Sustainable Energy and Financial Stability in European OECD Countries: An Analysis Based on GMM Dynamic Panel Estimation

Abstract

This study explores the effect of the energy transition on financial stability in the context of 13 OECD countries during the period from 2009 to 2019. In this sense, the soundness of the financial system is expressed through two dimensions: the Zscore and the volume of non-performing loans (NPLs). Using a dynamic panel estimation with the Generalized Method of Moments (GMM), the results highlight the complex effects of the energy transition on financial stability. Switching from fossil to clean energy improves the Zscore and reduces NPLs. In addition, the study reveals heterogeneous impacts depending on the renewable energy source involved. In fact, wind energy makes a positive contribution to both dimensions of financial stability. By linking the dynamics of the energy transition with the resilience of the banking sector, this study provides new insights into how sustainable energy policies can foster long-term financial sustainability. The effects of solar power and hydroelectricity, while positive overall, are not without nuances. Specifically, the former reduces the NPLs but also the Zscore, while the latter has the opposite effect on both aspects of financial stability. At this point, it is crucial to take into account the varying effects of different renewable energy sources when assessing the financial repercussions of the energy transition.

1. Introduction

In recent decades, climate change has become increasingly serious. The National Centers for Environmental Information have classified 2023 as the warmest year since 1850, i.e., the record start date. Its annual global report highlights the increase in the global land and ocean temperature, which is higher than the 20th-century average [1]. This context is associated with severe climatic events that generate negative externalities for the economic cycle. These facts have led central banks to consider climate risk in their assessment of financial stability. As fossil-based energy is often pinpointed as the main source, a reflection on appropriate energy models is required. In many contexts, the renewable energy sector has become more and more important. According to the Energy Institute statistics, the renewable energy consumption share of total energy increased from 13.43% to 22.26% in Norway and from 5.77% to 15.6% in Portugal between 2009 and 2019 [2]. If such a transition can reduce the negative environmental externalities of fuel use, it raises the question of its economic and financial implications.

To answer the question, it is necessary to highlight the relationship between greenhouse gas (GHG) emissions and the environment on one side and the financial sector on the other. The latter is often presented as a determinant of the former. In several contexts, the literature confirms the existence of the Financial Market Environmental Kuznets Curve (FMEKC). In other words, it shows the negative effect of financial development on environmental degradation. For instance, refs. [3,4] highlight the positive impact of financial development on CO₂ emissions, and consequently, the negative effect on environmental quality. However, the inverse effect is also possible. In fact, a section of the literature highlights the development of a financial system as a preliminary stage in the decarbonization process.

In this respect, refs. [5,6] underline the financial constraints of decarbonization. In this sense, financial development is a determinant of environmental quality, as shown by [7]. They suggest that the increase in the degree of openness and financial liberalization reduces environmental degradation. The results are explained by the positive effect of financial development on research and development and subsequently on the mitigation of CO₂ emissions. In other contexts, similar results have been found by [8,9,10,11].

Moreover, the opposite direction of the relationship, i.e., the effect of CO₂ emissions reduction policies on the financial system, remains possible. For example, ref. [12] show the negative relationship between GHG emissions and corporate financial performance. In addition, ref. [13] have demonstrated that the sustainable net-zero economy promotes financial performance by reducing costs [14]. Consequently, the return on assets (ROA) increases [12]. Therefore, the capacity of firms to repay their debts can be enhanced, and the default risk decreases. In summary, restricting GHG emissions can promote financial stability. In this sense, ref. [15] show a positive impact of physical risk on financial instability. At this juncture, the objective of financial stability can be achieved by reducing the risk associated with climate change, namely, greenhouse gas emissions.

In order for this transition to be successfully completed, the implementation of a clean energy model, as opposed to a fuel-based one, is encouraged. Formally, this is known as the energy transition. Nevertheless, the transition process can have a negative impact on the financial sector [16]. Furthermore, incentive policies for the energy transition, such as regulating CO₂ emissions, can reduce the value of firms. Carbon-intensive firms are the most affected by this effect [17]. Similarly, ref. [18] demonstrate that decarbonization policies significantly elevate the default risk of high-carbon enterprises. This, therefore, gives rise to a conflicting effect of energy transition. While it has the potential to mitigate greenhouse gas emissions and enhance financial stability, it could also counteract these beneficial outcomes. From a sustainability perspective, understanding how the energy transition simultaneously supports environmental goals and preserves financial stability is essential to achieving a resilient and inclusive low-carbon economy.

While a number of studies have highlighted the relationship between the financial system and the energy sector, certain aspects remain unexplored. More specifically, research into the implications of the energy transition in terms of financial stability remains limited. In addition, much of the literature focuses on the effects of transition incentives rather than the transition itself [19,20]. To the best of our knowledge, while the literature on climate risk and banks is growing [21,22,23], few studies have explicitly examined the direct effect of energy transition on financial stability. The present paper follows this line of research, but differs in several ways. First, it makes explicit the specific effect of renewable energy by source. Secondly, in addition to using the Zscore as in previous papers, it adds NPLs to represent financial stability. Methodologically, it adopts a GMM system method, whereas previous papers used an ARDL model and a logit/probit panel. While previous papers have considered the issue in the context of Asian, Sub-Saharan African, and emerging countries, the present study considers 13 OECD countries.

The remainder of the paper is as follows: Section 2 presents a brief literature review. Section 3 describes data and the econometric model. Section 4 analyses and discusses the results obtained. Section 5 offers concluding remarks.

2. Literature Review

The effect of the energy transition on the financial system, and more specifically on financial stability, involves two mechanisms. The first is that the adoption of the renewable energy model can have a direct impact on financial risk. The second highlights the indirect effect via GDP. In fact, the energy transition has an impact on GDP and, therefore, on companies’ ability to repay their debts and the quality of bank balance sheets.

2.1. Carbon Emissions, Energy Transition, and Financial Risk

The effect of the energy transition on financial risk remains ambiguous. While part of the literature reports a positive effect, other studies reveal a negative one. To identify this relationship, it is necessary to explore the link between financial risk and carbon emissions. In this context, the literature is wide-ranging. It distinguishes between two effects. The first links firms’ portfolios to carbon emissions. The second highlights the implications of the decarbonization policies and carbon regulations on firms’ financial position. In addition, the literature distinguishes between the response of high carbon-emitting industries and others.

Ref. [24] report the positive causality of carbon emissions on corporate credit default swap spreads in Japan. The results are explained by the “investor awareness” channel. Specifically, investors expect their profits to fall as a result of higher carbon taxation and increased regulatory costs [25]. In order to offset this possible decline, they require a higher risk premium for investments in sectors with a high level of carbon emissions. As a result, spreads differ across sectors and depend on the amount of carbon emitted. They are accentuated in sectors with high emissions. In summary, carbon emissions can have a positive impact on firms’ financing costs [26] and subsequently increase corporate default risk.

Along the same lines, ref. [27] found that the default probability increases in response to carbon regulation in China. The authors showed that the risk of environmental regulation, like carbon risk, is transmitted to the financial system through the debt-default channel. In fact, the energy transition is often associated with a set of regulatory measures such as carbon emission standards and carbon taxes. However, adapting to these constraints is not straightforward for companies and can result in increased costs and reduced profitability. As a result, repayment capacity tends to fall. In addition, banks compensate for carbon risk by charging a higher borrower rate for firms emitting carbon [28].

The same result was found by [25]. For a sample of 458 companies, they underlined a negative and significant effect of carbon emissions on the distance to default. More concretely, a 1 per cent increase in carbon emissions generates a 0.002 reduction in the distance to default. This was explained by the effect of carbon regulation. Following the Paris Agreement in 2015, climate policy became more stringent. Subsequently, the distance to default for high-carbon-emitting companies has become ever shorter. The aforementioned effect can also be explained by the energy transition channel. As the agreement requires the world to achieve zero net carbon emissions by 2050, the energy transition has become imperative. Consequently, the impact on financial risk remains conceivable.

The energy transition is driven by different factors. Some of these have already been explained above. In fact, regulation, i.e., decarbonization, is presented as a way of reducing carbon emissions. It encourages companies to use low-carbon technologies. As this measure is aimed at reducing carbon emissions and improving the quality of the environment, it appears to have a negative impact on the financial situation and the value of companies. At this point, debt costs appear to be rising [17]. This observation is often dictated by the effect of the carbon-reduction incentive policies on the probability of company default and on investors’ perception of risk. The effect is particularly marked in the high-carbon emissions sector. In this context, the measures related to the energy transition seem to have a negative impact on the probability of default. The increase in non-performing loans and financial risk is the resulting outcome. However, the opposite effect can be found in the literature. This is mainly the positive effect of the energy transition on the financial sector.

The model developed by [29] suggests that the adoption of renewable energy can improve the environmental dimension, as well as financial stability. In fact, an increase in the share of renewable energy in the overall energy supply reduces banks’ default risk [30]. While banks are financing the use of renewable energies, companies are becoming increasingly involved in the energy transition process. Furthermore, they meet the requirements of other stakeholders, such as the government. Since this situation reduces carbon emissions, it avoids penalties. In addition, it attracts government funding for environmentally friendly projects such as tax incentives and loan guarantees. Overall, companies’ profitability improves and their probability of default falls. As a result, banks’ balance sheets strengthen. Therefore, financial stability is improved.

Similar results were reported in [31]. The author points to the price of electricity as the link between the dynamics of the energy transition and financial stability. In other words, the use of renewable energies has an impact on electricity prices, which in turn are linked to financial stability [32]. The effect of electricity prices on the financial sector can be represented by two mechanisms. Firstly, a rise in electricity prices weighs on companies’ results. Subsequently, their likelihood of repaying loans falls. Therefore, banks’ profitability and balance sheets deteriorate. Secondly, the electricity price volatility leads to a high variability in the probability of repayment for businesses and banks’ balance sheets. However, the use of renewable energy can limit these two mechanisms. In fact, it reduces the level and volatility of electricity prices [33,34,35]. All in all, an increase in renewable energy can have a positive impact on financial stability.

In summary, there are diverging views on the implications of the energy transition for financial stability. While some authors point to a positive effect, others report a negative one. Notably, ref. [36] justified this discrepancy by the way the energy transition was conducted and the approach adopted. While a progressive and credible transition provides financial and economic stability, a disordered one has a damaging effect on the financial sector [37]. The former involves a gradual reduction in carbon emissions, according to a pre-announced schedule. It allows agents to distribute the cost of transition over a certain period. Thereafter, it allows aligning the response of the economy and ensures a flexible transition to a green economy. On the other hand, the second involves a sharp reduction in emissions in order to achieve a given target, which leads to a negative impact on financial stability.

The effect of the energy transition on the financial sector is also related to the nature of the implemented policies. In particular, ref. [36] prefer the introduction of additional regulatory capital in response to the increase in loans related to fossil energy rather than the support of green financing. The preference is justified by the benefits of conditional bank capital rather than rigid regulatory capital. In fact, the former allows banks to increase lending in a downturn. Subsequently, it allows the recovery of the economy and shortens the period of recession. In contrast, the second prevents banks from lending during a recession. As a result, the recession becomes increasingly severe.

2.2. Energy Transition, GDP and Financial Risk

The effect of the energy transition on financial stability can be represented by another channel through gross domestic product. It is structured in two stages. In the first, the energy transition impacts investment and gross domestic product [38,39]. In the second, GDP acts on non-performing loans, i.e., default risk [40,41,42,43,44,45]. As the second stage has been widely discussed in the literature, our attention is mainly focused on the first one.

In this regard, ref. [46] analyzed panel data from twenty Organization for Economic Cooperation and Development (OECD) countries between 1985 and 2005 and found a positive effect of renewable energy consumption on economic growth. In fact, a 1 per cent increase in the former generates an increase in GDP by 0.76 per cent. The impact of renewable energy on economic growth can be explained by its effect on real gross fixed capital formation and its implications on labor demand [47].

In the comparable economic context, ref. [48] found similar results. They analyzed annual data of nine developed countries: Canada, France, Japan, The Netherlands, Spain, Sweden, Switzerland, the United Kingdom and the United States. Therefore, they underlined the positive and significant effect of renewable energy technologies on economic growth. Similar results were reported for European countries. Using a time-varying fixed effects model, ref. [49] demonstrated a positive effect of renewable energy on economic growth.

In contrast, ref. [50] found opposing results for an African country, namely Ghana. They not only verified the effect of renewable energy on economic growth, but also highlighted the interdependence between the two variables. In other words, they argued that economic growth increases clean energy consumption. However, the feedback effect remains possible and significant. In more concrete terms, a 10 percent increase in renewable energy consumption decreases GDP by 3.97 percent. Although the consumption of renewable energy improves direct foreign investment, no indirect effect on economic growth was observed through this mediating variable.

According to the above, the impact of renewable energy consumption on GDP growth is not conclusive. For example, ref. [51] found an ambiguous effect. In fact, they highlighted a positive effect for 23 countries and a negative one for 9 countries. The disparity can be explained by country-level risk factors such as composite risk, political risk, financial risk, and economic risk [52]. In other words, the effect of renewable energy on the economic sector differs within economies depending on the level of risk. In general, a low level of country risk allows a greater contribution from renewable energy to economic development. More concretely, in a context of high-country risk, a 1% positive shock in renewable energy consumption generates a 0.0892% rise in economic growth. Whereas the equivalent increase in a low-risk context generates an increase of 0.0204% [52]. In summary, the contribution of renewable energy consumption to economic growth is contingent on the country’s level of risk.

Similarly, ref. [53] emphasized the non-linearity of the effect of renewable energy consumption on growth. They showed that the positive effect, in developing economies, is conditional on reaching a renewable energy consumption threshold. Under this threshold, the effect of renewable energy on growth remains negative. In brief, the indirect effect of the energy transition on the financial sector through economic growth remains ambiguous. It depends on the effectiveness of the impact of renewable energy consumption on economic growth. While the previous impact is positive in some contexts, it is negative in others. This paper attempts to extend the discussion in this direction and to contribute to understanding the nature of the implications of the energy transition on the economic sector.

3. Research Methodology

This section describes the data and presents the econometric method used to answer the research question.

3.1. Data

This paper is designed to answer the question of the implications of the energy transition for financial stability in a panel of OECD countries. It uses annual data from 2009 to 2019 for 13 countries: Belgium, France, Germany, Hungary, Ireland, Italy, The Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the United Kingdom. The period was carefully chosen to cover the recovery period following 2008. The analysis period was chosen based on data availability and to exclude the COVID-19 period, which could bias the results. As such, it incorporates the growth of climate policies and renewable energies, as well as providing comparable data within a relatively stable macroeconomic context. The choice of OECD European countries is justified by their progress in renewable energies and their relatively stable and well-regulated financial systems.

Table 1. Variables description.

Variable	Notation	Definition	Source
Financial stability
Zscore	Zscore	Bank Zscore	World Bank
Non-Performing Loans Ratio	npls	Bank nonperforming loans to total gross loans are the value of nonperforming loans divided by the total value of the loan portfolio (%)	International Monetary Fund
Macroeconomic conditions
Inflation	inf	Measured as annual percentage change in consumer price index (%)	International Monetary Fund
Real Gross domestic product per capita	rgdp	Gross Domestic Product divided by midyear population (Constant Local Currency Unit)	World Bank
Energy
Clean Energy	clean	Electricity generation from clean sources: wind, solar and biological (TWh)	International Energy Agency
Solar Energy	Solar	Electricity generation from solar (TWh)
Wind Energy	Wind	Electricity generation from wind (TWh)
Hydro Energy	hydro	Electricity generation from hydro (TWh)
Fossil Energy	fossil	Electricity generation from fossil sources (TWh)
Environment Conditions
CO2 Emissions	CO2	CO2 emissions in metric tons per capita	International Energy Agency
Banking Conditions
Capital Adequacy	capital	Bank regulatory capital to risk-weighted assets (%)	International Monetary Fund

Source: author’s development.

provides a detailed description of the variables used in this study, including their definitions, sources, and units of measurement. Two variables, Zscore and npls, are used to proxy financial stability as in [54,55,56,57,58,59,60,61]. The energy transition involves replacing fossil energies, which are responsible for climate change and pollution, with clean, renewable energies. In this respect, the energy transition is proxied by the variables: clean, solar, wind, hydro and fossil. This refers, respectively, to the volume of electricity generated from these sources: clean, solar, wind, hydro and fossil sources. Solar, wind and hydro power are chosen because of their significant contribution to renewable electricity generation capacity [62]. Although the energy transition concerns both production and consumption, this paper examines only the former. As financial stability cannot be attributed solely to the energy transition, macroeconomic conditions have been taken into account. These are rgdp and inf. The effect of the quality of the environment is measured by the variable CO₂. With the exception of inf and Zscore, a logarithmic transformation of the variables was applied in the econometric model. There are two reasons for this transformation: to reduce the scale of the variables and to allow the ability to interpret the estimated coefficients as elasticities. Finally, the ratio of non-performing loans and Zscore are considered to be endogenous variables while the remaining variables are treated as exogenous.

Table 2. Statistical summary of variables.

Variable	Obs	Mean	Std. dev.	Min	Max	Skewness (p Value)	Kurtosis (p Value)
Zscore	143	14.50746	6.74729	0.017333	37.7402	1.29277 (0.0000)	5.52945 (0.0003)
Ln(npls)	143	5.41812	5.52213	0.494908	25.7085	1.51978 (0.0000)	4.67732 (0.0040)
ln(wind)	143	1.85885	1.81055	−3.77226	4.83544	−0.95409 (0.0000)	3.59079 (0.1327)
ln(solar)	143	−0.58320	3.07225	−7.76522	3.81156	−0.58723 (0.0049)	2.41396 (0.0731)
ln(hydro)	143	1.96526	2.39540	−2.86470	4.96149	−0.62122 (0.0031)	1.88198 (0.0000)
ln(clean)	143	4.16071	1.21166	1.31372	6.27383	−0.24252 (0.2208)	2.32945 (0.0256)
ln(fossil)	143	3.95827	1.57592	0.615185	5.94135	−0.744243 (0.000)	2.42021 (0.0781)
ln(co2)	143	1.82459	0.28413	1.22424	2.33198	0.056969 (0.7711)	1.69151 (0.0000)
ln(rgdp)	143	11.21952	1.51630	9.72616	15.2537	1.41585 (0.0000)	3.79251 (0.0684)
Inf	143	1.27394	1.31879	−4.44754	5.65214	−0.04166 (0.8315)	5.11854 (0.0011)
capital	143	17.06079	3.75954	9.780442	26.9408	0.472697 (0.0207)	2.79822 (0.7674)

Source: author’s development.

shows the descriptive statistics for the variables used in this study. It includes the mean, standard deviation, minimum, maximum, as well as skewness and kurtosis measures, providing an overview of the distribution and dispersion of the data. The analysis reveals that the dataset exhibits a significant degree of skewness and mesokurtosis in the distribution of the variables. Specifically, 70% of the variables are skewed, 90% are mesokurtic, and 60% exhibit both characteristics.

3.2. Econometric Model

To address the research question, a panel regression analysis was employed within a system Generalized Method of Moments (GMM) approach. This approach was selected for its ability to control for endogeneity problems. The endogeneity problem is exacerbated in the context of the energy transition because of the multitude of factors at play, which can lead to biases due to omitted variables [63]. Similarly, it gives rise to heteroskedasticity and autocorrelation problems. This provides additional justification for the use of GMM in this study. Among the different versions of GMM, we have chosen the two-step system GMM for this analysis. Due to its higher efficiency, the System GMM is generally preferred over the Difference GMM. The latter involves differencing the model equations to eliminate unobserved fixed effects and using lagged values of the endogenous variables as instruments [64]. The former combines level and difference equations, allowing the use of a more extensive range of instruments [65,66,67]. It enables the System GMM to extract more information from panel data, particularly when the variables are highly persistent. The result is more accurate and robust estimators, with a reduction in finite-sample bias [67]. Compared to the one-step version, the use of two-step GMM can further enhance the robustness of the estimation [68].

Our study aims to measure the effect of the energy transition on financial stability. To this end, we developed six separate econometric models. These models analyze the impact of the volume of renewable energy production on indicators of financial health, such as the rate of non-performing loans and the Zscore. We have also extended the analysis by studying the effect of each source of renewable energy (hydroelectric, wind and solar) separately, in order to gain a better understanding of the specific mechanisms of the energy transition. In order to isolate the specific effect of the energy transition, we control for the influence of other macroeconomic, regulatory and environmental factors. The inclusion of $co2$ emissions make it possible to capture the indirect effects of the energy transition on financial stability, in particular those linked to climate policies and transition risks. In addition, we considered $rgdp$ to be a moderating variable. It appears to have an effect on $co2$ emitted and subsequently on financial stability. The term ${ln(co2}_t)\ast {ln(rgdp}_t)$ represents the joint effect of $co2$ and $rgdp$, respectively, on $zscore$ and $NPLs$. This suggests that the effect of environmental pressures on financial stability may be influenced by the level of economic development, as in [69]. In order to account for the effect of banking regulation on financial stability, we have included the variable ‘capital’, representing banking capital, in all our models as a control variable. Some potentially influential variables, such as bank liquidity, credit growth, or additional macroeconomic indicators, were not included in order to avoid model overfitting and small-sample bias, which could render the estimates unstable or distort significance tests. By instrumentalizing endogenous variables with lagged dependent variables and a diverse set of instruments, we address potential endogeneity bias. The validity of this instrumental variable strategy is evaluated using the Hansen and the Arellano–Bond $AR\left(2\right)$ tests.

To answer our research question, we built a set of econometric models. Models 1 and 2 (respectively, 4 and 5) estimate the effect of the energy transition on financial stability, using $NPLs$ (respectively, $zscore$) as proxies. They assess the consequences of switching from a fossil fuel energy model to a renewable energy model. To better understand the underlying mechanisms, models 3 and 6 focus on the specific impact of different renewable energy technologies (solar, hydroelectric and wind) on these same indicators. These econometric models are presented in the following equations:

$$\begin{array}{cc}\hfill \mathrm{l}\mathrm{n}\left({NPLs}_t\right)={\alpha }_1& + {\beta }_{10}\mathrm{l}\mathrm{n}\left({NPLs}_{t-1}\right)+{\beta }_{11}{\mathrm{l}\mathrm{n}(Clean}_t)+{\beta }_{12}{inf}_t+{\beta }_{13}{\mathrm{l}\mathrm{n}(co2}_t)\hfill \\ \hfill & + {\beta }_{14}{\mathrm{l}\mathrm{n}(rgdp}_t)+{\beta }_{15}{capital}_t+{\beta }_{16}{ln(co2}_t)\ast {ln(rgdp}_t)\hfill \\ \hfill & + {\epsilon }_{1t}\hfill \end{array}$$

(1)

$$\begin{array}{cc}\hfill \mathrm{l}\mathrm{n}\left({NPLs}_t\right)={\alpha }_2& + {\beta }_{20}\mathrm{l}\mathrm{n}\left({NPLs}_{t-1}\right)+{\beta }_{21}{\mathrm{l}\mathrm{n}(fossil}_t)+{\beta }_{22}{inf}_t+{\beta }_{23}{\mathrm{l}\mathrm{n}(co2}_t)\hfill \\ \hfill & + {\beta }_{24}{\mathrm{l}\mathrm{n}(rgdp}_t)+{\beta }_{25}{capital}_t+{\beta }_{26}{ln(co2}_t)\ast {ln(rgdp}_t)\hfill \\ \hfill & + {\epsilon }_{2t}\hfill \end{array}$$

(2)

$$\begin{array}{cc}\hfill \mathrm{l}\mathrm{n}\left({NPLs}_t\right)={\alpha }_3& + {\beta }_{30}\mathrm{l}\mathrm{n}\left({NPLs}_{t-1}\right)+{\beta }_{31}{\mathrm{l}\mathrm{n}(solar}_t)+{{\beta }_{32}\mathrm{l}\mathrm{n}(wind}_t)\hfill \\ \hfill & + {{\beta }_{33}\mathrm{l}\mathrm{n}(hydro}_t)+{\beta }_{34}{inf}_t+{\beta }_{35}{\mathrm{l}\mathrm{n}(co2}_t)+{\beta }_{36}{\mathrm{l}\mathrm{n}(rgdp}_t)\hfill \\ \hfill & + {\beta }_{37}{capital}_t+{\beta }_{36}{ln(co2}_t)\ast {ln(rgdp}_t)+{\epsilon }_{3t}\hfill \end{array}$$

(3)

$$\begin{array}{cc}\hfill {zscore}_t={\alpha }_4 +& {\beta }_{40}{zscore}_{t-1}+{\beta }_{41}{ln(Clean}_t)+{\beta }_{42}{inf}_t+{\beta }_{43}{ln(co2}_t)\hfill \\ \hfill & + {\beta }_{44}{ln(rgdp}_t)+{\beta }_{45}{capital}_t+{\beta }_{46}{ln(co2}_t)\ast {ln(rgdp}_t)\hfill \\ \hfill & + {\epsilon }_{4t}\hfill \end{array}$$

(4)

$$\begin{array}{cc}\hfill {zscore}_t={\alpha }_5+ & {\beta }_{50}{zscore}_{t-1}+{\beta }_{51}{ln(fossil}_t)+{\beta }_{52}{inf}_t+{\beta }_{53}{ln(co2}_t)\hfill \\ \hfill & + {\beta }_{54}{ln(rgdp}_t)+{\beta }_{56}{capital}_t+{\beta }_{56}{ln(co2}_t)\ast {ln(rgdp}_t)\hfill \\ \hfill & + {\epsilon }_{5t}\hfill \end{array}$$

(5)

$$\begin{array}{cc}\hfill {zscore}_t={\alpha }_6+ & {\beta }_{60}{zscore}_{t-1}+{\beta }_{61}{ln(solar}_t)+{{\beta }_{62}\mathrm{l}\mathrm{n}(wind}_t)\hfill \\ \hfill & + {{\beta }_{63}\mathrm{l}\mathrm{n}(hydro}_t)+{\beta }_{64}{inf}_t+{\beta }_{65}{ln(co2}_t)+{\beta }_{66}{ln(rgdp}_t)\hfill \\ \hfill & + {\beta }_{67}{capital}_t+{\beta }_{66}{ln(co2}_t)\ast {ln(rgdp}_t)+{\epsilon }_{6t}\hfill \end{array}$$

(6)

4. Results

The present section presents the empirical results of the energy transition’s impact on financial stability in OECD countries. To identify this effect, we consider the effect of the decrease (respectively, increase) in renewable (respectively, fossil) energy production on NPLs and Zscore. In this contexte, it uses a dynamic GMM specification to control the endogeneity and serial correlations in the data.

4.1. Baseline Model

Table 3. Baseline Models.

Dependent Variable		ln(npls)			Zscore
Independent Variable	M1	M2	M3	M4	M5	M6
Lag (1)	0.401 *** (0.000)	0.346 *** (0.000)	0.337 *** (0.000)	−0.135 ** (0.049)	0.073 (0.400)	0.327 *** (0.000)
ln(clean)	−0.523 *** (0.000)			3.084 *** (0.000)
ln(fossil)		0.252 *** (0.002)			0.044 (0.932)
ln(hydro)			−0.246 *** (0.000)			1.511 *** (0.000)
ln(wind)			0.118 ** (0.035)			2.101 *** (0.000)
ln(solar)			−0.072 ** (0.019)			−0.563 *** (0.001)
ln(gdp)	0.881 *** (0.007)	0.718 * (0.076)	−0.975 ** (0.030)	0.667 (0.719)	2.491 (0.285)	9.189 *** (0.000)
ln(co2)	7.122 *** (0.002)	3.949 (0.149)	−6.514 ** (0.041)	−8.820 (0.487)	8.981 (0.566)	51.503 *** (0.001)
capital	−0.048 ** (0.007)	−0.039 * (0.079)	−0.067 *** (0.001)	0.682 *** (0.000)	0.613 *** (0.000)	0.431 *** (0.000)
Inf	0.129 ** (0.014)	−0.049 (0.491)	0.000 (0.990)	−0.761 ** (0.018)	−0.157 (0.711)	−0.169 (0.529)
ln(co2) × ln(gdp)	−0.730 *** (0.000)	−0.443 * (0.064)	0.462 * (0.089)	−0.270 (0.809)	−2.066 (0.135)	−5.400 *** (0.000)
_cons	−4.473 (0.223)	−5.790 (0.210)	15.570 *** (0.004)	6.799 (0.745)	0.508 (0.985)	−92.42 *** (0.000)
R Adjusted	0.528	0.279	0.429	0.609	0.389	0.734
F Statistic (p-value)	21.98 (0.000)	8.28 (0.000)	11.97 (0.000)	30.22 (0.000)	12.94 (0.000)	41.25 (0.000)

Source: author’s development, ***: significant at 1%, **: significant at 5%, *: significant at 10%.

summarizes the baseline models. Ordinary least squares (OLS) estimation is used as the estimation method. Our results provide evidence of a positive effect of renewable energy on financial stability. In particular, we find that renewable energy improves Zscore and reduces NPLs, while the reduction in the use of fossil fuels significantly decreases the volume of NPLs. However, the effect of this reduction on the Zscore is not significant. In addition, the analysis highlights the effect of renewable energies according to their source. It demonstrates the positive effect of hydroelectric power on financial stability, with a negative effect on NPLs and a positive one on the Zscore. Despite its positive impact on NPLs, solar energy improves financial stability, as illustrated by its significant positive effect on the Zscore. In other words, as it increases the credit default probability, it improves banks’ ability to absorb this rise. On the other hand, solar energy improves loan repayment, but reduces banks’ ability to absorb credit losses.

OLS regression was estimated to obtain an initial assessment of the relationships between the variables. However, OLS assumes strict exogeneity of the explanatory variables, which may be questioned in a panel context, where endogeneity problems may arise, notably due to reverse causality, omitted effects or unobserved heterogeneity. To surmount these limitations, the Generalized Method of Moments (GMM) is recommended, as it allows potentially endogenous variables to be instrumented by using lagged values as instruments. Furthermore, GMM accounts for heteroscedasticity and error autocorrelation, making estimates more robust. In this way, the step from OLS to GMM improves the reliability of results by reducing endogeneity bias and ensuring more accurate inferences.

4.2. Cross-Sectional Dependency

In order to proceed with GMM model estimation, it is essential to examine the stationarity of the variables [70]. However, it is also crucial to determine the test hypothesis vis-à-vis cross-sectional dependence (CSD). To this end, cross-sectional dependence tests were applied [71]. The results are shown in

Table 4. Cross-sectional dependence test.

Variable	CD-Test	p-Value
Zscore	20.195 ***	0.000
npls	13.649 ***	0.000
ln(wind)	25.505 ***	0.000
ln(solar)	25.752 ***	0.000
ln(hydro)	3.702 ***	0.000
ln(clean)	13.979 ***	0.000
ln(fossil)	11.657 ***	0.000
ln(co2)	14.754 ***	0.000
ln(rgdp)	23.812 ***	0.000
Inf	13.708 ***	0.000
capital	21.028 ***	0.000

Source: author’s development, ***: significant at 1%.

. The latter shows the presence of statistically significant unobserved common shocks, indicating the existence of a CSD among all the variables.

4.3. Unit Root and Multicollinearity Tests

Given that the cross-sectional dependence hypothesis is confirmed, we proceed to the second-generation stationarity test, notably CADF and CIPS. This type of test, in contrast to traditional unit root tests, allows us to account for cross-sectional dependency of panel data. The results are shown in

Table 5. Unit root test.

	Level			First Difference
Variable	Deterministic Parameters	CADF	CIPS	Deterministic Parameters	CADF	CIPS
Zscore	CT	−3.131 ***	−2.569	C	−3.134 ***	−3.417 ***
ln(Npls)	CT	−2.197	−2.218	C	−2.350 *	−2.638 **
ln(wind)	CT	−3.356 ***	−2.450	C	−3.021 ***	−2.796 ***
ln(solar)	CT	−2.310	−2.489	C	−2.200 *	−2.706 ***
ln(hydro)	CT	−3.420 ***	−3.143 ***	C	−3.754 ***	−3.756 ***
ln(clean)	CT	−2.401	−3.134 ***	C	−2.295 *	−3.536 ***
ln(fossil)	CT	−2.202	−2.894 **	C	−2.406 **	−3.756 ***
ln(co2)	CT	−2.397	−2.788 *	C	−2.513 **	−3.021 ***
ln(rgdp)	CT	−2.221	−2.299	C	−2.210 *	−2.546 **
Inf	CT	−3.160 ***	−3.012 **	C	−2.886 ***	−3.477 ***
capital	CT	−2.723 *	−2.968 **	C	−2.542 **	−3.391 ***

Source: author’s development, ***: significant at 1%, **: significant at 5%, *: significant at 10%.

. It is clear that certain variables are not stationary at level, such as ln(Npls) ln(solar) and ln(rgdp), using both the CADF and CIPS tests. However, all variables are stationary after the first difference.

It is equally relevant to test the multicollinearity between the independent variables.

Table 6. The variance inflation factor (VIF).

Variable	M1 and M4 (VIF)	M2 and M5 (VIF)	M3 and M6 (VIF)
ln(gdp)	2.14	1.64	1.51
ln(solar)	1.70	-	-
ln(wind)	1.67	-	-
ln(co2)	1.45	1.50	1.30
ln(hydro)	1.35	-	-
inf	1.23	1.30	1.14
capital	1.15	1.17	1.09
ln(fossil)	-	2.09	-
ln(clean)	-	-	1.24
Mean VIF	1.53	1.54	1.25

Source: author’s development.

shows the variance inflation factor (VIF) values for each variable. Since the VIF values are all less than five, there is no multicollinearity issue [72].

4.4. GMM Estimation Results

Table 7. The results of GMM estimation.

	(1) Npls	(2) Npls	(3) npls	(4) Zscore	(5) Zscore	(6) Zscore
Lag (1)	0.090 *	0.478 ***	0.273 ***	0.059	−0.161 **	0.368 ***
	(0.094)	(0.000)	(0.000)	(0.396)	(0.020)	(0.002)
ln(clean)		−0.564 ***			3.036 ***
		(0.000)			(0.000)
ln(fossil)	0.938 ***			0.983 *
	(0.001)			(0.066)
ln(hydro)			−0.297 ***			1.799 ***
			(0.000)			(0.001)
ln(wind)			0.143 **			2.163 ***
			(0.004)			(0.000)
ln(solar)			−0.070 *			−0.853 ***
			(0.084)			(0.001)
ln(gdp)	1.537 *	0.628 ***	−1.608 ***	4.870 **	1.817 **	12.665 ***
	(0.062)	(0.001)	(0.000)	(0.020)	(0.046)	(0.000)
ln(co2)	4.686	5.470 ***	−10.584 ***	16.638	−1.645	77.215 ***
	(0.329)	(0.000)	(0.000)	(0.192)	(0.703)	(0.000)
capital	−0.035	−0.077 ***	−0.072 ***	0.653 ***	0.740 ***	0.401 ***
	(0.528)	(0.003)	(−0.007)	(0.007)	(0.000)	(0.002)
inf	−1.213 **	0.139 **	−0.021	−2.975 *	−1.001 *	−7.640 ***
	(0.030)	(0.041)	(0.802)	(0.054)	(0.052)	(0.981)
ln(co2) × ln(gdp)	−0.585	−0.594 ***	0.843 ***	−2.208 **	−0.747 **	−0.006
	(0.203)	(0.000)	(0.000)	(0.017)	(0.037)	(0.000)
_cons	−14.533 *	−0.832	22.620 ***	−23.456	−4.744	−133.944 ***
	(0.080)	(0.520)	(0.000)	(0.184)	(0.563)	(0.000)
AR(1)	0.008	0.006	0.002	0.002	0.002	0.005
AR(2)	0.163	0.374	0.790	0.188	0.154	0.759
Hansen test p-value	0.252	0.296	0.149	0.213	0.820	0.804

Source: author’s development, ***: significant at 1%, **: significant at 5%, *: significant at 10%.

presents a dynamic GMM estimate of the impact of the energy transition on financial stability. The former is captured by the change in fossil and renewable energy production. The latter is expressed in terms of bank credit default risk, using NPLs, and bank resiliency, in terms of the Zscore. While several diagnostic criteria are commonly used to assess the validity of dynamic GMM specifications, the Arellano-Bond autocorrelation test (AR2) is considered more relevant and robust in our model. To assess the validity of the instruments in the models, the Hansen test is used, which checks whether the instruments used are exogenous, i.e., whether they are correlated with the model errors.

4.4.1. Effect of Renewable Energy Deployment on NPLs and Zscore

The use of renewable energy, represented by the variable ln(clean), has a negative and significant effect on NPLs (−0.5644), indicating that increasing the amount of renewable energy reduces the probability of credit default. Concerning financial stability, as illustrated by the Zscore, the effect is positive (0.9833), suggesting that expanding renewables makes banks more resilient. All in all, these findings confirm that the energy transition to renewable sources contributes to a more stable financial sector. In this regard, they support the findings of [29,30,31,73] reporting that increasing the share of renewables in the energy mix can reduce financial risk. This effect is, in many cases, driven by the government incentives (such as subsidized loans and tax credits) that boost the profitability of firms and improve their solvency.

4.4.2. Effect of Fossil Energy Reductions on NPLs and Zscore

Reducing fossil energy intensity, quantified by the coefficient of ln(fossil), shows a conflicting effect on NPLs and bank Zscores. Firstly, there is a significant reduction in NPL associated with the reduction in fossil energy intensity. This result indicates that reduced exposure to fossil energy leads to a lower risk of credit default. Indeed, fossil energy is often considered a high-risk sector given its sensitivity to commodity price fluctuations and regulatory changes. Secondly, the effect on bank financial stability, as measured by the Zscore, is much more complex. Model 5 shows a positive relationship between the reduction in combustible energies and an increase in bank vulnerability. This may be explained by the fact that the disinvestment process from fossil energies exposes banks to new risks, particularly if they are faced with non-profitable investments or a shift in their assets towards emerging, less well-established and higher-risk sectors. This switch could temporarily destabilize banks if they fail to adapt their asset portfolios rapidly enough [36,37].

4.4.3. Effect of Specific Renewable Energy Sources on NPLs and Zscore

According to the results of models 1, 2 and 3, the effect of the energy transition on financial stability is relatively benign. This implies the question of the most beneficial energy source in terms of financial stability. Models 3 and 6 answer this question. With regard to financial stability in Zscore, hydropower and wind energy represent a significant positive effect. Moreover, the former is less efficient than the latter. This contrasts with [74] who show that hydroelectric energy is more efficient than wind energy in the context of OECD countries. However, the effect of solar energy remains significantly negative. This disparity in the contribution of renewable energies to financial stability, regardless of the technology involved, can be attributed to differences in energy efficiency.

However, when considering financial soundness from the point of view of NPLs, hydro and solar power improve financial stability by lowering NPLs. In addition, the effect of the former is relatively higher compared to the latter. This differential can be attributed to the comparative difference in efficiency between the two sources [75,76]. However, the effect of wind power on NPLs is positive. In this respect, and contrary to its positive impact on Zscore, it drives financial distortion. The mechanism behind this effect can be found in [77], which indicates that environmental, social, and governance (ESG) performance is positively associated with the default risk. Furthermore, this relationship follows an inverted U-shaped curve, the risk of failure decreasing when ESG performance exceeds a particular threshold.

For the fossil fuel and wind power sectors, the fact of observing both an increase in NPLs and an improvement in Zscore may seem counter-intuitive. However, this result can be explained by prudential and institutional mechanisms: certain energy exposures, although at the origin of non-performing loans, are lightly capital-weighted due to public guarantees or specific provisioning regimes. This means that banks can record more accounting defaults without any immediate deterioration in their soundness.

In the case of the solar sector, the opposite dynamic-declining non-performing loans and falling Zscores-may also come as a surprise. Solar energy projects are often subsidized or benefit from guaranteed tariffs, which reduces the risk of default. However, loans to these projects have limited margins, so the banks’ economic capital does not increase sufficiently, resulting in a lower Zscore.

In summary, the results reveal that an increase in renewable energies decreases NPLs and improves banks’ financial resilience, in particular through hydroelectricity and wind power. In contrast, reducing the use of fossil energies reduces NPLs but increases banks’ vulnerability. Finally, the impact of renewable energies varies according to their source, with hydropower being the most beneficial for financial stability.

5. Conclusions

This paper investigates the complex link between the energy transition and financial stability in a panel of 13 OECD countries between 2009 and 2019. Adopting a GMM systems approach to address issues of endogeneity and the dynamic panel nature of data, the impact of energy transition on key indicators of financial stability, non-performing loans (NPLs) and Zscores, is investigated. For this purpose, we have considered three designs for each of the two. In the first, we looked at the effect of lowering fossil energy. In the second, we examined the effect of increasing the use of clean energy, while in the third, we considered the impact of clean energy according to its source: hydro, wind and solar.

The GMM results, while based on the results of the baseline OLS model, demonstrate, in general, a positive effect of energy transition on financial stability. Excluding the downward effect of fossil energy reduction on Zscore, switching from fossil to clean energy improves Zscore and reduces NPLs. In addition, the study reveals heterogeneous impacts based on the respective renewable energy source. In fact, wind energy had a positive contribution to both dimensions of financial stability. The effects of solar and hydro power, while overall positive, showed some nuances. More specifically, the former reduces the default probability, i.e., NPLs, but also the resilience of banks, i.e., Zscore. The latter has the opposite effect on both aspects of financial stability. In other words, it enhances Zscore as well as non-performing loans. In general, these findings suggest that a well-planned energy transition contributes to financial stability and sustainability by reducing climate and financial risks while promoting sustainable economic and social development.

Given that the sample is restricted to 13 European OECD countries over the period 2009–2019, the findings should be interpreted as region-specific evidence rather than generalized global conclusions. Therefore, the external validity of the results is inherently limited and future research should extend the analysis to a broader set of economies.

In the last, the energy transition policies should provide better incentives for the deployment of renewable energy sources, particularly hydroelectricity and wind energy, which have been shown to promote financial stability. At the same time, they need to encourage continued investment in R&D, particularly in solar power technologies, so as to reinforce their efficiency and impact on financial soundness. Such polices are intended to smooth a gradual transition away from fossil energies, giving financial institutions sufficient time to adjust their portfolios and minimize their risk exposure to high-risk fossil assets.