Alleviating role of energy innovation on resource curse: a case of OECD countries

Abstract

The research motivates to provide some implications about the role of energy innovation (EINN) in the resource curse hypothesis. The significance of EINN is crucial for mitigating the economic and environmental damage caused by the excessive consumption of natural resources (NR). The study aims to inspect the effect of energy consumption (EC), EINN, and NR on the economic growth (EG) of OECD countries from 1990 to 2015. The study emphasizes the importance of EINN by incorporating the EINN in the empirical framework of the resource curse hypothesis (RCH) and contributes to the current research on RCH. The study has used the various advanced econometrics techniques that are robust in controlling the cross-section dependence (CD) and slope heterogeneity of the panel data. The results clarify that EINN is a decisive factor in the analysis of the RCH. The positive association between EINN and EG suggests that EINN is beneficial in improving the EG of OECD countries. The results reveal the negative impact of NR on EG; however, with the inclusion of EINN, the absolute value of the NR’s effect has declined significantly from −0.299% to −0.076% in the long run. The findings suggest that improvements in the EINN are crucial to increase the efficiency and productivity of NR and to avoid the RCH in the OECD countries. Thus, OECD countries should encourage energy innovation policies by promoting their application, so that replacement of traditional energy sources with new energy can be realized.

HIGHLIGHTS

• The importance of energy innovation is expressed in the context of the natural resource curse hypothesis (RCH).

• Advanced panel data method CS-ARDL is used to control the panel cross-sectional dependence and slope heterogeneity issues.

• Energy innovation plays a decisive role in alleviating the RCH in the OECD countries.

Keywords:

• Energy innovation
• natural resources
• CS-ARDL
• OECD

Introduction

Natural resources (NR) include all those means that are gifted by nature to be used to fulfill human wants. In addition, these objects are linked with nature and energy to generate economic growth (EG) and human welfare [1]. NR abundance not only provides a comparative advantage in the exports of primary goods but also facilitates capital and technology imports, which increase labor productivity as well [2]. All these factors have a positive impact on the quality and quantity of goods produced, and thus, support exports and assist in building foreign exchange reserves [3]. Furthermore, countries with abundant natural resources can provide energy for national economic expansion, as energy is crucial for economic growth [4]. Additionally, the price of natural resources has been impacted by the COVID-19 pandemic [5]. Therefore, it is a general opinion that high stocks of natural resources are vital for economic growth and human survival, as well as prosperity [6,7].

This opinion focuses on the extensive use of natural resources as a leading source of the industrial revolution [8]. Conversely, the conflicting evidence provided by some studies has changed the contribution of NR into a puzzle [9]. From the viewpoint of world economic development, it implies that countries with abundant natural resources may not inevitably follow good EG and even can have poor economic progress. This situation refers to the "Resource Curse Hypothesis (RCH)" [10–12]. RC generally refers to a situation when a country has abundant NR that, rather than supporting economic growth, becomes a hurdle and constraint to it [13]. This condition emerges due to the rapid extraction and consumption of NR [9].

The international extraction of NR is continuously growing. An unprecedented global surge was observed in demand for raw materials in the last few decades because of rapid industrialization and consistent high extraction of NR in developed countries [14]. While the world economy is projected to increase fourfold further and the world population to grow over 9.2 billion, there will be an added burden on the NR material and energy resources as the growing population need more food, additional manufacturing products, high-energy resources, and more water [15,16]. The total amount of materials acquired from the extraction of NR increased by 60% in 2008, whereas OECD nations comprised 38% of the domestic extraction of used material globally in 2008. The extraction and consumption of NR in OECD countries have enlarged but at a much slower pace than at the global level [14].

Conversely, the extensive use of NR to achieve high EG creates severe environmental concerns [8]. According to WDI , CO₂ emission from fossil fuels increased from 4.194 metric tons in 1990 to 4.981 metric tons in 2014. The extraction of NR is also a significant source of energy consumption (EC). Around 80% of global energy is consumed by the extraction of fossil fuels. Moreover, extracting industries are responsible for nearly 50% of the world’s CO₂ emission and 90% of biodiversity damage. Meanwhile, the depletion of NR has more than tripled in the previous five decades, which is consuming enormous energy [17]. The massive EC due to extracting NR and achieving high EG has serious implications for environmental safety [4]. This mounting EC is a universal concern and a critical threat to sustainable development because of environmental damage [18–21]. These challenges demand aggressive policies to encourage considerable growth in resources efficiency, mainly through technological improvement and innovations [20,22]. The solution to avoid ecological risk is conversion to a sustainable energy framework through innovations, which will ensure broader access to energy, environment sustainability, affordability, improved investment in clean technology, and developments in energy efficiency [23–25]. The speedy transformation toward energy-related technology is needed to fulfill the sustainable development goals set by United Nations, covering the clean environment challenge consistent with Paris Agreement IEA [26,27]. According to IEA [26], we need significant advancement in green and clean energy innovations to achieve net-zero emissions.

Considering the aforementioned, it can be inferred that the existing EC patterns may lead to RCH between NR and EG because of a worse impact on the environment and economy Canh, et al. [28]. The empirical findings that have described this curse impact already appeared in many countries [29–32]. Thus, if a curse arises, how this curse impact can be evaded or converted into a blessing is the main purpose of our study. Thus, the major objective of this paper is to discuss how to deal with a resource curse impact if it occurs or how it can be avoided or minimized. This study investigates the effect of energy innovation on the resource curse. The specific research question is whether EINN is a significant factor in controlling the resource curse?

EINN is an advanced process that produces innovative or better energy technologies that lead to more energy resources, improve the standard of energy services, and decrease ecological and financial costs related to energy supply and usage [33]. The EINN also proposes the prospects of multiple additional benefits, comprising a clean atmosphere, affordability of energy, more energy security, and improved energy efficiency because of enhancements in the energy system. Improvements in EINN reduce the monetary cost of energy and make it more cost-effective for consumers. EINN also decreases the political and environmental impact of an energy service. Progress in EINN increases the competitiveness of various products in both national and international markets, leading to the expansion of economic activities and speeding up EG. Over the past century and a half, advances in EINN have acknowledged a key solution for multiple issues. It has enhanced domestic oil production by reducing extraction and exploration cost, and increasing dependence on ample coal reserves while abating environmental emissions [34]. Thus, progress in ENIN is critical in reducing the economic and ecological effects of the extraction and consumption of natural resources.

RC hypothesis is established on the seminal modeling of Sachs and Warner [35], followed by plenty of studies. The empirical findings of these studies lead to contradictory findings, and some studies conclude in favor of this hypothesis [29–32], while others view resources as blessings [9,36–38]. A meta-analysis of 43 scholarships concluded that nearly 40% of studies were in line with RC hypothesis, 40% found no impact of NR on EG, and 20% found a positive relationship between the two variables [39]. These inconsistent experimental findings may arise by ignoring the critical role of some conclusive factors in the augmented production function.

Previous studies have explored various important factors that can contribute to RCH. These factors include NR [28,32,36,38], economic complexity [28], natural resources rent [9,31], institutional quality [40,41], energy use [6,42], energy prices [37,43], trade openness [9,38,44], CO2 emissions [30,40], GDP [9,29], gross fixed capital formation [28,31], financial development [9,37], human capital [11,32,45], and urbanization [40]—for details, see Table 1. Nevertheless, the existing literature ignored the role of EINN in investigating the RCH. EINN is a missing factor in empirical studies that can modify NR abundance and EG's association in the analysis of RCH. Improvements in ENIN help in reducing the economic and environmental effects of the extraction and consumption of NR. EINN is valuable to mitigate the curse impact of NR to achieve green and sustainable growth IEA [26].

Table 1. Literature review summary.

Author	Year	Country/Region	Variables	Methodology	Existence of RCH
M. Li et al. (2023)	1990-2020	G7 countries	COVID-19, natural resources rents, economic growth	CSARDL	No
Badeeb et al. (2023)	1990–2018	BRICS countries	Real exchange rate, natural resource, Economic growth, innovation		Mix results
Khan et al. (2023)	1990-2020	G7 Countries	Natural resources, economic growth, carbon emissions, human capital, green growth	MOM Quantile Regression	Yes
M. Li et al. (2023)	1990 - 2020	G7 Countries	Renewable energy consumption, research and development (R&D), gross fixed capital formation, natural resources	MOM Quantile Regression	No
Adabor et al. (2022)	2011 -2020	Ghana	Economic growth, oil resources rent, non-oil revenue, foreign direct investment, capital and interest rate	ARDL, Toda–Yamamoto test approach	NO
Deng et al. (2022)	1990–2020	BRICS countries	Economic risk, Renewable energy consumption, technological innovation	Panel quantile regression,	No
Khan et al. (2022)	1990–2020	Seven countries	Renewable energy consumption, political risk index, gross capital formation, economic growth, natural resources	Quantile regression	Mix results
Haseeb et al. (2021)	1970-2018	Top Asian economies (5)	NR, GDP	Quantile-on Quantile regression	No
	1970-2018	Saudi Arabia	Carbon emissions (CO2), Energy use (E), Fossil fuel energy consumption (FF), Population density (P), Natural resource (NR), Trade openness (TOP)	ARDL Cointegration, Impulse Response Function (IRF)	Mixed results
Dogan et al. (2020)	2001-2017	11 Developed countries	World governance index (WGI), financial development (FD), Natural resource rent (NRR), gross fixed capital formation (GFCF), GDP per capita, trade openness (TOP), index of financial openness (FOP)	Panel quantile regression	NO
	1970-2017	Iran	GDP per capita, NR, democracy, GFCF, FD	Vector error correction mechanism (VECM)	Yes
Jianqiang et al. (2020)	1990-2017	G7	NR, FD, Energy prices (EP), research and development (R& D)	Wasteland (2007) cointegration, CCEMG, CS-ARDL	No
Wei et al. (2020)	1990-2017	G7 countries	FD, NR, EP, TOP, R&D, GDP	Westerlund cointegration, Augmented Mean Group (AMG)	No
Badeeb (2020)	1970-2016	Malaysia	CO2, GDP, NRR,	ARDL bound test, Bayer and Hanck cointegration test	Yes
Anser et al. (2020)	1975-2017	Pakistan	FD, NRR, GFCF, Domestic credit (DC), market capitalization, TOP, GDP	ARDL cointegration, IRF	Yes
Atil et al. (2020)	1972–2017	Pakistan	FD, OP, GDP, economic globalization (EG),	Cross-quantilogram approach	No
Wang et al. (2020)	2006–2018	China	Resource consumption, land consumption, EC, SO2 emissions	Tobit regression	No RCH in eastern region RCH exists in the central and western regions
Y. Li et al. (2020)	1990 - 2017	N-11 countries	NRR, I, HC, FD, technological innovation (TI),	AMG, Mean Group (MG)	Yes
Ampofo et al. (2020)	1981 - 2017	Mineral-rich countries	GDP, NRR, TOP	Cointegration, Nonlinear ARDL(NARDL), Asymmetric causality	NA
	1995 - 2017	China	GDP, Renewable energy (RE), Non-renewable energy (NRE), P, TI	FMOLS	NA
Canh et al. (2020)	2002–2017	41 High Income Countries	NRR, GDP per capita, foreign direct investment (FDI), GFCF, Government consumption (G)	GMM, Sequential estimation of linear panel-data model (SELPDM)	Mixed results
Gritsenko et al. (2020)	2006–2015	Russian Arctic regions	OP, Exchange rate, C, Budget	regression	No
Kassouri et al. (2020)	1984-2016	Oil-exporting countries	FD, OP, GDP, G, I, financial development (FD)	Panel smooth transition regression	threshold level of democracy able to reverse the finance resource curse
Guan et al. (2020)	1971–2017	China	FD, NRR, Globalization (Glob), HC, GDP	ARDL bound test, Bayer and Hanck cointegration, FMOLS, DOLS, canonical cointegrating regression (CCR), Granger causality	Yes
Asif et al. (2020)	1975 to 2017	Pakistan	FD, GFCF, Private credit (CP), GDP, NRR	ARDL Cointegration, Impulse Response Function (IRF), Variance Decomposition Analysis (VDA)	Mixed results
Sun et al. (2020)	1990-2017	E-7	NRR, HC, TOP	Westerlund cointegration, Augmented Mean Group (AMG)	Yes
Gradstein et al. (2020)	1999–2012	5,565 municipalities in Brazil	GDP, nighttime light, oil, OP,	OLS regression	No
Tiba et al. (2019)	1990 to 2016	26 African countrie	GDP per capita, CO2, TOP, FDI, GFCF, Urbanization (URB), Human Development Index (HDI), Education (E), IQ	Fully modified OLS (FMOLS), Dynamic OLS (DOLS)	Yes
Shahbaz et al. (2019)	1976–2016	United States (USA)	GDP per capita, NR, GFCF, Labour (L), Oil prices (O),	VECM Granger causality	Yes
	2000-2015	29 countries	Natural resource (NR), GDP per capita, GFCF, Human capital (HC), Corruption (C), ethnic and religious conflicts, Institutions quality (IQ)	Conditional non-linear ARDL	Yes
Amini (2018)	1996-2010	83 Panel countries	Natural resource (NR), IQ, GDP	Fixed effect (FE), Random effect (RE)	Yes
Sepehrdoust et al. (2018)	2000-2016	OPEC countries	Social development, technological improvement, FD, GDP, L GFCF,	Generalized Least Square (GLS),	No
Allcott et al. (2018)	1969-2014	USA	Technology, economic recovery, reserves, Oils and gas endowement	FE regression	No
	1985–2010	China	URB, GDP, GFCF, E, Import(M), Export(X)	Panel Cointegration, Threshold estimation	NA

Moreover, the existing literature (for detail see Table 1) has examined the RCH in different regions while relatively less attention has been given to investigating RCH in the OECD countries. OECD countries are considered resource-rich countries with comparatively low population concentration. The Covid-19 epidemic and the Russian invasion of Ukraine have significantly disrupted energy and technology supply lines in OECD nations. The energy shift may be hampered by rising energy and material costs as well as shortages of essential minerals, semiconductors, and other components. While the extraction and consumption of NR in OECD countries have also enlarged [14]. Given the foregoing, it can be concluded that current EC patterns in OECD may result in RCH between NR and EG due to a negative impact on the environment and economy Canh et al. [28].

In this context, developing secure, resilient, and long-term clean energy supply chains is critical for OECD economies. Improvements in Energy innovation are helpful to change the curse effect between NR and EG [16]. OECD countries are increasing the production of clean energy technologies with the goals of accelerating zero-carbon transitions, bolstering energy security, and participating in the new global energy economy. Moreover, these countries supported clean and green energy innovations through various policy initiatives such as facilitating research and development activities since 1970 and attracting investment from various international platforms [46]. Energy technology innovation has far-reaching implications for OECD economies. Despite having a small GDP share, the extensive use of energy in OECD economies makes constant supply and stable cost of energy essential to maintain a stable growth path. Rapid growth in energy demand, combined with growing issue about energy security and the environment, raises concerns about the current energy system’s sustainability and asks for increased efforts to create and implement better energy technologies that can help in the development of a sustainable energy system IEA [47].

In this context, this paper studies the OECD countries as an example from the perspective of improvement in EINN. Therefore, the study aims to fill this gap in the literature by incorporating the EINN as a considerable factor in studying the RCH between NR and EG in OECD countries using data from 1990 to 2015. The study aims to examine the direct impact of EINN on EG and the indirect impact through NR.

The novelty or contribution of this paper to the current stream of literature goes as follows: (1) Systematic research on EINN is a relatively recent endeavor that has been ignored in earlier literature. The growth impact of EINN, which has not previously been studied in the empirical literature, was examined in this study. (2) The study has highlighted the importance of EIIN by incorporating EIIN into the empirical framework of RCH and assessing its impact on RCH. (3) Furthermore, the paper emphasizes the significance of EINN in mitigating the impact of the resource curse. The study has analyzed whether the RCH exists in OECD countries and suggested meaningful policies based on the empirical findings. (4) Furthermore, this study employed an advanced methodology, CS-ARDL, to address various econometric issues that commonly arise in panel data. CS-ARDL developed by Chudik and Pesaran [48] is an advanced method that addresses both cross-sectional dependence (CD) and slope heterogeneity issues. It is robust in the incidence of specification bias, potential endogeneity, autocorrelation, and non-stationarity issues and delivers efficient and unbiased estimates Ahmad et al. [49].

The structure of the paper is as follows. “Literature review” section provides a detailed literature review on the RCH. “Data, model, and methodology” section introduces the econometric methods and data. Results are analyzed and discussed in “Results and discussion” section. “Conclusion and policy recommendations” section concludes the paper and offers policy implications.

Literature review

Energy-growth nexus

Global energy consumption has increased dramatically since the second part of the twentieth century, and estimates indicate that this trend will continue for the next few decades [4,50]. Conclusively, the combined impact of global population growth, rapid economic expansion, and considerable structural changes is probably the main cause of this sharp rise in energy consumption [15,51]. There has been a lot of discussion on whether or not energy use contributes to economic expansion. Both the theoretical and empirical sides of this discussion have a long way to go [52,53].

There is a vast body of research examining the relationship between energy consumption (EC) and economic growth (EG). The causal association between EC and EG has been extensively researched in the energy economics literature. The causal relationship between EC and EG can be classified into four forms, each of which has significant consequences for energy policy [19]. Briefly, it can be summarized as follows.

Neutrality hypothesis

The hypothesis that there is no causal relationship between energy use and GDP is known as the neutrality hypothesis. It indicates that there is no relationship between energy use and GDP and that neither conservative nor expansive energy-use policies have any impact on economic growth [19]. Thus, the failure of a causal link between energy use and real GDP supports the neutrality theory. The studies of [54,55] supported this hypothesis. Eyuboglu and Uzar [56] also proved the neutrality hypothesis. Steve et al. [57] examined the neutrality hypothesis for African blocs.

Conservation hypothesis

The one-way causality that runs from GDP to energy use is known as the "conservation hypothesis." It implies that reducing energy consumption can be achieved with little or no negative impact on economic growth, as in a less energy-dependent economy. If a gain in real GDP generates a rise in energy consumption, the conservation hypothesis is supported [58]. The studies which proved this hypothesis are [59–62].

Growth hypothesis

The unidirectional causal link between energy consumption and GDP is called the “growth hypothesis.” It suggests that constraints on energy use may harm economic growth, but upsurges in energy promote economic growth. The growth hypothesis states that energy consumption plays a significant role in economic growth, both directly and indirectly, as a counterpart to labor and capital in the production procedure [58]. Consequently, we can conclude that energy is a constraint on economic growth and that disruptions in energy supply will have a damaging effect on economic growth [19]. This hypothesis is evidenced in the studies of [63–68] between total energy consumption and GDP. While [19,39,69–72] examined the growth hypothesis between renewable energy consumption and GDP.

Feedback hypothesis

Energy consumption and economic growth have bidirectional causality which is called the “feedback hypothesis.” In this situation, restrictive energy policies may harm economic growth, whereas expansionary energy policies promote economic growth and development [19]. [73–78] studied the Feedback Hypothesis between energy consumption and GDP growth. Whereas, [54,79–83] discovered two-way causality between renewable energy and GDP growth.

Resource curse hypothesis

Since Adam Smith’s time, countries that are blessed with NR have been a firm conviction, consuming these resources as a vital path to achieve sustained EG [84]. Moreover, today the role of energy in the development of a country is arguably different from its role in the 19th and early 20th centuries. Economists have analyzed that, countries with abundant NR have, on average, lower growth rates than the countries with limited or fewer natural resources [85]. These countries suffer from a "resource curse" (Badeeb et al. [10] states an inverse relationship between the endowment of NR and EG).

In their seminal work, Sachs and Warner [35] verified the inverse association between EG and the quantity of NR a country is bestowed with. This led to a theoretical puzzle and added the Dutch disease debate in the standard endogenous growth theory. It was assumed that resource richness must improve the investment and thus EG.

Six decades before, various studies used different econometrics approaches to investigate the RC phenomenon. Conversely, many studies applied the pioneer model of Sachs and Warner [35], intending to examine the effect of various measures of resource abundance to analyze the RC hypothesis. Other macro variables, such as education, institutional quality, investment, human capital, inflation, and education, were also employed, significantly affecting EG. However, all these studies confirmed the presence of the RCH. Despite this finding, further empirical studies found the positive effect of resource intensity on EG. In contrast, countries with limited or no access to NR achieved a high level of EG. The increase in EG was three times more speedy in resource-scarce countries than in resource-rich countries [35,86]. Then in the nineteenth century, the resource boom triggered EG in Latin American countries. At the start of the twentieth century, the resource curse subject emerged as a debatable fact in economics research [40].

The current literature on RCH can be outlined into two streams, with the first stream of literature emphasizing the relationship between NR and EG [31,87,88]. Khan, et al. [89] discovered that resources are both a curse and a blessing in the panel of the G7 countries. While Khan, et al. [90] examined that green growth exacerbates the natural resource curse in the G7 countries. While Li, et al. [91] discovered that NR are a blessing for the economies of G7 countries. Hordofa, et al. [92] found that NR rents improve economic growth. Khan, et al. [93] also analyzed that NR improves the economic performance of developed and developing economies. In contrast, the second stream has examined the RCH between NR and financial development (FD) [9,42,45]. Deng, et al. [94] found a positive impact of NR on the economic growth of BRICS countries. However, there are mixed and inconclusive results in the previous studies regarding the validation of RCH. Adabor, et al. [95] found some inconclusive findings about the impact of NR rent on economic growth. These mixed results can be due to the absence of a key factor in the framework of RCH or because of different methodologies.

Earlier studies have explored various significant determinants of RCH. However, these studies have not incorporated EINN in the investigation of the RCH, which is a meaningful factor to explain the RCH between NR and EG. Thus, these models suffer from missing variable bias, which puts the reliability of their findings under question and also shows that this research gap exists. While Badeeb, et al. [96] stated that innovations can replace unprofitable investments with more profitable investments and can reduce the natural resource curse effect by reducing the exchange rate appreciation.

Moreover, most of the existing studies have used various methods to estimate RCH, which includes ARDL [30,31,42], NARDL [44], FMOLS [32,97], DOLS [32,40], AMG [38,45], and VECM [29,88] (for details see Table 1). However, these approaches do not treat the cross-sectional dependence and slope heterogeneity of the panel data efficiently. Thus, conflicting results of RCH in existing studies can also be because of different estimation approaches that ignore CD and slope heterogeneity together. Therefore, our research fills this literature gap by introducing EI as a critical factor in explaining the RCH between NR and EG. To serve this purpose, the CS-ARDL method is employed, which is robust in the presence of CD and slope heterogeneity. Our study has examined the NR of OECD countries with the perspective of EINN, which is the main contribution of this research.

Data, model, and methodology

Data sources

This study covers the time period of 1990 to 2015 for 29 OECD countries to investigate the role of EINN in RCH. OECD countries are considered resource-rich countries with relatively low population density. These countries have started investing in green energy technology innovation since the 1970s and have structured various policies to encourage EINN [98]. The selection of time frame is subject to the availability of data of our focal variable EINN. The study has used two data sources to acquire the desired data. The complete data description is given in Table 2.

Table 2. Description of variables.

Variables	Measurement	Data source
Economic Growth (EG)	Real per capita income in constant 2010 US$	WDI (2020)
Capital stock (GFCF)	The share of Gross fixed capital formation in GDP.	WDI (2020)
Labor force (L)	Labor force participation between ages 15-64.	WDI (2020)
Natural resources (NR)	Total natural resources rent as a ratio of GDP.	WDI (2020)
Energy Innovations (EINN)	Energy Technology RD&D Budget in millions USD (2020 prices and PPP)	OECD (2020)
Energy Consumption (EC)	Energy use in kg of oil equivalent per capita.	WDI (2020)

Following [98–101], we have proxied EINN, by the Energy Technology RD&D Budget in millions USD. All the variables are converted into a natural logarithm to standardize the data in the same unit, as the variables are measured in different units. The log-linear transformation mitigates the likely distortion of the data; moreover, it provides consistent and efficient outcomes [12]. Descriptive statistics are presented in Table 3.

Table 3. Descriptive statistics.

Variables	Obs	Mean	Std.Dev	Min	Max	Skew.	Kurt.
EG	711	8.7	1.385	5.247	11.39	−.365	2.427
GFCF	470	23.566	2.657	18.421	29.299	.486	2.084
L	754	4.081	.096	3.808	4.226	−.718	2.683
NR	623	.611	2.433	−8.076	3.703	−1.574	5.533
EINN	577	3.132	1.989	−4.51	7.828	−.497	3.059
EC	754	8.263	.421	6.854	9.152	−.383	3.593

Theoretical framework and model

The neoclassical economists believe that the labor force and capital stock explain the pattern of long-run EG. Conversely, energy input is missing in the neoclassical framework. However, ecological economists postulate energy as a leading driver of EG in the aggregate production function, while capital stock and labor are considered moderators formed by energy [45].

Natural resources are considered a significant source of EG and national wealth for resource-rich countries. The utilization of national resources through NR decreases the debt burden of any country and provides the government more flexibility in national resources allocation to achieve the desired development goals. Moreover, nations blessed with NR develop faster than countries with less or without NR [28].

Since the pioneering study of Sachs and Warner [24], a number of researchers investigated the RC phenomenon between NR and EG in developed and resource-rich countries [3,26,29,35,46] but provided ambiguous results. These studies have also used various other economic, institutional, political, geographic, and ecological factors to explain this impression, but the findings are inconclusive. In fact, there is little consent on this argument that NR wealth super EG.

These contradictory results can be because of omitting some influential factors in the augmented production function, which may be crucial in explaining the relationship between NR and EG. None of the studies have highlighted the importance of EINN in the formation of RCH between NR and EC. While improvements in ENIN reduce the economic and ecological effects of the extraction and consumption of NR, it is also valuable to moderate the curse impact of NR to achieve green and sustainable growth IEA [16]. Considering the direct and indirect implications of EINN on EG, this study has incorporated the EINN as a decisive factor of NR and EG association in the augmented production function. Following the framework of [45,47], we incorporate the EINN, NR, and EC in the aggregate production function. First, we estimate Model 1, and then we estimate model 2 by incorporating EINN in the augmenting production function. The general construction of augmented production function is demonstrated as follows: (1) $${\mathit{Model}\mathrm{ }1\mathrm{ }EG}_t=q\left({\mathit{GFCF}}_t,\mathrm{ }{L}_t,\mathrm{ }{EC}_t,\mathrm{ }{NR}_t\right)\mathrm{ }$$(1) (2) $${\mathit{Model}\mathrm{ }2\mathrm{ }EG}_t=q\left({\mathit{GFCF}}_t,\mathrm{ }{L}_t,\mathrm{ }{EC}_t,\mathrm{ }{NR}_t,\mathrm{ }{\mathit{EINN}}_t\right)\mathrm{ }$$(2)

Following [12,47], we transform the general framework of the production function into log-linear modeling by taking the natural logarithm of all the selected variables in models (1) and (2) as follows: (3) $${\mathit{lnEG}}_t=\mathrm{ }{\beta }_1+\mathrm{ }{\beta }_K\mathrm{ }{\mathit{lnGFCF}}_t+\mathrm{ }{\beta }_L\mathrm{ }{\mathit{lnL}}_t+\mathrm{ }{\beta }_{EC}\mathrm{ }{\mathit{lnEC}}_t+\mathrm{ }{\beta }_{NR}\mathrm{ }{\mathit{lnNR}}_t\mathrm{ }$$(3) (4) $${\mathit{lnEG}}_t=\mathrm{ }{\beta }_1+{\beta }_K\mathrm{ }{\mathit{lnGFCF}}_t+{\beta }_L\mathrm{ }{\mathit{lnL}}_t+\mathrm{ }{\beta }_{EC}\mathrm{ }{\mathit{lnEC}}_t+{\beta }_{NR}\mathrm{ }{\mathit{lnNR}}_t+\mathrm{ }{\beta }_{\mathit{EINN}}\mathrm{ }{\mathit{lnEINN}}_t\mathrm{ }$$(4)

Here, ln refers to the natural log and ${EG}_t, {\mathit{GFCF}}_t, L_t, {EC}_t, {NR}_t, {\mathit{EINN}}_t$ represent the economic growth, capital stock, labor force, energy consumption, natural resources, and energy innovations, respectively. The log-linear conversion provides efficient and consistent estimates [12,47]. We exclude the other important factors of EG explained in existing literature because they are of limited interest regarding the focus of our study.

Empirical Methodology

Cross-sectional dependence (CD), slope heterogeneity, and unit root test (CIPS)

First-generation unit-root tests normally assume that panel data is cross-sectionally independent. However, in the case when the countries in the sample are socially, economically, and culturally connected, this assumption may not hold. The OECD countries have some identical attributes because of economic ties and cooperation. Therefore, we start our analysis by testing for CD in our panel data. Ignoring CD while analyzing short- or long-term relationships may lead to misleading and inefficient inferences [11,12]. Checking for CD is also crucial, as nearly all of the second-generation tests, such as Pesaran [48] CIPS assume the presence of CD. Therefore, checking for cross-sectional dependence is the first step of panel data analysis, which helps us select between first- and second-generation tests. The test statistic of the Pesaran [49] CD test is (5) $$CD=\sqrt{\frac{2T}{N\left(N-1\right)}\left(\sum _{i=1}^{N-1}\sum _{j=i+1}^N{\widehat{\rho }}_{ij}\right)}\sim N\left(0,1\right)\mathrm{ }$$(5) $$T=1,\mathrm{ }2,\mathrm{ }\dots ,\mathrm{ }N$$ (6) $$M=\sqrt{\frac{2T}{N\left(N-1\right)}\left(\sum _{i=1}^{N-1}\sum _{j=i+1}^N{\widehat{\rho }}_{ij}\right)\frac{\left(T-k\right){\widehat{\rho }}_{ij}^2-E\left(T-k\right){\widehat{\rho }}_{ij}^2}{\mathit{Var}\left(T-k\right){\widehat{\rho }}_{ij}^2}}\mathrm{ }$$(6) where ${\widehat{\rho }}_{ij}^2$ is residual pairwise correlation coefficient sample estimated, resulting from OLS. The null hypothesis of no cross-sectional dependence is tested against the alternative hypothesis.

Slope heterogeneity test

Since countries in the sample may differ in socioeconomic and demographic structures, slope coefficients could be heterogeneous. Therefore, we applied the slope heterogeneity test of Pesaran and Yamagata [50]. The empirical model for delta tilde (${\tilde{\Delta }}_{SH})$ and adjusted delta tilde (${\tilde{\Delta }}_{\mathit{ASH}})$ is: (7) $${\tilde{\Delta }}_{SH}=N^{\frac{1}{2}}{\left(2K\right)}^{\frac{1}{2}}\left(\frac{1}{N}\tilde{S}-k\right)\mathrm{ }$$(7) (8) $${\tilde{\Delta }}_{\mathit{ASH}}=N^{\frac{1}{2}}{\left(\frac{2k\left(T-k-1\right)}{T+1}\right)}^{-\frac{1}{2}}\left(\frac{1}{N}\tilde{S}-k\right)\mathrm{ }$$(8)

This test outperforms traditional slope heterogeneity tests as, unlike others, this test allows for cross-sectional dependence [51].

Cross-sectionally augmented IPS (CIPS)

After testing for CD and slope heterogeneity, we proceed to testing the stationarity of variables in our model. The conventional unit root tests, which are frequently used, do not consider the CD and heterogeneity of panel data. Therefore, we apply the second-generation panel unit root test “CIPS” developed by Pesaran [48]. CIPS is a standard approach that addresses the heterogeneity and CD of panel data while checking for unit root. The equations of this test can be stated as: (9) $$\Delta Y_{it}={\alpha }_i+{\beta }_i{Y}_{i,t-1}+{\mathrm{\gamma }}_i{\overline{Y}}_{t-1}+{\theta }_i\Delta {\overline{Y}}_t+{\mu }_{it}$$(9) where ${\overline{Y}}_t=N^{-1}{\sum }_{j=1}^N{Y}_{jt}$ is cross-section average.

The null hypothesis for the above-given equation is ${H_0: \beta }_i=0$ for all i, whereas alternative hypothesis is $H_1: {\beta }_i<0$ for some i’s. CIPS test is (10) $$\widehat{\mathit{CIPS}}=\frac{1}{N}\sum _{i=1}^p{t}_i(N,T)\mathrm{ }$$(10) where $t_i(N,T)$ indicates t-statistic for ${\beta }_i.$

CS-ARDL

There are various econometric approaches to investigate the short- and long-run relationships between variables in the panel data. However, in the presence of CD, using first-generation methods like FMOLS, DOLS, etc., will produce bias and inconsistent regression results [39,52]. CS-ARDL developed by Chudik and Pesaran [38] is an updated version of the pooled mean group (PMG), and it addresses both CD and slope heterogeneity issues. CS-ARDL is robust in the presence of specification bias, potential endogeneity, autocorrelation, and non-stationarity issues and provides efficient and unbiased results [28,39]. The test statistics of CS-ARDL are specified as (11) $${\mathit{GDP}}_{it}={\alpha }_0+\sum _{j=1}^p{\theta }_{it}{\mathit{GDP}}_{i,t-j}+\sum _{j=1}^p{\vartheta }_{it}^{\prime }{X}_{i,t-j}+\sum _{j=1}^p{\theta }_{it}^{\prime }{\overline{Z}}_{t-j}+{\mu }_{it}\mathrm{ }$$(11) where ${\overline{Z}}_{t-j}={\left(\Delta {\overline{\mathit{GDP}}}_t,{\overline{X}}_t^{\prime }\right)}^{\prime }$ and $X_{it}={\left(L_{it},+{\mathit{GFCF}}_{it}+{EI}_{it}+{NR}_{it}+{EC}_{it}\right)}^{\prime }.$ GDP is the logged value of GDP, which is the dependent variable of our study, and X represents all independent variables of our study, that is, logged values of GFCF, L, NR, EINN, and EC. For robustness analysis, this study uses Common Correlation Effect Mean Group (CCEMG).

Results and discussion

CD, slope heterogeneity, and CIPS

The results of the Pesaran [53] CD test are shown in Table 4. The null hypothesis states the cross-sectional independence is rejected at 1% level of significance for all the variables, which shows that the stochastic term across the cross-sections is homoscedastic. Thus, the results confirm the presence of CD in GFCF, L, NR, EINN, and EC. After confirming the CSD, we apply the slope heterogeneity test of Pesaran and Yamagata (2008) to test slope homogeneity between the cross-sections of panel data. This test provides better results in the presence of CD than other homogeneity tests [39]. The findings show that the null hypothesis is rejected at a 1% level of significance. This heterogeneity in slope coefficients may be because of social, economic, institutional, and demographic differences between the panel of OECD countries. The results are presented in Table 5.

Table 4. Pesaran (2004) CD test.

Variable	CD-test	p-value	abs(corr)
EG	8.170***	0.000	0.604
GFCF	32.240***	0.000	0.833
L	4.990***	0.000	0.494
NR	29.080***	0.000	0.513
EINN	32.370***	0.000	0.413
EC	25.210***	0.000	0.456

Notes: Under the null hypothesis of cross-section independence CD ∼ N(0,1).

Table 5. Slope heterogeneity test of Pesaran, Yamagata (2008).

Statistics	Model	p-value
$\nabla$	22.011***	0.000
${\nabla }_{\mathit{adjusted}}$	25.748***	0.000

H0: slope coefficients are homogenous.

This evidence of the presence of CD and slope heterogeneity assists in the use of CS-ARDL for further analysis. The findings also support the use of the CIPS test by Pesaran [48]. CIPS controls the CD and heterogeneity of the panel variables, including testing for unit root [12]. The test results of CIPS are shown in Table 6. The results show that we fail to reject the null hypothesis of no stationarity for all variables except for EINN only. Therefore, we conclude that only EINN is I(0), whereas all other variables are I(1). The results are reported in Table 6. We can now proceed for CS-ARDL to study the short- and long-run effects of the K, L, NR, EINN, and EC on EG.

Table 6. Pesaran (2007) Panel Unit-Root Test CIPS.

Variable	CIPS (Z(t-bar)) statistics (at Level)	CIPS (Z(t-bar)) statistics (at 1^st difference)
EG	−1.797	−4.317***
GFCF	−1.656	−4.164***
L	−1.613	−4.161***
NR	−1.354	−3.917***
EINN	−2.859***	−5.052***
EC	−1.924	−4.880***

Note. Results are calculated by authors and *** indicate p < 0.01.

CS-ARDL

The results of the Pesaran [53] CD test confirm that CD exists in our panel data. To control the CD of panel data, we have employed the CS-ARDL technique, which includes added lagged cross-sectional averages to control the CD of the residual [54].

Short- and long-run results of CS-ARDL are reported in Table 7. ECT in both models explains the error correction mechanism as the coefficient of ECT is significantly negative in both models, −0.700 in model (1) and −0.958 in model (2). This high value of ECT in both models shows the speedy recovery or correction for any departure from the long-run equilibrium, indicating that any shock on EG is fixed between 70% and 95.8%.

Table 7. CS-ARDL results.

Variables	Model 1 (Dependent variable: EG)	Model 2 (Dependent variable: EG)
Short-run results
GFCF	0.061*** (0.018)	0.056*** (0.009)
L	0.967*** (0.217)	0.251*** (0.067)
NR	−0.166*** (0.046)	−0.034** (0.015)
EINN		0.469*** (0.125)
EC	0.683*** (0.204)	0.383*** (0.107)
Error Correction (ECM)	−0.700*** (0.087)	−0.958*** (0.238)
Long-run results
GFCF	0.574*** (0.139)	0.391*** (0.096)
L	6.077** (2.251)	0.958*** (0.281)
NR	−0.299*** (0.074)	−0.076* (0.032)
EINN		0.168*** (0.041)
EC	1.154*** (0.315)	0.467*** (0.125)
Number of groups	29	29
CD-Test	19.46	22.37

Note. Estimations are based on the error correction model. ***, **, and * indicate p < 0.01, p < 0.05 and p < 0.10. The numbers in parentheses show the standard errors. All estimation results are generated by Stata’s xtdcce2 command.

Results of model 1 show that EG increases with an increase in GFCF, L, and EC, while it decreases with NR. In particular, an increase of 1% in GFCF increases EG by 0.061% in the short run and 0.574% in the long run. An increase of similar magnitude in the labor force results in increasing EG by 0.967% and 6.077% in the short and long run, respectively. Meanwhile, a 1% change in EC predicts a 0.683% change in EG in the short run and 1.154% in the long run, in a similar direction, whereas, ceteris paribus, 1% growth in NR shrinks the EG by 0.166% in the short run and 0.299% in the long run. This verdict validates the RCH in the OCED countries. These results are aligned with RCH, based on the work of Sachs and Warner [24]. These outcomes are in line with other studies in literature, such as those found RCH in Iran; Haseeb, Kot, Hussain and Kamarudin [25] investigated RCH for Asian economies. At the same time, the study of Shahbaz, Ahmed, Tiwari and Jiao [47] validates the RCH in the USA and Li, Naqvi, Caglar and Chu [37] confirmed the RCH for N-11 countries.

All of the above studies have reached the same conclusion of the existence of RCH. Natural resources are considered one of the most potent sources of EG. Therefore, the focus of the resource-rich countries’ investment is usually the extraction of NR, which is a critical area of environmental economists and natural scientists [55]. In contrast, extraction and depletion of NR is an important cause of EC. According to the research of Hussain, Khan and Zhou [9], nearly 80% of world energy is used to extract fossil fuel, and these extracting industries are also responsible for about half of the world’s CO₂ emission and 90% of biodiversity loss. The depletion of NR has also increased more than three times in the past five decades, consuming massive energy sources. Thus, extraction of NR ultimately damages the environment because of intensive EC. It is expected that demand shift to protect the environment brings more eco-intensive practices that broaden the ecological pressure by placing more burden on energy-intensive methods, which delays the view of sustainable development. It is argued that dependence on NR consumption persists the “RC” because it hampers the sustainability of a country by impeding its growth [56]. Thus, it is stated that countries with abundant NR experience relatively low growth in comparison with countries with scarce NR. Green energy resources have delivered an ultimate solution to meet energy demand with minimum burden on the environment; however, the installation process of green ventures is also energy-dependent [57]. Moreover, some green energy resources, such as biomass, waste, wood, and plants, also increase the burden on NR and their sustainable consumption [56]. Thus, the current pattern of EC is a major cause of RCH.

In model 2, we have added the EINN variable, which shows a positive relationship with EG. The nature of the relationship of other variables remains the same; however, there is some change in the magnitude. Specifically, with a 1% change in GFCF, EG changes by 0.0561% in the short run and by 0.391% in the long run in a similar direction. Dell’Anno [17] found similar effects for Iran and examined related effects for OPEC countries. The EG in OECD countries is a capital-driven growth [58]. For so long, physical capital has been considered a primary driver of EG [12]. Capital development helps in developing large manufacturing markets and enhances specialization. It also contributes to the consumption of NR and the growth of the industry that contribute to economic expansion [27]. It is also helpful in relieving the “RC” effects to some extent. The findings of our study also conclude the significant role of capital in promoting EG in OECD countries.

At the same time, 1% increase in L results in increasing EG by 0.251% in the short run and 0.958% in the long run. These findings are consistent with the study of Shahbaz, Ahmed, Tiwari and Jiao [41] for the USA. Young [59] found the same outcomes in Nigeria. These findings also endorsed the study of Peterson [60]. A necessary quantity of labor is required for the extraction of NR. If the country lacks the needed working population to a threshold level, it cannot enjoy the benefits extracted from NR. Currently, Canada is facing less population; thus, Canada has relaxed the migration policy to attract more human resources from the rest of the world [3]. Our results also conclude the positive effect of “L” on EG.

Similarly, 1% increase in EC results in EG by 0.383% in the short run and 0.467% in the long run, if there is no change in any other factor. The study results endorse the findings of Wang and Chen [1]; they found similar verdicts for China. Bekun, Emir and Sarkodie [61] also stated the positive effect of EC on EG. In comparison, Le and Tran-Nam [62] examined the same results for the Asian Pacific countries.

We have added the EINN variable in the model to show its impact on RCH. The model predicts a 1% increase in EINN improves EG by 0.469% in the short run and 0.168% in the long run. These outcomes are significant at a 1% level of confidence. It shows that increases or improvements in EINN technologies significantly positively affect the EG of OECD countries. The NR effect on EG is still negative, but this time the coefficient of NR is −0.034 in the short run and −0.076 in the long run, signifying that the absolute value of the coefficient has declined significantly from −0.166 to −0.034 in the short run and from −0.299 to −0.076 in the long run. These findings indicate that EINN has mitigated the “resource curse” effect substantially in the OECD countries. These outcomes are consistent with those of Popp [63], who claimed that innovations in energy are helpful to achieve greener growth. The findings are also aligned with the study of Henderson and Newell [64]; they support the EINN technologies to boost green growth with low carbon emissions. Moreover, our findings are also coherent with the findings of IEA [16], which explains that investment in EINN is essential to achieve SDG.

Energy is a crucial input in the production process. A smooth and uninterrupted supply of cheap energy is essential for rapid EG [65]. In this regard, many financial and human resources are being devoted to energy technology for bringing innovations that increase energy performance, reduce CO₂ emissions, and make it more cost-effective. Progress made on these lines results in cutting the cost of production and increasing the competitiveness of various products in both national and international markets. EINN results in the expansion of economic activities and speeding up economic growth. The traditional use of energy resources causes an RCH, as discussed above, because of increased economic, social, environmental, compliance, and psychic costs [56]. In contrast, innovations in energy because of dedicated efforts result in improving energy efficiency and promoting green energy and bringing down the economic and environmental cost of production [66]. Therefore, the adverse effects of extensive extraction and consumption of NR are mitigated to some extent by EINN [16].

Further, to confirm the strength of our findings, we have performed a robustness check. To meet this purpose, we have applied the Common Correction Effect-Mean Group (CCEMG) and Augmented Mean Group (AMG). The results are reported in Table 8, which confirm the nature and direction of the relationship presented in Table 7. So, our results are robust and consistent.

Table 8. Robustness of Results from CCEMG and (AMG).

	CCEMG		AMG
VARIABLES	Model 1	Model 2	Model 1	Model 2
GFCF	0.087* (0.049)	0.355*** (0.073)	0.220*** (0.0194)	0.320*** (0.0612)
L	0.099*** (0.032)	0.064* (0.029)	1.293* (0.748)	1.778** (0.791)
NR	−0.114*** (0.037)	−0.015*** (0.001)	−0.0452* (0.0245)	−0.0195* (0.0114)
EINN		0.024* (0.012)		0.0120** (0.0056)
EC	2.945* (1.674)	0.191* (0.105)	0.560*** (0.205)	1.802*** (0.249)
Number of groups	29	29	29	29

Note. ***, **, and *indicate the level of significance at 1%, 5%, and 10%, respectively. The numbers in parentheses show the standard errors.

Conclusion and policy recommendations

EINN is of great importance in reducing the economic and ecological effects of the extraction and consumption of NR. However, there is scarce empirical support on the growth effect of EINN in the systematic framework of RCH. This research investigates the effect of EC, NR, and EINN on the EG of OECD countries from 1990 to 2015. From methodological aspect, this study has used the CS-ARDL approach by Chudik and Pesaran [48], which controls the CD and slope heterogeneity issues of panel data. The study has used the second-generation econometric approaches of Pesaran [102] to test the CD of panel data; the Pesaran and Yamagata [103] test is used to check the slope heterogeneity of cross-sectional units. Moreover, Pesaran [104] CIPS is used to check the unit root properties of the panel variables. The robustness of the results is confirmed using the CCEMG method by Chudik and Pesaran [105].

The results provide sufficient evidence of the existence of CD and slope heterogeneity in the selected variables. These findings validate the use of the CS-ARDL method to test the association between GFCF, L, EC, NR, EINN, and EG. The findings clearly explain that EINN is a crucial factor in the study of RCH between NR and EG in the OECD countries. The positive association between EINN and EG implies that EINN is useful in improving the EG of OECD countries. Results indicate that a 1% rise in EINN expands the EG by 0.469% in the short and 0.168% in the long run, and these results are significant at 1% level of confidence. The results reveal the negative impact of NR on EG; however, with the inclusion of EINN in model 2, the absolute value of the coefficient of NR has declined significantly from −0.166 to −0.0344 in the short run and from −0.299 to −0.0760 in the long run. These results indicate that EINN has significantly mitigated the “resource curse” effect in the OECD countries. Moreover, the findings show that EC affects EG positively and significantly; a 1% surge in EC improves EG by 0.383% and 0.467% in the short and long run, respectively, while a rise in L and GFCF also increases the EG of OECD countries in both short and long run.

On the basis of empirical findings, the study offers significant policy implications. Since the results provide the existence of RCH in the OECD nations, so they should devote considerable efforts at the domestic and international levels to improve the NR productivity and efficiency to control the wastage of resources. These countries should monitor the improvement regarding resource efficiency and their sustainable consumption to improve the information related to ecological effects. About two-thirds of OECD countries have already started making efforts on the reporting of NR extraction and use. The results explain that EINN helps mitigate the RCH. Thus, OECD countries should encourage energy innovation policies by promoting their application, so that replacement of traditional energy sources with new energy can be realized. Moreover, environmental and innovations policies should be more rigorous and focused on long-run research and development (R&D) to encourage important innovations in energy sector. Both public and private sector investments in energy technology innovations are crucial to highlight the gaps and opportunities to enhance the efficiency of resources allocation. Both public as well as private sectors should work hand in hand to enhance investment in energy innovations as government-funded research provides in-depth knowledge, while the private sector plays an essential role in bringing innovations. Moreover, besides national efforts to promote innovations in energy, international cooperation is also needed to accelerate progress in EINN. OECD countries’ governments can work in collaboration to promote innovations in energy technology.

This research has some limitations because it is the earliest study done in the area of energy innovations. It may not be a broader study, hence more comprehensive measures of “EINN” can be used in future research for further analysis. Furthermore, due to data limitations, this study has concentrated on advanced countries such as the OECD. As a result, this research can be expanded to include empirical studies of developing countries. Besides, the current study used the novel CS-ARDL method to examine the relationship between EIN and RCH. However, there are additional empirical techniques that could be used in future studies to empirically examine the specific impact of energy innovation on the natural resource curse.

Abbreviations
CD	=	Cross-section dependence
EINN	=	Energy innovation
GFCF	=	Gross fixed capital formation
L	=	Labor force
RCH	=	Resource curse hypothesis
EC	=	Energy consumption
EG	=	Economic growth
GDP	=	Gross domestic product
NR	=	Natural resources
RD&D	=	Research development and Demonstration

Data availability statement

Data will be available on request.

Disclosure statement

No potential conflict of interest was reported by the authors.