THE STRATEGIC CHOICE OF MANAGERS AND MANAGERIAL DISCRETION

Transcription

1 THE STRATEGIC CHOICE OF MANAGERS AND MANAGERIAL DISCRETION XIANGKANG YIN* La Trobe Unversty Manageral dscreton s lkely to be benefcal to shareholders because of strategc cross-effects n an olgopoly. In certan crcumstances, shareholders delberately retan manageral dscreton n equlbrum even when the reducton of manageral dscreton s cost free. It s found that the postve effect of manageral dscreton on profts can only be created by power-buldng (shrkng) managers f quantty (prce) competton prevals n the market. Consequently, the domnant strategy for the owners of a quantty- (prce-) competng company s to employ a power-buldng (shrkng) manager. The strategc effect of such a match between the type of manager and the form of competton exsts for all manageral decsons as far as frms nteract wth each other. I. Introducton Corporate governance, as Shlefer and Vshny (1997) defne t, deals wth the ways n whch supplers of fnance to corporatons assure themselves of gettng a return on ther nvestment. It becomes a focus for economc and fnancal studes because of the separaton of corporate ownershp and effectve asset control. In modern corporatons, the owners usually delegate manageral tasks to professonal managers so that the managers effectvely control the allocaton of corporate assets. Snce they may not own, or may own just a tny share of the frm they are runnng, and snce they can derve prvate benefts from managng the frm, the managers are lkely to have dfferent nterests from the outsde shareholders, leadng to manageral decsons not maxmsng profts. Ths agency problem of conflctng nterests between owners and managers s a central ssue of corporate governance. The concerns surroundng the separaton of ownershp and control can be traced back to Adam Smth (1776), whle the modern economc theory s poneered by Berle and Means (193). The core of the agency theory of corporate governance s to tackle the ssue of how the prncpals (the shareholders of a company) defend ther nterests by montorng ther agents (managers) and compensatng manageral effort. The focus of the theory s on the two partes of a sngle company: shareholders and managers (see, Shlefer and Vshny (1997) and Trole (001)). A consensus among economsts s that manageral dscreton hurts shareholders so that shareholders should trade-off the gan from a reducton of manageral dscreton aganst the cost of dong so. 1 In a non-monopolstc envronment, t s wdely beleved that market competton reduces manageral dscreton and drves frms to perform better. Hart (1983) develops a hdden * Correspondence: Department of Economcs and Fnance, La Trobe Unversty, Bundoora, Vctora 3083, AUSTRALIA. Emal: x.yn@latrobe.edu.au. I thank Rchard Damana (managng Edtor) for hs valuable comments and Lee Smth for her excellent edtoral assstance. 1 Exceptons to ths consensus nclude Burkart et al. (1997) and Bolton and Scharfsten (1998), who argue that the constrants on managers through montorng are lkely to reduce entrepreneural and nnovatve efforts and n turn the reducton of manageral dscreton s not purely benefcal to shareholders.

2 374 AUSTRALIAN ECONOMIC PAPERS DECEMBER nformaton model n whch the prncpal of a manageral frm can nfer the manager s productvty through market sgnals and n turn competton unambguously reduces manageral slack. In Nalebuff and Stgltz (1983), the performance of compettors allows the prncpal to make nferences about common shocks. Then, an ncrease n competton generates addtonal nformaton whch can be used to mtgate agency problems. However, Scharfsten (1988) shows that Hart s result crucally reles on hs assumpton of manageral preferences. Wth more general manageral preferences, the effect of competton on manageral slack s ambguous. Decomposng the mpact of competton on manageral behavour nto four effects, Hermaln (199) fnds that each effect s of potentally ambguous sgn and therefore theory does not offer a defntve defense for a postve role of competton n reducng manageral slack. Smlarly, Schmdt (1997) argues that an ncrease n competton rases the probablty of lqudaton and stmulates manageral effort on the one hand, but t reduces frm s profts and makes hgh manageral effort less attractve on the other. Agan, the total effect s ambguous. Whle theory does not reach a defntve concluson on the effect of market competton on manageral dscreton, economsts fnd that manageral dscreton n terms of slack or other nonproft-maxmsng behavour does not always hurt the owners of an olgopolstc frm (Fershtman and Judd, 1987; Sklvas, 1987; Vckers, 1985). Qute often, a strategy not drectly maxmsng profts can lead to a better proft outcome through ts nteracton wth the actons of rval companes. The studes by Fershtman and Judd, Sklvas, and Vckers llustrate that n equlbrum a manager s pad n excess of hs/her decson s margnal proft n a Cournot-quantty game, but pad less than the margnal proft n a Bertrand-prce game. Thus, t s not necessarly true that maxmum profts are earned by frms whose objectves are proft-maxmsng. There are strategc effects n selectng manageral ncentves. Lke these studes, ths paper nvestgates the agency problem between owners and managers and the strategc effects of corporate governance. However, t develops a model wth an addtonal dmenson the nteracton between manager type, manageral dscreton and competton form. The ntuton of our analyss s as follows. Manageral dscreton makes manageral decsons devate from a proft-maxmsaton objectve. However, these non-proft-maxmsng decsons can hurt a rval company more than the company tself under certan condtons. In the end, a certan degree of manageral dscreton s lkely to be proft-enhancng. In other words, the shareholders may beneft from manageral dscreton and they should not reduce t to a mnmum even f such a reducton s cost free. Whether manageral dscreton can create postve effect on profts or not, however, depends on the combnaton of the manager type and competton form n the product market. We fnd that when quantty competton prevals n the product market the lkely postve effects of manageral dscreton can only be generated by power-buldng managers, whle when prce competton prevals the postve effects can only be generated by shrkng managers. The reason for ths s straghtforward. A power-buldng manager attempts to produce more than the proft-maxmsng output because of the prvate benefts that extra output yelds to the manager. Because output decsons are strategc substtutes n quantty competton, a larger output forces the rval frm to reduce output and n turn brngs the manager s own company a greater market share. Thus, a manageral decson wth an agency problem s lkely to result n more profts than ts counterpart wthout an agency problem. The same ratonale s behnd prce competton, snce prce decsons are strategc complements and a shrkng manager benefts from makng less effort n management. As a result, the preferred manager type for shareholders depends on the form of market competton. Vckers (1985) consders manager heterogenety but does not formally analyse the equlbrum choces of ncentves and manager type.

3 003 THE STRATEGIC CHOICE OF MANAGERS 375 An mportant dfference between our model and those of Fershtman and Judd (1987), Sklvas (1987) and Vckers (1985) s that we explctly consder the shareholders costs of montorng and compensatng managers whle n those models the costs are neglected. The ntroducton of these costs not only makes the model more realstc but also makes the concept of shareholders exertng effort to choose managers more comprehensble and meanngful. By choosng a sutable manager, the shareholders of a frm can sgnfcantly save on the costs of montorng and compensaton. It can be shown that the domnant strategy n manager selecton s hrng a powerbuldng (shrkng) manager f the frm s nvolved n quantty (prce) competton. In two related works, Lambertn (000a, b) models the nterplay between market competton and the nternal organsaton of a frm. These studes analyse whether frms set prces or quanttes, move smultaneously or sequentally, and choose to be manageral or entrepreneural n a multstage game. In contrast to Lambertn, we assume the owners of a frm defntely delegate manageral tasks to professonal managers. The focus of our model s on the owners choces of the type of managers runnng the frm and the extent of manageral dscreton. As ponted out earler, the model establshes a lnk between these decsons and the forms of competton. Moreover, manageral decsons n our model are not confned to prces and quanttes. The dscusson n Secton IV shows that our analyss of the strategc choce of managers s vald for all manageral decsons as far as there s some strategc nterplay between frms. In the remander of the paper, Sectons II and III analyse the choces of the manager type and manageral dscreton n equlbrum when frms compete n the product market wth quantty and prce, respectvely. Besdes prce and quantty decsons, other manageral decsons of a company, such as R&D nvestment, producton capacty, product range and servce qualty, also nteract wth other frms manageral decsons n an olgopolstc ndustry. Secton IV comments on a framework to accommodate general manageral decsons wth strategc effects. The fnal secton concludes the paper. II. The Equlbrum Choce of Managers and Manageral Dscreton wth Quantty Competton Consder two companes n an ndustry where ther shareholders delegate the manageral tasks to managers. Focusng on the agency problem between shareholders and ther managers, we assume that the shareholders of each company have a perfect congruence of nterests and there s no free-rder problem n montorng the managers. All agents n the model are rsk neutral. The demand functon for each company s product s uncertan and unverfable so that the shareholders cannot verfy whether ther manager adopts a proft-maxmsng manageral decson or not, although profts are observable and verfable ex poste. We consder a three-stage game between the two companes. In the frst stage, the shareholders of each company hre a manager to run ther busness. Then they choose an ncentve package to compensate the manager and decde the extent to whch the manager wll be montored. These decsons determne the algnment of the manager s nterests wth those of the shareholders. In the thrd stage, the manager of each company makes a manageral decson to compete wth the rval company n the product market. 3 The equlbrum concept n the paper s pure strategy subgame perfect equlbrum. 3 In our model, the competton form s exogenously gven. However, t s possble to extend the model to allow the managers to determne not only the levels of quantty or prce but also the competton form (.e., the choce between quantty and prce beng the manageral decson) as Lambertn (000a, b) does. Snce such an extenson sheds no more lght on the ssue, we confne our analyss to the exogenous determnaton of competton form.

4 376 AUSTRALIAN ECONOMIC PAPERS DECEMBER In the market, nverse demand functons are assumed to be lnear and uncertan, p = α β q γ q j + ε, j,, j = 1, (1) where p s the prce of company s product and q s the quantty, α, β, and γ are postve constants and satsfy γ mn{β 1, β }. The random parameter ε represents the uncertanty of the demand, whch has a zero mean. Ths demand system mples that the two companes supply dfferentated products to the market when γ < mn{β 1, β }. If γ = β 1 = β, ther products are homogenous and f γ = 0 each frm s a monopolst. Ther correspondng demand functons are where q = a b p + cp j + e, j,, j = 1, () a = (α β j α j γ )/(β 1 β γ ), b = β j /(β 1 β γ ), c = γ /(β 1 β γ ), e = (β j ε γε j )/(β 1 β γ ). Assume each company has a constant margnal cost m and no fxed costs, the proft obtaned from product sales of can be charactersed as π = (p m )q. (3) In the process of producton, the manager of a company obtans a net prvate beneft from managng the frm, B, whch s assumed to be proportonate to the output level; 4.e. B = t q. (4) The prvate benefts of management consdered n the model are the usual perks and psychologcal benefts from empre-buldng so they are small relatve to profts and they do not result n a proft reducton. 5 On the other hand, wth a larger output the manager has to devote more effort to management, whch can be nterpreted as an extra cost. In equaton (4), B measures, n monetary terms, the dfference between these benefts and costs. Because managers have dfferent preferences and they value the benefts and costs dfferently, the net beneft depends on t the preferences or type of frm s manager. Thus, we call a manager a power-buldng manager f t = t p > 0 and a shrkng manager f t = t s < 0. The objectve of company s manager s V = δ π + ν B = δ (π + θ B ), (5) where δ and ν are varables chosen by the shareholders of company n the second stage. 0 < δ < 1 can be nterpreted as the compensaton mechansm n company that rewards the manager for ncreasng company profts and 0 ν < 1 represents the montorng mechansm that reduces the prvate benefts of the manager. As (5) shows, t s not each ndvdual δ and ν but ther rato, θ ν /δ that nfluences the manager s manageral decson. Thus, θ measures not only the ntensty of montorng and compensaton provson but also the manageral dscreton of the company, and θ = 0 ndcates that the manager s nterests are completely n lne wth the shareholders. For the analyss below, we defne 1 as the value of θ when the shareholders of frm nether montor the manager nor provde an ncentve package to the manager. 6 It s worth notng 4 Ths assumpton s the same as Vckers (1985). If the prvate benefts are proportonate to sales as assumed n Fershtman and Judd (1987) and Sklvas (1987), our results wll not be changed qualtatvely. 5 However, the prvate benefts can also be thought of as managers stealng a large share of the cash flow produced by the company assets as modeled n Bolton and Scharfsten (1990) and Hart and Moore (1998). In ths case, the prvate benefts wll reduce a company s profts, whch s not explctly modelled n the paper. 6 But the manager s work s stll compensated by salary or other lump-sum payment.

5 003 THE STRATEGIC CHOICE OF MANAGERS 377 that δ π n (5) needs not to be the drect payoff to the manager. Thus, although a company does not provde ts manager wth an ncentve package lke a proft bonus, stock optons or equty shares, the manager may stll look after the profts of the company for the sake of hs/her long-term career reputaton or stock market montorng (Holmstrom and Trole (1993)). In other words, δ 0 even though there s no drect proft transferrng from the shareholders to the manager. In the thrd stage, the manager of each company has to make a manageral decson. If they use the quantty of output as a manageral decson varable, the frst-order condton (FOC) maxmsng the expected value of (5) leads to Eπ / q + θ t = 0 or α β q γ q j m + θ t = 0, j,, j = 1, (6) where E s the expectaton operator. From these condtons, we can mmedately obtan output reacton functons q = (α γ q j m + θ t )/β. (7) These reacton functons are smlar to those obtaned from a Cournot olgopoly except for an addtonal term θ t. Ths s not surprsng because n our model the manager s objectve functon can be wrtten as [ p (m θ t )]q. Thus, consderng (m θ t ) as margnal cost, the reacton functon (7) s exactly the same as the conventonal Cournot reacton functon. Solvng (7) for q 1 and q we obtan equlbrum output for each company n the thrd stage, 7 q * = [β j (α + θ t m ) γ (α j + θ j t j m j )]/ L, (8) where L 4β 1 β γ. Agan, when θ 1 = θ = 0, the managers maxmse profts n makng manageral decsons and (8) leads to the conventonal Cournot output levels. Expectng that the managers wll make manageral decsons n the thrd stage accordng to (8), the shareholders of each company can adjust ther ncentve and montorng varable, θ, to maxmse ther net profts (net of compensaton and montorng costs) n the second stage. But before analysng ths, let us frst nvestgate how the proft (gross of compensaton and montorng costs) changes as θ ncreases. Substtutng equlbrum output (8) nto proft functon (3) and takng dervatve wth respect to θ yelds E( π q*, q*) βθ j t γ t q* 1 = +, (9) θ L where E( π s the expected proft condtonal on the outputs of (8). 8 q1*, q*) Notng that L and other parameters n (9) except for t are postve, the followng concluson s obvous. Proposton 1. The proft functon monotoncally decreases as the company s manageral dscreton ncreases f ts manager s of the shrkng type. But the functon has a maxmum f ts manager s of the power-buldng type. Proof. Snce L, β j, θ and q * are postve, (9) shows that E( π q1*, q*)/ θ < 0 holds f t < 0. When t > 0, substtutng (8) nto (9) we get E( π q1*, q*)/ θ < 0. Settng (9) equal to zero yelds the proft-maxmsng θ, whch s charactersed by 7 To avod the trval stuaton of no producton n equlbrum, we assume that β j (α + 1 t s m ) > γ (α j + 1 j t p m j ) so that equlbrum outputs are postve. 8 Wthout losng generalty, q* s used for ether the amount of equlbrum output when the shareholders decsons are specfed or the equlbrum producton rule as a functon of shareholders decsons.

6 378 AUSTRALIAN ECONOMIC PAPERS DECEMBER Fgure 1. where S β 1 β γ. θ = γ [β j (α m ) γ (α j + θ j t j m j )]/4 S β j t, (10) The proposton llustrates that manageral dscreton defntely makes the shareholders worse-off f the manager s shrkng and the frms play a quantty game. But f the manager s power-buldng, a certan degree of manageral dscreton s benefcal to the shareholders. Ths result s contradctory to the agency problem n a monopolstc envronment, where manageral dscreton usually has adverse effects on the shareholders. To see why, let us go back to (9). In the equaton, the term βθ j t π q = < 0, q θ L represents the own-effect of a margnal ncrease n manageral dscreton on own output and consequently on the proft, and s defntely negatve. However, an ncrease n manageral dscreton also affects the rval company s output and consequently own proft. Ths cross-effect s represented by the term γ tq * π q j =. q θ L Ths effect rases the company s proft f the manager s power-buldng. The total effect as a combnaton of these two effects s postve when θ s suffcently small and t s postve. Fgure 1 above llustrates these effects graphcally. The Cournot reacton curve for company 1 s the reacton functon (7) wth θ t = 0 ( = 1). If the manager of company 1 maxmses profts, he/she should respond to company s output accordng to ths curve and the equlbrum pont s at Q a. However, when θ 1 > 0, he/she s not a proft maxmser and hs/her reacton curve, determned by (7), s an outward shft of the Cournot reacton curve f t 1 > 0. The equlbrum s establshed at pont Q c. We can decompose the move from Q a to Q c nto two steps: from Q a to Q b and from Q b to Q c. Notcng that the soproft curve s tangent to the horzontal lne at Q a and the slope of soproft curve s negatve on the rght of Q a, t s not hard to fnd that the frst move leads to a reducton n proft whle the second produces an j

7 003 THE STRATEGIC CHOICE OF MANAGERS 379 ncrease. 9 However, the reacton curve shfts n the opposte (nward to the orgn) drecton f t 1 < 0 (not llustrated n Fgure 1), resultng n an ntersecton of the two reacton curves above Q a. Thus, company 1 defntely cannot obtan more profts f a shrkng manager devates from a Cournot reacton curve. The proof of Proposton 1 shows that (10) s the shareholders reacton functon of choosng the compensaton and montorng level when the cost of montorng and compensaton provson s ether neglgble or lumpy. Therefore, the followng corollary s mmedate from Proposton 1. Corollary. When the costs of ncentve provson and montorng the manager are ether neglgble or lumpy, the shareholders choose θ = 0 f they have hred a shrkng manager. But f they have hred a power-buldng manager and ther rval s manager s also power-buldng, they choose 10 γ [ L( α m) βγ( α j mj)] θ =. j,, j = 1, (11) ( β1 β β1βγ + γ ) t If ther manager s power-buldng and the rval company s manager s shrkng, they choose θ = γ [ β ( α m ) γ( α m )]/ 4 β t. j j j S j j,, j = 1, (1) Proof. If t < 0, E( π q1*, q*)/ θ < 0 θ = 0 maxmses proft. If t > 0 and t j > 0, solvng (10) yelds (11). If t > 0 and t j < 0, then θ j = 0 and (10) leads to (1). Recallng that 1 s the value of θ when the shareholders of frm nether provde any compensaton to algn ther manager s nterests wth thers nor montor the manager, t s plausble to assume that 1 s greater than max { θ, θ }. However, we have no economc reason to exclude the possblty that 1 s smaller than mn { θ, θ }. If that happens, the shareholders actually prefer that the power-buldng manager s objectve s more devated from proft maxmsaton. Corollary 3. If the manager s power-buldng and 1 < mn { θ, θ }, the best opton for the shareholders s to adopt certan measures to ncrease θ rather than to reduce t. In the rest of the paper, we assume manageral dscreton s suffcently large wthout ncentve provson and montorng by the shareholders;.e., Wth ths assumpton and (10), t s obvous that E( π q1*, q*)/ θ s negatve at θ = 1 gven any value of θ j. On the other hand, the cost to reduce θ s assumed to be a quadratc functon of θ that where µ s a postve constant, ndcatng the expensveness of manageral compensaton and montorng. Snce the shareholders n the second stage maxmse the net expected proft Nπ E( π q1*, q*) C( θ), and Nπ / θ s negatve at θ = 1, equlbrum cannot be at the corner θ = 1. In other words, the shareholders are defntely wllng to spend some money n reducng θ. 9 An nteror soproft curve represents a hgher proft level. 10 Wth a mld assumpton that the dfference between (α 1 m 1 ) and (α m ) s small, we can ensure θ and θ are postve f t > 0. 1 > θ. C θ = µ θ ( ) ( 1 ) /, (13) (14)

8 380 AUSTRALIAN ECONOMIC PAPERS DECEMBER Proposton 4. Assume the cost of manageral compensaton and montorng s as (14). The equlbrum level of manageral dscreton can be charactersed by θ 1 * and θ * below f they are postve, 3 γ t[ L( tj + β jµ )( α m) γ( β tj + Lµ )( αj mj)] + [( 4β t j S + Sµ ) 1 γ tt µ θ j L * 1 1 ] =. ( β1 β β1βγ + γ ) t1 t + 4( βt1 + β 1t ) L Sµ + 3 Lµ j,, j = 1, (15) If θ * s postve but θ j * s negatve, then equlbrum s θ j = 0 and θ equal to γ t[ βj( α m) γ ( α j mj)] + Lµ 1 θ * =. j,, j = 1, (16) 4β jt S + Lµ If both θ 1 * and θ * are negatve, then equlbrum s θ 1 = θ = 0. Proof: N π βθ j t γ tq* Notce = + + µ ( 1 θ). Defne θ L ˆ 1 Lµ 1 θj γ( αj mj) βj( α m). tj γ t Then Nπ / θ < 0 f θ j > 9 j and t 1 t > 0 or f θ j < 9 j and t 1 t < 0. The correspondng reacton functon s θ = 0. Otherwse, solvng Nπ / θ = 0 to obtan the reacton functon γ t[ βj( α m) γ( αj + θ j tj mj)] + Lµ 1 θ =. j,, j = 1, (17) 4β j t S + Lµ Fgure below llustrates possble confguratons of the reacton curves. Routne calculaton shows the soluton to the two equatons of (17) s (15). If θ 1 * and θ * are postve as n Fgure (a) or (c), they form the equlbrum of the second stage. If θ * s postve but θ j * s negatve as n Fgure (b), then equlbrum s θ j = 0 and θ = θ *. If both θ 1 * and θ * are negatve, Fgure (d) shows equlbrum s θ 1 = θ = 0. It s clear that f µ = 0, that s the montorng and manageral compensaton are cost free, then θ * and θ * n Proposton 4 collapse to θ and θ n (11) (1), respectvely. Implctly assumng that managers are dentcal and power-buldng, Fershtman and Judd (1987) and Sklvas (1987) obtan the smlar equlbrum ncentves as Proposton 4. But our model llustrates how the shareholders of a frm compensate and montor the manager when the rval frm employs a same or dfferent type of manager. Ths complexty enables us to analyse the equlbrum choce of managers. Turnng to the frst stage, we assume there s a pool of managers whose type s charactersed by ether t s or t p, where t s = t p < Through ntervewng, tracng the track record of managers and other methods, the shareholders can dentfy the type of manager and choose the rght type they want. 1 To explctly express the values of θ 1, θ, t 1 and t mpled n E( π q1*, q*) by (8), the expected proft functon s wrtten as E( π q*, q*, θ, θ, t, t ). 1 j j 11 The assumpton t s = t p < 0 s more than requred. To ensure pure strategy equlbrum n the frst stage, we only requre that the dfference between the absolute values of t s and t p s not too large. 1 Ths, of course, s smplfcaton. More generally, we should assume that ntervews and track records can only provde an mperfect sgnal of the type of manager. However, such complexty does not shed more lght on the ssue we focus on. Thus, we assume n the text that the shareholders can obtan a perfect sgnal of manager type through ther nvestgaton.

9 003 THE STRATEGIC CHOICE OF MANAGERS 381 Fgure. Proposton 5. The shareholders defntely choose a power-buldng manager for ther company, n spte of the rval company s choce of manager. Proof. For t = t s and any t j, let the values of consequent equlbrum θ s be and j. Eq. (8) llustrates that the proft functon of company s ndependent of the manager type when θ = 0; that s, s p s E( π q1*, q*, 0, j, t, tj) = E( π q1*, q*, 0, j, t, tj ). Eq. (9) shows that E( π q1*, q*, θ, j, t, tj) p declnes faster than E( π q*, q*,, 1 θ θj, t, tj) as θ ncreases (the latter ncreases when θ s s p suffcently small). Thus, E( π q1*, q*,, j, t, tj) C( ) < E( π q1*, q*,, j, t, tj) C( ). By optmal choce of θ n the second stage, the shareholder of company defntely can obtan p more net proft than E( π q1*, q*,, j, t, tj) C( ) f they hre a power-buldng manager n the frst stage. When companes compete wth quantty, Fgure 1 llustrates that an output response, beng a lttle bt more aggressve than the Cournot reacton, s more proftable than the Cournot reacton. Thus, manageral dscreton n the form of power-buldng s better than shrkng. By choosng a power-buldng manager rather than a shrkng manager, the shareholders can save ther costs of montorng and manageral compensaton.

10 38 AUSTRALIAN ECONOMIC PAPERS DECEMBER To fnd out the equlbrum of the whole game, we only need to substtute t 1 = t = t p nto (15) or (16) to obtan equlbrum manageral dscreton, and then usng (8) to obtan the equlbrum outputs. After obtanng the equlbrum of quantty competton wth manageral dscreton, t s nterestng to compare t wth a standard Cournot equlbrum. Proposton 6. Quantty competton wth manageral dscreton leads to more outputs but lower expected prces and profts than Cournot competton. Proof. Notng that t 1 = t = t p > 0 and θ 0 n equlbrum and the Cournot outputs are the specal case of (8) wth θ 1 = θ = 0, the concluson of larger outputs s mmedate. Substtutng equlbrum outputs of both companes nto (1) yelds lower expected prces. For the expected proft, we have E q qj E q q C j E q C q C C C ( π *, *) < ( π *, ) < ( π, j ), where q and q j are the Cournot outputs. The frst nequalty s due to Eπ / q j < 0 and qj q C C * > j and the second s due to the noton that q s the C best response to q j. The ntuton behnd ths proposton s straghtforward. As Proposton 5 shows, shareholders prefer a power-buldng manager to a shrkng manager f quantty competton prevals n the market. The reason s that gven the competng company s manageral dscreton, an expansonary devaton from the proft-maxmsng output level s better than a shrnkng devaton. But after power-buldng managers are hred n both companes, ther self-nterests nduce them to produce more than the Cournot output, resultng n a collapse n prces and n turn a reducton n profts. Ths outcome of lower profts arses because of a prsoner s dlemma problem: when both frms produce more aggressvely t ultmately reduces both frms profts (gross of the costs of manageral compensaton and montorng). We should dstngush lower equlbrum profts n Proposton 6 from the lkely postve effects of manageral dscreton on profts n Proposton 1. Proposton 1 shows that profts may ncrease as a company s own manageral dscreton rses gven the rval company s manageral dscreton and manager type. However, when the shareholders of both companes prefer powerbuldng managers and allow for a certan degree of manageral dscreton, t ends up wth lower profts than the Cournot profts. The result n Proposton 6 s also obtaned by Fershtman and Judd (1987), Sklvas, (1987) and Vckers (1985). The man dfference s that the managers n ther models are assumed to be power-buldng whle n our model the power-buldng managers employed by quantty-competng frms are endogenously determned. III. The Equlbrum Choce of Managers and Manageral Dscreton wth Prce Competton 13 When prce competton prevals n the market, the managers of both frms use prces as decson varables. The maxmsaton of expected value of (5) leads to Eπ / p θ t b = 0 or a b p + cp j (θ t m ) b = 0 ( j,, j = 1, ) (18) 13 Ths secton only outlnes the major results because the devaton of these results s smlar to that of Secton II. The dervaton detals are avalable upon request or at staffhp/xyhp/xk-publcaton.html.

11 003 THE STRATEGIC CHOICE OF MANAGERS 383 These condtons yeld the prce reacton functons, p = [a + cp j (θ t m ) b ]/b. (19) The soluton to these equatons gves the equlbrum prces, p = { b j [a (θ t m )b ] + c [a j (θ j t j m j ) b j ]}/ D L, (0) where D L (4 b 1 b c ). In the second stage of the game, substtutng (0) nto proft functon (3) we fnd the equlbrum level of manageral compensaton and montorng can be charactersed by θ 1 * and θ * below f they are postve, 3 c [( 4bb DStj DLµ m bjc tjmj bjbjdltj DLµ a m c bb jt j DLµ aj m θ 1 + ) + ( + )( + ) ( + )( + j)] * = 4 ( 16b 1b 1b1bc + c ) bt ( bt j j) 3 [( 4bDt j S j + DLµ ) 1 + ct1j] DLµ +. j,, j = 1, (1) 4 ( 16bb 1 1bbc 1 + c) bbtt 1 j If θ * s postve but θ j * s negatve, then equlbrum s θ j = 0 and θ equal to bc t[ DLm bj( a + bm ) c( aj + bjmj)] + DLµ 1 θ * =. j,, j = 1, () 4bbtD j S + D Lµ If both θ 1 * and θ * are negatve, equlbrum s θ 1 = θ = 0. Back to the frst stage, the concluson analogous to Proposton 5 s that shareholders defntely choose a shrkng manager for ther company regardless of the rval company s choce of manager. In comparson wth conventonal Bertrand game, prce competton wth manageral dscreton leads to hgher equlbrum prces. When the demand functons and margnal costs of the two companes are symmetrc, t also leads to lower expected outputs. Moreover, each company can earn more proft (gross of the cost of manageral ncentve provson and montorng) than the Bertrand proft f [a (b c) m ] c > b (b c)1t s. IV. Comments on Extensons of the Model The analyss of prevous sectons s not confned to quantty or prce competton n the product market. In fact, the strategc effects of corporate governance exst n all manageral decsons as far as nter-frm rvalry should be taken nto consderaton. For a more general framework, let π (s 1, s ) be the proft functon, where s 1 and s are manageral decsons of the two frms. So, s 1 and s can be ether quantty and prce decsons as analysed n the prevous two sectons or other decsons such as product range, servce qualty, nvestment n producton capacty, R&D and new product desgn. A t -type manager receves net prvate beneft B (s 1, s, t ) = t O (s 1, s ) from hs/her management. Here O > 0 represents the operaton scale of the frm, of whch the manager s n charge. In such a more general settng, the manager stll uses crteron (5) to determne manageral decsons. Assumng the second-order condton holds, there s a unque manageral reacton functon that s = R (s j, θ t ). Thus, f R / θ s postve for one type of manager, t s defntely negatve for the other type. Ths mples that as manageral dscreton (θ ) ncreases, one type of manager wll shft the

12 384 AUSTRALIAN ECONOMIC PAPERS DECEMBER reacton curve outwards whle the other wll shft the curve nwards to the orgn as llustrated by Fgures 1 and 3. In other words, shareholders can hre a type of manager, whose reacton to a rval company s competton defntely leads to a postve strategc effect;.e., ( Eπ / s j )( s j / θ ) > 0. Because of the postveness of the strategc effect, the shareholders delberately keep a certan amount of manageral dscreton even f the cost of reducng manageral dscreton s neglgble. Moreover, by explotng ths effect they can sgnfcantly save ther expendture on compensatng and montorng managers. V. Concludng Remarks Although orthodox economc theory adopts the proft-maxmsaton hypothess, many economsts have questoned the valdty of the hypothess for large corporatons (for nstance, Baumol, 1958). More recent studes fnd that the owners of a frm may delberately make ther managers devate from the objectve of proft maxmsaton, even though the reducton of manageral dscreton s cost free (Fershtman and Judd, 1987; Sklvas, 1987; Vckers, 1985). In ths paper, we ntroduce the heterogenety of managers nto the analyss. It s found that when the type of manager s endogenously determned n equlbrum, there s a match between the type of manager and the form of market competton. Frms have an ncentve to hre power-buldng (shrkng) managers when quantty (prce) competton n the product market prevals. More generally, shareholders ntend to explot the postve strategc effect of manageral decsons by properly choosng ther managers. Wth the rght choce of manager type, manageral dscreton s proft enhancng to a certan extent. Therefore, the owners can cut the costs of reducng manageral dscreton by strategcally choosng managers. Our fndng of the match between the managers type and competton form has an mportant mplcaton for the labour market of managers; that s, the employment of managers s determned by ther companes forms of competton. But whether such a specalsaton of managers exsts stll remans to be tested n emprcal studes. A related ssue s the sgnallng of manager type, whch s not nvestgated n ths paper. When a manager fnds that a potental employer does not apprecate hs/her type, there s an ncentve to send a false sgnal. Such sgnallng can be ntegrated nto the current model n a further study. Another mplcaton of our fndngs s that of Bertrand prcng. The crtque to the homogeneous Bertrand competton s ts unrealstc outcome of margnal cost prcng. The well-known solutons to the crtque nclude the ntroducton of product dfferentaton, capacty constrants and search/swtch costs. Our fndngs provde another soluton to the Bertrand paradox. Snce the managers hred by prce-competng frms are shrkng, they defntely adopt a prce hgher than the proft-maxmsng prce. 14 So, we never observe margnal cost prcng when tasks are delegated to managers. References Baumol, W. 1958, On the Theory of Olgopoly, Economca, vol. 5, no. 99, pp Berle, A. Jr. and Means, G. 193, The Modern Corporaton and Prvate Property, Commerce Clearng House, Chcago. 14 Though ths concluson s obtaned from a dfferentated olgopoly, t s also vald for a homogenous olgopoly.

13 003 THE STRATEGIC CHOICE OF MANAGERS 385 Bolton, P. and Scharfsten, D. 1990, A Theory of Predaton Based on Agency Problems n Fnancal Contractng, Amercan Economc Revew, vol. 80, no. 1, pp and 1998, Corporate Fnance, the Theory of the Frm, and Organzaton, Journal of Economc Perspectves, vol. 1, no. 4, pp Burkart, M., Gromb, D. and Panunz, F. 1997, Large Shareholders, Montorng, and the Value of the Frm, Journal of Quarterly Economcs, vol. 11, no. 3, pp Fershtman, C. and Judd, K. 1987, Equlbrum Incentves n Olgopoly, Amercan Economc Revew, vol. 77, no. 5, pp Hart, O. 1983, The Market Mechansm as an Incentve Scheme, Bell Journal of Economcs, vol. 14, no., pp and Moore, J. 1998, Default and Renegotaton: A Dynamc Model of Debt, Quarterly Journal of Economcs, vol. 113, no. 1, pp Hermaln, B. 199, The Effects of Competton on Executve Behavor, Rand Journal of Economcs, vol. 3, no. 3, pp Holmstrom, B. and Trole, J. 1993, Market Lqudty and Performance Montorng, Journal of Poltcal Economy, vol. 101, no. 4, pp Lambertn, L. 000a, Strategc Delegaton and the Shape of Market Competton, Scottsh Journal of Poltcal Economy, vol. 47, no. 5, pp b, Extended Game Played by Manageral Frms, Japanese Economc Revew, vol. 51, no. 3, pp Nalebuff, B. and Stgltz, J. 1983, Informaton, Competton, and Markets, Amercan Economc Revew, vol. 73, no., pp Scharfsten, D. 1988, Product-Market Competton and Manageral Slack, Rand Journal of Economcs, vol. 19, no. 1, pp Schmdt, K. 1997, Manageral Incentves and Product Market Competton, Revew of Economc Studes, vol. 64, no., pp Shlefer, A. and Vshny, R. 1997, A Survey of Corporate Governance, Journal of Fnance, vol. 5, no., pp Sklvas, S. 1987, The Strategc Choce of Manageral Incentves, Rand Journal of Economcs, vol. 18, no. 3, pp Smth, A. 1776, An Inqury nto the Nature and Causes of the Wealth of Natons, Modern Lbrary, New- York, Trole, J. 001, Corporate Governance, Econometrca, vol. 69, pp Vckers, J. 1985, Delegaton and the Theory of the Frm, Economc Journal, vol. 95, no. 0, Supplement, pp