«Three essays on corporate boards R. Øystein Strøm A dissertation submitted to BI Norwegian School of Management for the degree of Dr.Oecon SERIES ...»
JEL Classiﬁcation codes: G32, G34 1 I have beneﬁted from discussions with Øyvind Bøhren. Pål Rydland and Bernt Arne Ødegaard have guided me to data.
4.1 Introduction A primary function of the board is to appoint and dismiss the CEO (Monks and Minow, 2001, p. 200). When the board is not able to fulﬁl this function, it increases the potential for agency costs associated with the separation of ownership and control (Berle Jr. and Means, 1932). Therefore turnovers of CEO and directors provide natural settings for studying manifestations of agency problems. Goyal and Park (2002) exploit this in a study of CEO turnover when the CEO holds the joint ofﬁce of chairman, while Falaye (2007) investigates the CEO turnover for staggered boards. In this paper, I argue that the timing of CEO departure relative to board enlargements and director substitutions can likewise shed light on agency problems, since different board control types give diverging predictions on the relative timing of departures. The data is from Norway, where neither CEO-chairman duality nor staggered boards are allowed (Aarbakke et al., 1999). Thus, CEO turnover and board changes are unhampered by CEO or director protection.
The results in Goyal and Park (2002) and Falaye (2007) indicate that the regulatory conditions give the CEO control over the board, since the turnover of CEOs in the ﬁrms is lower than in comparable ﬁrms. Hermalin and Weisbach (1998) present a model where the CEO gradually gathers control over the board’s composition due to former good ﬁrm performances. A prediction in the model is that directors leave during the CEO’s tenure, as the CEO is able to negotiate for laxer monitoring. While in the former case the CEO is entrenched due to board regulations, in the Hermalin and Weisbach (1998) model the CEO becomes entrenched due to his former ﬁrm performance success. Thus, when regulations favouring the CEO are not in place, we should observe the Hermalin and Weisbach (1998) situation. If we still cannot observe the model’s predicted pattern of CEO departures and board changes, control over the board must lie with either shareholders or with the CEO and the board together.
Thus, I differentiate between three board control types. The ﬁrst is the CEO control, where the CEO in effect elects his own directors. The implication for timing is that board changes occur during the CEO’s tenure when ﬁrm performance is good. Under the second control type, the board fulﬁls its primary function of hiring and ﬁring the CEO. I call this “shareholder control”. In this case, the turnovers of CEO and directors are unrelated. The third control type is “joint control”, when the board and the CEO together form a team, and are jointly responsible for ﬁrm performance. Since they are jointly responsible, a prediction is that the CEO and
4.1. INTRODUCTION 107director turnovers will be simultaneous. Thus, the fact that the three control types imply different patterns of timing in CEO turnover and board changes can be used to examine whether the typical director election is under shareholder, CEO, or joint control.
No earlier authors have utilised the relative turnover timing of CEOs and directors to investigate the manifestations of agency problems. In recent papers, Farrell and Whidbee (2000); Yermack (2004) and Fich and Shivdasani (2006) all show that outside directors are more likely to leave when a new CEO takes ofﬁce. But none of the authors make the timing of CEO and director departures the central issue of study. Of course, many studies deal with the separate CEO and director turnovers. Typically, these are related to ﬁrm performance and to board or CEO characteristics. For CEO turnover two results stand out. The ﬁrst is that weak performance increases the likelihood of CEO dismissals (Coughlan and Schmidt, 1985; Warner et al., 1988; Weisbach, 1988; Kaplan, 1994a,b) and Parrino (1997). The second is that an independent board is more likely to dismiss the CEO than a board ﬁlled with corporate insiders (Weisbach, 1988; Borokhovich et al., 1996; Hadlock and Lumer, 1997; Huson et al., 2001; and Dahya et al., 2002). Thus, a time-based test of board and CEO turnover timing and its relation to board control is lacking. In this paper, I deal explicitly with these joint departures.
To do so, data for several periods is needed. Since this paper employs panel data spanning fourteen years (1989 to 2002), I may study board changes in relation to CEO turnover. In order to incorporate both board size changes and director departures, I develop a board turbulence measure. My approach is to study the matter from two angles. First, the lagged and contemporaneous turbulence measure is used in a probit study to predict CEO turnover. Second, director turbulence is regressed against lagged and contemporaneous CEO turnover. Notice that these tests are not merely mirror regressions involving the same variables. In the CEO turnover regression, board turbulence precedes CEO turnover, while in the second, CEO turnover precedes board turbulence.
The different control types yield predictions not only for timing, but also for the monitoring role of shareholders. Under CEO control the shareholders play no role in the election of directors. Thus, if they do the control type must be of the joint control or the shareholder control type. The monitoring role of shareholders is recognised in models such as in Shleifer and Vishny (1986) and Bolton and von Thadden (1998), and conﬁrmed in the empirical literature (Morck et al., 1988; McConnell and Servaes, 1990).
CHAPTER 4. BOARD CONTROLFirm performance (ROA and stock return) and board independence belong to the set of explanatory variables. The Bøhren and Strøm (2007) proxy for board independence, that is, the tenure difference of board and CEO, is used. I include a variable for the board’s network from Bøhren and Strøm (2007) taking account of both direct and indirect links to other companies’ boards through multiple directorships, relying upon social network theory (Wasserman and Faust, 1994). Last, entrenchment variables such as the number of CEOs on the board may measure CEO-board collusion (Tirole, 2006), for instance the existence of friends on the board (Gilson and Kraakman, 1991). These variables should have a negative relation to CEO turnover in the CEO control type.
The sample comprises all non-ﬁnancial ﬁrms listed on the Oslo Stock Exchange (OSE) at year-end at least twice during the period 1989 to 2002.
The public securities register provided the ownership data, accounting and share price data is from the OSE’s data provider, and board data was collected manually from Kierulf’s Håndbok and a public electronic register from 1995.
I ﬁnd that the evidence supports joint control. CEO turnover and board turbulence are simultaneous; ownership matters in that outside ownership concentration is positively associated with board turbulence, but not with CEO turnover; and that entrenchment variables are either non-signiﬁcant or have opposite signs to those predicted under CEO control. In a robustness test, I construct a busy director indicator to follow up Ferris et al. (2003) and Fich and Shivdasani (2006). I ﬁnd that when the board is busy, it is less likely to ﬁre the CEO. Thus, a busy director is a less effective monitor. The results show little variation across speciﬁcations of ﬁrm performance and CEO turnover. I conclude that board control tends to be of the joint control type.
The paper proceeds as follows. The following section brieﬂy refers to earlier empirical investigations. Section 4.3 discusses theories behind the three control types and spells out speciﬁc hypotheses. In section 4.4 I explain the deﬁnition of CEO turnover and the construction of the board turbulence variable. Section 4.5 explains the panel data tobit and GMM regressions, and also gives an overview of the data used. Then section 4.6 shows descriptive statistics and ﬁgures of CEO-director turnovers. Section
4.7 shows results for CEO turnover, chairman turnover and board turbulence. The section also contains tests of differences in board turbulence between ﬁrms that have improved performance after CEO turnover compared to ﬁrms that have not. The ﬁnal section 4.8 presents my concluding
4.2. FORMER EMPIRICAL LITERATURE 109 remarks.
4.2 Former empirical literature Most empirical work in the area has been done on the CEO turnover or the top management team. The top management team is usually deﬁned as CEO and other executives or as CEO and chairman (Warner et al., 1988).
The triggering mechanism is usually weak ﬁrm performance, beginning with the papers of Coughlan and Schmidt (1985); Warner et al. (1988) and Weisbach (1988), and later Parrino (1997), who controlled for industry homogeneity. The role of the board was taken up in Weisbach (1988);
Borokhovich et al. (1996); Hadlock and Lumer (1997); Huson et al. (2001) and Dahya et al. (2002). The results are in general that weak ﬁrm performance leads to CEO turnover, and the more independent the board is, the more likely it is to dismiss a CEO. Denis and Denis (1995) downplay the role of the board, observing that over two-thirds of the forced resignations are due to blockholder pressure, takeover attempts, ﬁnancial distress, and shareholder lawsuits. In fact, Barclay and Holderness (1992) document that only 26% of the original CEOs were retained two years after a blockholding stake had changed hands. They also ﬁnd that the chairman is changed just as often.
In comparison, few studies have been done on director turnover. An early Hermalin and Weisbach (1988) study of inside and outside director departures around the CEO turnover event ﬁnds that inside directors tend to leave, but that the result is not conﬁrmed for outside directors. Kaplan (1994a,b) looked at both CEO and director turnovers in Germany and Japan, but examined both separately, not in conjunction. Other studies of director turnover are often limited to extraordinary circumstances such as company distress (Gilson, 1990) and takeovers (Franks and Mayer, 1996;
and Harford, 2003). The Franks and Mayer (2001) study of board turnover in German companies encompasses both a general study of turnover and one related mainly to takeovers. The focus in Farrell and Whidbee (2000) is the career concerns of outside directors who ﬁre a CEO, while Yermack (2004) includes new directors as a control variable to control for any simultaneity effect in CEO turnover. Kaplan and Minton (1994) and Yermack (2004) ﬁnd that CEO and director turnover are more likely to take place after weak ﬁrm performance. Fich and Shivdasani (2006) use a new CEO in regressions of director appointments and departures, yet their focus is not upon directors as such, but upon busy directors. Likewise, they
CHAPTER 4. BOARD CONTROLinclude busy directors in regressions of forced CEO departures.
Thus, a study that takes the simultaneity of CEO and director turnovers as its primary target of investigation seems to be missing. No study has exploited the information in the sequence of CEO and director departures to differentiate between various CEO-board control types.
4.3 Theory and hypotheses
CEO and director turnovers constitute a natural setting for studying agency problems (Goyal and Park, 2002). This applies both individually and for the relative timing of departures of the CEO and directors. For instance, a board that is unable to dismiss an underperforming CEO, but is instead subjected to changes dictated by the same CEO, is clearly a sign of agency problems. The board is not fulﬁlling its primary function to “(s)elect, regularly evaluate, and, if necessary, replace the chief executive ofﬁcer” (Monks and Minow, 2001, page 200). Such a situation has implications for CEO turnover relative to board changes, since directors tend to leave during the CEO tenure. But in other cases the board is fulﬁlling its primary function, and the timing of CEO turnover relative to board changes is different from the above. Thus, detecting a speciﬁc pattern in CEO turnover relative to board changes may show the existence of agency problems. I set out the various implications of timing below.
Director changes encompass both turnover and board size. Together they constitute the measure board turbulence, which is described in section 4.4. I ﬁrst set out hypotheses regarding CEO turnover and board turbulence in section 4.3.1 and add hypotheses concerning other variables in section 4.3.2.
4.3.1 Board control and timing
In the formal sense, the shareholders elect a board and the board hires or ﬁres the CEO. The board evaluates the CEO and dismisses him or her when necessary. The incumbent CEO is compared to alternatives, and if the rival is judged to be better, the incumbent is replaced. In the Almazan and Suarez (2003) model the size of the incumbent’s severance pay and his ﬁrm speciﬁc investment slow down the decision to replace. A weak board tends to grant a higher severance pay and to demand less ﬁrm speciﬁc investment. The relationship between the CEO and the board may be termed shareholder control. Here, the CEO leaves independently of
4.3. THEORY AND HYPOTHESES 111 director turnover. When this is so the CEO’s scope for extracting private beneﬁts is small.
The other extreme of the CEO-board relationship is CEO control, that is, the state when the CEO effectively appoints the board of directors. Berle Jr. and Means (1932, p. 82) state that this comes about when management appoints the proxy committee that nominates directors. Hermalin and Weisbach (1998) formalised their theory to explain endogenous determination of board composition, endogenous in the sense that board composition comes to depend on the CEO’s success. I give a verbal exposition of the model and discuss implications.
A fundamental assumption of the model is that shareholders’ inﬂuence is disregarded due to the institutional fact of dissipated ownership.