of point and diffuse resources to the best of our knowledge. Equally, the economy is considered as Africa miniature and blessed with the abundance of both point and diffuse resources. This section highlights the sources of data, estimation techniques and model specification.

3.1. Nature and sources of data

In order to explore the effect of renewable and non-renewable natural resources on economic growth in Cameroon, we used secondary annual data from the world development indicators (WDI) from 1977 to 2018. The WDI database is a yearly compilation of aggregated data by the world bank for most economies of the world and available for usage by researchers. The study period is chosen due to data availability and it equally falls within the period of time that Cameroon started exploiting crude oil. We capture economic growth by GDP per capita growth in annual percentage (gdpc), such a measure has been used within literature ([[67], [68], [69], [70]]). Renewable natural resource rents are proxied by Forest rents as a percentage of GDP (rnr) and non-renewable natural resource is proxied by Oil rents as a percentage of GDP (nrnr) in line with [71]. We equally make use of other explanatory variables like Domestic credit to private sector as a percentage of GDP (crdpsec), Inflation GDP deflator annual percentage (inf) and Exports of goods and services as a percentage of GDP (xpot).

3.2. Estimation techniques and model specification

To attain the objectives of the study, this study adopts the time series ARDL estimation technique developed by Ref. [72] and the dynamic ARDL technique developed by Ref. [30]. We adopt this technique because of its numerous advantages. Firstly, the ARDL approach can be used for the series which are both stationary at level, I(0) and first differences, I(1). Secondly, the ARDL approach controls for the endogeneity of modelled variables ([[73], [74], [75], [76]]) and permits the simultaneous identification of the short and long run effects of the estimated variables ([77]). Finally, it uses the bounds test which yields more desirable effects and therefore is used commonly for empirical modelling. In this study, we adopt the model proposed by Ref. [53] with an adjustment on the partition of natural resources into renewable (diffuse) and non-renewable (point) resources and the adjustment of other explanatory variables. The estimated ARDL model is given as:

$${\Delta \text{GDPC}}_{\mathrm{t}}={\mathrm{\alpha }}_0+{\mathrm{\alpha }}_{1\mathrm{i}}\sum _{\mathrm{i}=1}^{\mathrm{n}}{\Delta \text{GDPC}}_{\mathrm{t}‐1}+{\mathrm{\alpha }}_{2\mathrm{i}}\sum _{\mathrm{i}=1}^{\mathrm{n}}\Delta {\text{RNR}}_{\mathrm{t}‐1}+{\mathrm{\alpha }}_{3i}\sum _{i=1}^n\Delta {\text{NRNR}}_{\mathrm{t}-1}{\mathrm{\alpha }}_4\mathrm{i}\sum _{i=1}^n\Delta {\text{CRDPSEC}}_{\mathrm{t}-1}+{\mathrm{\alpha }}_5\mathrm{i}\sum _{i=1}^n\Delta {\mathrm{INF}}_{\mathrm{t}-1}+{\mathrm{\alpha }}_6\mathrm{i}\sum _{i=1}^n\Delta {\text{XPOT}}_{\mathrm{t}-1}+{\mathrm{\lambda }}_1{\text{RNR}}_{\mathrm{t}-1}+{\mathrm{\lambda }}_2{\text{NRNR}}_{\mathrm{t}-1}+{\mathrm{\lambda }}_3{\text{CRDPSEC}}_{\mathrm{t}-1}+{\mathrm{\lambda }}_4{\mathrm{INF}}_{\mathrm{t}-1}+{\mathrm{\lambda }}_5{\text{XPOT}}_{\mathrm{t}-1}+{\mathrm{\xi }}_{\mathrm{t}}$$ (1)

where Δ is the first difference operator, α₁ to α₆ are the short run parameters while λ₁ to λ₅ are the long run parameters, α₀ is a constant and ξ_t is the disturbance term containing all the unobserved characteristics of the study and errors in measurement. After estimating the above model, different diagnostics test is performed to ensure that the output is robust notably, The Jarque-Bera Normality test to evaluate the normality of the residual, given that if the residual of the estimated model is not normally distributed, the output can lead to misleading inferencing. Further, conventional problems like autocorrelation (testing if the mathematical expectation of the error term is equal to zero) and heteroskedasticity (testing if the variance of the error term of the model is constant) are equally examined using the Breusch-Godfrey Lagrange Multiplier (LM) test for autocorrelation and the ARCH test for Heteroskedasticity all done in a bit to avoid spurious regression. To check for misspecification problem of the model, we employ the Ramsey RESET test. Furthermore, the cumulative sum (CUSUM) and the cumulative sum of square (CUSUMQ) are equally used to test for stability of the parameters. It helps to verify the constancy of the regression coefficients over time.

4. Empirical results

4.1. Correlation between variables

Before proceeding to estimate the components of our model, we first looked at the correlation between the variables in order to establish pre-estimation relations among these variables.

From the correlation results presented in Table 1, we notice a positive high correlation between non-renewable resource rent and export with growth of 0.30 and 0.41 respectively. Credit to private sector is equally positive but relatively low that is 0.09. Meanwhile, we observe a negative link between oil resource rent and inflation on growth of −0.24 and −0.16 respectively.

Table 1.

Pairwise Correlation matrix.

	GDPC	RNR	NRNR	CRDPSEC	INF	XPOT
GDPC	1.0000
RNR	0.2950	1.0000
NRNR	−0.2381	−0.4305	1.0000
CRDPSEC	0.0890	0.0023	−0.0080	1.0000
INF	−0.1595	0.0428	−0.0835	0.1873	1.000
XPOT	0.4118	0.0338	0.3913	0.1590	0.0871	1.0000

GDPC = Gross domestic product per capita, RNR = renewable natural resources, NRNR = Non-renewable natural resources, CRDPSEC = Credit to private sector, INF = Inflation, XPOT = Export of goods and services.

4.2. Unit root test

A key issue to address when using the ARDL estimation technique is to verify the order of integration of the variable. The ARDL technique requires that order of integration of all the variables should either be I (0) or I (1) or both I (0) and I (1), but should not exceed I(1). Since if the estimated variables are I (2) and above, the model will not be valid. To achieve this, we apply the Augmented Dickey-Fuller (ADF), the Philip Perron (PP) and the Kwiatkowski-Phillips-Schmidt-Shin (KPSS) unit root test to verify the order of integration. The reason for the multiplicity of test was to ensure robustness of the stationarity condition of the variables.

The unit root results in Table 2 show that, the null hypothesis that the variables gdpc, rnr, nrnr, inf and xpot contained unit root at level (for the ADF and PP test) is rejected at level, indicating that these variables are all stationary at level, implying that they are I(0) variables. This outcome is validated for the KPPS test given that the LM test statistics for the aforementioned variables are all less than the 5 % critical value (0.1460), this equally shows that the null hypothesis of trend stationarity is accepted. However, cps is stationary at first difference for all the three unit tests employed, showing that it is an I(1) variable. This shows that the ARDL technique is appropriate for the data used in this study.

Table 2.

Unit root test results.

Variables	Augmented Dickey fuller		Phillip Perron		KPSS (5 % CV(0.1460))	Integration order
	Coefficient	p-value	Coefficient	p-value	LM-Stat
GDPC	−4.5869	0.001y	−4.6196	0.001x	0.1261y	I(0)
RNR	−2.7552	0.073z	−2.8207	0.064z	0.1017y	I(0)
NRNR	−4.185350	0.002y	−4.019871	0.003y	0.1115y	I(0)
CRDPSEC	−1.0612	0.722	−1.2488	0.644	0.2717	I(1)
CRDPSEC	−4.7656	0.000x	−4.6872	0.001y	0.1230y
INF	−5.0583	0.000x	−5.0694	0.000x	0.0914y	I(0)
XPOT	−3.2736	0.022y	−3.2752	0.022y	0.1327y	I(0)

^{x,y and z} are the respective significant levels at 1 %, 5 % and 10 %. GDPC = Gross domestic product per capita, RNR = renewable natural resources, NRNR = Non-renewable natural resources, CRDPSEC = Credit to private sector, INF = Inflation, XPOT = Export of goods and services.

4.3. Cointegration and diagnostic tests

The ARDL estimation technique is very good when it comes to cointegration analysis since it makes use of the Bounds test approach which yields more desirable estimators compared to the Johansen cointegration test, developed by Ref. [78]. As such, this study employs the Bounds test to cointegration. From the results presented in Table 3, the F-statistic of the Bounds test is 6.9476 which exceeds the highest upper bound value of 4.68 at the 1 % significance level. Consequently, we reject the null hypothesis of no long run relation, thereby confirming cointegration of the ARDL model and hence the existence of a long run relation.

Table 3.

Cointegration and diagnostic test results of the estimated model.

Description	Test statistics
Cointegration test
Selected ARDL model: SIC	(3,0,3,2,4,3)
K	5
F statistic	6.9476
Critical values	I(0) Bound	I(1) Bound
1 %	3.41	4.68
5 %	2.62	3.79
10 %	2.26	3.35
Diagnostic test
Test	Statistic	p-value
R2	0.901
Adjusted R2	0.824
F-statistic	11.745	0.000x
LM Test	1.6015	0.449
ARCH Test	0.2682	0.605
RESET Test	1.2537	0.226
Normality Test	0.1450	0.930

^x1 % significant level. ARDL = autoregressive distributive lag, LM = Langrange multiplier, ARCH = autoregressive conditional heteroskedasticity, SIC = Schwarz Information criteria.

In order to confirm good fit of our model, we equally carry out diagnostic tests on the estimated model. From the results of the Jarque-Bera Normality test, with null hypothesis of normally distributed errors, the p-value of the chi-square statistic is insignificant (0.930), implying that we accept the null hypothesis and conclude that the errors are normally distributed. Equally, the Breusch-Godfrey Lagrange Multiplier (LM) test of autocorrelation rejects the null hypothesis of autocorrelation, implying that the estimated model does not suffer from issues of autocorrelation of residuals. The ARCH test also confirms the absence of Heteroskedasticity or variance variation of the residuals. Finally, the Ramsey RESET test is used to analyse whether or not the ARDL model has been correctly specified. Based on the results of the Ramsey RESET test, the null hypothesis of correct specification cannot be rejected, implying the ARDL model is correctly specified.

4.4. Discussion of empirical results

After verifying for cointegration and good fit of our model, Table 4 presents the estimated results of the short and long run effects. The results indicate that in the short run, both renewable and non-renewable natural resources exert a positive and significant effect on economic growth in Cameroon at 10 % and 1 % respectively. From the results, a 1 % increase in renewable natural resource rent will lead to 1.04 % increase in economic growth in Cameroon while, a 1 % increase in non-renewable natural resource rent will lead to an increase in economic growth by 1.43 %. The results indicate that in the short run, the exploitation of natural resources (renewable and non-renewable) is a blessing to the economy of Cameroon and this is in line with the views of Rostow who opined that natural resource wealth can propel the transition of countries from underdevelopment to industrial take-off. Although these results are contrary to the recent findings in the context of sub-Saharan Africa ([6]), they are however consistent with the natural resource blessing hypothesis, thereby corroborating the works of [20,46,52,53,55,61]. Besides authenticating the resource-blessing hypothesis, the results equally indicate that there exist a positive and significant short run effect of credit to private sector and inflation on economic growth while export exerts a negative effect in the short run. Hence, it can be claimed that along with renewable and nonrenewable natural resources, it is also necessary for Cameroon to enhance credit access to the private sector and maintain a stable state of inflation. However, in the short run the current state of short run export activities does not translate to economic benefits.

Table 4.

Estimated result of the ARDL (3,0,3,2,4,3) model.

Variables	Coefficients	Std. error	t-statistic	p-value
Short run
D(RNRT)	1.0425	0.576	1.811	0.086z
D(NRNRT)	1.4394	0.256	5.615	0.000x
D(CRDPSEC)	1.0757	0.252	4.269	0.000x
D(INF)	0.122	0.0462	2.650	0.016y
D(XPOT)	−0.6950	0.1652	−4.207	0.001y
ECT(-1)	−1.8153	0.281	−6.467	0.000x
Long Run
RNRT	0.5743	0.311	1.847	0.080z
NRNRT	−1.7731	0.155	−11.445	0.000x
CRDPSEC	−0.1099	0.0307	−3.576	0.002y
INF	0.1023	0.076	1.512	0.147
XPOT	1.0786	0.076	14.176	0.000x
C	−16.7820	2.350	−7.140	0.000x

^x,y,z are the respective significant levels at 1 %, 5 % and 10 %. GDPC = Gross domestic product per capita, RNR = renewable natural resources, NRNR = Non-renewable natural resources, CRDPSEC = Credit to private sector, INF = Inflation, XPOT = Export of goods and services.

The results equally show that the error correction term or the adjustment term from short run disequilibrium to long run equilibrium is negative as expected and statistically significant. This shows that there is adjustment from short run disequilibrium to long run equilibrium in our estimated model. Specifically, the ECT(-1) coefficient of – 1.815, implying that the deviation from long-run equilibrium of real GDP of the previous period is corrected by about 181.5 % in the current period to restore equilibrium. Hence, this high speed of adjustment is appropriate in restoring long-run equilibrium, since it lies between −1 and −2 (80; [28]).

As to what concerns the long run relation presented in Table 4, even if the short run effect of renewable and non-renewable resources are similar in the short run, there exist some disparities in the long run relation. From the results, renewable natural resources continue to exert a positive and significant effect on economic growth in Cameroon. Thus, a 1 % increase in renewable energy exploitation will lead to a 0.57 % increase in long run economic growth. It can be said that renewable natural resource of Cameroon is being utilised in a suitable manner that transmute to long run economic benefits. This result is in concordance with some previous findings ([15,22,29,46]). Conversely, we found a negative and significant effect of non-renewable natural resources on economic growth in the long run. It shows that a 1 % increase in non-renewable resource rent will reduce economic growth by about 1.77 % in Cameroon. Hence, this specific outcome validates the authenticity of the resource curse hypothesis in the Cameroonian framework in the long run for nonrenewable natural resource. This negative result corroborates the findings of ([37,60]) [50,53,62]. Equally in the long run, credit to private sector and exports respectively have significant negative and positive effects on economic growth.

Finally, we perform the Cumulative sum (CUSUM) test which helps to identify systematic changes in regression coefficients and the cumulative sum of squares (CUSUMQ) test which detects the sudden changes from the constancy of the regression coefficients. The result of the CUSUM and CUSUMQ test presented in Figs. 1 and 2 respectively indicates the absence of any instability of the coefficients since the plots of the CUSUM and CUSUMQ statistics fall inside the critical bands of the 5 % confidence intervals of the parameter stability. Therefore, stability of coefficients over the sample is confirmed.

Using the novel dynamic autoregressive distributive lag (DynARDL) technique proposed by Ref. [30], we examine the responsiveness of growth from a negative shock of renewable and non-renewable natural resources. The dynamics of the shocks are predicted by the flow of the impulse response graphs. The impulse response graphs presented in Fig. 3 for the non-renewable resources show that, as a result of a negative shock to point resources at time t = 10, economic growth increases for about two periods, and greatly decreases in response to the negative point resources shock, resulting to a new equilibrium predicted just above 5. This indicates that point resource shocks generally dampen the growth path of the Cameroonian economy. In the same light, a negative shock of the diffuse resources at time t = 10 shown in Fig. 4, indicates a slight decrease in economic growth for about two periods and an increase afterwards. This leads to a new equilibrium just below 18. The impulse response analysis confirms a long run wild decrease in economic growth due to a negative shock in point resources and a long run mild increase in economic growth due to a negative shock from diffuse resources. The outcome of the impulse response graphs indicate that the hypothesis of resources curse is more common with point resources whereas the resource blessing hypothesis is more common with diffused resources.

5. Conclusion and policy recommendation

This paper revisits and examines the short and long run effect of renewable and non-renewable natural resources on economic growth in Cameroon from 1977 to 2018. To this end, we employed the ARDL and the Dynamic ARDL models that allow for long run and short run relation between variables and equally simulations. The results obtained indicate that indeed natural resource type matters from the perspective of economic growth. From the results, Cameroon has significantly benefited both from point and diffuse resource abundance in the short run confirming the natural resource blessing hypothesis. However, in the long run, the non-renewable resources authenticate the resource curse hypothesis while the renewable resources continue to show resources blessing. The respective positive and negative long run effects of diffuse and point resources are confirmed by the impulse response graphs that indicate a long run fall in economic growth due to a negative shock of point resources and a long run increase in growth due to a negative shock of diffuse resources.

Contingent on these findings, the study recommends that it is necessary for the government of Cameroon to ensure a proper allocation of natural resource revenues (especially point resource rents) to growth-inducing investment. These resource rents could be directed especially to social overhead capital such as transport infrastructures and power sectors, as this is likely to propel sustainable growth and development. Equally, the Cameroon government should improve the control of corruption within the natural resource sector, as this will act as an incentive to prospective investors.

While this article has bridged a gap in the existing research on the impact of natural resources on growth, there are still some short comings. For instance, the study only focuses on the Cameroonian economy and involves a limited time period wherein data was sufficiently available. However, follow up research can be engaged in other developing economies in the world with an expansion of the time period. Further, the use of oil rent and forest resource rents to capture point and diffused resources is relatively restricted. This is because other natural resource rent exists such as iron ore, bauxite, coal and natural gas whose data availability are limited for short time periods. Follow up research could consider this other dimension of natural resource rents probably within a panel framework which will help increase the observation. The article did not conduct any fit back (causality) analysis due to the scope of its set objectives. In follow up research, causal relation between the variables can be studied using methodologies like the Toda Yamamotto vector autoregressive (VAR) approach and the Bayesian VAR approach. Moreover, further research could be conducted to incorporate recent trends in the resource-growth nexus following the availability of more recent data in the study area.

Data availability statement

The data associated with this study has not been deposited into any publicly available repository. However, data will be made available upon reasonable request from the corresponding author.

CRediT authorship contribution statement

Gildas Dohba Dinga: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Ndam Mama: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Conceptualization. Elvis D. Achuoc: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix 1.

Fig. 1.

COSUM residual test for stability.

Fig. 2.

COSUMSQ residual test for stability.

Fig. 3.

Impulse response of economic growth to shocks from point resources.

Fig. 4.

Impulse response of economic growth to shocks from diffuse resources.

Appendix 2. List of Abbreviations

2SLS

Two Stage Least Square

ADF

Augmented Dickey-Fuller

ARCH

Autoregressive Conditional Heteroscedasticity

ARDL

Autoregressive Distribution Lag Model

COP

Conference Of The Parties

CUSUM

Cumulative Sum

CUSUM

Cumulative Sum of Squares

DynARDL

Dynamic Autoregressive Distribution Lag Model

GCC

Gulf Cooperation Council

GDP

Gross Domestic Product

GMM

Generalised Method of Moments

KPSS

Kwiatkowski-Phillips-Schmidt-Shin

LM

Lagrange Multiplier

PMG

Pooled Mean Group

PP

Philip Perron

SSA

Sub-Saharan African

VAR

Vector Autoregressive Model

VECM

Vector Error Correction Model

WDI

World Development Indicator

==========