«Three essays on corporate boards R. Øystein Strøm A dissertation submitted to BI Norwegian School of Management for the degree of Dr.Oecon SERIES ...»
1.2. ALIGNED, INFORMED, AND DECISIVE 5 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 (1.1) is positive (negative). According to Hermalin and Weisbach (1998), the board is more independent the higher the value of (1.1).
The board may have access to better information when the CEO is also a board member, when the CEO is a board member in other companies (exported CEO), when CEOs of other companies are directors (imported CEO), and when directors who are not CEOs are directors in other companies and meet directors in yet other companies. This so-called small world property of interlocking corporate boards (Conyon and Muldoon,
2006) is approached with network theory (Wasserman and Faust, 1994), that is, with the information centrality measure.
Finally, the board’s decisiveness is approached with the board size and board heterogeneity. Decisions are supposed to be made more easily the smaller and more homogeneous is the group (Mueller, 2003). Yermack (1996) and Eisenberg et al. (1998) show that smallness is a desirable property in boards. Gender, the age dispersion of directors, and the fraction of employee directors constitute the heterogeneity measures. Most studies of the gender effect in boards, for instance Shrader et al. (1997) and Smith et al. (2006), conclude that increased gender heterogeneity means lower ﬁrm performance. Likewise, evidence in FitzRoy and Kraft (1993);
Gorton and Schmid (2000) and Falaye et al. (2006) show a negative relation between employee directors and ﬁrm performance. The age dispersion is new to the literature. Notice that employee board representation is imposed through the political process, while gender diversity became strongly recommended during the period under study, but no concern is given to age dispersion.
In the main relation Tobin’s Q, measured as the ﬁrm’s market value over its book value of assets, is the dependent variable whose variation we seek to explain using the above board characteristics. Panel data give repeated observations of the same ﬁrm, so that data need to be transformed in order to remove the serial dependence in the error terms. We choose the ﬁxed effects method (Woolridge, 2002) to do so. An advantage of this method is that the need for control variables disappears. Furthermore the general method of moments (GMM) estimation method gives further advantages compared to rival methods, as it is free from assumptions of the variable’s distribution and its non-linearity.
We ﬁnd that ﬁrm performance improves with higher director ownerCHAPTER 1. OVERVIEW ship and network score, while a larger and more heterogeneous board reduces ﬁrm performance. Gender and employee directors are negatively related to ﬁrm performance. Thus, our results are consistent with what is observed in other investigations. We subject the results to robustness tests, in that different deﬁnitions of e.g. the network variable is used. Still, we ﬁnd that the results withstand these tests.
A common objection to performance studies is that board characteristics are a result from earlier performance. Hermalin and Weisbach (1998) hold that the board is endogenously chosen. Thus, in their theory, there is a “reverse causation” that runs from ﬁrm performance to board characteristics. Furthermore, board characteristics may be internally related, either as complements or as substitutes to each other (Agrawal and Knoeber, 1996). A long-time panel data set alleviates endogeneity problems (Bhagat and Black, 1999), since ﬁrm performance is related to different board characteristics across ﬁrms at a given time, but also contains variations in these relationships in the same ﬁrms across time. Furthermore, most boards are elected in the spring and early summer, and the ﬁrm performance recorded at year-end. Thus, the same-year board impact of the board’s characteristics should become evident. Nevertheless, we test for endogeneity explicitly in two ways. First, a lagged Tobin’s Q is included in the basic regression, and second, important board characteristics are assumed to be dependent.
The tests reveal some endogeneity. In particular, and remembering that a higher network score means better ﬁrm performance, we ﬁnd that well-performing ﬁrms tend to attract well-connected directors. We also ﬁnd that board characteristics are complements rather than substitutes, if they are internally related. However, when we include the preceding period’s performance none of the earlier results are materially upset, indicating that the reverse causation hypothesis receives conﬁrmation to a minor degree only.
Our results have relevance for the ongoing debate about corporate governance in many countries. In particular, we ﬁnd no economic argument for mandating independence or diversity, such as requiring by law or code that a minimum fraction of directors are independent, employees, or of a given gender. If anything, regulatory implications are the opposite of what is argued in the public domain: Regulators should encourage more owners in the boardroom, more directors with multiple seats, fewer employee directors, less gender diversity, and smaller boards. Independence is already at its optimal level and needs no regulatory pressure
1.3. CO-DETERMINATION 7 of the type which is currently introduced worldwide. In fact, regulation aimed at preventing costly scandals in a small number of ﬁrms may end up destroying more value in the vast majority of ﬁrms.
1.3 Better ﬁrm performance with labour on the board?
While the “Aligned, informed, and decisive” paper presents a broad overview of board issues, the second paper looks at the relationship between ﬁrm performance and employee directors in more detail. Although codetermination (employee board representation) exists in many European countries, the issue is under-researched. Jensen and Meckling (1979) predict that codetermination needs backing in laws, and this is indeed the case for Norway, where codetermination was enacted in 1972 (Aarbakke et al., 1999).
The central idea of the paper is that the presence of employees on the board has both a direct and indirect effects upon ﬁrm performance. The indirect effects come about when shareholders adjust unregulated governance mechanisms in order to compensate for perceived negative consequences of employee representation. Buchanan and Tullock (1962) predict such an outcome in the political arena, and in Germany much effort has been done to avoid codetermination effects, for instance, by limiting the number of meetings in the supervisory board.
Former studies mostly ﬁnd a negative relationship between ﬁrm performance and employee board representation. This applies to Germany (FitzRoy and Kraft, 1993; Schmid and Seger, 1998; and Gorton and Schmid,
2000) and to Canada (Falaye et al., 2006), in addition to our own results in the “Aligned, informed, and decisive” essay. However, Fauver and Fuerst (2006) report a positive effect in information-intensive industries in Germany. None of the studies investigate the indirect effects employee directors may have upon other governance variables, and only our own uses panel data. Also, compared to the German and Canadian experience the Norwegian institutional setting offers some advantages. First, employee directors sit on the board, while they are present in the rarely convening and little informed supervisory board only in Germany. Second, the employee directors are elected due to their employment in the ﬁrm, not because they are owners as in Canada, or as union representatives, as in Germany where one third of the labour representation is reserved for union representatives. Third, in all size classes and in all industries a dichotomy arises between those ﬁrms that have employee directors and
CHAPTER 1. OVERVIEWthose that have none. This is useful for testing as it allows the creation of sub-samples with e.g. codetermined ﬁrms with more than 200 employees. In all, the Norwegian institutional framework should give a sharper picture of the employee director effect than former studies.
The relationship between employee directors and ﬁrm performance is treated in different strands of literature? From a property rights perspective Becht et al. (2003) argue that board representation can be a guarantee against shareholders’ expropriation of employees’ rents through their “urge to dismiss”, and that this may bring forth employees’ investments in ﬁrm-speciﬁc human capital and the cessation of industrial action. The ﬁrm-speciﬁc human capital makes the employee a residual claimant on par with investors. The stakeholder theory position is that employees are superior monitors of management, since they observe day-to-day action (Blair and Stout, 1999). Board representation implies that owners may get access to the information. Both these positions would predict a positive relationship to ﬁrm performance.
A negative relationship may result from the collective choice problems that parties with divergent interests induce in decision-making (Tirole, 2001 and Hansmann, 1996). The shareholders’ objective is to maximise ﬁrm value, while the employees’ objectives are to maximise wages and to protect ﬁrm-speciﬁc human capital. The resultant mixed ﬁrm objective means longer decision time and compromise decisions. The CEO tends to be a compromise maker rather than a shaper of the ﬁrm under a clear objective (Tirole, 2002). The implied consensual decision model in co-determination means that the ﬁrm pursues stability and predictability instead of bold new moves (Siebert, 2005).
Furthermore, Pistor (1999) and Hopt (1998) point out that employees’ monitoring may not be truthfully revealed, as in economising on the supply of internal information to the board, and also in leakage of sensitive board information. Employees may also use moral arguments against for instance plant closures or high management pay. The shareholder elected directors may have trouble withstanding such arguments, since they may experience large personal costs and small personal gains from taking adverse employee decisions (Baker et al., 1988). Thus, even though the employees are in a minority position in the board, they may inﬂuence board decisions to their advantage.
When shareholders fear such effects of co-determination, they may adjust governance mechanisms in order to neutralize the codetermination impact imposed through regulation by placing a heavier weight on the
1.3. CO-DETERMINATION 9 unregulated (Buchanan and Tullock, 1962). The employee directors variable’s effect upon board characteristics is one such neutralizing action, another is to increase leverage so as to give the CEO less scope for wastage (Easterbrook, 1984 and Jensen, 1986). These previously unexplored indirect effects make a simultaneous equations approach necessary. I term this relationship the co-determination hypothesis.
The model tests the hypothesis that codetermination is negatively related to ﬁrm performance, but it has relevance for the endogeneity debate as well (see Hermalin and Weisbach, 2003 and Bhagat and Black, 1999).
Hermalin and Weisbach (1998) suggest that governance mechanisms are endogenously determined by former ﬁrm performance. Thus, a reverse causation is proposed, since the lagged ﬁrm performance should have a negative relationship to governance mechanisms. With good ﬁrm performance, the CEO has bargaining power to reduce monitoring intensity. A rival endogenous hypothesis is that good ﬁrm performance induces even better monitoring, since the ﬁrm needs to improve its governance to maintain good performance. I test for this, and term this the reverse causation hypothesis.
Finally, a link between the board index and the leverage reveals whether the governance variables are complements or substitutes (Agrawal and Knoeber, 1996).
Instead of using individual board characteristics, I construct a board index from signiﬁcant variables in the AID essay. The variables directors’ ownership, network, board size, and gender receive equal weights in the index, and they keep their estimated sign from the AID essay paper. Following Bertrand and Mullainathan (2001), the variables are standardised to have zero mean and standard deviation of one in order to add them.
The econometric estimation is carried out with simultaneous equations regressions using the three-stage least squares (3SLS)4 methodology (Greene, 2003). I use ﬁxed effects estimation (Woolridge, 2002) in regressions for the whole sample and for sub-samples of co-determined and shareholder determined ﬁrms, and for co-determined ﬁrms with more than 200 employees. I also perform robustness tests.
First of all, the co-determination hypothesis ﬁnds conﬁrmation in the negative, direct effect upon form performance, and in the indirect impact upon the board index and average wage, which are both positive.
4 The 3SLS is an instrumental variable estimation method. The instruments are the predicted values of the dependent variable from a regression on all the explanatory variables in the system (Greene, 2003, p. 398). Thus, regressions are undertaken in several steps, or stages, hence the name.
CHAPTER 1. OVERVIEWThus, shareholders try to adjust board characteristics so as to neutralize the harmful direct effect of co-determination. But the indirect effect of the board index turns out to be economically weaker than the direct effect of employee directors, so that the shareholders are only partially able to neutralize the negative direct effect. The overall effect of co-determination is therefore negative. The results become progressively stronger in regressions from the overall sample to the sub-sample of co-determined ﬁrms and then to co-determined ﬁrms with more than 200 employees.
I also conﬁrm the positive impact of employee directors upon leverage.
On the other hand, higher leverage means lower ﬁrm performance, a fact that is often found in the empirical literature (Barclay et al., 1995; Rajan and Zingales, 1995; and Brick et al., 2005), but contradicts predictions in theory (Easterbrook, 1984 and Jensen, 1986) that a higher leverage brings better ﬁrm performance. Thus, the employee director impact upon ﬁrm performance is even more negative when leverage is taken into account.