Work Ethic, Employment Contracts, and Firm Value

Transcription

1 Work Ethi, Employment Contrats, and Firm Value BRUCE IAN CARLIN and SIMON GERVAIS (forthoming in the Journal of Finane) ABSTRACT We analyze how the work ethi of managers impats a firm s employment ontrats, riskiness, growth potential, and organizational struture. Flat ontrats are optimal for diligent managers beause they redue risk-sharing osts, but they attrat egoisti agents who shirk and unskilled agents who add no value. Stable, bureaurati firms with low growth potential are more likely to gain value from managerial diligene. Firms that hire from a virtuous pool of agents are more onservative in their investments and have a horizontal orporate struture. Our theory also yields several testable impliations that distinguish it from standard ageny models. Brue Carlin is from the University of California at Los Angeles; Simon Gervais is from Duke University. The authors would like to thank Alon Brav, Pino Lopomo, and David Robinson for helpful disussions throughout the projet. Also providing useful omments and suggestions were Mihael Brandt, Amil Dasgupta, Paolo Fulghieri, Shimon Kogan, Tray Lewis, Rih Mathews, Curtis Taylor, two anonymous referees, the Ating Editor, and seminar partiipants at Duke University, Singapore Management University, National University of Singapore, Hong Kong University of Siene and Tehnology, the University of California at Irvine, INSEAD, the University of North Carolina at Chapel Hill, and the annual onferene of the Finanial Researh Assoiation. All remaining errors are the authors responsibility.

2 Ever sine Berle and Means (1932) reognized that the separation of ownership and ontrol impats firm value, eonomists have foused on ways to mitigate the ageny problems that arise between shareholders and managers. Ageny theory is now ubiquitous in orporate finane as an important determinant of firm size, apital struture, orporate governane, and firm value. 1 Traditionally, ageny theory is founded on the priniple that managers are egoisti and must be given inentives to at in the best interests of the firm. 2 However, this framework largely ignores the possibility that some managers are innately diligent and do not pose a moral hazard threat. This shortoming is highlighted by Brennan (1994), who argues that there is a signifiant differene between being rational and being self-interested. That is, to assume that rational beings are neessarily egoisti in an ageny framework needlessly narrows the sope of the analysis. Indeed, as far bak as Aristotle (in Niomahean Ethis), it was proposed that individuals in a ivilized soiety inorporate ethial standards into the deisions that they make. Similarly, Akerlof (2007) suggests that soial norms an impat the overall eonomy by affeting the hoies that people make. This interplay between personal inentives and soietal pressures (or morals) has been studied in the fields of psyhology (e.g., Judge and Ilies (2002)), law (e.g., Bohnet, Frey, and Huk (2001) and Shavell (2002)), politial siene (e.g., Rose-Akerman (1999)), and eonomis (e.g., Frank (1987) and Sen (1987)). The ommon theme in all of this literature is that ethial individuals make deisions within a self-imposed moral onfine. 3 Applying this to an ageny framework, this means that diligent employees have a self-imposed moral onstraint that prevents them from shirking, but they rationally internalize the utility osts of this onstraint when they aept the job. With this in mind, we study how managerial diligene and employee work ethi affet the preditions of traditional ageny theory and its impliations for firm value. We show that the propensity for virtue impats the employment ontrats that firms offer, the expeted growth and riskiness of firms, and their orporate struture. More speifially, we address the following questions: When managers are possibly virtuous, what is the optimal employment ontrat? What is the value of virtuous behavior for the firm and for the agent? Is it possible to sreen managers for virtue? How does the manager s level of dediation affet the firm s projet hoie, growth potential, and riskiness? How does diligene impat inentives and profitability in projets that require teamwork? To address these questions, we onsider a prinipal-agent model in whih agents may be either egoisti or virtuous. The egoisti agent is the lassi manager who ats in his own best interest and requires inentives to exert high effort. The egoisti agent is subjet to both a partiipation and 1

3 an inentive ompatibility onstraint. In ontrast, the virtuous agent always exerts high effort and does not require extra inentives: his word is his bond. This does not mean that virtuous agents give eonomi surplus away; rather, they antiipate and internalize the fat that they will exert ostly effort in some situations when it is not optimal for them to do so. As suh, they harge the firm ex ante for this expeted effort. From the firm s perspetive, hiring a virtuous agent amounts to hiring one whose inentive ompatibility onstraint is never binding. However, the partiipation onstraint, whih ensures that the virtuous agent does not foolishly give up eonomi surplus, still has to be satisfied. In equilibrium, the firm optimally offers an agent known to be virtuous an employment ontrat that is different from that offered to an agent that is known to be egoisti. Sine a virtuous agent is not bound by an inentive ompatibility onstraint, the firm an redue the inentive portion of ompensation and offer a larger fixed wage (i.e., offer a flatter ompensation ontrat), whih saves the risk-neutral firm the osts of imposing ompensation risk onto a risk-averse agent. In this ase, firm value inreases and the virtuous agent is equally well-off (beause his partiipation onstraint must still be satisfied). First-best is ahieved and the firm aptures the inrease in surplus. With an egoisti agent, the firm annot offer suh a ontrat as the agent would shirk, dereasing the probability that the firm s endeavors are suessful. In general, the firm does not know ex ante whether the agent is virtuous or egoisti beause the agent s morality is private information. The firm also faes an adverse seletion problem based on the skill of the agent. Thus, agents an be skilled or unskilled, in addition to virtuous or egoisti. We show that firms may use inentive ontrats to sreen for skill and to guarantee the provision of effort by the managers they do hire. This is not possible with a fixed-wage ontrat beause suh a ompensation ontrat is appealing to all types of workers. We also show that it is impossible for the firm to profitably sreen for virtuous agents. That is, the firm an never be sure that all managers will exert effort unless it provides them with the appropriate inentives. Therefore, when the firm hooses the optimal employment ontrat to offer an agent, it faes a tradeoff. Contrats that rely on agents ethis (i.e., fixed-wage ontrats or ethis-based ontrats) have better risk-sharing properties, whereas inentive ontrats sreen out unskilled workers and motivate the skilled, egoisti workers to exert effort. There exist onditions under whih eah type of ontrat is optimal. Inentive ontrats are preferred to fixed-wage ontrats when ageny osts are relatively low, when the moral standards of the labor population are relatively poor, and when the firm s prodution relies heavily on labor input, espeially for jobs that require speialized 2

4 skills. As a result, our model predits that ompensation ontrats should be more sensitive to performane in small, high-growth firms heavily invested in researh and development (R&D). In ontrast, ompensation ontrats in large, mature, low-growth firms with large investments in property, plant, and equipment (PP&E) should offer flatter ontrats. After analyzing how these tradeoffs affet the firm s optimal employment ontrat, we onsider the effet of virtue on the firm s hoie of projets and potential for growth. Surprisingly, we find that employee virtue drives firm onservatism. That is, firms are less aggressive in their hoie of projets when they know that the pool of agents from whih they hire is highly ethial. Beause virtuous agents are better mathed with safer projets, firms swith from a high-risk strategy that requires inentive ontrats and ostly risk-sharing to a low-risk strategy and heaper ethis-based ontrats when the labor population s ethis improve. Thus, firms loated in areas or ountries known for their high morality standards should not only offer flatter ompensation ontrats, but also pursue a real investment strategy that is less aggressive and less volatile. Next we onsider the situation in whih a firm must hire two agents for prodution. When the prodution tehnology requires ooperation between the agents (i.e., when the effort of one agent makes the effort of the seond agent more produtive), the firm is less likely to offer ethis-based ontrats to both agents. This is beause, for suh a firm, little is gained from only one agent s effort, and so the opportunity ost of shirking is high. To avoid this possibility, the firm relies on inentive ontrats that ensure joint effort. The opposite is true when the agents efforts are substitutes for one another. In this ase, the firm gains a lot from the first agent who exerts effort, but muh less from the effort of a seond agent. The firm is then more likely to rely on the virtue of its agents and offer fixed-wage ontrats. Indeed, bureaurati firms that essentially require the effort of only one of their agents for suessful prodution an offer fixed-wage ontrats and hope to have at least one virtuous agent in their ranks. Thus, whereas inentive ompensation drives value in firms that require a onerted effort from their agents, virtue is more likely to drive value in more bureaurati firms. Finally, we onsider optimal employment ontrats in a ompetitive labor market. We analyze two versions of the model namely, a sequential hiring game and a simultaneous hiring game. In both ases, firms are less likely to use ethis-based ontrats beause this type of employment agreement makes them vulnerable to predatory hiring by their ompetitors. We show that if one firm offers a fixed-wage ontrat, their rivals an ream-skim away the virtuous types. The intuition is that when a firm offers a fixed-wage ontrat, the expeted utility of egoisti agents is higher than that 3

5 of otherwise idential but virtuous agents. Indeed, both types of agents reeive the fixed wage payment but egoisti agents shirk and save on effort osts. This wedge between the reservation utility of the two types of agents then makes it possible for ompetitors to steal the virtuous types by offering a ontrat that meets their reservation utility but not that of the egoisti workers. This leaves the initial firm with egoisti agents who shirk under an ethial ontrat. Realizing this threat ex ante, the initial firm is more likely to offer inentive ontrats in the first plae. Therefore, ompetition for skilled, sare labor redues the possibility that ethis-based ontrats survive and makes it less likely that virtue reates value in the eonomy. At a general level, our results are onsistent with Bewley s (1995, 1999) survey evidene that firms use ontrats that rely on workers ethis, and with Shmidt and Hunter s (1998) empirial evidene that employee ethis affet job performane. The novel ontribution of our model is to establish a link between population ethis, ompensation ontrats, and firm harateristis. Indeed, many of the model s empirial impliations annot be generated by lassi prinipal-agent models. For example, firms that draw their labor fore from a more diligent population are predited to use flatter ompensation ontrats, adopt a safer investment poliy, observe a lower rate of growth, and implement a more bureaurati organizational struture. 4 To our knowledge, these preditions have never been formally tested previously. Although we do not perform suh tests in the paper, we provide the following guidane for future empirial work. The psyhology literature has developed several measures of integrity, 5 and these measures have been shown to proxy for diligene (e.g., Murphy and Lee (1994) and Marus, Lee, and Ashton (2007)). Importantly, these measures also appear to be unorrelated with skill, job omplexity, and the general mental aptitude of employees (e.g., Ones, Viswesvaran, and Shmidt (1993) and Mount, Barrik, and Strauss (1999)), whih has the benefit of alleviating some of the potential endogeneity problems that may arise in empirial testing. Our theory predits that measures of integrity in the population are orrelated with firm deisions suh as ontrat hoie and investment poliy. 6 Ideally, however, to isolate work ethi as an important determinant of ageny relationships and to further differentiate our model from standard ageny theory, one would want to ontrol for the degree of moral hazard aross firms when performing these tests. 7 One way to proeed would be to study firms that employ agents in many loales to perform similar tasks with similar employee disretion. Another test might involve ontrasting the deisions and inentives of a firm s agents who have similar jobs in the same loation, but who originate from different regions. 8 Another novel predition of our model is that ompetition in the labor market may lead to 4

6 higher-powered inentives. That is, although fixed-wage ontrats an be optimal when the firm has a monopoly on labor, performane-based ontrats are more likely to survive as ompetition inreases. This predition is onsistent with the observation made by Shleifer (2004) that ompetition disourages firms from relying on ethis. However, whereas Shleifer (2004) postulates that this is due to an inrease in unethial behavior, we show that this may arise from the optimal behavior of firms ompeting for labor. In other words, ompetition redues the eonomi value of virtue, but not neessarily the extent of it. Our paper is losely related to the work of Noe and Rebello (1994), but is distint in several respets. They fous on how ethis evolve dynamially over time in business situations. In ontrast, we address the issues of optimal employment ontrats, projet hoie, firm strategy and struture, and firm value. Noe and Rebello model ethial behavior as a large disutility for shirking one a projet is undertaken by an agent. We instead follow Etzioni (1988) and Sen (1997) and model diligene as a self-imposed onstraint. Aording to Rabin (1995), either modelling tehnique is aeptable. We have made sure that this distintion does not hange the results in our paper. Our model of virtue is also similar to how Somanathan and Rubin (2004) model honesty in the sense that agents keep their ommitments. However, the value of this ommitment is assumed exogenously in their work, whereas the value of diligene emerges endogenously through ontrating in ours. In our model, diligene arises from a moral obligation. Another ommon soure of diligene is from intrinsi motivation. The two onepts are related but affet the deisions that people make in different ways. Intrinsi motivation aptures the idea that some people find effort-ostly ativities rewarding, even in the absene of external rewards (e.g., DeCharms (1968) and Dei (1975)). That is, people who are intrinsially motivated derive extra utility from merely partiipating in the ativity. 9 In ontrast, diligene that originates from morality arises beause of self-imposed onstraints, and does not diretly affet the amount of utility that a person ahieves. In partiular, the utility ost that a moral agent inurs when performing an effort-intensive job is the same as that for any other agent. Work ethi and intrinsi motivation also differ in how they are affeted by external inentives. Psyhologists (e.g., Dei (1971) and Lepper and Greene (1978)) and eonomists (e.g., Baron and Kreps (1999) and Bénabou and Tirole (2003)) have shown that extrinsi (e.g., monetary) rewards an rowd out intrinsi motivation. 10 In ontrast, Etzioni (1986, 1988) argues that morality annot be traded off with monetary payoffs. Thus, whereas external rewards an make agents disinterested and remove the impetus for them to innately exert effort, motivating bonuses do not affet a person s ethis. 5

7 The rest of the paper is organized as follows. In Setion I, we set up our benhmark prinipalagent model and formally introdue the notion of a virtuous agent. We also derive the equilibrium ontrats that the firm offers, first in the ase in whih agent types are ommon knowledge, and then in the ase in whih types are privately known by agents. Setion II analyzes how a population s work ethi an affet a firm s hoie of projet and growth prospet. Setion III studies the role of work ethi when multiple agents are required and must interat for prodution. Setion IV studies the effet of ompetition on the survival of ethis-based ontrats. Setion V onludes. The Appendix ontains all the proofs. A. Shareholders and Managers I. The Model Consider an unlevered firm, owned by risk-neutral shareholders (the prinipal), whih initially onsists of F dollars in ash and a risky projet that expires at the end of one period. shareholders have the opportunity to hire a manager (the agent) to potentially improve the projet s expeted profitability. The shareholders fae an adverse seletion problem beause managers may be skilled or unskilled, and only a skilled agent an improve the probability that the projet sueeds. A manager s ability (or skill) is a random variable denoted by Ã, whih takes a value a (0, 1) with probability φ a and zero with probability 1 φ a. Managers privately observe their ability, but annot redibly ommuniate it to the shareholders. The shareholders also fae a moral hazard problem beause a manager s ability gets impounded into firm value only if they exert effort, whih is unobservable. The end-of-period payoff of the firm s projet is given by { σ, prob. Ãẽ ṽ = 0, prob. 1 Ãẽ, (1) where ẽ {0, 1} is the manager s hoie of effort. 11 Only skilled managers (Ã >0) who exert effort (ẽ = 1) inrease the likelihood that the firm s projet is suessful. Otherwise, if the manager is unskilled or does not exert effort, the projet fails with ertainty. In this speifiation, σa is the expeted ontribution of a skilled agent to firm value. Beause σ is a harateristi of the firm and not the agent, we interpret it as the extent to whih the firm depends on labor for its prodution (as opposed to other fators of prodution that are not modelled here). For example, a firm with a large σ ould be one that relies heavily on R&D, whose assets are less tangible, and/or whose produts are more servie-oriented. In ontrast, a firm with 6 The

8 a small σ is one in whih less innovation takes plae, and/or that is more heavily invested in PP&E. Given the empirial evidene that human apital is a key soure of growth (e.g., Shultz (1960) for some early evidene, and Glaeser et al. (2004) for more reent evidene), we sometimes refer to σ as a measure of a firm s growth potential. Parameter a is speifi to the agent and measures the potential impat that he an have on firm value, given the firm s growth potential. Beause the impat that an employee an have on the firm s prodution varies aording to his position, we an also think of a as the agent s rank in the orporate hierarhy of the firm. Managers also differ in their work ethi (or ethis for short). They may be virtuous ( t =1)or egoisti ( t = 0) with probabilities φ t and 1 φ t. Managers privately observe their level of work ethi, but annot redibly ommuniate it ex ante to the shareholders. If hired, egoisti managers only exert effort when the benefits exeed their ost of effort, >0. They are free to hoose ẽ {0, 1} based on the inentives offered by the firm (to be desribed shortly). In ontrast, virtuous agents fae a moral obligation to hoose ẽ = 1 and to inur the utility ost of their effort one they aept the position. Like egoisti agents, virtuous agents aept the firm s employment ontrat only if their expeted utility from doing so exeeds their best alternative, that is, if their partiipation onstraint is satisfied. In making this assessment, virtuous agents aount for the instanes in whih they expet to exert effort without proper inentives to do so. In this model, every agent s type is a pair {Ã, t}, whih is unobservable by the firm. This twodimensional type spae, and the dual adverse seletion problem that aompanies it, ontrasts an agent s potential to ontribute to firm value (Ã) and his innate propensity to do so ( t). Although our analysis fouses on t throughout the paper, we will show that the addition of t affets the solution to the more traditional adverse seletion problem that only onsiders agents skills. In partiular, ontrats tailored for high ethis types ( t = 1) tend to attrat low skill employees (Ã = 0). As suh, the firm must trade off the gains from the natural propensity of some agents to ontribute and the deadweight loss imposed by agents whose potential to ontribute is low. In modelling the agent s preferenes and ethis this way, we are guided by the work of Etzioni (1986, 1988), who argues that morality annot be ranked or traded off like the pleasure that one gets from goods or (lottery) payoffs. Instead, he, Prele (1991), and Sen (1997) suggest that values and priniples ome in the form of self-imposed onstraints in the maximization of one s utility. In partiular, the pain and assoiated utility ost of effort is felt by all; only the propensity to exert effort varies aross individuals. 12 Our representation of virtuous behavior as a morality onstraint is also onsistent with the work of Shelling (1960) and Frank (1987), who argue that 7

9 the onsiene an at as a ommitment devie (in our ase, the ommitment to exert effort). As pointed out by Frank (1987), it is not enough that one has a onsiene to solve any ommitment problem; indeed, it is ruial that others know about it as well. In our model, this last aspet is aptured by φ t, whih measures the population s moral standards and the extent to whih the firm an rely on the agent s ethis. Beause the agent s hoie of effort is unobservable, it is nonontratible. This means that the firm an only offer ontrats that speify the agent s ompensation in the two possible end-of-period states of the world. We denote an employment ontrat in this setting as the pair C {ω, β}, where the manager reeives ω in the bad state (ṽ = 0) and ω + β in the good state (ṽ = σ). As suh, ω is the manager s fixed wage and β is a lump-sum inentive bonus if the projet sueeds. 13 We assume that the agent has no initial wealth and has limited liability, so we restrit ω and ω + β to be nonnegative. In the spirit of Jensen and Mekling (1976) and Treynor and Blak (1976), we assume that beause the manager s human apital is less diversified than that of the shareholders, he is risk averse about the outome of his ompensation. The manager s utility from a ompensation ontrat C = {ω, β} is ũ ω +(1 r)1 {ṽ=σ} β ẽ, (2) where the parameter r (0, 1) aptures the manager s risk aversion. The manager reeives ω for sure, but only reeives β if the projet sueeds and only inurs the ost if effort is exerted (ẽ = 1). Multiplying the seond term by (1 r) measures how muh less utility the manager gets from a stohasti bonus versus ertain ompensation. As r 0, the manager approahes risk neutrality, and as r 1, he only values the riskless portion of his ompensation and effetively beomes infinitely risk averse. This utility funtion is similar to that used by Dow (2004) in that it is pieewise linear with a single hange of slope. It aptures all of the effets that risk aversion has on the agent s hoies while maintaining the model s tratability. Finally, we assume that all managers have a reservation utility of ū>0. B. Equilibrium with Observable Types From the firm s perspetive, the ompensation that it pays the manager is a random variable w ω + 1 {ṽ=σ} β. For simpliity, we assume that the risk assoiated with the firm s projet is idiosynrati and that the risk-free rate is zero. Therefore, the value of the firm at the beginning 8

10 of the period, V, is simply F plus the expeted value of π ṽ w, given the endogenous ations taken by the agent if he is hired by the firm. Without a manager, the firm pays no wage (i.e., w = 0) and is assumed to be unskilled (i.e., à = 0), so that firm value is simply V = F. For the firm to onsider hiring the manager, it is suffiient that σa be large enough relative to and ū. We assume this to be the ase as, otherwise, no hiring ever takes plae. 14 The following proposition derives the optimal ontrats and resultant firm value when agent types are observable. Proposition 1: Suppose that the firm an observe the agent s type {Ã, t}. The firm never hires an agent with à =0. If à = a and t =0, then the firm offers an inentive ontrat C I = {ω I,β I }, where ω I =ū and β I = a(1 r), and the value of the firm is V = F + σa ū 1 r. (3) If à = a and t =1, then the firm offers an ethis-based ontrat C E = {ω E,β E }, where ω E =ū + and β E =0, and the value of the firm is V = F + σa ū. (4) Clearly, the firm never hires an unskilled agent, as suh an agent an never improve the firm s value. When the firm knows it is hiring a skilled agent, it maximizes value by finding the heapest way to ompensate the agent and make sure that he hooses ẽ = 1. For an egoisti agent, the optimal ontrat involves the usual tradeoff between inentives and risk-sharing. The ontrat provides the agent with enough inentive to exert effort. As suh, β I is stritly positive. Sine ompensation is not the same in both states of the world, risk is transferred from the risk-neutral firm to the risk-averse agent, whih is ostly to the firm. The tradeoff between inentives and risk-sharing does not apply when the firm knows it is hiring a virtuous agent. Suh an agent does not require inentives to exert effort one he aepts the position. This is why β E = 0. This does not mean that the virtuous agent gives up surplus; rather, he antiipates his non-inentivized effort and requires a larger fixed wage. This is why ω E >ω I. Ex ante, both agents expet the exat same utility, whih the firm restrits to be exatly ū to meet their partiipation onstraint. A simple omparison of (3) and (4) shows that the loss to r risk-sharing is 1 r. This loss is higher when agents are more risk averse and fae a larger ost of effort. 9

11 C. Equilibrium with Unobservable Types By Proposition 1, the firm prefers to hire a skilled, virtuous agent and eonomize on risk-sharing osts. However, the firm does not observe the skill or the ethis of the agent. Instead it only knows the distribution of the various types. As suh, ontrats are not only used to motivate effort but also to sreen agent types. Our first result shows that the firm annot profit from sreening agents based on their ethis. This is due to the fat that any ontrat that is attrative to a skilled, virtuous agent is also attrative to an otherwise idential but egoisti agent. First, when suh a ontrat inludes a bonus inentive large enough to motivate the egoisti agent to exert effort, both types reeive the same expeted utility from the ontrat. Seond, when the ontrat does not motivate the egoisti agent to exert effort, the expeted utility of the egoisti agent is larger than that of the virtuous agent. This is beause the egoisti agent saves on effort osts and loses an expeted bonus that is worth less than (sine the bonus does not meet his inentive ompatibility onstraint). In short, if the partiipation onstraint of the virtuous type is satisfied, so is the egoisti type s. This is formalized in the following proposition. Proposition 2: Any ontrat that attrats a virtuous agent of a given skill also attrats an egoisti agent of the same skill. All ontrats that attrat an egoisti agent of a given skill but do not attrat a virtuous agent of the same skill do not motivate any effort provision. It is possible for the firm to offer a ontrat that is attrative to the egoisti agent, but not to an otherwise idential virtuous agent. 15 As Proposition 2 shows, all suh ontrats fail to meet the egoisti agent s inentive ompatibility onstraint. This means that the firm ould selet out egoisti agents and hire them, but the ontrat that aomplishes this would indue the agent to shirk. Realizing this, the firm never offers these ontrats as they represent a deadweight loss. 16 Given Proposition 2, the firm an limit its hoie of ontrats to either C I or C E from Proposition 1. First, sine only agents with à = a add value, the firm will hire a skilled type only if the ontrat meets their partiipation onstraint. That is, it must be that E[ũ à = a] ū. Seond, firm value is reated only if these skilled agents exert effort. We already know from Proposition 1 that both C I and C E ahieve this in the heapest manner for eah of the two ethis types. The following proposition haraterizes the tradeoffs faed by the firm when it hooses the ontrat to offer a prospetive agent. 10

12 Proposition 3: When the firm does not know the agent s skill and ethis, it offers the inentive ontrat C I as long as (1 φ a )(ū + )+φ a (1 φ t )σa > φ ar 1 r. (5) Otherwise, the firm offers the ethis-based ontrat C E. The tradeoff in (5) an be appreiated as follows. The two left-hand terms measure the relative benefits of an inentive ontrat (over an ethis-based ontrat), whereas the right-hand side measures its relative ost. The first term on the left-hand side measures the firm s gain from sreening for skill. 17 This benefit is omputed as the probability that agents are unskilled (1 φ a ) times the ineffiient wage that would be lost if the firm had offered an ethis-based ontrat ω E =ū + to an unskilled agent. The seond term on the left-hand side of (5) measures the extra value reated by induing skilled, egoisti agents to exert effort. This benefit is alulated as the probability that agents are skilled and egoisti φ a (1 φ t ) times the expeted value of the projet in their presene, σa. Finally, the right-hand side of (5) measures the loss in value due to suboptimal risk-sharing. Beause the unskilled agents are sreened away when the firm offers the inentive ontrat, the loss from suboptimal risk-sharing is only inurred with skilled agents (i.e., with probability φ a ). As suh, the expeted loss from risk-sharing is alulated as φar 1 r. Thus, the firm s deision to offer an inentive ontrat versus an ethis-based ontrat weighs the benefits of sreening and inentives against the ost of suboptimal risk-sharing. Upon inspetion of (5), it is lear that the use of inentive ontrats inreases with low r,, and φ t, and with high σ and a. If the agent has a low ost of effort (low ) and a low level of risk aversion (low r), risk-sharing is inexpensive and so inentive ontrats are more valuable. If the fration of agents available who are virtuous is low (low φ t ), then it is foolish for the firm to rely on agent ethis for value reation. Finally, if the expeted value of the projet is high (large σ and large a), then shirking represents a big opportunity ost for the firm. In this ase, an inentive ontrat is preferred. These omparative statis are quite instrutive about the types of firms that will be haraterized by flatter ontrats, whih rely mostly on the ethis of agents. Firms whose prodution depends more on labor inputs (large σ) will offer ontrats that have a stronger dependene on firm performane. Suh firms simply annot afford to rely on the expeted ethis of their employees as their main fator of prodution. Similarly, agents who an affet a firm s prodution more signifiantly (large a), either through their skill or their rank in the firm s hierarhy, will reeive a larger 11

13 fration of their ompensation in the forms of stoks, options, and performane-related bonuses. Finally, jobs that require skill-speifi labor will be assoiated with inentive ontrats whereas less speialized labor will be paid with fixed wages. This an be seen in (5) if we derease φ a while keeping aφ a onstant. These preditions are largely onsistent with the existing literature on ompensation ontrats. 18 Mehran (1995), Aggarwal and Samwik (1999), Core and Guay (2001), and Palia (2001) all find a positive relationship between inentive ompensation and various proxies for growth opportunities, inluding R&D, market-to-book, and the intangibility of assets as measured by the (inverse-) ratio of PP&E to assets. Ittner, Lambert, and Larker (2003) also find that new eonomy firms (whih they define as firms in the omputer, software, internet, teleommuniation, and networking industries) provide their non-exeutive labor fore with more equity grants. Beause the profitability of suh firms is largely driven by the quality and effort of labor, relying on virtuous behavior is too risky and done to a lesser extent. The empirial evidene that managerial ompensation in larger firms is less sensitive to performane, as doumented by Demsetz and Lehn (1985), Hall and Liebman (1998), and Core and Guay (2002), is also onsistent with the idea that any one agent annot affet the profitability of these firms as muh as that of small firms. As suh, the reliane on ethial behavior beomes more appealing for large firms. Proposition 3 also yields preditions that further differentiate our model from traditional ageny models. In partiular, the use of ethis-based ontrats does not depend only on the potential ontribution of an agent to firm profitability; through φ t, it also depends on the moral harateristis of the population from whih the firm draws its labor. This has a number of impliations. First, the shape of ompensation ontrats offered for the same job and the same required skill should vary with the soial ontext of the firm. 19 Seond, when φ t is large, we see from (5) that the minimum skill a that justifies an inentive ontrat is larger. This implies that the skill premium, the total ompensation of skilled labor (above a threshold ā, say) relative to that of unskilled labor (below ā), should be smaller in morally sound ultures. 20 II. Ethis, Projet Choie, and Growth So far, we have assumed that the growth potential of the firm σ is given exogenously. In this setion we onsider what happens when the firm has aess to multiple projets with various levels of reward and risk. In partiular, we allow the firm to pik the projet profile that puts the harateristis of the agent population to their most profitable use. Counterintuitively, we find that 12

14 virtue leads to firm onservatism. That is, we show that if the agent population is highly virtuous, it is optimal for a firm to be onservative and hoose low-growth, low-risk projets. The opposite holds for a firm that only has aess to an agent population that is highly egoisti. Beause we fous the rest of our analysis on the effets of privately known ethis, we assume hereforward on that all agents are equally skilled. That is, we assume that φ a = 1, so that sreening for skill is no longer required (Ã = a) and an agent s type is simply t. Also suppose that, instead of being assigned the projet ṽ desribed by (1), the firm an pik one projet from a family of available projets ṽ µ indexed by µ [0, 1 a] and that this hoie ours before the agent is hired. For any given µ, the end-of-period payoff is { σ µ, prob. aẽ + µ ṽ µ = 0, prob. 1 (aẽ + µ), where σ µ is alulated so that the mean payoff of any projet is the same given that the agent exerts effort. More preisely, E[ṽ µ ẽ =1]=(a + µ)σ µ is equal to the same value m for all µ [0, 1 a]. This means that σ µ = m a+µ for any µ. Even though all of the projets have the same expeted value given that the agent exerts effort, they differ along three important and related dimensions: the amount of firm growth when the projet sueeds, the riskiness of the projet, and the total ost that the firm inurs to motivate the agent. Projets with a high µ are less risky, as Var [ ṽ µ ẽ =1 ] = σµ(a 2 + µ) [ 1 (a + µ) ] ( ) 1 = m 2 a + µ 1 is dereasing in µ. However, these projets yield a lower payoff when suessful, as σ µ is dereasing in µ. We assume that the firm is ommitted to its hoie of projet µ when it makes its ontratual offer to the agent, that is, the firm does not revise its hoie of µ one the agent aepts the ontrat. Otherwise, the agent ould, in his deision to aept or rejet the ontratual offer, antiipate and undo any projet-shifting behavior on the part of the firm. In this sense, the hoie of µ aptures the idea that firms set their general strategy (e.g., the industry in whih they operate or their loation) before aquiring all the labor resoures that they require for prodution. As in Setion I.C, for any projet µ the firm may limit its sope of ontrats to one inentive ontrat, whih we denote by C I (µ), and the ethis-based ontrat derived in Proposition 1. The argument for this follows a similar logi as before. (6) Speifially, C I (µ) is the heapest inentive ontrat that indues all agents to exert high effort for a given µ. If the firm intends to rely only on 13

15 virtuous agents, then the most effiient ontrat is C E. The following lemma formally establishes this result and derives the C I (µ) ontrat. Lemma 1: Conditional on undertaking projet µ, the firm may limit its ontrat hoie to the following ontrats: an inentive ontrat C I (µ) {ω I (µ),β I } with ω I (µ) =ū µ a and β I = a(1 r), (7) or the ethis-based ontrat C E of Proposition 1. It is lear from Lemma 1 that the firm s hoie of µ will affet the total ost of the assoiated inentive ontrat as well as the relative ontributions of the fixed wage and bonus to this ost. For a given µ, the expeted wage that is paid with C I (µ) is E [ w ] = ( ū µ ) + (a + µ). a a(1 r) The fixed wage portion of the firm s ost (the first term) is monotonially dereasing in µ, whereas the expeted bonus (the seond term) is monotonially inreasing in µ. This is beause any inentive reated through β omes with some expeted bonus ompensation that is not tied to the agent s effort but to other fators of prodution impliitly inluded in µ. More preisely, the smallest bonus that motivates an egoisti agent s effort, β I, implies that the agent not only reoups his effort ost of but also reeives an expeted bonus of β I µ. When µ is high, more of the agent s ompensation and expeted utility omes from this expeted bonus. This makes risk-sharing more ostly and the firm more likely to use ethis-based ontrats. In ontrast, when µ is lose to zero, the bonus β I required for effort inentives is lower as the agent s effort is the main input to prodution and the agent annot free-ride on other fators. As suh, the risk-sharing osts of induing an egoisti agent to exert effort are smaller for suh projets, and the firm is more likely to use inentive ontrats. This relationship between projets and ontrats impats the firm s optimal projet hoie, given the population of agents it faes. For a given projet µ, the firm will hoose C I (µ) and C E based on the tradeoff between the benefits of inentives and the ost of risk-sharing. Then, in hoosing the optimal µ to maximize firm value, the firm not only onsiders the harateristis of the agent population, but also whih employment ontrat maximizes value. The following proposition haraterizes the firm s optimal strategy. Proposition 4: The firm hooses the projet µ =0and offers the manager the inentive ontrat 14

16 C I (0) = C I if and only if r m> a(1 r)(1 φ t ). (8) Otherwise, the firm hooses the projet µ =1 a and offers the manager the ethis-based ontrat C E. Condition (8) tells us that the firm will hoose to be a risky growth firm (µ = 0) when agents are highly skilled (large a), have low aversion to risk (low r), and a low ost of effort (low ). More interestingly and less intuitively, it also tells us that suh projets are less likely to be undertaken when the firm an rely on a highly ethial population of potential agents (large φ t ). In that ase, the firm prefers the better risk-sharing properties and lower ost of the ethis-based ontrat, and aordingly hooses a projet that is safer and does not depend as heavily on the agent s effort for its suess. In other words, the firm hooses to rely on the effort of virtuous types as well as on the fat that the projet µ =1 a still sueeds with probability 1 a when the egoisti agent shirks. In short, virtue leads to firm onservatism. Interestingly, empirial support for this predition an be found in the work of Hilary and Hui (2006), who show that firms whose headquarters are loated in areas of high religion membership tend to be more onservative on average in their investment and finaning deisions. We an see from (8) that, as before, a virtuous population (large φ t ) will tend to be assoiated with a flatter assoiation between wages and performane. Proposition 4 also shows that suh a population will prompt loal firms to adopt a safer investment poliy and, given our previous arguments, these firms will experiene lower growth on average. Again, this is onsistent with the aforementioned empirial evidene of a positive relationship between inentive ompensation and growth. However, the impliations of our model an now be further refined: in high-virtue ultures, we should see a smaller proportion of growth firms, slower eonomi growth, and a smaller fration of individuals with high-powered inentives. Also, sine a smaller fration of skilled employees will reeive bonus ompensation, the ross-setional distribution of wages should be more skewed and the skill premium should be lower. As in Prendergast (2002), our theory predits that the relationship between risk and inentives is affeted by other deisions that are endogenously made by the firm. Whereas he posits that the monitoring mehanisms that the firm puts in plae will differ aording to environment risk, we argue that the atual hoie of making the environment risky will be endogenously driven by the morality of the labor pool. Indeed, firms that have aess to an ethial population of agents will 15

17 make their operations less risky and rely on flat ontrats. Firms in less morally driven ultures must motivate their workers through inentives and these inentives are more powerful when the firms underlying operations are riskier in the sense that they depend more on labor and less on other fators of prodution. III. Multi-agent Projets and Firm Organization Up to this point, we have foused on a situation in whih only one agent is required to improve the firm s profitability and value. In this setion we analyze how ethis affet the firm s hoie of employment ontrats when multiple managers work towards a ommon goal. When agents work in teams, they not only affet eah other s produtivity, but also their inentives to exert high effort (e.g., Anderson and Shmittlein (1984) and Holmström and Milgrom (1991)). In what follows, we onsider two general types of projets: those that require ooperative effort (synergyintensive projets) and those in whih synergy is not required to reate value. As we show, the proportion of virtuous agents in the population and the synergy required in the projet drive both the employment ontrats that are offered and the firm s organizational struture. As in Setion I, let us assume that the firm has F dollars in ash and that it owns a projet with a random payoff whose prospet an be affeted by labor. The firm now must hire two agents, i {1, 2}, to implement a projet with a payoff given by { σ, prob. aẽ γ ṽ = 0, prob. 1 aẽ γ, where γ>0, ẽ = 1 2ẽ ẽ2, and ẽ i {0, 1} represents the effort deision of agent i. As suh, the prospets of the projet depend both on the effort exerted by the two agents and the degree of synergy required, whih is parameterized by γ. When γ>1, the effort of the first agent is less valuable than the effort of the seond agent, as ( ) 1 γ 2 < 1 2 but 1 ( ) 1 γ 2 > 1 2. When γ is large, the projet requires ooperation by the agents sine the effort of only one agent is worth very little to the firm. The opposite is true for γ<1. For example, as γ 0, the effort of one agent is enough for ẽ = ( 1 γ 2) to approah one, and the seond agent s effort then has very little impat on the projet s probability of suess. As before, beause the effort of eah agent is unobservable, the ompensation ontrats that the firm offers eah of the two agents only depend on the two possible realizations of ṽ. Thus, eah agent i reeives a ontrat C i {ω i,β i } that speifies the fixed-wage (ω i ) and bonus (β i ) (9) 16

18 portions of their ompensation. The agents simultaneously aept or rejet these ontrats, and we assume that the terms of both ontrats are ommon knowledge to eah agent when they make their deisions. Beause the effort of one agent affets the produtivity of the other, the two agents effetively engage in a oordination game. As we know from standard game theory, multiple equilibria often our in these irumstanes. 21 In partiular, the usual definition of a Nash equilibrium often allows for both of the agents to work or both of the agents to shirk. 22 We adopt the onvention that, in these irumstanes, agents oordinate to pik the Pareto-dominant equilibrium. As our analysis will show, when multiple equilibria are possible, the two agents always get symmetri expeted utility in all equilibria. As suh, the Pareto-dominant equilibrium ould arise simply from Shelling s (1960) foal-point effet. The following lemma defines the hoie set that the firm faes when it deides what types of ontrats to offer eah agent. To make the notation simpler, we define the parameter g 2 γ as the group effort ẽ when only one of the two agents exerts effort. Lemma 2: When hoosing the ontrats to offer the two agents, the firm an restrit its deision to three sets of ontrats: (i) C EE {C E, C E }. Two ethis-based ontrats, as defined in Proposition 1. (ii) C EI {C E, C I1 }. One ethis-based ontrat as defined in Proposition 1, and one inentive ontrat with ω I1 =ū φ tg α where α φ t (1 g)+(1 φ t )g. (iii) C II {C I2, C I2 }. Two inentive ontrats with and β I1 = aα(1 r), (10) ω I2 =ū g 1 g and β I2 = a(1 g)(1 r). (11) As before, when offering inentive ontrats, the firm always uses the smallest possible bonus required to motivate an egoisti agent to exert effort and otherwise uses fixed-wage ompensation that satisfies the agents partiipation onstraint. If the firm offers both agents the same fixed-wage ontrat with C EE, it essentially treats them as equals. In this ase, the organizational struture of the firm is stritly horizontal. In ontrast, when the firm offers C EI, it does not treat its agents equally and the firm has a more hierarhial organizational struture. More speifially, the firm 17

19 indues one agent to work hard by paying an inentive bonus, but pays the other a fixed wage and relies on his virtue. The agent reeiving the bonus an be onsidered the leader in the organization, whereas the agent with flatter inentives is the underling. With C II, the firm treats the two agents the same by offering them the same inentive ontrat. In this ase, there is no natural leader, but the bonus inentives provided to eah agent are reminisent of those seen in partnerships, as the agents share the value that their joint effort reates. It is easy to verify that β I2 >β I1 if and only if g> 1 2, that is, if the firm s projet does not require muh synergy between the two agents (i.e., when γ<1). Beause bonus inentives redue the effiieny of risk-sharing between the firm and its agents, this means that it will be expensive to sustain ooperation between the two agents when that ooperation is not ritial to the firm s suess. We therefore expet firms with low levels of synergy aross their agents to rely more heavily on ethis-based ontrats. This is formalized in the following proposition, whih haraterizes the onditions under whih the firm optimally hooses the different ontrats. Proposition 5: The firm hooses to offer C EI over C EE if and only if σa(1 r) r > φ t +(1 φ t )g (1 φ t )α 2 Σ EI EE, (12) it hooses to offer C II over C EE if and only if σa(1 r) r > 2 [ 1 φt (α + g) ] (1 g) ΣII EE, (13) and it hooses to offer C II over C EI if and only if σa(1 r) r > φ t(1 g)+(1 φ t )g(1 + g) (1 φ t )(1 g) 2 α Σ II EI. (14) Moreover, there is a unique ĝ (0, 1) suh that Σ II EI > Σ II EE > Σ EI EE for g>ĝ, and Σ II EI < Σ II EE < Σ EI EE otherwise. It follows from Proposition 5 that as the ratio (denoted by Σ) of projet value (σa) to the ost of inentive provision ( r 1 r ) rises, the firm tends to inrease the inentives that it provides to its agents. Holding all else onstant, when this ratio is low, the firm tends to offer both agents an ethis-based ontrat. As the ratio rises, the firm offers one agent an inentive ontrat as long as g>ĝ, that is, as long as the synergy between the two agents is low (low γ). Eventually, regardless of the level of synergy between the two agents, if Σ rises suffiiently, the firm provides both agents an inentive ontrat. 18

20 To better grasp the results of Proposition 5, it is useful to illustrate the regions in whih eah set of ontrats is used by the firm. We do this in Figure 1, whih has three graphs orresponding to different levels of φ t (0.3, 0.5, 0.7). The synergy parameter γ is measured on the x-axis and the ratio Σ = σa(1 r) r is measured on the y-axis. As an be seen from the three graphs, the firm will tend to rely more heavily on ethis as γ and Σ derease and as φ t inreases. That is, firms whose prodution relies mainly on the effort of only one of its workers (i.e., firms with low γ) will offer ethis-based ontrats and take the hane that one of the two agents is virtuous. Firms that an only be suessful when both of its agents exert effort (i.e., firms with high γ) do not have the same luxury. Unless φ t is lose to one, they must use inentive-based ontrating to ensure that both Insert Figure 1 here agents exert effort as, otherwise, little is gained from the agents presene. With multiple agents, it is the nature of the prodution proess that predits how muh eonomi value is reated by virtue. Firms that do not require muh ooperation between their agents tend to offer flatter ontrats and rely on diligene. As a result, virtue is a key driver of performane for these firms. In ontrast, firms that strongly depend on the synergisti interations of their agents prefer to offer inentive ontrats and rely less on diligene. This should be the ase for firms that depend on the joint reputation of their human apital (e.g., law firms), whose divisions depend on eah other s prodution (e.g., vertially integrated firms), or are involved in strategi allianes. For these firms, virtue is not likely to be a signifiant driver of value. Note that the integrity measures disussed in the introdution ould serve to test this predition as well. Indeed, when measures of inentives and integrity are used to explain prodution in a regression model, we would expet inentives (integrity) to dominate the relationship for firms that are more (less) synergy driven. IV. Ethis and the Competition for Labor In previous setions, we analyze the tradeoffs between inentive and ethis-based ontrats, given that the firm is a monopolist in the labor market. In this setion, we onsider how firms optimally hoose their employment ontrats when they ompete for labor. We show that ompetition makes it harder for ethis-based ontrats to survive as it renders ethis sreening possible. Speifially, ompetition makes it possible for firms to steal the agents that are virtuous. We demonstrate this first with a sequential hiring game in whih an entrant firm arrives with positive probability and has the option to steal an agent from the inumbent firm. Subsequently, we onsider a simultaneous game in whih both firms bid for the servies of a single agent. 19