Economic Complexity and Renewable Energy Effects on Carbon Dioxide Emissions: a Panel Data Analysis of Middle East and North Africa Countries

Abstract

With the rising economic complexity of the world economy, applied research has been focusing on the economic complexity-environmental quality nexus. Many of the Middle East and North Africa (MENA) countries and Turkey show relatively low economic complexity with undiversified product spaces. The present study is thereby motivated to investigate the economic complexity-environmental quality relation for the MENA countries and Turkey. Using panel data for eighteen MENA countries and Turkey over the period 1990–2020, the fully modified ordinary least squares method of estimation is employed in two separate models, and Granger causality tests are performed. Each model includes the relation of economic complexity to carbon dioxide (CO₂) emissions, one form of energy (renewable or non-renewable), population density and variables to validate the environmental Kuznets curve (EKC) hypothesis. In model (1), a negative and statistically significant renewable energy consumption has a mitigating effect on CO₂ emissions, though economic complexity and its interaction with renewable energy consumption are not statistically significant. In model (2), a positive and statistically significant non-renewable energy indicates that increased fossil fuel reliance increases CO₂ emissions (deteriorating environmental quality), while a negative and statistically significant economic complexity and population density reduce CO₂ emissions (improving environmental quality), respectively. The EKC hypothesis is validated for the MENA countries and Turkey in both models (1) and (2), and Granger causality test results are given. Among the key policy implications are that the MENA countries and Turkey must recognize that increasing GDP and industrialization must be coupled with moving from less- to more-knowledge-based structures.

Introduction

Concern over economic complexity as a determinant of energy demand and also as a determinant of economic growth and of environmental quality has been recently rising. On the one hand, economic complexity is defined as a change in the product characteristics of an economic system where it moves from simple, lower knowledge-content products to more knowledge-based, skill-based, and sophisticated products. An increasing economic complexity index (ECI) (with details of its calculation given below) would reveal a structural transformation from relatively simple low-technological-content to more sophisticated high-technological-content products and from high to low energy-intensive industries. It is also expected to be associated with higher levels of productivity, as well as an increased use of more sophisticated factors of production. On the other hand, the degraded quality of the environment is demonstrated by various indicators: rising carbon dioxide (CO₂) emissions; heightened greenhouse gas, methane and nitrous oxide emissions; greater exposure to particles in the PM_2.5 size range; and rising ecological footprint (EF) with more land areas needed to support the forest, pasture, farmland and marine resources consumed.

Indeed, rising CO₂ emissions are highly relevant to Middle East and North Africa (MENA) countries and Turkey. Some of the heavy-oil-producing countries (e.g. Qatar and Kuwait) rank high on the CO₂ emissions levels for the region as a whole in 2018 (Source: DataBank (World Bank)). The need to study the effect of economic complexity and explore the effect of the productive structure of countries as a source from which CO₂ emissions originate (Doğan et al., 2020) remains highly pertinent to these countries. Also of relevance to them is the implication of economic complexity and the increased use of renewable energy for their achievement of the Sustainable Development Goals, with emphasis on goals 7, 9, and 12 (affordable and clean energy, industry innovation and infrastructure, responsible consumption and production, respectively).

The economic complexity-CO₂ emissions nexus has been very sparsely studied for the MENA countries and Turkey. The present study thus aims to make a contribution to applied research with relevance to these countries on several levels, providing stylized facts about the countries’ economic complexity and how it has evolved, as well as individual MENA countries and Turkey’s pattern of CO₂ emissions over time, testing a model that captures the effect of economic complexity, renewable and non-renewable energy consumption on CO₂ emissions and validating the environmental Kuznets curve (EKC) hypothesis for the MENA countries and Turkey, as well as testing for the causal relations between the variables. This is believed not to have been previously performed for the MENA countries and Turkey.

Following the introduction, the study is organized as follows: the ‘Stylized Facts of Economic Complexity and Environmental Quality in the MENA Countries and Turkey’ section gives some stylized facts of economic complexity and environmental quality for the MENA countries and Turkey; the ‘Literature Review’ section gives a review of theoretical and applied literature; the ‘Data and Methodology’ section outlines data and variable description and presents the model specification; the ‘Results and Discussion’ section gives results and discussion; and the ‘Conclusion and Policy Implications’ section outlines some policy recommendations based on the research findings.

Stylized Facts of Economic Complexity and Environmental Quality in the MENA Countries and Turkey

Giving key stylized facts of economic complexity and environmental quality in the MENA countries and Turkey, we start with the average economic complexity index for the period 1990–2020, with the MENA countries and Turkey arranged in descending order of complexity, and with indication of whether a country’s economic complexity has improved/deteriorated over the period of study (Table 1).

Table 1 Economic complexity index (ECI) for MENA countries and Turkey arranged in descending order, 1990–2020 period average

The MENA countries and Turkey’s ECI have not merely improved or deteriorated over the period of study but have shown some pattern of evolution as portrayed in Fig. 1 with countries arranged in descending 1990–2020 period average ECI and their ECIs at 10-year intervals:

Fig. 1

Source: Author’s computation based on MIT Observatory of Economic Complexity (OEC)

ECI by MENA country and Turkey for the years 1990, 2000, 2010 and 2020. Notes: MENA countries and Turkey arranged in descending 1990–2020 period average ECI.

As evident from Fig. 1, the top-ranked MENA countries and Turkey by ECI are as follows: Turkey, Lebanon, Saudi Arabia and Jordan with positive values of ECI, thereby reflecting these countries’ relatively diversified product space. With reference to ECI absolute value and sign, the absolute value of a country’s ECI measures the distance of that country to the boundary between two clusters (positive and negative value ECI clusters). The ECI’s absolute value would thus reveal the extent to which a country deviates from the respective boundary, and +/− sign indicating being lower than/higher than the boundary ECI value.¹ In this regard, and as indicated in Fig. 1, the U.A.E. and Kuwait have turned from negative to positive ECI values in the year 2020, marking progress towards increased complexity. The remaining countries have all fallen in the negative ECI value range, albeit with Tunisia, Bahrain, Egypt, Oman, Iran and Algeria showing ECI values signaling improvement to lower negative ECI values over time. This may mark a relative increase in their economic complexity. Iraq has the lowest negative ECI values for the four reported years, though with the year 2020 marking an improvement.

With respect to CO₂ emissions, Fig. 2 gives the average CO₂ emissions per capita for the world and for the MENA countries and Turkey over the period spanning 1990 to 2019. It conveys a clear picture of the MENA countries’ CO₂ emissions exceeding those of the world in all the years of study. Moreover, the gap between the average per capita MENA countries and Turkey CO₂ emissions as a percent of the world average CO₂ emissions is as high as 178% in 2002 when it is at its peak, going down to 106% and 110% in 2018 and 2019, respectively. Despite dropping by the end of the study period, the gap remains substantial.

Fig. 2

Source: Author’s computation based on World Bank World Development Indicators data

World average CO₂ emissions per capita, MENA countries average CO₂ emissions per capita, and gap (% of world average), 1990–2019. Notes: 2020 was excluded for lack of world CO₂ emissions per capita data in DataBank, World Development Indicators of World Bank.

A closer examination of the period average CO₂ emissions per capita by country is presented in Fig. 3. Qatar is seen to have the highest period average CO₂ emissions per capita, followed by U.A.E., Kuwait, Bahrain and Saudi Arabia. The data may thereby point to the high-ranking emissions stemming from those countries with heavy reliance on oil production. In addition, these countries have relatively lower populations compared to the rest of the MENA countries, thereby contributing to the high CO₂ per capita data. Yemen has the lowest CO₂ emissions per capita probably stemming from low production levels under the country’s current political and economic conflicts.

Fig. 3

Source: Author’s computation based on World Bank World Development Indicators data

Average CO₂ emissions per capita for MENA countries and Turkey, 1990–2020.

Having examined the MENA countries and Turkey stylized facts of economic complexity and environmental quality, one detects an economic complexity that leans heavily towards the negative range and with modest improvements in most of the countries. There is also an overall deterioration of the environmental quality of the MENA countries and Turkey. This calls to question the extent to which economic complexity is a determining factor in the environmental quality of these countries.

Literature Review

Based on the concept of economic complexity developed by Hausmann and Hidalgo (2011), Can and Gozgor (2017) define economic complexity as an ‘an indicator which demonstrates the product characteristics of an economic system by taking into consideration the knowledge-based, skill-based, and sophisticated production of a country’ (Can & Gozgor, 2017:16365). Romero and Gramkow (2021) refer to a more sophisticated production of a country as comprising the information required for employing a clean manufacturing technology. Economic complexity is measured by the economic complexity index (ECI) which is constructed in a manner that captures the economic structure of a country and the level of technical knowledge of its product manufacturing. Accordingly, a rise in the value of a country’s index reveals structural transformation from relatively simple low-technological-content products to more sophisticated high-technological-content ones. The rise may also reveal transforming from energy-intensive industries (e.g. textiles) to industries of lower energy intensity and higher knowledge content (e.g. semiconductor chips). The rise is also expected to be associated with higher levels of productivity and a deeper capability of the country to manufacture sophisticated products, as well as increased use of more sophisticated factors of production. With transformation underway, the country’s human capital, legal system, institutional system (such as property rights) and infrastructure will further come to assume bigger roles in its economic system.

The ECI is calculated from a matrix that records the ‘export product categories’ of world countries’ tradable products. Row elements of the matrix give data by country, and column elements give data by categories of products. However, data are not recorded in absolute export values but in the revealed comparative advantage (RCA) of each country in each product category benchmarked against a ‘predetermined’ threshold level. Thus, matrix elements for a product category take the value one if the country has an RCA exceeding a 0.5 threshold level, zero otherwise. Thereby, the sum over column gives the ubiquity of a product category as it quantifies the number of countries which possess an RCA above the threshold level in that category. The sum over row quantifies the number of product categories in which a country has an RCA above the threshold level, hence measuring the respective country’s degree of product diversity. The ECI is obtained through a solution of the system of equations of the average ubiquity of the products of a country and the average diversity of its products (for a detailed discussion of the ECI calculation, see Can et al., 2022; Can & Gozgor, 2017; Hausmann & Hidalgo, 2011). The higher the ECI, the greater the level of its economic complexity evidenced by the export of higher value-added, higher technological content product categories. However, it is important to note that such products would be characterized by being ‘less’, as opposed to ‘more’, ubiquitous in the sense that they are exported by a more limited number of countries possessing the knowledge and skill capabilities required for their production (Can & Gozgor, 2017).

Economic complexity was described as a ‘useful metric for predicting economic growth, geographic differences in income inequality, gender inequality, human development, output volatility, productivity, health and greenhouse gas emissions’ (Hidalgo, 2021:92-94). The effect of economic complexity on environmental degradation comes via the effect of economic growth on degradation. The ecological modernization theory proposes that as countries experience earlier stages of development, they tend to rely on energy-intensive low-technology products and will tend to favour growth over environmental sustainability, hence experiencing deteriorating environmental quality (manifested in rising CO₂ emissions, among other indicators). As they subsequently grow, and with structural transformation towards high-technology products and a service-based economy (Sadorsky, 2014), not only would countries’ labour and capital productivities be expected to grow but also their environmental productivity. The latter would be evident in a more efficient use of resources (of which is energy), in the introduction of product and process innovations, and in the use of clean technologies and clean sources of energy (Boleti et al., 2021). In short, the process of product complexity, diversification, and structural transformation has an improving effect on environmental quality.

However, the above trajectory is not standard for all countries. In some cases, raising economic growth and economic complexity may be associated with deteriorating environmental quality even in the later stages of development. This may be explained by economic growth affecting the environment via three key effects: scale, composition and technique. Under the scale effect, and for a given state of technology, a large scale of production, a more extensive use of industrial resources (particularly energy) and depletion of natural resources would raise CO₂ emissions (Dinda, 2004; Martins et al., 2021; Tsurumi and Managi, 2010). Under the composition effect, the structural transformation from low- to high-technology and information-and-knowledge-based industries and service activities, as well as increased demand for health, education, energy security, and energy efficiency, and rising overall environmental sensitivity, CO₂ emissions may be reduced (Boleti et al., 2021; Cialani, 2007; Martins et al., 2021; Zafar et al., 2020). Furthermore, the technique effect is such that countries adopting new and advanced/enhanced technology, environmentally friendly production techniques, as well as rising efficiency, will find their CO₂ emissions reduced. However, if the scale effect outweighs the composition and technique effects, higher emissions may result even in later stages of development (Neagu and Teodoru, 2019).

Economic complexity is further associated with energy intensity. The nature of this association is explained by hypotheses of ‘neutrality’, ‘growth’, ‘conservation’ and ‘feedback’ (Apergis & Payne, 2011). The ‘neutrality’ hypothesis proposes no causal relationship between economic complexity and energy consumption, while the ‘growth’ hypothesis proposes that the relation runs from energy consumption in the direction of economic complexity, implying that the gradual increase in the use of renewable energy raises economic complexity. A causal relationship running from economic complexity to energy consumption is proposed by the ‘conservation’ hypothesis where economic complexity may increase, but not in support of greater renewable energy use but rather of fossil fuel energy use. The ‘feedback’ hypothesis implies the existence of a causal relation between economic complexity and energy consumption running in both directions, suggesting a feedback effect (Doğan, 2015).

Empirically, the relation of economic complexity to environmental degradation (deteriorating environmental quality) has been studied by many authors. Perhaps a seminal work is that of Can and Gozgor (2017) who studied the effect of ECI and CO₂ emissions, using time series data for France over the period 1964–2016. Employing the dynamic ordinary least squares method of estimation, the authors find ECI to have a negative and statistically significant effect on CO₂ emissions. These results are corroborated by Doğan et al. (2020) in a study for a panel of 28 OECD countries over the period 1990–2014. Employing the fully modified ordinary least squares method of estimation, the authors conclude that CO₂ emissions are reduced under higher economic complexity. Leitao et al. (2021) with application to the BRICS economies over the period 1990–2015 and using the fully modified ordinary least squares method of estimation, find ECI to also have a negative and statistically significant effect on CO₂ emissions.

With respect to other factors affecting CO₂ emissions, Can and Gozgor (2017) find energy consumption to have a positive (worsening) effect on them, while Doğan et al. (2020) and Leitao et al. (2021) find renewable energy consumption to have a negative (lessening) effect.

Other key empirical works have arrived at conclusions which have differed by ‘category/group’ of countries. Doğan et al. (2019) for a panel of 55 world countries over the period 1971–2014 use the panel quantile regression fixed effects method of estimation and find ECI to have a positive (worsening) and statistically significant effect on CO₂ emissions for lower-middle and higher-middle income countries. However, the effect is |negative (lessening) and statistically significant for high-income countries. In this regard, the high-income countries are believed to be gradually diversifying into the production of environmentally friendly goods and services and importing the goods and services that increase pollution (Doğan et al., 2019:31904). Neagu and Teodoru (2019) use panel data for 25 countries of the European Union over the period 1995–2016, under a heterogeneous panel technique of estimation. The authors find that in the subpanel of countries with lower levels of complexity and higher energy consumption (e.g. Poland, Romania and Bulgaria), there is a high and statistically significant positive effect of economic complexity on environmental degradation (deteriorating environmental quality), indicating the dominance of the scale effect in this subpanel. That effect is lessened in the subpanel of countries with higher levels of economic complexity, lower energy consumption (with increased use of renewable energy), and the dominance of the composition and the technique effects (e.g. Ireland, Austria and the Netherlands).

Studying the long-run relation of economic complexity to eco-efficiency, Chen et al. (2022) employed a cross-sectional autoregressive distributed lag method of estimation using a panel of the world’s highest ten polluted economies over the period 1990 to 2019. With co-efficiency encompassing economic, resource, environmental, and social dimensions and defined as the ability to create more products or services with fewer resources and minor environmental damage, the authors find the squared term of economic complexity to positively affect eco-efficiency (following EKC hypothesis in its effect).

Other studies have found evidence of a positive association between ECI and environmental degradation (deteriorated environmental quality). Khezri et al. (2022) use panel data for 29 Asia Pacific countries over the period 2000–2018 and apply the panel fully modified ordinary least squares method of estimation. The authors find the ECI to have a positive and statistically significant effect on CO₂ emissions. Their findings also validate the EKC hypothesis for the panel countries, and the higher ECI further shifts the EKC upwards. This result implies that increasing economic complexity enhances the scale effect and hence increases CO₂ emissions. However, when interacted with the energy intensity, ECI and energy intensity variable is found to have a negative and statistically significant effect on CO₂ emissions implying that higher complexity will mitigate the positive effect of energy intensity on emissions (which was evident for the latter variable alone having a positive and statistically significant effect).

Peng et al. (2022) use a panel of the BRICS countries (Brazil, Russia, India, China and South Africa—which together produce 41% of world CO₂ emissions) over the period 1984–2017 and employ the fully modified ordinary least squares method of estimation to study the effect of economic complexity and energy consumption on CO₂ emissions. Their findings indicate a positive effect of growth and economic complexity on emissions, implying that complexity intensifies/worsens CO₂ emissions. Meanwhile, ECI squared has a negative effect on emissions, thereby validating the EKC hypothesis of the panel countries. Similarly, Doğan et al. (2022) in a panel study of the G7 countries over the period 1994–2014 validate EKC hypothesis for the G7 countries, explaining that as the countries of the G7 initially transitioned from agricultural to industrial-based economies their environmental quality deteriorated. However, beyond a certain threshold level, with higher complexity and diversification, CO₂ emissions are reduced.

Similarly, Martins et al. (2021) for a panel of the top seven high ECI countries over the period 1993–2018, employing the autoregressive distributed lag method of estimation, find ECI to have a positive and statistically significant effect on carbon emissions. The study further verifies the existence of the EKC hypothesis for GDP and the square of GDP. The findings thereby indicate that the high ECI countries have favoured growth over environmental degradation, and that the scale effect has outweighed the composition and technique effect in these countries.

Other studies have investigated the relation of ECI to ecological footprint (EF), where EF measures human pressure on the widespread use of natural resources and is calculated through global hectares of water and land required to waste absorption and goods production. Moreover, it measures ocean, grazing land, forest products, croplands, carbon footprint and built-up land (Rafei et al., 2022:102852). Rafei et al. (2022), using a panel vector autoregressive model, studied the relation of ECI, natural resources, renewable energy consumption and foreign direct investment to the EF for a panel of weak, medium and high levels of institutional-quality countries over the period 1995 to 2017. The authors found ECI to positively affect EF (deteriorating environmental quality) in all countries with different institutional-quality levels. Rafique et al. (2022) arrive at a similar conclusion while employing the fully modified ordinary least squares method of estimation for a panel of the top 10 ECI countries over the period 1980–2017. The authors find ECI to have a positive and statistically significant effect on EF, while the consumption of renewable energy has a negative and statistically significant effect on it. Both results of Rafei et al. (2022) and Rafique et al. (2022) indicate the scale effect at work. Contrary to both works, Ahmed et al. (2022), in a stochastic impact regression on population, affluence and technology (STIRPAT) framework, use an autoregressive distributed lag model (ARDL) on time series for India over the period 1970–2017 and found ECI to have a negative (lessening) and statistically significant effect on EF, while economic growth has a positive effect on it. When interacting ECI with economic growth, ECI mitigates/moderates the positive effect of growth on EF.

With application to the MENA countries, an empirical study of Yalta and Yalta (2021) has explored the determinants of economic complexity for the MENA countries, albeit with no focus on the ECI-emissions nexus. The authors employed a generalized method of moments approach, using data for a panel of twelve countries over the period 1970–2015, finding evidence that primary and secondary education enhances economic complexity. They further find evidence that rent from natural resources (given the high natural resource intensity of many of the panel countries, especially Gulf Cooperation Council Countries) negatively affects economic complexity, while foreign direct investment and terms of trade have no significant effect on it.

The theoretical and empirical reviews depict relatively strong evidence that the relationship between economic complexity and environmental degradation has been quite inconclusive. There is also very sparse exploration of that relation for the MENA countries and Turkey. That directs attention to a much-needed study of that nexus, and the present study aims to fill a gap in the empirical literature in this regard.

Data and Methodology

Data

The CO₂ emissions per capita equation is estimated for the MENA countries and Turkey (excluding Djibouti, Malta, West Bank and Gaza).² A total of eighteen countries are included in the dataset over the period 1990–2020. For those countries that were excluded, data are unavailable for various key variables, of which are as follows: CO₂ emissions per capita and the economic complexity index. We note, however, that several countries had missing observations for several variables. Missing observations were, thereby, linearly interpolated so as to complete a balanced dataset in the interest of obtaining more efficient parameter estimates. All data, except for the ECI, are drawn from the World Bank Open Data (World Development Indicators, several years). ECI data are drawn from the MIT Observatory of Economic Complexity (OEC).

The dependent variable is CO₂ emissions per capita measured as the emissions stemming from the burning of fossil fuels and the manufacture of cement, in metric tonnes. These emissions include CO₂ produced during the consumption of solid, liquid, and gas fuels and gas flaring.

The following are the independent variables used in the estimated equations:

• 1. The index of economic complexity is calculated as explained in the ‘Literature Review’ section above. A higher level of economic complexity is expected to have a negative relationship with CO₂ emissions per capita if product complexity, diversification and structural transformation have improving effects on environmental quality. However, the said relationship may be positive (has a deteriorating effect on environmental quality) if rising economic growth and economic complexity are associated with a more extensive use of industrial resources (particularly energy) and the depletion of natural resources, thereby raising CO₂ emissions. Also, if the scale effect of growth overpowers the composition and technique effects, the effect on CO₂ emissions is expected to be an aggravating one.

• 2. Gross domestic product (GDP) per capita (GDP/capita) is defined as GDP in constant 2010 US dollars divided by midyear population. GDP is measured as the sum of gross value added by all resident producers in the economy plus indirect taxes minus subsidies not included in the value of the products. GDP per capita is used to reflect the state of economic development, and its relation to CO₂ emissions per capita may either be positive or negative. İf growth is associated with more energy-intensive processes leaning towards greater fossil fuel consumption at the expense of renewable energy, the effect on CO₂ emissions is expected to be positive (deteriorating environmental quality). İf, however, renewable energy use prevails with increased growth, the effect on carbon emissions is expected to be a negative one (thereby, improving environment qualtiy). This would be in line with the proposition of the Ecological Modernization Theory whereby higher modernization levels associated with growth help shift communities’ rationality from being economic-based to being sustainability-based.

The square of GDP per capita is used to test for the EKC hypothesis. A positive value of the coefficient of GDP per capita and a negative value of the coefficient of the square of GDP per capita would indicate that countries’ growth is marked by higher CO₂ emissions and environmental quality deterioration in early stages of development and with lower CO₂ emissions and environmental quality improvement in later stages.

• 3. Population density is measured as midyear population divided by land area in square kilometres. Population includes all residents regardless of their legal status or citizenship, within the physical boundaries of a country and under the jurisdiction of that country’s political control. The relationship of population density to CO₂ emissions is expected to be negative if growth and the scale effect are outweighed by an advancement in technology and an efficiency in the use of public services such as waste disposal, sanitation, heating, rail and bus-based collective transportation, among other public services (Ahmed et al., 2019).

• 4. Consumption of renewable energy is defined as the share of renewable energy in total final energy consumption. The expected relationship of consumption of renewable energy to CO₂ emissions is negative (having the much-expected and desired improvement of environmental quality).

• 5. Consumption of fossil fuels is defined as the share of fossil fuels in total final energy consumption, and fossil fuels comprise coal, oil, petroleum and natural gas products. The greater the share of fossil fuels in energy consumption, the greater the expected deterioration of environmental quality, hence a positive effect on CO₂ emissions.

• 6. ECI*renewable energy consumption is defined as the interaction of ECI with renewable energy consumption. The expected relationship of this variable to CO₂ emissions is negative, with indication that renewable energy consumption would reduce CO₂ emissions further in countries with higher economic complexity.

• 7. ECI*fossil fuel energy consumption is defined as the interaction of economic complexity with fossil fuel energy consumption. The expected relationship of this variable to CO₂ emissions is negative, with indication that higher economic complexity would mitigate the positive effect of greater fossil fuel consumption on CO₂ emissions.

The descriptive statistics for the dependent and independent variables are given in Table 2.

Table 2 Descriptive statistics

The CO₂ emission equation is estimated in logs in order to minimize the variation in all variable observation values as evident from the standard deviation (particularly high for GDP per capita and population density). Estimation is performed using Eviews statistical software.

Methodology

To test the effect of economic complexity, GDP per capita, and use of renewable and fossil fuel energy and population density on CO₂ emissions, the following models are estimated:

$$\log\;{CO}_{2it}=\alpha+\beta_1\;{ECI}_{it}+\beta_{2\;}\log\;{GDP}_{it}+\beta_{3\;}\log{\;{GDP}^2}_{it}+\beta_4\;\log\;{RECons}_{it}+\beta_5\;\log{\;Pop}_{it}\beta_6\;({ECI}_{it}\;x\log{RECons}_{it})+\varepsilon_{it}$$

(1)

$$\log\;{CO}_{2it}=\alpha+\beta_1\;{ECI}_{it}+\beta_2\;\log\;{GDP}_{it}+\beta_3\;\log\;{GDP}_{2it}+\beta_4\;\log\;{FFCons}_{it}+\beta_5\;\log\;{Pop}_{it}\beta_6\;({ECI}_{it}\;x\log{FFCons}_{it})+\varepsilon_{it}$$

(2)

where $i=\mathrm{1,2},..,18,t=\mathrm{1,2},\dots ,30.$

However, estimations of models (1) and (2) are preceded by key tests, the first of which is testing for the existence of cross-sectional dependence in the panel using the Peasaran cross-sectional dependence test. Rejection of the null hypothesis of cross-sectional independence would indicate cross-sectional dependence among panels. The Breusch-Pagan LM test for cross-sectional independence is also performed as a robustness check. Subsequently, it is key to test for non-stationarity (unit root) of cross-sectionally dependent panels. The latter test is performed for all variables using the cross-sectionally augmented Im, Peasarn, Shin unit root test, where H_o, all panels have unit root, and H₁, some panels are stationary.

Upon establishing the non-stationarity of the variables (or some of which), it would be key to verify if a cointegrating relation exists for the non-stationary panel series. This is done by performing the Pedroni panel cointegration test, where H_o, no cointegration in panel series, and H₁, cointegrated panel series. Subsequently, the equations are estimated as cointegrated regressions using the fully modified ordinary least squares (FMOLS) method of estimation. Among the advantages of the FMOLS, which uses a nonparametric approach to tackle endogeneity of the regressors and autocorrelation of the error term, is that it provides optimal estimates of cointegrating regressions. Philipps (1995) further notes that the endogeneity of the regressors is itself a result of the existence of a cointegrating relationship (Philipps, 1995:1). Hence, FMOLS would lead to reliable parameter estimates in a cointegrating regression (upon identification of the existence of a cointegrating relation). Also, Bashir and Siam (2014) point to the reliability of panel FMOLS for small sample sizes, this being the case in the estimated equation with n < t. Finally, the Dmitrescu-Hurlin panel causality test is employed to verify the existence and direction of panel causality, where H₀, no Granger causality, and H₁, Granger causality.

Results and Discussion

This section gives the results of the Peasaran test for cross-sectional dependence (and Breusch-Pagan LM test), the test for the stationarity of the cross-sectionally dependent panel series, the test for the existence of a cointegrating relation, and the long-run estimated equation of CO₂ emissions in the MENA region and Turkey.

Table 3 results point to rejecting the null hypothesis of cross-sectional independence, thereby indicating cross-sectional dependence, which further implies proceeding to perform a stationarity test for panel, in level and first difference, with test results presented in Table 4.

Table 3 Results of tests for cross-sectional independence

Table 4 Results of test for panel unit root (level and first difference)

Im-Peasaran-Shin test for panel unit root indicates the stationarity (I(0)) of CO₂ emissions per capita, renewable energy consumption and economic complexity at 1% and 5% levels of significance, respectively. Otherwise, GDP/capita, fossil fuel energy consumption and population density are non-stationary in level and stationary in first difference (I(1)).

With the establishment of a panel unit root process for some of the variables and stationarity of other variables, the subsequent test is of panel cointegration, with results given in Table 5.

Table 5 Result of Pedroni panel cointegration test

It is thus evident from Table 5 that there exists a cointegration relation between panel variables. We thereby proceed to examine the long run relation using the panel FMOLS method of estimation. In model (1) the effect of renewable energy consumption on CO₂ emissions is tested, while in model (2) the effect of fossil fuel energy consumption is tested, respectively. Estimation results are given in Table 6.

Table 6 Results of Panel FMOLS estimation

The empirical results show that per model (1), the coefficient of GDP/capita (elasticity of CO₂ per capita with respect to growth/development level of panel countries), is positive and statistically significant, with a 1% GDP growth yielding a 1.7% increase in CO₂ per capita level. The results further indicate the validity of the EKC hypothesis for the MENA countries and Turkey: CO₂ emissions per capita increase at lower levels of GDP per capita, and they subsequently decrease at higher levels. A 1% increase in the square of GDP per capita yields a 0.6% reduction in CO₂ emissions per capita. It further points to the MENA countries and Turkey first experiencing a scale effect of growth outweighing composition and technique effects. In later development stages, the technique effect is such they would be adopting new and advanced/enhanced technology and environmentally friendly production techniques with a CO₂ emission-reducing effect. This finding is in line with empirical findings of Doğan et al. (2022), Boleti et al. (2021), Can and Gozgor (2017), Kherzi et al. (2022) Martins et al. (2021).

Concerning ECI, though not found to be statistically significant in model (1), the negative sign of the coefficient may point to higher levels of complexity, associated with reduced CO₂ emissions. With respect to the statistically significant consumption of renewable energy variable, a 1% increase in renewable energy consumption decreases CO₂ emissions per capita by 0.06%. Thereby, the increased consumption of renewables has the expected deteriorating effect of environmental quality. Furthermore, though the coefficient of the interaction variable of economic complexity and renewable energy consumption is not statistically significant, it may indicate that perhaps there are insufficient policies fostering economic complexity and renewable energy consumption together. Also, the negative sign on both coefficients would suggest that renewable energy consumption would reduce CO₂ emissions further in countries with higher economic complexity.

In model (2), the EKC is also found to be valid for the MENA countries and Turkey. Furthermore, the coefficient of ECI is statistically significant and with a negative sign, such that a unit increase in ECI, and reduces log of CO₂ per capita by 0.05 units. Thereby, higher levels of complexity are associated with reduced CO₂ emissions indicating that economic complexity enhances the technique and composition effects in the MENA countries and Turkey, thus having a mitigating effect on CO₂ emissions. This result is in line with the findings of Leitao et al. (2021) with application to the BRICS economies, Doğan et al. (2019) with respect to high-income countries and Can and Gozgor (2017) with respect to France. Further results of model (2) show that a 1% increase in the consumption of fossil fuels increases CO₂ emissions by 0.2%, portraying the expected deteriorating effect of the consumption of fossil fuels on environmental quality.

In model (2), population density is also statistically significant and has a negative (improving) effect on CO₂ emissions per capita, with a 1% increase in population density, reducing the log of CO₂ emissions per capita by 0.006%. This result is in line with the findings of Ahmed et al. (2019) who find that the scale effect of growth may be outweighed by an advancement in technology and an efficiency in the use of public services such as waste disposal, sanitation, heating, rail and bus-based collective transportation, among other public services which render population density having a negative effect on CO₂ emissions (Ahmed et al., 2019).

Furthermore, Table 7 gives results of the Dmitrescu-Hurlin panel causality test results.

Table 7 Dmitrescu-Hurlin panel causality test of carbon emissions, economic complexity, fossil fuel consumption and population density

Panel causality test results convey the existence of a bidirectional causality between CO₂ emissions per capita and GDP growth. This suggests that GDP growth can predict CO₂ emissions such that higher stages of development may be associated with emissions reduction. In the other direction of causality, CO₂ emissions would predict GDP growth, suggesting that the emissions increase is expected to have adverse effects on growth. There is also bidirectional causality between population density and CO₂ emissions which suggests that population density can predict CO₂ and vice versa. As such, higher population density may increase CO₂ emissions per capita, but higher emissions may also propagate greater population density. Meanwhile, there is a unidirectional relationship from ECI to CO₂ per capita emissions which suggests that ECI increase may well predict lower emissions. There is similarly a unidirectional relationship from CO₂ emissions per capita to the consumption of fossil fuels, which suggests that increased CO₂ emissions have the undesired effect of propagating fossil fuel consumption.

Conclusion and Policy Implications

Having investigated the effect of ECI, GDP per capita, the validity of the EKC hypothesis, consumption of renewable energy and of fossil fuel energy, as well as population density on per capita CO₂ emissions, estimation results for models (1) and (2) confirm the validity of the EKC hypothesis for the MENA countries and Turkey. In model (1), renewable energy consumption is found to have a statistically significant mitigating effect on CO₂ emissions, and though economic complexity and its interaction with renewable energy consumption coefficients are not statistically significant, each bears a negative sign. This may indicate that perhaps there are insufficient policies fostering economic complexity and renewable energy consumption together. In model (2), statistically significant fossil fuel energy consumption is found to lead to environmental quality deterioration. Meanwhile, statistically significant economic complexity and population density indicate that increased complexity and increased population density may be associated with improving environmental quality. Granger causality test results show that there is a bidirectional causality between CO₂ emissions, GDP growth and population density, respectively. There is only a unidirectional relationship from ECI towards CO₂ emissions and from CO₂ emissions towards the consumption of fossil fuel. In total, the findings further nominate increased economic complexity, increased renewable energy use and lessened reliance on non-renewable energy, as well as increased population density work towards the MENA countries and Turkey’s improvement of environmental quality and achievement of SDG goals 7, 9, and 12.

Perhaps a key policy implication relates to the validity of the EKC hypothesis in conjunction with the negative effect of economic complexity on CO₂ emissions. The MENA countries and Turkey must, thereby, recognize that as their incomes increase and their industrialization is further promoted, their economic structures should evolve from a less-to-a-more knowledge-based structures (with increased economic complexity), with possible CO₂ implications for CO₂ emissions reduction.

And since the boosting of economic complexity requires technological competencies, this calls for MENA countries and Turkey’s investments to be directed towards industries of relatively higher technology levels. This, in fact, comes in line with the COP27 goals and implementation plans which call for developed and developing countries needing to cooperate on technology development, its transfer and innovation, all towards emissions reduction. Perhaps subsidizing research and development conducted by developed countries in technology and its subsequent transfer to the developing countries may catalyze that process. The shift to digitalization is also recommended as a step in that direction.

When it comes to economic complexity, it is of great importance that the MENA countries and Turkey recognize that it would be a mistake for them to use a one-size-fits-all mechanism for increasing economic complexity. Each country’s industries and capabilities must be closely examined so as to nominate candidate industries whose promotion would aid in achieving higher complexity. We also point to the need to examine productivity constraints in each country so as to foster greater complexity.

From an energy perspective, the MENA countries and Turkey may be well-advised to work towards greater energy efficiency, thereby reducing the overwhelming fossil fuel consumption in order to mitigate CO₂ emissions effects. Descriptive statistic given in Table 2 shows the average share of fossil fuels in total energy consumption is 96%, against a mere of 3.9% average renewable energy. There is much room for renewable energy with a higher share of solar, wind, biofuels and green hydrogen renewables in the countries’ energy mix. Many of the MENA countries are not only rich in oil but also in solar energy.

As to the population density having a negative effect on CO₂ emissions level, this would suggest that higher population density in the region comes along with more rail and bus transportation, reduced energy use and reduced private vehicle use. This would suggest that substantial attention be given to sustainable transportation in these countries as an option to further mitigate negative effects on environmental quality.

Perhaps a limitation of this study is that the two estimated models were of the panel homogeneous slopes type. With a heterogeneous slope specification (and it is validated as an appropriate model specification), the researcher may obtain slope coefficients for individual MENA countries and thereby differentiate between the different panel members. Not only are the results expected to be panel member-specific but also may further allow for that large MENA countries’ group to be classified into smaller sub-groups. This should allow for drawing policy implications more relevant and specific to each subgroup. As much as this is a limitation of the study, it may also be a direction for future applied research on the region.

Funding

No funding was received to assist with the preparation of this manuscript.