«Three essays on corporate boards R. Øystein Strøm A dissertation submitted to BI Norwegian School of Management for the degree of Dr.Oecon SERIES ...»
Therefore the CEO turnover and board changes are the outcome of the interactions of CEO and the board. The model is initiated when the CEO is hired, and after one period earnings are realised. At this point, the board updates its judgement of the CEO. If performance is weak, the CEO is dismissed, if it is good the board and CEO negotiate over a ﬁxed wage and monitoring intensity, the latter approximated by the fraction of independent directors on the board2. If the negotiations fail, the CEO leaves. If they succeed, the continuing board is less independent than the preceding, that is, independent directors leave, or the board is enlarged with inside or gray directors. In the third stage, the board may receive a private signal, and the CEO may be discharged on this new information. Thus, the CEO is supposed to leave at three points; after weak performance, after failed negotiations, and after the board obtains adverse private information. On the other hand, a board member may leave after successful negotiations, that is, after good performance. However, the two never leave at the same time. Alternatively, the CEO uses his negotiating power to enlarge the board with dependent or gray directors. If the CEO continues to turn in good performances, board independence is gradually dissipated. In consequence, we should observe a steady stream of director departures or board enlargements during the CEO’s tenure.
2 An independent director is commonly deﬁned as a person without family ties to the CEO and business ties to the ﬁrm. Similarly, a director from management is dependent.
A gray director delivers business advisory and ﬁnancing services to the ﬁrms, such as management consultants and bankers. An independent board has at least 50 per cent independent directors.
CHAPTER 4. BOARD CONTROLNeedless to say, the potential for agency costs is high if this is the case.
When performance is weak, but the CEO’s accumulated negotiating position is strong, the absence of owners means that the CEO stays, and that board independence is unchanged. Unchanged board independence is again the outcome when the CEO’s negotiating position is weak, but then the CEO may be discharged. One might expect that the board would become more independent after a weak performance, that is, board turbulence would increase after such performance. But the model does not predict such a result, since this would require a third party intervening in the director elections. The third party, shareholders, is disregarded. Thus, the model says that directors leave after the CEO has assumed ofﬁce, and that the CEO will be discharged after weak performance and when the board is independent.
Hermalin and Weisbach (1998) make the caveat that the assumed equal monitoring intensity for all directors does not hold. If directors free-ride on other directors’ diligence, the old board will have an incentive to want diligent new directors. Then the board may remain independent, even though directors leave or the board is enlarged during the incumbent CEO’s tenure. However, this caveat does not upset the timing of CEO and director departures.
The authors arrive at their results on the institutional assumption that owners are dissipated and therefore unable to exercise their voice alternative (Hirschman, 1970). But shareholders may also be prevented by a regulatory regime allowing such CEO protection as CEO-chairman duality and a staggered board, studied in Goyal and Park (2002) and Falaye (2007). If CEO control is observed, it is uncertain whether this is due to endogeneity in the Hermalin and Weisbach (1998) sense, or if it is due to CEO protection. However, if CEO protection is not allowed, the endogeneity hypothesis of CEO control may be tested. Since Norway has a low level of CEO protection (Aarbakke et al., 1999), the data in this paper allows such testing.
A middle position between the extremes of CEO and shareholder control is the joint control of CEO and the board. The Adams and Ferreira (2007) model, termed the friendly board, may serve as a point of departure. They do not model CEO or director turnover, but the information exchange between board and the CEO. If the CEO supplies information, the board responds with better advice, but also better monitoring. The CEO appreciates advice, but dislikes monitoring. A balance between the two is found, with the board becoming ‘friendly’ in not requiring too much inforTHEORY AND HYPOTHESES 113 mation. When the CEO and the board share information in this friendly way, they also become jointly responsible for the ﬁrm’s performance. Joint responsibility should also imply that their departure will be simultaneous.
Under joint control, the board gives up some monitoring in exchange for CEO information. Thus, a potential for agency costs is allowed for the greater beneﬁt of a more informed board, and therefore better joint board and CEO decisions.
However, instead of co-operating for the good of the company, the CEO and the board may collude for their own private beneﬁt. Tirole (2006, p. 356) describes collusion as a state when the board is lenient towards the CEO who reciprocates by tunnelling corporate resources to his monitors.
But if this is the case, outside ownership concentration should have no impact upon board departures. Thus, if such an association does appear, the joint control is not of the collusion kind.
In summary, under shareholder control the CEO and directors leave independently of each other; under CEO control directors leave after the CEO has assumed ofﬁce; and under joint control CEO and directors leave simultaneously. Let us look more speciﬁcally at CEO turnover. Under shareholder control we should expect no association between CEO turnover and board turbulence. In the case of CEO control lagged or simultaneous board turbulence should show no effect upon CEO turnover. When it comes to joint control, however, board turbulence should predict a simultaneous CEO turnover.
Now we turn to board turbulence. Under CEO control, directors should leave after the CEO has entered ofﬁce. Thus, a lagged CEO turnover should have a signiﬁcant impact upon board turbulence. Under board and joint control, this aspect should be absent. Instead, if joint control is the case, a simultaneous effect should be apparent.
4.3.2 Ownership and independence The discussion in the last section concerns not only the timing of CEO turnover and board turbulence. Also, ﬁrm performance, board independence, and ownership belong to the determining variables. I look at these in turn, and then expand my discussion to include related explanations and to control variables.
For reference, table 4.1 reports variable deﬁnitions and main statistical properties.
Table 4.1 CHAPTER 4. BOARD CONTROL
CEO turnover and board turbulence are explained more fully in section 4.4.
An important feature in the Hermalin and Weisbach (1998) model is the relation between board turbulence and ﬁrm performance. Under CEO control, the better the ﬁrm performance, the higher board turbulence is expected to be. The prediction is unlike the one for shareholder control, as the better the ﬁrm performance, the lower is board turbulence. This last prediction is common with joint control. For CEO turnover, a weaker ﬁrm performance implies a higher likelihood of departure.
As already noted, in the Hermalin and Weisbach (1998) model the shareholders play no monitoring role due to the separation of ownership and control. When this is the case the shareholders’ “voice” alternative (Hirschman, 1970) is closed. From a shareholder value perspective this may seem strange, as it removes the ordinary line of authority between shareholders and the board. Furthermore, the underlying assumption that dissipated ownership remains ﬁxed if the ﬁrm sees continued weak performance is unrealistic. Such a situation would seem to welcome takeover specialists, who could build a majority stake and then replace the board, whereupon the board replaces the CEO (Scharfstein, 1988; Jensen and Ruback, 1983 and Jensen, 1988). A larger shareholder will internalise more of the beneﬁts of oversight. Therefore a plausible prediction is that the monitoring intensity of shareholders increases with outside ownership concentration as the Shleifer and Vishny (1986) and Bolton and von Thadden (1998) models suggest.
I predict that higher ownership concentration has no effect upon CEO turnover under any control type. For board turbulence, I predict that under CEO control the effect is insigniﬁcant, while it is positive under board and joint control. The effects of external ownership concentration are, therefore, important in differentiating between control types.
An independent board may be a monitoring substitute for a large shareholder (Fama, 1980 and Fama and Jensen, 1983). In the literature on CEO turnover the more independent the board is the more likely it is to discharge a CEO. Thus, if independence is important in explaining CEO turnover, it cannot be CEO controlled. This rather increases the case for shareholder control. However, if the board is jointly controlled, board independence should be insigniﬁcant in explaining CEO turnover. It is also likely that board turbulence decreases with board independence, since the independent board will perform the monitoring that shareholders are looking for.
4.3. THEORY AND HYPOTHESES 115 Thus, the timing of CEO and director turnover, ﬁrm performance, outside ownership, and independence may differentiate between control types.
But the CEO and the directors may be entrenched for various reasons. Under CEO control the prediction is that the effects are positive.
First, if the CEO or the board themselves are owners, they are at least partly shielded from shareholders’ discipline. Thus, a higher CEO ownership is likely to lead to lower likelihood of own turnover, while a higher board ownership stake should likewise be associated with lower board turbulence.
Second, CEO entrenchment is associated with lower CEO turnover and higher director turnover. I use three indicators of CEO entrenchment from Bøhren and Strøm (2007) to investigate this further. First, independent directors could be friends of the CEO, or at least, have the same background, and therefore, will be less effective monitors (Gilson and Kraakman, 1991 and Tirole, 2006). Alternatively, this ‘imported CEO’ may carry information from similar business conditions, and as such be a better monitor.
Thus, the ﬁrst is a colluding interpretation, the second an information interpretation of imported CEOs. Second, the CEO may be a director of other ﬁrms. The direction of effect may be uncertain here as well. External board directorships in isolation are a private beneﬁt to the CEO and a sign of entrenchment, indicating a lower CEO turnover. But the ﬁrm’s performance may suffer with a busy CEO and increase his likelihood for dismissal. Third, the CEO may be a member of the board. Goyal and Park (2002) ﬁnd that when the CEO holds the dual role of chairman, CEO turnover is less sensitive to performance. By law, this duality is not allowed in Norway (Aarbakke et al., 1999). Thus, these entrenchment variables may alternatively be seen as information variables. Under CEO control we should expect the variables to be negatively related to CEO turnover and board turbulence.
Third, I would expect that a well networked board would be less likely to be ﬁred, since such a board has been shown to create shareholder value (Ferris et al., 2003 and Bøhren and Strøm, 2007), but could be more likely to ﬁre the CEO, since it is in a better position to compare outside candidates for the CEO post with the performance of the incumbent CEO. Thus, the networked board is a strong board in the Almazan and Suarez (2003) sense. Notice that all CEOs are eliminated from the network measure, which furthermore is constructed from social network theory (Wasserman and Faust, 1994), and shows direct (to neighbours) and indirect (to neighbours’ neighbours) links to other ﬁrms through multiple directorships.
CHAPTER 4. BOARD CONTROLThus, the measure represents the global properties of the whole board network (Conyon and Muldoon, 2006).
The network variable may gauge director busyness rather than information. If this is the case, I expect a higher score on the network variable to be associated with lower likelihood or CEO turnover. Fich and Shivdasani (2006) ﬁnd that busy outside directors have a lower tendency to replace a non-performing CEO than a board with independent, but nonbusy directors. They use the number of other directorships as the measure of busyness, creating a dichotomous variable with a cut-off point at three additional directorships. However, this is different from my network variable. Thus, my variable has more to do with board information than director overstretching. In order to compare, I include measures of the number of other directorships similar to the Fich and Shivdasani (2006) busy directors in separate regressions.
Employees and incumbent managers are natural allies against shareholders in the models of Coffee (1990); Pagano and Volpin (2005) and also in Roe (2003). If this is true, employees can be counted among the managers’ friends. In the Norwegian data I may study this aspect, since nearly half of the observations in the sample have employee representatives on the board. This representation is due to legislation stipulating that ﬁrms with more than 200 employees are obliged to have employee directors, in the 30 to 200 range they are obliged only if an employee majority vote is in favour, and in both cases only if the industry is not exempted (Aarbakke et al., 1999). Since further compromises with the functioning CEO and board are more likely than with a new, employee directors may be opposed to change. This implies that CEO turnovers and board turbulence are less frequent in co-determined ﬁrms.
The higher is the age of the CEO and director, the more likely they are to leave their positions. This variable is routinely included in turnover studies.
Finally, ﬁrm size and systematic risk constitute control variables. In probit regressions, year and industry dummies are added. This is further explained in section 4.5.