Product Market Competition, Managerial Incentives, and Firm Valuation

Transcription

1 Product Market Competton, Manageral Incentves, and Frm Valuaton Stefan Bener Markus Schmd Gabrelle Wanzenred 04-1 July 004 Dskussonsschrften Unverstät Bern Volkswrtschaftlches Insttut Gesellschaftstrasse Bern, Swtzerland Tel: 41 (0) Web:

2 Product Market Competton, Manageral Incentves, and Frm Valuaton Stefan Bener a, Markus Schmd b, and Gabrelle Wanzenred c, * Ths verson: July 004 Abstract Our paper s a further contrbuton to the stll very small emprcal lterature on the effects of competton on manageral ncentve schemes. Based on a theoretcal model that ncorporates both strategc nteracton between frms and a prncpal agent relatonshp, we nvestgate the relatonshps between product market competton, ncentve schemes and frm valuaton n a multvarate OLS framework. To take nto account a possble endogenety of the ncentves provded to managers and Tobn s Q, our measure of frm value, we addtonally employ a smultaneous equatons framework. The sample comprses 156 Swss frms for the year 00, and the compensaton data refer to the frms top management and the board of drectors n total. Our results suggest that more ntensve product market competton s assocated wth stronger ncentve schemes for managers and a lower frm value. Key Words : Product Market Competton; Manageral Incentves; Frm Valuaton; Smultaneous Equatons System JEL Classfcaton: C31, J33, L1 a Stefan Bener, Saïd Busness School, Unversty of Oxford, Park End Street, Oxford OX1 1HP UK, and Department of Fnance, Unversty of Basel, Holbenstrasse 1, 4051 Basel, Swtzerland, Mal: stefan.bener@unbas.ch. b Markus Schmd, Department of Fnance, Unversty of Basel, Holbenstrasse 1, 4051 Basel, Swtzerland, Phone: , Mal: markus-max.schmd@unbas.ch. c Gabrelle Wanzenred, Haas School of Busness, Unversty of Calforna Berkeley and Department of Economcs, Unversty of Bern, Mal: gabrell@haas.berkeley.edu. * We thank Stefan Duffner, Ken Okamura and Edth Bernhard for support wth data provson. Bener acknowledges fnancal support from the Swss Natonal Scence Foundaton (SNF). All errors reman our responsblty. 1

3 1. Introducton What are the effects of product market competton on manageral ncentves? Do managers work harder when the frm s envronment s more compettve,.e. do competton and ncentve schemes substtute each other? And what are the mpacts on frm value? Whle these and related questons are at the heart of an ongong debate about corporate governance ssues, the underlyng mechansms are only partly understood, and there s a serous lack of emprcal evdence on these ssues. The effects of competton on ncentve schemes and frm valuaton are not only nterestng from a purely academc pont of vew. These ssues are also hghly relevant for publc polcy makers. Durng the last decade, there has been an ncreasng nfluence of governments and non-governmental organzatons on corporate governance rules. 1 As Kheman and Leechor (001) outlne, much of the attenton has focused on the frms and the regulatons that protect shareholder rghts and govern the conduct of management. However, the envronment n whch busness s conducted, such as the degree of competton among frms, entry and ext rules, and the openness of the economy, requres close consderaton. Competton s needed for a culture of good corporate governance to thrve. Competton polcy helps to ncrease effcency, reduce prce dstortons, lower the rsk of poor nvestment decsons, promote greater accountablty and transparency n busness decsons, and lead to better corporate governance. Consequently, the desgn of effectve corporate governance rules necessarly has to take nto account the compettveness of markets. Whle questons about the role of publc polcy are clearly beyond the scope of our paper, we should keep n mnd that there may exst mportant nteracton effects between competton and corporate governance rules. The theoretcal lterature on the lnks between product market competton and manageral ncentves can bascally be dvded nto two man strands. A frst strand analyzes the effects of product market competton on manageral ncentves, but compensaton contracts are not allowed to affect competton. Whle the earler lterature nformally argues that competton reduces manageral slack (e.g., Machlup, 1967), Hart (1983) s the frst to formalze ths dea by modelng the effect of competton on the agency problems between a frm s owner and a manager. Subsequent research shows, however, that the relatonshp between competton and managers effort level s ambguous (e.g., Scharfsten, 1988; Hermaln, 199; and Grazano 1 The OECD s about to revse ts Prncples of Corporate Governance that were adopted n In the USA, the Sarbanes-Oxley Act, whch renforces the frms transparency requrements among others wth respect to executve compensaton, came nto power n 00. In Germany, the German Corporate Governance Code, a smlar set of transparency rules that s however not compulsory, has recently been mplemented by the German government. In Swtzerland, the Swss Code of Best Practce has become effectve n 00. The authortes of the European Unon dscuss the adopton of a corporate governance codex. See also Emons and Wanzenred (003).

4 and Parg, 1998). Whle these studes rely on the nformaton effect of competton, whch means that competton nduced by many frms n the market may gve more precson to ncentves based on relatve performance evaluaton, Schmdt (1997) uses the dea that more competton ncreases the probablty of frms gong bankrupt. He shows that the effects of competton on managers effort level and the strength of ther ncentve schemes crucally depend on managers outsde optons. In partcular, an ncrease n the product market competton s more lkely to result n stronger ncentves n case managers have good outsde optons. Rath (003) examnes how the degree of competton among frms n an ndustry wth free entry and ext affects the ncentves for ther managers. Such a setup wth free entry and ext mples that changes n the nature of competton lead to changes n the equlbrum market structure. Hs results suggest an unambguous postve relatonshp between competton and ncentves. The second strand of the theoretcal lterature on competton and ncentves s based on the dea that precommtment to manageral ncentve contracts can alter the strategc competton between rvals. 3 Aggarwal and Samwck (1999) extend the lterature by consderng compensaton contracts based on relatve performance evaluaton. 4 The emprcal papers that relate product market competton to compensaton are not very numerous and are mostly n lne wth Aggarwal and Samwck (1999), Keda (003) and Joh (1999). These studes explctly take nto account strategc nteractons and the structure of product markets to explan manageral compensaton contracts. In partcular, they use these aspects to address the relatve performance evaluaton puzzle, whch s the fact that emprcal studes do not seem to fnd any role for relatve performance evaluaton n ncentve contracts. Another recent paper by Funk and Wanzenred (003) provdes some evdence for Schmdts (1997) hypothess that the relatonshp between competton and strength of ncentves depends on the managers outsde optons. Fnally, Cuñat and Guadalupe (004) study the mpact of product market competton on the compensaton packages n the US bankng and fnancal ndustry. Scharfsten (1988) reconsders Hart s model whle relaxng the assumpton of nfntely rsk-adverse managers. Hermaln (199) consders addtonal effects of competton on the agency problem, all of whch are of potentally ambguous sgn. Therefore, he concludes that theory cannot offer a defntve answer to the queston of whether competton reduces manageral slack. Grazano and Parg (1998) analyze the relatonshp between product market competton and manageral effort n a lnear prncpal agent model. Whle ncreasng competton stemmng from a lower degree of product market dfferentaton reduces the manager s optmal effort level and the optmal pece-rate, an ncrease n the number of frms has an ambguous effect on effort and pece-rate. 3 See, e.g., Vckers (1985), Fershtman and Judd (1987), Sklvas (1987), and Fumas (199). 4 They examne compensaton contracts for managers n mperfectly compettve product markets and show that strategc nteractons among frms can explan the lack of relatve performance-based ncentve schemes for whch compensaton decreases wth rval frm performance. They fnd that frms n more compettve ndustres place more weght on rval frm performance relatve to own frm performance. Ther study s one of the very few papers that emprcally test the relatonshp between ncentves and competton. 3

5 There s very few emprcal evdence on the relaton between product market competton and frm value. Grffth (001) argues on page 1 that the drecton of the effect that product market competton should have on frm value s ambguous: On the one hand ncreasng competton lowers frm s profts and thus reduces ncentves to exert effort (the Schumpeteran effect), on the other hand t reduces agency costs (or ncreases the rsk of bankruptcy) thus ncreasng ncentves to exert effort. However, the emprcal lterature s manly concerned wth the effect of product market competton on productvty growth nstead of frm value. For example, Nckel et al. (1997) fnd that product market competton has a postve mpact on total factor productvty. 5 Consstently, Januszewsk et al. (00) fnd for a sample of almost 500 German frms n the manufacturng ndustry that frms experence hgher productvty growth when operatng n markets wth ntense competton. Grosfeld and Tressel (001) analyze the mpact of competton and corporate governance on frm performance for frms lsted on the Warsaw Stock exchange. They fnd that product market competton and good governance tend to renforce each other. 6 One excepton s Habb and Ljungqvst (004) nvestgatng the effect of product market competton, as measured by a Herfndahl ndex based on four-dgt SIC codes, on frm value. They provde evdence that frm value s postvely related to product market competton. The am of our paper s to contrbute to the stll very small emprcal lterature on the relatonshp between product market competton and manageral ncentves. Based on a theoretcal model that ncorporates both strategc nteracton between frms and a prncpal agent relatonshp, we use a multvarate OLS framework to emprcally nvestgate the relatonshps between product market competton, ncentve schemes, and frm valuaton for Swss companes. We consder a prncpal-agent model n a Cournot olgopoly setup. Such a setup not only takes nto account the classcal moral hazard problem wthn the frm, whch s nduced by the unobservablty of the manager s effort, but t also ncorporates strategc nteracton between the frms. The model s structured as follows: At stage one, the frm owner hres a manager whose task s to reduce the frm s costs. At stage two, the manager decdes on hs unobservable effort level. At the last stage, fnally, the frms compete wth each other on output markets. The theoretcal predctons of the model are threefold. Frst, the relatonshp between the strength of the ncentve scheme and the ntensty of competton depends on the 5 See, e.g., Nckell et al. (1997) for a summary of emprcal evdence on the effect of product market competton on productvty performance. 6 See, e.g., Carln and Horvath (000) for a summary of emprcal evdence on the mpact of competton on frm performance n transton economes. 4

6 absolute level of competton. For low levels of competton, more competton leads to weaker ncentves. For hgher levels of competton, however, a hgher ntensty of competton results n stronger ncentves. Second, the relatonshp between the ntensty of competton and the strength of the ncentve schemes ncreases n the level of competton. Thrd, the effect of competton on frm value s negatve, meanng that frms n more compettve envronments realze lower profts. Our data comprses nformaton about frm characterstcs and top management s and board of drectors compensaton of 156 frms quoted at the Swss Exchange (SWX) for the year 00. In addton, we use three dfferent varables for measurng the ntensty of competton on product markets, namely one frm-specfc and two ndustry-specfc measures. The frmspecfc measure s based on frms rents from producton and other busness actvtes. The two ndustry-specfc competton measures are a sales-based Herfndahl ndex and an ndcator that measures the ntensty of regulatons for enterng a new ndustry. We use a multvarate OLS regresson analyss to nvestgate the relatonshp between product market competton, manageral ncentves and frm value. Based on Agrawal and Knoeber (1996) and Bener et al. (004), we addtonally apply a smultaneous equatons framework n order to take nto account a potental endogenety problem. The emprcal results reveal that n general a more ntense product market competton s assocated wth stronger ncentve schemes for managers as measured by the fracton of sharebased to cash compensaton. Ths result s consstent wth the frst hypothess of our theoretcal model and suggests that frms are operatng n compettve envronments on average. Moreover and consstent wth our second hypothess, we fnd the postve nfluence of competton on ncentve schemes to be stronger for frms operatng n a hgh competton envronment. Fnally and consstent wth the thrd hypothess of our theoretcal model, we fnd a postve but not statstcally sgnfcant effect on frm value for all three competton proxes ndcatng that a hgher product market competton s assocated wth a lower frm value. Thus, the negatve effect of lower economc rents seems to outwegh the postve effect of reducng manageral slack and ncreasng the managers effort by provdng addtonal montorng and ncreasng the threat of lqudaton. The new aspects of our paper are the followng ones: It s the frst study that provdes emprcal evdence on the effects of competton on manageral ncentves and frm valuaton for Swss companes. Furthermore, our emprcal framework based on a smultaneous equatons system suggests a way to deal wth the potental endogenety problem between 5

7 competton, ncentves and frm valuaton. Fnally, our theoretcal prncpal-agent model explctly takes nto account the strategc nteractons between frms wthout relyng on the nformaton effect of competton and the relatve performance evaluaton. Ths feature of the model allows us to formulate our hypothess n terms of observable varables, whch makes t partcularly attractve for emprcal tests. The paper s structured as follows. The theoretcal model and our man hypotheses are n secton. Secton 3 descrbes the data. The emprcal analyss s n secton 4, and secton 5 comprses some robustness tests. Secton 6 concludes.. Theoretcal model and man hypotheses.1. The setup The purpose of our model s to nvestgate the effect of product market competton on the ncentve schemes for managers and the value of the frm when there are strategc nteractons between the market players. 7 We consder a prncpal-agent model wthn a Cournot olgopoly setup, where the owner of the frm hres a manager to reduce margnal costs. In contrast to other work, our model nether reles on the nformaton effect of competton, nor on relatve performance evaluaton, whch both mpose rather strong constrants n terms of observablty of certan varables. The model has three stages. At stage one, the owner of frm hres a manager who has to reduce the costs of the frm. At stage two, the manager provdes effort that affects the frm s margnal producton cost. The manager s effort decson s unobservable to the owner, who can only observe the cost reducton. At stage three, the owner decdes on the output level, profts are realzed and the manager gets pad. Each frm has constant margnal costs gven by c = ( c e u ), where c s a constant, e s the effort level exerted by the manager, and u s a random term that s assumed to be normally dstrbuted wth zero mean, varance σ, and s ndependent of the other frms shocks. The manager s effort level s not observable. The owner of the frm can only observe the realzed costs c, whch are also contractble. There are no fxed costs. The owner of the frm offers the manager a lnear compensaton scheme that s a functon of the observed cost reducton,.e., 7 Our model s smlar to Rath (003), who consders the effects of competton on ncentves wthn a crcular cty model wth an endogenous determnaton of market structure. 6

8 w = α + β ( c c ) (1) The parameter α denotes the fxed part of the salary, and β s the pece rate that tes the manager s wage to the performance of the company, and ( c c ) s the observed cost reducton. Gven that the cost reducton affects the proftablty of the frm, we can nterpret β also as pay-for-performance senstvty. 8 The utlty of the manager s gven by exp { r[ w g( )]} e manager s degree of rsk averson, whch we assume to be constant, and dsutlty of exertng effort, wth, where r, wth r>0, s the s hs k >1. The expected value of the manager s wage s thus ge ( ) = ke / α + β wth varance of e manager can be wrtten as n (),.e., β σ. Gven the normal dstrbuton of u, the utlty of the U rβ σ ke = α + βe () The manager accepts any contract ( α, β ) that gves hm an expected utlty of at least hs reservaton utlty, whch we normalze to zero. The nverse demand functon of frm s gven by (3): where a, wth p( q, q ) = a bq d jqj, j = 1,..., N, j (3) j a > 0 and a > c, s the sze of the market, b s a postve constant, and q s frm s output. The varable q j s the output of frm s rval j. The coeffcent, wth 0 < dj <b, captures the degree of product dfferentaton between the products on the market. The larger d, the closer substtutes the products are. The parameter d s commonly used to j measure the degree of competton n a market, where hgher values mply a more ntensve competton. 9 To keep thngs smple, we set b=1 and d = d = d, j. We further assume that there are only two frms n the market. We are lookng for the subgame perfect equlbrum of the game. Therefore, we solve the model by backwards nducton. j j d j 8 See, e.g., Jensen and Murphy (1990) and Murphy (1999). 9 See, e.g., Grazano and Parg (1998). 7

9 .. The frm s output decson At t=3, the frms smultaneously choose ther output levels. The proft of frm gross of manageral compensaton s gven by (4). π = ( p c ) q = ( a q dq c ) q,, j = 1, (4) j From maxmzng (4) wth respect to q and solvng for q we get the frms reacton functon,.e., a c de( qj) q( qj) =,, j = 1, (5) If frm s rval s expected to set a quantty of Eq ( j ), the resultng proft of frm s as n (6). ( a c de( qj)) π( c, E( qj)) =,, j = 1, (6) Smultaneously solvng the system of two equatons as gven by (5) yelds the equlbrum quanttes of the thrd stage as a functon of the frm s own margnal costs expected costs ( ) Ec j,.e., c and the rval s ( a c ) ( ( )) *** d a E c j q =,, j = 1, (7) (4 d ) From substtutng (7) nto (3) we obtan the equlbrum prce gven by (8) and can compute the expected gross profts as gven by (9): p *** j ( a+ c ) d( a E( c )) d c =,, j = 1, (4 d ) (8) ( ) ( ( )) *** a c d a E c j π =,, = 1, (4 ) j (9) d.3. The manager s effort decson At t=, the manager of frm chooses hs effort level by maxmzng hs utlty gven n (): 8

10 1 k max U = α + βe rβ σ e, j = 1, (10) e Dfferentatng (10) wth respect to e yelds the effort level as a functon of the compensaton parameter β,.e., β e( β ) =, j = 1, (11) k The ndvdual ratonalty constrant (IRC) of the manager s gven by 1 k α + βe rβ σ e 0, j = 1, (1) where the manager s outsde utlty s normalzed to zero. Assumng compettve labor markets, the (IRC) s bndng, whch also means that (1) holds wth equalty. Ths allows us to calculate the fxed salary component α the manager has to be pad n order to have a reservaton utlty of zero. krσ ** β (1 ) α ( β) =, j = 1, (13) k The manager s wage as a functon of β s then gven by ** β (1 krσ ) β β w ( ) = + ( c c ), j = 1, (14) k.4. The optmal ncentve scheme At the frst stage of the game at t=1, the owner of the frm chooses the ncentve scheme for the manager. He maxmzes hs expected proft net of manager s wage, whch s gven by (9) mnus (14). Usng ( c c ) = e + u, e( β) = β / k, and Eu ( ) = 0, the net expected proft s gven by (15). β ( a ( c )) + d( E( c j ) a) (1 ), ( ) k β krσ β π net β = + β,, j = 1, (15) (4 d ) k k Dfferentatng (15) wth respect to β and solvng for β leads to 9

11 4k(( a c) + d( E( c j ) a)) β =,, j = 1, (16) ( k + k rσ )(4 d ) 8 In a symmetrc equlbrum, all frms choose the same pece rate β, and each manager chooses the same effort level e. Accordngly, E( c j ) = c e = c β / k. Substtutng ths expresson nto (16) and solvng for β leads to (17), the optmal ncentve parameter. * 4k( a c) β = (17) 3 ( k + rσ k )( d d + 4d + 8) 4 To fnd the equlbrum quantty and proft net of manageral compensaton, we plug (17) nto the correspondng second-stage equlbrum values, whch yelds the followng results: q * ( k + rσ k )(4 d = 3 ( k + rσ k )( d d )( a c) + 4d + 8) 4 (18) π * net (1 + = rσ k)( a c) k[ ( k rσ + k)( d 4) 8] 3 [( k rσ + k)( d + d 4d 8) + 4] (19).5. The effects of competton on the strength of ncentve schemes and frm value How does competton affect the optmal pay-for-performance senstvty β * and frm value? Followng Grazano and Parg (1998), we use the degree of product dfferentaton d as a proxy for the ntensty of competton. The larger d, the closer substtutes the products are, and the hgher the ntensty of competton. As to frm value, we look at the proft net of manager s compensaton. From dfferentatng the optmal pay-for-performance senstvty as gven by (17) wth respect to competton measure d we obtan * β d = 4k ( a c)( d + )(3d )( krσ + 1) 3 [( k rσ + k)( d + d 4d 8) + 4] (0) To obtan the sgn of ths expresson, we only need to look at the numerator snce the denomnator s always postve. Gven that a > c by assumpton, ths expresson s postve 10

12 ff d > /3. It follows that the owner of the frm more closely tes the manager s wage to the performance of the company once the ntensty of product market competton has reached a certan level. Ths leads us to our frst hypothess. Hypothess 1: A hgher ntensty of product market competton, as measured by the degree of product dfferentaton d, leads to stronger ncentve schemes for the manager n case the ntensty of product market competton has reached a certan level,.e., * β d > 0 for d>/3. Obvously, there are dfferent effects at work. Frst, there s a busness stealng effect: a hgher value of d mples a more elastc demand, whch makes t easer for a frm wth a cost advantage to take away busness from ts rval. Accordngly, for a gven quantty of ts rval, a more ntensve competton ncreases a frm s margnal beneft of reducng ts costs. Gven ths frst effect, the frm wants to gve stronger ncentves to ts manager wth ncreasng competton, leadng to lower margnal costs. However, there s a second effect at work that can be denoted as a scale effect: a hgher value of d also leads to a fall n frm s output. 10 Ths decreases the frm s gan from reducng ts costs and leads the frm to gve weaker ncentves to the manager when competton s ncreasng. Whle the second effect, the scale effect, s domnatng for lower values of d, the formerly descrbed busness stealng effect starts to domnate once the degree of competton has reached a certan level,.e., for values of d>/3. Accordngly, for lower values of d, the ncentve parameter β s decreasng when the ntensty of competton s ncreasng; for hgher values of d, n contrast, the ncentve parameter β s ncreasng n the ntensty of competton parameter. In other words: For values of d>/3, frms provde stronger manageral ncentves because greater competton ncreases the value of makng good decsons (.e., the busness stealng effect ncreases). To understand the underlyng mechansms from a formal pont of vew, we can look at frm s margnal gan of reducng ts costs,.e., let us dfferentate (9) wth respect to c : π *** c 4 ( a c) d( a E( cj)) = ( d ) ( d + ) (1) In a symmetrc equlbrum, expresson (1) s clearly negatve. Ths reflects the fact that the frm can ncrease ts proft by lowerng ts costs. To see how the margnal proft of a cost reducton moves wth the ntensty of competton, whch s really our man nterest, we go 10 Ths can best be seen by dfferentatng the equlbrum output level as gven by (18) wth respect to d, whch s clearly negatve. From an economc pont of vew, the wllngness to pay for the product of frm decreases wth a hgher value of d,.e., the closer substtutes the products are. As we can see from frm s reacton functon as gven by (5), a hgher value of d leads to a lower output for frm. Ths s to compensate the fall n profts due to the lower prce. 11

13 one step further and dfferentate (1) wth respect to the degree of product dfferentaton d that yelds (): *** π c 4 (3d 4)( E( cj) a) 8 d( a c ) + + = 3 3 d ( d ) ( d + ) () In a symmetrc equlbrum expresson () s postve for d</3, whereas () s negatve for d>/3. A postve sgn of () means that the margnal proft of a cost reducton, whch s a negatve value, becomes less negatve and thus smaller n absolute terms when d s ncreasng. Ths reflects the fact that the scale effect s domnatng and the frm lowers the ncentve parameter β when the ntensty of product market competton s ncreasng. The negatve sgn of expresson () for d>/3, n contrast, mrrors the domnance of the busness stealng effect: The margnal proft of a cost reducton becomes larger n absolute terms wth a hgher ntensty of competton d, and ths nduces the frm to gve stronger ncentves to ts manager. To see how the relatonshp between the ncentve parameter β and d changes wth dfferent levels of competton, we go another step further and dfferentate (0) wth respect to d,.e., * β d d 16k = ( a c 3 4 )(1 + krσ )[( k rσ + k)(16 + 8d + 3d ) 4 6d ] 3 [( k rσ + k)( d d + 4d + 8) 4] 3 (3) Expresson (3) s postve, whch also means that the margnal effect of competton on the ncentve parameter becomes stronger wth ncreasng competton. To see ths, we only need to look at the square brackets n the numerator snce all other expressons are postve. Wthn the square bracket, the product s always equal to or bgger than 16, snce k s equal to or bgger than one and r s postve. Therefore, the consdered expresson n the square brackets s postve for all values of d (to remember: 0 d 1). Ths expresson becomes even larger wth hgher values of k. These consderatons lead us to our second hypothess: Hypothess : The margnal effect of competton on the ncentve parameter β ncreases wth the ntensty of product market competton, as measured by the degree of product ( β ) dfferentaton d,.e., d > 0 d * for d. As to the effect of competton on frm value, we dfferentate the net proft as gven by (19) wth respect to d, whch yelds (4). 1

14 π d * net = (1 + rσ k) k ( a 4 3 c) ( d + ) [( k rσ + k)( d 4d + 16d 16) 8d + 8] 3 [( k rσ + k)( d d + 4d + 8) 4] 3 (4) Expresson (4) s clearly negatve: From before, we know that the denomnator s always postve. As to the numerator, we only need to look at the square brackets snce all other expressons are postve. Wthn the square bracket, the product s always equal to or bgger than -16, snce k s equal to or bgger than one and r s postve. Snce k > 1 and -8d + 8 s equal or smaller than 8 for all values of d, the expresson n the square bracket s always negatve. These consderatons lead us to our thrd hypothess. Hypothess 3: A hgher ntensty of product market competton, as measured by the degree of product dfferentaton d, leads to a lower net proft,.e. * net π d < 0 for d. The explanaton of ths result s straghtforward and stands n lne wth standard olgopoly models. The closer we move to perfect competton n terms of havng more homogenous products, c.p., the lower the profts of the frms are. The hypotheses derved from our theoretcal model are subject of our emprcal tests n sectons 4 and Data and sample 3.1. Defnton of varables Product market competton In ths secton we provde a detaled descrpton of the varables we use n our emprcal analyss. Smlar to Nckell (1996), Nckell et al. (1997) and Grosfeld and Tressel (001), our standard measure of product market competton s a frm s rents from producton and other busness actvtes, Rents, whch can be nterpreted as an ex-post measure of market power. The motvaton for usng ths measure s that frms operatng n less compettve markets should be able to sell ther products well above margnal costs and, therefore, earn hgher rents after coverng ther expenses. We defne Rents as profts before nterest payments, tax, and deprecaton (EBITDA) mnus the costs of captal (cc) multpled by total assets (TA) and standardzed by the company s sales (SA): ( EBITDA cc TA) SA Rents = /. 13

15 The costs of captal (cc) are defned as follows: f ( r r ) cc = r + δ + λ β, m f where r f s the rsk free rate, δ s the rate of deprecaton, λ s equal to the equty rato of the frm, β s the estmated market beta of the frm s stock, and r m s the return to a broad market ndex. The rsk free rate s calculated as the average one month Swss Interbank Rate over 60 monthly values from January 1997 to December 001 and amounts to 1.9%. Followng Nckell (1996), the deprecaton rate s assumed to be constant at 4 percent. 11 The equty rato, λ, s calculated as 1 mnus the rato of total (non-equty) labltes to total assets. 1 The market beta, β, s 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). 13 The rsk premum s equal to the average return of the Pctet-Rätzer Index, a broad Swss stock market ndex, less the average short-term nterest rate (the one month Swss Interbank Rate). The man drawback of ths type of measure of ex-post monopoly power s that t s clearly strongly correlated not only wth market power but also wth proftablty, whatever the precse defnton chosen (see also Nckell, 1996). Snce we analyze the mpact of product market competton on frm valuaton and frm value s expected to be postvely correlated wth proftablty, we may obtan a postve bas n our results. In fact, as our emprcal results n Secton 4 reveal, Rents affects frm value postvely. To mtgate ths potental bas, we control for the nfluence of proftablty on Tobn s Q by ncludng an alternatve measure, the return on assets, nto our regresson analyss. In addton to the frm-specfc competton ndcator Rents, we employ two alternatve ndustry-level measures of product market competton that are presumably not afflcted wth ths problem. Specfcally, we use a sales-based Herfndahl ndex, Herf, and an ndcator measurng the level of regulaton for new ndustry entrants, Reg, as addtonal proxy varables for competton. Herf s calculated as follows: 11 Alternatvely, to test the robustness of our results to ths assumpton, we apply a second measure of rents based on a rate of deprecaton of 8 percent, labeled as Rents8. The results change only mmaterally (see Tables 8 and 9 n the robustness secton). 1 Followng Nckell (1996), we also apply an alternatve measure of rents, RentsL1, where λ s set equal to one. The results reman bascally unchanged as Tables 8 and 9 n the robustness secton reveal. 13 For frms wth more than one share category all varables related to stock return data are weghted based on nomnal values. Frms wth return data not avalable for the full perod of 60 months are not excluded from our sample f return data could be obtaned for at least 9 months. 14

16 N j N j Herf = SA j SA j, = 1 = 1 where SA j s the sales attrbutable to frm n ndustry group j, where ndustry groups are based on the classfcaton of the Swss Exchange (SWX). Each ndustry group comprses all quoted Swss frms and not only the frms n our sample. As argued by Aggarwal and Samwck (1999), the Herfndahl ndex can be used as a proxy for product substtutablty. Of course, there are also some problems assocated wth the use of Herf as a measure of market power. Frst, Herf does not take nto account foregn compettors, a problem, whch s lkely to be especally severe n a small open economy as Swtzerland. Second, our classfcaton of ndustry groups s arbtrary and may not represent anythng lke the relevant product market for the frms ncluded n the respectve ndustres. 14 Thrd, actual as well as potental competton nfluences the market power of frms wthn an ndustry and Herf does clearly not take nto account the latter. However, Rents s supposed to be much less afflcted wth these problems. Our thrd competton proxy, Reg, measures the ntensty of legal regulatons for frms enterng a new ndustry. Ths ndcator s unque n the sense that s only avalable for Swtzerland n ths form. It goes back to a study that was commssoned by the Swss government and s descrbed n detal n Ledergerber et al. (1998). The purpose of the study was to buld a smple ndcator that measures the ntensty of regulatons by law area and by ndustry n Swtzerland. The ndustry classfcaton ncludes 3 ndustral sectors and s based on the NOGA Industry Classfcaton System. 15 Wthout gong too much nto detals, the constructon of the ndcator can be descrbed as follows: In a frst step, the relevant regulatons mposed by federal law were classfed nto the fve categores Informaton Rules, Qualtatve Standards, Quanttatve Standards, Permts, and Implementaton by Cantons. Ths categorzaton goes back to Ogus (1994) and reflects the strength of regulatons, n ascendng order, and the mpled costs for frms. The category Informaton Rules s the weakest form of nterventons, whereas Permts represents the strongest regulaton category. 16 In a second step, the number of regulatons by law area and by regulaton category was derved. In addton, the ndustres affected by each regulaton were dentfed. Fnally, the ndcator was bult by frst allocatng specfc weghts to the number of regulatons by categores, where 14 To nvestgate ths ssue, we use an addtonal varaton of Herf, whch s based on an alternatve ndustry defnton, n the robustness secton. In fact, we fnd some of the results to depend on the precse defnton of Herf. 15 NOGA stands for Nomenclature Générale des Actvtés Economques. 16 The category Implementatons by Cantons was ntroduced to ft the specfc characterstcs of the Swss system. It s parallel to the other categores and does not represent the strongest form of regulaton. 15

17 stronger regulatons got larger weghts, and then by buldng the sum of these weghted numbers. The hgher the value of the ndcator, the stronger s the mpact of regulatons for frms. The regulaton ndcator Reg that we use n our study s a varaton of the general ndcator as descrbed above. It s based on a smaller set of regulatons that are specfcally relevant for market entry. The weakest forms of nterventons are categorzed under Market Regulatons, whch nclude actvtes related to certfcatons, regstratons or type tests of new products. The second category refers to Permts the frms have to apply for by the governmental authortes. The fnal category s denoted by Professonal Lcences, whch are a recognzed certfcaton of professonal capabltes. The hgher the value of the ndcator Reg, the stronger s the mpact of the regulatons for frms enterng a new market. Gven that these regulatons are barrers to entry, fewer frms are expected to enter n hghly regulated ndustres. Therefore, we nterpret a hgh value of Reg as a low ntensty of product market competton. Note that we dvded the orgnal values as reported n the study by Ledergerber et al. (1998) by Also, the nformaton refers to the tme perod before We are aware of the fact that ths ndcator s not perfect, and t s vulnerable to all sorts of crtcsm. The weghtng procedure, for nstance, s not based on some objectve crtera and s therefore rather arbtrary. Furthermore, the analyss s lmted to federal law only, and the consdered regulatons manly refer to small and medum companes. Nevertheless, ths regulaton ndcator s stll useful n the sense that t gves us some ndcaton about the costs assocated wth enterng a new ndustry or market, and ths type of nformaton s elsewhere hardly avalable. Snce all three proxy varables for product market competton measure dfferent aspects of competton and, hence, are afflcted wth other problems, t makes perfectly sense to nclude all three varables smultaneously nto the emprcal nvestgatons. To cope wth a potental endogenety problem related to our competton measures, we use lagged values for Rents, Herf, and Reg. That s, n contrast to the other varables, whch n general refer to the reportng perod from January 00 to December 00, our competton proxes are based on 001 data The results obtaned by usng 00 data for the competton proxes are reported n Tables 8 and 9 n the robustness secton. 16

18 3.1.. Measurng ncentves for managers To measure the ncentve schemes provded to managers, we use the percentage value of shares alloted n 00 to the frm s offcers and drectors n total relatve to cash compensaton pad durng the same year, Srato. Besdes the fracton of share-based to cash compensaton to the frm s offcers and drectors, ths paper consders four addtonal corporate governance mechansms, whch are assumed to provde ncentves to managers and therefore allevate the agency problems between managers and shareholders (see Bener et al., 004). Stocksod s the sum of all shares owned by offcers and executve as well as nonexecutve members of the board dvded by the total number of shares outstandng. 18 Blocko denotes the percentage of cumulated votng rghts exercsed by large outsde nvestors wth votng rghts exceedng 5%. Outsder refers to outsde membershp on the board, measured by the percentage of board seats held by drectors wthout any executve functon. Leverage denotes frm leverage and s calculated as the rato of total (non-equty) labltes to total assets. Besdes these corporate governance mechansms, we employ eght dfferent control varables n ths paper. Frm sze s measured by the natural logarthm of total assets and s labeled Lnassets. As a measure of proftablty, we nclude the return on assets, ROA, whch s calculated as operatng proft n 00 dvded by the average of the 00 startng and endng value of total assets. Pgrowth s the average annual sales growth over the past three years (000-00). As t s standard n the lterature on the relatonshp between pay and performance, we use the change n shareholder value, CSV, as a measure of frm performance n our nvestgatons related to Srato. Followng Jensen and Murphy (1990) and Funk and Wanzenred (003), we defne CSV as the return on equty multpled by the market value of equty n the prevous perod. Stdv s the standard devaton of 60 monthly returns of a frm s stock. 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). 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. To control for ndustry effects, we nclude 9 dummy varables, labeled Industry, whch are equal to one f the frm belongs to a partcular economc sector based on the classfcaton of the Swss Exchange (SWX) and zero otherwse. 18 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 (004) shows that for the 116 frms of hs total sample of 145 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. 17

19 Fnally, our measure of frm valuaton s Tobn s Q, alternatvely smply labeled as Q. As suggested by Chung and Prutt (1994), Perfect and Wles (1994), Agrawal and Knoeber (1996), Kang and Stulz (1996), and Loderer and Peyer (00), 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 frms equty nfluence our results, we follow Bener et al. (004) and Schmd (004) and compute the market value of equty as the mean of daly observatons durng Defntons of all varables employed n ths study are also provded n Table 1. [Insert Table 1 about here] 3.. Sample descrpton As a startng pont we target all 75 frms quoted at the Swss Exchange (SWX) by the end of 00. The excluson of nvestment companes leaves us wth a sample of 35 frms. Compensaton data (Srato, Stocksod) s avalable for 171 frms. Another 13 frms must be dropped because no data on Rents, our standard measure of product market competton, s avalable. Fnally, the excluson of two obvous outlers concernng ther value of ROA and Rents leaves us wth a sample of 156 frms for our cross-sectonal regresson analyss. Data has been collected from dfferent sources and generally refers to the reportng perod from January 00 to December 00. The necessary data to compute Rents, Herf, Q, Leverage, Lnassets, ROA, Pgrowth, CSV, Stdv, and Beta were obtaned from Datastream and Worldscope. Data for the varables Blocko, Bsze, Outsder, and CEOP stems from the webste of Fnanz & Wrtschaft 0 and the Swss Stock Gude 00/003. However, for most varables data was not avalable for all frms n our sample. Mssng values were obtaned from the companes annual reports. Srato and Stockod have been drectly collected from the 00 annual reports of the companes covered n ths study. Reg s taken from the report Regelungsdchte nach Branchen by Ledergerber et al. (1998). Fnally, the varable Industry s based on the classfcaton of the Swss Exchange (SWX) and stems from the SWX webste For some frms daly stock prce data s not avalable for all share categores on Datastream, and we replace them by the mean of the 001 and 00 year end values of total market captalzaton obtaned from Worldscope. 0 The webste of Fnanz und Wrtschaft, Swtzerland s major fnancal newspaper, s: 1 The webste of the Swss Exchange s: 18

20 3.3. Descrptve statstcs Table shows descrptve statstcs of all varables ncluded n our analyss. The negatve average value of Rents ndcates that Swss frms burn captal n 001, on average. However, the medan value s postve and amounts to.9%. The mean and the medan of Herf are qute smlar; 0.41 and 0.409, respectvely. Reg takes on average a value of 1.7. It has a low of 0.7 for the nformaton technology ndustry, and a hgh of 3.6 for the travel ndustry, whch faces the hghest market entry regulatons. The average value of Tobn s Q s 1.41, and the medan s 1.13, ndcatng that Swss frms, on average, nvest n postve NPV projects. Concernng our measure of ncentve schemes provded to managers, Srato, we fnd that total share-based compensaton to offcers and drectors amounts to 5.7% of cash-based compensaton, on average, whle the medan value s zero. Ths qute small values are not surprsng as Murphy (1999) shows, that stock (and opton) partcpaton plans for the top management are relatvely rare n Swtzerland and account for a much smaller fracton of total compensaton than n most other countres and especally as compared to the US. 3 Table further shows several other nterestng results, whch we only brefly summarze: Offcers and drectors hold on average 16.5% of the equty of a frm. However, the medan of.6% 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 samples of U.S. frms used by Loderer and Martn (1997) and Anderson et al. (000) confrms that average nsder shareholdngs are even slghtly hgher n Swtzerland than n the U.S. However, the medan s a lot smaller n our sample and, hence, nsder shareholdngs are much more skewed n Swtzerland. Many other frm characterstcs are comparable to those reported by other studes n ths area. However, the mean value of Blocko of 8% s much larger than the value of 7.6% reported by Anderson et al. (000) for the U.S. Smlarly, the average value of Outsder s 87.9%, whch strongly dffers from the much lower values of 54% and 60% reported by Yermack (1996) and Barnhart et al. (1994), respectvely, for U.S. companes and 44% reported by Peasnell et al. (003) for U.K. companes. Ths fndng s especally surprsng, because foundng famles are stll regarded as an mportant factor n corporate Swtzerland. Two dfferent reasons may be responsble for ths somewhat surprsng fndng. Frst, 001 was a worldwde down-market and many frms experenced substantal losses. Second, our measure of profts ncluded n the calculaton of Rents (EBITDA) contans a number of balance sheet tems that can potentally dstort the economc content of ths varable, resultng n values of EBITDA that are downward-based measures of raw operatng surplus (e.g., see Januszewsk et al., 00). 3 Murphy (1999) compares the level and structure of CEO pay n 3 countres based on data reported n Towers Perrn s 1997 Worldwde Total Remuneraton report. 19

21 [Insert Table about here] 3.4. Correlatons Table 3 shows the correlaton coeffcents between all varables ncluded n the emprcal analyss of Secton 4. Of specal nterest are the correlaton coeffcents between Rents, Herf, Reg, Srato, and Q. Consstent wth the mplcatons of our theoretcal model, we fnd a postve correlaton coeffcent between Rents and Q and a negatve and sgnfcant coeffcent between Rents and Srato. Based on our theoretcal model, ths latter result suggests that frms may be operatng n compettve envronments on average, where more ntensve product market competton s assocated wth stronger ncentves provded to managers by ncreasng the fracton of share-based compensaton (see hypothess 1). The relatvely hgh and statstcally sgnfcant postve correlaton coeffcent between Rents and ROA s expected as mentoned n Secton 3.1. Consstent wth the man hypothess of Rath (003), the correlaton coeffcents between all three competton proxes and Beta and Stdv are negatve. Rath (003) argues that frms operatng n an ntensve competton envronment experence a hgher volatlty of profts, whch, n turn, leads to hgher stock prce volatlty and arguably hgher market betas. Wth respect to the two other measures of competton, we fnd a negatve correlaton coeffcent between Herf and Q and a sgnfcant postve relaton between Reg and Q, ndcatng that the dfferent product market competton proxes measure dfferent aspects of competton. The correlaton between Herf and Srato s postve and that between Reg and Srato negatve but both coeffcents are not statstcally sgnfcant. Furthermore, Rents s negatvely correlated wth Herf and postvely correlated wth Reg. However, both correlatons are weak and not statstcally sgnfcant, confrmng the conjecture that all three proxy varables for competton measure dfferent aspects of product market competton. 4 Thus, we conclude from ths correlaton analyss that our emprcal nvestgaton should nclude both frm-level as well as ndustry-level measures of competton. Let us fnally look at the corporate governance mechansms. Srato and Stocksod are postvely correlated wth Q, whle Blocko, Leverage, and Outsder are negatvely correlated wth Q. Rents s only very weakly correlated wth all corporate governance mechansms (besdes Srato) and no coeffcent s statstcally sgnfcant. The same s true for Herf and 4 Hence, collnearty problems should not be lkely when usng all three competton measures smultaneously n the emprcal analyss. 0

22 Reg wth the excepton of sgnfcant correlaton coeffcents between Stocksod and Herf and Stocksod and Reg. Srato s negatvely correlated wth Blocko ndcatng possble substtuton effects between the montorng by large outsde blockholders and usng share-based compensaton to te the personal wealth of offcers and drectors to ther company s performance. As the postve and statstcally sgnfcant coeffcent between Srato and Lnassets ndcates, share-based compensaton seems to be more common n large than n small frms. Interestngly, a CEO who s by the same tme presdent of the board seems to be assocated wth hgher values of Srato as well. Fnally and consstent wth earler emprcal fndngs (e.g., see Jensen and Murphy, 1990, or Funk and Wanzenred, 003), we fnd a sgnfcantly postve correlaton coeffcent between Srato and frm performance as measured by the change n shareholder value (CSV). [Insert Table 3 about here] 4. Emprcal analyss 4.1. Comparsons of frms operatng n ntensve competton envronment and other frms We begn our emprcal analyss by nvestgatng whether there are systematc dfferences wth respect to the varables employed n ths study between frms operatng n an ntensve competton envronment and frms whch do not. Table 4 presents comparsons of mean and medan values between frms wth a value of Rents equal to or above the medan value (noncompetton frms) and frms wth a value of Rents below the medan value (competton frms). Most mportantly, we fnd that non-competton frms have sgnfcantly hgher values of Tobn s Q than competton frms ndcatng that the lower economc rents, and therefore profts, assocated wth a hgher product market competton outwegh the potental benefts of reducng manageral slack. Ths fndng s consstent wth our thrd hypothess. Furthermore and consstent wth our theoretcal model, Srato s hgher for competton frms. Even though the dfference s not statstcally sgnfcant at any conventonal level, ths outcome may provde some evdence for our theoretcal fndngs that the relatonshp between the strength of ncentves and the ntensty of competton dffers dependng on the level of product market competton. In partcular, t s consstent wth our second hypothess, statng that the relatonshp between ncentves and competton becomes steeper wth ncreasng competton. 1

23 Table 4 furthermore reveals a sgnfcantly hgher proftablty, as measured by ROA, of noncompetton frms. Ths result s expected snce ROA and Rents are correlated by constructon. 5 For the same reason, the sgnfcantly hgher mean and medan values of the change n shareholder value CSV, whch s also some measure of proftablty, of noncompetton frms as compared to competton frms comes as no surprse. The fndng of sgnfcantly hgher leverage ratos of competton frms stands n lne wth Lord and McIntyre Jr. (003), who provde evdence for leverage ncreasng wth mport competton n the textle and apparel ndustry. 6 Fnally, competton frms have sgnfcantly hgher mean and medan values of Stdv and Beta. A possble reason for that fndng s that frms operatng n an ntensve competton envronment experence a hgher volatlty of profts (e.g., see Rath, 003), whch, n turn, leads to hgher stock prce volatlty and arguably hgher market betas. [Insert Table 4 about here] 4.. Multvarate OLS-regressons In ths secton, we nvestgate the nfluence of product market competton on Srato and Tobn s Q by controllng for dfferent governance mechansms and control varables n a multvarate regresson framework. Frst, we motvate the regresson equaton amed to nvestgate the determnants of Srato and report the results of OLS estmatons. Then, we examne the effect of competton on Tobn s Q The effect of product market competton on ncentve schemes Snce our man nterest s to nvestgate the effect of product market competton on ncentve schemes, we nclude Rents, Herf, or Reg as a frst explanatory varable nto our regresson equaton wth Srato as a dependent varable. Because managers are more lkely to accept share-based compensaton when they are confdent that ther company wll do well and t s benefcal for them to partcpate on ths success, we nclude Tobn s Q as a forward-lookng performance measure of the frm. To nvestgate whether there are any nterrelatons between Srato and other governance mechansms (e.g., see Bener et al., 004), we also nclude Stocksod, Blocko, Leverage, and Outsder. 5 As Table 3 reveals, the correlaton coeffcent between ROA and Rents s 0.34 and statstcally sgnfcant at the 1% level. 6 Other work on the relatonshp between leverage and competton nclude Brander and Lews (1986), Maksmovc (1988), Chevaler (1995), Kovenock and Phllps (1995), Phllps (1995), and Zngales (1998).