How Is the ESG Reflected in European Financial Stability?

Abstract

Environmental, social, and governance (ESG) factors are increasingly analysed to identify the risks and opportunities in contemporary economies. The banking sector influences the whole economy through the credit channel and balances its stability. The interplay of these elements motivated our main question, whether ESG scores impact European financial stability, measured for the banking sector. To this aim, we employ the cross-quantilogram methodology, which explores dependences at all levels of the distributions of two random variables. To determine the quantile dependence, we resort to methods of measuring systemic risk (Marginal Expected Shortfall—MES, CoVaR, and ΔCoVaR) for all commercial banks listed on European stock exchanges. While our approach provides a dashboard for analysis of the dependence of financial stability on ESG pillars, our findings indicate that such a connection is valid and cannot be identified with standard approaches that explore average distribution levels. We also document the differences in these impacts across the ESG pillars.

1. Introduction

The turmoil and distress sometimes created in financial markets have led to a growing interest in investigating systemic risk, as an exponent for financial stability, in sundry ways. Analysing the specialised literature, we observed some main topics approached over the decades, which also developed depending on the major economic events. In this respect, we identified several pieces of research that were published in the 1980s on liquidity [1,2] or leverage [3], emphasising the importance of these elements for the banking system. In the next decade, the crises that arose in 1997 and 1998 contributed to the enrichment of the economic literature with the analysis of the contagion problem [4,5,6,7], which is a triggering element for financial instability. A new crisis that emerged in 2007–2008 and was continued by a sovereign debt crisis in Europe (2010–2012) has been reflected in blooming literature that addressed new topics of high importance for systemic risk, such as tail risk [8,9], or continued the research lines on liquidity [10,11,12], contagion [13,14,15], or amplification of risks [16,17]. The use of quantile regression [18] and entropy proceeding [19] was implemented to assess systemic risk.

While a widely accepted definition of systemic risk does not exist, after studying 220 published works on this issue over 35 years, Benoit et al. have concluded that, in a very general form, it can be acknowledged as being the risk of severe losses simultaneously felt by numerous financial market participants which then propagate through the entire system [20]. The capital requirement regulations valid in the European Union (known as CRR) [21] emphasise the need to mitigate systemic risks, defined in Article 3(1) of European Union Directive 2013/36 [22] as “a risk of disruption in the financial system with the potential to have serious negative consequences for the financial system and the real economy”, a similar view to that provided by Benoit et al.

On the other hand, in recent years, the ESG criteria have become more and more attractive for investors as they are used for companies’ evaluation. The financial risks associated with environmental and social issues may be identified and avoided. As reflected in the literature, the trends for energy demands are growing in modern economies [23]. According to the latest report, United Nations experts judged that the next few years (until 2030) are critical to narrowing down global warming, which could affect all components of the ESG; moreover, the social issue has been approached concerning reducing inequalities [24]. The financial risks related to climate and their connection with financial stability are being studied more intensively, involving new methodologies and approaches [25]. In the light of recent financial crises and the intensification of the inclusion of environmental, social, or governance criteria in the assessment of banks, the idea of an ethical bank is developing and is growing more than the conventional model [26].

Investing in ESG assets had a continuous ascending trend in previous years (USD 22.8 trillion in 2016, USD 30.6 trillion in 2018, and over USD 35 trillion in 2020), with an estimation of above USD 41 trillion in 2022 and around USD 50 trillion by 2025 [27]. The European market was a leader in this area until 2018, after which the U.S. market surpassed it, but it continues to be an essential player with unique features and perspectives. According to the latest report published by the European Securities and Markets Authority, which analyses the European market in 2020, equity, bond, and mixed funds that had an ESG strategy performed better than non-ESG ones [28]. Half of the listed companies in the old Member States of the European Union published information on ESG from 2002 to 2019 [29], allowing an assessment of their performance. Moreover, companies’ market capitalisation factor is directly related to the quality of ESG reporting [29], which is relevant information when deciding to analyse the influence of ESG on European financial stability in the banking sector.

An increasing amount of assets with ESG scores has facilitated research opportunities in the field. Some studies focus on risk and return assessments, others on investment strategies timelines, given that this kind of investment is oriented toward the long term. In the European banking context, the ESG principles have been strengthened by economic policies, as summarised by Bruno and Lagasio (2021) [30]. The authors also refer to the requirements of the European Banking Authority, as a consequence of the new banking regulatory package (CRR2 and CRDV). The growing relevance of ESG investment motivated us to study the relation between ESG scores and European financial stability, an important aspect when looking at the development of both parts. The interplay with the banking sector is crucial because actually, this influences the entire economy through the credit channel. Our main question is whether ESG scores have an impact on European financial stability, measured for the banking sector.

To capture some features of companies with ESG scores in relation with financial (in)stability, we may look at their performance in equilibrium (lower prospective returns, strengthened by the fact that investors look for a long-term investor and protection against climate risks) and when a positive shock from ESG factors occurs (better performance) [31]. Furthermore, in times with lower volatility for asset markets, the funds with high ESG scores have lower losses than those with low ESG scores; the authors conclude that contagions are less potent for the first category [32]. A similar conclusion was reinforced when applying neural architecture for mutual funds with different ESG scores: an increasing level of ESG helps to protect against contagion [33]. When looking at different scores of ESG for European banks, it was concluded that investors value the environmental implications more and do not appreciate the involvement of banks in social and governance aspects [34].

Several studies deal with the link between ESG scores and the stability of the European banking system but use methodological packages different from those proposed by this research. After implementing a regressive panel model for 243 European banks and their ESG performance from 2002 to 2018, Toth et al. concluded that the latter is associated with a decrease in non-performing loans and positively impacts regulatory capital [35]. Financial stability for banks is considered to be a function of specific financial indicators for profitability, liquidity, regulatory capital, etc., while the ESG scores are those available on the Refinitiv Eikon platform; the main conclusion is that ESG performance has a positive effect on financial stability in the banking sector [35]. Toth et al. suggested that a further direction for the research is the quantile approach, a methodology that we propose to address the issue of ESG scores and financial stability, filling in this way the theoretical gap existing in the literature.

The influence of ESG on bank stability was considered in the literature from the perspective of periods with financial disturbances. From this perspective, Chiaramonte et al. analysed the European banks’ stability under ESG scores, considered as a whole unit or as distinct elements, for the period 2005–2017. The crisis period was considered between 2008 and 2012, representing both the global financial crisis and the European debt crisis. To measure financial stability, a distance to default approach was chosen. The authors have shown that high ESG scores increase banks’ resilience in turbulent times, noting differences between banks depending on their characteristics and the environment in which they operate [36]. In an indirect way, in a study on risk-taking analysis and its relationship to ESG scores, the authors suggest that banks with ESG activity and small and diverse executive boards support financial stability [37].

The previous studies used different methods and information and their results do not reflect the same conclusions regarding the relation between ESG scores and financial stability. However, they stress that the great importance of systemic risk, the possibility of new data, the new regulations regarding ESG that determine banks to act differently, and the importance of further research using different methods are reasons for addressing this issue. Moreover, the importance of studying the connection between ESG and financial stability is the role played by ESG disclosure for increasing the stock prices informativeness [38].

Given this gap in the literature, we developed our research to test the following hypotheses:

Hypothesis 1 (H1). 

ESG scores of European banks listed on stock exchanges influence their financial stability.

Hypothesis 2 (H2). 

The influence of financial stability on ESG scores for the listed European banks is nonlinear.

Hypothesis 3 (H3). 

The impact rendered by the ESG scores on financial stability of European banks is varies across the E, S, and G pillars and it is different from the aggregated ESG score.

Given the above formulation of hypotheses, we can state that the main objective of this research is to determine the quantile dependence between ESG scores and measures of financial stability; we intend to prove that we have a clear dependency and also to produce a framework for the attempt to quantify it. To this end, we consider addressing this challenge by applying admitted methods of measuring systemic risk (Marginal Expected Shortfall—MES, CoVaR, and ΔCoVaR) [39,40] combined with a cross-quantilogram analysis to determine the connections following the methodology of Han, Linton, Oka, and Whang [41]. The analysis was conducted for all commercial banks listed on European exchanges, considering closing prices with daily frequency from January 2010 to February 2022. The special features of the European financial markets (more banking oriented, new ESG regulations) motivated us to choose this context for our analysis. Our key result is that the financial institutions’ systemic risk depends on previous ESG pillars levels in the European market, which is a validation of hypothesis 1. Extending our investigation, we also found evidence that validates both hypotheses H2 and H3.

After the presentation of the purpose of our research, the theoretical background and the main empirical evidence acknowledged in the literature the article is further organised as follows. Section 2 describes the data used and the methodology. Section 3 introduces the results, while in Section 4, findings are discussed, and we conclude.

2. Materials and Methods

As mentioned in the literature [42], the bank sector is prone to higher risk in terms of possible contagion that is finally negatively affecting its stability. This is occurring faster than in other sectors, has an ample spread, and is scattering beyond the financial sector.

After the crisis of 2007–2008, research intensified looking for practical methods of measuring systemic risk. The issue was addressed both in theory and practice, and is an aspect for which different controlling and regulating measures were adopted, with the final scope to enforce financial stability. New models emerged that take into account the contribution of individual parts to the systemic risk. Several methods such as z-score, distance from insolvency, or first-to-default probability were criticised for different reasons (contagion or interconnection is not taken into account, the difference between effects produced by small and big firms, etc.). On the other hand, some developed methodologies gained visibility and were furthermore applied for different aspect, periods, markets, or regions. Among them, Adrian and Brunnermeier [39,40] stood out through the recognition received from the academic community.

To achieve our research goals, firstly, we calculated the indicators constructed to measure the financial stability of 110 European banks, all listed on the European exchanges. Some of the frequently used standards to catch this stance are CoVaR and ΔCoVaR employed by Adrian and Brunnermeier [39]. Their method considers that the tail co-movement increase in case of financial disturbances, announcing increased systemic risk. In their vision, $CoVa{R}_q^{j|C\left(X^i\right)}$ represents the VaR for the entire system of companies taken into account, considering that the one for which the value is calculated is affected by a shock $C\left(X^i\right)$. In our case, the probability q to have the systems’ yieldingness below CoVaR is 1%:

$$Pr\left(X^j|C\left(X^i\right)\le CoVa{R}_q^{j|C\left(X^i\right)}\right)=1\%.$$

(1)

The difference between this variant of conditional CoVaR and that obtained for a usual situation represented by the median yield is called ΔCoVaR and is calculated using the formula:

$$\mathsf{\Delta }CoVa{R}_q^{j|i}=CoVa{R}_q^{j|X^i=Va{r}_q^i}-CoVa{R}_q^{j|X^i=Va{r}_{50}^i}$$

(2)

In addition, the methodology developed by Acharya, Pedersen, Philippon, and Richardson [40] was applied to assess financial stability. In this regard, MES was calculated applying the formula:

$$ME{S}_{\alpha }^i=\frac{\partial E{S}_{\alpha }}{\partial y_i}=-E\left[r_i|R\le -Va{R}_{\alpha }\right]$$

(3)

This indicator measures the mean of the marginal loss that appeared after a shock greater than a particular probability threshold, named expected shortfall (ES). This represents the mean of logarithmic returns below the calculated VaR value for α confidence level:

$$E{S}_{\alpha }=-E\left[E|R\le -Va{R}_{\alpha }\right].$$

(4)

In the second part of the research, a cross-quantilogram methodology developed by Han, Linton, Oka, and Whang [41] was adapted to evidence the connections between ESG scores and financial stability measured as described above. The methodology allows for quantifying the strength and the span of the connections between markets, for different lags and distributions. It was previously successfully implemented in financial market analysis; it was used to detect the connection between indices of stock markets and gold [43], emerging stock markets’ connectedness under COVID-19 influence [44], and tail dependence of main currencies [45]. Moreover, it was considered to analyse the green finance issues: the link between green bonds and different commodities [46], the connections between green and other stock under the influence of financial crisis [47], or green bonds and other investment assets [48], among others.

The returns of stocks are represented by the time series y_it, with i representing the returns of one asset (i = 1, 2) and t representing the time, with t = 1, ..., T. The F_i(·) and f_i(·) are the notations for the distribution and the density functions for the time series y_it, while α are the quantiles with a range α_i ∈ (0, 1) with the following function:

$$q_{it}\left({\mathsf{\alpha }}_i\right)=inf\left\{\vartheta :F_i\left(\vartheta \right)\ge {\mathsf{\alpha }}_i\right\}$$

(5)

The cross-quantilogram for two events described as $\left\{y_{1t}\le q_{1t}\left(\mathsf{\alpha }{\tau }_1\right)\right\}$ and $\left\{y_{2t-k}\le q_{2t-k}\left({\mathsf{\alpha }}_2\right)\right\}$, with lag k, and a quantile-hit process ${\phi }_a\left(u\right) \equiv 1\left[u<0\right]-a$, will take the form:

$${\rho }_{\tau }\left(k\right)=\frac{E \left[{\phi }_{{\mathsf{\alpha }}_1}\left(y_{1 t}-q_{1 t}\left({\mathsf{\alpha }}_1\right)\right){\phi }_{{\mathsf{\alpha }}_2}\left(y_{2 t-k}-q_{2 t-k}\left({\mathsf{\alpha }}_2\right)\right)\right]}{\sqrt{E\left[{\phi }_{{\mathsf{\alpha }}_1}^2\left(y_{1 t}-q_{1 t}\left({\mathsf{\alpha }}_1\right)\right)\right]} \sqrt{E \left[{\phi }_{{\mathsf{\alpha }}_2}^2\left(y_{2 t-k}-q_{2 t-k}\left({\mathsf{\alpha }}_2\right)\right)\right]}}.$$

(6)

The computations were performed for every component of ESG scores, i.e., environment (E), social (S), and governance (G) scores, and completed with the general ESG scores. The source of our data for these scores is Refinitiv Eikon, one of the largest and well positioned in the economic literature. Because of the standardisation process applied, the information may be compared across companies. Their final scores are obtained after including 186 different metrics, gathered in ten categories.

Each pillar (E, S, and G) has more components that receive scores based on the company’s achievements and have a certain weight in the calculation of the pillar score. In the environment pillar are three categories out of ten: resource use (related with the company’s reduction of using resources), emissions (refers to company’s willingness and ability to reduce emissions), and innovation (includes innovations oriented towards reducing environmental costs for clients). The social one has four categories, the largest number of items included in a pillar score, and a corresponding consistent weight: workforce (ensuring good working conditions, satisfaction, and opportunities for employees), human rights (compliance with international human rights regulations), community (respect for the community, promoting business ethics), and product responsibility (the company offers quality, safety, privacy). The last pillar, related to governance, includes management (observance of corporate governance principles), shareholders (shareholder compliance), and corporate social responsibility strategy (communication and implementation of economic, social, and environmental objectives in current activities).

In order to obtain the ESG score, each pillar acquires a certain weight, which varies depending on the industry where the company is active, except for the governance score, whose weight remains constant. As the calculation methodology shows [49], for the banking sector, the weights are 14.4% for environmental pillar, 49.6% for the social pillar, and 36% for the governance pillar. The scores take values from 0 to 100, the highest reflecting the peak level for ESG performance.

3. Results

We use stock market data from Datastream, covering closing prices with a daily frequency from January 2010 to February 2022, for all commercial banks listed on European exchanges. For these companies, we also queried the values corresponding to their environmental, social, and governance pillars. After data cleaning, which mainly consisted of eliminating those companies for which we did not have data for the whole time interval, we ended up with 110 companies.

Given the methodology published by Refinitiv Eikon [49], we computed values for the ESG indicator by using the weights of 14.4% for the environmental pillar, 49.6% for the social pillar, and 36% for the governance pillar.

The first step of our analysis consisted of the computation of financial stability for the commercial banks in our sample. As mentioned in the previous section on methodology, to proxy financial stability, we will use the main systemic risk values, i.e., CoVaR and ΔCoVaR, as in the methodology developed by Adrian and Brunnermeier [39] and MES as in Acharya, Pedersen, Philippon, and Richardson [40]. For the computation of these metrics, we used the STOXX 600 index for the same period, yielding daily values for these systemic risk measures. Figure 1 shows their dynamics as averages across all banks in our sample, at a daily frequency.

The values for the ESG pillars are only changing on a yearly basis, and therefore, we decided to move our analysis for the annual frequency. Consequently, for each bank in our database, we had five annual time series: on one hand, the yearly values for the systemic risk measures computed as averages across the daily data, and on the other hand, the annual indicators for the environmental, social, and governance pillars, to which we also added the composite ESG indicator. Figure 2 depicts the empirical distributions of these indicators at the annual frequency across all banks.

In order to test for the three hypotheses, the next step consisted of performing the cross-quantilogram methodology for each bank. We investigated the extent to which each of the E, S, G, and ESG indicators had an impact on the dynamics of the systemic risk measures (quantified separately by CoVaR, ΔCoVaR, and MES) for 10 values of quantiles (all percentiles every 10% ranging from 10% to 100%) for a two-lag dependence.

The choice of the cross-quantilogram methodology sheds light on the extent to which hypotheses H1 and H2 are valid. Finding significant results for any quantile validates the first hypothesis. Additionally, retrieving significance for the 50% quantiles (i.e., the median) corresponds to a linear dependence, while significance for other quantiles proves that this connection is nonlinear, which validates the second hypothesis. Given the complexity of nonlinearity, searching connections for any combination of the ten quantiles ensures that we investigate the dependence on a grid of 100 such possible dependences. We therefore cover almost the entire spectrum of nonlinear dependences to allow for the test of hypothesis H2.

Designing this research for each of the E, S, and G pillars as well as for the ESG score allows us to investigate the validity of hypothesis H3.

Therefore, we performed for each bank the cross-quantilogram methodology for the tenth percentile twelve times, i.e., each of the four ESG indicators on each of the three systemic risks.

Given a large number of computations, we present a synthesis of our results in the following figures.

Figure 3, Figure 4 and Figure 5 depict the number of significant connections obtained by applying the cross-quantilogram methodology for all pairs of quantiles and for each bank, as described above. Thus, Figure 3 shows these results for the CoVaR values for each bank in our database. We notice that the highest numbers of significant dependences are found for the 60% percentile in the environmental and social pillars on several percentiles for the CoVaR series. We found similar results for the ESG index. Given the one-lag analysis, we can conjecture that for levels of ESG at their 60% percentile, financial stability tends to react in the following year, at almost all quantiles. Therefore, the distribution of financial stability is dependent on the extent to which the ESG is situated at their 60% percentile. An important impact also happens when the ESG pillars are situated at their lowest values, under the 10% percentile. Looking at the number of significant situations, the social pillar tends to have the highest impact, which is also perceived when the analysis includes the ESG index.

Figure 4 reports the number of significant connections obtained by applying the same methodology for all pairs of quantiles and for each bank for the ΔCoVaR set. We notice similar situations as in the case of the CoVaR analysis. Slightly more significant connections were identified in the case of the governance pillar, which also culminates at the 60% level. Given the difference between the two systemic risks, we can interpret this result as proof that the average intensity of contribution to systemic risk is to a small extent, more dependent on ESG values from the previous year.

Figure 5 displays the number of significant connections for all pairs of quantiles and for each bank for the MES series in the cross-quantilogram methodology framing. We notice that in some situations, the numbers are larger than before, which indicates that this systemic risk measure that quantifies the impact of each bank to the extreme value of the market is mostly affected by the ESG measures. We notice larger values for the environmental pillar, particularly at the 10% level, which shows that these extreme values tend to be connected. These results show the benefit of our analysis, which explores all percentile levels and produces results that are not visible with standard tools.

Observing the large number of significant connections between several indicators from the ESG group and the financial stability measures, we can conjecture that hypothesis H1 is validated by these results. Additionally, noting the fact that the largest number of significant dependences are not situated at the level of the 50% quantile, we can also admit that we have evidence in favour of hypothesis H2. Therefore, the dependences that we investigate are nonlinear.

Hypothesis H3 is also valid given the diverse manners in which ESG indicators impact financial stability. We further this analysis by trying to provide a framework to visually depict the differences across the ESG pillars.

As such, Figure 6 shows how the ESG pillars affect the three measures of systemic risk also by tracking the number of significant relations. We notice that, overall, the highest number of significant relations is recorded for the ESG Index for all systemic risk measures.

On the other hand, the environmental pillar has the lowest number of significant relations, but in general, these numbers are quite close. Moreover, the impact on MES yields the larger number of significance dependences.

Figure 7 provides further insight into the quantiles for which we find the greatest number of significant dependences. We notice that the smallest number of significances belongs to the 50% quantile (median), while the 40% and 60% quantiles produce larger numbers of significant dependences. This result shows the value of the cross-quantilogram methodology, since different quantiles yield different numbers of dependences.

4. Discussion and Conclusions

Increased awareness of environmental issues along with social and governance issues shed new light on the financial system. On the other hand, systemic risk, as an exponent for financial stability, has grown in importance following the financial turmoil of the past two decades. In our research, we have tried to connect the two relevant contemporary issues: ESG performance and financial stability.

Our results suggest that each component (E, S, and G) and the overall ESG score influence the calculated financial stability of European banks, which is a validation of our first hypothesis (H1). This is in line with previous research which demonstrates that ESG performance has a positive effect on financial stability in the banking sector [35]. The sensitivity of each pillar was previously suggested by the results obtained [50] when considering the financial materiality. Given the fact that these dependences tend to be located at various values for the quantiles, we found evidence in favour of hypothesis H2, according to which the manner in which ESG indicators influence financial stability for European commercial banks is nonlinear.

The heat-maps presented in Figure 3, Figure 4 and Figure 5 provide dashboard views for these dependencies, which can be useful for regulatory investigation of the extent to which possible unbalances might be expected in the future. We noticed slightly different impacts for each component, with the strongest dependence for the case of MES and ESG index. As such, we have identified evidence in favour of hypothesis H3.

Our proposed methodology intrinsically aims at revealing the connections for all levels of the distribution, exploring the dependence at the lower and higher percentiles. This unveiled some connections at the tails, showing that there are high chances for low levels of instability when ESG values are slightly higher than the median, around their 60% percentile levels, which is close to the regular evolution of these indicators.

The results obtained are all the more important as the new banking regulations refer to the ESG performance of banks. The European regulations [22], which are periodically amended and corrected, allow the European Banking Authority to make assessments on ESG risks. One of the considered elements to be affected by these risks is financial stability, for which evaluation criteria need to be built. From this perspective, our results may have managerial and policy implications for both financial institutions and regulation authorities.

Supervisory authorities (such as the European Systemic Risk Board at the European level) closely monitor financial stability. Our findings provide evidence in support of the fact that financial stability is related to sustainability as accounted for by sustainable investments. Designing instruments for monitoring financial stability will have to keep track of the extent to which banks’ activities are sustainable. Policies in favour of sustainable activities for banks are therefore prone to achieve financial stability, too. This supports the view that coordination of authorities in developing policy measures would achieve better levels of desired results.

The difficulties of research on various aspects of ESG are deepened by the findings of recent studies which show that there are differences in used criteria and ratings [51,52]. However, regulations that would unify the approach are being announced, at least on the reporting side, and make things more interpretable.

Another limitation of ESG studies is the relatively short period in which information is provided for the calculation of ESG scores and the restriction to voluntary adoption for companies.

Despite these restraints, the results provide a methodological framework for the analysis of systemic risk. Considering the connections with the ESG pillars, market authorities may monitor the financing structure of the economy by repetitive computations of these dependencies and can propose measures suited for a smooth transition to a green economy. To this extent, we consider that one possible pathway to future research could envisage the connections of systemic risk produced by other sectors of the economy with the ESG pillars. These dependences might be correlated with the current financing structure, which is the main driver for the systemic risk produced by financial institutions.