Board members' educational background and financial performance: Evidence from Eurozone banks

Pereira, Vítor Manuel Morais; Filipe, José António Article Board members' educational background and financial performance: Evidence from Eurozone banks Journal of Central Banking Theory and Practice Provided in Cooperation with: Central Bank of Montenegro, Podgorica Suggested Citation: Pereira, Vítor Manuel Morais; Filipe, José António (2022) : Board members' educational background and financial performance: Evidence from Eurozone banks, Journal of Central Banking Theory and Practice, ISSN 2336-9205, Sciendo, Warsaw, Vol. 11, Iss. 3, pp. 203-227, https://doi.org/10.2478/jcbtp-2022-0030 This Version is available at: https://hdl.handle.net/10419/299058 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks 203 Vítor Manuel Morais Pereira *, José António Candeias Bonito Filipe ** Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks Abstract: This study examines the impact of the quality of board members’ educational background on the financial performance of a group of Eurozone banks. Return on average assets (ROAA), Return on average equity (ROAE), and Tobin’s Q are used as measures to assess bank financial performance. Three indices are used as proxies for board members’ educational qualifications: Eduindex, for all academic qualifications in areas such as business or economics; Eduversal, for all qualifications from business schools ranked by Eduniversal; and EduFT, for all qualifications from business schools ranked by the Financial Times. Our study results offer relevant policy implications. Generally, there is a significant negative effect from Eduversal and EduFT qualifications on bank financial performance. This effect can be explained by the fact that some well-qualified board members use their expertise for their own interest, which, in most cases, is not favourable for bank financial performance. The implication is that the European Central Bank needs to implement more rigorous measures than those currently imposed to control bank board member behaviour and reduce agency problems. Keywords: Bank financial performance; corporate governance; board of directors; educational background; Eurozone banking. JEL Classification: C10, G20. * CEFAGE-ISCAL, ISCAL – Instituto Superior de Contabilidade e Administração de Lisboa, Instituto Politécnico de Lisboa, Portugal E-mail: [email protected] * BRU-IUL – Business Research Unit, ISTAR-IUL – Information Sciences and Technologies and Architecture Research Center, Instituto Universitário de Lisboa (ISCTE-IUL), Portugal E-mail: [email protected] Journal of Central Banking Theory and Practice, 2022, 3, pp. 203-227 Received: 13 July 2021; accepted: 11 November 2021 UDK: 336.711(4) DOI: 10.2478/jcbtp-2022-0030 Journal of Central Banking Theory and Practice 204 1. Introduction In the last decade, banks have remained a focus of attention for international financial institutions, governments, and their national central banks. Many believe that corporate governance remains fundamental to maintaining the stability of the financial system. For instance, The Basel Committee on Banking Supervision (2015) has stated that effective corporate governance is essential to guarantee the normal functioning of the banking sector. In addition, the European Banking Authority and European Securities and Markets Authority (2017) has indicated that weaknesses in corporate governance have contributed to the excessive and imprudent risk-taking that has caused the failure of several institutions. There are several real cases among the Eurozone banks that have had significant problems where one relevant cause is bad corporate governance practice. The following examples of banks with relevant problems in the Eurozone include: Banco Espírito Santo in Portugal (2014); Hypo Real Estate in Germany (2008); and Allied Irish Banks in Ireland (2009). Notably, banks have their own particularities based on the opacity of many services and the complexity of some products they sell. However, as stated by Adams and Mehran (2012), many governance reforms have not taken into consideration the particularities in the banking sector, which can increase the level of risk. John, De Masi, and Paci (2016) argue that banks have special features that increase the probability of governance problems. These authors illustrate this situation by referring to the high leverage of banks; and, comparing this to the situation in nonfinancial companies, the main providers of capital to banks are depositors and other debt holders. One relevant characteristic of bank boards is the financial expertise and professional experience of each bank board member. As posed by John et al. (2016), there are several questions that have been asked by scholars and regulators in terms of bank boards. For example, is the quality of training of the board members relevant? If a bank board member holds a specific degree (PhD or MBA), will that improve bank performance? The literature has documented some previous research questions in this context. Hau and Thum (2010) discovered that the education of the board members of German banks did not have a significant correlation with bank losses. By contrast, Nguyen, Hagendorff, and Eshraghi (2015) found opposite results claiming that the education of the executives in the US banking sector created shareholder wealth. We propose a study that considers the impact of bank board members’ Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks 205 education on bank financial performance. The literature on this topic is scarce and other relevant literature on this issue has been applied only to mutual fund managers. For instance, Chevalier and Ellison (1999) found that fund managers who had MBAs did not present a significantly better performance compared with those who did not have MBAs. Another stream of research has considered several board characteristics and their effects in terms of bank performance. In particular, the following variables have been emphasized: board size; percentage of directors who are independent; annual compensation of the board of directors; percentage of foreign directors; and director age. In general, the impact of these board characteristics on bank performance has been contradictory in several empirical studies, which we will discuss in the literature review. Taking into consideration the state-of-the-art corporate governance policies of banks, we identify a gap in the literature on the effects of the quality of training of bank board members on bank financial performance in the context of the main banks operating in the Eurozone. There is no consensus in the literature that human capital resources can predict bank risk-taking or performance. This topic is the subject of a study by Minton, Taillard, and Williamson (2014) indicating that independent directors with financial expertise supported increased risk taking prior to the financial crisis. However, there are few studies that consider banks in their samples. Thus, our study aims to expand the extant literature on corporate governance of banks and contribute to a better understanding of the effects of the board members’ quality of training on bank financial performance. This study contributes to the literature in the following ways. First, the study represents a new perspective specifically applied to the banking sector. Second, our study presents a broader view of corporate governance by considering the main Eurozone banks. Third, compared with V. Pereira and Filipe (2016), we present a new perspective in terms of the quality of the training investigated by considering the Financial Times Business School Rankings. Our results have implications for the prudential supervision that needs to be developed by national central banks and the European Central Bank. This study uses quantitative and qualitative perspectives. In terms of board member qualifications, the study considers the period between 2011 and 2013. Additionally, measures of operating performance (e.g. ROAA) and stock return performance are used as dependent variables. In terms of independent variables, we construct indices measuring the quantity and quality of the qualifications of the board members. Journal of Central Banking Theory and Practice 206 The study design includes several means to validate the results. First, as suggested by Boyd, Adams, and Gove (2017) we apply a power analysis to consider the small effect sizes found in governance studies. Second, research methods in corporate governance must control possible effects of endogeneity. Accordingly, we include a lagged design to protect against possible reverse causality effects. The remainder of the paper proceeds as follows. Section 2 presents the literature review and the hypotheses development. Section 3 discusses our data and methodology. Section 4 presents the empirical results. Section 5 provides the robustness tests; and finally, Section 6 offers our conclusions. 2. Literature review and hypotheses development 2.1. Board member education level and bank financial performance The literature on the corporate governance of banks includes a specific stream of research that studies the association between board member education and bank performance. However, when considering specific samples of banks, there are few empirical studies in this stream. The most relevant studies consider samples of mutual funds, as in the case of Golec (1996) who argued that investors could expect a better risk-adjusted performance from younger managers with MBA degrees. By contrast, a study made by Chevalier and Ellison (1999) on mutual funds claimed that managers with MBAs did not show a meaningfully better performance than those without them. Another study, by Graham and Harvey (2001) discovered that CEOs with MBAs had a higher probability of using the net present value, which can be considered a sign of more financial expertise. Bertrand and Schoar (2003) argued that CEOs with MBAs related to higher operating returns on assets. In the case of banks, we found a study by Kauko (2009) arguing that the impact of age on efficiency depended on education. Kaukoalso stated that a university degree was useful mainly in the case of the largest banks in the sample. Another study from Hau and Thum (2010) indicated that, on average, board member education did not have a significant correlation with bank losses. In their study, Berger, Kick, and Schaeck (2014), using a sample of German banks, discovered that the presence of executives with PhD degrees related to a decline in portfolio risk. Moreover, Nguyen et al. (2015) claimed that in the US banking sector, the education of executives generated shareholder wealth. Furthermore, V. M. M. Pereira and Filipe (2015) identified a significant effect of bank board member education on a bank`s ROAE. Additionally, King, Srivastav, and Williams (2016) Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks 207 emphasized that CEO education mattered, both in terms of level and quality, for bank performance. Another study by Fernandes, Farinha, Martins, and Mateus (2017) did not identify a statistically significant effect of education on the banks’ financial performance. Finally, Gande and Kalpathy (2017) discovered that banks with CEOs with MBAs from a top 20 schools registered improvements in terms of ‘buy and hold’ returns. Notably, the majority of these studies that associate board member education and bank performance generally used US bank samples and considered only the CEO’s education. John et al. (2016) stated that current literature on bank corporate governance was indecisive when considering the effects of financial expertise on bank performance. Therefore, considering the aforementioned, we believe that board member education is important to guarantee that board members are able to do a good job in terms of their managing capability. Further, we believe that the resource dependence theory applies here to our first hypothesis, as we think bank board members use their resources (e.g. education) to make the best decisions to positively affect bank performance. Accordingly, our first hypothesis (H1) is formulated as follows: H1: Degrees and/or executive development in the areas of economics/business of board members are positively related to bank financial performance. 2.2. Board member education from prestigious business schools and bank financial performance In terms of the relationship between board member education and bank financial performance, not only is it relevant to consider the number of degrees each board member has, but also the quality of the education. This relevant research topic captures the quality of the training of the board members. Gottesman and Morey (2006) argued that mutual fund managers who had MBAs from schools ranked in the top 30 of the Business Week rankings of MBA programs revealed better performance. Moreover, King et al. (2016) showed that bank CEOs with a higher level of quality of management education usually influenced better firm performance. However, they considered only undergraduate study and PhDs did not seem to be significant. Here, we investigate whether the quality of the education of bank board members is relevant to bank financial performance. However, in this study, we consider all board members, not only CEOs, in contrast to the study of King et al. (2016). In Journal of Central Banking Theory and Practice 208 our study, we consider a sample of Eurozone banks, offering a diverse perspective when compared with traditional US bank samples. We conjecture that bank board members who hold degrees from business schools considered in the Financial Times Business Rankings, or in the Eduniversal Ranking, may have better technical skills. This, then, must imply better management decisions, which ultimately will affect bank performance positively. Hence, our second hypothesis (H2) is the following: H2: Degrees and/or executive development from business schools considered in the Financial Times Business Ranking, or in the Eduniversal Ranking, of bank board members are positively related to bank financial performance. Again, in this second hypothesis, we assume the resource dependence theory point of view, supposing that education will be used in a positive way by bank board members to improve the bank’s financial performance. 3. Data and Methods 3.1. Timeline The current study considers corporate governance data for the period between 2011 and 2013. We choose that period because it involved several measures and modifications in terms of banking systems in several Eurozone countries. This may be considered a natural experiment considering the corporate governance of banks. To avoid possible endogeneity effects, we use a lagged design in the proposed econometric model. We consider corporate governance data from the period 2011-2013 and the effects in terms of bank financial performance in the period 2012-2014. The visualization of data allows us to identify two general aspects. First, as referenced by Adams (2017), many corporate governance variables do not change considerably over time, for example, the variable of ownership. In this study, we identify the same situation in the particular case of board member education. Second, board members are typically nominated for a period of three years. As there is a high rotation of board members, we consider three years for the analysis. For illustration, we assume that, in general, a bank board member is nominated for a period of three years and will not stay on that bank board after that period. Moreover, after being nominated as a board member, the possible impact of Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks 209 his/her decisions in terms of bank performance will take at least a year to register. Thus, the aforementioned aspect is the main reason why we choose a lagged design. 3.2. Bank Sample Initially, our sample included the major financial institutions of the various countries of the Eurozone. We decided on the Eurozone region for several different reasons. First, relevant corporate governance research is focused mainly on US samples; thus, we wanted to test previous empirical results in a different region. Second, the Eurozone banking system is supervised by the European Central Bank, which guarantees a higher level of comparison that is critical in a multi-country study. Third, the Eurozone banking system has imposed several constraints in the last decade, which allows us to consider this regional banking system as a natural experiment for the current research development. However, ultimately, we opted only for countries where there was a significant market share of state-owned banks. The reason for this relates to the fact that we wanted to analyse whether there are relevant differences between governmentcontrolled banks and non-government-controlled banks in terms of board member education. For that purpose, we needed a sample of banks from countries where there was a representative presence of government-controlled banks. We followed the study by Čihák, Demirgüç-Kunt, Pería, and Mohseni-Cheraghlou (2012) to define the cut-off percentage for selecting countries where supposedly government control of banks is relevant. We decided to use a cut-off percentage of 10% in terms of banking system’s assets that are government controlled. Taking into consideration the most recent Bank Regulation and Supervision Survey (World Bank) from 2012, which covered the time period 2008-2010, we identified that in 2010, there were seven Eurozone countries that fulfilled the cut-off percentage of 10%: Austria, Germany, Greece, Ireland, Netherlands, Portugal, and Slovenia. However, considering that there are no data from the Boardex for the case of Slovenia, we excluded banks from this country. We also obtained all available information on board members from the Boardex database, annual reports, interim reports, and press releases. Moreover, financial data were obtained from the Bankscope database (currently named Moody’s Analytics BankFocus) and from annual reports. Notably, only major banks, in general, have the detailed information necessary for our study; this was one reason why these banks were the ones considered in Journal of Central Banking Theory and Practice 210 each country. In fact, usually only major banks had detailed corporate governance information in their annual reports. Another reason was the fact that this task of hand-collecting data is extraordinarily time consuming and thus, it is not feasible to include a larger number of banks. Given that we found some errors in the databases, we manually checked data to prevent big differences in the final results. For that, we used other databases: Zoominfo and Bloomberg. Initially, board member data included 326 elements, but in the final phase we could use only 302 elements as in the other cases there was no available information. 3.3. Construct measurement As previously stated, our study uses data from six Eurozone countries: Austria, Germany, Greece, Ireland, the Netherlands, and Portugal. As is common in corporate governance research, when the sample includes various countries, researchers need to be aware of possible problems with comparative situations. Consequently, taking this into consideration, we selected a group of countries in the Eurozone; this means that these banks respond, in general, to common legislation, procedures, and rules, allowing comparative evaluations. In fact, the Eurozone banking system is supervised by a common institution: the European Central Bank (ECB). However, each country has leeway to implement the European rules, which allows some differences among the countries in the Eurozone. To deal with the specificities of each country, we consider the inclusion of country fixed effects. 3.4. Main variables 3.4.1. Bank performance measures In this study, we consider two types of measures of bank performance. First, we consider a market performance measure as the sample includes publicly traded banks. Therefore, we use a similar methodology to that of Adams and Mehran (2012), which considered a proxy for Tobin’s Q. However, in this case, we use the Moody’s Analytics BankFocus methodology directly considering the ratio market capitalization to total assets. We also consider two accounting measures for bank performance, which are essential due to the inclusion of state-owned banks in the sample. First, we consid- Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks 217 Table 2: Summary Statistics Variable Obs Mean Std. Dev. Min Max  ROAA 96 -.1784167 2.381599 -20.62 4.429 ROAE 96 6.769333 43.26251 -106.628 322.87 Tobin’s Q 79 .311281 .4188497 .005 1.0661 Eduindex 135 1.180928 .4180613 .3333333 2 Eduversal 138 .9873188 .2046579 .75 1.375  Eduft 129 .3790567 .3716312 0 1.25 Banksize 90 10.96948 1.619581 7.180831 14.5148 Government 135 .4 .4917225 0 1 PublicTraded 135 .637037 .4826452 0 1 Tier 1 ratio 76 13.40368 3.535967 8.5 28  Riskroaa 138 1.049077 1.393172 .0431292 6.828177 Beta 69 .9816116 3.254256 -12.8759 7.2953 ROAE Crisis 132 -13.31189 93.05915 -992.293 18.204 ROAA Crisis 135 -.2218741 1.840748 -12.367 1.714 Tobin’s Q Crisis 48 .6457833 .4709016 .005 1.0861 Table 3 shows the Pearson correlation coefficients between independent variables and control variables. We identify one case where there is a possible problem of multicollinearity: there is a value greater than 0.8 in the correlation between the variables ROAACrisis and ROAECrisis. Table 3: Correlation Matrix Eduindex Eduversal EduFT Banksize Government Publictraded Tier1ratio RiskROAA Beta ROAECrisis ROAACrisis Tobin's Q Crisis Eduindex 1.0000 Eduversal 0.4140 1.0000 EduFT 0.5111 0.2014 1.0000 Banksize 0.1011 -0.1726 -0.1299 1.0000 Government 0.1192 -0.0540 -0.1400 0.4714 1.0000 Publictraded 0.2909 0.4751 0.2008 -0.0693 -0.4214 1.0000 Tier1ratio 0.1063 0.1196 0.0009 -0.2354 -0.1859 0.0222 1.0000 RiskROAA 0.4465 0.2796 0.1109 0.0558 0.3703 0.0303 -0.2005 1.0000 Beta -0.1440 -0.1364 0.1202 0.0045 0.1104 -0.2565 -0.0843 -0.1331 1.0000 ROAECrisis -0.0483 -0.1541 0.0573 -0.0619 -0.2205 -0.0318 0.0367 -0.2550 -0.0105 1.0000 ROAACrisis -0.0821 -0.1594 0.0868 -0.1436 -0.2936 0.0105 0.0106 -0.3005 -0.0720 0.8202 1.0000 Tobin's Q Crisis -0.2368 -0.4447 -0.1973 0.1036 -0.1404 -0.1360 0.0729 -0.0723 -0.2275 0.2486 0.3217 1.0000 Journal of Central Banking Theory and Practice 218 Notably, the Pearson correlation coefficients are more appropriate for variables that follow a normal distribution (Clark (2013)). In this study, three variables assume a normal distribution using the Shapiro-Wilk Normality test: Eduindex, Government, and Publictraded. Consequently, the Pearson’s correlation should not be used with the other variables; as a viable alternative the Spearman’s Rank Correlation can be used. We detect some relevant relations in terms of the following variables using Spearman’s correlation: EduFT vs. Eduindex with a value of 0.7233; ROAECrisis vs. RiskROAA with a value of -0.8505; ROAACrisis vs. RiskROAA with a value of -0.8143; and ROAACrisis vs. ROAECrisis with a value of 0.9591. We also intend to verify if a monotonic relationship exists between variables that do not follow a normal distribution by using a two-way graph. In fact, we only detect one possible monotonic relationship between the variables ROAECrisis and ROAACrisis. Similarly, as referenced in V. M. Pereira and Filipe (2018), for non-monotonic relationships, it may be a good idea to implement Hoeffding’s D measure. However, we cannot implement this in the Stata program. 5.2. Empirical results To test the effects of board members’ characteristics on bank financial performance we first consider a traditional ordinary least squares (OLS) regression with different control variables (see Table 4). Moreover, we identify a significant negative effect of Eduversal on Banks’ ROAA, namely, an increase in the number of board members with degrees from business schools ranked at Eduniversal might have a negative effect on banks’ ROAA. This result is counter to our expectations, meaning that the methodology we used might not have been the best option for the current situation. Moreover, the other variables Eduindex and EduFT did not have significant effects on ROAA, ROAE, or Tobin’s Q. Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks 219 Table 4: OLS Performance Regressions with Control Variables ROAA ROAA ROAA ROAE ROAE ROAE Tobin's Q Tobin's Q Tobin's Q Intercept -1.21152 0.1504581 -1.231024 -16.79549 -29.61411 -18.70641 2.196371*** 1.683326*** 1.702485** (-0.93) (0.10) (-0.95) (-0.24) (-0.36) (-0.27) (3.62) (3.08) (2.56) Eduindex 0.0041983 4.109641 0.2663637 (0.01) (0.26) (1.49) Eduversal -1.43639*12.00381 -0.0514536 (-1.75) (0.27) (-0.14) EduFT -0.0832908 -10.45995 0.0239104 (-0.28) (-0.67) (0.11) Banksize 0.0489528 0.0155418 0.0508281 -1.215473 -0.6487595 -0.8318646 -0.1165273*** -0.0796184** -0.0825795* (0.51) (0.17) (0.55) (-0.24) (-0.13) (-0.17) (-2.89) (-2.40) (-1.95) Government -0.5069273 -0.5476709*-0.4885564 17.87601 18.37295 20.58774 -0.1025894 -0.0124605 -0.0259126 (-1.63) (-1.80) (-1.54) (1.08) (1.11) (1.22) (-0.78) (-0.09) (-0.19) Publictraded -0.4039202 -0.2825176 -0.3652354 7.833344 8.424181 13.48235 -0.4872532 -0.2482142 -0.2735959 (-1.16) (-0.86) (-1.04) (0.42) (0.47) (0.71) (-1.55) (-0.84) (-0.86) Tier 1 ratio 0.0820953** 0.0987473*** 0.0823408** 0.8839899 0.8465778 0.9966214 -0.049112*** -0.0327345** -0.0341621** (2.56) (3.08) (2.63) (0.52) (0.49) (0.61) (-2.89) (-2.09) (-2.33) RiskROAA 0.1614337 0.3436486** 0.1559466 13.47323*12.3719 13.08321* (1.20) (2.08) (1.17) (1.95) (1.41) (1.92) ROAACrisis 0.1240823** 0.1313866** 0.1248142** (2.22) (2.40) (2.24) ROAECrisis 0.117292** 0.1169728** 0.1196951** (2.26) (2.25) (2.32) Beta 0.0033203 0.003219 0.0031664 (0.19) (0.17) (0.17) Tobins' Q Crisis 0.2628858** 0.1807662 0.1918562* (2.50) (1.59) (1.92) Obs. 76 76 76 73 73 73 33 33 33 Adj-R2 0.1127 0.1510 0.1137 0.0628 0.0629 0.0682 0.2472 0.1813 0.1811 The regressions controls are bank size measured by the natural logarithm of total assets, Tier 1 ratio. The table reports the regression coefficients, t-statistics (in parentheses), number of observations (N) and adjusted R2. * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level. Second, we consider a different methodology that uses heteroskedastic linear regressions (see Table 5). We find a significant negative effect of EduFT on ROAA. Moreover, we are not able to perform the regressions that use ROAE and Tobin’s Q as dependent variables as the convergence cannot be achieved in the Stata software. Journal of Central Banking Theory and Practice 220 Table 5: Heteroskedastic Linear Regression ROAA# ROAA# ROAA# ROAE& ROAE& ROAE& Tobin's Q& Tobin's Q& Tobin's Q& Intercept 0.8094535 0.6751355 0.442707 (1.43) (0.89) (1.05) Eduindex 0.0107243 (0.08) Eduversal -0.2150003 (-0.40) EduFT -0.1921018* (-1.90) Banksize -0.0484286 -0.0371524 -0.0254039 (-1.37) (-1.02) (-1.04) Government -0.1910882*-0.152043 -0.1264587 (-1.77) (-1.03) (-1.04) Publictraded -0.1865278** -0.1449578 -0.0890837 (-2.00) (-0.93) (-0.84) Tier 1 ratio 0.009908 0.0222431 0.0159954 (0.59) (1.13) (1.18) RiskROAA -0.1584233 -0.1922381 -0.1624031 (-0.45) (-0.62) (-0.72) ROAACrisis 0.147791 0.1360004 0.1480051 (1.16) (1.25) (1.62) ROAECrisis Beta Tobins' Q Crisis Obs. 76 76 76 Wald chi2(8) 24.52 21.30 32.32 Prob > chi2 0.0009 0.0033 0.0000 The regressions controls are bank size measured by the natural logarithm of total assets, Tier 1 ratio. The table reports the regression coefficients, z-statistics (in parentheses), number of observations (N). * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level. # In this case, the likelihood-ratio test reported at the bottom of the table shows that the model of the variance fits the data better than a model in which the variance is constant. & In this case, the convergence is not achieved. Third, we include a linear regression with panel-corrected standard errors (PCSE) in the empirical analysis, which can consider disturbances that are not assumed to be independent and identically distributed. First, we detect a significant negative effect of Eduversal and EduFT on bank ROAA. Second, we see a significant Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks 221 negative effect of EduFT on ROAE. Third, we see a significant negative effect of Eduversal on Tobin’s Q as well. Table 6: Linear Regression with Panel-Corrected Standard Errors (PCSE) ROAA ROAA ROAA ROAE ROAE ROAE Tobin's Q Tobin's Q Tobin's Q Intercept -1.975003*** -0.6456164 -2.120044*** -50.17831*** -63.12878*** -50.78874*** 1.87371*** 1.631757*** 1.465164*** (-2.63) (-0.98) (-2.86) (-3.57) (-3.24) (-4.20) (5.00) (11.41) (8.68) Eduindex -0.0890973 2.953715 0.1423056 (-0.46) (0.53) (1.05) Eduversal -1.617571*15.52272 -0.1669045** (-1.92) (0.99) (-1.83) EduFT -0.1171203*** -11.70007*** -0.0503578 (-6.00) (-4.03) (-0.76) Banksize 0.0824805 0.0490949*0.0843555*-0.5941634 -0.0587525 -0.2962687 -0.0927282*** -0.0708331*** -0.0650567*** (1.56) (1.70) (1.94) (-0.60) (-0.15) (-0.45) (-3.68) (-11.95) (-5.98) Government -0.4408951 -0.4743156 -0.4083278 26.09568*** 25.87066*** 29.00413*** -0.0702166 -0.0020793 -0.0172941 (-1.20) (-1.07) (-1.01) (7.57) (5.30) (6.33) (-1.22) (-0.07) (-0.77) Publictraded -0.3487226** -0.232551 -0.3316216 15.29796*** 14.467*** 21.08117*** -0.3878063*** -0.2344489*** -0.2454787*** (-2.34) (-1.40) (-1.36) (4.92) (3.58) (3.42) (-3.23) (-25.14) (-6.37) Tier 1 ratio 0.1130725*** 0.1376985*** 0.1171523*** 2.37104*** 2.123985*** 2.402081*** -0.0403658*** -0.0264037*** -0.0293495*** (5.35) (6.96) (6.54) (3.27) (3.01) (3.82) (-3.98) (-4.80) (-5.96) RiskROAA 0.1795493 0.3695339 0.1650782 13.41147*** 11.85109*** 12.66567*** (0.94) (1.28) (0.89) (3.17) (3.34) (3.03) ROAACrisis 0.1699699*** 0.1882*** 0.1772759*** (10.13) (-0.98) (10.48) ROAECrisis 0.1158364*** 0.1157811*** 0.1166948*** (9.37) (8.77) (9.02) Beta 0.00208 0.0000978 0.0010725 (0.35) (0.03) (0.38) Tobins' Q Crisis 0.1882069*** 0.1003522*** 0.1191189*** (2.92) (3.74) (4.37) Obs. 76 76 76 73 73 73 33 33 33 R-squared 0.2359 0.2751 0.2443 0.1204 0.1232 0.1232 0.3715 0.3470 0.3428 Wald chi2(3) 38.25 304.21 94.71 391.59 11.60 316.96 94.26 867.96 309.62 Prob > chi2 0.0000 0.0000 0.0000 0.0000 0.0089 0.0000 0.0000 0.0000 0.0000 rho 0.2704816 0.3554793 0.3130976 0.4751402 0.4348918 0.4720382 0.3084633 0.5643973 0.5226201 The regressions controls are bank size measured by the natural logarithm of total assets, Tier 1 ratio. The table reports the regression coefficients, z-statistics (in parentheses), number of observations (N). * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level. Fourth, a random-effects model is included in the analysis as it is common in corporate governance research. Moreover, as referenced by Torres-Reyna (2017), fixed effects should not be used with slow-changing variables over time. Accord- Journal of Central Banking Theory and Practice 222 ingly, we consider a random-effects model as the best methodology for the econometric model, as board member education is considered as having little time variation. Furthermore, we use the Breusch-Pagan Lagrange multiplier (LM) to define whether it is worth considering a random-effects regression or a simple OLS regression. The analysis shows a negative and significant effect of Eduversal on banks’ ROAA. Finally, according to data, in most cases, the random-effects model is not a better option than the OLS regression (Table 7). Table 7: Random Effects Regression (Breusch and Pagan Lagrangian Multiplier Test) ROAA ROAA ROAA ROAE ROAE ROAE Tobin's Q Tobin's Q# Tobin's Q# Intercept -1.530739 0.035579 -1.536203 -16.79549 -29.61411 -18.70641 0.9866033*** 1.298684*** 0.9305407*** (-0.97) (0.02) (-1.00) (-0.56) (-0.94) (-0.61) (3.74) (2.92) (4.24) Eduindex -0.0277765 4.109641 -0.1498241 (-0.10) (0.43) (-0.79) Eduversal -1.479588** 12.00381 -0.4064646 (-1.99) (0.49) (-1.23) EduFT -0.0834019 -10.45995 -0.1521954 (-0.46) (-1.59) (-0.67) Banksize 0.0634272 0.0208295 0.0629278 -1.215473 -0.6487595 -0.8318646 -0.0491005*** -0.0534382*** -0.0490802*** (0.64) (0.24) (0.68) (-0.62) (-0.31) (-0.41) (-2.62) (-2.95) (-2.79) Government -0.5071127*-0.5487601*-0.4895604 17.87601 18.37295 20.58774 0.0400683 0.0414293 0.0209419 (-1.67) (-1.74) (-1.59) (1.09) (1.16) (1.28) (0.66) (0.67) (0.32) Publictraded -0.3945561 -0.2783464 -0.3692399 7.833344 8.424181 13.48235 -0.1643931 -0.1971408*-0.2304383* (-1.39) (-1.07) (-1.28) (0.54) (0.62) (0.97) (-0.86) (-1.83) (-1.74) Tier 1 ratio 0.0948365* 0.1051687** 0.0942841*0.8839899 0.8465778 0.9966214 0.0002947 0.0003304 0.0002972 (1.93) (2.31) (1.96) (0.75) (0.69) (0.96) (0.15) (0.17) (0.15) RiskROAA 0.1799483 0.3551963*0.1710246 13.47323** 12.3719** 13.08321** (0.95) (1.69) (0.94) (2.24) (2.47) (2.08) ROAACrisis 0.1281839 0.1330274 0.128426 (1.44) (1.52) (1.45) ROAECrisis 0.117292** 0.1169728** 0.1196951** (2.43) (2.46) (2.42) Beta -0.0002428 -0.0002078 -0.0002404 (-1.03) (-0.92) (-1.04) Tobins' Q Crisis 0.0000155 -0.0000149 0.00000762 (0.00) (-0.00) (0.00) Obs. 76 76 76 73 73 73 33 33 33 R-squared (overall) 0.1945 0.23 0.1955 0.1539 0.154 0.1588 0.1433 0.2113 0.1369 The regressions controls are bank size measured by the natural logarithm of total assets, Tier 1 ratio. The table reports the regression coefficients, z-statistics (in parentheses), number of observations (N) and adjusted R2. * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level. # In this case, the random effects model fits better than OLS. Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks 223 Remembering that most of our research variables are not normally distributed, it is worth considering a nonparametric regression that is not subject to misspecification errors. Therefore, we perform nonparametric regressions using dependent variables (ROAA, ROAE, and Tobin’s Q) and independent variables (Eduindex, Eduversal, and EduFT). Consequently, we include control variables in the nonparametric regression, but the bootstrap does not work due to insufficient observations. Therefore, our results suggest that EduFT affects the banks’ Tobin’s Q significantly and negatively. Finally, we recognize that these results have limitations. Table 8: Nonparametric Regression ROAA ROAA ROAA ROAE ROAE ROAE Tobin's Q Tobin's Q Tobin's Q Dependent variable mean -0.1784167 -0.1784167 -0.1430705 6.769333*8.21051 6.769333*0.3253504*** 0.3395149*** 0.2694477*** (-0.79) (-0.76) (-0.61) (1.77) (1.54) (1.65) (8.30) (6.99) (7.56) Eduindex -0.9216678 11.17186 0.0139132 (-1.00) (1.16) (0.11) Eduversal 0.8002118 48.3279 -0.3821559 (0.81) (1.31) (-1.64) EduFT -0.8897961 -5.274097 -0.3174367*** (-0.92) (-1.34) (-3.47) Obs. 96 96 96 96 96 96 79 79 76 R-squared 0.0292 0.0042 0.0538 0.0130 0.0628 0.0022 0.2132 0.1087 0.0927 The regressions controls are bank size measured by the natural logarithm of total assets, Tier 1 ratio. The table reports the regression coefficients, z-statistics (in parentheses), number of observations (N). * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level. Furthermore, we include a Hausman-Taylor regression in the analysis. As referred to by V. M. Pereira and Filipe (2018) this regression considers that covariates can correlate with unobserved individual-level random effects. The results in Table 9 do not reveal any significant effect of Eduindex, Eduversal, and EduFT on bank performance measured by ROAA, ROAE, and Tobin’s Q. Journal of Central Banking Theory and Practice 224 Table 9: Hausman-Taylor Regression ROAA ROAA ROAA ROAE ROAE ROAE Tobin's Q Tobin's Q Tobin's Q Intercept -26.51289 -32.69192 -44.11273*-36.87272 -32.52721 242.8973 1.25877*** 1.810014** 0.9682293*** (-1.59) (-1.40) (-1.78) (-0.26) (-0.23) (0.24) (3.31) (2.48) (3.98) Eduindex -13.04344 11.44542 -0.6545753 (-1.32) (0.10) (-1.12) Eduversal 8.093455 -8.485038 -0.8967529 (0.74) (-0.04) (-1.34) EduFT -5.981572 321.7828 -0.5048545 (-0.52) (0.24) (-0.83) Banksize 2.875366*1.994994 3.576365*-0.6118763 0.2113671 -25.93309 -0.0536761*** -0.0578434*** -0.0517745*** (1.65) (1.53) (1.75) (-0.04) (0.05) (-0.29) (-2.70) (-3.24) (-2.95) Government -1.489051 -1.523575 -1.860457 18.15053 18.44339 -38.38957 0.2147593 0.0842383 0.1124147 (-0.66) (-0.95) (-0.61) (0.94) (1.13) (-0.13) (0.96) (0.82) (0.59) Publictraded 6.712925 -0.1509423 4.308044 8.964789 14.32175 -130.9359 0.1666176 -0.1805864 -0.1444085 (1.27) (-0.08) (0.71) (0.17) (0.56) (-0.22) (0.31) (-1.00) (-0.44) Tier 1 ratio 0.2427759** 0.236251** 0.2404936** 1.163332 1.33893 1.279547 0.0004491 0.0004314 0.0003884 (2.18) (2.05) (2.26) (0.26) (0.34) (0.30) (0.23) (0.22) (0.20) RiskROAA 1.473178 -0.2814041 0.4031144 12.8609** 14.72085 36.95322 (1.21) (-0.23) (0.35) (2.11) (0.75) (0.32) ROAACrisis 0.1994084*0.1892482*0.2051048** (1.95) (1.90) (1.97) ROAECrisis 0.1203349*** 0.121297** 0.1177335*** (2.65) (2.56) (2.75) Beta -0.0002035 -0.0001694 -0.000217 (-0.78) (-0.71) (-0.89) Tobins' Q Crisis -0.0002734 -0.0001829 -0.0001671 (-0.04) (-0.03) (-0.03) Obs. 76 76 76 73 73 73 33 33 33 rho 0.99441201 0.98688975 0.99972494 0.72108312 0.68481828 0.99347977 0.99967857 0.99965609 0.9996318 The regressions controls are bank size measured by the natural logarithm of total assets, Tier 1 ratio. The table reports the regression coefficients, z-statistics (in parentheses), number of observations (N). * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level. 6. Conclusions As indicated by V. M. Pereira and Filipe (2018), the influence of the educational attainment of board members on bank financial performance is scarce in the literature. Therefore, there is a need to expand the existing research to multiple countries and consider other business school rankings, such as the Financial Times Ranking. Our study responds to these gaps in the literature. Board Members’ Educational Background and Financial Performance: Evidence from Eurozone Banks 225 Our results reveal that, in most cases, there is a significant and negative effect of Eduversal and EduFT on bank financial performance. These results differ from our expectations from a resource dependence perspective. We believe that these results might be explained by the fact that some well-qualified board members use their expertise for their own interest, which, in most cases, is not favourable for the banks’ financial performance. We believe our research findings are particularly relevant. As such, we recommend that the European Central Bank implement more rigorous measures to control bank board member behaviour and to reduce agency problems issues. We suggest that future research consider whether the current mechanisms of controlling agency problems in banks are truly efficient and, if not, what can be done to improve such mechanisms. Acknowledgements The authors acknowledge financial support from the Instituto Politécnico de Lisboa, research project IPL/2017/CGB/ISCAL and research project IPL/2018/ IIPB_ISCAL. Journal of Central Banking Theory and Practice 226 References 1. Adams, R. (2017). The ABCs of empirical corporate (governance) research. Corporate Governance: An International Review, 25(6), 461-464. 2. Adams, R., & Mehran, H. (2012). Bank board structure and performance: Evidence for large bank holding companies. Journal of Financial Intermediation, 21(2), 243. doi:10.1016/j.jfi.2011.09.002 3. Basel Committee on Banking Supervision. (2015). Corporate Governance Principles for Banks. Retrieved from https://www.bis.org/bcbs/publ/d328.pdf 4. Berger, A. N., Kick, T., & Schaeck, K. (2014). Executive board composition and bank risk taking. Journal of Corporate Finance, 28, 48-65. 5. Bernerth, J. B., & Aguinis, H. (2016). A Critical Review and Best-Practice Recommendations for Control Variable Usage. Personnel Psychology, 69(1), 229-283. doi:10.1111/peps.12103 6. Bertrand, M., & Schoar, A. (2003). Managing with style: The effect of managers on firm policies. The Quarterly Journal of Economics, 118(4), 11691208. 7. Boyd, B. K., Adams, R., & Gove, S. (2017). Research methodology of governance studies: Challenges and opportunities. Corporate Governance: An International Review, 25(6), 382-383. 8. Chevalier, J., & Ellison, G. (1999). Are some mutual fund managers better than others? Cross‐sectional patterns in behavior and performance. The Journal of Finance, 54(3), 875-899. 9. Čihák, M., Demirgüç-Kunt, A., Pería, M. S. M., & Mohseni-Cheraghlou, A. (2012). Bank Regulation and Supervision around the World: A Crisis Update. 10. Clark, M. (2013). A comparison of correlation measures. Center for Social Research, University of Notre Dame, 4. 11. European Banking Authority and European Securities and Markets Authority. (2017). Joint ESMA and EBA Guidelines on the assessment of the suitability of members of the management body and key function holders under Directive 2013/36/EU and Directive 2014/65/EU. Retrieved from 12. Fernandes, C., Farinha, J., Martins, F. V., & Mateus, C. (2017). Supervisory boards, financial crisis and bank performance: do board characteristics matter? Journal of banking regulation, 18(4), 310-337. 13. Gande, A., & Kalpathy, S. (2017). CEO compensation and risk-taking at financial firms: Evidence from U.S. federal loan assistance. Journal of Corporate Finance, 47, 131-150. doi:10.1016/j.jcorpfin.2017.09.001 14. Garcia-Meca, E., Garcia-Sanchez, I.-M., & Martinez-Ferrero, J. (2015). Board diversity and its effects on bank performance: An international analysis. Journal of Banking & Finance, 53, 202-214. doi:10.1016/j.jbankfin.2014.12.002