Is Broad Size An in Dependance Corporate Governace Machnisim

KYKLOS, Vol. 57 2004 Fasc.
3, 327356
Is Board Size an Independent Corporate Governance Mechanism?

Stefan Beiner, Wolfgang Drobetz, Frank Schmid and Heinz Zimmermann*
I. INTRODUCTION The board of directors plays a pivotal role in the governance of widely held corporations. It is generally acknowledged that the legal and contractual setting as well as the structure and activities of the board of directors have a non-negligible impact on the agency costs to which rms are exposed. At least in theory, the board of directors is one of the most important corporate governance mechanisms ensuring that managers pursue the interests of shareholders. Its task is to monitor, discipline, and remove ineffective management teams. One may suspect that several aspects and mechanisms are important in increasing the effectiveness of boards, such as board composition, board independence, and board size. While there is ample empirical evidence on board composition and board independence, there is only little international empirical evidence on the relationship between board size and rm valuation. Most importantly, previous papers do not take into account that there are possibly complex interrelationships between different corporate governance mechanisms, i.e., they neglect that board size is only one mechanism out of a wide menu of choices a rm faces. Yermack (1996) was the rst to report a negative relationship between board size and rm valuation. In general, an important issue in all empirical
* Sad Business School, University of Oxford, Park End Street, Oxford OX1 1HP, England, and Department of Finance, University of Basel, Holbeinstrasse 12, 4051 Basel, Switzerland, email: stefan.beiner@sbs.ox.ac.uk; Department of Corporate Finance, University of Basel, Petersgraben 51, 4003 Basel, Switzerland, email: wolfgang.drobetz@unibas.ch; Department of Economics, University of Bern, Vereinsweg 23, 3012 Bern, Switzerland, email: frank.schmid@ vwi.unibe.ch; Department of Finance, University of Basel, Holbeinstrasse 12, 4051 Basel, Switzerland, email: heinzz@bluewin.ch. We thank two anonymous referees, Yakov Amihud, Wolfgang Bessler, Stefan Duffner, Reiner Eichenberger, David Rey, and Markus Schmid for valuable comments. Financial support from the National Center of Competence in Research Financial Valuation and Risk Management (NCCR FINRISK) is gratefully acknowledged. The NCCR FINRISK is a research program supported by the Swiss National Science Foundation. Beiner acknowledges nancial support from the Swiss National Science Foundation (SNF).
327
G4_HS:Auftrge:HEL002:15491_SB_Kyklos_2004-03:15491-A02:Kyklos_2004-03_S-303-472
28.5.04
22. Juli 2004 13:07
STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN
work is the question whether board size is endogenous. Firm performance is both a result of the actions of previous directors and itself a factor that inuences the choice of subsequent directors. While previous studies with mainly US evidence have been careful to control for endogeneity, they did not account for any interrelationships between different, and possibly alternative, corporate governance mechanisms. Specically, these studies only looked at simultaneous equations with board size and rm valuation as the dependent variables, thereby ignoring the possibility that different governance mechanisms are substitutes, at least to some extent. The underlying notion of the latter presumption is that where one mechanism is used less, others could be used more, resulting in the same valuation effects. Accordingly, the relationship between board size and rm valuation could be driven by some other (so far unidentied) governance mechanism correlated with both board size and rm valuation. Therefore, in this paper we explicitly model the interrelationships between board size, board composition, leverage, and ownership structure in order to reexamine the issue of causality between board size and rm valuation in a more realistic setting. For example, the number of board outsiders may be positively related with board size. Therefore, it may be the case that any board size effect actually relates to the composition of the board, and possibly also to other governance mechanisms. To explore these interrelationships is an important extension to the previous literature and specically addresses the question whether board size is really an independent governance mechanism. Our second line of argument relates to previous work by Jensen (1993) and Lipton and Lorsch (1992), who were the rst to express the notion of an optimum board size of around 7 or 8 directors. Intuitively, they suggest that smaller boards lack the necessary management capacity. In contrast, large boards can be less effective than small boards, presuming that the emphasis on politeness and courtesy in boardrooms is at the expense of truth and frankness. Specically, when boards become too big, agency problems (e.g., director free-riding) increase and the board becomes more symbolic and neglects its monitoring and control duties. Moreover, large boards may reect an inadequate perception of the true executive function, particularly in rms with public involvement. Unfortunately, the ad-hoc claim of the existence of an optimal board size is inconsistent with previous empirical evidence. For example, Yermacks (1996) US sample includes rms with very large boards (with a mean of 12.3 board members), and the Finnish sample in Eisenberg, Sundgren and Wells (1998) is dominated by small- and medium-sized companies (with a mean of 3.7 board members). Interestingly, both papers nd a negative relationship between board size and rm valuation, which is clearly inconsistent with a unique optimal board 328
28.5.04
22. Juli 2004 13:07
IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?
size. The specic characteristics of our sample, consisting of publicly listed Swiss rms, allow us to shed more light on these contradictory results. As in the United States, in Switzerland the board of directors (Verwaltungsrat) shares top executive responsibility. Note that this contrasts with, e.g., the German system, where the Aufsichtsrat has a purely supervisory function. Interestingly, average board size in our sample is 6.6, which is just about the size suggested to be optimal by Jensen (1993) and Lipton and Lorsch (1992). This allows us to further explore the hypothesis of the existence of an optimal board size. Intuitively, in the optimum, the size of the board would be increased up to the point where marginal benet just equals marginal cost. In this case, as we will further discuss below, a carefully specied cross-sectional regression should nd no relationship between rm valuation and board size. This also applies to other governance mechanisms, such as outside representation, debt, and ownership structure. It is also important to note that the mere fact that the average board size in our sample is similar in magnitude to what Jensen (1993) and Lipton and Lorsch (1992) suggested, does not necessarily drive our results. Even though the average board size in our sample is close to optimal, this may only be a coincidence, and any seemingly validating results are driven by some other governance mechanism, which itself is related to rm valuation. This again demonstrates the importance of estimating a simultaneous system of equations with a wide range of governance mechanisms. To test our argument, we proceed in two steps. In a rst step, we examine whether board size is an independent governance mechanism. If this is not the case, all previous papers lack from an important shortcoming, i.e., they look at board size as the only governance mechanism and neglect important interrelationships. Estimating a simultaneous system of equations, we nd that board size is indeed an independent governance mechanism out of several from which a rm can deliberately choose. This implies that any relationship between board size and rm valuation could in fact be interpreted as causal. In a second step, we additionally include a measure of rm valuation to our simultaneous system of equations. In contrast to previous literature, we do not nd a signicant relationship between board size and rm valuation. Our results, therefore, support the Jensen (1993) and Lipton and Lorsch (1992) hypothesis. The remainder of this paper is as follows. Section II gives an overview of the related literature on board size and rm valuation. Section III explains the empirical approach. Section IV describes our sample of Swiss rms. Section V presents the empirical results, and section VI concludes.
329
28.5.04
22. Juli 2004 13:07
II. PRIOR RESEARCH ON BOARD SIZE Probably the most widely discussed question in the literature on boards of directors is whether having more outside directors increases corporate performance. Unfortunately, there is numerous empirical evidence, which is mixed at best (e.g., Morck, Shleifer and Vishny 1988, Hermalin and Weisbach 1988, and Bhagat and Black 1999). Interestingly, the maybe more obvious question about the size of the board of directors has received much less attention. Therefore, our focus is on board size, or more specically, on the proposal for limiting the size of boards in order to improve their effectiveness. One potential approach is to regard boards simply as a product of regulation. However, as Hermalin and Weisbach (2003) note, if boards simply satisfy regulatory requirements, they would represent deadweight costs to rms, which subsequent lobbying presumably would have eliminated, at least somewhere in the world. In any case, we should observe boards of minimum size. In practice, however, boards are generally larger than required by law. Therefore, a more plausible hypothesis is that boards are part of the equilibrium solution (but still second-best) to the contracting problem between dispersed shareholders and management (e.g., Hart 1995). Viewed from this perspective, boards should be interpreted as an endogenously determined institution that helps to alleviate agency problems in large rms. Jensen (1993) and Lipton and Lorsch (1992) were the rst to hypothesize that board size affects corporate governance. The underlying notion is that at some point the coordination problems outweigh the advantages from having more people to draw on. Jensen (1993) notes that when boards get beyond seven or eight people, they are less likely to function effectively and are easier for the CEO to control. Yermack (1996) presents empirical support for this proposition. Using a sample of large US public corporations, he reports an inverse association between board size and rm valuation. This nding is not necessarily clear ex ante; an alternative hypothesis could be that larger boards bring together specialists from various functional areas and therefore contribute to higher rm values. However, his results are robust to a variety of control variables, such as company size, growth opportunities and ownership structure. He carefully accounts for possible endogeneity problems. Causality runs from board size to rm valuation, and there is no evidence that companies change board size as a result of past performance. Specically, the results are inconsistent with the conjecture that board size increases after poor performance in an attempt to increase management capacity. Yermack (1996) also reports that smaller boards are more likely to dismiss CEOs for poor performance, and that CEO compensation is less dependent on performance if board size increases. 330
28.5.04
22. Juli 2004 13:07
Huther (1997) analyzes the relationship between board size and operating costs of US electricity companies. He also reports a negative board size effect. Nevertheless, because the possible endogeneity of board size is not properly taken into account, his results must be interpreted with due care. Conyon and Peck (1998) test the board size effect for a sample of rms from the United Kingdom, France, the Netherlands, Denmark, and Italy. Their results show a negative relationship between board size and rm valuation and the return on equity, respectively. However, the statistical signicance is rather mixed. In a related study for a sample of Dutch rms, Postma, van Ees and Sterken (2003) report a negative relationship between board size and the market-to-book ratio, but again they do not properly account for the endogeneity problem. Eisenberg, Sundgren and Wells (1998) nd a negative relationship between board size and rm valuation for a sample of small Finnish rms. The average number of board members is 3.7 in their sample, as compared to 12.25 in Yermack (1996). This is interesting because average board size in their sample is signicantly below the critical number of board members of about 7 or 8, which has been put forth by Lipton and Lorsch (1992) and Jensen (1993). Taken together, the results are clearly inconsistent with the existence of an optimal board size. To avoid biased estimates in ordinary least square regressions in the presence of endogeneity, Eisenberg, Sundgren and Wells (1998) test a simultaneous system of two equations, modeling board size and return on assets as the dependent variables 1. However, while this is an appropriate methodology, they do not account for possible relationships between different governance mechanisms. Finally, Loderer and Peyer (2002) analyze the effects of board overlap and seat accumulation on share prices for a panel of Swiss rms. They report that seat accumulation is negatively related to rm value. Board size is used as a control variable in pooled regressions. Their results document that larger board size is associated with lower rm valuation. Given that we also apply Swiss data but obtain contradicting results, it is important to note that their data set differs from that in this paper. Loderer and Peyer (2002) use a panel of all rms traded on the Swiss stock exchange in 1980, 1985, 1990 and 1995. In contrast, we use data as of end 2001. This is an important distinction because Loderer and Peyer (2002) report that average board size declined from 10.5 in 1980 to 8.5 in 1995. The corresponding number in our sample is 6.6. Hence, average board size has decreased even further (see Section IV). This number is close to the optimal board size put forth by Lipton and Lorsch (1992) and Jensen
1. They also report that while bad performance implies larger board changes, there is no signicant relation between the lagged return on assets and the net change in board size.
331
28.5.04
22. Juli 2004 13:07
(1993). Most important, in contrast to the interpretations in previous studies, Loderer and Peyer (2002) are hesitant to draw nal conclusions. They argue that the problem with the traditional argument is that it cannot explain why marginally larger boards should impair performance even if boards are small to begin with, which is the case in the Eisenberg, Sundgren and Wells (1998) study for Finnish rms. In addition, one could generally expect that executive committees are an effective way to increase board effectiveness and conceivably correct the drawbacks of large boards. However, Loderer and Peyer (2002) nd that the delegation of activities and responsibilities in companies with large boards to executive committees does not impact rm value. Therefore, they come up with an alternative explanation, suggesting that it is not so much that large boards make it harder to run rms properly, but rather that rms, whose governance systems are not working properly, are also characterized by larger boards. If this is the case, large board sizes simply identify rms that are run inefciently, but they are not the main reason for those inefciencies. Their considerations build an optimal starting point for our empirical analysis, since they seem to suggest that board size is not an independent corporate governance mechanism. Table 1 provides an overview of the related literature and summarizes the discussion.
III. EMPIRICAL APPROACH 1. Theoretical Considerations Although striking, we think that the results in the previous literature raise an important question. If boards are detrimental to rm value, why do we still see large boards? Or as Hermalin and Weisbach (2003) put it, why hasnt economic Darwinism eliminated this unt organizational form? 2 To address this question, it is important to understand two problems, which plague virtually all empirical work on boards of directors. First, most of the variables are endogenous, i.e., the econometrician faces the problem of joint endogeneity. A plausible alternative to the Jensen (1993) and Lipton and Lorsch (1992) hypothesis is that troubled rms expand their board in response to poor past performance in order to increase managerial capacity 3. Therefore, rm performance is both a result of the actions of previous directors and itself a factor that
2. See Hermalin and Weisbach (2003), p. 13. 3. However, an alternative hypothesis would be that rms attempt to wash out responsibilities over a larger number of directors.
332
28.5.04
22. Juli 2004 13:07

Table 1
Overview of previous literature on board size Sample Yermack (1996) Board Size Performance Measure Tobins Q Findings Signicant negative board size effect Smaller boards re CEOs more frequently CEO compensation less performance dependent if board size is large Large changes in board size (> 3) have signicant price impact Signicant cost increasing impact of board size Signicant negative board size effect Bad performance implies larger board changes Negative board size effect in all countries ROE: signicant negative board size effect in 3 countries Tobins Q: signicant negative board size effect in 2 countries
Panel of 452 4 to 34 large US rms Mean: 12.25 (1984 1991) Median: 12
Huther (1997) US electricity companies Eisenberg, Sundgren and Wells (1998) Conyon and Peck (1998) 785 small Finnish rms (1992 1994) 2886 rms from UK, 360 from France, 186 from Netherlands, 132 from Denmark, 126 from Italy (1990 1995)
3 to 16 Mean: 9 Mean: 3.7
Total variable costs ROA
Means: ROE, UK: 8.5 Tobins Q France: 10.5 Netherlands: 10.3 Denmark: 10.7 Italy: 11.8
Postma, 94 Dutch van Ees and rms (1996) Sterken (2003) Loderer and Peyer (2002) Panel of 169 Swiss rms
Mean 4.95 Median: 5
Market-tobook ratio
Negative board size effect
Means: betTobins Q ween 8.5 and 10.5 (depending on year) Medians: between 7 and 9 (depending on year)
Signicant negative board size effect Committee work has no impact of rm value; thus new interpretation large board is a sign of bad overall governance system
inuences the choice of subsequent directors. Second, as forcefully argued in Hermalin and Weisbach (2003), many empirical results can be interpreted as either equilibrium or out-of-equilibrium phenomena. Viewed as an out-of-equilibrium phenomenon, a negative relationship between board size and rm per333
28.5.04
22. Juli 2004 13:07
formance implies that limits on board size should be encouraged 4. In contrast, the equilibrium interpretation of this result is that some other factor is causing both board size and rm performance. In this case, any correlation between the two variables is purely spurious. An important policy implication, for example, is that regulation would be at best useless and possibly counterproductive. Or put differently, does board size really constitute an independent governance mechanism that is directly responsible for rms inefciencies and malfunctions? Alternatively, do larger board sizes simply identify rms that are not run as effectively as other rms? And in the latter case, what is the other factor driving the results? Sorting out the appropriate interpretation is an important task, which previous studies could not answer due to their limited focus on board size as the single governance mechanism. Specically, these studies only looked at simultaneous equations with board size and some measure of rm performance as the dependent variables (e.g., Eisenberg, Sundgren and Wells 1998). While this is an appropriate way to control for the possible endogeneity of board size and rm valuation, it does not account for the possibility that a rm can choose from a wide menu of different governance mechanisms. However, one may suspect that there exist important substitution effects between the different governance mechanisms. Given a menu of alternative governance mechanisms, the greater use of one mechanism needs not be positively related to rm valuation; where one mechanism is used more, others may be used less, resulting in the same valuation effects. Accordingly, the relationship between board size and rm valuation may possibly be driven by some other (so far unidentied) governance mechanism correlated with both board size and rm performance. In the terminology of Hermalin and Weisbach (2003), this would indicate an equilibrium situation.
2. Benets of Estimating a Simultaneous System of Equations Our empirical approach accounts for the endogeneity problem in a simultaneous system of equations and carefully chooses appropriate controlling variables to distinguish between the two types of interpretation. Specically, we analyze four corporate governance mechanisms: board size, outside representation on the board, ownership structure, and debt. In a rst step, we are interested in
4. This is particularly interesting in light of the result of Wu (2000), who reports that board sizes over the 19901995 period decreased in the US, among other things, due to pressure from active investors such as CalPERS.
334
28.5.04
22. Juli 2004 13:07
whether there are any interdependencies among these mechanisms. We include board size to examine whether it constitutes an independent governance mechanism, or whether it is simply related to the other mechanisms, but with no independent impact on other rm characteristics. With the exception of ownership structure, these are all mechanisms chosen internally by the rms decision makers. Therefore, as argued by Agrawal and Knoeber (1996), all of the costs and benets should be considered. Optimal choice requires that the use of a mechanism is increased until marginal benet just offsets marginal cost. In this case, a carefully specied cross-sectional regression should nd no relation between rm performance and the use of internally chosen mechanisms (i.e., board size, outside representation, and debt). This, in turn, does not imply that these mechanisms are ineffective. Conversely, for a governance mechanism chosen by outside parties, such as ownership structure, part of the costs or benets may be borne by parties other than those choosing the use of the particular mechanism. As a consequence, these choices need not maximize rm value. If a rm altered the use of one of its internally chosen mechanisms, this would likely lead to a change in managerial behavior and a change in the rms performance. If these governance mechanisms are chosen optimally, any crosssectional variation in their use should primarily reect differences in rms underlying environments, but not mistaken choices. If these differences are controlled for, then there should be no cross-sectional relation between the extent to which these mechanisms are used and rm valuation. In contrast, variation across rms in the use of external mechanisms may reect both differences in rms environment and non-value maximizing choices. Hence, regression analysis may reveal a cross-sectional relation between rm performance and the extent to which an external mechanism is used. Overall, therefore, our empirical analysis addresses three main questions. First, is board size an independent governance mechanism, or is it merely dependent on the three other mechanisms? In other words, one goal is to shed light on the possibly complex interrelationships between a set of governance mechanisms. We implicitly assume that different combinations of governance mechanisms can result in similar valuation effects. Second, do rms with smaller boards exhibit higher valuations? Taken together, the rst two questions are directly related to an interpretation of the results in previous empirical studies as equilibrium or out-of-equilibrium situations. Third, are the four governance mechanisms selected optimally? If this is indeed the case, any cross-sectional variation in their use reects differences in rms underlying environment, but not inefcient choices. In contrast, if these mechanisms are not selected optimally, changes in their use alleviate governance malfunctions and increase rm 335
28.5.04
22. Juli 2004 13:07
value. With respect to our particular interest in board size, the results from estimating a simultaneous system of equations can shed light on the hypothesis of the existence of an optimal board size, as put forth by Jensen (1993) and Lipton and Lorsch (1992).
3. Setting up a Simultaneous System of Equations To account for the problem of joint endogeneity, we test several equations separately as well as in a system of simultaneous equations 5. The four corporate governance mechanisms we explore empirically are board size (BOARDSIZE), outsider representation on the board (OUTSIDER), ownership structure (OWNERSHIP) and leverage (LEV) (see Table 2 for a description). Each governance mechanism appears on the left-hand side of one equation and the right-hand side of each of the others. Our rst regression equation contains BOARDSIZE as the dependent variable, where BOARDSIZE refers to the number of directors on the board of the company. Assuming that all relations are linear, we have 6: BOARDSIZEi = 0 + 1 OUTSIDERi + 2 LEVi + 3 OWNERSHIPi + 4 SIZEi + 5 GOVi + 6 ROAi + 7 INDi + i (1) Large rms will naturally have larger boards. Hence, we expect a positive relationship between BOARDSIZE and SIZE, where SIZE is measured as the natural logarithm of total sales. The dummy variable GOV is one if the state owns more than 5% of the rms equity, and zero otherwise. This variable accounts for the possibility that political inuences lead to presumably larger boards with a disproportionate number of government representatives. As hypothesized by Hermalin and Weisbach (1991) and Yermack (1996), small boards could contribute to better performance, or companies might adjust board size in response to past performance in order to increase managerial capacity. While causation is not clear ex ante, Yermacks (1996) results provide no evidence that boards either expand or contract in response to past performance. Nevertheless, to capture possible relationships between operating performance and board size, we also include current-year ROA. This variable is dened as operating in5. See Hermalin and Weisbach (2003) and Agrawal and Knoeber (1996) for a generalized description of this approach. 6. We assume linear relations to avoid econometric problems, which may occur in simultaneous equation systems with nonlinear (endogenous) variables (see Davidson and MacKinnon 1993, chapter 18).
336
28.5.04
22. Juli 2004 13:07
come over total assets, where the latter is a simple average of the year 2001 starting and ending values. Finally, to control for industry effects, we use an industry dummy, labeled as IND, as an exogenous variable. The industry classication is from Swiss Exchange (SWX). We measure the extent of outsider membership on the board as OUTSIDER, i.e., the percentage of board seats held by non-ofcers and members without relationship to the founding family (if any). This is the dependent variable in our second equation: OUTSIDERi = 0 + 1 BOARDSIZEi + 2 LEVi + 3 OWNERSHIPi + 4 CEOCOBi + 5 ROAi + 6 GOVi + 7 INDi + i (2) In addition to ROA and IND, we use a dummy variable called CEOCOB as a control variable that is one if the CEO is also Chairman of the Board (COB), and zero otherwise. Being CEO and COB at the same time alleviates coordination and communication problems. On the other hand, a major conict within the boardroom is between the CEO and the directors. The CEO has the incentive to capture the board, so as to ensure that he can keep his job and increase the other benets he derives from being CEO. Directors have the incentives to maintain their independence and monitor the CEO. Shivdasani and Yermack (1999) suggest that a situation where the CEO is also the COB leads to a concentration of power and the election of less independent board members. Accordingly, we expect a negative relationship between OUTSIDER and CEOCOB. Finally, to control for government ownership, we again include GOV as an exogenous explanatory variable. Ex ante, one would suspect that rms with large government stakes have fewer outsiders on the board. Our third internal governance mechanism, rm leverage, denoted as LEV is , the ratio of total (non-equity) liabilities to total assets. We use this governance mechanism as the dependent variable in our third regression equation: LEVi = 0 + 1 BOARDSIZEi + 2 OWNERSHIPi + 3 OUTSIDERi + 4 SIZEi + 5 AGEi + 6 GROWTHi + 7 INDi + i (3) The control variables are chosen following previous results. Recently, Drobetz and Fix (2003) report that Swiss rms with more growth opportunities have less leverage. Hence, we expect a negative relationship between LEV and GROWTH, where the latter is the average annual sales growth over the past three years. In contrast, their results indicate that larger and more mature rms exhibit higher leverage ratios, implying a positive relationship between LEV and both SIZE and AGE, where AGE is the number of years since inception as 337
28.5.04
22. Juli 2004 13:07
a private limited company. This is also consistent with Jensens (1986) free cash-ow hypothesis, where mature rms with substantial cash-ows use more debt to discipline managers. To control for industry effects, we again include the industry dummies, labeled as IND. Our only external governance mechanism is ownership structure. Specically, the variable OWNERSHIP denotes the percentage of cumulated voting rights exercised by large investors with > 5% of voting rights 7. For rms with unitary shares, this is equivalent to the percentage of actual stockholdings. The fourth regression specication is: OWNERSHIPi = 0 + 1 BOARDSIZEi + 2 LEVi + 3 OUTSIDERi + 4 SIZEi + 5 RESTRi + 6 GROWTHi + 7 INDi + i (4) We include SIZE as an exogenous variable, because one might suspect that larger rms have less ownership concentration. Voting restrictions could allow some shareholder groups (e.g., founding families) to practically dominate the rm even if they own less than 50% of the rms stock. This allows them to pursue their own interest at the expense of other shareholders with less board inuence. One may therefore suspect a positive relationship between OWNERSHIP and RESTR, where the latter variable is a dummy variable that is one if the rm has different share categories with different voting rights and, hence, deviates from the one-share-one-vote principle. Finally, Zeckhauser und Pound (1990) argue that the higher the R&D intensity, the more closed is the information structure, and the more difcult is outside monitoring. Large investors will recognize the problems associated with asymmetric information and, hence, ownership concentration should be lower for rms with a more diffuse information structure. We therefore use GROWTH as a proxy for asymmetric information (i.e., the difculty to monitor) and expect a negative relationship with OWNERSHIP. To summarize, each governance mechanism depends on all the others as specied in equations (1)(4). To estimate these relationships empirically, we adopt a simultaneous equations framework and apply the weighted two-stage least squares (W2SLS) procedure. This procedure also accounts for heteroscedasticity. There are eight exogenous variables in the four equations. Therefore, at least three of the exogenous variables must be excluded from any
7. Unfortunately, the Aktienfhrer Schweiz 2002/2003 (Swiss Stock Guide) by Finanz und Wirtschaft only includes voting rights. Therefore, we also use the variable RESTR (i.e., a dummy variable that is 1 if the rm has different share categories with different voting rights attached, and 0 otherwise) as a controlling variable for OWNERSHIP. In our sample, 24% of the rms have this form of voting restrictions.
338
28.5.04
22. Juli 2004 13:07

Table 2
Summary of variables Endogenous variables Tobins Q Ratio of market value to book value of assets. Market value of assets is computed as market value of equity plus book value of assets minus book value of equity Number of directors on the board of the company Outsider membership on the board, measured by the percentage of board seats held by non-ofcers without relationship to the founding family (if any) Leverage, measured as the ratio of total (non-equity) liabilities to total assets Percentage of cumulated voting rights exercised by large investors with > 5 % of voting rights
BOARDSIZE OUTSIDER LEV OWNERSHIP
Exogenous variables SIZE GOV ROA CAR AGE RESTR CEOCOB GROWTH IND Firm size, measured by the natural logarithm of sales Dummy variable = 1 if state owns > 5 % of the rms equity, = 0 if otherwise Ratio of operating income and total assets (return on assets) Cumulated risk-adjusted abnormal return over the last 24 months (using weekly data) Number of years since inception as private limited company Dummy variable = 1 if the rm has different share categories with different voting rights attached, = 0 if otherwise Dummy variable = 1 if the CEO is also COB, = 0 if otherwise Average annual growth of sales over the past three years (1999 2001) Dummy variable = 1 if rm belongs to a particular industry (15 sectors as classied by SWX), = 0 if otherwise
single equation to identify the system. The development of equations (1)(4) is motivated partially by the need for these exclusion restrictions to be met, but most of them seem non-controversial. For example, AGE is likely to be related to leverage, but it is unlikely to be related to all other governance mechanisms. Similarly, CEOCOB should only be important for the OUTSIDER equation. Excluding ROA from the LEV equation seems somewhat more problematic. The ndings in Drobetz and Fix (2003) indicate that ROA is an important variable to explain cross-sectional differences in leverage of Swiss rms. However, this exclusion is driven by the requirement that any single equation of the system must be identied 8.
8. Note that in the system of equations (1) (4) we could add ROA as an additional variable, still leaving the system identied. However, when we expand the system and add the Tobins Q re-
339
28.5.04
22. Juli 2004 13:07
In our nal regression equation, in addition to the interrelations among the governance mechanisms, we also examine the cross-sectional relationship between these mechanisms and rm valuation, measured by Tobins Q 9. Tobins Q is dened as the ratio of market value to replacement value. As an approximation, the market value of assets is usually computed as market value of equity plus book value of assets minus book value of equity. The replacement value is taken as the book value of assets. A Tobins Q ratio greater than 1 indicates that the rm has done well with its investment decisions, i.e., it has invested in positive net present value projects. In contrast, a value of Tobins Q lower than 1 indicates that the company did not even earn its rm-wide cost of capital. The empirical specication is as follows: Qi = 0 + 1 BOARDSIZEi + 2 LEVi + 3 OUTSIDERi + 4 OWNERSHIPi + 5 SIZEi + 6 GROWTHi + 7 ROAi + 8 INDi + i (5) In addition to the four governance mechanisms, we include controls for other variables that might affect Tobins Q directly. Following Morck, Shleifer and Vishny (1988), we control for growth opportunities and expect a positive relationship between Tobins Q and GROWTH. We also control for SIZE, since growth opportunities should be lower for larger rms. On the basis of simple valuation models, ROA seems a purely analytical control variable; one expects a positive relationship. We rst estimate equation (5) using ordinary least square. This allows us to examine the effect of all governance mechanisms together, but treats each of them as exogenous. If the previous analysis showed that all four governance mechanisms were independent, this would be an adequate and sufcient procedure. Nevertheless, in a second step we include equation (5) along with (1)(4) in a simultaneous system of equations, adding Tobins Q as an independent variable, and again apply weighted two-stage least square (W2SLS) to estimate the system. This treats Tobins Q as endogenous along with the governance mechanisms, allowing each of the mechanisms to affect Tobins Q but also allowing Tobins Q to affect the choice of each mechanism. Finally, note that this system is just identied equation-by-equation.
gression in equation (5), we must exclude at least four exogenous variables from each equation to identify the system. In untabulated tests we nd that using ROA in equation (3) instead of AGE (which we think could most easily be replaced) does not qualitatively change our results. 9. Alternatively, Eisenberg, Sundgren and Wells (1998) and Conyon and Peck (1998) use return of assets (ROA) and return on equity (ROE), respectively.
340
28.5.04
22. Juli 2004 13:07
IV. DATA DESCRIPTION We potentially target all 267 rms contained in the Swiss Performance Index. We exclude all banks, insurances, and other nancial services rms. This is standard in the empirical corporate nance literature due to problems in computing appropriate valuation measures. After inspecting the collected data, we drop another six rms, which we regard as obvious outliers 10. This leaves us with a nal sample of 165 publicly listed rms. Our primary sources of data are Datastream and Worldscope. Several data had to be collected manually from Aktienfhrer Schweiz 2002/2003 and the website of Finanz und Wirtschaft. Most of the data are as of end 2001, i.e., they refer to the reporting period from January 2001 to December 2001. Table 3 shows descriptive statistics of Tobins Q and the four governance mechanisms. Following Loderer and Peyer (2002), we dene Tobins Q as the ratio of market value to book value of assets. Market value of assets is computed as market value of equity plus book value of assets minus book value of equity. To avoid that daily uctuations inuence our results, we compute the market value of equity as the mean of daily observations during 2001. The average value of Tobins Q is 1.50, the median is 1.24. This indicates that Swiss rms, on average, invest in positive NPV projects. The average board size in our Swiss sample is 6.59. This is only half of the size of boards studied in Yermack (1996) for the US, but signicantly larger than those in Eisenberg, Sundgren and Wells (1998) for a Finnish sample. It is also important to note that the average board size of Swiss rms has decreased over time. Using data from all rms traded on the Swiss stock exchange in 1980, 1985, 1990 and 1995, Loderer and Peyer (2002) report that average board size declined from 10.5 in 1980 to 8.5 in 1995. Since then, board size further decreased to an average of 6.6, which is just the board size hypothesized to be optimal by Lipton and Lorsch (1992) and Jensen (1993). However, as one could have expected, there are pronounced differences between rms with and without government inuence. Specically, there are seven rms in our sample where some state authority owns more than 5% of the rms equity. On average, board size for these rms is 11.71. The difference in board size is statistically signicant on the basis of a simple t-test (with a tvalue of 6.69).
10. Excluded rms have values of Tobins Q above 6.
341
28.5.04
22. Juli 2004 13:07

Table 3
Descriptive statistics Variable Tobins Q BOARDSIZE BOARDSIZE if GOV = 1 OUTSIDER LEV OWNERSHIP Mean 1.50 6.59 11.71 0.87 0.55 0.47 Median 1.24 6.00 11.00 0.89 0.57 0.50 S. D. 0.86 2.33 3.64 0.15 0.18 0.26 Maximum/Minimum 5.70/0.40 19/3 19/8 1/0.2 0.99/0.043 1/0
Notes: The table contains the descriptive statistics of 165 Swiss rms listed on Swiss Exchange (SWX), excluding banks, insurance and nancial services companies. Tobins Q is the ratio of market value to book value of assets. Market value of assets is computed as market value of equity plus book value of assets minus book value of equity. BOARDSIZE is the number of directors on the board of the company. GOV is a dummy variable that is one if the state owns > 5 % of a rms equity, and zero otherwise. OUTSIDER is measured by the percentage of board seats held by non-ofcers without relationship to the founding family (if any). LEV is computed as the ratio of total (non-equity) liabilities to total assets. OWNERSHIP is the percentage of cumulated voting rights exercised by large investors with > 5% of voting rights. S. D. is the abbreviation for standard deviation.
Lipton and Lorsch (1992) and Jensen (1993) argue that for board sizes beyond seven or eight the benets of increasing monitoring capacities are outweighed by costs such as slower decision-making, less-candid discussions of managerial performance, and biases against risk-taking. Roughly 85% of all rms in the full sample have a board size equal to or less than eight. Figure 1 shows the mean and median values of Tobins Q for companies sorted by board size. Interestingly, there is no obvious relationship between the two variables, i.e., in a univariate analysis larger board sizes are not associated with lower valuations. This is in sharp contrast to the results in Yermack (1996) for US data. Plotting board size against average Tobins Q as in Figure 1, he nds that Tobins Q values decline almost monotonically over the range of board size 11. For board sizes below six, however, he also reports no consistent association between board size and rm value. In contrast, Eisenberg, Sundgren and Wells (1998) present a gure with a negative relationship between the two variables for their sample of Finnish rms, even though average board size is only 3.7. Nevertheless, as argued above, the relationship depicted in Figure 1 does not drive our empirical results; it is necessary to account for possible interrelationships between different governance mechanisms.
11. See Yermack (1996), Figure 1, p. 193.
342
28.5.04
22. Juli 2004 13:07

Figure 1
Tobins Q and board size
Table 3 further shows that the average value of OUTSIDER is 0.87, which is also in contrast to the results in Yermack (1996). He reports a much lower value of 0.54 for US rms. This is a surprising result for two reasons. First, similar to the United States (but in contrast to Germany), Switzerland has a one-tier board system. And second, founding families are still regarded as a non-negligible factor in corporate Switzerland. The data for the variable OUTSIDER is taken from the website of Finanz und Wirtschaft 12. The average leverage ratio (LEV) is 55%, which closely corresponds to the gures recently reported in Drobetz and Fix (2003). The average of OWNERSHIP is 0.47. The maximum of 1 for this variable is due to Hilti, where all registered shares with voting rights are in the hands of the founding family, while all publicly traded participation notes have no voting rights. Finally, Table 4 shows the correlation coefcients between Tobins Q and the four governance mechanisms. All correlations are relatively small. The only noteworthy correlation is that between BOARDSIZE and OUTSIDER. The re12. See http://www.nanzinfo.ch.
343
28.5.04
22. Juli 2004 13:07
spective value is 0.275, which is slightly higher than in Yermack (1996). This is an important result, because it indicates that part of the board size effect reported in earlier studies may relate to the composition of the board. Note that we account for this possibility in our empirical setup, having both OUTSIDER and BOARDSIZE as explanatory variables in equations (1) and (2), respectively.
Table 4
Correlation matrix between governance mechanisms BOARDSIZE Tobins Q BOARDSIZE LEV OWNERSHIP 0.018 LEV 0.261 0.075 OWNERSHIP 0.207 0.012 0.020 OUTSIDER 0.072 0.275 0.005 0.100
Notes: The table shows the correlation coefcients between Tobins Q and the four governance mechanisms: BOARDSIZE, LEV OWNERSHIP, and OUTSIDER. Tobins Q is the ratio of market , value to book value of assets. Market value of assets is computed as market value of equity plus book value of assets minus book value of equity. BOARDSIZE is the number of directors on the board of the company. OUTSIDER is measured by the percentage of board seats held by non-ofcers without relationship to the founding family (if any). LEV is computed as the ratio of total (nonequity) liabilities to total assets. OWNERSHIP is the percentage of cumulated voting rights exercised by large investors with > 5% of voting rights. The sample contains 165 Swiss rms listed on Swiss Exchange (SWX), excluding banks, insurance and nancial services companies.
Our controlling variables are gathered from different sources. SIZE is measured as the natural logarithm of sales in 2001 13. We dene GROWTH as the average annual sales growth over the past three years (19992001). For rms which went public after 1999, we use average sales growth since IPO 14. ROA is dened as the ratio of operating income to total assets, where operating income is measured as of end 2001, but total assets are a simple average of the respective starting and ending values. The corresponding list of rms with voting restrictions for the variable RESTR can directly be found in the Aktienfhrer Schweiz 2002/2003 15. The data for the dummy variable CEOCOB is from the website of Finanz und Wirtschaft. Finally, we apply the industry classication, as provided
13. We also experimented with the logarithm of total (book) assets as of end 2001, but the results are qualitatively similar. 14. We also excluded particular rm-years for Allreal, Bon Apptit, Sulzer und Zblin due to rmspecic events, such as acquisitions and spin-offs. 15. See Aktienfhrer Schweiz 2002/2003, p. 468.
344
28.5.04
22. Juli 2004 13:07
by Swiss Exchange (SWX) 16. For each of the 15 industries we construct a dummy variable, denoted as IND.
V. EMPIRICAL RESULTS This section describes our empirical results. In section V.1 we examine the interdependencies among the four governance mechanisms. In section V.2 we extend the simultaneous system of equations and analyze the relationship between rm valuation and the different governance mechanisms. Section V.3 contains several robustness tests.
1. Relationships Between the Governance Mechanisms To examine the relationship between the four governance mechanisms, we estimate (1)(4) as a system of linear equations using weighted two-stage least square (2SLS). Each governance mechanism appears on the left-hand side of one equation and the right-hand side of each of the others. We are primarily interested in the null hypothesis that board size is not an independent governance mechanism. If this is true, any ndings that board size is negatively related to rm valuation may be spurious, and some other factor determines both board size and rm valuation. The null hypothesis must be rejected if the coefcients on the other three governance mechanisms in the BOARDSIZE equation are insignicant. The results of the weighted 2SLS estimation are displayed in Table 5. The coefcients of the exogenous variables in the lower part of Table 5 generally have the predicted sign, but they are often statistically insignicant. As hypothesized, looking at board size (equation (1)), rms with government ownership have larger boards. On average, a rm where the state owns a signicant portion of stock has four additional board members, all else equal. The coefcient on GOV is signicant at the 1% level. The coefcients on SIZE and ROA are positive and negative, respectively, but both are insignicant. With respect to board composition (equation (2)), we nd evidence for a potential conict within the boardroom. There is a negative relationship between OUTSIDER and CEOCOB, indicating that outsiders seek independence from the CEO and prohibit the CEO to take the role as COB at the same time. Looking at leverage (equation (3)), the signs of the coefcients on SIZE and AGE are as predicted
16. See http://www.swx.com.
345
28.5.04
22. Juli 2004 13:07
ex ante. Large rms carry higher debt levels, and the coefcient on SIZE is signicant at the 1% level. Older rms (measured by the variable AGE) are also more levered, but the relationship is not signicant. Surprisingly, rms with more growth opportunities exhibit higher leverage, but the coefcient on GROWTH is again insignicant. Finally, with respect to ownership structure (equation (4)), the only signicant exogenous variable is RESTR, indicating that a rm has different share categories with different voting rights attached. As hypothesized, larger rms have less ownership concentration, although the relationship is insignicant. We now look at the coefcients for the endogenous variables in the upper part of Table 5. Evidently, there are no interdependencies between any of the four governance mechanisms; the respective coefcients are all insignicant. Therefore, we conclude that the null hypothesis must be rejected. Board size should be regarded as an independent governance mechanism, and potential correlations with rm valuation are not merely spurious. Most important, any potential board size effect is not (solely) related to board composition. Of course, with no interdependencies at all, the other three mechanisms board composition, ownership structure, and leverage also constitute independent governance mechanisms. However, given the results from previous studies, this should come as no surprise. Finally, a Wald test for the simultaneous signicance of all the coefcients (except the constant and the industry dummies) always rejects the null hypothesis that they are jointly zero, except for the OUTSIDER equation.
2. Firm Valuation and Governance Mechanisms In this section we examine the null hypothesis that board size has no inuence on rm performance against the alternative hypothesis that reductions in board size increase rm performance. For this purpose, we estimate the extended system of simultaneous equations (1)(5). The null hypothesis must be rejected if the coefcient on BOARDSIZE in the regression containing Tobins Q as the dependent variable (in equation (5)) is signicant. A nal (and closely related) question is whether the four governance mechanisms are selected optimally. As hypothesized, this is the case if the marginal benet of a mechanism just offsets marginal cost. This hypothesis can be tested in the same system of equations (1)(5). Specically, it must be rejected if a signicant relationship between rm valuation and the use of internally chosen mechanisms is uncovered. In contrast, insignicant coefcients on the respective variables in the regression involving Tobins Q as the dependent variable (in equation (5)) are a sign of op346
28.5.04
22. Juli 2004 13:07

Table 5
Results from weighted 2SLS regressions of governance mechanisms Dependent variable (N = 165) Independent variable Constant BOARDSIZE OUTSIDER LEV OWNERSHIP SIZE GOV AGE GROWTH RESTR CEOCOB ROA Industry Wald test 0.116 0.9596 included 56.72 (6) 0.0000 0.109** 0.0215 0.042 0.8614 included 8.49 (6) 0.2042 included 33.49 (6) 0.0000 included 33.04 (6) 0.0000 0.794 0.8187 5.170 0.4529 0.075 0.9557 0.216 0.4307 4.284*** 0.0000 0.348 0.5316 0.0003 0.1419 0.033 0.6582 0.053 0.7331 0.309*** 0.0000 0.641 0.6468 0.053 0.7176 0.121 0.2043 0.027*** 0.0089 0.0005 0.9879 BOARDSIZE 0.151 0.9663 OUTSIDER 0.627** 0.0361 0.092 0.4851 LEV 0.131 0.6601 0.016 0.3113 0.050 0.8734 OWNERSHIP 0.329 0.5897 0.020 0.5885 0.698 0.3253 1.343 0.1006
Notes: The table shows the results from a weighted two stage least square (W2SLS) estimation of equations (1) (4) of the system of linear equations described in section III.3. The dependent variable in column (1) is BOARDSIZE, which is the number of directors on the board of the company. The dependent variable in column (2) is OUTSIDER, which is measured by the percentage of board seats held by non-ofcers without relationship to the founding family (if any). The dependent variable in column (3) is LEV which is computed as the ratio of total (non-equity) liabilities to total , assets. The dependent variable in column (4) is OWNERSHIP, which is the percentage of cumulated voting rights exercised by large investors with > 5% of voting rights. The description of the exogenous variables can be found in Table 2. The sample contains 165 Swiss rms listed on Swiss Exchange (SWX), excluding banks, insurance and nancial services companies. A Wald test is performed for the simultaneous signicance of all coefcients (except the constant and the industry dummies). The numbers beneath the coefcients are probability values for two-sided tests. The symbols ***, **, * denote statistical signicance at the 1%, 5 %, 10 % level, respectively.
347
28.5.04
22. Juli 2004 13:07
timal choice. An exception is ownership structure, which constitutes an externally chosen mechanism. In this case, variation across rms may reect both differences in rms environment and non-value maximizing choices. Hence, even a carefully specied regression may reveal a cross-sectional relation between rm valuation and the extent to which an external mechanism is used. Table 6 shows the empirical results. In the rst column of Table 6 we test equation (5) using simple OLS, with Tobins Q as the dependent variable. Given that all four mechanisms are unrelated and independent, we can treat them as exogenous in this simplest specication. Note that this setup does not allow for any interdependence in the choices of governance mechanisms, but it does allow for the availability of alternative governance mechanisms. To control for a possible nonlinear relationship between board size and rm valuation, we add a quadratic term (BOARDSIZE^2) in the second column. As one may expect, the respective coefcient is negative, but insignicant. In the third column we also add an interaction term, OWNERSHIP*RESTR, to account for the strong relationship between the two variables. In all three equations BOARDSIZE has a negative coefcient, but they are far from being statistically signicant. Therefore, we cannot reject the null hypothesis that board size has no inuence on Tobins Q. This conrms the initial intuition from a mere visual inspection of Figure 1, where board size is plotted against average Tobins Q. Because OWNERSHIP constitutes an external mechanism, it should not be surprising that the corresponding coefcients in columns (1) and (2) are significant at the 5% level. What is surprising, however, is the negative sign of the coefcient, i.e., a more dispersed ownership leads to a higher Tobins Q. When we include the interaction term OWNERSHIP*RESTR in column (3), it becomes evident that not ownership concentration itself drives the results, but instead the existence of shares with different voting rights. While the coefcient on OWNERSHIP becomes insignicant, the coefcient on the interaction term is negative and signicant at the 1% level. This result has an important practical implication. It clearly indicates that the interests of dominating shareholders (possibly in combination with some voting restrictions) and other shareholders are not always in line (see Morck, Shleifer and Vishny 1988). In fact, many Swiss rms have abandoned their common dual class shares and introduced unitary shares only in recent years (see Kunz 2002). This seems to be an important step towards overall rm value maximization. In column (4) we include both board size squared as well as the interaction term, but again only the latter is estimated signicantly. The regression setup in columns (1)(4) assumes that protability is held constant. Accordingly, we test a direct relationship between board size and rm 348
28.5.04
22. Juli 2004 13:07

Table 6 Results from OLS and W2SLS regressions using Tobins Q
Dependent variable: Tobins Q (N = 165) (2) (3) (4) (5) (6) (7) OLS OLS OLS OLS OLS W2SLS 1.181 0.767 0.721 0.569 0.527 1.187 0.1171 0.3037 0.3420 0.4550 0.4896 0.4639 BOARDSIZE 0.004 0.016 0.013 0.026 0.002 0.036 0.9646 0.4576 0.8739 0.2639 0.9770 0.6705 BOARDSIZE^2 0.001 0.002 0.002 0.7838 0.6664 0.6816 LEV 0.371 0.372 0.477 0.479 0.609 0.609 2.732 0.3390 0.3389 0.2164 0.2150 0.1375 0.1375 0.3634 OWNERSHIP 0.449** 0.449** 0.039 0.036 0.310 0.310 1.250** 0.0451 0.0458 0.9005 0.9070 0.1885 0.1886 0.0184 OWNERSHIP*RESTR 0.570*** 0.573*** 0.0058 0.0058 OUTSIDER 0.228 0.233 0.334 0.342 0.256 0.263 0.122 0.5038 0.5002 0.3373 0.3286 0.4552 0.4491 0.9360 SIZE 0.038 0.036 0.064* 0.062 0.081** 0.078* 0.118 0.3215 0.3633 0.0991 0.1231 0.0446 0.0659 0.2810 GROWTH 0.555 0.554 0.575 0.574 0.675 0.674 0.528 0.1688 0.1705 0.1400 0.1416 0.1199 0.1214 0.1687 ROA 2.046*** 2.042*** 1.957*** 1.951*** 1.702* 0.0034 0.0035 0.0028 0.0029 0.0733 Industry included included included included included included included Wald test 20.62 (7) 21.54 (8) 34.29 (8) 34.53 (9) 12.18 (6) 13.99 (7) 25.86 (7) 0.0044 0.0059 0.0000 0.0001 0.0582 0.0513 0.0005 Adjusted R2 0.459 0.456 0.482 0.479 0.412 0.408 Notes: The table shows the results from OLS regressions of Tobins Q on all governance mechanisms along with different combinations of the exogenous control variables included in equation (5), as described in section III.3. Tobins Q is the ratio of market value to book value of assets. Market value of assets is computed as market value of equity plus book value of assets minus book value of equity. The nal column (7) contains the results from a weighted two stage least squares (W2SLS) estimation of the complete system, as described in equations (1) (5). Only the coefcient estimates for equation (5) are reported. BOARDSIZE is the number of directors on the board of the company. OUTSIDER is measured by the percentage of board seats held by non-ofcers without relationship to the founding family (if any). LEV is computed as the ratio of total (non-equity) liabilities to total assets. OWNERSHIP is the percentage of cumulated voting rights exercised by large investors with > 5% of voting rights. SIZE denotes the natural logarithm of sales, GROWTH the average annual growth of sales over the past three years (1999 2001), and ROA the ratio of operating income and total assets (return on assets). RESTR is a dummy variable equal to 1 if the rm has different share categories with different voting rights attached, and 0 otherwise. The sample contains 165 Swiss rms listed on Swiss Exchange (SWX), excluding banks, insurance and nancial services companies. A Wald test is performed for the simultaneous signicance of all coefcients (except the constant and the industry dummies). The numbers beneath the coefcients are probability values for two-sided tests. The symbols ***, **, * denote statistical signicance at the 1%, 5 %, 10 % level, respectively. Independent variable Constant (1) OLS 1.210 0.1046 0.015 0.5086
349
28.5.04
22. Juli 2004 13:07
valuation, independent of a rms protability. This is the implicit notion that underlies the Jensen (1993) and Lipton and Lorsch (1992) argument of coordination problems and director free-riding, among others, with increasing boards. However, the analysis may ignore a potentially important indirect relationship between board size and rm valuation. Specically, smaller boards could affect protability, which then in turn leads to higher valuations 17. To allow for this indirect relationship between board size and rm valuation through increased protability, we exclude ROA as a control variable. The results are shown in column (5). Interestingly, the coefcient of board size changes only marginally and remains statistically insignicant. Even when we add the quadratic BOARDSIZE^2 variable in column (6), the results are unchanged. To further explore the possibility of an indirect effect between board size and Tobins Q, one may also apply ROA as the dependent valuation measure and explore whether board size affects protability. We postpone this analysis until section V.3. Finally, in column (7) we treat all four governance mechanisms as endogenous and allow for possible interrelations. The system of equations (1)(5) is again estimated simultaneously using weighted 2SLS. For the sake of brevity, we only report the results from equation (5), with Tobins Q as the dependent variable 18. The coefcient on BOARDSIZE is negative, but again insignicant. Similarly, the coefcients on the other internal governance mechanisms, LEV and OUTSIDER, are also insignicant. This is in contrast to Agrawal and Knoeber (1996), who report a persistent effect of board composition on rm performance. Our results are consistent with optimal choice of all internal mechanisms. Hence, the cross-sectional variation in the use of these mechanisms to a large extent reects differences in rms underlying environments, rather than mistaken choices, which need to be reversed. This also implies that any changes in their use leave rm valuation unaffected at best, but possibly lead to a decrease in Tobins Q. But in this case, it is not because some other factor is responsible for a spurious correlation between any of the internal mechanisms (e.g., board size) and rm valuation, but it is because the mechanisms are already in optimal use. One obvious explanation for this result is that average board size in our sample (6.59) is just below the critical value of 78, as postulated by Lipton and Lorsch (1992) and Jensen (1993). Accordingly, on average, the number of board members in our sample of Swiss rms is just about optimal, i.e., the advantages of larger boards from increased management ca17. The problem of fragile empirical linkages, where small changes in the conditioning information set can lead to extremely different policy implications, was discussed by Levine and Renelt (1992) in their study of the determinants of long-run macroeconomic growth. 18. The coefcients on Tobins Q in equations (1) (4) are always insignicant.
350
28.5.04
22. Juli 2004 13:07
pacities are offset by the potential disadvantages from coordination, communication, and decision-making problems. Taken together, our results in Table 5 and 6 suggest that board size is an independent governance mechanism, and rms choose board size just optimal. This interpretation is consistent with Harts (1995) notion that boards are part of the equilibrium (but still secondbest) solution to the contracting problem in rms with dispersed shareholders. The nal column in Table 6 also reveals a signicant negative relationship between OWNERSHIP and Tobins Q, i.e., more cumulated voting rights exercised by large investors lead to lower rm valuations. This is inconsistent with the notion in Stiglitz (1985) and Shleifer and Vishny (1986) that a natural way to improve corporate governance is to ensure that a company has one or more large shareholder. Empirically, however, Morck, Shleifer and Vishny (1988) show evidence that large shareholders have a mixed role; there is a non-monotonic relationship between Tobins Q and the fraction of company stock owned by insiders 19. A possible explanation for the signicant negative relationship in our Swiss sample could be that large shareholders use their voting power to improve their own position at the expense of other shareholders. Recall, however, that we presumed that OWNERSHIP was an external governance mechanism. The crosssectional differences in OWNERSHIP and the negative relationship with Tobins Q may at least partly reect non-value maximizing choices, which are out of direct control by rms decision makers. Finally, as expected, current level of profitability, measured by return on assets, has a positive association with Tobins Q. The explanatory powers of our cross-sectional regressions are reasonably high, the adjusted R-squares range between 0.456 and 0.482. Again, Wald tests for the null hypothesis that all coefcients are jointly zero (except the constant and the industry dummies) strongly reject. 3. Robustness Tests To make sure our results are accurate, we conduct several robustness tests. First, in untabulated regressions we reproduced all our results using three-stage least square (3SLS), but they are qualitatively the same. Because the expanded system of equations (1)(5) is just identied equation-by-equation, the coefcient estimates in weighted 2SLS and 3SLS are identical in this case. Second,
19. Note that the relationship between OWNERSHIP and Tobins Q is supposedly highly non-linear. For low and moderate OWNERSHIP levels, incentive-alignment effects result in a positive relationship. But as managerial ownership increases, entrenchment-related costs will dominate and produce a negative relationship between OWNERSHIP and Tobins Q. Obviously, our 5 % criteria with regards to voting rights exercised by large investors is above the threshold.
351
28.5.04
22. Juli 2004 13:07
instead of the return on assets (ROA) we used cumulated risk-adjusted returns (CAR) over the 24 months period from 2000.01 to 2001.12 (using weekly returns) as a controlling variable, but again the results did not change. Third, we used alternative measures of rm valuation. In the spirit of Eisenberg, Sundgren and Wells (1998), we compute an industry-adjusted Tobins Q. This variable is dened as the residual between a rms and the corresponding industrys median Tobins Q. Alternatively, we also use the ratio of market-tobook equity as a measure of rm value. Our results are robust in both cases, and the interpretations do not change. Finally, we use ROA as the dependent variable. Given the discussion about direct and indirect relationships between board size and rm valuation in section V.2, this is a particularly interesting robustness check. Smaller boards may affect protability, which then in turn leads to higher valuations. To further explore this possibility of an indirect relationship, we apply ROA as our nal measure of rm valuation and estimate the impact of board size. Table 7 reveals that our results are still robust, i.e., the coefcients on BOARDSIZE and BOARDSIZE^2 remain insignicant in all specications 20. We conclude that there is no unobserved indirect effect between board size and rm valuation through increased protability that drives our main results. The only noteworthy changes in results are that the coefcient on OWNERSHIP is now positive and the coefcient on LEV becomes signicant. The negative sign of the coefcient on LEV is consistent with Myers (1984) pecking order theory of the capital structure, where more protable rms use internal sources to realize value-enhancing projects and can avoid approaching the capital market. To make sure causality runs from board size to rm valuation, Yermack (1996) and Eisenberg, Sundgren and Wells (1998) test whether companies expand board size in response to bad performance. They both nd that poor performance leads to both more departures of board members and more appointments to the boards, but net board size remains constant. While we think that our approach appropriately controls for endogeneity problems, we test for this possibility in a very simplistic way by including lagged values of the return on assets in the BOARDSIZE equation (equation (1)). We use lags of one and two years, but in both cases the coefcients are insignicant. To double-check the validity of the results in Table 6, we replicate the OLS regressions with Tobins Q as the dependent variable separately for each governance mechanism. We use the same exogenous variables (SIZE, GROWTH, ROA and IND) in each of the four regressions. As before, only the coefcient
20. Because our initial system in Table 6 was just identied, we cannot provide W2SLS estimates in Table 7 without an additional exogenous variable.
352
28.5.04
22. Juli 2004 13:07

Table 7
Results from OLS regressions using ROA Dependent variable: ROA (N = 165) Independent variable Constant BOARDSIZE BOARDSIZE^2 LEV OWNERSHIP OWNERSHIP*RESTR OUTSIDER SIZE GROWTH Industry Wald test Adjusted R2 0.014 0.8617 0.021*** 0.0024 0.059 0.3212 included 18.17 (6) 0.0058 0.215 0.015 0.8509 0.021*** 0.0055 0.059 0.3258 included 21.85 (7) 0.0027 0.210 0.116* 0.0507 0.068** 0.0318 (1) OLS 0.313** 0.0385 0.005 0.1516 (2) OLS 0.320** 0.0317 0.001 0.9615 0.0003 0.6189 0.116* 0.0513 0.068** 0.0320 0.120** 0.0481 0.085* 0.0663 0.023 0.4362 0.018 0.8180 0.022*** 0.0030 0.060 0.3168 included 19.89 (7) 0.0058 0.213 (3) OLS 0.330** 0.0354 0.005 0.1552 (4) OLS 0.338** 0.0289 0.0002 0.9861 0.0003 0.5909 0.120** 0.0485 0.085* 0.0658 0.024 0.4270 0.019 0.8060 0.022*** 0.0061 0.060 0.3216 included 23.75 (8) 0.0025 0.208
Notes: The table shows the results from OLS regressions of ROA on all governance mechanisms along with different combinations of exogenous control variables. The dependent variable in all four columns is ROA, which is the ratio of operating income to total assets (return on assets). BOARDSIZE is the number of directors on the board of the company. OUTSIDER is measured by the percentage of board seats held by non-ofcers without relationship to the founding family (if any). LEV is computed as the ratio of total (non-equity) liabilities to total assets. OWNERSHIP is the percentage of cumulated voting rights exercised by large investors with > 5% of voting rights. SIZE denotes the natural logarithm of sales, and GROWTH is the average annual growth of sales over the past three years (1999 2001). RESTR is a dummy variable equal to 1 if the rm has different share categories with different voting rights attached, and 0 otherwise. The sample contains 165 Swiss rms listed on Swiss Exchange (SWX), excluding banks insurance and nancial services companies. A Wald test is performed for the simultaneous signicance of all coefcients (except the constant and the industry dummies). The numbers beneath the coefcients are probability values for two-sided tests. The symbols ***, **, * denote statistical signicance at the 1%, 5 %, 10 % level, respectively.
353
28.5.04
22. Juli 2004 13:07
on OWNERSHIP is signicant. BOARDSIZE, LEV and OUTSIDER are estimated insignicantly in the corresponding regressions, respectively.
VI. CONCLUSION Boards of directors are an economic institution, which, in theory, helps to solve the agency problems inherent in public corporations. While there is only limited theoretical work, the empirical literature is well developed. Most previous studies have focused on specic board characteristics, such as board composition and board dynamics, and their relationship with rm valuation. Surprisingly, there has been relatively little work on the relationship between the size of boards and rm valuation. We contribute to this literature by exploring whether board size constitutes an independent governance mechanism and is related to Tobins Q. This is an interesting research question with potentially important practical implications. Nevertheless, by focusing on board size as the only governance mechanisms in isolation and ignoring that rms can choose from a wide menu of different governance mechanisms with possibly similar valuation effects, previous studies with mainly US evidence are clearly limited. Using a comprehensive set of companies listed at Swiss Exchange (SWX), our results support the notion that board size is indeed an independent governance mechanism. In particular, board size is independent from board composition. However, in contrast to previous studies we do not nd a signicant relationship between board size and rm valuation, as measured by Tobins Q. This suggests that Swiss rms, on average, choose their number of board members just optimally. In a simultaneous system of equations we carefully control for the endogeneity problems inherent in empirical studies on governance. Also in contrast to most of the literature, this empirical setup allows us to distinguish between equilibrium and out-of-equilibrium interpretations of estimation results. Cross-sectional variations in board size to a large extent reect differences in rms underlying environment, and not mistaken choices, which should be reversed. Accordingly, even though we nd that board size constitutes an independent corporate governance mechanism, any changes leave rm valuation unaffected at best, but more probably lead to a decrease in Tobins Q. This is consistent with the observation that the average board size in our sample of Swiss rms (6.59) is just below the optimal value of 78, as hypothesized by Lipton and Lorsch (1992) and Jensen (1993). The advantages of larger boards from increased management capacities may just offset the potential disadvantages from coordination, communication, and decision-making problems. 354
28.5.04
22. Juli 2004 13:07

REFERENCES
Agrawal, Anup and Charles R. Knoeber (1996). Firm Performance and Mechanisms to Control Agency Problems between Managers and Shareholders, Journal of Financial and Quantitative Analysis. 31: 377 397. Bhaghat, Sanjai and Bernard Black (1999). The Uncertain Relationship Between Board Composition and Firm Performance, Business Lawyer. 54: 921 963. Conyon, Martin and Simon I. Peck (1998). Board Size and Corporate Performance: Evidence from European Countries, European Journal of Finance. 4: 291 304. Davidson, Russell and James G. MacKinnon (1993). Estimation and Inference in Econometrics. Oxford: Oxford University Press. Drobetz, Wolfgang and Roger Fix (2003). What are the Determinants of the Capital Structure? Some Evidence for Switzerland, Working Paper 4-03, University of Basel. Eisenberg, Theodore, Stefan Sundgren and Martin T. Wells (1998). Larger Board Size and Decreasing Firm Value in Small Firms, Journal of Financial Economics. 48: 35 54. Hart, Oliver (1995). Corporate Governance: Some Theory and Implications, The Economic Journal. 105: 678 689. Hermalin, Benjamin E. and Michael S. Weisbach (1988). The Determinants of Board Composition, Rand Journal of Economics. 19: 589 606. Hermalin, Benjamin E. and Michael S. Weisbach (1991). The Effects of Board Composition and Direct Incentives on Firm Performance, Financial Management. 20: 101112. Hermalin, Benjamin E. and Michael S. Weisbach (2003). Board of Directors as an Endogenously Determined Institution: Survey of the Economic Literature, FRBNY Economic Policy Review. 9: 7 26. Huther, Jeff (1997). An Empirical Test of the Effect of Board Size on Firm Efciency, Economics Letters. 54: 259 264. Jensen, Michael C. (1986). Agency Costs of Free Cash Flows, Corporate Finance, and Takeovers, American Economic Review. 76: 323 329. Jensen, Michael C. (1993). The Modern Industrial Revolution, Exit, and the Failure of Internal Control Systems, Journal of Finance. 48: 831 880. Kunz, Roger (2002). Simplication of Equity Capital Structure and Market Value, Financial Markets and Portfolio Management. 16: 30 52. Levine, Ross and David Renelt (1992). A Sensitivity Analysis of Cross-Country Growth Regressions, American Economic Review. 82: 942 963. Lipton, Martin and Jay Lorsch (1992). A Modest Proposal for Improved Corporate Governance, Business Lawyer. 48: 59 77. Loderer, Claudio and Urs Peyer (2002). Board Overlap, Seat Accumulation and Share Prices, European Financial Management. 8: 165 192. Morck, Randall, Andrei Shleifer and Robert W. Vishny (1988). Management Ownership and Market Valuation: An Empirical Analysis, Journal of Financial Economics. 20: 293 315. Myers, Stewart C. (1984). The Capital Structure Puzzle, Journal of Finance. 39: 575 592. Postma, Theo J. B. M., Hans van Ees and Elmer Sterken (2003). Board Composition and Firm Performance in The Netherlands, Eastern Economic Journal. 29: 41 58. Shivdasani, Anil and David Yermack (1999). CEO Involvement in the Selection of New Board Members: An Empirical Analysis, Journal of Finance. 54: 1829 1854. Shleifer, Andrei and Robert W. Vishny (1986). Large Shareholders and Corporate Control, Journal of Political Economy. 94: 461 488. Stiglitz, Joseph E. (1985). Credit Markets and the Control of Capital, Journal of Money, Credit, and Banking. 17: 133 152. Wu, YiLin (2000). Honey, I Shrunk the Board, Working Paper. University of Chicago. Yermack, David (1996). Higher Market Valuation of Companies with a Small Board of Directors, Journal of Financial Economics. 40: 185 211.
355
28.5.04
22. Juli 2004 13:07

Zeckhauser, Richard J. and John Pound (1990). Are Large Shareholders Effective Monitors? An Investigation of Share Ownership and Corporate Performance, in: Glenn R. Hubbard (ed.), Asymmetric Information, Corporate Finance, and Investment. Chicago: University of Chicago Press: 149 180.
SUMMARY Using a simultaneous equations framework with a comprehensive set of publicly listed Swiss companies, our ndings suggest that the size of the board of directors is an independent corporate governance mechanism. This implies that any potential relationship between board size and rm valuation is indeed causal. However, in contrast to previous studies, we do not uncover a signicant relationship between board size and rm valuation, which can be interpreted as support for the hypothesis of the existence of an optimal board size. On average, rms choose the number of board members just optimally. This indicates that cross-sectional variations in board size to a large extent reect differences in rms underlying environment, and not mistaken choices.
ZUSAMMENFASSUNG Die Ergebnisse der Schtzung eines simultanen Gleichungssystems mit einer reprsentativen Stichprobe brsengehandelter Schweizer Unternehmen zeigen, dass die Grsse des Verwaltungsrates einen eigenstndigen Corporate Governance Mechanismus darstellt. Damit kann ein mglicher Zusammenhang zwischen der Grsse des Verwaltungsrates und dem Unternehmenswert als kausal interpretiert werden. Die empirischen Ergebnisse zeigen aber keine Evidenz fr einen derartigen Zusammenhang, was die Hypothese einer optimalen Grsse des Verwaltungsrates sttzt. Im Durchschnitt weist der Verwaltungsrat Schweizer Unternehmen die optimale Grsse auf. Die Unterschiede in der Grsse der Verwaltungsrte der Stichprobenunternehmen knnen zum grossen Teil durch unternehmensspezische Einussgrssen erklrt werden, und sind nicht auf eine falsche Besetzung des Verwaltungsrates zurckzufhren.
RSUM En utilisant un systme dquations simultan, estim sur lensemble des entreprises inscrites la bourse suisse, nos rsultats montrent que la taille du conseil dadministration rsulte dun mcanisme indpendant de gouvernance dentreprise. Ceci implique quil existe vritablement un lien causal entre la taille du conseil dadministration et la valeur de lentreprise. Nanmoins, contrairement aux tudes prcdentes, nous navons pas dcouvert de relations signicatives entre ces deux lments, ce que lon peut interprter comme support pour lhypothse de lexistence dune grandeur optimale du conseil dadministration. Le nombre de membres du conseil dadministration est en moyenne choisi de manire optimale. Ceci indique que la variation en coupe transversale de la taille du conseil dadministration rete en majeure partie des diffrences dans lenvironnement de chaque entreprise plutt que des mauvaises dcisions.
356
28.5.04
22. Juli 2004 13:07

Is Broad Size An in Dependance Corporate Governace Machnisim

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Is Broad Size An in Dependance Corporate Governace Machnisim

Uploaded by

Copyright:

Available Formats

KYKLOS, Vol. 57 2004 Fasc.

Is Board Size an Independent Corporate Governance Mechanism?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

Panel of 452 4 to 34 large US rms Mean: 12.25 (1984 1991) Median: 12

3 to 16 Mean: 9 Mean: 3.7

Total variable costs ROA

Mean 4.95 Median: 5

Negative board size effect

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

BOARDSIZE OUTSIDER LEV OWNERSHIP

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

10. Excluded rms have values of Tobins Q above 6.

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

11. See Yermack (1996), Figure 1, p. 193.

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?

22. Juli 2004 13:07

STEFAN BEINER/WOLFGANG DROBETZ/FRANK SCHMID/HEINZ ZIMMERMANN

22. Juli 2004 13:07

IS BOARD SIZE AN INDEPENDENT CORPORATE GOVERNANCE MECHANISM?