“Revisiting the environmental Kuznets curve: Evidence from West Africa”

This paper analyzes the revenue-pollution relationship by revisiting the Environmental Kuznets Curve (EKC) hypothesis for West African countries over the period of 1980– 2014. The study approximates the income measurement by GDP per capita and uses carbon dioxide (CO2), nitrogen oxide (NO2), and methane emissions as various environmental quality measures. The paper uses parametric and non-parametric estima- tion techniques to test the EKC hypothesis. The results support the existence of the U-inverted relationship between income and methane emission, on one hand, and be- tween income and nitrogen dioxide emission on the other. The estimates also show a mixed result for the U-inverted hypothesis between income and carbon dioxide emis- sions. Thus, the verification of the curve depends on the estimation techniques and the measurement of the pollutant used. The obtained results led to the conclusion that the EKC hypothesis is validated for West African countries. effects models, Autoregressive


INTRODUCTION
Since the early work (Grossman & Krueger, 1991;Holtz-Eakin & Selden, 1995), the Environmental Kuznets Curve (EKC) hypothesis has given rise to several empirical works in economics and other related sciences. For some authors, the growth of economic activities requires more significant inputs of energy and materials. This growth is driven by industrialization, which is expected to generate more substantial quantities of liquid, gaseous, and solid waste by-products. The accumulation of these wastes contributes to environmental degradation and, thus, to the degradation of the well-being of earth's inhabitants, despite the steady rise in countries' incomes. This work proves that the degradation of the resource base will eventually put economic activity at risk. Thus, to save both the economy and the environment, countries need to slow down their growth, and the world needs to transition to a sustainable economy.
On the other hand, other researchers argue that the fastest route to an improved environment is economic growth. With higher incomes, there is an increased demand for less material-intensive goods and services and a demand for enhanced environmental quality, leading to the adoption of environmental protection measures. The goal of any economy is to achieve a desirable and sustainable economic growth and development level. However, the environment's quality varies according to the various stages, patterns, and structures of development that each country wishes to achieve. Thus, although economic growth usually leads to environmental deterioration in the early stages of the development process, the best way to achieve a decent environment in most countries is to become rich. The central issue is the evidence of the Environmental Kuznets Curve (EKC), which assumes that at low-income levels, an increase in national income is accompanied by increased environmental pressure in the form of an inverted U-shape. This relationship thus described is generally known as the Environmental Kuznets Curve (EKC).
Empirical studies support an inverted U-shaped relationship between environmental degradation and economic growth (Grossman & Krueger, 1991). On the other hand, others highlight the existence of a bell-shaped relationship between levels of environmental quality and per capita income. Azomahou et al. (2006) and Stern (2004) find that GDP growth rates do not play significant role in environmental quality dynamics.
In the context of African countries, studies on the EKC hypothesis are far from reaching a consensus. In the case of organic pollutants in water, their paper suggests a positive connection when Osabuohien et al. (2014) establish the existence of an Environmental Kuznets curve for CO2 emissions.
Over the period 2010-2018, the Member States of Economic Community of West African States (ECOWAS) achieved average real GDP growth of 3.8% against 2.7% for the whole of Africa (WDI, 2020). The rates for energy consumption and carbon dioxide emission for the ECOWAS region also exploded over this period. Data from the Notre Dame University Global Climate Change Adaptation Indicator (ND-GAIN) confirms that despite notable progress since 2008, the ECOWAS region is highly vulnerable to climate change and has very low resilience to climate-related risks. For instance, the vulnerability of ECOWAS in 2018 was 0.547 against 0.442 for the world, while the resilience to climate risks was 0.309 versus 0.427 for the world (ND-GAIN Country Index, 2018). Therefore, there is a need for ECOWAS countries to pursue policies within the framework of the commitments made in the United Nations 2030 agenda and the Paris Climate Agreement for economic development with less polluting gases.
However, given the mixed results of empirical studies on the link between economic growth and the emission of polluting gases, it is necessary to investigate this link within the ECOWAS framework to assess the policies' effectiveness. In doing so, is the hypothesis of the environmental Kuznets curve in inverted U shape verified in ECOWAS countries? This paper's main objective is to determine the relationship between GDP, energy intensity, sectoral value added in services, industry, CO2, NO2, and Methane emissions in ECOWAS countries. At the end of the study, it emerges that the environmental Kuznets curve is verified for ECOWAS. However, the results are dependent on the pollutant used and the estimation techniques implemented.

LITERATURE REVIEW
Since Grossman and Krueger (1991), several empirical studies have tested the relationship between growth and the environment in several other countries by including other indicators. The research hypothesis in these various studies is the "Environmental Kuznets Curve (EKC)" hypothesis. This hypothesis postulates that the relationship between income (often captured by real Gross Domestic Product (GDP) per capita) and pollution approximated by the concentration of a par-ticular pollutant is an inverted U-shaped curve. According to Kuznets (1955), increasing income levels increase environmental pollution in the first phase of the development process. Beyond an income threshold, the second phase is marked by a decrease in ecological degradation (pollution) due to the adoption of efficient technologies in the production process.
At the empirical level, the analysis of the EKC hypothesis has provided mixed results (Effiong & Oisaozoje, 2016) due to economies' divergence. On one hand, empirical studies provided unclear results and, on the other hand, increasingly intense and precise criticisms of the estimation techniques. One reason put forward by Effiong and Oisaozoje (2016) to justify these mixed results is the diversity of variables used as a proxy of environmental quality and the econometric estimation techniques employed. Findings can be grouped into two categories.
The first group of studies has concluded that the EKC hypothesis has been verified in certain countries or groups of countries (Lamla, 2009

Empirical work using nonparametric or semi-parametric estimation models
In parallel to studies that used parametric estimation methods, other papers tested the pollution-income nexus through more flexible estimation techniques (non-parametric and semi-parametric). The choice of an estimation methodology has sometimes been criticized in the literature. Stern (2004) details why the statistical and econometric analysis on which the Environmental Kuznets Curve is based is not robust. He highlighted heteroskedasticity problems, simultaneity, cointegration, or bias due to omitted variables. Other issues arise from the reduced-form analysis. The choice of functional form (quadratic or cubic) also affects the result obtained and determines the number of inflection points that appear (Kijima et al., 2010). Indeed, if the estimators are significant and have the expected sign, a quadratic form implies an environmental Kuznets curve. In contrast, a cubic shape imposes the N-shaped curve. Thus, several works overcome this restriction of the functional form by using non-parametric methods (semi-parametric panel with fixed effects), which have the advantage of being more robust.
Using this method, Bertinelli

Theoretical framework and empirical specification of the model
The paper uses the Ehrlich and Holdren (1971) model as a theoretical framework to analyze the relationship between income and environmental quality measured by CO2, NO2, or methane emissions. This model makes it possible to describe ecological changes induced by human activities by relating environmental quality to variables such as population, income, and technological level. This model is translated into the following equation: , where I -the variable measuring environmental impact; P -the population's size; A -the variable measuring human economic activity; and T -the technological level per unit of consumption and production. This model has been improved by Dietz and Rosa (1997) and York et al. (2003). Indeed, the IPAT model has been criticized as an identity mathematical model and rigid in proportionality restrictions between variables. As a result, its first improvement proposed by Dietz and Rosa (1997) resulted in a flexible, more improved model called STIRPAT, which makes it possible to analyze environmental impacts. The equation translates this: The variables I, P, T, and A remain the same as those in the IPAT model. But a, b, c, and d represent the model parameters. ε i is the idiosyncratic error term; i refers to the units of observation. By taking the logarithm of equation (2), a relationship is obtained, which is translated by the following equation (3): .
where C it -is the variable for measuring the quality of the environment; E it -is the total final energy consumption (Terajoules); Indva it -is the value-added of the industrial sector as a proportion of GDP; Serva it -is the value-added of the service sector as a proportion of GDP. Yit -real GDP per capita; α i represents the specific effect to each country; τ t -time effect variable to account for omitted time-varying variables and stochastic shocks common to all nations; i -is the individual dimension representing countries; and t represents time.
The income's effect on the quality of the environment distinguishes two (2) cases according to the sign and significance of the parameters β 4 and β 5 . The first case is where β 4 > 0 (β 5 < 0 respectively) and β 5 = 0, then the relationship increases monotonically (respectively decreases). The second case is the one in which we have β 4 = 0 and β 5 < 0, on the one hand, and β 4 > 0 and β 5 < 0, on the other. In this case, an inverted U-shaped relationship is observed between income and CO2 emission, and the turning point is given by:

Econometric estimation methods
The study uses fixed-effect (FE) models, random-effect models, and a semi-parametric estimation technique (SEMI-PARA) based on panel data. The advantage of using a semi-parametric estimation technique is that it makes no assumptions about 2 For the estimation method of semi-parametric partially linear models at effets fixes, refer to Baltagi and Li (2002).
the functional form between income and the measurement of environmental degradation (CO2 emissions, for example). The relationship tested is more flexible because the functional form's estimation depends on the sample data. Following the example of Effiong and Oisaozoje (2016), let us specify a semiparametric partially linear fixed-effects model as follows: ( ) where m (.) is an unknown smooth function that is sought to be estimated and depends only on income while the other variables in the specification are specified parametrically. Equation (6) is estimated by the method of Baltagi and Li (2002), which is carried out in two steps. In the first step, the parameters of the variables specified in the model's parametric way are estimated. Parameters estimated in the first step allow the m (.) curve estimate by the kernel's regression method or the spline 2 .

EMPIRICAL RESULTS
This section is devoted to the presentation and discussion of the results. These are the panel estimation results of Ordinary Least Squares (OLS) model, the fixed-effects model, the random-effects model, and the semi-parametric fixed-effects model. The estimation results of these different models are presented in Tables 4, 5, and 6.

Modelling using CO2 emissions as a measure of environmental degradation
When CO2 emissions are considered ecological quality indicators, total energy consumption is significant for all estimation techniques (at the 1% and 10% threshold). This result is consistent with those of Effiong and Oisaozoje (2016), showing that an increase in energy consumption contributes to a rise in CO2 emissions and environmental quality destruction. As for the sector added values considered, they are only significant in the semi-parametric estimates. Indeed, this implies that industrialization favors CO2 emissions and thus contributes to environmental degradation ( Table 4).
Effects of income and income square on the CO2 emission were done with the OLS model. Then, the nature of the income-CO2 emission link was tested in the semi-parametric model by analyzing the graph describing the shape of the relationship between these two variables ( Figure 1). Two observations can be made.
First, OLS estimation shows that income is positive and significant at the 1% threshold, and its quadratic term is negative and significant at the 5% threshold only (Table 4). Thus, the income-CO2 emission relationship in the ECOWAS zone follows a positive relationship, indicating that higher-income contributes to environmental quality degradation.   Secondly, concerning the semi-parametric estimation technique, the curve in Figure 1 shows the absence of a U-inverted relationship contrary to the estimation with OLS model. The semi-parametric estimation fails to validate the EKC hypothesis leading to mixed results depending on the estimation technique. Not surprisingly, these results confirm the work of Stern (2004) and Effiong and Oisaozoje (2016), who conclude whether or not the EKC hypothesis is verified depends on the estimation techniques used.
3.2. Modelling using NO2 emissions as a measure of environmental degradation Table 5 presents, as previously, the results of the estimates of the linear fixed-effect and semi-parametric models using the nitrogen dioxide (NO2) emission variable. The variables of energy consumption and value-added of the industrial sector are significant at the 1% threshold with a positive and negative sign. Energy consumption contributes more to nitrogen dioxide emissions and harms the environment. As for the verification of the EKC hypothesis, the OLS model shows that it is well verified in the context of nitrogen dioxide. The GDP per capita threshold corresponding to the maximum emission is set at e (-7.441/2(-0.550)) = 866.57 USD and above this threshold, any increase in per capita income translates into a reduction in the level of nitrogen dioxide emissions. The average value of ECOWAS per capita GDP (USD 883) being above this threshold, ECOWAS is in the decreasing phase of the EKC curve. Semi-parametric modeling also confirms the existence of this EKC relationship in the ECOWAS zone. Indeed, starting from no functional form between NO2 emission and GDP, the method suggests an inverted U-shape between nitrogen dioxide emission and GDP per capita ( Figure 2). From these two analyses, it emerges that the parametric and semi-parametric methods converge on the validation of the EKC hypothesis in the ECOWAS region by considering nitrogen dioxide to measure environmental pollution.

Modelling using methane emission as a measure of environmental degradation
The results of the methane emission estimates are reported in Table 6. The OLS model estimation showed that all variables are significant at the 5% threshold. Energy consumption and sectoral added values were found to be substantial and of positive sign. Thus, an increase in energy consumption and an increase in industrial activity and services contribute to a rise in methane emissions. The real GDP per capita and its quadratic term proved significant with a positive and negative sign, verifying the EKC hypothesis. The threshold of GDP per capita equivalent to the maximum methane emission in the ECOWAS zone is e (-3.257/2(-0.307)) = 199.95 USD and indicates that the ECOWAS countries are on average in the decreasing phase of the Environmental Kuznets Curve.   The semi-parametric estimate does not establish a relationship between methane emission and energy consumption variables and sectoral added values. Still, it confirms, through Figure 3, the verification of the EKC hypothesis in the ECOWAS zone. Thus, parametric and semi-parametric methods make it possible to validate the EKC hypothesis in the ECOWAS zone when methane emissions are used as a measure of degradation of environmental quality.

CONCLUSION
This paper aims to revisit the Environmental Kuznets Curve (EKC) to determine the revenue-pollution relationship's functional form using a sample of 14 ECOWAS member states over the period 1980-2014. The paper approaches income by GDP per capita and uses three indicators as proxies for pollution measurement: CO2, NO2, and methane emissions in parametric and semi-parametric modeling.
It emerges that for ECOWAS, the Kuznets environmental curve's hypothesis is confirmed if methane and nitrogen dioxide are used as degradation of the quality of the environment independently of the adoption of parametric or semi-parametric modeling. These results are consistent with those of Effiong and Oisaozoje (2016) and Keho (2017). However, with carbon dioxide as an indicator for measuring environmental pollution, the results are contradictory depending on whether the modeling is parametric or semi-parametric. The EKC curve is verified in parametric modeling and invalidated in semi-parametric modeling.
These results require awareness of the need to reduce CO2 emissions, which will continue to grow without clean production techniques. The ECOWAS Commission should promote measures among States and citizens of the Community to reduce CO2 emissions and preserve the ozone layer. These measures could concern the promotion of alternative or renewable energy sources such as solar, wind, and geothermal energy. Other actions could include combating deforestation to increase the absorption of CO2 released into the atmosphere.