Board governance and dominant shareholders: substitutes or complements? Sridhar Arcot and Valentina Brunoy Current version: February 2013

JEL Classi…cation: G32, G34, G38, K22

We appreciate comments on an earlier version from Jan Bena, Stijn Claessens, Phil English, Mara Faccio, Antoine Faure-Grimaud, Jose-Miguel Gaspar, Yaniv Grinstein, Denis Gromb, Sudip Gupta, Robert Hauswald, Henry Hu, N.R. Prabhala, Michel Robe, Michela Verardo, and participants at various seminars and conferences (World Bank, Bank of England, London School of Economics, SEC, American University, ESSEC, Universita’Cattolica, Bocconi University, the International Conference on Corporate Governance in Birmingham, CAF Summer Research Conference, Annual Conference on Empirical Legal Studies at Austin). We thank Dragana Cvijanovic, Terence Teo and Zhigang Qiu for excellent research assistance. Sridhar Arcot acknowledges support from the FMG and LSE. Valentina Bruno acknowledges …nancial support from Ente Luigi Einaudi. All errors are of course our own. y Sridhar Arcot: ESSEC Business School, Department of Finance, 95021 Cergy Pontoise Cedex, France, Tel: +33 (0) 1 34 43 30 77, Email: [email protected]. Valentina Bruno: American University, Kogod School of Business, Department of Finance, Washington DC, USA, Tel: +1 202 885 1899, Email: [email protected]

Board governance and dominant shareholders: substitutes or complements?

Abstract We study the governance dynamics surrounding …rms with dominant shareholders in a voluntary regulatory arena where we can directly observe the impact of …rm ownership on corporate governance practices pertaining to the composition of the board of directors. We …nd that …rms with a dominant shareholder are more likely to deviate from standards of best practice in corporate governance. However, lesser governance standards in …rms where a dominant shareholder is the monitor in-place are not associated with lower performance. In contrast, governance practices and disclosures matter in widely-held …rms where they alleviate the con‡icts between managers and dispersed shareholders. Overall, our results suggest that controlling shareholders and board practices are substitute governance mechanisms.

1

Introduction

Studies have shown that large shareholders are common in many countries, both developed (Faccio and Lang, 2002) and developing (Claessens et al., 2002), and that a …rm’s ownership structure has an important but ambiguous governance role (Shleifer and Vishny, 1997). However, the interdependence between ownership and other governance mechanisms is relatively unexplored in the literature (Judge, 2011). At the same time, the academic literature on the relationship between board structure and …rm performance has shown mixed or yet inconclusive evidence (Wintoki, Link and Netter, 2012). The purpose of this paper is to examine the relationship between …rm ownership structure and board corporate governance practices and its impact on …rm performance. In particular, we ask whether the presence of a dominant shareholder acts as an alternative governance mechanism to board governance practices. If so, do board governance practices matter more in widely-held …rms than in …rms with a large monitor in-place? Inside a corporation, two potential con‡icts coexist: the …rst is among shareholders and the second is between shareholders and managers. In …rms with a dominant individual shareholder, the potential con‡ict between the controlling shareholder and minority shareholders is more severe than the potential con‡ict between managers and shareholders because the dominant shareholder typically controls the board of directors. The large shareholder has the incentive to collect information and the power to monitor managers. He thus substitutes for the monitoring role of the board, alleviates the con‡icts between owners and managers (Shleifer and Vishny, 1986; Bolton and Von Thadden, 1998), and acts as an alternative governance mechanism (Aghion and Tirole, 1997).1 If the individual large shareholder can better substitute for the monitoring and advising role of the board, then standard corporate governance practices, like independent boards, board committees or higher quality disclosures, would matter less for performance in …rms where a dominant shareholder is present (the monitoring 1

Because large shareholders have the incentive to collect information and the power to monitor the manager, the large shareholders substitute for the monitoring role of the board, choose their preferred project and formal authority prevails. In contrast, shareholders who are not informed lose control and the manager has real authority (e¤ective control) (Aghion and Tirole, 1997).

hypothesis). The dominant shareholder, however, may also have incentives to expropriate minority shareholders and/or entrench himself in managerial positions (Shleifer and Vishny, 1997). If the dominant shareholder extracts …rm resources at the minority shareholders’expense, then he will alter corporate governance practices and transparency to entrench himself and achieve greater ease in extracting private bene…ts (the entrenchment hypothesis). If so, strong board practices would matter for performance in …rms where a dominant shareholder is present. Both the monitoring and the entrenchment hypotheses imply that …rms with dominant shareholders will adopt lesser corporate governance practices and disclosures than widely-held …rms. We di¤erentiate between these two competing hypothesis by estimating the relationship between corporate governance practices and performance. In widely-held …rms, the potential con‡ict arising from the separation of ownership and management is more severe than the con‡ict among shareholders. Because of the free-rider problem, dispersed shareholders have fewer incentives to monitor the management, who in turn can pursue self-serving goals that do not maximize shareholder value. At such …rms, therefore, better and more transparent corporate governance practices should align managers’incentives with …rm value maximization and should be associated with better performance. We test these theories in an institutional environment where companies are not mandated to adopt prescribed governance practices. We choose the United Kingdom as the setting of our study because the UK pioneered a corporate governance approach characterized by voluntary compliance with governance standards coupled with mandatory disclosure. In such a setting, companies can deviate from governance standards listed in a Code of Best Practice as long as they explain the reasons for deviating as part of compliance with the Listing Requirements (the “Comply or Explain” approach). The UK corporate governance approach enables the board to retain ‡exibility in the way in which it organizes itself and exercises its responsibilities. Based on the information contained in the corporate governance section of UK annual reports, we construct a unique dataset by handcollecting details of both compliance with the UK corporate governance Code and explanations

2

given for noncompliance by FTSE350 non-…nancial companies between 1998 and 2004. We choose this period because the Code of Best Practice was in e¤ect continuously and with no amendments. The non-mandatory governance setting considered in this paper provides an extraordinary laboratory to test how di¤erent ownership structures are associated with di¤erent governance choices and to assess their impact on performance. Such a test would pose serious challenges if performed in a mandatory regulatory setting. In the United States, for example, the 2002 Sarbanes-Oxley Law and NYSE listing standards require that a majority of the board of directors be independent and that board committees (audit, monitoring, and compensation) be comprised entirely of independent directors. By contrast, UK …rms do not have to conform to such requirements, provided they explain their reasons for doing so. By looking at both the "comply" and the "explain" aspects of corporate governance, we …ne-tune our identi…cation strategy and detect whether companies deviate from governance standards due to di¤erent monitoring needs or self-serving goals. Our analysis shows that ownership structure a¤ects the decision to comply with corporate governance provisions. In particular, we …nd that …rms with a dominant individual shareholder are less likely to comply with corporate governance standards and explain their governance choices. Such companies do not publicly communicate the reasons why commonly considered good practices in corporate governance are not optimal for the company. Thus, the dominant shareholder behaves more like an insider and discloses as little as possible to other investors. However, lesser governance standards and lower disclosure in …rms with an individual large shareholder are not associated with lower performance. The dominant shareholder endogenously chooses the company’s optimal governance structure, which may not conform to standard governance practices recommended or prescribed by law. We argue that standard governance practices, that mostly empower the board of directors with monitoring roles, are less relevant when a large shareholders is the monitor in-place. This result shows that the monitoring hypothesis prevails over the entrenchment hypothesis. Because the individual shareholder’s wealth is linked to the company’s welfare, the large shareholder has strong incentive to act in the company’s best interest rather than expropriate minority shareholders.

3

In contrast, governance practices and disclosures do matter for performance in widely-held …rms. Speci…cally, we …nd that better governance practices and higher governance transparency are associated with better performance. This result suggest that in companies where the freerider problem is more serious better corporate governance standards and disclosures substitute the monitoring role of dispersed shareholders, limit the self-serving goals of managers and incentivize managers to e¢ ciently use company resources. We supplement these …ndings by examining the relationship of board practices with two alternative indicators of agency problems: earnings quality and acquisitions activity. The accounting literature has found that earnings quality is worse when agency-problems are worse. In the corporate governance literature, acquisitions activity is a widely used proxy for agency costs of managerial discretion. We …nd that widely-held …rms are associated with higher earnings quality and lower acquisitions activity only if they have stricter governance practices and better disclosures in place. In contrast, we do not …nd signi…cant di¤erences in …rms with a dominant shareholder. This evidence thus complements the evidence on performance that stricter corporate governance standards and better governance disclosures are more relevant in widely-held companies. Taken together, the evidence above suggests that corporate governance practices and disclosures bear a di¤erent impact depending on the ownership structure of the company. Speci…cally, better corporate governance standards and disclosures are more e¤ective in solving the con‡ict between shareholders and management in companies with dispersed shareholding. This paper contributes to the literature in at least three ways. First, we focus on the unique governance dynamics surrounding …rms with dominant shareholders and on the role of governance for dispersed shareholders. Second, we perform this combined analysis of corporate governance practices and ownership structures in a non-mandatory setting where companies can choose di¤erent governance structures. Such a setting allows us to observe the corporate governance practices and disclosures endogenously chosen by …rms. Third, our …ndings hold important policy implications in light of the widespread adoption of corporate governance Codes modeled on the UK system. This corporate governance system assigns a monitoring role to shareholders, but monitoring incentives

4

vary across shareholders. This study bridges three strands of the literature. First, the corporate governance literature highlights the costs and bene…ts of insider and large shareholder monitoring (e.g., Shleifer and Vishny, 1986; Aghion and Tirole, 1997; Bolton and Von Thadden, 1998; Anderson and Reeb, 2003). In the UK setting, Fidrmuc, Goergen, and Renneboog (2006) and Goergen, Renneboog, and Zhang (2008) …nd that individual controlling shareholders mitigate problems caused by asymmetric information and agency. Second, the literature on disclosure (e.g., Ali, Chen, Radhakrishnan, 2007; Anderson, Duru, and Reeb, 2009) highlights the association between di¤ering disclosure decisions and ownership characteristics. Third, the literature on corporate governance codes of best practice (e.g., Conyon and Peck (1998); Dedman (2003), Dahya and McConnell (2007)) focuses on the aspect of compliance with some governance standards, whereas we also take into account the possibility that a company can choose not to adopt governance standards if that is more appropriate to its circumstances. Our results bring together, complement and extend these strands of the literature, and show that the impact of corporate governance practices and disclosures is the result of complex governance interdependences. In particular, our results suggest that one corporate governance mechanism (controlling shareholders) acts as a substitute for another (related to boards). The structure of the paper is as follows. Section 2 describes the sample selection, the corporate governance indicators, ownership and …nancial variables. Section 3, the multivariate analysis, Section 4 the robustness tests and Section 5 summarizes and concludes.

2 2.1

Sample and Data The "Comply or Explain" approach

In 1992 a committee chaired by Sir Adrian Cadbury issued a series of recommendations, known as the Cadbury Report, related to the relationship between the chairman and chief executive, the role of non- executive directors and internal control reporting. A requirement was added to

5

the Listing Rules of the London Stock Exchange that companies should report whether they had followed the recommendations or, if not, explain why they had not done so (this is known as "Comply or Explain"). In essence, the "Comply or Explain" system identi…es good governance standards relating to, for example, the board and its committees and risk management and internal control, but companies can choose to adopt a di¤erent approach if that is more appropriate to their circumstances. Where they do so, however, they are required to explain the reasons for their deviations from the corporate governance standards. It is the role of shareholders to scrutinize and monitor the corporate governance practices adopted by the company. The role of shareholders is therefore crucial for the system to work. In 1998 two reports advocating best practices in corporate governance were brought together in a single code (known as the Combined Code).2 We download annual reports of companies belonging to the FTSE 350 index for the period of 1998

2004 from Mergent Online. We choose this period because the Combined Code was in

force continuously and with no amendments for the entire duration, after which it was subsequently updated by the Higgs Committee recommendations. Each annual report contains a corporate governance section that states the level of compliance with all the Combined Code provisions and an explanation in the case of noncompliance with any of its provisions. Provisions listed in the Combined Code are related to board governance practices such as a separate Chairman and CEO, a balance of executive and independent non-executive directors, strong and independent audit, nomination and remuneration committees (see Appendix).3 Since the Combined Code was in e¤ect from December 31st, 1998 to June 30th, 2004, we exclude annual reports of companies with …nancial year endings before and after those dates, respectively. As is common in the literature, we exclude 106 …nancial companies (SIC codes 60

69) (banks,

insurance, REITs, equity investment trusts, etc.) because the regulatory environment for those 2

See FRC "The UK approach to corporate governance" http://www.frc.org.uk The Combined Code consists of eleven provisions, of which we analyze eight. We did not include in our analysis the provisions related to the directors’re-election, pay linked to performance, and the quality of the internal control systems, because all companies in the sample complied or stated their intention to comply with these provisions. The inclusion of these provisions would not have a¤ected our results. Furthermore, judging the e¤ective level of compliance of the provisions pertaining to pay-linked to performance and internal controls would have required additional information which was not available to us. 3

6

companies di¤ers signi…cantly from that of non-…nancial companies. Some annual reports are not listed in Mergent Online. We are therefore left with 1275 total company-year annual reports.

2.2

Corporate Governance Indicators

We look at a company’s overall level of compliance with all the provisions of the Code, as it is stated in the annual reports, and we construct the following dummy variable: Comply equal to 1 if a company is fully compliant with all the Combined Code provisions, 0 otherwise. Because the "Comply or Explain" system allows companies to depart from corporate governance provisions provided an explanation is given, the assessment of the level of compliance with certain provisions does not alone measure the quality of an individual company’s corporate governance. When analyzing the companies’ annual reports, we …nd instances of non-compliance statements where there is no explanation of the reasons why a company decided to opt out of certain governance standards. Such companies not only fail to follow corporate governance standards, they also do not provide any justi…cation to their shareholders of why the adoption of a certain corporate governance provision is not optimal to them. The absence of any explanation is an objective fact, and it can be easily detected in the corporate governance section of the annual report.4 Lack of disclosure is generally an indicator of bad governance. For instance, Leuz, Nanda, and Wysocki (2003) argue that managers and controlling owners manage the level and variability of reported earnings in order to mask true …rm performance and to conceal their private control bene…ts from outsiders. Lang, Lins, and Miller (2004) …nd that …rms with governance problems are likely to be less forthcoming in terms of disclosure, giving analysts less information to work with in assessing investment potential. Kothari, Shu, and Wysocki (2009) show that managers tend to withhold information if it is bad news. 4

All companies incorporated in the UK and listed on the Main Market of the London Stock Exchange are required under the Listing Rules to report on how they have applied the Combined Code in the annual report. However, the Code does not propose any explicit penalty for not providing explanations.

7

We thus look at the company’s noncompliances that are not explained. Of the 853 company-year noncompliances observations, we …nd 197 instances when a company omit any form of explanation for at least one non compliant provision. We construct the following dummy variable: Comply or Explain equal to 1 if a company complies with all the governance standards and/or explain each noncompliance, and 0 otherwise. The dummy equal to 1 identi…es those companies that have adopted either high corporate governance standards or high levels of governance disclosures or both. The dummy equal to 0 identi…es those companies that do not comply with governance standards and do not disclose information to shareholders about at least one of their governance decisions. Of the 197 instances where there is at least one noncompliant provision that is not explained, we have 157 cases of exactly one unexplained provision, 22 cases where the company does not provide an explanation for two provisions, 11 cases of three provisions that are not explained and 7 cases of four provisions that are not explained. We therefore construct the following indicator: # No Explanations, that counts the number of noncompliances that are not explained.

2.3

Financial Data

We use return on assets (ROA) as our measure of performance since we are interested in investigating how heterogenous corporate governance choices re‡ect on the e¢ cient use of the company assets. We also focus our analysis on operating performance rather than stock market performance because we do not know the exact day corporate governance announcements are made public; also, public information about governance may not be impounded in stock prices in a timely manner (see Core et al., 2006), whereas weak corporate governance may be associated with an ine¢ cient use of company’s resources that leads to poor operating income. As discussed in Barber and Lyon (1996) and Core et al. (2006), return on assets is a preferred measure of operating performance because it is not a¤ected by leverage, extraordinary items, and other discretionary items; it also has more desirable distributional properties than return on equity. 8

Furthermore, as pointed out by Wintoki, Link and Netter (2012, page 591) "Many studies that examine the governance/performance relation use Tobin’s Q as a measure of …rm performance. This can be a problem for a number of reasons. Tobin’s Q (usually de…ned as the market-to-book ratio) is a proxy for growth opportunities, and there is strong theoretical reason to expect that growth opportunities are a cause, rather than a consequence, of governance structures." However, for robustness we also use the market to book ratio as alternative performance measure. Following Gompers et al. (2003), we de…ne the return on assets as the ratio of earnings before interests and taxes (EBIT) to total assets. We then compute the industry-adjusted measure by subtracting the return on assets of each company in each year with the median return on assets of its respective Fama-French industry group. Accounting information comes from Worldscope, whereas monthly stock market data is from Datastream. We also use in the analysis the following variables that have been used in similar studies (e.g., Klapper and Love, 2004; Dahya and McConnell, 2007; Boone et al., 2007) investigating the determinants of corporate governance choices and their performance impact: the logarithm of the book value of assets (Size) or of the market value of assets (Market Capitalization), the logarithm of …rm age (Age), the logarithm of the ratio of market to book value of equity (Growth Opportunities), the ratio of long term debt to total assets (Leverage), the ratio of EBIT to sales (Pro…tability), and the ratio of property, plant and equipment (PPE) to sales (Capital Intensity). Due to the presence of signi…cant outliers, we winsorize all of the control variables, with the exception of Age, at the 1% level.

2.4

Ownership Data

Ownership data is downloaded from Thomson One Banker (Ownership Module). Following Faccio and Lang (2002) and Dahya, Dimitrov, and McConnell (2008); companies are divided between individually-owned and widely-held based on the dominant shareholder at the 10% threshold. A …rm has a dominant individual shareholder if an individual, family, or privately held …rm owned by an individual or family, controls at least 10% of the voting rights in the …rm. We checked 9

for the presence of dual-class shares and pyramid structures. We construct a dummy Dominant Shareholder that takes the value 1 if a company has an individually-owned dominant shareholder and 0 if a …rm is widely-held. Since 10% of voting rights is frequently su¢ cient to exert control, this cut-o¤ is used extensively (e.g., La Porta et al., 1999, Dahya et al., 2008). If ownership data is not available for a particular company in Thomson One Banker, we hand-collect information from Hemscott’s Corporate Register.

2.5

Summary Statistics

Panel A of Table 1 provides summary statistics of corporate governance characteristics. Of the total 1261 company-year observations for which we could …nd …nancial data, we have 408 cases (32%) of companies complying with all the provisions of the Code. On average, 84:4% of …rms either comply or explain. Compliance with all the provisions of the Combined Code increases steadily from less than 10% in 1998 to almost 60% in 2004. Conversely, there are 197 instances (15:6%) where companies do not provide any explanation for at least one of their noncompliances. The average number of No Explanations is 0:208, computed as the number of noncompliances that are not explained divided by the total number of observations. The average number of noncompliances is 1:81 per non-compliant company; the median number of noncompliances is 1 (not reported). For our sample of 1261 company-year observations with corporate governance and …nancial information we could …nd 1135 with company-year ownership information. We have 200 companyyear observations related to the presence of a dominant shareholder at the 10% threshold (18% of the sample) which is consistent with the summary statistics reported in Faccio and Lang (2002) relative to the sample of companies (FTSE 350 non …nancial companies) we consider in our analysis. Panel B of Table 1 shows some …nancial characteristics of our sample. Because the companies belong to the FTSE350 index they are, on average, pro…table in terms of ROA (4:9%), large (assets of GBP 3002 millions), relatively old (40 years old) and not highly levered (20%). Panel C of Table 1 presents di¤erence of means tests between …rms with a dominant shareholder and widely-held …rms. Widely-held …rms comply and explain more than …rms with a dominant 10

shareholder. Firms with a large shareholder are smaller in size, with a lower market capitalization, sales and leverage. They also tend to be younger, though the di¤erence is not statistically signi…cant. There is no statistically signi…cant di¤erence in ROA, Market to Book ratio or Capital Intensity between the two groups of …rms.

3

Multivariate Analysis

3.1

Corporate Governance and Dominant Shareholders

We …rst focus our attention to the decision of comply with governance standards. We investigate whether the probability to comply with governance standards is a function of the ownership structure. We run a logit estimation and we regress the dummy Comply on the Dominant Shareholder variable and the control variables Market Capitalization, Age, Leverage, Pro…tability, Capital Intensity. Regressions are run with industry and time dummies and standard errors are robust and clustered at the …rm-level. The results presented in Table 2, column 1, show that companies with a dominant shareholder comply less with corporate governance standards than widely-held companies. This result shows that companies with a dominant shareholder rely less on board monitoring and advising. This re‡ects the large shareholder’s discretion in the management of the company. In contrast, companies with dispersed shareholders alleviate the free-rider problem by adopting higher corporate governance standards. We next turn our attention to the decision to comply or explain. We run a logit estimation and we regress the dummy Comply or Explain on the Dominant Shareholder variable, a set of company-level variables as de…ned before, with industry and time dummies and robust standards error clustered at the …rm level. Column 2 of Table 2 shows the logit results. Companies with a dominant individual shareholder are less likely to comply with governance standards and to omit explanations for their noncompliances. This evidence is consistent with the existent literature showing that concentrated family ownership exhibits less governance disclosure because large shareholders

11

behave more like insiders and have fewer incentives to disclose public information (Anderson, Duru, and Reeb, 2009; Chen, Chen, and Cheng, 2008). In fact, because large shareholders are actively involved in management, the information asymmetry between owners and managers is lower and the dominant shareholder thus prefers less public disclosure.

3.2

Corporate Governance and ROA

The results before show that …rms with a dominant large shareholder tend to adopt fewer governance standards and are less transparent about their governance choices. This evidence is consistent with both the monitoring hypothesis (the large shareholder is in the position to better monitor and discipline managers) and the entrenchment hypothesis (the large shareholder can extract private bene…ts of control). The governance-performance regressions will tell which of the two hypotheses prevails in …rms with an individual large shareholder. In widely-held …rms, because of the free-rider problem, higher corporate governance standards should better substitute the monitoring role of the shareholders and limit self-serving goals by managers. Hence, in widely-held …rms better corporate governance practices and disclosures should align managers’incentives to …rm value maximization and should be associated with better performance. 3.2.1

Base case regression

We run the following OLS regression:

ROAi;t+1 =

+

CGIndicatori;t +

Controlsi;t + "i;t

(1)

where ROA is the next year industry-adjusted ROA and CGIndicator is either the dummy Comply or the dummy Comply or Explain or the index #No Explanations. We use future performance to reduce endogeneity issues. The regression is run with time dummies and robust standard errors clustered at the …rm level. We cannot include company dummies because the various governance indicators are either invariant or change slowly from year-to-year.

12

Table 3, column 1, shows that the coe¢ cient of the dummy Comply is not signi…cant. The decision to comply or not comply with governance standards does not have a signi…cant performance impact. This could be the results of endogenous optimal choices by companies or simply evidence that there is no signi…cant association between compliance with governance standards and performance. The UK corporate governance system allows companies not to comply with the governance standards provided they give an explanation. We then separate the group of companies that do not adopt the governance standards and are not transparent about their governance choices from those companies that follow what prescribed by the corporate governance system and we regress ROA on the dummy Comply or Explain. If a shareholder of a manager does not adopt corporate governance standards for self-serving reasons, then we should observe worse performance. Companies omitting explanations for their noncompliances are more likely to fall in this category because lack of disclosure may indicate bad governance. Column 2 shows that the coe¢ cient of the dummy Comply or Explain is positive and has a p-value of 0:12. When using the index #No Explanations as the corporate governance indicator, we …nd that a higher number of noncompliances that are not explained is signi…cantly associated with worse performance in the overall sample (p-value= 7.1%, column 3) and in the subsample of noncompliant companies only (p-value=6.9%, column 4). These results show that noncompliance with governance standards does not necessarily mean bad governance, provided companies are transparent in their governance disclosures. In contrast, noncompliant companies with poor governance disclosure appear to be associated with worse performance. This evidence thus suggests that omitting corporate governance information is associated with agency problems and private bene…ts extraction. Overall, these results are consistent with the literature showing that corporate governance practices and better governance disclosure contribute directly to performance by disciplining the management, promoting better use of the assets in place, enabling better project selection, and reducing expropriation of investors’wealth (Shleifer and Vishny, 1997; Bushman and Smith, 2003; Klapper and Love, 2004; Durnev and Kim, 2006;

13

Hope and Thomas, 2008). 3.2.2

Dominant Shareholders

We now investigate whether corporate governance has a di¤erent impact on widely-held …rms versus …rms with a dominant shareholder. In untabulated regressions we add the dummy Dominant Shareholder to speci…cation 1 and we do not …nd any signi…cant association with performance. We then divide the sample of companies based on their ownership (…rms with a dominant shareholder and widely-held …rms) and run speci…cation (1) for each subsample of companies. Table 4, column 1 shows that the dummy Comply or Explain is not signi…cant in the subsample of …rms with a dominant shareholder. This suggests that higher corporate governance standards and governance disclosures are not signi…cantly associated with better performance in the presence of a dominant shareholder. In contrast, the dummy Comply or Explain is positive and signi…cant in the subsample of widely-held …rms (column 3). This result suggests that better corporate governance standards and disclosures are important in widely-held companies because noncompliant widelyheld companies with lack of governance disclosure have worse performance. Columns 2 and 4 con…rm that worse governance disclosure proxied by a higher number of non disclosures (# No Disclosures) is associated with lower performance in companies with dispersed shareholding only. The above results show the existence of a di¤erent impact of compliance and disclosure across ownership. Departure from standard governance practices and lack of governance disclosures are not associated with worse performance in …rms with a dominant shareholder. This evidence suggests that the large shareholder substitutes the monitoring role of the board and endogenously chooses the optimal company governance structure, which may not conform to the standard governance practices recommended or prescribed by law. This result shows that the monitoring hypothesis prevails on the expropriation hypothesis. Because the shareholder’s wealth is linked to the company welfare, the dominant shareholder has strong incentives to monitor managers and minimize the freerider problem rather than expropriate minority shareholders. This result is consistent with the evidence found in Fidrmuc, Goergen, and Renneboog (2006)

14

and Goergen, Renneboog, and Zhang (2008). They …nd that dominant individual shareholders reduce the market reaction to both directors’purchases and sales. This suggests that these types of shareholders engage in monitoring and thereby decrease the informational value of directors’ dealings. They conclude that these blockholders engage in monitoring of the …rms and ensure that the management focuses on value maximization. High legal protection of minority shareholders in the UK is an important condition underlying the prevalence of the monitoring hypothesis over the expropriation hypothesis in …rms with a large shareholder and may provide a competitive advantage to the …rm (Burkart, Panunzi, and Shleifer, 2002). Our results are consistent with the evidence found by La Porta et al. (2000), who …nd that the ability of the controlling shareholder to extract resources from the …rm depends on the protection a¤orded to outside investors. In a cross-country study, Dahya, Dimitrov, and McConnell (2007) …nd that the relationship between the percentage of independent directors and performance in …rms with dominant shareholders is stronger in countries with weaker legal protection. They do not …nd any signi…cant association between independent boards and performance in …rms with dominant shareholders in countries with high legal protection. Ellul, Guntay, and Lel (2009) …nd that family …rms originating from countries with high investor protection have lower cost of debt …nancing than non-family …rms. In contrast, noncompliant widely-held companies with a lack of governance transparency have worse performance. Because of the free-rider problem, better corporate governance practices and disclosures align managers’ incentives to …rm value maximization and are associated with better performance. Taken together, the evidence above suggests that, in the presence of high legal protection, as in the UK, corporate governance practices and disclosures have a di¤erent impact depending on the ownership structure of the company. Speci…cally, they matter more for widely-held companies than for …rms with an individual large shareholder. In untabulated regressions, we divide the companies that do not comply and do not explain based on the presence of a large dominant shareholder and construct dummy variables equal to 1 (and 0 otherwise) for each of the above categories. We then run speci…cation 1 with the above-mentioned

15

two dummies for the entire sample of companies (1135 observations). We …nd that lack of disclosure is signi…cantly associated (at the 10% level) with worse performance in companies with dispersed shareholding, but not in …rms with a dominant shareholder. These results estimated for the entire sample of …rms con…rm the results reported in Table 4 that lack of transparency in governance practices in widely-held …rms hides self-serving goals by managers, but it is not associated with entrenchment activities by the large shareholder. For the ease of exposition we leave these results untabulated.

4

Robustness Tests

4.1

Endogeneity

Endogeneity concerns are endemic in the corporate governance literature. If companies endogenously choose their governance structure and disclosure to maximize pro…tability, we should not observe any relationship with performance (Hermalin and Weisbach, 2003). By contrast, if managers or the controlling shareholder have opaque governance disclosures and adopt corporate governance practices di¤erent from the standards for reasons other than pro…t maximization, they should be less pro…table. The UK system allows us to tackle this aspect of the endogeneity problem because we can detect whether a company provides an explanation or not when departing from governance standards. Because lack of disclosure may be an indicator of bad governance, we can thus di¤erentiate between companies that depart from governance standards for value-maximizing reasons or self-serving reasons. Other possible concerns could be related to unobserved factors that are driving both governance and performance. Fixed-e¤ects estimations would ameliorate the bias arising from unobservable heterogeneity, but we cannot use them because the various corporate governance indicators are either invariant or change slowly over time. Omitted variables constitute another problem. We add several control variables that proxy for speci…c …rm characteristics (ownership, size, growth opportunities, leverage, and capital intensity) and …nd that our governance results are not spuriously

16

caused by these omitted variables. There is an additional source of endogeneity that arises from the possibility that governance choices may be a function of past performance. Wintoki et al. (2012) demonstrate that a panel dynamic GMM estimator can be used to control for the dynamic nature of the performance-governance relationship. An important aspect of this methodology is that it relies on a set of internal instruments, i.e. past values of governance and performance, which eliminates the need of external instruments. In the absence of appropriate exogenous instruments, the dynamic GMM estimation methodology is our best attempt at statistically addressing endogeneity. We re-estimate speci…cation 1 in a dynamic panel setting by using dynamic GMM estimations. As Wintoki et al. (2012) argue, any bias arising from the underlying dynamic nature of governance appears to be more important in regressions of governance on performance than in regressions of …rm characteristics on governance. We estimate a dynamic model that includes a lag of ROA as the explanatory variable and that uses the system GMM estimator. The system GMM estimator uses the levels equation to obtain a system of two equations: one di¤erenced and one in levels. By adding the second equation additional instruments can be obtained. Thus the variables in levels in the second equation are instrumented with their own …rst di¤erences. This usually increases e¢ ciency. The rule of thumb is to avoid a number of instruments larger then the number of cross-sections to prevent that the Sargan test is weak. We thus lag variables at most four periods to arrive at valid instruments for this dynamic model.5 All explanatory variables are treated as endogenous variables. Table 5, columns 1 to 6, report the results. The de…ned GMM estimation passes all the relevant speci…cation tests, i.e., the AR(2) second-order serial correlation tests and the Hansen J test of over-identifying restrictions. The AR(2) tests yield p-values well above the 10% threshold across all the speci…cations, which means that we cannot reject the null hypothesis of no second-order 5

Our regressions are executed using xtabond2 in Stata. The large number of endogenous variables and hence of instruments is problematic within the subsample of …rms with a large dominant shareholder. In this case, due to the small number of observations, the number of instruments is larger than the number of …rms thus weeaking the Sargan test. For this subsample of …rms we we use the "collapse” option of xtabond2. The collapse option speci…es that xtabond2 should create one instrument for each variable and lag distance, rather than one for each time period, variable, and lag distance. This option avoids instrument proliferation. For more details, see Wintoki et al. (2012).

17

serial correlation. Furthermore, the J-statistics have p-values above 0.10 and as such, we cannot reject the hypothesis that our instruments are valid. The results in Table 5 thus con…rm the earlier evidence in Tables 3 and 4. In untabulated results, we further deal with endogeneity by instrumenting the endogenous variable #No Explanations and running 2SLS estimations. A good instrument should be correlated with the endogenous variable but uncorrelated with the error term in the structural equation. We use the average #No Explanation of …rms in the same industry, based on Fama-French industry groups and years, which captures industry factors explaining corporate governance.6 The rationale for using this instrument is that corporate governance choices of the industry peer group are likely to in‡uence the corporate governance choices of an individual …rm. Thus, we would expect that this instrumental variable would have a positive relationship with the corporate governance indicator in our …rst stage regressions. Since our dependent variable measure (ROA) is already adjusted for industry e¤ects, it is reasonable to assume that the instrument is not correlated with the error term. Such an instrument has been widely used in the existing corporate governance literature (e.g., John, Litov, and Yeung, 2008) with the usual caveat related to its true "exogeneity" which, unfortunately, we cannot statistically test.7 Untabulated regressions con…rm the OLS results observed in Tables 3 and 4. With the usual caveats and limitations of the various methodologies, the above tests give us some con…dence in limiting concerns about endogeneity when interpreting the causal relationship between corporate governance and performance.

4.2

Ownership

We use a dummy variable speci…cation to identify the presence of a dominant individual shareholder. In untabulated regressions, we replicate Table 2 regressions by using the fractional level of family ownership in the company. The evidence is con…rmed. 6 We exclude industries with only one company observation. Our results are robust also to the exclusion of industries with less than …ve companies. 7 In the …rst stage regression the instrument coe¢ cient estimate shows a positive and signi…cant (at 1%) relationship, while the Cragg-Donald Wald F statistics is 173:06. Based on the Stock et al. (2002) critical values we can statistically reject the null hypothesis of weak instruments.

18

Dominant ownership is computed at the 10% threshold. Since 10% of voting rights is frequently su¢ cient to exert control, this cut-o¤ is used extensively (e.g., La Porta et al., 1999, Dahya et al., 2008). The results presented in Table 2 and Table 4 are qualitatively con…rmed when using the 5% and 15% threshold (results not shown). The results within the widely-held subsample do not hold when using the 20% dominant ownership threshold. At the 20% threshold we only have 120 company-year observations of …rms with a large shareholder (10% of the sample). This is consistent with the descriptive statistics provided by Faccio and Lang (2002) who shows that 90% of the UK largest …rms are widely-held at the 20% threshold. In essence, when computing dominant ownership at the 20% threshold many companies with a large dominant shareholder e¤ectively "controlling" the company at the 10% threshold would fall in the widely-held category, thus generating insigni…cant estimations.

4.3

Alternative proxies

We have found that …rms with a dominant shareholder tend to be more opaque in their governance structure because they comply less and disclose less. The literature has developed other indexes of corporate transparency related to di¤erent types of information dissemination structures. For instance, Anderson, Duru and Reeb (2009) proxy corporate opacity by using the bid-ask spread or the total number of equity analysts following each …rms. Bid-ask spreads are a widely-held used proxy for information asymmetry among investors, similarly to the number of analysts variable which is used to capture the intensity of market scrutiny. We …nd that …rms with a dominant shareholder tend to be followed by a lower number of analysts and have higher bid-ask spreads, which indicates a higher level of corporate opacity. The above results are not reported but they are available upon request.

19

4.4

Alternative speci…cations

4.4.1

Quality of earnings

An alternative indicator of agency problems in a …rm is the quality of reported earnings. In this subsection we investigate the e¤ect of corporate governance on reported earnings quality. We follow the accounting literature and use the Dechow and Dichev (2002) model, as modi…ed by Ball and Shivakumar (2006), as a proxy for earnings quality. We estimate the following equation using data for all available companies in the UK from Worldscope at the two-digit SIC industry level.

ACCt =

+

1 CFt

+

2 CFt 1

+

3 CFt+1

+

4 DCFt

+

5 DCFt

CFt + "t

(2)

where: ACCt = total accruals at t, scaled by average total assets at t; total accruals are earnings before extraordinary items minus operating cash ‡ows; CFt = operating cash ‡ows at t, scaled by average total assets at t; CFt

1=

operating cash ‡ows at t-1, scaled by average total assets at t;

CFt+1 = operating cash ‡ows at t+1, scaled by average total assets at t; DCFt = one if the change in cash ‡ows at t is less than zero (CFt

CFt

1

< 0), and zero

otherwise. We then use the above industry estimates to calculate the residual for companies in our sample. The residual captures the unexpected portion of total accruals that deviate from economic transactions. The absolute value of the residuals, i.e., abnormal accruals, from equation (2) is our proxy for the quality of earnings. The higher the value of abnormal accruals is, the worse the quality of reported earnings is. Table 6, Panel A, shows OLS results when ACCt is used as the dependent variable for the subsamples of widely-held and non widely-held …rms, with controls8 and robust standard errors. Columns 2 and 4 show that widely-held compliant or explaining …rms have lower accruals, i.e. 8

We use Size, Leverage, Capital Intensity and ROA as control variables.

20

higher quality of reported earnings, than noncompliant and not explaining widely-held ones. There is no signi…cant association between stricter governance practices or better disclosures and earnings quality in …rms with a dominant shareholder (columns 1 and 3). This is consistent with the evidence described in Wang (2006) and Ali, Chen, Radhakrishnan (2007) that, for a sample of US …rms, earnings quality is worse in non-family …rms because agency problems are worse than in family …rms. The above results thus complement our earlier …ndings and provide further evidence that corporate governance and disclosure reduce agency problems between managers and shareholders in widely-held companies, but they have little impact in …rms with a controlling shareholder. 4.4.2

Acquisitions activity

Acquisition activity might proxy for agency costs in a company (see e.g., Gompers et al. (2003)). Managers of …rms, in the absence of oversight, are more likely to engage in higher acquisition activity for empire building reasons or simply to entrench themselves. We investigate whether there is an association between governance practices and acquisitions activity by using the next year ratio of net assets from acquisitions to total assets as a proxy for acquisitions activity. Net assets from acquisitions (Worldscope item # 04355) represent assets acquired through pooling of interests or mergers. We regress it on the dummy Comply or Explain or the index # No Explanations and various control variables9 . Standard errors are robust clustered. Table 6, Panel B, shows that the dummy Comply or Explain is negative and signi…cant and the index # No Disclosures is positive and signi…cant within the subsample of widely-held …rms, i.e., widely-held …rms with stringent governance practices and better disclosures are associated with a lower assets from acquisitions ratio. This result shows that widely-held …rms with weak boards (i.e., not conforming to high corporate governance standards and lacking of transparency) engage in higher acquisition activity. In contrast, we do not observe such a behavior in …rms with a dominant shareholder. This result is consistent with previous evidence that managers will engage in such activities to derive private bene…ts of control if not monitored by the board or the dominant shareholder. 9

We use the growth rate of Sales and the logarithm of the ratio of the Market to Book value of equity as control variables.

21

4.5

Alternative performance variable

The performance analysis uses ROA as the dependent variable because operating performance re‡ects the e¢ cient use of company assets. For robustness, we use the logarithm of the marketto-book ratio as an alternative measure of performance in lieu of ROA.10 Table 6, Panel C, shows that the market-to-book results are very comparable with the evidence on ROA shown in Table 4: stricter corporate governance practices and better quality disclosure have a signi…cant impact in widely-held companies but not in …rms with a dominant shareholder.

5

Summary and Discussion

This paper examines how corporate governance choices and transparency di¤er between …rms with concentrated or dispersed shareholders, and their consequences for performance. Our analysis takes advantage of an institutional setting in which corporate governance standards are recommended but not legally mandated and companies have ‡exibility in their governance choices. We …nd that …rms with a dominant shareholder are more likely to deviate from corporate governance standards and to disclose less information about their governance choices, which is consistent with the role of the dominant family shareholder as the monitor in-place. We show that the impact of governance and disclosure varies across …rms: better governance practices and disclosures are associated with better performance in widely-held companies, but they do not have a signi…cant impact on performance in …rms with a dominant shareholder. This di¤erence suggests that corporate governance practices matter little in …rms with dominant shareholders because the large shareholder acts as an alternative governance mechanism, alleviating free-rider problems and e¤ectively monitoring managers. In contrast, corporate governance practices matter in widely-held companies where boards adhering to better corporate governance practices and disclosures align managers’incentives to an e¢ cient use of the company resources. Our results show that the corporate governance of …rms is the results of complex interdepen10

As in Durnev and Kim (2005), we use the following control variables: Sales, growth rate of Sales, R&D Expenses/Assets, Capital Expenditures/Assets.

22

dences that go beyond a one-size-…ts-all perspective. For instance, Chhaochharia and Grinstein (2007) …nd that the 2002 US governance rules had a di¤erent impact across …rms and some provisions were detrimental to small …rms. When judging and valuing a …rm’s corporate governance, we must consider the role and in‡uence of ownership structure on corporate governance practices and outcomes. Our study further suggests that ownership, corporate governance practices and disclosures are substitute governance and monitoring mechanisms that reduce agency costs. Hence, noncompliance with some corporate governance practices does not necessarily imply bad governance in a …rm. The UK corporate governance regime designed on the "Comply or Explain" system a¤ords companies discretion in their choices. Because the dominant shareholder optimally substitutes the monitoring role of the board and operates in an environment that guarantees high legal protection to minority shareholders, company-level governance practices do not have an additional e¤ect on performance. High legal protection of minority shareholders in the UK is thus an important condition underlying the prevalence of the monitoring hypothesis over the expropriation hypothesis by dominant individual shareholders. In such an environment, the issue of "who monitors the controlling shareholder" is solved by the investor protection laws (La Porta, Lopez-de-Silanes, Shleifer and Vishny, 2000). At the same time, the absence of mandated regulations and formal monitoring of disclosure gives discretion to companies, which some of them take advantage for self-serving goals. This suggests that voluntary regulation may have drawbacks in countries with weak legal protection where companies are less inclined to be transparent and are more likely to take advantage of the allowed ‡exibility for expropriation and entrenchment purposes (Claessens et al., 2002). The question of whether mandatory regulation would help solve the agency con‡icts and be preferred to voluntary regulation is left for future research.

23

Appendix - The Combined Code provisions 1. Separation of Chairman and CEO: There should be a clear division of responsibilities at the head of the company (Provision A.2.1); 2. Senior Non-executive Director (SNED): Whether the posts (Chairman/CEO) are held by di¤erent people or by the same person, there should be a strong and independent non-executive element on the board (SNED), with a recognized senior member other than the chairman to whom concerns can be conveyed (Provision A.2.1); 3. Non-executive Directors: They should comprise not less than one third of the board (Provision A.3.1); 4. Independent Non-Executive Directors: The majority of non-executive directors should be independent of management and free from any business or other relationship (Provision A.3.1); 5. The Term of Service Contracts: it should be one year or less (Provision B.1.7); 6. Nomination Committee: it should be established to make recommendations to the board on all new board appointments. A majority of the members of this committee should be non-executive directors (Provision A.5.1); 7. Compensation Committee: it should be established for a transparent policy on executive compensation. It should consist exclusively of independent non-executive directors (Provisions B.1 and B.2.2); 8. Audit Committee: it should be established for transparent arrangements on the …nancial reporting. The committee should consist of at least three non-executive directors, a majority of whom should be independent (Provision D.3).

24

References [1] Aghion, P., and J. Tirole, 1997, Formal and Real Authority in Organizations, Journal of Political Economy, 105, 1-29 [2] Anderson, R., and D. Reeb, 2003, Founding family ownership and …rm performance: Evidence from the S&P 500, Journal of Finance, 58, 1301-1328. [3] Anderson, R, A. Duru, and D. Reeb, 2009, Founders, heirs, and corporate opacity in the United States, Journal of Financial Economics, 92, 205-222. [4] Ali, A., T.-Y Chen, and S. Radhakrishnan, S. 2007, Corporate disclosures by family …rms, Journal of Accounting and Economics, 44, 238-286. [5] Ball, R., and L. Shivakumar, 2006, Role of Accruals in Asymmetrically Timely Gain and Loss Recognition, Journal of Accounting Research, 44(2), 207-242 [6] Barber B. M., and J. D. Lyon, 1996, Detecting long-run abnormal operating performance: The empirical power and speci…cation of test statistics, Journal of Financial Economics, 41, 359-399 [7] Bolton P., and E. Von Thadden, 1998, Blocks, liquidity and corporate control, Journal of Finance, 53, 1–25. [8] Boone, A., L. C. Field, J. M. Karpo¤, and C. G. Raheja, 2007, The Determinants of Corporate Board Size and Composition: An Empirical Analysis, Journal of Financial Economics, 85, 66-110. [9] Burkart, M., F. Panunzi, and A. Shleifer, 2003, Family Firms, Journal of Finance, 58, 21672201 [10] Bushman, R.M., and A.J. Smith, 2003, Transparency, …nancial accounting information, and corporate governance, FRBNY Economic Policy Review, 65-87 [11] Chen, S., X. Chen, and Q. Cheng, 2008, Do Family Firms Provide More or Less Voluntary Disclosure?, Journal of Accounting Research, 46, 499-536. [12] Chhaochharia, V., and Y. Grinstein, 2007, Corporate Governance and Firm Value - The Impact of the 2002 Governance Rules, Journal of Finance 62.4, 1789-1825. [13] Claessens, S., S. Djankov, J.P. Fan, and L.H. Lang, 2002, Disentangling the incentive and entrenchment e¤ects of large shareholdings, Journal of Finance, 57, 2741-2771. [14] Committee on Corporate Governance, 1998, Final Report, Gee Publishing, London. [15] Conyon, M. J. and S. I. Peck, 1998, Board control, remuneration committees and top management compensation, Academy of Management Journal, 41, 146-157.

25

[16] Core, J. E., W. R. Guay and T. O. Rusticus, 2006, Does Weak Governance Cause Weak Stock Returns? An Examination of Firm Operating Performance and Investors’Expectations, Journal of Finance, 61, 655-687 [17] Dahya, J., O. Dimitrov and J. J. McConnell, 2008, Dominant Shareholders, Corporate Boards and Corporate Value: A Cross-Country Analysis, Journal of Financial Economics, 87, 73-100 [18] Dahya, J., and J. J. McConnell, 2007, Board Composition, Corporate Performance and the Cadbury Committee Recommendation, Journal of Financial and Quantitative Analysis, 42, 535-564. [19] Dechow, P., and I. Dichev, 2002, The Quality of Accruals and Earnings: The Role of Accrual Estimation Errors, The Accounting Review, 77, 35–59. [20] Dedman, E., 2003, Executive turnover in UK …rms: the impact of Cadbury, Accounting and Business Research, 33, 33-50. [21] Durnev A., and E. H. Kim, 2005, To steal or not to steal: Firm attributes, legal environment and valuation, Journal of Finance, 60, 1461-1493. [22] Ellul A., L. Guntay and U. Lel, 2009, Blockholders, Debt Agency Costs and Legal Protection, FRB International Finance Discussion Paper No. 908 [23] Faccio, M., and L. H. P. Lang, 2002, The ultimate ownership of western European corporations, Journal of Financial Economics, 65, 365-395. [24] Fidrmuc, J., M. Goergen, and L. Renneboog, 2006, Insider trading, news releases and ownership concentration, Journal of Finance, 61, 2931-2973. [25] Financial Reporting Council, 2009, Review of Combined Code. Progress Report and Second http://www.frc.org.uk/corporate/reviewCombined.cfm

the E¤ectiveness of the Consultation, available at

[26] Gompers, P., J. Ishii and A. Metrick, 2003, Corporate governance and equity prices, The Quarterly Journal of Economics, 118, 107-156. [27] Goergen, M., L. Renneboog and C. Zhang, 2008, Do UK institutional investors monitor their investee …rms? Journal of Corporate Law Studies, 8, 39-56. [28] Hermalin, B. E. and M. S. Weisbach, 2003, Boards of Directors as an Endogenously Determined Institution: A Survey of the Economic Literature, Economic Policy Review, 9, 7-26. [29] Hope, O-K., and W. B. Thomas, 2008, Managerial Empire Building and Firm Disclosure, Journal of Accounting Research, 46, 591-626. [30] John, K., L.P. Litov. and B. Y. Yeung, 2008, Corporate Governance and Risk Taking, Journal of Finance, 63, 1679-1728.

26

[31] Judge, W., 2011, The Pivotal Role of Ownership, Corporate Governance: An International Review, 19(6), 505-506. [32] Klapper, L., and I. Love, 2004, Corporate governance, investor protection, and performance in emerging countries, Journal of Corporate Finance, 10, 703-28. [33] Kothari, S., S. Shu, and P. Wysocki, 2009, Do managers withhold bad news?, Journal of Accounting Research, 47, 241-276. [34] Lang, M., K. Lins, and D. Miller, 2004, Concentrated control, analyst following, and valuation: Do analyst matter most when investors are protected least?, Journal of Accounting Research, 42, 589-623. [35] La Porta R., F. Lopez-de-Silanes and A. Shleifer, 1999, Corporate ownership around the world, Journal of Finance, 54(2), 471-517. [36] La Porta R., F. Lopez-de-Silanes, A. Shleifer and R. Vishny, 2000, Investor protection and Corporate Governance, Journal of Financial Economics, 58, 3-27. [37] Leuz, C., D. Nanda, and P. Wysocki, 2003, Earnings Management and Investor Protection: An International Comparison, Journal of Financial Economics, 69, 505-527. [38] Shleifer, A., and R.W. Vishny, 1986, Large Shareholders and Corporate Control, Journal of Political Economy, 94, 461-88 [39] Shleifer, A., and R.W. Vishny, 1997, A Survey of Corporate Governance, Journal of Finance, 52, 737-84. [40] Stock, J.H., J.H. Wright, and M. Yogo, 2002, A survey of weak instruments and weak indenti…cation in generalized method of moments, Journal of Business and Economics Statistics, 20, 518-529. [41] Wang, D., 2006, Founding Family Ownership and Earnings Quality, Journal of Accounting Research, 44(3), 619-656. [42] Wintoki, M.B., J.S. Link and J. M Netter, 2012, Endogeneity and the dynamics of internal corporate governance, Journal of Financial Economics, 581-606.

27

Table 1: Summary statistics This table shows statistics for our sample of UK companies analyzed over the period 1998-2004. Panel A shows the corporate governance indicators gathered from the annual reports of companies. Comply is a dummy variable that takes value 1 if a company complies with all the provisions of the Combined Code, and 0 otherwise. Comply or Explain is a dummy that takes value 1 if the company is either compliant with all the provisions of the Code or provides an explanation for all its noncompliances, and 0 otherwise. # No Explanations counts the number of noncompliances that are not explained. Dominant Shareholder is a dummy equal to 1 if a company has a family, individual or privately held …rm owned by an individual or family a shareholder with at least 10% controlling stake, and 0 otherwise. Panel B shows the …nancial characteristics of …rms. Widely-held are …rms that do not have a family, individual, or strategic entity as a shareholder with at least 10% stake. Panel C reports the di¤erence in mean tests between …rms with no dominant shareholders (Widely-held) and …rms with a dominant shareholder. N

Mean

Std. Dev

P25

Median

P75

Panel A: Corporate governance indicators Comply

1261

0.324

0.468

0

0

1

Comply or Explain

1261

0.844

0.363

1

1

1

# No Explanations

1261

0.208

0.565

0

0

0

1135

0.176

0.381

0

0

0

Dominant Shareholder

Panel B: Firm characteristics ROA

1261

0.049

0.099

0

0.033

0.088

Age (Years)

1137

39.29

34.45

12.000

25

65

Size (GBP million)

1261

3002.24

5490.78

451.144

1022.8

3050

Market capitalization (GBP million)

1255

3714.32

12575.18

405.430

856.58

2727.70

Sales (GBP millions)

1261

2798.11

8069.99

407.965

954.19

2586

Market to book

1192

4.31

7.04

1.340

2.22

3.85

Leverage

1261

0.19

0.16

0.07

0.17

0.28

Capital intensity

1261

0.66

1.05

0.12

0.26

0.61

Panel C: Di¤erence of means test Widely

Dominant

t-stat

Held

Shareholder

Comply

0.338

0.245

Comply or Explain

0.853

0.770

2.92***

# No Explanations

0.190

0.340

-3.32***

ROA

0.049

0.046

0.38

Age Total assets (GBP million)

2.56***

39.674

35.453

1.46

3069.180

1398.460

4.29***

Market capitalization (GBP million)

3522.287

1419.413

2.72***

Sales (GBP millions)

2965.462

1518.421

2.26**

Market to book

4.208

4.484

-0.50

Leverage

0.208

0.163

3.58***

Capital intensity

0.636

0.673

-0.48

935

200

Number of observations

28

Table 2: Corporate Governance and Dominant Shareholders This table shows the results of a logistic regression with Comply and Comply or Explain as dependent variables in columns (1) and (2), respectively. Comply is a dummy variable that takes the value 1 if a company complies with all the provisions of the Combined Code, and 0 otherwise. Comply or Explain is a dummy that takes value 1 if the company is either compliant with all the provisions of the Code or provides an explanation for all its noncompliances, and 0 otherwise. Dominant Shareholder is a dummy equal to 1 if a company has a family, individual or privately held …rm owned by an individual or family a shareholder with at least 10% controlling stake, and 0 otherwise. Market capitalization is the logarithm of the market capitalization of the company. Age is the logarithm of the number of years since incorporation. Leverage is the ratio of long term debt to the book value of assets. Capital intensity is the ratio of net property, plant, and equipment to sales. Pro…tability is the ratio of earnings before interest and taxes to sales. T-statistics are in parentheses. ***, ** and * denote signi…cance at the 1%, 5%, and 10% levels, respectively.

Dominant Shareholder

(1)

(2)

Comply

Comply or Explain

-0.679*

-0.680**

(-1.88)

(-1.98)

0.072

0.259*

Market capitalization Age Leverage Capital intensity

(0.70)

(1.69)

-0.260**

-0.235

(-2.11)

(-1.24)

-1.457*

-0.967

(-1.82)

(-1.02)

0.305**

0.152

(2.06)

(0.92)

0.507

0.276

(0.64)

(0.22)

Pro…tability Intercept

0.549

0.302

(0.34)

(0.11)

Industry and year dummies

Yes

Yes

N

961

961

Pseudo R-squared

0.20

0.16

29

Table 3: Corporate Governance and ROA This table shows results of ordinary least squares regressions with one year ahead industry-adjusted ROA as the dependent variable, where ROA is de…ned as the ratio of earnings before interest and taxes to book value of assets adjusted for the median from the respective Fama-French industry group. Comply is a dummy variable that takes value 1 if a company complies with all the provisions of the Combined Code, and 0 otherwise. Comply or Explain is a dummy that takes value 1 if the company is either compliant with all the provisions of the Code or provides an explanation for all its noncompliances, and 0 otherwise. # No Explanations counts the number of noncompliances that are not explained. Size is the logarithm of book value of assets. Leverage is the ratio of long term debt to the book value of assets. Capital intensity is the ratio of net property, plant, and equipment to sales. T-statistics are in parentheses. ***, ** and * denote signi…cance at the 1%, 5%, & 10%, respectively. (1) Comply

(2)

(3)

(4)

-0.011*

-0.011*

-0.000 (-0.01)

Comply or Explain

0.015 (1.56)

# No Explanations Size Leverage

(-1.82)

(-1.82)

-0.010**

-0.011**

-0.011**

-0.010*

(-2.11)

(-2.28)

(-2.33)

(-1.79) 0.016

0.001

0.001

0.000

(0.02)

(0.04)

(0.01)

-0.4

Capital intensity

-0.015***

-0.015***

-0.015***

-0.019***

(-3.56)

(-3.59)

(-3.59)

(-3.30)

Intercept

0.195***

0.205***

0.209***

0.178**

(2.93)

(3.10)

(3.14)

(2.37)

Yes

Yes

Yes

Yes

N

1,261

1,261

1,261

853

R-squared

0.05

0.06

0.06

0.06

Sample

All

All

All

NC

Time dummies

30

Table 4: Corporate Governance, Dominant Shareholders and ROA This table shows results of ordinary least squares regressions with one year ahead industry-adjusted ROA as the dependent variable, where ROA is de…ned as the ratio of earnings before interest and taxes to book value of assets adjusted for the median from the respective Fama-French industry group. Regressions are run within the subsample of …rms with a dominant shareholder (Dominant Shareholder ) and with no dominant shareholders (Widely-Held ). Comply or Explain is a dummy that takes value 1 if the company is either compliant with all the provisions of the Code or provides an explanation for all its noncompliances, and 0 otherwise. # No Explanations counts the number of noncompliances that are not explained. Size is the logarithm of book value of assets. Leverage is the ratio of long term debt to the book value of assets. Capital intensity is the ratio of net property, plant, and equipment to sales. T-statistics are in parentheses. ***, ** and * denote signi…cance at the 1%, 5%, & 10%, respectively. (1) Sample Comply or Explain

(2)

(3)

(4)

Dominant

Dominant

Widely

Widely

Shareholder

Shareholder

Held

Held

-0.010

0.021*

(-0.42)

(1.86)

# No Explanations

0.004

-0.016*

(0.46) Size Leverage

(-1.94)

-0.016*

-0.016*

-0.009

-0.010

(-1.89)

(-1.83)

(-1.59)

(-1.64)

0.119

0.122*

-0.025

-0.025

(1.67)

(1.71)

(-0.60)

(-0.60)

Capital intensity

-0.025*

-0.025*

-0.014***

-0.014***

(-1.85)

(-1.86)

(-3.09)

(-3.10)

Intercept

0.284**

0.287**

0.166**

0.171**

(2.51)

(2.39)

(2.28)

(2.35)

Time dummies

Yes

Yes

Yes

Yes

N

200

200

935

935

R-squared

0.10

0.09

0.05

0.06

31

Table 5: Arellano-Bond estimations This table reports the results of Arellano-Bond GMM regressions for the entire sample as well as within subsamples of …rms with a dominant shareholder (Dominant Shareholder ) and with no dominant shareholders (Widely-Held ). Comply or Explain is a dummy that takes value 1 if the company is either compliant with all the provisions of the Code or provides an explanation for all its noncompliances, and 0 otherwise. # No Explanations counts the number of noncompliances that are not explained. Control variables used are ROAt , Size which is the logarithm of book value of assets, Leverage which is the ratio of long term debt to the book value of assets and Capital intensity which is the ratio of net property, plant, and equipment to sales. T-statistics are in parentheses. ***, ** and * denote signi…cance at the 1%, 5%, and 10% levels, respectively.

Sample ROAt Comply or Explain

(1)

(2)

(3)

(4)

(5)

(6)

All

Dominant

Widely

All

Dominant

Widely

Shareholder

Held

Shareholder

Held

0.740***

0.145

0.657***

0.727***

0.231

0.642***

(12.13)

(0.25)

(9.73)

(11.72)

(0.35)

(9.66)

0.017**

-0.008

0.017*

(2.24)

(-0.18)

(1.72) -0.012***

0.004

-0.013**

(-3.01)

(0.09)

(-2.21)

# No Explanations Size Leverage Capital intensity Intercept Time dummies N

-0.012*

-0.155

-0.013**

-0.016**

-0.148

-0.014**

(-1.96)

(-1.55)

(-2.17)

(-2.56)

(-1.55)

(-2.19)

0.056

0.097

0.086

0.066

0.157

0.083

(1.15)

(0.39)

(1.53)

(1.41)

(0.48)

(1.42)

-0.011*

-0.027

-0.01

-0.011

-0.025

-0.011

(-1.92)

(-1.12)

(-1.21)

(-1.63)

(-1.15)

(-1.30)

0.162*

2.157

0.176**

0.219**

2.075

0.194**

(1.96)

(1.63)

(2.17)

(2.55)

(1.54)

(2.24)

Yes

Yes

Yes

Yes

Yes

Yes

1,261

200

935

1,261

200

935

Test for AR(1) p-value

0.00***

0.87

0.00***

0.00***

0.68

0.00***

Test for AR(2) p-value

0.32

0.28

0.32

0.35

0.34

0.36

Hansen Test p-value

0.31

0.57

0.34

0.36

0.90

0.44

32

Table 6: Alternative speci…cations and performance variable This table reports results of ordinary least squares regressions within subsamples of …rms with a dominant shareholder (Dominant Shareholder ) and with no dominant shareholders (Widely-Held ). Comply or Explain is a dummy that takes value 1 if the company is either compliant with all the provisions of the Code or provides an explanation for all its noncompliances, and 0 otherwise. # No Explanations counts the number of noncompliances that are not explained. The following dependent variables are used: the absolute value of abnormal accruals (Panel A), the ratio of net assets from acquisitions to total assets (Panel B), and the logarithm of the market to book ratio (Panel C). T-statistics are in parentheses, and ***, ** and * denote signi…cance at the 1%, 5%, and 10% levels, respectively.

Panel A: Quality of Earnings Sample Comply or Explain

(1)

(2)

(3)

(4)

Dominant

Widely

Dominant

Widely

Shareholder

Held

Shareholder

Held

0.026

-0.047**

(0.54)

(-2.28) -0.012

0.032**

# No Explanations

(-0.47)

(2.28)

0.00

0.57***

-0.01

-0.57***

(-0.00)

(3.48)

(-0.05)

(-3.46)

Controls, Time, Industry dummies

Yes

Yes

Yes

Yes

N

184

837

184

837

0.02

0.02

0.02

0.06

Intercept

R-squared

Panel B: Acquisitions Activity Sample Comply or Explain

(1)

(2)

(3)

(4)

Dominant

Widely

Dominant

Widely

Shareholder

Held

Shareholder

Held

0.018

-0.021**

(1.65)

(-2.35) -0.014*

0.018**

(-1.93)

(2.51)

# No Explanations Intercept Controls, Time, Industry dummies

0.039

0.037***

0.053

0.014

(1.04)

(2.78)

(1.48)

(1.52)

Yes

Yes

Yes

Yes

N

180

805

180

805

R-squared

0.04

0.07

0.05

0.08

33

Table 6: Alternative speci…cations and performance variable (continued)

Panel C: Market to Book ratio Sample Comply or Explain

(1)

(2)

(4)

Dominant

Widely

Dominant

Widely

shareholder

Held

shareholder

Held

0.511

1.560*

(0.24)

(1.75) -0.552

-1.621**

# No Explanations Intercept

(3)

(-0.42)

(-2.35)

9.272

2.528

9.819

4.805

(0.77)

(0.60)

(0.76)

(1.15)

Controls, Time, Industry dummies

Yes

Yes

Yes

Yes

N

182

809

182

809

R-squared

0.18

0.20

0.18

0.20

34