The Impact of CSR on Default and Systemic Risk in the
European Banking Sector
Author: Sophie van ‘t Klooster (s2358107)1 Supervisor: prof. dr. L.J.R. Scholtens
MSc Economics and MSc Finance Combined Master’s Thesis (EBM000A20)
January 19th, 2018
Abstract: This paper empirically examines whether corporate social responsibility affects a bank’s default risk and systemic risk contribution. I find evidence that corporate social responsibility of a bank results in reduced bank default risk. I also find this negative relationship on a bank’s systemic risk contribution based on the SRISK measure, but not based on the Marginal Expected Shortfall measure. Examining this relationship for the banking sector on a national level, I find strong support for a positive impact of a banking-sector increase in corporate social responsibility on the sector’s systemic risk contribution. This implies that the impact of corporate social responsibility on individual systemic risk contribution does not correspond to the impact on national systemic risk contribution.
JEL classification: G21, G23, M14
Keywords: Corporate social responsibility, Banking, Risk, Systemic Risk
1 Correspondence information: University of Groningen, Faculty of Economics and Business,
2
1. Introduction
The growing realization of the importance of corporate social responsibility (CSR) in society has led to a new era in which firms increasingly integrate CSR into their decision-making and advocate it in their policies. CSR can broadly be defined as the obligation to undertake actions to improve and protect the welfare of society and interest of organizations (Sun and Cui, 2013). In addition, the firm is able to build a strong image and reputation. Prior research suggests that CSR efficiently improves a firm’s performance. Examples are on a firm’s financial performance (e.g. Oikonomou et al., 2012; Cornett et al., 2016), its costs reduction (El Ghoul et al., 2011; Schröder, 2014) and enhanced reputation (Shen et al., 2016). These strands of research focused on all firms jointly, including the banking sector, and thus did not make a distinction between sectors. However, particularly for the banking sector, the general opinion stresses the importance to engage in CSR, because the banking system plays an important role in economic development as its safety and stability create external benefits for society (Wu and Shen, 2013). Correspondingly, studies reported a positive impact of CSR on performance indicators for the banking sector like financial performance (Wu and Shen, 2013; Cornett et al., 2016) and on return on assets (Shen et al., 2016). Given the mounting evidence of the importance of CSR on performance indicators, it is highly surprising that there is little knowledge about its impact on a bank’s default risk, the risk of a bank being unable to fulfil its obligations of repaying its debt. This is a critical indicator by which to measure a bank’s own financial health and stability and is crucial as risk is an integral part of a bank’s business and its activities. Its actions could not only affect itself but even more so the entire financial system and society. Therefore, it is essential for financial stability to investigate whether CSR has an impact on a bank’s default risk and its contribution to systemic risk of the financial system.
3
and vision, leads to the question what kind of effect this not only has on a bank’s default risk, but even more so on its contribution to systemic risk of the financial system.
This paper studies the impact of CSR policy implementation in the banking sector on both a micro- and macroprudential level, while incorporating the impact of recent crises as systemic shocks. To investigate the impact of CSR, this paper uses data from the ESG Asset4 Database, which contains data upon a firm’s environmental, social and corporate governance performance rating. This will be used for banks within the Euro area starting 2002 up until 2016. In this manner, the pre- mid- and post- crises years will be included. A Euro area perspective will give a more realistic representation of a systemic financial system due to a common currency and an integrated market, both politically and economically. I find that a bank’s CSR rating does have a positive effect on its individual solvency measured as by the Z-Score, an inverse measure of bank default risk. This result holds when using an alternative measure for default risk, being the standard deviation of the return on equity of a bank. When looking at a bank’s CSR rating and its impact on systemic risk contribution, I find a significant negative effect based on the SRISK measure for systemic risk contribution, but no evidence when using the Marginal Expected Shortfall measure. Incorporating the bank’s default risk measure to the systemic risk equation confirms that CSR does not only have an effect through bank default risk as a transmission channel, but also directly. A country-level perspective of an aggregate CSR rating for the banking sector finds that aggregate systemic risk contribution increases following an increase in CSR. These findings are in contrast to the individual effect. This suggests that the impact of CSR on systemic risk contribution on an individual level does not lead to the same effect on systemic risk contribution for the banking sector on a country level and that synergies exist.
4
default risk as well to control whether default risk only serves as an intermediary for CSR or that CSR has a separate, direct effect on systemic risk contribution. Third and last, I will bring the systemic risk contribution analysis to a country-sectoral level in which I would like to see whether the effect of CSR intensifies, and find synergies, when aggregating a bank’s individual’s CSR rating and systemic risk contribution for the whole banking sector on a country level. In this sense, I will also be the first to aggregate CSR for the banking sector on a country level and to investigate whether synergies exist for CSR.
The findings could have the following potential implications. The finding that CSR decreases a bank’s default risk and individual systemic risk contribution, but increases systemic risk contribution when viewed from a national banking-sectoral perspective, implies that more research should be focused on the specifics of how banks improve their CSR rating. Such information could be worthwhile to find out what kind of CSR-related factors drive systemic risk of the financial system. Moreover, although CSR is increasingly implemented, this should be done carefully. A clear distinction between firms in general and the banking sector should be made as they have differing influences on the economy. Likewise, governmental policies should be designed while focusing on both an individual and national level. A shared CSR related vision and strategy among banks can be a blessing in a sense that it enhances a society’s wellbeing, but can also banks expose themselves to similar and maybe even an increasing amount of risks, thereby harming financial stability. Therefore, governments should start monitoring how banks implement CSR and ensure by means of regulations that banks diversify less similarly.
This paper continues as follows; Section 2 contains a literature review on the motives and effects of CSR, factors influencing not only bank default risk but also systemic risk and the relation of risk to CSR in the banking sector. Section 3 describes the methodology used in this paper, after which Section 4 discusses the data that has been used. Next, Section 5 elaborates upon the results. Finally, section 6 provides a conclusion on the results, discussion and interesting paths for future research.
2. Literature Review
5
its potential impact on the banking sector in particular. Therefore, this literature section tries to shed more light on the relationship between CSR and risk.
In recent years, growing evidence in the field of CSR shows that CSR activities have an impact on a firm’s performance (e.g. El Ghoul et al., 2011; Goss and Roberts, 2011; Schröder, 2014). Goss and Roberts (2011) examined the link between CSR and bank debt. They found that firms with social responsibility concerns had to pay between 7 and 18 basis points more for their bank loans than firms that were more responsible. Accordingly, also Schröder (2014) found that firms having on average higher CSR ratings had lower financing costs. The higher the CSR rating, particularly environmental criteria and employee relationships, the lower the default risk of loans and corporate bonds and the lower the costs of obtaining debt capital. Another paper by Bauer and Hann (2010) addressed a similar relationship in which proactive environmental practices were associated with lower costs of debt, whereas environmental concerns were associated with higher costs of debt and lower credit ratings. Moreover, El Ghoul et al. (2011) found that higher CSR ratings contributed to lower costs of equity capital of a company, except for controversial industries. Scholtens (2008) recognized the conflicting views related to the causality between corporate social performance and corporate financial performance. According to his research, financial performance on a risk and return level precedes social performance in general. However, not all different themes of CSR necessarily have the same type of interaction with financial return and risk.
6
credit rating for a firm also reflect a firm’s default risk as perceived by the debt holder. As has been previously noted (e.g. Goss and Roberts, 2011; Bauer and Hann, 2010; Scholtens, 2008), there exists a positive relationship between CSR and credit ratings and a negative relationship between CSR and financing costs and risk. Such results are indicators of a negative relationship between a firm’s CSR rating and its default risk.
When it comes to CSR in the banking sector, implementation is not always apparent to observe. Yet, banks are aware of their role in economic development, because its safety and stability are highly valued by society. As banks use substantial resources from society for their day-to-day operating activities, they are more than most other industries required to provide the community transparency about their asset management (Wu and Shen, 2013). Broadly spoken, there are three motives for CSR engagement in the banking sector: strategic choices, altruism, and greenwashing. The degree to which banks engage in CSR highly differs between the motives. Wu and Shen (2013) found that the relationship between CSR and financial performance is positive for the strategic choice motive, non-negative as for the altruism motive, and even non-existent for the greenwashing motive. The positive effect is reflected in higher returns on assets and higher returns on equity. In that light, according to Shen et al. (2016) most banks seem to engage in CSR primarily out of a strategic motive. Moreover, CSR engaged banks turn out to strongly outperform non-CSR engaged banks in terms of return on assets and on equity when comparing the two groups. A study among US commercial banks showed that financial performance was strongly positively related to CSR scores (Cornett et al., 2016). Moreover, this effect proved to be even stronger during the crisis, which could be an indication of lower bank default risk as especially during the crisis they were hit less. The above suggests that CSR practices are pursued by bank managers as a long-term investment strategy. As such, transparency and accountability are essential to stakeholders, because a higher earnings quality provides better information about the underlying features of a bank’s financial performance. CSR has shown to have a positive influence on earnings quality (Sánchez and García-Meca, 2017). In addition, it seems that larger banks are more CSR minded and that banks are increasingly willing to act more socially responsible to enhance their competitive advantage when the market becomes more competitive (Chih et al., 2010).
7
shows that increased competition enhances a bank’s financial stability (Fiordelisi and Mare, 2014).
Although CSR in general has not been investigated among banks as having an impact on a bank’s default risk, several indicators of CSR have been addressed in a few studies. A closer look at the influence of indicators of corporate governance, one of the CSR pillars, on bank default risk shows that more powerful owners tend to take greater risks, and that government regulation can have an either positive or negative impact on a bank’s risk-taking, depending on its ownership structure (Laeven and Levine, 2009). Next to that, executive board composition also has an effect, younger boards and a larger share of women for instance are reflected by higher bank risk (Berger et al., 2014). Such outcomes could be an indication of the CSR pillar corporate governance as a whole to have an effect on bank default risk. As has been noted before, there does exist a relationship between CSR and default risk on a firm level. However, a distinction between the banking sector and overall firm level has not yet been made. A further investigation of default risk on a bank level is important though due to its importance for the society at large. Based on the above results of an effect of CSR on an overall firm level and several categories associated with CSR to have an effect on a bank’s default risk, it leads to the following hypothesis:
Hypothesis 1: CSR affects a bank’s default risk.
8
could indicate that banks diversify in the same manner, exposing them to similar risks, increases systemic risk contribution, whereas solely observing an individual level reflects a low default risk. In addition, the comparison by López-Espinosa et al. (2013) between the determinants of bank default and systemic risk finds that higher unstable sources of funding intensify bank default risk and increase the risk of spill-overs to the financial system. The same authors also found that a higher loan-to-deposits ratio, which is another proxy for funding, exacerbates both bank default- and systemic risk contribution.
Accordingly, it does not have to hold that CSR has a similar effect on bank default risk as it has on systemic risk contribution. CSR comprises a series of categories in which banks can obtain a high rating in several ways. This can have differing implications on either default or systemic risk. An example is corporate governance. Anginer et al. (2017) found that shareholder-friendly corporate governance is associated with higher bank default risk as well as systemic risk. However, Berger et al. (2014) studied the composition of executive boards and found that a bank with more officers with PhD degrees faced lower risks, whereas banks with teams with younger or more female executives faced higher risks. Furthermore, it is plausible that CSR in general only has an effect through its influence on bank default risk on systemic risk contribution. In this case, bank default risk could be perceived as a transmission channel for CSR on systemic risk contribution. Therefore, I include a bank’s default risk in my estimation to verify whether CSR only has a mediating effect through bank solvency or directly affects systemic risk contribution as well.
Hypothesis 2: CSR affects a bank’s systemic risk contribution.
9
on the influence of organizational structure can be related to the corporate governance pillar and provides some insight in the effects CSR could have on systemic risk contribution.
Not only on an individual level does CSR of a firm have an effect, but when taking a larger perspective other effects seem to be in place as well like enhanced competitiveness and economic growth. Therefore, it would not be surprising that the impact of CSR on systemic risk contribution would have an enlarging effect if the CSR impact on a bank would be taken to a sectoral level, the entire banking sector in a country. Here, both the CSR rating and the systemic risk contribution will be aggregated to a national banking sector level. However, since the impact of CSR on systemic risk contribution has not yet been investigated, let alone on aggregate systemic risk contribution, I introduce a new path of research. An enhanced CSR policy by only one bank cannot change systemic risk significantly, but if the banks jointly enhance their CSR policy it could increasingly impact the stability of the financial system. For instance, if all banks in a certain country collectively enhance their employment quality, a category which is a component of the social pillar, this could enhance trust of the society in the financial system, making the bank less sensitive to systemic risk. Similar categories that enhance a bank’s image could have such an effect as well. In addition, if banks in a country share the same vision in which excessive risk-taking is outrageous, this could directly affect the sector’s contribution to systemic risk. If just one bank would enforce this vision, it would still be highly exposed to defaulting as all banks share the same primary domestic market, accompanied with mostly similar market risks. Therefore, the third hypothesis is as follows:
Hypothesis 3: On a national level, a banking sector’s aggregate CSR affects the aggregate systemic risk contribution.
The following section will provide an econometric structure which is essential in testing the three hypotheses.
3. Research Design 3.1 Empirical Model
The empirical model consists of several regressions, ranging from an individual to a systemic level. First, I will investigate the relationship between CSR and individual bank default risk, in which the following setup is applied:
𝐷𝑒𝑓𝑎𝑢𝑙𝑡 𝑅𝑖𝑠𝑘𝑖,𝑡 = 𝛼 + 𝛽1𝐶𝑆𝑅𝑖,𝑡−1+ 𝛽2𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡−1+ 𝛽3𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑐,𝑡−1
10
Where the measure for individual bank default risk refers to bank i at time t, CSR to the prior year’s CSR rating of bank i, controls include bank-specific characteristics of bank i at time t-1 and country-specific characteristics of country c at time t-1, 𝜔𝑡 are time fixed effects, 𝛾𝑖 bank fixed effects and 𝜀𝑖,𝑡 represents the error term.
The second regression of interest considers the relationship between CSR and a bank’s contribution to systemic risk:
𝐶𝑜𝑛𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛 𝑡𝑜 𝑆𝑦𝑠𝑡𝑒𝑚𝑖𝑐 𝑅𝑖𝑠𝑘𝑖,𝑡 = 𝛼 + 𝛽1𝐶𝑆𝑅𝑖,𝑡−1+ 𝛽2𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡−1+ 𝛽3𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑐,𝑡−1+ 𝜔𝑡+ 𝛾𝑖+ 𝜀𝑖,𝑡
Here, the dependent variable will be systemic risk for bank i at time t, which is the only difference from the first regression.
Lastly, the prior mentioned regression will be taken to an aggregate level. The explanatory variable of interest is an aggregated CSR ranking for all banks in a specific country. The same holds for the systemic risk contribution. All individual systemic risk contribution levels for the banks will be aggregated to a joint country-level systemic risk contribution for the banking sector.
𝐶𝑜𝑛𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛 𝑡𝑜 𝑆𝑦𝑠𝑡𝑒𝑚𝑖𝑐 𝑅𝑖𝑠𝑘𝑐,𝑡 = 𝛼 + 𝛽1𝐶𝑆𝑅𝑐,𝑡−1+ 𝛽2𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑐,𝑡−1+
𝛽3𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝐶𝑆𝑅 𝑅𝑎𝑡𝑖𝑛𝑔𝑐,𝑡−1+ 𝛾𝑖 + 𝜀𝑖,𝑡 (3)
This regression brings both the bank’s contribution to systemic risk and CSR to a country level, in which country-specific characteristics will be included as controls as well as a country-level CSR Rating variable, which differs from the aggregated CSR rating which only includes the banking sector.
As the bank-specific variable Assets appeared to be non-stationary after running a Fisher-type unit root test, it has been transformed into stationary variables to make sure the model is correctly specified, see Table A.3 in the Appendix for the specifics. The model for both Equation (1) and Equation (2) are estimated using a fixed effects regression, based on the results of the Hausman test which tests for either the use of fixed or random effects of which the outcomes can be found in Table A.3 in the Appendix. Moreover, the standard errors are clustered on a bank level to correct for both heteroskedasticity and autocorrelation. Also, in line with the Hausman test, Equation (3) will be based on a pooled OLS regression in which Newey-West standard errors are included. In addition, time-fixed effects have been added for all three equations to take into account time or seasonal trends and effects of the Global
11
Financial and Sovereign Debt Crisis. In addition, the explanatory variables are all lagged by one year to control for possible endogeneity. According to Garcia-Castro et al. (2010) previous research has not properly addressed endogeneity problems which significantly biased results regarding CSR and financial performance throughout the years. They state that most of these research results turn into a negative CSR - financial performance relationship when corrected for this.
3.2 Methodology
To analyse the effect of CSR on systemic risk contribution of the banking sector in the Euro area, this paper makes use of several methodologies based on bank default risk at an individual level and at level of the entire banking sector on a country-level. First, two measures for default risk at the bank-level will be calculated being the Z-Score as proposed by Berger et al. (2016), and the standard deviation of return on equity measure (Laeven and Levine, 2009; Berger et al., 2016). the Z-Score serves a proxy for individual bank stability, being extensively used in the literature of banking (Boyd et al., 2007; Laeven and Levine, 2009; Fiordelisi and Mare, 2014). This is my main variable of interest whereas the volatility of return on equity measure serves as a control for the robustness of the Z-Score. Both are performance metrics, contrary to for example the Tier-1 capital ratio which is a regulating metric. Second, measures for systemic risk contribution of the banking sector will applied. These are the Marginal Expected Shortfall (MES) as suggested by Acharya et al. (2012) and the SRISK, firstly introduced by Brownlees and Engle (2016). Hereafter, an elaboration will follow upon the methodology regarding aggregating both the CSR score and systemic risk contribution for the banking sector on a country-level.
Here, the Z-Score is computed as the sum of the return on average assets (ROAA) and the capital-asset-ratio (CAR) divided by the standard deviation of the return on assets over a pre-specified amount of years(σROA). It represents the number of standard deviations by which returns have to decline in order to deplete a bank’s equity capital, and is defined as follows:
𝑍-𝑆𝑐𝑜𝑟𝑒𝑖,𝑡 =𝑅𝑂𝐴𝐴𝑖,𝑡+𝐶𝐴𝑅𝑖,𝑡
𝜎(𝑅𝑂𝐴)𝑖,𝑡 (4)
12
As an alternative for the Z-Score, this paper makes use of the standard deviation of the return on equity 𝜎𝑅𝑂𝐸 which is measured as the net income of a bank i in year t over its total equity in the same year t as in Laeven and Levine (2009) and Berger et al. (2016). Analogous to the computation of the standard deviation of 𝑅𝑂𝐴𝑖,𝑡 for the Z-Score, the standard deviation of 𝑅𝑂𝐸𝑖,𝑡 is calculated over a three-year rolling time window.
To calculate the systemic risk contribution of banks, this paper utilizes the SRISK measure, firstly introduced by Brownlees and Engle (2016). The calculations necessary for this method are based on readily available balance sheet information. SRISK is defined as the expected capital shortfall of a bank during a period of distress in which the financial market declines substantially, or the capital that a bank is expected to need in case of a financial crisis. That is,
𝑆𝑅𝐼𝑆𝐾𝑖,𝑡 = 𝐸𝑡−1(𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑆ℎ𝑜𝑟𝑡𝑓𝑎𝑙𝑙𝑖| 𝐶𝑟𝑖𝑠𝑖𝑠)
Formally, it will look as follows:
𝑆𝑅𝐼𝑆𝐾𝑖,𝑡 = 𝑘𝐷𝑖𝑡− (1 − 𝑘)𝑊𝑖,𝑡(1 − 𝐿𝑅𝑀𝐸𝑆𝑖,𝑡+ℎ|𝑡(𝐶𝑡+ℎ|𝑡)) (5)
Where k is the minimum fraction of capital as a ratio of total assets that each bank needs to hold. I will set this prudential capital fraction equal to -8 % just like Brownlees and Engle (2016). 𝐷𝑖𝑡 and is the book value of the firm’s debt, total liabilities, and 𝑊𝑖𝑡 its market value of equity. Moreover, 𝐿𝑅𝑀𝐸𝑆𝑖𝑡 represents the Long-Run Marginal Expected Shortfall. In calculating this, I follow Acharya et al. (2012), where I define 𝐶𝑡+ℎ|𝑡 as the market decline below a threshold C, equal to -40%, over a time horizon h which I set equal to 180 days. Next, the following approximation based on Acharya et al. (2012) will be used to compute the Long-Run Marginal Expected Shortfall (LRMES):
𝐿𝑅𝑀𝐸𝑆𝑖,𝑡+180|𝑡(𝐶𝑡+180|𝑡) = 1 − 𝑒𝑥𝑝(−18 * 𝑀𝐸𝑆𝑖,𝑡+180|𝑡(𝐶𝑡+180|𝑡)) (6)
As appears from Equation (6), the LRMES is based on the one-day Marginal Expected Shortfall (MES). The MES is defined as the tail expectation of the firm’s equity return conditional on a market decline of -2% on that same day. The one-day MES is calculated as the average MES over a year t. More specifically,
𝑀𝐸𝑆𝑖,𝑡+1|𝑡(𝐶𝑡+1|𝑡) = −𝐸𝑡(𝑅𝑖,𝑡+1|𝑡|𝑅𝑚,𝑡+1|𝑡 < 𝐶) (7)
13
part of the SRISK measure, the one-day average MES serves as an alternative measure for systemic risk contribution. In constructing the SRISK measure, I do not limit SRISK to a threshold of zero as in Acharya et al. (2012), but allow for negative values too. The philosophy behind this is that highly capitalized banks with a large capital buffer can absorb systemic shocks that will reduce the overall systemic risk of the financial system (Laeven et al., 2016). After constructing SRISK across all banks over period t, I want to create an overall measure of systemic risk contribution that is for the entire banking sector in a country, 𝑆𝑅𝐼𝑆𝐾𝑐,𝑡, hereafter called aggregate systemic risk contribution. In this sense, the total amount of financial risk in the country-level’s financial system is measured as the sum of individual banks’ contribution to systemic risk, conditional on the bank being situated in country c:
𝑆𝑅𝐼𝑆𝐾𝑐,𝑡 = ∑𝑁𝑖=1𝑆𝑅𝐼𝑆𝐾𝑖,𝑡 (8)
Instead of reflecting aggregate systemic risk contribution by a currency value, it can often be more insightful to show it in a share of total aggregate systemic risk for the entire system:
𝑆𝑅𝐼𝑆𝐾 𝑆ℎ𝑎𝑟𝑒𝑐,𝑡 = 𝑆𝑅𝐼𝑆𝐾𝑐,𝑡
𝑆𝑅𝐼𝑆𝐾𝑡 (9)
The total aggregate systemic risk measure for the entire system can be regarded as the total amount of capital that has to be provided by the government(s) to bail out the financial system in case of distress of the system (Brownlees and Engle, 2016). In their paper, Brownlees and Engle (2016) argue that in case of a crisis it will be unlikely that surplus capital, reflected by a negative SRISK, will easily be used in terms of mergers or loans by those banks, to support failing banks. Therefore, negative contributions to systemic risk will not be included in the aggregate measure. However, I argue that banks with abundant capital do in most cases acquire to some extent the client base of the failing bank. This would subtract systemic risk from the financial system. In this decision, I follow Laeven et al. (2016). I do admit that this does not hold in every case, where the government has to intervene and thus the actual aggregate systemic risk would be slightly higher.
3.3 Bank and country characteristics as determinants of risk
14
the deposit ratio. First, I control for bank size, because I expect larger banks to be less fragile by virtue of a higher degree of diversification and higher systemic relevance leading to more intense monitoring by supervisors. On the other hand, the “too big to fail” argument could lead to increased fragility as larger banks take more risks, knowing they will become government support in times of distress (Eichler and Sobanski, 2016; Laeven et al., 2016). In addition, larger banks tend to have a higher social impact and to draw higher levels of attention from the public. Therefore, larger banks are increasingly likely to be engaged in CSR activities due to public awareness, especially after the crisis (Cornett et al., 2016). In addition, the Tier-1 capital ratio will be added, which is the ratio of core equity capital to total risk-weighted assets. The Basel III capital ratio was introduced to increase a bank’s solvency, where a high Tier-1 capital ratio would reflect high bank solvency. However, as banks will be forced to have a minimum amount of capital, they could also be incentivized to take more risk. Also, having abundant capital also leads to having more funds available to pursue CSR activities (Cornett et al., 2016). Subsequently, the short-term wholesale funding ratio has been included. This variable is expected to have a positive relationship to systemic risk. Banks with excessive short-term funding ratios are usually more interconnected to other banks, are more vulnerable to market conditions, liquidity risk and exposed to a high degree of maturity mismatch (López-Espinosa et al., 2013). To describe the type of business the bank is mainly engaged in, I introduce the share of non-interest income to total income. This reflects how exposed a bank is to riskier non-commercial banking activities like trading. However, it could also indicate that a bank is better diversified, has a more innovative business model and therefore has a lower risk exposure. To measure the influence of a bank’s capital structure, I include the deposit ratio in my regression. Traditional commercial banks typically engage in savings and loan activities, and are usually less financed via securities or the capital market. That makes them less interconnected to other banks, and would therefore have a negative influence on systemic risk.
15
the economy. This could increase systemic risk contribution (Eichler and Sobanski, 2016). Lastly, an additional control factor has been created that measures a country’s general CSR rating. This rating is based on three different measures, which are further explained in the data section. The country’s general CSR rating is to be included in the aggregate bank-CSR regression to control for potentially varying trends in CSR in the sample countries.
4. Data 4.1 Sources
My analysis begins by identifying a hypothetical financial system in which to analyse systemic risk. In this regard, the sample will be restricted to banks situated in countries part of the Euro area starting January 20022. Of these countries, 11 out of 12 agreed to join the Euro Area in 1999 by introducing the Euro as a virtual currency, whereas Greece had been formally accepted to the Euro area as of January 2001. The dataset ranges from 2002 until 2016 as 2002 was the year in which the Euro cash changeover took place, making sure that differences among these 12 economies are minimized as they face the same currency-related risks. By taking this range, a pre- mid- and post-crisis period have been included allowing me to analyse the effect of a systemic crisis, being the Global Financial and Sovereign Debt Crisis.
The measurement of CSR is of special importance for the reliability of my results. As CSR-related topics have gained significant interest throughout the years, an increasing number of organizations provide CSR data nowadays as there is no univariate approach for the construction of such a rating. Some of the most well-known suppliers are MSCI, Thomson Reuters and Sustainalytics (see Schäfer et al. (2006) for a concise overview of ratings). This study makes use of Thomson Reuters DataStream to obtain ESG ratings from the ASSET4 database. The ESG rating comprises an environmental, social and corporate governance pillar, and is used to measure CSR. The choice for this database has been made based on its reputation of being one of the most reliable and comprehensive databases for financial firms. It combines financial and non-financial publicly available information with the intention to create an integrated view of corporate performance. It is used to fill 226 CSR-related key performance indicators (KPI’s), being distributed among three pillars, being an environmental, social, and corporate governance pillar. The environmental pillar has been composed of the categories emission reduction, product innovation and resource reduction. The social pillar consists of the categories community, diversity, employment quality, health and safety, human rights, product responsibility, and training and development. Lastly, the corporate governance pillar contains the following categories: board functions, board structure, compensation policy, shareholder
2 Countries that were part of the Euro area starting 2002 included: Austria, Belgium, Finland, France,
16
rights, and vision and strategy. My final rating will comprise of an aggregate equal-weighted measure of the three pillars. After matching Thomson Reuters ASSET4 ESG ratings with these banks my final sample includes 43 banks in 10 countries. A list of all banks and corresponding countries can be found in the Appendix in Table A.1. As the CSR rating is the main variable of interest in this paper, banks of which no observations of CSR ratings are available have been excluded. Still, all banks that have been marked as globally systemically important by the European Banking Authority in one of these countries are part of this sample.
Research into the country-level relationship between aggregate CSR and aggregate systemic risk contribution further restricts the sample. As several banks do not have CSR data in the year 2002, this year will be excluded. The Netherlands will be excluded as the majority of years the systemic risk level only consists of one bank, ING Group. This would give an unrealistic reflection of the systemic risk level of the banking sector in that country. Moreover, banks which lack more than two years of data on either CSR or systemic risk will be excluded as well. By means of linear interpolation the remaining one or two years will be calculated. A total of 7 banks will be excluded by these restriction, leading to a country level sample in which 36 banks are involved, see also Table A.1 in the Appendix.
All financial data has been obtained from Datastream, as well as bank-specific control variables and the country-specific control variable market-capitalization rate as well. The remaining country-specific control variables have been obtained from the World Bank World Development Indicators. Table A.2 in the Appendix provides a complete overview of all variables and related definitions that have been used in this study.
The country’s general CSR rating consists of three proxies for the three individual pillars. First, the proxy for the environmental pillar is represented by the Yale Environmental Performance Index. Second, the social pillar has been matched to the Human Development Index which is created by the United Nations. Third, the corporate governance factor is represented by the World Bank’s Worldwide Governance Indicator control for corruption. Just like the bank-level CSR rating, all three pillars are equally weighted and combined form one score. This leads to the additional country’s general CSR rating.
4.2 Summary statistics
17
calculated a value of 36.053, suggesting that the European banks in my dataset are more stable. The difference becomes even more apparent when comparing my sample to the European cooperative bank sample of Fiordelisi and Mare (2014), who report a mean value of 17.257 for the period 1998 to 2009. This holds for the standard deviation of equity as well, which is more than twice as high in their sample, 0.035. The CSR Rating ranges from 13.45 to 288.14, and with a mean of 186.412 indicates extreme values. Yet, the existence of potential outliers will be dealt with in a robustness test whether they have an impact. As for the variable size, the natural logarithm of a bank’s total assets is taken to transform it into a stationary variable.
Table 1. Descriptive Statistics
N Mean Stdev. Min Max Skewness Kurtosis
Z-Score 536 57.456 97.464 -2.403 1189.020 5.275 45.193 Stdev. ROE 580 0.168 0.573 0.000 5.869 6.362 49.423 MES 605 0.002 0.002 0.000 0.010 1.762 6.787 SRISK (€) 600 1.29‧107 2.71‧107 -4.52‧107 1.61‧108 2.504 9.793 CSR Rating: Individual 562 185.870 80.307 13.450 288.140 -0.660 2.099 Size 622 18.701 1.400 14.708 21.509 -0.088 2.897 Tier-1 Ratio (%) 557 10.029 3.081 -7.300 21.400 -0.169 6.468 Deposit Ratio 579 0.440 0.152 0.023 0.831 -0.593 3.466 Short-Term Funding 622 0.189 0.102 0.000 0.589 0.806 4.046 Non-Interest Income 586 0.569 0.863 -0.157 11.783 7.795 81.504 GDP Growth Rate (%) 645 0.883 3.218 -9.132 26.276 1.760 20.411 Public Debt to GDP 532 83.895 33.284 26.920 151.824 0.107 2.072 Inflation Rate (%) 645 1.780 1.428 -4.480 4.880 -0.557 3.786 Market Capitalization 623 52.655 24.985 11.740 121.660 0.533 2.484 Notes: For the exact definition of the variables, see Table A.2.
18
Table 2. Correlation Matrix
1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 Z-Score 1.000 2 Stdev. ROE -0.158 1.000 3 MES -0.198 0.109 1.000 4 SRISK -0.039 -0.039 0.326 1.000 5 CSR Rating: Individual -0.050 -0.001 0.174 0.398 1.000 6 Size -0.045 -0.034 0.229 0.682 0.661 1.000
7 Tier-1 Capital ratio -0.069 0.032 0.022 0.150 0.114 0.169 1.000
19
5. Empirical results 5.1 Main results
In this section, Equation (1) has been used to estimate the effect of CSR on bank default risk, reflected by two different risk measures: The Z-Score and the standard deviation of return on equity. This equation tests the first hypothesis, stating that CSR affects default risk of a bank. Second, Equation (2) tests the second hypothesis. This hypothesis states that CSR affects a bank’s systemic risk contribution. As such, the variables MES and SRISK reflect the bank’s systemic risk contribution. Third, the effect of CSR on systemic risk contribution on the banking at a country level has been estimated by Equation (3). An aggregate banking sector CSR rating and aggregate systemic risk contribution will be constructed to perform this analysis. The third equation tests the third hypothesis that aggregate CSR affects systemic risk contribution as a result of potential synergies.
Table 3 reports the regression results of Equation (1). The first hypothesis predicted that CSR impacts bank default risk. The test confirms this hypothesis by using two different measures for default risk. The estimation results for the Z-Score show that a bank’s CSR level is positively related to the Z-Score. An increase of the CSR rating by 1 unit, increases the bank’s Z-Score by 0.303 units, which means that it is more solvent, as the Z-Score is an inverse measure for default risk. Moreover, the CSR level is negatively related to the standard deviation of the return on equity at a ten percent level. An increase in the CSR rating by 1 unit decreases the standard deviation of ROE by 0.001, meaning a decrease in bank default risk. In line with the literature on a firm level, CSR is negatively related to bank default risk (Jo and Na, 2012; Oikonomou et al, 2012; Sun and Cui, 2014).
20
Table 3. Empirical Analysis Results Bank Default Risk
Z-Score Z-Score Stdev. ROE
Full model Full Model Robust-cluster Full Model Robust-cluster CSR Rating: individual 0.303** 0.303* -0.001* (0.142) (0.154) (0.001) Size 14.813 14.813 -0.186* (25.663) (23.502) (0.106)
Tier-1 Capital Ratio 2.511 2.511 -0.024**
(2.810) (2.893) (0.011) Deposit Ratio -87.834 -87.834 -1.227*** (81.290) (115.657) (0.420) Non-Interest Income 7.138 7.138 -0.113** (17.153) (18.798) (0.051) Short-term Funding -167.629** -167.629* -0.465 (77.524) (89.988) (0.280) GDP Growth Rate 1.595 1.595 -0.013 (1.920) (1.269) (0.010) Public Debt to GDP -0.075 -0.075 0.000 (0.447) (0.480) (0.001) Inflation Rate 2.264 2.264 -0.022 (6.817) (7.910) (0.063) Market Capitalization -0.308 -0.308 -0.006 (0.581) (0.551) (0.004)
Time Fixed Effects yes yes yes
Bank Fixed Effects yes yes yes
Adjusted R-Squared 0.138 0.138 0.174
Notes: For the exact definition of the variables, see Table A.2. All explanatory variables are lagged by one year. Robust standard errors are reported in the parentheses. *indicates significance at the ten percent level, ** indicates significance at the five percent level and ***indicates significance at the one percent level.
21
Table 4. Empirical Analysis Results Systemic Risk Contribution
SRISK % SRISK% SRISK% MES%
Full model Full Model Robust-cluster Full Model Robust-cluster Full Model Robust-cluster CSR Rating: individual -1.552* -1.552* -1.457** 0.018 (0.921) (0.855) (0.655) (0.212) Z-Score -0.269 0.019 (0.234) (0.057) Size 38.433 38.433 75.023 103.921*** (163.509) (140.065) (85.580) (37.136)
Tier-1 Capital Ratio -7.044 -7.044 -3.645 3.164
(18.924) (16.219) (15.287) (4.790) Deposit Ratio -286.236 -286.236 -255.494 117.269 (522.864) (471.565) (477.694) (132.478) Non-Interest Income -42.932 -42.932 -37.768 -6.923 (110.353) (52.857) (60.078) (20.364) Short-term Funding 125.656 125.656 -65.981 171.733 (510.723) (406.504) (389.476) (105.286) GDP Growth Rate -17.966 -17.966 -20.890 5.306** (12.601) (16.671) (17.318) (1.971) Public Debt to GDP 1.483 1.483 0.847 -0.708 (2.955) (2.657) (2.428) (0.518) Inflation Rate 103.288** 103.288** 98.524** 9.563 (44.182) (47.074) (44.453) (10.121) Market Capitalization -12.367*** -12.367* -10.101 -0.248 (3.667) (6.603) (6.159) (1.004)
Time Fixed Effects yes yes yes yes
Bank Fixed Effects yes yes yes yes
Adjusted R-Squared 0.092 0.092 0.087 0.712
Notes: For the exact definition of the variables, see Table A.2. All explanatory variables are lagged by one year. Robust standard errors are reported in the parentheses. *indicates significance at the ten percent level, ** indicates significance at the five percent level and ***indicates significance at the one percent level.
22
good performance and indicates that CSR has contrasting effects. The third hypothesis predicted that on a national level the aggregate CSR rating of the banking sector has a larger impact on the aggregate systemic risk contribution than the banks individually. The test strongly confirms this on a 1% level. The outcome shows that on a national level the CSR rating has a positive effect on systemic risk contribution. This contrasts with the outcome of the second hypothesis in which I found a negative relationship on an individual level.
Table 5. Empirical Analysis Results Aggregate Systemic Risk
SRISK Share SRISK Share Pooled OLS Pooled OLS
Newey-West CSR Rating: sectoral 0.208*** 0.208*** (0.050) (0.053) GDP Growth Rate -0.004 -0.004 (0.005) (0.005) Inflation Rate 0.004 0.004 (0.021) (0.023) Public Debt to GDP 0.001 0.001 (0.001) (0.001) Market Capitalization 0.000 0.000 (0.001) (0.001) Country CSR Rating 0.022*** 0.022** (0.008) (0.010)
Year Fixed Effects yes yes
Adjusted R-Squared 0.226 0.226
Notes: For the exact definition of the variables, see Table A.2. All explanatory variables are lagged by one year. Robust standard errors are reported in the parentheses. *indicates significance at the ten percent level, ** indicates significance at the five percent level and ***indicates significance at the one percent level.
5.2 Further analyses and robustness tests
In this section, I report the results from the regressions on the three individual CSR pillars and several robustness checks on the models of interest.
23
a significant influence on bank default risk or systemic risk contribution based on the Z-Score and SRISK measure. Next, I run the model for the aggregate systemic risk contribution and the three individual CSR pillars. The results that have been reported in Table A.5 in the Appendix show that all three pillars have a significant effect on the systemic risk contribution share. The social pillar has the largest coefficient estimate of 0.295, significant at a 1% level. Moreover, the environmental pillar is also significant at a 1% level, with a corresponding estimate of 0.205. The corporate governance pillar has a considerately lower estimate of 0.099 and is significant at a 5% level. On a larger or aggregate scale, corporate governance seems to have more impact on risk contribution than on an individual risk level as its effect was not significant at an individual level. Especially the effect of the social pillar is strongly significant at the aggregate level and seems to gain from synergies.
Outliers in the dataset can significantly influence the results and potentially give biased results. To deal with biased results, extreme values from the main explanatory variable, the CSR score, will be removed by taking out the top and bottom 1% of CSR scores from the models based on Equation (1), (2) and (3). The results remain robust as has been reported in Table A.6 and Table A.7 in the Appendix. The CSR rating coefficient for the Z-Score has become slightly smaller from 0.303 in Table 3 to 0.300 in Table A.6, whereas the CSR rating coefficient for the SRISK systemic risk contribution becomes slightly larger at -1.485 in Table A.6.
Adding additional bank-specific control variables to Equation (1) and (2), particularly the return-on-assets ratio and the loan-to-deposit ratio instead of the deposit ratio, and other country-specific control variables, being the unemployment ratio and the natural logarithm of GDP, leads to the results that can be found in Table A.8 in the Appendix. The return-on-assets ratio has been included, because it controls for potential importance of bank profitability on risk (Jo et al., 2015; Black et al., 2016; Shen et al., 2016). Moreover, the loan-to-deposit ratio has been included as it measures to what extent loans are financed with deposits, perceived as a more stable source of funding (López-Espinosa et al., 2013; Black et al., 2016). The natural logarithm of GDP of a country has been added, to control for the country’s overall size of the economy (Beltratti and Stulz, 2012). The unemployment rate has been included to control for differences in economic conditions across countries (Boyd et al., 2007; Engle et al., 2014).
24
6. Conclusion
This paper stressed the importance of finding out whether CSR has an impact on bank default risk and systemic risk contribution, as the banking sector plays an essential role in stability of the financial system and society. For this purpose, the paper investigated the effect of CSR on bank default risk and systemic risk contribution for 43 banks headquartered in the Euro area. Using a dataset for the period 2002 to 2016 I find evidence that CSR negatively affects bank risk. This result is confirmed using two different estimates for bank default risk, both showing signs in favour of this relationship. Therefore, this evidence confirms hypothesis 1 that CSR affects bank default risk. A further investigation of CSR on a systemic risk level provides evidence of a negative effect of CSR on the contribution to systemic risk when measured by SRISK, but not when measured by the MES. Here, incorporating a bank’s default risk in combination with CSR shows even more evidence of a negative effect of CSR on systemic risk contribution. This result suggests that bank default risk does not only serve as a transmission channel for the effect of CSR, but that CSR does have a direct effect on systemic risk contribution as well. Therefore, I can confirm hypothesis 2 based on the SRISK measure. The third hypothesis explored the relationship between CSR and systemic risk contribution for the entire banking sector in a country. Aggregating bank-level CSR ratings and systemic risk contribution to a country-sectoral level provides evidence in favour of a positive relationship between CSR and systemic risk contribution. This result contradicts to the finding that CSR negatively influences a bank’s default risk. It suggests that when taking an individual perspective, a bank’s CSR rating does have a positive impact on systemic risk contribution, whereas the rating for the banking sector jointly has a negative impact on joint systemic risk contribution. The finding can be explained by a similar phenomenon where a diversified bank seems highly solvent, but as all banks diversify similarly they are still exposed to the same risks. This suggests the same relationship for a bank’s CSR rating.
The robustness tests confirm my findings. Separating CSR into the three individual pillars shows that the social pillar is the only driving pillar behind the decrease in bank default risk. However, on a bank’s systemic risk contribution both the social and environmental pillar have a decreasing effect, where the effect of the social pillar is more significant. On an aggregate level, all three pillars have a significant positive effect, where both the social and environmental pillar are strongly significant. This last result shows that the effect on an aggregate level for CSR becomes more apparent as even corporate governance shows to be of effect, indicating synergies from CSR on an aggregate level on aggregate systemic risk contribution.
25
light on the impact of CSR in the banking sector. This analysis has certain shortcomings. First, most of my data was only available on a yearly basis. Therefore, the measures for the standard deviation of bank default risk were only calculated over a small sample. This makes it likely that large variations in the Z-Score are driven by the volatility of ROA. This holds for the alternative measure as well, the standard deviation of return on equity. This appears to be a shortcoming too in the estimation of the systemic risk measure. As a financial breakdown occurs suddenly, yearly data cannot accurately assess the effect on a bank its systemic risk contribution. The availability of quarterly or monthly data could greatly enhance the quality of the research into the effect of CSR on bank risk, especially with the Global Financial and Sovereign Debt Crisis that affected European banks. Furthermore, simply aggregating all bank’s individual CSR ratings and systemic risk contribution numbers to arrive at a country level measure does have its limitations. As I wanted to verify whether synergies could exist at a country level, by means of stronger evidence of the impact of the CSR coefficient, the systemic risk measure itself could due to synergies combined also be larger in theory for example. However, identifying this goes beyond the scope of my research. As to my knowledge other research has not come up with a solution either for the aggregation of similar measures.
The findings in this paper have several implications. First, the finding that CSR decreases a bank’s default risk and individual systemic risk contribution, but increases systemic risk contribution when viewed from the joint banking sector in a country, implies that more research should be focused on the specifics of how banks improve their CSR rating. Such studies could be worthwhile to find out what kind of CSR-related factors drive systemic risk of the financial system. The finding that the environmental and especially the social pillar have a significant effect, suggests that research should concentrate more on these two. Accordingly, it may help central banks in their quest to lower systemic risk and circumvent another crisis. Second, although CSR is increasingly implemented, governmental policies on the promotion of CSR should be carefully considered. A clear distinction between the banking sector and a general firm level should be made as they play different roles in the economy. Banks are more interconnected with society than a regular firm, implying that a bank’s default has more severe consequences. Likewise, governmental policies for the banking sector in particular should be designed while focusing on both an individual and national level. A shared CSR related vision and strategy among banks can be a blessing in a sense that it enhances a society’s wellbeing, but could expose themselves to comparable risks as a result of their vision and strategy, harming financial stability. Therefore, governments should also start monitoring how banks diversify risks and CSR and implement regulations that ensure a more diverse diversification of CSR among banks.
26
27
7. References
Acharya, V., Engle, R., & Richardson, M. (2012). Capital shortfall: A new approach to ranking and regulating systemic risks. The American Economic Review, 102(3), 59-64.
Anginer, D., Demirgüç-Kunt, A., Huizinga, H., & Ma, K. (2017). Corporate Governance and Bank Insolvency Risk: International Evidence. European Banking Center Discussion Paper Series No. 2014-010 (No. 2014-053). CentER Discussion Paper Series.
Bauer, R. & Hann, D. (2010). Corporate environmental management and credit risk. Working Paper. Maastricht University.
Beale, N., Rand, D. G., Battey, H., Croxson, K., May, R. M., & Nowak, M. A. (2011). Individual versus systemic risk and the regulator's dilemma. Proceedings of the National Academy of Sciences, 108(31), 12647-12652.
Beltratti, A., & Stulz, R. M. (2012). The credit crisis around the globe: Why did some banks perform better?. Journal of Financial Economics, 105(1), 1-17.
Berger, A. N., Kick, T., & Schaeck, K. (2014). Executive board composition and bank risk taking. Journal of Corporate Finance, 28, 48-65.
Berger, A. N., El Ghoul, S., Guedhami, O., & Roman, R. A. (2016). Internationalization and bank risk. Management Science.
Black, L., Correa, R., Huang, X., & Zhou, H. (2016). The systemic risk of European banks during the financial and sovereign debt crises. Journal of Banking & Finance, 63, 107-125.
Boulouta, I., & Pitelis, C. N. (2014). Who needs CSR? The impact of corporate social responsibility on national competitiveness. Journal of Business Ethics, 119(3), 349-364.
Boyd, J. H., De Nicolò, G., Jalal, A. M., & Boyd, H. (2007). Bank risk-taking and competition revisited: New theory and new evidence”. IMF Working Paper No. 06/297. In Journal of Finance.
28
Brunnermeier, M. K., Dong G. N., & Palia D. (2012). Banks Non-Interest Income and Systemic Risk. Working Paper. Princeton University.
Chih, H. L., Chih, H. H., & Chen, T. Y. (2010). On the determinants of corporate social responsibility: International evidence on the financial industry. Journal of Business Ethics, 93(1), 115-135.
Cornett, M. M., Erhemjamts, O., & Tehranian, H. (2016). Greed or good deeds: An examination of the relation between corporate social responsibility and the financial performance of US commercial banks around the financial crisis. Journal of Banking & Finance, 70, 137-159.
De Jonghe, O. (2010). Back to the basics in banking? A micro-analysis of banking system stability. Journal of financial intermediation, 19(3), 387-417.
Eichler, S., & Sobański, K. (2016). National politics and bank default risk in the eurozone. Journal of Financial Stability, 26, 247-256.
El Ghoul, S., Guedhami, O., Kwok, C. C., & Mishra, D. R. (2011). Does corporate social responsibility affect the cost of capital?. Journal of Banking & Finance, 35(9), 2388-2406.
Engle, R., Jondeau, E., & Rockinger, M. (2014). Systemic risk in Europe. Review of Finance, 19(1), 145-190.
Fiordelisi, F., & Marques-Ibanez, D. (2013). Is bank default risk systematic?. Journal of Banking & Finance, 37(6), 2000-2010.
Fiordelisi, F., & Mare, D. S. (2014). Competition and financial stability in European cooperative banks. Journal of International Money and Finance, 45, 1-16.
Garcia-Castro, R., Ariño, M. A., & Canela, M. A. (2010). Does social performance really lead to financial performance? Accounting for endogeneity. Journal of Business Ethics, 92(1), 107-126.
29
Goss, A., & Roberts, G. S. (2011). The impact of corporate social responsibility on the cost of bank loans. Journal of Banking & Finance, 35(7), 1794-1810.
Jo, H., & Na, H. (2012). Does CSR reduce firm risk? Evidence from controversial industry sectors. Journal of Business Ethics, 110(4), 441-456.
Jo, H., Kim, H., & Park, K. (2015). Corporate environmental responsibility and firm performance in the financial services sector. Journal of business ethics, 131(2), 257-284.
Kleinow, J., & Moreira, F. (2016). Systemic risk among European banks: A copula approach. Journal of International Financial Markets, Institutions and Money, 42, 27-42.
Köhler, M. (2014). Does non-interest income make banks more risky? Retail-versus investment-oriented banks. Review of Financial Economics, 23(4), 182-193.
Laeven, L., & Levine, R. (2009). Bank governance, regulation and risk taking. Journal of financial economics, 93(2), 259-275.
Laeven, L., Ratnovski, L., & Tong, H. (2016). Bank size, capital, and systemic risk: Some international evidence. Journal of Banking & Finance, 69, S25-S34.
López-Espinosa, G., Moreno, A., Rubia, A., & Valderrama, L. (2012). Short-term wholesale funding and systemic risk: A global CoVaR approach. Journal of Banking & Finance, 36(12), 3150-3162.
López-Espinosa, G., Rubia, A., Valderrama, L., & Antón, M. (2013). Good for one, bad for all: Determinants of individual versus systemic risk. Journal of Financial Stability, 9(3), 287-299.
Luciano, E., & Wihlborg, C. (2018). Financial synergies and systemic risk in the organization of bank affiliates. Journal of Banking & Finance, 88, 208-224.
Luo, X., & Bhattacharya, C. B. (2009). The debate over doing good: Corporate social performance, strategic marketing levers, and firm-idiosyncratic risk. Journal of Marketing, 73(6), 198-213.
30
Schäfer, H., Beer, J., Zenker, J., and Fernandes, P. (2006). Who is who in corporate social responsibility rating? A survey of internationally established rating systems that measure corporate responsibility. Bertelsmann Foundation, pp.120.
Scholtens, B. (2008). A note on the interaction between corporate social responsibility and financial performance. Ecological economics, 68(1), 46-55.
Scholtens, B. (2009). Corporate social responsibility in the international banking industry. Journal of Business Ethics, 86(2), 159-175.
Schröder, M. (2014). Financial effects of corporate social responsibility: a literature review. Journal of Sustainable Finance & Investment, 4(4), 337-350.
Shen, C. H., Wu, M. W., Chen, T. H., & Fang, H. (2016). To engage or not to engage in corporate social responsibility: Empirical evidence from global banking sector. Economic Modelling, 55, 207-225.
Škare, M., & Golja, T. (2014). The impact of government CSR supporting policies on economic growth. Journal of policy modeling, 36(3), 562-577.
Sun, W., & Cui, K. (2014). Linking corporate social responsibility to firm default risk. European Management Journal, 32(2), 275-287.
Weiß, G. N., Bostandzic, D., & Neumann, S. (2014). What factors drive systemic risk during international financial crises?. Journal of Banking & Finance, 41, 78-96.
31
8. Appendix
Table A.1 Banks included in dataset
Country Bank
Austria Erste Group Bank
Raiffeisen Zentralbank Österreich
Belgium Dexia KBC Ancora KBC Group Crédit Agricole* BNP Paribas* France Société Générale* Natixis** Germany Commerzbank Deutsche Bank* Deutsche Postbank Alpha Bank Bank of Greece Bank of Piraeus Eurobank Ergasias Greece
National Bank of Greece
Ireland Allied Irish Banks
Bank of Ireland Group Permanent TSB Group
Italy Banca Carige
Banca Monte Dei Paschi
Banca Piccolo Credito Valtellinese
Banca Popolare
Banca Popolare di Sondrio
Banco BPM
BPER Banca
Intesa San Paolo Mediobanca UniCredit*
Unione di Banche Italian
Netherlands ABN AMRO Group
ING Group*
Banco Portugues de Investimento Portugal
Banco Comercial Portugues Banco Esprito Santo
Spain Banca Popular Espanol
Banco de Sabadell Banco Santander
Bankia
Bankinter
Banco Bilbao Vizcaya Argentaria
Caixabank
32
Table A.2 Definitions and data sources of variables used in panel regressions.
Variable Definition Data Source
Risk Measures
Z-Score ROAA + ETA/volatility of ROA, where
volatility is based on a 3-year rolling window
Datastream, own calc.
Stdev. ROE The standard deviation of the return on equity based on a 3-year rolling time window (see e.g. Berger et al., 2016)
WC01651, WC03501
MES The MES is defined as the tail expectation of
the firm’s equity return conditional on a market decline of -2% on that same day. The one-day MES is calculated as the average MES over a year t.
Datastream, own calc.
SRISK Defined as the expected capital shortfall of a financial firm during a period of distress in which the financial market declines substantially
Datastream, own calc.
CSR Rating An equal-weighted measure of the 3 ESG pillars environmental, social and corporate governance
Datastream: ENVSCORE, SOCSCORE, CGVSCORE Bank Control Variables
Size Natural logarithm of total assets WC02999
Tier-1 Ratio Basel III Tier 1 capital/Risk-Weighted Assets WC18157
Deposit Ratio Total deposits/total liabilities WC03019, WC03351
Short-Term Funding Short-term borrowings/total assets WC03019, WC02999 Non- Interest Income Non-interest income/total interest income WC01021, WC01016 Return On Assets Ratio Net income before preferred dividends/total
assets
WC01651, WC02999
Loan-Deposit Ratio Total loans/total deposits WC02271, WC03019
Country Control Variables
GDP Growth Rate Annual percentage growth rate of GDP at market prices based on constant local currency
World Development Indicators
Public Debt to GDP The annual share of public debt as a percentage of its GDP
World Development Indicators
Inflation Rate As measured by the annual implicit GDP deflator in %
World Development Indicators
Market Capitalization The market value of a country of domestically listed companies as a percentage of its GDP
Datastream Unemployment Rate Total unemployment as percentage of total
labor force
World Development Indicators
GDP Gross Domestic Product of a country denoted
in local currency
World Development Indicators
Country CSR Rating An equal-weighted measure of CSR for a country itself, based on proxies for the ESG pillars
Own calc., Yale Environmental
33
Table A3. Empirical Test Statistics
Fisher Type Unit-Root Test H0: All panels contain unit roots
Assets Modified inv. Chi-squared=-1.6493 Prob>chi2= 0.9505
Test: Time-Fixed Effects H0: All time dummies are zero
(1): Z-Score F(22, 325)= 2.37 Prob>F= 0.001
(1): Stdev. ROE F(9, 364)= 3.77 Prob>F= 0.000
(2): SRISK% F(23, 331)= 1.45 Prob>F= 0.084
(2): MES F(10, 319)= 13.21 Prob>F= 0.000
(3): SRISK Share F(14, 49)= 1.02 Prob>F= 0.448
Woolridge Hausman Test H0: Difference in coefficients not systematic
(1): Z-Score* Chi2(17)= 18.67 Prob>chi2= 0.3480
(1): Stdev. ROE Chi2(9)= 15.32 Prob>chi2= 0.0825
(2): SRISK% Chi2(10)= 19.09 Prob>chi2= 0.039
(2): MES Chi2(10)= 299.78 Prob>chi2= 0.000
34
Table A.4 Empirical Analysis Individual CSR Pillars
Z-Score Z-Score Z-Score SRISK% SRISK% SRISK%
Full Model Robust-cluster Full Model Robust-cluster Full Model Robust-cluster Full Model Robust-cluster Full Model Robust-cluster Full Model Robust-cluster CSR Rating: environmental 0.312 -3.698* (0.216) (1.896) CSR Rating: social 0.769** -3.596** (0.355) (1.665) CSR Rating: Governance 0.546 -0.267 (0.357) (1.924) Z-Score -0.305 -0.261 -0.323 (0.239) (0.227) (0.247) Size 25.065 15.839 26.088 70.168 73.578 -11.142 (21.495) (22.638) (21.363) (92.240) (91.026) (85.975)
Tier-1 Capital Ratio 2.031 2.810 2.523 -1.169 -4.473 -0.000
(2.828) (2.850) (2.866) (14.890) (15.758) (14.510) Deposit Ratio -82.423 -74.277 -80.442 -216.306 -324.897 -323.016 (115.668) (109.304) (114.987) (465.978) (493.975) (498.439) Non-Interest Income 5.721 8.643 6.264 -30.324 -47.153 -32.089 (19.944) (19.279) (18.468) (58.044) (62.179) (60.679) Short-term Funding -141.921 -165.732* -168.550* -160.728 -110.304 -159.327 (84.656) (86.587) (91.878) (398.742) (399.369) (439.804) GDP Growth Rate 1.846 1.399 1.770 -21.834 -20.242 -22.021 (1.253) (1.272) (1.251) (16.112) (17.903) (17.808) Public Debt to GDP -0.137 0.022 -0.092 1.301 0.358 0.983 (0.511) (0.467) (0.491) (2.496) (2.394) (2.426) Inflation Rate 3.772 1.923 2.993 92.936** 100.591** 90.543** (7.709) (7.828) (7.909) (42.646) (45.315) (42.682) Market Capitalization -0.313 -0.278 -0.379 -10.334* -10.206 -10.125 (0.569) (0.545) (0.541) (6.098) (6.095) (6.124)
Time Fixed Effects yes yes yes yes yes yes
Bank Fixed Effects yes yes yes yes yes yes
Adjusted R-Squared 0.120 0.140 0.133 0.089 0.087 0.081
35
Table A.5 Empirical Analysis Aggregate Systemic Risk Contribution Individual Pillars
SRISK Share SRISK Share SRISK Share
Pooled OLS Newey-West Pooled OLS Newey-West Pooled OLS Newey-West Log(CSR Rating: Environmental) 0.205***
(0.057)
Log(CSR Rating: Social) 0.295***
(0.053)
Log(CSR Rating: Governance) 0.099**
(0.039) GDP Growth Rate -0.004 -0.004 -0.004 (0.004) (0.005) (0.005) Inflation Rate 0.001 0.011 -0.008 (0.021) (0.019) (0.025) Public Debt to GDP 0.001 0.002** -0.001 (0.001) (0.001) (0.001) Market Capitalization 0.000 -0.000 0.001 (0.001) (0.001) (0.001) Country CSR Rating 0.019* 0.040*** 0.008 (0.010) (0.010) (0.009)
Year Fixed Effects yes yes yes
Adjusted R-Squared 0.224 0.409 0.076
36
Table A.6 Empirical Analysis Z-Score and SRISK% Removal Extreme CSR Values
Z-Score SRISK% Full Model Robust-cluster Full Model Robust-cluster CSR Rating: individual 0.300* -1.485** (0.156) (0.660) Z-Score -0.309 (0.236) Size 16.079 79.460 (23.232) (85.998)
Tier-1 Capital Ratio 2.343 -3.051
(2.903) (15.240) Deposit Ratio -86.567 -242.068 (115.651) (478.770) Non-Interest Income 7.242 -36.502 (18.433) (60.246) Short-term Funding -169.813* -71.552 (90.109) (389.233) GDP Growth Rate 1.399 -20.807 (1.293) (17.337) Public Debt to GDP -0.081 0.786 (0.477) (2.424) Inflation Rate 2.351 99.531** (7.887) (44.620) Market Capitalization -0.234 -10.230 (0.525) (6.236)
Time Fixed Effects yes yes
Bank Fixed Effects yes yes
Adjusted R-Squared 0.139 0.088
37
Table A.7 Empirical Analysis Aggregate Systemic Risk Removal Extreme CSR Values
SRISK Share Pooled OLS Newey-West CSR Rating: sectoral 0.208*** (0.053) GDP Growth Rate -0.004 (0.005) Inflation Rate 0.004 (0.023) Public Debt to GDP 0.001 (0.001) Market Capitalization 0.000 (0.001)
Country ESG Rate 0.022**
(0.010)
Year Fixed Effects yes
Adjusted R-Squared 0.226
38
Table A.8 Empirical Analysis Z-Score and SRISK% Additional Control Variables
Z-Score SRISK% Full Model Robust-cluster Full Model Robust-cluster CSR Rating: individual 0.275* -1.250** (0.152) (0.614) Z-Score -0.291 (0.235) Size 19.907 25.954 (20.600) (104.079)
Tier-1 Capital Ratio 1.149 -1.330
(3.212) (17.670) Loan-to-Deposit Ratio 4.284 -22.678 (5.270) (20.469) Non-Interest Income 5.155 -55.203 (17.784) (69.051) Short-term Funding -135.768* 156.289 (67.844) (418.719)
Return on Assets Ratio 581.012 4,639.873*
(550.186) (2,515.468) GDP Growth Rate 0.820 -11.030 (1.773) (18.009) Public Debt to GDP 0.082 -1.922 (0.626) (3.547) Inflation Rate 0.958 80.491* (8.528) (46.878) Market Capitalization -0.419 -9.486* (0.549) (5.479) Unemployment Rate -0.329 -23.143* (2.140) (13.542) Log(GDP) 70.667 -2,305.939 (127.618) (1,587.864)
Time Fixed Effects yes yes
Bank Fixed Effects yes yes
Adjusted R-Squared 0.139 0.108
