THE INFLUENCE OF INTERNAL AND EXTERNAL CORPORATE GOVERNANCE ON FIRM VALUE: EVIDENCE FROM STOXX EUROPE 600 INDEX MEMBERS

How to cite this paper: Stender, P., & Rojahn, J. (2020). The influence of internal and external corporate governance on firm value: Evidence from STOXX Europe 600 index members. Corporate Ownership & Control, 18(1), 152-162. http://doi.org/10.22495/cocv18i1art12 Copyright © 2020 The Authors This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0). https://creativecommons.org/licenses/by/


INTRODUCTION
In recent years, compliance with corporate governance has become an important factor to shareholders in identifying potential investment opportunities and aligning their investment strategies (Tseng et al., 2019). However, prior research on the effects of corporate governance on firm value provides inconsistent findings. Gompers, Ishii, and Metrick (2003), Aggarwal and Williamson (2006), and Ngoc Phi Anh (2016) among others, reveal a positive impact of corporate governance on financial performance indicators like sales growth, profit, and firm valuation. Gupta, Kennedy, and Weaver (2009) as well as Hassouna, Ouda, and Hussainey (2017) find no influence on valuation, and Daines, Gow, and Larcker (2010) report mixed evidence for the influence on firm value depending on the specific governance rating they apply.
One reason for the contradicting findings may stem from the application of different corporate governance measures by the respective studies (Bhagat & Bolton, 2019). As Bebchuk, Cohen, and Ferrell (2009) demonstrate, the influence of governance on firm value depends on the respective set of governance provisions included in a governance score. Apparently, the diverse compositions of these scores measure different aspects of corporate governance quality, showing ambiguous influences on firm value (Daines et al., 2010). Even more importantly, using a single corporate governance index may cause biased estimations (Agrawal & Knoeber, 1996), especially if there are interactions between the scores (Donker & Zahir, 2008).
Early research on the impact of corporate governance on firm value is dominated by studies on the US market. These studies predominantly use scores that measure governance quality through the protection of shareholder rights, adopting an external approach to corporate governance. However, Aggarwal and Williamson (2006) and Schäuble (2019) underline the importance of internal governance mechanisms that go beyond a sole shareholder orientation. When examining the relation between corporate governance and firm value, stakeholder-oriented internal governance provisions should also be taken into consideration. Assuming convergence of external corporate governance across countries, for example, due to investor demand, best practice transfer, or international harmonization of disclosure requirements (Yoshikawa & Rasheed, 2009), internal corporate governance offers firms opportunities to differentiate themselves.
Thus, this paper seeks to contribute to the ongoing debate on the governance-firm value relationship in the following aspects. First, instead of using a single governance score, we apply a more holistic measurement approach. For this purpose, we reconstruct the most frequently cited corporate governance scores according to a Web of Science and Google Scholar review on a common database. To test whether the scores capture different aspects of corporate governance, we perform a principal component analysis (PCA). To our best knowledge and based on our literature review, such an approach to measure corporate governance quality has not yet been applied. The PCA reveals that the governance scores load on two general factors that represent internal and external governance quality. Second, most studies on the impact of corporate governance on firm value focus on the US market. Such findings may not be generalizable because corporate governance structures vary by nature across countries due to legal, cultural, and other factors. Instead, our sample is comprised of the largest non-financial European firms included in the STOXX® Europe 600. Particularly in continental Europe, corporate governance is more oriented toward stakeholder interests than in the US (Bottenberg, Tuschke, & Flickinger, 2017). Hence, capturing both internal and external corporate governance quality may provide additional insights. We do so by performing a fixed-effects panel data regression and an instrumental variable (IV) regression to account for endogeneity concerns.
The remainder of this paper is organized as follows. Section 2 reviews related literature on different approaches to measure corporate governance and address endogeneity problems. Section 3 describes the dataset and discusses the results of the PCA to construct our composite governance measures. Section 4 provides and discusses the empirical analysis results, including the IV regression, to account for endogeneity. Section 5 concludes.

Measuring corporate governance quality
Approaches to measuring governance quality are typically based on a set of individual governance provisions to construct a firm-specific rating. While most ratings apply pass-fail scoring methodologies, they substantially differ on the number, weighting, and choice of provisions (Louizi & Kammoun, 2016). There are two general categories of provisions: those that measure external and those that measure internal governance quality (Al-Najjar & Clark, 2017). Scores that predominantly include provisions related to shareholder rights and anti-takeover provisions are considered ratings of external governance (Schäuble, 2019). It is assumed that shareholders will pay a premium for firms that protect their rights through governance regulations (Aggarwal, Erel, Ferreira, & Matos, 2011). In theory, this should increase the valuation of firms with good external governance (Durnev & Kim, 2005).
More holistic measurement approaches contain additional provisions of internal governance, thereby emphasizing the importance of internal control (Brown & Caylor, 2006). In this context, provisions of internal governance relate to the characteristics of the board of directors, internal audit, director's compensation, or a firm's ownership structure (Wintoki, Linck, & Netter, 2012). Ratings that consist of internal governance provisions go beyond a sole orientation toward the shareholder by including provisions that address all stakeholders (Acharya, Myers, & Rajan, 2011). Zagorchev and Gao (2015) show that governance scores which comprise provisions of internal governance such as board structure, audit, compensation, and ownership structure are associated with higher Tobin's Q, which is commonly used in literature to represent firm or market value of a company.
To analyze the composition of governance scores with regard to internal and external provisions, we conduct a Web of Science and Google Scholar ranking to identify prevailing academic ratings for governance quality. 1 Our ranking results include the Gompers et al. (2003) G-Index (2,636 citations) underlining the score's widespread recognition by media and academia (Khanchel, 2007), the Bebchuk Gompers et al. (2003) to construct the Entrenchment Index that uses only those 6 out of the original 24 provisions from the same database that are specifically related to anti-takeover and protection provisions. Brown and Caylor (2006) use the Institutional Shareholder Services database to set up the GOV-Score and Parsimonious Index. Whereas the GOV-Score includes all 51 firm-specific provisions 3 from such a database, the Parsimonious Index only contains those 7 provisions that drive the positive impact between governance and firm value. To provide a complete picture of all scores discussed in this paper, we extend our analysis by the Eikon ESG Shareholder and Eikon ESG Management scores provided by Refinitiv Eikon 4 to represent established commercial ratings of corporate governance. Following the shareholder-and stakeholdercentric approach by Shahzad, Rutherford, and Sharfman (2016) we set up three categories for provisions of similar content, which are shareholder rights, board of directors, and audit and disclosure. Our analysis of the different scores is summarized in Table 1. Provisions included in the G-Index, the entrenchment index, and the Eikon ESG Shareholder score predominantly fall in the shareholder rights category. Consequently, these scores can be considered as shareholder-oriented ratings of external corporate governance.
Provisions used to compute the GOV-Score and Parsimonious Index contain internal and external provisions. In the GOV-Score, weighting of internal and external provisions is balanced, while external provisions prevail in the Parsimonious Index. The Eikon ESG Management score can be clearly identified as a stakeholder-oriented rating of internal corporate governance quality as it only comprises provisions from the board of directors and audit and disclosure categories. Our comparison between the selected governance scores indicates that approaches to measuring governance quality are based on different sets of underlying provisions among researchers. Consequently, it can be assumed that differences in computing scores to assess governance quality may contribute to the mixed research results (Bhagat & Bolton, 2019).

Endogeneity and the governance-firm value relation
More recent publications raise concerns about the governance-firm value relation being affected by endogeneity (Hassouna et al., 2017). Specifically, the results on the governance-firm value relation may be biased due to self-selection whenever non-random samples are analyzed (Iyengar & Zampelli, 2009). Hence, it is challenging to determine whether the effect of proxies for corporate governance on firm value is causal, or whether issuers choose the level of corporate governance that meets their respective goals or preferences (Wintoki et al., 2012). For instance, firms trading at discounts might strive to improve their governance quality to increase valuation and send out positive signals to investors (Renders, Gaeremynck, & Sercu, 2010). Moreover, for firms that operate in highly competitive industries, compliance with corporate governance structures may be an effective way to differentiate themselves or improve reputation (Bear, Rahman, & Post, 2010). However, when managers are not able to add value through core business activities, but only by choosing certain governance practices, a sustainable effect on firm valuation is questionable (Schultz, Tan, & Walsh, 2010).
When managers purposely select a governance structure to optimize market value, it requires an appropriate methodology to reveal the actual influence of good governance on valuation (Iyengar & Zampelli, 2009 In a fixed-effects panel regression, the "within" transformation wipes out any individual effects (Ebbes, Papies, & van Heerde, 2017). Fixed-effects panel models, therefore, produce consistent parameters in the case of unobserved heterogeneity but are not optimal to control for endogeneity when data is affected by a self-selection bias (Secchi, Tamagni, & Tomasi, 2016). An alternative to the fixed effects models is the use of external instrumental variables that are correlated with the endogenous regressor but uncorrelated with the error term (Renders et al., 2010). Such a technique has been employed by Drobetz, Schillhofer, and Zimmermann (2004)  Our initial sample consists of 3,600 annual observations from 600 firms. First, we exclude 134 firms with 804 annual observations, which are considered financials according to the ICB sector classification. As we control for debt to total assets, regression results are likely to be distorted when financials remain in the sample. Furthermore, financials are strictly regulated, which potentially affects both their performance and corporate governance regulations. In the second step, we remove 247 annual observations because of missing data. Among these, 166 observations lack data on corporate governance, while data on the controlling variables are incomplete for 81 observations. Finally, 109 outliers are dropped. An observation is defined as an outlier if the Cook's distance exceeds a value of 4/n, where n denotes the number of observations. This leaves us with an unbalanced panel dataset comprising 2,440 annual firm observations for 437 firms over a six-year period.

Recreation of common corporate governance scores
We recreate the corporate governance scores outlined in Sub-section 2.1 on a common data basis. Our source of governance data is the Refinitiv Eikon ESG database, which is easily accessible and provides a broad set of provisions covering internal and external governance characteristics. A set of 65 provisions relevant to reflect governance quality 5 Survivorship bias describes a situation where historical performance is overestimated as a dataset only considers the "survivors" at the end of a period while companies which have failed or do not longer exist are neglected.
of a firm is identified, 41 of which relate to internal governance, that is the Eikon Management category, while 24 cover external governance, i.e., the Eikon Shareholder category.
As Eikon provides this separation of provisions, we subdivide the GOV-Score into two separate scores, one based on the Eikon management and one based on the shareholder category. This leaves us with a total of seven governance scores.
Following the approach of Brown and Caylor (2006), we use the most recent version of the Institutional Shareholder Services governance methodology guide to determine whether a provision contributes positively or negatively to corporate governance quality (Institutional Shareholder Service, 2020). When a firm fulfills a provision that positively affects governance quality, it is coded as one, and zero otherwise. If a provision is stated as a number or percentage, it is deemed to be fulfilled if the threshold as proposed by the Institutional Shareholder Service (2020) methodology guide can be achieved without making any further gradations (Gompers et al., 2003). For instance, a board meeting attendance of 75% or more is proposed as a threshold for good governance. The recreation of the scores then follows a straightforward scoring model that sums the binary value of each provision contained in the score. Each score is scaled as a percentage showing the ratio of the sum of total provisions fulfilled to the sum of provisions included in the respective score. Compared to the composition of the scores in the original papers, the following modifications are necessary due to the use of the Eikon database:  GINDEX refers to the recreated G-Index and is based on a total number of 12 Eikon provisions that are identified as similar or overlapping with 17 out of the 24 original IRRC provisions used by Gompers et al. (2003).
 ENTRM denotes the recreated Entrenchment Index by Bebchuk et al. (2009), which we base on all provisions in the Eikon dataset related to anti-takeover or protection mechanisms. 6 Our Entrenchment Index includes a total set of 14 entrenching Eikon provisions related to limitations of shareholder voting rights, anti-takeover provisions, and classified or staggered board structure.
 GOV-SH represents the Brown and Caylor (2006) shareholder GOV-Score and comprises all 24 provisions from the Eikon shareholder category.
 GOV-MNG represents the management GOV-Score and is based on the 41 management provisions provided by the Eikon database.
 PARS refers to the Parsimonious Index for which we use a set of similar governance provisions as identified by Brown and Caylor (2006). The reconstructed score includes six of the original seven Institutional Shareholder Services provisions, as the relevant provision for "Option Burn Rate" is not available in the Refinitiv Eikon dataset.
The Eikon ESG Shareholder score (EIKON-SH), including provisions capturing external governance quality, and the Eikon ESG Management score (EIKON-MNG), including provisions capturing internal governance quality, are based on the original Eikon score without further modifications. Table 2 summarizes the descriptive statistics. In line with previous findings (Daines et al., 2010), correlation coefficients between our seven corporate governance scores reported in Table 3, however, indicate that different aspects of corporate governance quality are measured. While coefficients between scores mainly capturing either external or internal corporate governance tend to be rather high, correlations between external and internal governance scores are low and occasionally even negative.

Principal component analysis (PCA)
To identify common factors on which our governance measures load, we perform a PCA similar to Larcker et al. Prior to the PCA, we test pre-conditions through the Kaiser-Mayer-Olkin criterion with a value of 0.7097 being well above the critical value of 0.5, and the Bartlett test being statistically significant at the 95% confidence interval level (Maddala, 2008). We extract the number of relevant factors for our PCA using the Very Simple Structure (VSS) criterion. Based on a varimax rotation, the VSS criterion indicates a two-factor model that accounts for 67.89% of the total variance. Table 4 reveals that scores which mainly include external provisions related to shareholder rights show high loadings on factor 1, while ratings that predominantly consist of provisions related to internal governance load on factor 2. These results correspond to prior findings by Louizi and Kammoun (2016) and suggest that approaches to measuring governance quality need to differentiate between the aspects of internal and external governance. In addition to the reconstruction of the seven corporate governance indicators on a common data basis and the choice of the corporate governance measures, high loadings on the external governance factor may be driven by the influence of institutional investors. Especially institutional investors from countries with strong shareholder rights export corporate governance practices  Based on the PCA results, we construct two composite measures of corporate governance quality, one for external (EXT) and one for internal governance quality (INT). To compute EXT and INT, all governance scores are z-standardized. These normalized values are summed up and divided by the number of scores comprised, that is five for factor 1 (EXT) and two for factor 2 (INT). Summary statistics on EXT and INT are presented in Table 5.  , 2006). Q is defined as the market value of equity and total assets less common equity and deferred taxes divided by total assets. Panel A of Table 6 provides descriptive information on Q.
Our set of controlling variables is based on the insights of prior research on the impact of corporate governance on firm valuation. We use SIZE as total assets to represent firm size (Drobetz et al., 2004). To account for risk, we consider DEBT, i.e., the ratio of total debt to total assets, (Chhaochharia & Laeven, 2009) whereas the return on assets ROA, captured by gross income to total assets, is added as a measure of a firm's profitability (Bebchuk et al., 2009). To control for growth potential, we follow Aggarwal, Erel, Stulz, and Williamson (2010) and add CAPEX, i.e., the capital expenditures to total assets, as an indicator for investment intensity. Consistent with Flammer (2015), we include CASH, i.e., cash holdings as the ratio of cash and short-term investments to total assets and further account for corporate risk by including RESVOL, i.e., the residual volatility ) and information asymmetries denoted by INTANGIBLES, i.e., the ratio of intangible assets (Khanchel, 2007). For the regression analysis, we use the natural logarithm to transform the controlling variables SIZE and RESVOL.

Basic results
We test for the impact of EXT and INT on Q while controlling for the variables outlined in Sub-section 3.3. To account for unobservable time and individual effects, we use the "within" estimator with year and firm fixed effects, as indicated by the Hausman test. All variance inflation factors fall below the critical threshold of five so that results reported in Table 7 are not affected by multicollinearity between the independent variables. Significance tests of the regression coefficients rely on robust Huber-White standard errors to account for heteroscedasticity (Huber, 1967;White, 1980). Our regression model accounts for 27.9% of the total variance reported by the overall R-squared value. Regression results for the controlling variables mostly correspond to expectations and prior findings in the literature. The positive impact of the controlling variables on Q is reported for ROA and CASH, while DEBT, CAPEX, and INTANGIBLES are not significant. For the remaining controlling variables, we report a significant negative impact of ln(SIZE) and ln(RESVOL). The negative effect of ln(SIZE) on Q is surprising as prior studies by Rizqia and Sumiati (2013) reveal a positive relationship between firm size and Q due to eased access to external funding, lower transaction cost, and more spotlight from investors compared to smaller firms. However, our results are in line with prior studies by Brown and Caylor (2006), Chen, Chung, Hsu, and Wu (2010) or Jayachandran, Kalaignanam, and Eilert (2013) who also find a negative impact of firm size on Q. Our findings on a negative ln(SIZE)-Q relationship are probably driven by the sample characteristics. Among the largest companies measured by total assets, industry sectors such as industrials, basic materials, or energy account for a large fraction, while the technology sector, where higher valuation levels can frequently be observed, is underrepresented. The negative influence of ln(RESVOL) on valuation is as expected, as the residual volatility can be used to represent information asymmetries (Krishnaswami, Spindt, & Subramaniam, 1999).
Regarding our independent variables, we reveal a positive influence of EXT on Q, while Q non-significantly decreases with increasing INT. The market value of public companies is determined by investors' demand for equity. Hence, we assume that external governance plays a more important role for shareholders than internal governance due to its direct impact on shareholder rights. Moreover, several studies disclose a negative influence of certain internal governance provisions on valuation.
With regard to board size, Yermack (1996) argues that smaller boards of directors can work more effectively and at a lower cost. Governance regulations prescribing a mandatory number of board members may lead to an inverse relationship between internal governance and valuation. Durden and Pech (2006) claim that internal governance regulations have negative consequences on management performance as they potentially hinder managers creating value for the business. Excessive internal governance obligations might distract and pre-occupy management resources. When managers cannot effectively operate the business due to internal governance obligations, firm value may be negatively influenced (Durden & Pech, 2006).

IV regression
To mitigate endogeneity concerns caused by a self-selection bias, we use an IV regression model (Chung & Zhang, 2011). We follow Durnev and Kim (2005) and use lagged values of the endogenous independent variables EXT and INT as instruments, since the natural instruments are either already considered as control variables, such as firm size (Black et al., 2006), or by the sample selection process, such as index membership (Drobetz et al., 2004). The instruments are denoted as EXT t-1 and INT t-1 .
We apply an IV regression including year and firm fixed effects. To verify the consistency of the model, we first test the relations between the instrumental variables and the endogenous regressors. Table 8 reports the first-stage of the IV regression results showing that EXT is correlated with EXT t-1 , and INT is correlated with INT t-1 , which fulfills the relevance criterion (Ebbes et al., 2017). First-stage F-statistics of excluded instruments are well above the threshold of ten according to Stock and Yogo (2005). The results of the second stage of the IV regression are revealed in Table 9. Applying the Cragg-Donald and Kleibergen-Paap Wald F-statistics, our set of instruments is identified as valid and strong instruments for EXT and INT with F-statistics well above the critical value according to the Stock-Yogo weak ID test (Bhagat & Bolton, 2019). As the endogeneity test shows that the EXT and INT cannot be treated as exogenous, endogeneity is of concern, so the IV regression results from Table 9 are more consistent than the regression results from our fixed effects regression in Sub-section 4.1. Signs of the estimates and significance of the controlling variables from our second stage IV regression are comparable to those of the fixed effects regression. However, by including the instrumental variables, the influence of EXT on Q remains significant and positive, whereas the relation between INT and Q is significantly negative. As outlined in the previous section, the negative influence of INT on Q can be explained by costs related to good internal governance structures such as board committees, board size, or frequency of board meetings as well as by distractions of the management from core business activities (Durden & Pech, 2006). The overall cost to implement good internal governance might negatively affect the market value of a company, while shareholder-oriented external governance creates value for shareholders and thereby has a positive effect on firm value. These results correspond with Stiglbauer and Velte (2012).

CONCLUSION
Using a sample of non-financial firms included in the STOXX® Europe 600 index over a period from 2012 through 2017, this study investigates the influence of internal and external corporate governance on firm valuation. The level of corporate governance quality is measured by recreating the most frequently cited corporate governance scores from the literature on a common database while following the originally applied methodologies. Using a PCA, we find that our set of seven governance scores loads on two different factors. We identify such factors to represent internal and external governance quality and construct two composite governance measures for each of the factors.
We apply fixed effects panel regression techniques to gain insights into the governance-firm value relation. Accounting for endogeneity, we further use a two-stage IV regression model including lagged values of the independent variables as instruments. Controlling for generally accepted variables derived from literature, we find that external governance quality is positively linked to firm valuation, whereas internal governance shows a significant negative impact when endogeneity is addressed. The positive influence of external governance quality can be explained by its direct impact on shareholder rights. As firm valuation reflects the investor's demand in the firm, the protection of shareholder rights through good external governance may be perceived as a positive signal to investors. While it is plausible that internal governance plays an important role in good corporate management, potential costs from the creation of board committees, large boards of directors, frequent board meetings, and the use of management resources may negatively affect valuation. Our findings allow concluding that one possible reason for mixed evidence on the governance-firm value relation in the literature might stem from the inverse influence of internal and external governance.
Overall, the study suggests that both internal and external corporate governance play an essential role in the responsible management of a firm. Compliance with governance principles nowadays is a mandatory requirement for firms as governance criteria are increasingly included in capital investment decisions and became an important input to define investment strategies. Our findings indicate that differences in the influence of internal and external factors of corporate governance on the governance-firm value relation should be considered. Moreover, the findings emphasize the requisite of addressing endogeneity concerns when analyzing the influence of corporate governance on firm value. The limitations of this paper mainly relate to the data sample used. Due to the availability of governance data at company level, our study focusses on large companies with high market capitalization, while data for small and medium-sized companies could not be obtained. Broader studies in different countries and markets could therefore produce more generalizable results. It could also be of interest for the academic community to gain more insights on why internal governance negatively influences on firm value and if there are differences between industry sectors.