«Three essays on corporate boards R. Øystein Strøm A dissertation submitted to BI Norwegian School of Management for the degree of Dr.Oecon SERIES ...»
Our second contribution is new empirical proxies for board independence and for director network. Important parts of the board literature and most countries’ corporate governance codes classify directors as dependent if they are afﬁliated, i.e., have past or present business or family relations to the ﬁrm. A possible reason why the literature has produced inconclusive evidence on how performance relates to board independence is that the independence proxy is theoretically ad–hoc. According to the Hermalin and Weisbach (1998) model, what matters for director independence is not afﬁliation. Rather, it is the relative timing of entry, i.e., whether the director was appointed before or after the current CEO took ofﬁce. Our independence proxy reﬂects this characteristic. Similarly, the existing literature measures director network simply by the number of board seats directors hold in other ﬁrms. This deﬁnition assumes every board seat is equally important as an information source, and it doublecounts when more than one of the ﬁrm’s directors sit on the same outside board. Our network proxy avoids double-counting and treats each seat individually according to its information centrality, accounting for both the direct information effect of sitting on another ﬁrm’s board and the indirect effect of meeting directors on that board who hold seats in still other boards.
Third, we try to control for endogeneity as efﬁciently as possible. The endogenous relationship between performance and independence was modeled theoretically by Hermalin and Weisbach (1998), who recently used this framework to show how board design by owners can be understood as a response to regulatory change (Hermalin and Weisbach, 2006). However, whereas endogeneity is difﬁcult to control for with the classic simultaneous equations methodology, equation-by-equation estimation with GMM is a more robust alternative for two reasons. First, because the true system of simultaneous equations is unknown, the coefﬁcients will be biased if the simultaneously estimated system is misspeciﬁed. In particular, if one equation is misspeciﬁed, the estimates of the remaining equations in the system will be contaminated as well. Unlike in the classic simultaneous equations approach, estimating each regression separately with GMM keeps the misspeciﬁcation local (Woolridge (2002)). Second, GMM can handle correlation between the error term and the independent variINTRODUCTION 25 ables, which is typical in endogenous relationships.
Our ﬁnal contribution comes from the regulatory setting, which offers an opportunity to explore the role of three board mechanisms that are often addressed by regulators and public opinion. These are independence, employee directors, and gender diversity. First, Norwegian corporate law rules that the ﬁrm’s CEO cannot be its chairman. Thus, decisions on the CEO-chairman duality have been moved from owners to regulators, who mandate separation for all ﬁrms. Although not by law, it is an empirical fact that non-CEO members of the management team are never directors in their ﬁrm. Even the CEO is not on the board in roughly two thirds of our sample ﬁrms. Thus, regulation and a voluntary restriction on board composition jointly produce boards that are at least formally less controlled by management than most boards in other countries. These exogenous restrictions increase the power of tests that relate corporate performance to board independence. The second unusual characteristic is that when a listed Norwegian ﬁrm employs more than 200 people, the employees choose one third of the directors. Since roughly 40% of our sample ﬁrms have employee directors, the cross-sectional variation of this board characteristic allows us to analyze the performance effect of mandatory employee directors, which is quite unexplored in the literature.3 Finally, both independence and gender diversity in the boardroom were heavily discussed in the second half of our sample period, and gender diversity was mandated two years after its end.4 Since we know each director’s gender, we can explore whether gender diversity has unexploited economic potential that owners will not capture unless regulators mandate it.
We ﬁnd that corporate performance as measured by Tobin’s Q is signiﬁcantly higher when insider ownership is high (i.e., aligned) and when directors have wide networks through seats they hold in other ﬁrms (informed). Firms with small boards, low gender diversity, and no employee directors are more valuable than others (decisive). These relationships are statistically signiﬁcant at standard levels, and the economic signiﬁcance is stronger for the alignment and information mechanisms than for decisiveness. In contrast, we ﬁnd no evidence that independence relates systematically to performance. These results, which control for endogeneity, 3 Firms in the newspaper, shipping, petroleum extraction, and ﬁnancial service industries are exempted. 62% of the sample ﬁrms have more than 200 employees, and two thirds of them have employee directors.
4 A corporate governance code for listed ﬁrms issued in 2004 recommends at least 50% independent directors. A law passed in 2004 mandates at least 40% directors of each gender in listed ﬁrms from 2006 on.
CHAPTER 2. ALIGNED, INFORMED, AND DECISIVEare robust to alternative model speciﬁcations. Consistent with theoretical predictions, we do ﬁnd that board design mechanisms are endogenous, both relative to each other and to performance. For instance, higher gender diversity increases board size (i.e., mechanisms drive each other), and directors with strong information networks both improve performance and gravitate towards well-performing ﬁrms (two-way causation between mechanisms and performance). Moreover, board mechanisms are complements rather than substitutes. For instance, the decisiveness mechanisms show that boards with low gender mix are smaller and have less age diversity, which all contribute to more homogeneity. Also, and as expected, board-external ﬁrm characteristics matter for board composition. For instance, higher risk generates lower insider holdings, more networked directors, and less gender diversity. Finally, the ﬁnding that several board mechanisms relate signiﬁcantly to performance in an endogenous system apparently shows that owners do not design their board optimally (Demsetz and Lehn, 1985). Such a conclusion may be premature, since the regulator does not allow the owners of our sample ﬁrms to freely design the optimal board. Mandatory employee directors in large ﬁrms is an example of such an exogenous, binding restriction on board design.
These ﬁndings imply that well-functioning directors are not necessarily independent of the CEO, which runs counter to conventional wisdom behind recent corporate governance regulation. However, it is line with most existing empirical results and also consistent with the theoretical model of Adams and Ferreira (2007). The evidence suggests that owners are able to trade off a director’s role as hands-off monitor against the role as hands-on adviser, and that stronger emphasis on independence may hurt not just the advice function, but also the value of monitoring, which is the only function captured by the Hermalin and Weisbach (1998) model and also the one receiving all the attention in current board regulation. Similarly, it seems directors have multiple seats not because they elbow themselves into the board room, but due to the valuable information network they bring along. Also, the negative association between performance and diversity in terms of more gender mix, larger board size, and the use of employee directors does not support the claim that director heterogeneity is an underexploited resource, and that owners will not capture its economic value unless regulators help them. Finally, the endogeneity of board mechanisms is a real-world phenomenon, making both research and regulation of board design more difﬁcult.
The rest of the paper is organized as follows. Section 2.2 reviews the litTHEORY, EVIDENCE, AND METHODOLOGY 27 erature and explains how our methodology deviates from the one used by others. Section 2.3 describes the institutional framework, explains the data selection procedure, and presents the descriptive statistics. We formally test the relationship between board design and economic performance in section 2.4, whereas section 2.5 provides robustness checks. Section 2.6 summarizes and concludes.
2.2 Theory, evidence, and methodology
Becht et al. (2003) recently concluded that the theory of board design is grossly underdeveloped. This characteristic of a young, immature paradigm is problematic for empiricists. Although the board design problem is multidimensional, each theory is partial and addresses one or a few board design mechanisms. Thus, theory cannot predict what the full set of value-creating board mechanisms looks like in equilibrium. Neither can it specify the expected internal relationship between major mechanisms, such as the endogeneity between insider equity holdings and board size.
Consequently, estimated relationships between the mechanisms and how they drive performance should be considered stylized facts rather than tests of well-founded hypotheses.
In the following, we explain our choice of focus and methodology by reviewing the existing literature. We organize the discussion around the three major concerns underlying the choice of any speciﬁc board design mechanism, which are to align the interests of principals and agents (section 2.2.1), provide information for monitoring and support (2.2.2), and to enhance the board’s effectiveness as a decision-maker (2.2.3).
2.2.1 Interest alignment Interest alignment in a board context concerns the ﬁrm’s ownership structure and the degree of independence between monitoring directors and monitored ofﬁcers.
The theory of corporate governance argues that ownership concentration matters for interest alignment by inﬂuencing the principal’s incentives and power to monitor the agent (Shleifer and Vishny, 1986). Both properties are stronger the higher the ownership concentration, and inside ownership concentration (equity holdings by ofﬁcers and directors) is more powerful than outside concentration because inside owners are better informed and have direct access to the ﬁrm’s decision-making. However, because powerful insiders may entrench themselves and exploit their
CHAPTER 2. ALIGNED, INFORMED, AND DECISIVEoutside co-owners, the expected relationship between inside ownership concentration and market value is positive at low concentration levels and declining after a certain point which generally cannot be speciﬁed ex ante.
The empirical evidence on the relationship between outside concentration and ﬁrm performance is mixed and inconclusive (Gugler, 2001). As for inside concentration, which is the more relevant ownership characteristic in a board setting, the predicted curvilinear relationship has received consistent support by studies that mostly ignore other board design mechanisms than insider ownership. However, this clean result does not carry over to the board literature, where the models include more board characteristics than just insider holdings. Hermalin and Weisbach (1991); Byrd and Hickman (1992); Yermack (1996); Cotter et al. (1997) and Bhagat and Black (2002) all ﬁnd a positive relationship, but it is only signiﬁcant in Hermalin and Weisbach (1991) and Yermack (1996). Thus, adding more mechanisms to a board design model than just ownership structure may easily blur the mostly clean empirical relationship between insider ownership and ﬁrm performance found in simpler models. Our comprehensive model allows us to study this issue more closely. We measure outside ownership concentration by the Herﬁndahl index based on all outside owners5. Insider ownership is proxied for by the directors’ aggregate equity fraction in the ﬁrm.
The board literature and existing corporate governance codes argue that monitoring quality is higher the stronger the independence between directors and managers. Such independence is generally thought to reﬂect the directors’ ability to monitor without feeling pressure from the monitored CEO. Arguing that this issue involves more than just outside vs.
inside directors, Byrd and Hickman (1992) introduce a ﬁner partition by distinguishing between inside, afﬁliated outside, and independent outside directors. Only the latter type has no past or present business or family ties to the ﬁrm.
The empirical evidence on the relationship between such independence measures and ﬁrm performance is inconclusive. Baysinger and Butler (1985) estimate a ten-year lagged positive effect, Hermalin and Weisbach (1991) ﬁnd no signiﬁcant link, while the relationship is negative and significant in Yermack (1996); Agrawal and Knoeber (1996); Klein (1998); Bhagat and Black (1999), and Bhagat and Black (2002). One possible reason for this 5 The Herﬁndahl index for outside ownership concentration is the sum of squared ownership fractions across all the ﬁrm’s outside owners. Its maximum value is one (a single investor owns every share held by the outsiders), approaching its minimum value of zero as the ownership structure gets increasingly diffuse.
2.2. THEORY, EVIDENCE, AND METHODOLOGY 29 low consistency is the missing theoretical justiﬁcation for the afﬁliationbased independence measure. To increase the power of the test, we base our independence measure on the Hermalin and Weisbach (1998) model, where the CEO’s ability to recruit dependent directors increases with the ﬁrm’s past performance. This model predicts that the longer the history of good performance under the current CEO, the less independent the current board. Thus, the key independence criterion is not afﬁliation, but whether the director was appointed before or after the CEO took ofﬁce.6 Consistent with the Hermalin-Weisbach model, we measure a board’s independence as the difference between the average tenure of its non-CEO
directors and the tenure of the CEO:
1n ∑ non-CEO director tenurei − CEO tenure (2.1) Independence ≡ n i=1 where non-CEO director tenurei is the number of years since non-CEO director i entered ofﬁce, and n is the number of shareholder–elected directors. The average director has longer (shorter) tenure than the CEO when expression (2.1) is positive (negative). According to Hermalin and Weisbach (1998), the board is more independent the higher the value of (2.1)7.