An Integrated Framework of Corporate Governance and Firm Valuation Evidence from Switzerland

Transcription

1 An Integrated Framework of Corporate Governance and Frm Valuaton Evdence from Swtzerland Stefan Bener 1, Wolfgang Drobetz 2, Markus Schmd 3, and Henz Zmmermann 4, Verson: October 2003 Abstract Recent emprcal work shows evdence of a postve relatonshp between frm-specfc corporate governance and frm valuaton, as measured by Tobn s Q. Instead of lookng at a sngle control mechansm, we use a broad corporate governance ndex and addtonal varables related to ownershp structure, board characterstcs, and leverage to provde a comprehensve descrpton of frm-level corporate governance for a broad sample of Swss frms. To avod a problem nherent n most of the prevous lterature, we carefully control for the endogenety of these control mechansms. We develop a system of smultaneous equatons, where all control mechansms are allowed to affect each other as well as Tobn s Q, whle at the same tme Tobn s Q s also allowed to affect the choce of each mechansm. Our results support the wdespread hypothess of a postve relatonshp between corporate governance and Tobn s Q. An ncrease n our corporate governance ndex by one pont (where the ndex ranges from 1 to 100) causes an ncrease of the market captalzaton by roughly 8%, on average, of a company s book asset value. We also report a number of other nterestng results on the relatonshp between Tobn s Q and the dfferent control mechansms. Keywords: Corporate governance; Ownershp structure; Frm valuaton JEL Classfcaton: G12, G32, G34, G38 1 Department of Fnance, Unversty of Basel, Holbenstrasse 12, 4051 Basel, Swtzerland, and Sad Busness School, Unversty of Oxford, Park End Street, Oxford OX1 1HP UK, Mal: stefan.bener@sbs.ox.ac.uk 2 Department of Corporate Fnance, Unversty of Basel, Holbenstrasse 12, 4051 Basel, Swtzerland, Phone: , Fax: , Mal: wolfgang.drobetz@unbas.ch 3 Department of Fnance, Unversty of Basel, Holbenstrasse 12, 4051 Basel, Swtzerland, Phone: , Fax: , Mal: markus-max.schmd@unbas.ch 4 Department of Fnance, Unversty of Basel, Holbenstrasse 12, 4051 Basel, Swtzerland, Phone: , Fax: , Mal: henz.zmmermann@unbas.ch We thank Manuel Berger and Nco Waldmeer for helpful comments and Max Schmd for hs large contrbuton to the collecton of compensaton and shareholdng data from the annual reports. Fnancal support from the Natonal Center of Competence n Research Fnancal Valuaton and Rsk Management (NCCR FINRISK) s gratefully acknowledged. The NCCR FINRISK s a research program supported by the Swss Natonal Scence Foundaton. Bener acknowledges fnancal support from the Swss Natonal Scence Foundaton (SNF).

2 2 1. Introducton Ths paper addresses the queston whether good corporate governance has a postve mpact on the valuaton of lsted companes n Swtzerland. There has been an ongong debate n the lterature how to measure the qualty of frm-level corporate governance. In the meantme, many countres have adopted new standards, rules, or codces of best practce to establsh gudelnes for publcly lsted companes n an attempt to mprove the overall level of corporate governance. In Swtzerland, for example, the Drectve on Informaton Relatng to Corporate Governance and the Swss Code of Best Practce have become effectve as of July 1, Smlarly, the German Corporate Governance Code has been put n place n early 2002 and contans recommendatons and suggestons for good governance practces for lsted frms. Whle there are numerous emprcal studes on the lnk between governance practces and wealth effects for the US stock market, the emprcal evdence for European markets s stll scarce. Swtzerland s a partcularly nterestng case to analyze. The nsttutonalzaton of shareholdngs,.e., the accumulaton of stocks by professonal asset managers, had a partcular strong effect on the structural changes of the equty market after penson plans became mandatory n the md-eghtes and emerged as the major domestc nvestment force thereafter. Moreover, n the course of globalzaton of equty markets many restrctons protectng the management of Swss frms were abandoned, such as restrctons on the transferablty or ownershp of shares ( Vnkulerung ) or multple share classes wth lmted or unequal votng rghts. These developments make t nterestng to nvestgate the role of specfc control mechansms n more detal. Although the market for corporate control has stll developed slowly durng the nnetes (see Loderer and Zgraggen (1999) for a remarkable takeover attempt), there s a clear trend to adopt nternatonally recognzed governance prncples n recent years. We suspect that corporate governance s ncreasngly understood among polcymakers as a value enhancng strategy n a compettve envronment rather than an oblgaton. The mplct hypothess underlyng our analyss s that better corporate governance wll ncrease frm value, as measured by Tobn s Q. The noton behnd ths proposton s that agency problems may affect the value of companes through two dstnct channels: () the expected cash flows accrung to nvestors and () the cost of captal.

3 3 Frst, agency problems make nvestors pessmstc about future cash flows. Based on ths smple dea, the model n La Porta, Lopez-de-Slanes, Shlefer, and Vshny (2002) predcts that nvestors bd up stock prces, because wth better legal protecton, more of the frm s profts would come back to them as nterest or dvdends as opposed to beng exproprated by the entrepreneur who controls the frm. 5 Second, good corporate governance decreases the cost of captal,.e., the expected return on equty, to the extent that t reduces shareholders montorng and audtng costs. Ths dea has been formalzed n a model by Lombardo and Pagano (2002). Specfcally, they extend the Captal Asset Prcng Model (CAPM) to account for the expected agency costs caused by the conflct of nterest between nsders and outsde shareholders. In order to test the hypotheszed relatonshp between governance qualty and frm valuaton, we construct a frm-specfc Corporate Governance Index (CGI) based on a broad survey among all lsted companes on the Swss Stock Exchange (SWX), excludng nvestment companes. Ths approach has been popular n the lterature only recently. For example, Black, Jang, and Km (2003) and Drobetz, Schllhofer, and Zmmermann (2003) construct a surveybased governance ndex and report that a better frm-level corporate governance s assocated wth hgher frm valuaton. The ndex we use n ths paper s based on the recommendatons and suggestons of the Swss Code of Best Practce. It serves as a broad measure of frmspecfc corporate governance qualty and reflects 38 dfferent governance attrbutes, whch are not (yet) legally requred but consdered as good corporate governance practce by nternatonal standards. To provde an ntegrated framework, we also nvestgate other crucal control mechansms not contaned n the ndex, namely stock ownershp by offcers and drectors, outsde blockholdngs, leverage, board sze, and the fracton of outsde drectors on the board. A problem whch plagues vrtually all emprcal studes n corporate governance s endogenety. The queston s whether good corporate governance causes hgher frm valuatons. Alternatvely, frms wth hgher market values could smply be more lkely to choose better gov- 5 See LaPorta, Lopez-de-Slanes, Shlefer, and Vshny (2002), p We also refer to the theoretcal model n Shlefer and Wolfenzon (2003).

4 4 ernance structures. Clearly, f corporate governance was endogenously determned, we could not make an assessment of the causal connecton. In our emprcal analyss we are therefore careful to control for possble nterrelatonshps between the dfferent control mechansms and Tobn s Q. One may suspect mportant substtuton effects between the dfferent governance mechansms,.e., snce alternatve control mechansms exst, greater use of one mechansm needs not be postvely related to frm valuaton. Specfcally, where one mechansm s used less, others may be used more, resultng n the same valuaton effects. The exstence of alternatve control mechansms and ther possble nterdependences make regressons relatng the use of any sngle mechansms to frm valuaton dffcult to nterpret. To avod spurous regresson results, followng Agrawal and Knoeber (1996) we develop a comprehensve system of smultaneous equatons and apply three-stage least squares (3SLS). Our results clearly support the wdespread hypothess of a postve relatonshp between corporate governance and frm value. Most mportant, an ncrease n the corporate governance ndex by one pont (where the ndex ranges from 1 to 100) causes an ncrease of the market captalzaton by roughly 8%, on average, of a company s book asset value. We also report a number of other nterestng results on the relatonshp between Tobn s Q and the dfferent control mechansms. For example, frm valuaton sgnfcantly ncreases wth hgher shareholdngs of offcers and drectors. In addton, hgher shareholdngs of offcers and drectors are assocated wth a lower fracton of outsders on the board, whch ndcates possble substtuton effects. The remander of the paper s organzed as follows. Secton 2 gves a motvaton for the use of each control mechansm employed n ths paper and revews some earler emprcal evdence. Secton 3 presents the methodologcal approach. Secton 4 defnes the varables and descrbes the data. The emprcal results are presented n Secton 5. Fnally, Secton 6 concludes. 2. Control mechansms and frm value Theoretcal consderatons and emprcal evdence In the theoretcal lterature dfferent control mechansms have been suggested to allevate the well-known agency problems between managers and shareholders. Accompanyng ths classcal strand of the corporate fnance lterature, an extensve emprcal lterature has developed wth the avalablty of approprate frm-level data only recently. An aspect, whch attracted partcular attenton, has been the emprcal relatonshp between corporate governance and frm valuaton. Unfortunately, vrtually all prevous studes concentrated on a partcular as-

5 5 pect of governance, such as takeover defenses (e.g., Gompers, Ish and Metrck (2003)), executve compensaton (e.g., Loderer and Martn (1997)), blockholdngs (e.g., Demsetz and Lehn (1985)), board sze (e.g., Yermack (1996) and Esenberg, Sundgren, and Wells (1998)) or board composton (e.g., Hermaln and Wesbach (1991) and Bhagat and Black (2002)). 6 Nevertheless, t seems crucal to account for the fact that all these control mechansms whle valuable per se can be adopted alternatvely and are possbly substtuted for each other to some extent. Therefore, n our emprcal analyss, we use an extensve set of governance varables. As a frst varable we construct a broad corporate governance ndex, whch s manly based on the Swss Code of Best Practce (e.g., see Black, Jang and Km (2003) and Drobetz, Schllhofer, and Zmmermann (2003) for comparable ndces). To gve a comprehensve pcture of frm-level corporate governance, our study ncludes fve addtonal (unrelated) control mechansms: share ownershp by offcers and drectors, shareholdngs by large outsde blockholders, board sze, leverage, and outsde representaton on the board. In ths secton we provde a short motvaton for the use of these control mechansms and revew some earler emprcal evdence. 2.1 Corporate Governance Index (CGI) There s substantal evdence n the law and fnance lterature that varatons n country-level legslatve rules affect the market valuaton of ndvdual frms. Supportng the noton that ncreased shareholder actvsm, tghtened rules and regulaton, and addtonal self-regulaton delver hgher shareholder value, La Porta, Lopes-de-Slanes, Shlefer and Vshny (2002) fnd for a sample of 539 frms of 27 wealthy economes that better protecton of mnorty shareholders s assocated wth a hgher valuaton. However, there s less evdence on the effect of frm-specfc varaton n corporate governance practces wthn a sngle jursdcton on frm valuaton. For example, Black (2001) examnes the relatonshp between corporate governance and frm value for a sample of 21 Russan frms. He constructs a frm-specfc corporate governance rankng and shows that a one standard devaton change n the governance rankng predcts a seven-fold ncrease n frm value. 7 For a much broader sample of 859 frms n 27 countres Durnev and Km (2002) fnd that companes wth a better corporate governance and better dsclosure standards have, on average, hgher Tobn s Qs and nvestments. They 6 See Shlefer and Vshny (1997) for a survey. 7 However, these results should be nterpreted wth due care. Frst, the sample sze s very small. Second, Black (2001) does not control for endogenety.

6 6 report that a 10 pont ncrease (out of 100) n the Credt Lyonnas Securtes Asa (CLSA) corporate governance ndex ncreases a frm s market value by 13.3%, whle a 10 pont ncrease (out of 98) n the S&P dsclosure and transparency ndex ncreases a frm s market value by 16.3%. Klapper and Love (2002) support these fndngs for the CLSA ndex usng a sample of 374 large frms n 145 emergng markets. Smlarly, Gompers, Ish, and Metrck (2003) construct a governance ndex based on takeover defenses for a sample of about 1500 US frms. They report that frms wth a better corporate governance receve hgher market valuatons and have hgher profts, hgher sales growth, and lower captal expendtures. In a smlar ven, Drobetz, Schllhofer and Zmmermann (2003) document a postve relatonshp between governance practces and frm valuaton for German publc frms by constructng a broad corporate governance ratng related to the German Corporate Governance Code. They report that for the medan frm a one standard devaton change n the governance ratng results n a 24% ncrease n the value of Tobn s Q. However, most of these studes do not drectly address the possble endogenety of corporate governance mechansms,.e., most of the results just descrbed can really be nterpreted as partal correlatons only wthout clear ndcaton of causalty. An excepton s Black, Jang, and Km (2003), who fnd a postve relaton between ther corporate governance ndex and Tobn s Q for a sample of 526 Korean publc companes. Ther ndex s prmarly based on responses to an extensve survey among Korean lsted companes and conssts of sx subndces for shareholder rghts, board of drectors n general, outsde drectors, audt commttee and nternal audtor, dsclosure to nvestors, and ownershp party. To control for a possble endogenety, they use a three-stage least square (3SLS) smultaneous equatons approach and show that a 10 pont ncrease (out of 100) n the governance ndex causes a 19.4% ncrease n Tobn s Q. 2.2 Offcers and drectors fnancal commtment to ther frm Probably the most promnent nstrument to algn the nterests of managers and shareholders s to gve managers an equty stake n the frm. Another mportant control mechansm whch ensures that managers pursue the nterests of shareholders s the montorng of managers by the board of drectors. But who montors the montors? Recently, stock partcpaton plans have been extended to drectors of the board to ensure that they montor and dscplne managers, thereby protectng the nterests of shareholders. However, n spte of the pervasve

7 7 theoretcal clams, the emprcal evdence on the relaton between nsder stockholdngs and frm performance s mxed and somewhat contradctory. For example, Morck, Shlefer, and Vshny (1988) fnd a nonlnear relaton between the fracton of stock held by the members of the board of drectors and Tobn s Q, and a less sgnfcant relaton when frm performance s measured by the accountng rate of return. At least, when the fracton of shares held by the board s small, as s the case for most frms n ther sample, greater board shareholdng mproves performance. McConnell and Servaes (1990) fnd a smlar nonlnear relaton between nsder shareholdngs and frm valuaton. Mehran (1995) also fnds that frm performance, as measured by Tobn s Q or the return on assets, s postvely related to both the percentage of executve compensaton whch s stock-based and the percentage of equty held by the management. However, dfferent fndngs are reported n the more recent lterature, whch employs smultaneous systems of equatons to control for a possble endogenety of manageral share allocaton and ownershp. For example, Chung and Prutt (1996) fnd a postve nfluence of CEO equty ownershp on Tobn s Q. In contrast, also applyng a smultaneous equatons framework, Loderer and Martn (1997) fnd no evdence that larger manageral stockholdngs lead to a better frm performance. However, frm performance seems to affect how much stock executves want to hold n ther frm. Hmmelberg, Hubbard, and Pala (1999) control for the endogenety of ownershp by usng panel data and also fnd lttle evdence that changes n manageral ownershp affect performance. Fnally, Schmd (2003) uses a smultaneous equatons system and fnds a postve effect of the percentage share ownershp of offcers and drectors on frm valuaton for a sample of 134 Swss frms. By the same tme hs results suggest that Swss managers are more lkely to hold equty when they are confdent that ther company wll do well and, hence, t s attractve for them to partcpate. 2.3 Blockholdngs The ncluson of outsde blockholdngs as an addtonal control mechansm s motvated by the well-known freerder problem assocated wth wdely dspersed shareholders. In contrast, an ownershp structure where one or more shareholders own large blocks of stock should enable closer montorng of management behavor. In fact, Shlefer and Vshny (1986) present a model n whch a block shareholder effectvely montors management by vrtue of representng a credble takeover threat. In addton, Dens, Dens, and Sarn (1997) show that top executve turnover s more senstve to poor performance n frms wth outsde blockholders

8 8 than n frms wthout outsde blockholders. They nterpret ths fndng as evdence for a montorng functon of managers by outsde blockholders. However, the overall emprcal evdence on the effects of block ownershp on frm value s mxed. Mkkelson and Ruback (1985) document postve abnormal returns followng the announcement of the acquston of a 5% or greater stake n one frm by another frm. Smlarly, McConnell and Servaes (1990) fnd a postve relaton between nsttutonal ownershp and Tobn s Q. In contrast, however, Demsetz and Lehn (1985) fnd no cross-sectonal relatonshp between the concentraton of shareholdngs and accountng rates of return. Agrawal and Knoeber (1996) also fnd no evdence for a postve nfluence of outsde blockholdngs on frm performance. Ther results are robust, usng a standard OLS procedure and also estmatng a smultaneous system of equatons to control for possble endogenety. Usng a smlar approach, Schmd (2003) fnds no statstcally sgnfcant effect of outsde blockholdngs on frm valuaton for hs sample of 134 Swss frms. Interestngly, Bener, Drobetz, Schmd, and Zmmermann (2003) even fnd a sgnfcant negatve relaton between blockholdngs and frm performance for ther sample of Swss frms Board sze Consderng ts role to drectly montor and dscplne a frm s management, the board of drectors s another mportant control mechansm, whch ensures that managers pursue the nterests of shareholders. Lpton and Lorsch (1992) and Jensen (1993) were the frst to hypothesze that board sze s an ndependent control mechansm. Specfcally, they argue that large boards may be less effectve than small boards. The underlyng noton s that large boards can make coordnaton, communcaton, and decson makng more cumbersome than n smaller groups. Jensen (1993) suggests an optmal board sze of seven or eght drectors. Yermack (1996) was the frst to nvestgate ths proposton emprcally. In fact, usng a sample of large US publc corporatons, he reports an nverse relatonshp between board sze and frm value, as measured by Tobn s Q. Controllng for possble endogenety, he fnds that causalty s runnng from board sze to Tobn s Q. There s no evdence that frms change board sze as a reacton to past performance. Although he does not account for a possble endogenety of 8 These seemngly dfferent and contradctory fndngs are not surprsng because Bener, Drobetz, Schmd, and Zmmermann (2003) do not separate between nsder and outsder shareholdngs. Hence, stock holdngs by offcers and drectors exceedng 5% are also ncluded n ther proxy for blockholdngs.

9 9 board sze, Huther (1997) confrms these fndngs for a sample of US electrcty companes. Esenberg, Sundgren, and Wells (1998) also fnd a negatve sze effect for a sample of small Fnnsh frms. Most recently, usng a smultaneous equatons approach for a sample of Swss companes Bener, Drobetz, Schmd, and Zmmermann (2003) cannot detect a sgnfcant relatonshp between board sze and frm value. They nterpret ths fndng as evdence that Swss frms, on average, choose the number of board members just optmally. In fact, average board sze n ther sample s 6.6, whch s close to the optmal board sze put forth by Lpton and Lorsch (1992) and Jensen (1993). 2.5 Leverage Jensen (1986, 1993), Stulz (1990), and Hart and Moore (1995), among others, suggest that debt helps to dscourage overnvestment of free cash flow by self-servng managers. Debt can also create value by gvng the management the opportunty to sgnal ts wllngness to dstrbute cash flows and to be montored by lenders. Hence, we also nclude leverage as an addtonal control mechansm. McConnell and Servaes (1995) fnd that book leverage s postvely correlated to frm value when nvestment opportuntes are scarce, whch s consstent wth the hypothess that debt allevates the overnvestment problem. In contrast, Agrawal and Knoeber (1996) and Bener, Drobetz, Schmd, and Zmmermann (2003) fnd no relaton between leverage and frm performance and argue that leverage s just employed optmally n conjuncton wth the other control mechansms consdered n ther studes. 2.6 Outsde representaton on the board In addton to board sze, the composton of the board has also been suggested as an mportant determnant of frm performance. The board of drectors s responsble for evaluatng the senor management of a corporaton and to replace t f t does not pursue shareholders nterests. Because nsde drectors careers are ted to the CEO s, they are generally unable or unwllng to remove ncumbent CEOs and, hence, ths task s lkely to fall on outsde drectors. Moreover, outsde drectors are often respected leaders from the busness or academc communty, whose reputatons suffer when they are assocated wth falng companes. It s wdely beleved that ths creates an ncentve to prevent severe governance malfunctons. Consstently, Rosensten and Wyatt (1990) provde evdence that the proporton of outsde drectors postvely affects shareholder wealth. In fact, they document a postve stock prce reacton upon announcement of the appontment of an addtonal outsde drector. Addton-

10 10 ally, the fndngs of Wesbach (1988) suggest that frms wth outsder-domnated boards are sgnfcantly more lkely to remove the CEO on the bass of bad performance than frms wth nsder-domnated boards. In a related study, Hermaln and Wesbach (1988) fnd that outsders are more lkely to jon a board after a frm performs poorly or leaves an ndustry. In contrast, Yermack (1996) fnds no assocaton between the percentage of outsde drectors and frm performance. Ignorng any nterdependences among the dfferent mechansms, Agrawal and Knoeber (1996) fnd that representaton of outsders on the board of drectors, debt polcy, and corporate control actvty are related to Tobn s Q. When they account for the nterdependences among these mechansms n a smultaneous equatons system, only the percentage of outsder board representaton remans sgnfcantly related to frm performance. These fndngs are agan consstent wth the proposton that all control mechansms are just chosen optmally, except for board composton. In contrast, Bener, Drobetz, Schmd, and Zmmermann (2003) fnd no relaton between the fracton of outsde board members and Tobn s Q for ther sample of Swss frms. 3. Methodologcal approach A common approach to estmate the mpact of frm-level corporate governance practces on frm value s to regress Tobn s Q on a corporate governance ndex and to nclude some (exogenous) control varables. However, the exstence of alternatve control mechansms, such as nsder stock ownershp or outsde blockholdngs, whch are usually not ncluded n the ndex may lead to an omtted varables bas. Hence, regressons, whch relate the use of any sngle control mechansm to frm value, are dffcult to nterpret. Furthermore, because such regressons fal to account for possble nterrelatons among the dfferent control mechansms, the results suffer from endogenety problems. Therefore, followng Agrawal and Knoeber (1996), Bener, Drobetz, Schmd and Zmmermann (2003) and Schmd (2003) we specfy a smultaneous equatons system where each control mechansm s the dependent varable n one of the equatons. The choce of any of the sx control mechansms may depend upon choces of the fve other control mechansms, but these choces wll depend on other (exogenous) factors as well. Thus, each equaton ncludes all other control mechansms and addtonal exogenous explanatory varables. To nvestgate the effect of the dfferent control mechansms on frm valuaton, a seventh equaton wth Tobn s Q as the dependent varable s added to the system. At the same tme, Tobn s Q s ncluded as an explanatory varable n the other sx equatons

11 11 of the system to allow for possble nterrelatons wth the sx control mechansms. In ths secton we provde a detaled descrpton of each equaton n our system. The frst equaton has our corporate governance ndex, denoted as CGI, as the dependent varable. We provde a more detaled descrpton of CGI n secton 4.1, but t s mportant to note that the ndex for the most part refers to the recommendatons and suggestons n the Swss Code of Best Practce. It s therefore not drectly related to the other control mechansms we use n our analyss,.e., share ownershp by offcers and drectors, shareholdngs by large outsde blockholders, board sze, leverage, and outsde representaton on the board. Nevertheless, we expect that CGI depends on choces of the other control mechansms as well as on addtonal exogenous control varables. As a frst varable, we nclude frm sze, as measured by the natural logarthm of total assets, Lnassets, as exogenous varable. From a theoretcal pont of vew, the effect of sze on a frm s corporate governance s ambguous. On the one hand, large frms may have more severe agency problems because t s harder to montor them or because of Jensen s (1986) free cash flows argument. Therefore, they may voluntarly choose strcter governance rules to avod hgh agency costs. On the other hand, small frms may have better growth opportuntes and, hence, greater needs for external fnancng and better control mechansms. However, emprcal studes provded convncng evdence of a postve effect of sze on the qualty of a frm s corporate governance prncples. 9 As a second varable, we nclude a measure of growth opportuntes. Frms wth good growth opportuntes wll need to rase external fnancng n order to expand and may therefore fnd t optmal to mprove ther corporate governance. The underlyng noton s that better governance and better mnorty shareholder protecton wll lkely lead to lower cost of captal. 10 Lackng alternatve measures of growth, we follow Klapper and Love (2002) and use the average annual sales growth over the past three years, Growth, as an arguably mperfect measure of future growth opportuntes. To capture a possble nterrelaton between operatng performance and frm specfc corporate governance, we also nclude the return on assets, ROA. Klapper and Love (2002) argue that the composton of a frm s assets wll affect ts contractng envronment. Intutvely, t s easer to montor and harder to steal fxed assets (e.g., machnery and equpment) than soft captal (e.g., ntangbles and R&D captal). Accordngly, a frm operatng 9 See for example Klapper and Love (2002) and Drobetz, Schllhofer, and Zmmermann (2003). 10 See Lombardo and Pagano (2000) and Hmmelberg, Hubbard, and Love (2001), among others, for the relatonshp between nvestor protecton and the cost of captal.

12 12 wth a hgher proporton of ntangble assets may fnd t optmal to adopt strcter control mechansms to prevent msuse of these assets,.e., we should observe a postve relaton between CGI and the proporton of ntangble to total assets, Intang. We also nclude SMI, a dummy varable, whch s equal to one f a frm s ncluded n the Swss Market Index (SMI), comprsng of the 26 largest Swss frms, as of end In addton, due to stronger nvestor pressure from abroad, one would also expect a postve effect of the dummy varable SMI on our corporate governance ndex. As also argued by Klapper and Love (2002), there are reasons to presume that frms, whch trade n the US, should have better corporate governance rankngs. Frst, frms lsted on a US exchange are requred to comply wth US GAAP accountng standards, whch mght mprove ther transparency. Second, frms wth a lstng on a US stock exchange are subject to many SEC laws and regulatons that protect mnorty shareholders. Thus, we add a dummy varable ndcatng whether a frm trades Amercan Depostory Recepts (ADRs) n the Unted States, labeled as ADR, and expect t to have a postve effect on CGI. Fnally, to control for ndustry effects, we nclude 12 dummy varables, Industry, n all seven equatons of the system. Assumng that all subsequent relatons are lnear and denotng the other sx corporate governance mechansms (also ncludng Tobn s Q) as CGM, the frst equaton of our system s: (1) CGI = α = 1 α SMI CGM 12 7 ADR Lnassets + 12 = 1 α Industry Growth + ε 9 ROA 10 Intang The second equaton of our system uses the percentage of total shares owned by offcers and drectors, labelled as Stocksod, as the dependent varable. We expect Stocksod to be less where the costs of such shareholdngs are greater. These costs arse manly from holdng an undversfed portfolo. As suggested by Agrawal and Knoeber (1996), we use the volatlty of stock returns, Vola, and frm sze, Lnassets, as ndcators of the assocated costs. Thus, we expect Stocksod to be negatvely related to Vola. Because very large amounts of money are necessary to hold a sgnfcant fracton of a large company, Stocksod s expected to be negatvely related to Lnassets as well. In contrast, we nclude Growth as an ndcator of expected growth opportuntes, whch ncreases the attractveness of holdng shares of the company. Addtonally, we nclude the natural logarthm of the age of a frm, Lnage. One would expect

13 13 shareholdngs of offcers and drectors to be hgher n younger frms. Votng restrctons could allow some shareholders to practcally domnate a frm even f they own sgnfcantly less than 50% of the frm s stock. We therefore expect a postve relatonshp between Stocksod and a dummy varable for the exstence of more than one share category, denoted as Scat. Fnally, we nclude a dummy varable whch s equal to one f the CEO or the presdent of the board s also the founder of the frm and zero otherwse. The symbol for ths dummy varable s Founder. We expect Stocksod to be postvely related to Founder. Summarzng, the second equaton of our system s: (2) Stocksod = α = 1 α Lnage CGM 11 7 Scat Vola 12 8 Lnassets Founder + 12 = 1 α Growth Industry + ε The dependent varable n the thrd equaton of our system s Blockout, whch s the percentage of cumulated votng rghts exercsed by large outsde blockholders ownng 5% or more of the frm s equty. Smlar to equaton (2), we expect Blockout to be negatvely related to the costs of outsder shareholdngs. Agan, we use Vola and Lnassets as ndcators of these costs. Thus, we expect Blockout to be negatvely related to both Vola and Lnassets. For the same reason as for Stocksod, we expect Blockout to be postvely related to Growth. Zeckhauser and Pound (1990) argue that the hgher a frm s R&D ntensty, the more closed s the nformaton structure and the more dffcult s outsde montorng. Large nvestors wll recognze the problems assocated wth asymmetrc nformaton and, hence, Blockout s expected to be negatvely related to R&D ntensty. Because R&D data for the year 2002 was not avalable for the frms n our sample, we use the rato of ntangble to total assets, Intang, as a proxy for nformaton asymmetry. As for Stocksod, we agan expect a postve relatonshp between Blockout and a dummy varable for the exstence of more than one share category, Scat. Fnally, we nclude the number of outsde blockholders, Blockonr, as a control varable n our system. Summarzng, the thrd equaton s: (3) Blockout = α = 1 α Intang CGM 11 7 Scat Vola 12 8 Lnassets Blockonr + 12 = 1 α Growth Industry + ε

14 14 The fourth equaton n our system has the number of drectors on the board, labeled as Bsze, as the dependent varable. As exogenous control varables we nclude frm sze, Lnassets, a dummy varable that s one f the state owns more than 5% of the frm s equty and zero otherwse, denoted as Sown, and the return on assets, ROA. We expect large frms to have larger boards of drectors and, hence, a postve relatonshp between Bsze and Lnassets should emerge. Followng Bener, Drobetz, Schmd, and Zmmermann (2003), Sown should be ncluded to account for the possblty that poltcal nfluences lead to presumably larger boards wth a dsproportonate number of government representatves. Thus, Sown s expected to be postvely related to Bsze. As hypotheszed by Yermack (1996), small boards could ncrease frm performance, or dependng on the drecton of causalty, frms mght adjust board sze n response to past performance. To capture a possble nterrelaton between operatng performance and board sze, we nclude ROA. Hence, the fourth equaton n our system s: (4) Bsze = α = 1 α 6 = 1 9+ α CGM Industry + ε 7 Lnassets 8 Sown 9 ROA The dependent varable n the ffth equaton of our system s LV, as measured by the rato of total (non-equty) labltes to total assets (Rajan and Zngales (1995)). As suggested by Jensen s (1986) free cash-flow hypothess, we expect mature frms wth substantal cash-flows to use more debt to dscplne managers. We use two dfferent varables to proxy for the maturty of a frm. The frst s Lnassets, expectng that more mature frms tend to be larger, and the second s the frm s age, Lnage. Addtonally, we nclude a dummy varable, whch s one f the frm pad a dvdend n 2003 (based on the earnngs of 2002) and zero otherwse, denoted as Dv. Because the avalablty of nternal funds provdes an alternatve to debt fnancng, we expect a negatve relatonshp between Dv and LV. If a frm has expected future growth opportuntes, debt servcng requrements can lmt management s ablty to pursue postve net present value projects, leadng to ex-post undernvestment (Myers (1977)). Consstently, Drobetz and Fx (2003) report lower leverage ratos for Swss frms wth favorable growth opportuntes. Accordngly, we expect a negatve relatonshp between Growth and LV. To capture possble relatonshps between operatng performance and leverage, we also nclude ROA. Summarzng, the ffth equaton n our system s:

15 15 (5) LV = α = 1 α Growth CGM 11 7 ROA Lnassets + 12 = 1 α Lnage Industry + ε 9 Dv Fnally, the percentage of outsde (non-executve) drectors on the board, Outsder, s the dependent varable n the sxth equaton of our system. The frst control varable we nclude s Ceop, a dummy varable, whch s equal to one f the CEO s the presdent of the board at the same tme. Whle ths helps to allevate coordnaton and communcaton problems between the CEO and the board of drectors, t clearly cannot ensure an ndependent montorng of the CEO and the top management by the board of drectors. Shvdasan and Yermack (1999) argue that a stuaton where the CEO s at the same tme presdent of the board leads to a concentraton of power and the electon of less ndependent board members. Accordngly, we expect a negatve relatonshp between Ceop and Outsder. We also nclude the Founder dummy varable because foundng CEOs and presdents of the board may wthdraw from ther professonal postons but retan ther stock holdngs of the frm, stll havng enough power to nfluence the composton of the board. To control for the effect of government ownershp on board composton, we also nclude the dummy varable Sown. Fnally, we nclude Growth and ROA as exogenous explanatory varable n the sxth equaton, whch can formally be wrtten as: (6) Outsder = α = 1 α Growth CGM 11 7 ROA Ceop + 12 = 1 α Sown Industry 9 Founder + ε To examne the cross-sectonal relaton between the control mechansms and frm performance, the dependent varable n the last equaton of our system s Tobn s Q. Followng Yermack (1996), Bener, Drobetz, Schmd, and Zmmermann (2003), and Schmd (2003) we nclude two varables to control for growth opportuntes: Lnassets and Growth. We expect a postve relatonshp between Growth and Q and a negatve nfluence of Lnassets on Q because growth opportuntes should be lower for larger frms. Based on smple valuaton models, Q may addtonally depend on ROA and Beta. Summarzng, the fnal equaton n our system s:

16 16 (7) Q = α = 1 α Beta CGM + 12 = 1 α Lnassets Industry + ε 8 Growth 9 ROA To nvestgate the senstvty of our results we are careful to properly account for the possble endogenety of the dfferent control mechansms as well as Tobn s Q. To analyze the resultng dfferences, we proceed n three steps. Frst, we estmate sx OLS regressons of Tobn s Q on each control mechansm separately. Second, we estmate equaton (7) by OLS to examne the effect of all control mechansms together. Fnally, we nclude equaton (7) along wth equatons (1) (6) n a smultaneous equatons system and use three stage least squares (3SLS) to estmate the system. Ths treats Q as endogenous along wth the sx control mechansms, allowng each of the mechansms to affect Q, but also allowng Q to affect the choce of each mechansm. Our system of equatons ncludes 15 exogenous and seven endogenous varables. The order condton for dentfcaton states that f an equaton s to be dentfed, the number of predetermned varables excluded from the equaton must be greater than or equal to the number of ncluded endogenous varables mnus one. The lst of ncluded endogenous varables contans varables on the left-hand sde and the rght-hand sde of the equaton. Therefore, at least sx of the exogenous varables must be excluded from any sngle equaton to dentfy the system. However, our development of equatons (1) (7) s motvated ndependently of the requrement for these dentfcaton restrctons to be met. In fact, all equatons n our system are overdentfed and at least three varables could be ncluded to any equaton wthout jeopardzng the dentfcaton. 4. Defnton of varables and data 4.1 Defnton of Varables In ths secton we provde a detaled descrpton of the varables we use n our emprcal analyss. Frst, our Corporate Governance Index (CGI) s based on responses to a detaled questonnare, whch was manly based on the suggestons and recommendatons of the Swss Code of Best Practce. Smlar ndces have been constructed by Black, Jang, and Km (2003) and Drobetz, Schllhofer, and Zmmermann (2003). The survey was sent out to all Swss frms quoted at the Swss Stock Exchange (SWX) wth the excepton of nvestment compa-

17 17 nes and was completed between May and July Where necessary, the data was supplemented and verfed on the bass of annual reports and web pages. Our ndex conssts of 38 governance attrbutes dvded nto the followng fve categores: (1) corporate governance commtment, (2) shareholders rghts, (3) transparency, (4) management and supervsory board matters, and (5) audtng. We assume that all fve categores have the potental to mtgate hdden nformaton as well as moral hazard n a company. To qualfy for ncluson nto the CGI, an attrbute must refer to a governance element that s not (yet) legally requred and needs to be consdered as an nternatonal market practce of good corporate governance. Hence, all of the 38 governance attrbutes can be ntated and mplemented by the frm s decson makers. A sample of representatve questons for each category s lsted below: Corporate governance commtment: Is there a governance offcer who reports regularly to the board of drectors? Shareholders rghts: Does the frm strctly follow the one share-one vote prncple (e.g., no preferental shares and partcpaton certfcates outstandng)? Transparency: Is the agenda of the general meetng, the annual report, any counter proposals, management comments and votng results of all shareholders avalable on the nternet n a tmely manner (n Englsh and German, French or Italan)? Management and supervsory board matters: Are there frm-specfc rules to handle conflct of nterest and own account tradng for managers and members of the board of drectors? Audtng: Are there frm-specfc rules to ensure that the audtor does not perform other servces for the frm (e.g., consultng)? The constructon of our CGI s straghtforward and transparent: Frst, each of the 38 governance attrbutes s assgned a value between 1 (mnmum) and 5 (maxmum). One pont s added for each subsequent acceptance level of the respectve attrbute n a fve-scale answerng range. A hgher acceptance level can be nterpreted as an (earler) actve move by the frm s decson makers to mprove ts corporate governance system. Ths dstncton n the governance qualty s straghtforward n almost all cases. Second, a smple sum over the The questonnare s avalable from the authors upon request (n German or French).

18 18 attrbutes s computed. Whle such a smple weghtng scheme makes no attempt to accurately reflect the relatve mportance of the ndvdual governance attrbutes, t has the advantage of beng transparent and allows easy nterpretaton. Fnally, the overall CGI s normalzed to have a value between 0 and 100, wth better-governed frms havng hgher ndex scores. The percentage of shares owned by offcers and drectors, Stocksod, s the sum of all shares owned by offcers and executve as well as non-executve members of the board dvded by the total number of shares outstandng. For companes wth more than one share category, whch apples for about 22% of our sample, the ownershp of dfferent share categores s weghted by ther respectve nomnal values. 12 Blockout denotes the percentage of cumulated votng rghts exercsed by large outsde nvestors wth votng rghts exceedng 5%. Bsze s the number of drectors on the board of the frm. Outsder refers to outsde membershp on the board, measured by the percentage of board seats held by drectors wthout any executve functon. LV denotes frm leverage and s calculated as the rato of total (non-equty) labltes to total assets. In addton to the sx endogenous control mechansms, 15 exogenous control varables are ncluded n our system of equatons. ADR s a dummy varable ndcatng f a frm trades Amercan Depostory Recepts (ADRs) n the Unted States. Beta s the market beta estmated by regressng the frm s monthly stock returns over the past fve years on the respectve returns of the market as proxed by the Swss Performance Index (SPI). Frms wth return data not avalable for the full perod of 60 months are not excluded from our sample f we have return data for at least 9 months. Blockonr s the number of outsde shareholders wth a stake exceedng 5% of the frm s equty. Ceop s a dummy varable whch s equal to one f the chef executve offcer (CEO) s also presdent of the board of drectors and zero otherwse. Dv s a dummy varable that s equal to one f the frm pad a dvdend n 2003 (based on the earnngs of 2002) and zero otherwse. Founder s a dummy varable that s equal to one f the CEO or the presdent of the board s the founder of the frm and zero otherwse. Growth s the average annual sales growth over the past three years ( ). Intang s defned as the rato of ntangble to total assets. A hgh rato mplctly ndcates the presence of ntangble 12 Weghtng has to be based on nomnal values because market values are not avalable for all share categores of the frms n our sample. However, Schmd (2003) shows that for the 116 frms of hs total sample of 134 Swss frms for whch he has market values for all exstng share categores, the values of Stocksod are very close to those obtaned by usng nomnal values to wegh the ownershp of dfferent share categores.

19 19 assets lke frm-specfc human captal, technology, or market power. Lnage s the natural logarthm of the number of years snce ncepton of the frm as a prvate lmted company. Frm sze s measured by the natural logarthm of total assets, Lnassets. As a measure of proftablty, we nclude the return on assets, ROA, whch s calculated as operatng proft n 2002 dvded by the average of the 2002 startng and endng value of total assets. Scat s a dummy varable whch s equal to one f the frm has dfferent share categores wth dfferent votng rghts attached and zero otherwse. SMI s a dummy varable that s equal to one f a frm s ncluded n the Swss Market Index comprsng of the 26 largest Swss frms at the end of 2002 and zero otherwse. Sown s a dummy varable, whch s equal to one f the state owns more than 5% of the frm s equty and zero otherwse. Vola s the standard devaton of 60 monthly returns of a frm s stock. Fnally, to control for ndustry effects we nclude 12 dummy varables, Industry. Our measure of frm valuaton s Tobn s Q, alternatvely labeled as Q. As suggested for example by Chung and Prutt (1994), Perfect and Wles (1994), Agrawal and Knoeber (1996), Kang and Stulz (1996), and Loderer and Peyer (2002), among others, Tobn s Q s estmated as the rato of the market value of equty plus the book value of debt to the book value of total assets. To avod that fluctuatons n the market value of the frms equty nfluence our results, we follow Bener, Drobetz, Schmd, and Zmmermann (2003) and Schmd (2003) and compute the market value of equty as the mean of daly observatons durng Unfortunately, for some frms daly stock prce data s not avalable for all share categores on Datastream and had to be replaced by the mean of the 2001 and 2002 year end values of total market captalzaton obtaned from Worldscope. Defntons of all varables employed n ths study are summarzed n Table 1. [Insert Table 1 about here] 4.2 Sample and data descrpton As a startng pont we target all 275 frms quoted at the Swss Stock Exchange (SWX) by the end of The excluson of nvestment companes leaves us wth a sample of 235 frms recevng our questonnare. From these 235 frms 120 returned our questonnare, whch accounts to a response rate of 51.06%. Another 9 frms must be dropped because no data on stock ownershp of offcers and drectors s avalable. Fnally, the excluson of two obvous

20 20 outlers concernng ther value of Tobn s Q and return on assets leaves us wth a sample of 109 frms for our cross-sectonal regresson analyss. Data has been collected from dfferent sources besdes the questonnare and generally refers to the reportng perod from January 2002 to December Stockod has been collected from the 2002 annual reports of the companes covered n ths study. Data for the varables Blockout, Bsze, Blockonr, Ceop, Lnage, Scat, and Sown stem from the webste of Fnanz & Wrtschaft 13 and the Swss Stock Gude 2002/2003. The necessary data to compute Q, LV, Beta, Dv, Growth, Intang, Lnassets, ROA, and Vola were obtaned from Datastream and Worldscope. However, for most varables data was not avalable for all frms n our sample. Mssng values were obtaned drectly from the companes annual reports. We dentfed whether a frm trades ADRs on the NYSE, AMEX, or NASDAQ usng the JP Morgan webste. 14 SMI consttuents are checked from the webste of the Swss Stock Exchange (SWX). Fnally, the varables CGI, Industry, Founder, and Outsder are based on the questonnare answers. Table 2 shows descrptve statstcs of all varables ncluded n our analyss. The average value of Tobn s Q s 1.34 and the medan s 1.06, ndcatng that Swss frms, on average, nvest n postve NPV projects. The dstrbuton of our corporate governance ndex s dsplayed n Fgure 1. The mean of CGI s and the medan 59.21, ndcatng a relatvely symmetrc dstrbuton. Addtonally, Fgure 1 reveals that there are substantal dfferences n frm level corporate governance between the 109 frms n our sample (the mnmum value s and the maxmum value 90.13). Ths suggests that our governance proxes are adequately selected to reach a suffcently wde dstrbuton, whch mtgates a possble selecton bas n our results. Fnally, Fgure 1 shows that, as expected, SMI frms (n dark grey) have sgnfcantly hgher values of CGI than the other frms n our sample. As Table 2 reveals, offcers and drectors hold on average 12.10% of the equty of a frm. However, the medan of 0.60% s much smaller, ndcatng that there are some frms n our sample where offcers and drectors hold very large fractons of total equty. A comparson of these values to the sample of US frms used by Loderer and Martn (1997) confrms that aver- 13 The webste of Fnanz und Wrtschaft, Swtzerland s major fnancal newspaper, s: 14 The webste s:

21 21 age nsder shareholdngs are smlar n Swtzerland and the US. However, the medan s a lot smaller n our sample and, hence, nsder shareholdngs are much more skewed n Swtzerland. The mean of Blockout s 32.74%. The average board sze s 7.35 n our sample, whch roughly equals the optmal sze of seven or eght drectors, as has been hypotheszed by Lpton and Lorsch (1992) and Jensen (1993). Ths s only about half of the average board sze reported by Yermack (1996) for US frms, but larger than that reported by Esenberg, Sundgren, and Wells (1996) for ther Fnnsh sample. The average leverage rato of 62.34% s very smlar to the values reported by Bener, Drobetz, Schmd, and Zmmermann (2003), Schmd (2003), and Drobetz and Fx (2003) for Swtzerland and Peasnell, Pope, and Young (2003) for the UK. The average value of Outsder s 89.27%, whch strongly dffers from the much lower values of 54% and 60% reported by Yermack (1996) and Barnhart, Marr, and Rosensten (1994), respectvely, for US companes and 44% reported by Peasnell, Pope, and Young (2003) for UK companes. Ths fndng s especally surprsng because foundng famles are stll regarded as an mportant factor n corporate Swtzerland. Table 2 further shows several other nterestng results, whch we only brefly summarze: about 22% of the frms n our sample have more than one share category, n 15% of the frms the CEO s also presdent of the board, n 10% of the frms the CEO or the presdent of the board founded the frm, about 11% of the frms trade Amercan Depostory Recepts (ADRs) n the Unted States, and the state s a sgnfcant shareholder of about 12% of the frms. [Insert Table 2 about here] [Insert Fgure 1 about here] Table 3 shows the correlaton coeffcents between Tobn s Q and the sx control mechansms. Of specal nterest, and consstent wth our general noton, s the postve correlaton coeffcent of 0.24 between our corporate governance ndex, CGI, and Tobn s Q. In fact, ths s the hghest postve correlaton coeffcent between Tobn s Q and any other varable. Stocksod and Bsze are also postvely correlated wth Tobn s Q. In contrast, Blockout, LV, and Outsder are all negatvely correlated wth Q. Clearly, the negatve correlaton coeffcent between Stocksod and Blockout occurs by constructon and, hence, should not come as a surprse. It s also mportant to note that the CGI s a proxy whch ncorporates a broad range of governance ssues. However, as already mentoned above, t does not nclude the addtonal