Banking sector development and environmental degradation in the Economic Community of West African States: do technology effects matter?

Abstract

This paper contributes to the discussion on environmental degradation by exploring the connection between banking sector development and environmental degradation in the ECOWAS. In addition, we investigate the direction of causation between environmental degradation and its drivers and the technological effect of banking sector development on environmental degradation. We rely on a balanced panel dataset of 11 ECOWAS nations from 1990 to 2019. We present the following conclusions using the AMG estimator and the Driscoll–Kraay panel regression model. First, banking sector development reduces environmental degradation. Second, banking sector development has a deleterious technological effect on environmental quality. Thirdly, population and affluence were found to significantly promote environmental degradation, while the impact of technology was inconclusive. We further demonstrate a unidirectional causation association between the development of the banking sector and environmental degradation using the Dumitrescu and Hurlin causality analysis. Based on the study conclusions, numerous policy ramifications have been suggested for the ECOWAS nations to mitigate environmental degradation.

Introduction

Environmental quality (EQ) issues have gained worldwide attention over the past decade. Many academics and decision-makers in developed and developing nations have reiterated the importance of reducing greenhouse gas emissions (GHGs), major contributors to climate change and global warming [1, 2]. These stated hazards are mostly linked to population growth, economic growth, technological advancements, and other economic activities linked to an increase in fossil fuel utilisation, of which banking sector development (BSD) is no exception. In this paper, we review issues concerning the association between BSD and environmental degradation, perform an empirical exercise concentrating on long-run impacts utilising information from ECOWAS, and elaborate on the findings’ policy ramifications. Financial sector development (FSD) happens when the costs that financial market players incur for information, transactions, and enforcement relating to the operation and monitoring of the financial system are lowered [3]. FSD can but does not invariably, have a favourable impact on economic growth [4]. Unsurprisingly, FSD is most frequently investigated regarding how it impacts economic growth, given the proven favourable social well-being consequences linked to economic growth.

Climate change issues in the ECOWAS region have gone unaddressed until recently, while policies promoting economic growth have received much attention. Owing to that, environmental degradation in ECOWAS has worsened due to rising CO₂ emissions. Figure 1 shows that CO₂ emissions have surged from 1990 to 2019 with few instances of decline. The increases in non-renewable energy consumption might explain this trend, as increases in non-renewable energy consumption dampen EQ by releasing greenhouse gases into the atmosphere [5]. Many studies have examined how FSD affects EQ and energy consumption. Considering the underdeveloped nature of the banking sector in the ECOWAS sub-region, supporting the adoption and implementation of green technologies would be herculean. This is because research and development (R&D) efforts vital for adopting and implementing eco-friendly technologies get inadequate funding in economies with underdeveloped banking sectors. Even after embarking on several reforms to accelerate banking sector development, ECOWAS still lags behind other regions in this aspect.

Fig. 1

The trend of CO₂ emissions in ECOWAS from 1990 to 2019

The paper concentrates on the causal impacts of BSD on environmental degradation (ED), which is vital for at least three reasons. First, since these impacts are ambiguous from a theoretical standpoint, it is important to research to identify the empirical causal links between BSD and ED. Prior studies have shown that the impact of FSD on EQ has varied by the dimension of FSD examined, the country or region, and the period studied. As noted, these impacts can be positive [6,7,8], negative [9,10,11], and neutral [12, 13]. Second, some studies have established a bidirectional causality between FSD and economic growth [14, 15]. Thus, comprehending the observed causal association between ED and FSD sheds new light for better appreciating the nexus between economic growth and ED. Third, as relevant social and economic policies are more impacted by environmental factors, which are receiving larger community attention, this understanding is becoming even more crucial.

Despite much research on the connection between FSD and EQ, this paper expands the corpus of knowledge in three areas. Firstly, a few studies try to determine how the BSD has impacted EQ, for example, Samour et al. [1] for South Africa, Radulescu et al. [10] for OECD countries, and Mehmood [16] for a panel of N-11 economies. It is crucial to highlight that none of these studies examined how BSD affected EQ in the ECOWAS region. As a consequence, the novelty of this paper is to examine how BSD influences ED in ECOWAS. Secondly, this study departs from prior studies by exploring the technological effects of BSD on ED. BSD may have a technological impact because credit provided by the banking sector often facilitates technological advancements. These technological improvements may lead to adopting and implementing ecologically friendly technologies [11] or drive energy utilisation [17]. Finally, most existing empirical literature fails to account for cross-sectional dependence (CSD) and slope heterogeneity, which may invalidate empirical findings if present in the data. This study addresses this gap by utilising the second-generation stationarity test, the Westerlund cointegration test, the augmented mean group (AMG) estimator, and the Driscoll–Kraay panel regression technique, which offer reliable estimations when slope and CSD heterogeneity are present.

Following this introduction, the remainder of the paper comprises the following sections: A succinct summary of the pertinent literature is provided in “Literature review” section. The data and econometric techniques are covered in “Methodology” section. The empirical findings are presented and discussed in “Results and discussion” section. “Conclusion and policy ramifications” section presents the conclusion and policy ramifications.

Literature review

During the last decades, many scholars have explored the factors influencing EQ, CO₂ emissions in particular. The seminal work of Grossman and Krueger [18] demonstrated an inverted U-shaped association between economic growth (income) and ED. Environmental Kuznets curve (EKC) is the term given to this connection [19,20,21]. It contends that income is positively associated with ED at the early phases until a certain threshold, beyond which the association becomes negative. Also, A plethora of studies have explored the factors influencing ED on different dimensions like the impact of income (affluence) on EQ [21,22,23], the effect of population on EQ [24,25,26,27,28] and the interaction between technology and EQ [29,30,31] using a variety of time series and panel models with a range of sample sizes. For instance, on the affluence-ED nexus, Phadkantha and Tansuchat [32] using the Markov switching model indicate that rises in income increase ED in Thailand. Similarly, Chen et al. [33] found that affluence boosts ED while technology mitigates ED in Bangladesh from 1972 to 2020. In the case of Egypt, Raihan et al. [34] validate the harmful impact of affluence on EQ from 1990 to 2019. Similar findings were established by Raihan and Tuspekova [35] for Kazakhstan, Malik et al. [36] for Pakistan, Udeagha and Breitenbach [37] for South Africa, Patel and Mehta [38] for India, and Liu et al. [39] for China.

Turning to the population-ED nexus, Yang et al. [40] and Li et al. [41] demonstrate that low EQ is associated with increasing population size in China. On the contrary, employing the ARDL technique, Ali et al. [42] found that increasing population size significantly improved EQ in China from 1990 to 2019. Using the log-polynomial estimating approach, Yeh and Liao [43] established a positive association between population and ED in Taiwan from 1990 to 2014. Between 1999 and 2014, Hashmi and Alam [26] examine the nexus between population and ED in OECD nations and found that population growth harms EQ. Studies conducted by Kwakwa and Alhassan [44], Anser et al. [45] and Liu et al. [46] further reinforced the harmful effect of increasing population size on EQ. However, Begum et al. [47] document that population growth does not affect EQ in the Malaysian context.

Regarding the impact of technology on ED, Hussain and Dogan [48] conducted a panel study of BRICS nations using data from 1992 to 2016. Using the cross-section augmented ARDL approach, their findings suggest that technology negatively influences ecological footprints. The implication is that technology drives EQ. A recent study by Chu [49] establishes that technology and renewable energy usage are critical in attaining sustainable development as they reduce environmental degradation. Huo et al. [50] explored the nexus for China using data covering 1991–2017 and the ARDL technique. Technology was found to harm EQ. Employing the quantile QARDL model and data from Pakistan from 1980 to 2019, Chien et al. [51] found a negative association between technological innovations and ED. The inhibiting effect of technology on ED is supported [31, 52,53,54].

Theoretically, the association between FSD and ED is contentious among scholars with conflicting viewpoints. Indeed, a relevant issue that has received much empirical attention over the past few decades is how banking sector development (i.e., FSD) affects ED. For instance, Samour et al. [1] using the ARDL approach, conclude that BSD significantly dampens EQ in South Africa for the period covering 1986–2017. Therefore, an increase in credit from the banking sector will result in investments in projects and expansions and an enhancement of risk mitigation mechanisms, which would influence economic growth and energy demand, increasing CO₂ emissions. Mehmood [16] using data from N-11 countries and the cross-sectional ARDL approach, established that BSD has a significant positive association with CO₂ emissions. Regarding OECD economies, Radulescu et al. [10] demonstrated that BSD has a deleterious impact on environmental sustainability. Employing the Methods of Moments—Quantile Regression, Chien et al. [55] found that FSD and economic growth significantly harmed EQ in BRICS from 1995 to 2018. This finding is validated by Yang et al. [17] for the period 1990–2016. Similarly, Baloch et al. [56] established that FSD increases ecological footprint using the Driscoll–Kraay panel regression model for 59 Belt and Road countries (BRIC). Similar outcomes of the adverse influence of FSD on EQ were established by Ibrahiem [57] for Egypt, Avom et al. [11] for Sub-Saharan Africa (SSA), Charfeddine et al. [58] for the MENA region, Zakaria and Bibi [59] for South Asia and Musah et al. [9] for West Africa.

On the contrary, Shahbaz et al. [60] showed that FSD lessens CO₂ emissions, reducing environmental deterioration in France. Aluko and Obalade [8] highlight that FSD contributes to an environmental decline in Sub-Saharan Africa (SSA). However, FSD has a dampening technological effect on EQ. Park et al. [61] suggested that FSD reduces ED by hindering CO₂ emissions in European Union (EU) economies. Regarding Asia Pacific Economic Cooperation (APEC) countries, Usman and Hammar [6] demonstrated that FSD and renewable energy usage positively impact EQ. Saidi and Mbarek [62] found that FSD mitigates ED in emerging economies. Applying the ARDL model, Salahuddin et al. [63] found an inverse association between FSD and ED. Dogan and Seker [64] discovered that FSD had a dampening influence on CO₂ emissions in top renewable energy countries. Focusing on G-7 and N-11 nations, Zafar et al. [7] found evidence that BSD promoted EQ. Using data from 54 developing economies, Yang et al. [65] demonstrate that financial instability enhances EQ by reducing CO₂ emissions.

Some studies also established the neutral impact of FSD on ED. For example, Adams and Klobodu [66] using a panel of 26 African nations, concluded that FSD is an insignificant determinant of EQ. Studies like Omri et al. [67] for MENA countries, Jamel and Maktouf [13] for European countries, and Maji et al. [12] for Malaysia confirm the neutral effect of FSD on EQ. The empirical works of Acheampong et al. [68] and Acheampong [69] demonstrate that the impact of FSD on EQ is contingent on the measure of FSD, the sampled period, and the countries under study. Fakher and Ahmed [70] observed that FSD magnifies the positive impact of technology on EQ using a sample of 25 OECD nations.

Thus, there is no unanimity among the findings of different studies worldwide. Furthermore, there is a paucity of studies that address this critical connection between BSD and EQ, as well as the technological impact of BSD on EQ generally and in the ECOWAS region specifically. The authors are aware of no empirical work in ECOWAS that examines these nexuses. Additionally, most of these studies fail to consider critical panel data issues like CSD and slope homogeneity, which might generate biased conclusions. This study, therefore, examines the case of ECOWAS while accounting for cross-sectional dependence and slope homogeneity by using appropriate second-generation and long-run estimators such as the augmented mean group estimator and the Driscoll–Kraay panel regression model.

Methodology

Empirical model

The study adheres to the conceptual framework of the STIRPAT model [71]. The STIRPAT argues that environmental degradation results from affluence, technological changes, and population. The STIRPAT model is an improvement on the IPAT model [72] by transforming it into a stochastic model. The IPAT model is a mathematical identity that presupposes proportionality among these determinants in the functional association and does not permit hypothesis testing [73]. This limitation was overcome by the STIRPAT model, which also offered a theoretical foundation for understanding how population growth, technological advancements, and affluence affect EQ. The STIRPAT model has been modified in several studies to include additional critical determinants of environmental degradation [8, 11, 29, 74,75,76,77]. The following equation gives the basic STIRPAT:

$$ I_{it} = \alpha P_{it}^{\partial } A_{it}^{\varphi } T_{it}^{\omega } \varepsilon_{it} $$

(1)

In model 1, I is environmental degradation, α is the constant, while $\partial$, $\varphi$ and $\omega$ denote elasticities of population (P), affluence (A), and technology (T), respectively. $\varepsilon$ is the residual term, and subscripts i and t are the cross-sectional unit (i.e., country) and time period, respectively. The STIRPAT model is linearised by taking the natural logarithm (ln) form, which is written as:

$$ \ln I_{it} = \ln \alpha + \partial \ln P_{it} + \varphi \ln A_{it} + \omega \ln T_{it} + \varepsilon_{it} $$

(2)

We modify the STIRPAT model by integrating lnBSD to reflect the impact of BSD on ED. The modified STIRPAT model is stated as follows:

$$ \ln {\text{ED}}_{it} = \ln \alpha + \partial \ln P_{it} + \varphi \ln A_{it} + \omega \ln T_{it} + \rho \ln {\text{BSD}}_{it} + \varepsilon_{it} $$

(3)

To explore the technological effect of lnBSD on lnED, we examine if lnT moderates the lnBSD—lnED nexus. Model 3 was thus modified by incorporating an interactive term of lnBSD and lnT ($\ln {\text{BSD}} \times \ln T$):

$$ \ln {\text{ED}}_{it} = \ln \alpha + \partial \ln P_{it} + \varphi \ln A_{it} + \omega \ln T_{it} + \rho \ln {\text{BD}}_{it} + \vartheta \left( {\ln {\text{BSD}} \times \ln T} \right)_{it} + \varepsilon_{it} $$

(4)

In this study, the impacts of banking sector development (lnBSD), population (lnP), affluence (lnA), technology (lnT), and technology effect of BSD (lnBSD × lnT) on environmental degradation (lnED) are empirically examined with the aid of AMG estimator [78, 79] which accommodates a common dynamic mechanism and takes into account CSD $\hat{\mu }_{t}^{*}$ in the country regression. Additionally, the AMG estimator exhibits acceptable bias and RMSE performance and is resilient to CSD and nonstationary data. The AMG estimate goes through two phases:

$$ {\text{Phase 1}}:\;\Delta Y_{it} = \beta^{\prime}\Delta X_{it} + \mathop \sum \limits_{t = 2}^{T} c_{t} \Delta D_{t} + \varepsilon_{it} \to \hat{c}_{t} = \hat{\mu }_{t}^{*} $$

(5)

$$ {\text{Phase 2}}:\;Y_{it} = \emptyset_{i} + \beta^{\prime}X_{it} + d_{i} \hat{\mu }_{t}^{*} + \varepsilon_{it} ; \hat{\beta }_{AMG} = N^{ - 1} \mathop \sum \limits_{i} \hat{\beta }_{i} $$

(6)

where $\Delta$ is the first difference operator, $\emptyset_{i}$ is the intercept, $\varepsilon_{it}$ is the residual term and $\hat{\beta }_{{{\text{AMG}}}}$ denotes the AMG estimation. $\hat{\mu }_{t}^{*}$ is the time dummy parameter. Lastly, the path of causality was examined using the Dumitrescu and Hurlin [80] panel causality test, which produces unbiased outcomes in small and large heterogeneous panels and is robust to CSD.

Data and variables

In this study, panel data are employed because they can be used to analyse outcomes that are difficult to identify in time series or cross-section data [81]. We used a balanced dataset for eleven Economic Community of West African States (Benin, Burkina Faso, Cabo Verde, Cote d’Ivoire, Ghana, Guinea, Niger, Nigeria, Senegal, Sierra Leone, and Togo) covering the period 1990–2019. Only ECOWAS nations with data for the chosen variables were captured in this paper. The WDI database is the source of all data. Environmental degradation was measured using CO₂ (emissions per capita) [24, 57, 82,83,84], banking sector development is proxied by domestic credit to the private sector by banks (% of GDP), GDP per capita (constant 2015 US$) measures affluence [8, 29, 85, 86]. The technological effect (T) is the industry sector value added (% of GDP). Modern technologies are introduced during industrialisation to produce new and old goods. As a result, industrialisation has frequently been employed as a technology indicator [8, 73, 82, 87]. The technological effects of BSD (lnBSD $\times$ lnT) are the interaction between lnBSD and lnT. In sync with most empirical papers, the aggregate population of a nation is used as a proxy for population. ED is more pronounced in economies with high populations [88]. All variables were transformed into their natural logarithmic forms for easy analysis and interpretation of results.

Table 1 presents the descriptive statistics of the variables. lnA and lnBSD are the only variables that are not negatively skewed. All variables, except population, which has a leptokurtic distribution, can be seen to have a platykurtic distribution. The correlations among the study variables are highlighted in Table 2. All the regressors are positively correlated with lnED. However, an insignificant correlation was established between lnP and lnED. A strong correlation coefficient of 0.832 was observed between lnA and lnED, indicating a strong correlation between these variables. A moderate correlation was established between lnBSD and lnED. The explanatory variables have modest correlation coefficients, with the maximum correlation coefficient being 0.505. This indicates the absence of multicollinearity issues among the variables. According to Gujarati [89], the problem of collinearity is present if the correlation coefficient exceeds 0.8. Figure 2 illustrates the associations between the variables in a graphical format.

Table 1 Descriptive statistics

Table 2 Pairwise correlations

Fig. 2

Scatter plot of the lower triangular matrix, kernel density of regression line variables

Results and discussion

Preliminary analysis

CSD may arise from globalisation which has brought dependence on economies in one region. Regression estimations could be biased and inconsistent if this problem is not resolved. This study begins its empirical analysis by conducting a CSD test for all three functional forms. Three different CSD approaches, Pesaran et al. [90] LM test, Breusch and Pagan [91] LM test, and Pesaran [92] CSD test were applied in this study. The outcomes of the CSD tests are reported in Table 3. All three tests reveal no cross-sectional independence in the panel models. The outcomes suggest that CSD exists; hence, employing an estimating approach that considers CSD is imperative. It is, therefore, appropriate for us to apply the AMG estimator.

Table 3 Outcomes of CSD tests

The presence of CSD rules out the use of first-generation panel stationarity tests owing to its stringent requirement of cross-section independence. Hence, using the first-generation stationarity tests might generate biased outcomes. Thus, we utilise a second-generation panel unit root test known as the cross-sectionally augmented IPS (CIPS) proposed by Pesaran [93], which checks the unit root properties of variables under the assumption of CSD. As displayed in Table 4, we failed to confirm the H₀ of nonstationary for all the variables except lnA, which became stationary after the first differencing. This shows that none of the variables are I(2). The AMG estimator may produce spurious results in the presence of I(2) variables.

Table 4 CIPS stationarity test

After establishing the presence of an I(1) variable in the panel series, the next logical step is to test for cointegration among the models. The Westerlund [94] cointegration test is based on four different types of statistics: group statistics (G_t and G_a) and Panel Statistics (P_t and P_a). From Table 5, the Westerlund panel cointegration test results are mixed in all panels and groups for the three models. The results of G_a and P_a confirmed the H₀ of no cointegration. At the same time, G_t and P_t failed to confirm the H_0, indicating the existence of cointegration in the STIRPAT and modified STIRPAT models. This shows that the regression results are valid and that there are long-term interactions among the variables. However, It is worth noting that the lack of cointegration does not impose restrictions on the AMG estimator [78, 95].

Table 5 Westerlund test for cointegration

Table 6 summarises the results of the AMG estimator. The findings suggest that lnP significantly positively affects lnED, as shown in columns 1–3. A 1% rise in lnP increases lnED by 0.586–0.833%. Thus, a surge in lnP promotes lnED because a rise in lnP drives the demand for non-renewable energy consumption, increasing the volume of GHGs emitted into the atmosphere [5]. This outcome of the study confirms those of Ghazali and Ghulam [25], Hashmi and Alam [26], Acheampong et al. [96], Opoku and Aluko [29], Aluko and Obalade [8], and Li and Lin [73] which shows a rise in population lower EQ. Franklin and Ruth [97] argued that lnP and other socioeconomic factors exert massive pressure on natural resources leading to ED. lnA is positively associated with lnED at least at the 5% significance level. This supposes that a rise in lnA harms EQ. From the analysis, a percentage rise in lnA will result in a 0.852–1.246% increase in lnED. The harmful effect of affluence suggests that a rise in the income levels of citizens increases the demand for goods and services, driving energy consumption which consequently drives the degradation of the environment in ECOWAS. This evidence corroborates the findings of Mohsin et al. [98], Pham et al. [99], Wang and Dong [100], and Anwar et al. [101], which also found an unfavourable impact of lnA on lnED. Adams and Klobodu [66] attributed the harmful impact of lnA on lnED to the prevalence of less effective production methods and procedures on the African continent.

Table 6 AMG test results

We found a neutral effect of lnT on lnED in columns 1 and 3. However, in Column 2, we established a positive but marginally significant (p value < 0.1) association between lnT and lnED. The outcome indicates that lnED will likely increase by 0.221% should lnT increase by 1%. This finding corroborates the findings of Usman and Hammar [6], Gu and Wang [102], Kivyiro and Arminen [103] and Yongping [104], which claimed that technology might promote ED due to inadequate funds allocated to research and development (R&D) for the desired level of technology to be attained. This finding also supports Jevon’s Paradox [105] which posits that technological advancements lead to a rise in energy consumption, thereby increasing CO₂ emissions. Also, this finding resonates with the rebound effect theory, a scenario in which technological advancements result in lower energy prices for energy-related products and services—thereby stimulating demand.

lnBSD significantly hinders lnED in both Columns 1 and 2. A percentage rise in lnBSD will significantly reduce lnED by 0.120–1.327%. Thus, we discover evidence that BSD amplifies EQ in ECOWAS nations. This finding agrees with Aluko and Obalade [8] for SSA countries and Usman and Hammar [6] for APEC nations. However, it contradicts the findings of Mehmood [16] and Samour et al. [1] which found a deleterious impact of FSD on EQ for South Africa and N-11 economies, respectively. As posited by Jiang and Ma [106], developments in the financial sector increase lnED by making funds/loans available to individuals and households that promote the acquisition of energy-consuming appliances such as automobiles and equipment. Mehmood [16] contends that increased bank lending to the private sector will boost economic activity and increase energy usage. The outcome of this paper, however, suggests that lnBSD can enhance EQ in ECOWAS by channelling funds toward investments in environmentally friendly technologies.

Focusing on the interactive term of BSD and technology (lnBSD $\times$ lnT), we establish a significant positive association between lnBSD × lnT and lnED. The finding demonstrates an adverse technological effect of lnBSD and shows that lnT complements lnBSD to raise lnED. Financial sector developments may promote technological advancements driving energy demand [5, 70, 107]. lnBSD may channel its dampening effect on lnED through lnT by increasing firms’ ability to access low-cost financing to expand their operations, acquire more machines and plants and create more jobs leading to a rise in energy consumption [108]. This may have dire consequences on EQ. This empirical outcome is similar to the work of Aluko and Obalade (2020), which reflects the harmful technological effect of FSD on EQ. Regarding the diagnostic tests, the Wald ${x}^{2}$ for all the estimations are significant at 1%, indicating that the results are valid for making predictions. The residuals are also free from strong CSD. Strong CSD in residuals may render statistical conclusions invalid.

To ascertain the robustness of the AMG estimates, we have applied the Driscoll and Kraay [109] standard error test, which is resilient to cross-sectional dependence and heteroscedasticity [21] and Table 7 presents the findings. As is evident, the outcomes corroborate the AMG estimates.

Table 7 Driscoll–Kraay standard errors test results

The AMG estimator only estimates the magnitude of the long-run association among the variables. Knowing the path of causality of these associations (+ or −) is critical to suggest some practical policy ramifications. In the presence of CSD, we applied the Dumitrescu and Hurlin [80] Granger non-causality test which accounts for CSD. The results of the D-H panel causality test of lnED and determining factors (lnP, lnA, lnT and lnBSD) are reported in Table 8. The findings indicate a two-way causality between lnP and lnED and lnA to lnED. The directional relationship between lnA and lnED is in sync with Ibrahiem [57] which supports the interdependence between lnA and lnED. lnA causes lnED, and lnED causes lnA in ECOWAS, and this is consistent with the literature because ECOWAS member states are developing nations with low-income levels. The unidirectional causality from lnBSD to lnED was anticipated. Advancement in the banking sector enables industries to access cheaper sources of funds for expansion and purchase of machinery which promotes economic growth but drives energy usage, consequently releasing harmful emissions. Furthermore, the analysis discovered a one-way causality from lnED to lnT. lnED causes lnT, and technology encourages using clean energy sources, promoting EQ by mitigating CO₂ emissions.

Table 8 Dumitrescu and Hurlin Granger non-causality test outcomes

Conclusion and policy ramifications

The globe has seen increased attempts to stem the rising tide of environmental degradation in an effort to attain the Sustainable Development Goals (SDGs). As scholars are finding ways to improve environmental quality, we explored the modified STIRPAT model by incorporating banking sector development and the interactive term of technology and banking sector development using a balanced panel of 11 ECOWAS member states for the period 1990–2019 using the AMG Estimator and Kraay standard error test. In addition, we explored the path of causality between environmental degradation and its determinants.

After considering cross-sectional dependence, the empirical findings demonstrate cointegration in all STIRPAT and modified STIRPAT models. The findings suggest that population and affluence are positively associated with environmental degradation. Thus, ECOWAS countries with larger populations and greater affluence are associated with higher rates of environmental degradation. However, we discovered that the model determined how technology affected environmental deterioration. Technology is, therefore, not a robust determinant of environmental degradation in ECOWAS. Banking sector development is negatively associated with environmental degradation, and this indicates that banking sector development improves environmental quality in ECOWAS. The results further suggest that technology moderates banking sector development to increase the intensity of environmental degradation. The technological effect of banking sector development is, therefore, harmful to environmental quality. Additionally, the findings of the Dumitrescu Hurlin panel Granger non-causality tests showed that population, affluence, and banking sector development could all be used to predict future environmental quality in the sub-region.

In view of policy ramifications, ECOWAS nations must promote banking sector development by implementing policies intended to enhance their banking institutions and the financial sector as a whole since this has the potential to lessen environmental degradation. Considering the harmful technological effect of banking sector development on environmental quality, it would be crucial for banking institutions to ensure that funds are made available for firms at a cheaper rate to encourage them to adopt energy-efficient technologies. Governments can also support innovative ideas to install ecologically friendly technologies. This will undoubtedly boost the use of contemporary clean energy instead of non-renewable sources. Future environmental regulations should also mandate that businesses and industries reveal their environmental performance. The study also suggests that firms in the nations, especially large multinational firms, integrate sustainable development practices into their reporting cycle per SDG 12. Environmental regulators could also employ other tools to mitigate environmental degradation including emissions trading, caps, and taxes. Also, it is necessary to control population growth to keep it from exceeding the ecosystem’s carrying capacity. This can be done by creating a demographic policy. Finally, ensuring that country-specific heterogeneous effects are considered when executing an environmental quality strategy is essential. The ramifications of policy extend beyond the ECOWAS nations to other developing regions. If considered, the abovementioned policy ramifications will assist ECOWAS in addressing SDG 13, which addresses climate change and its dire effects.

Notwithstanding the considerable methodological and policy ramifications, this study has shortcomings worth addressing. ECOWAS countries were the primary focus of the paper. However, as some other countries are also dealing with high population growth, rising levels of wealth, and advanced technology, the underlying analytical framework and line of inquiry may be extended to other countries, particularly emerging economies and regions. The study only focused on how bank-based financial development influenced environmental degradation in ECOWAS. It would be interesting to comprehend how market-based development and the overall development of the financial sector also influence environmental quality in ECOWAS.