CEO Compensation, Change, and Corporate...

24
CEO Compensation, Change, and Corporate Strategy James Dow London Business School Clara C Raposo ISCTE 27 January 2003 Abstract We study how CEO compensation can inuence the kinds of strate- gies that rms adopt. With performance-related compensation, CEO’s might have a tendency to look for overly ambitious, hard to implement, strategies. They do so in order to boost their own compensation, and not necessarily shareholder value. At a cost, shareholders can curb this tendency for excessively dramatic strategy choice, by pre-commiting to a regime of CEO overcompensation in highly changeable environments. Another way to avoid overambition would be to adopt a regime of com- mitment to low pay. This however might be infeasible with renegotiation, and highly costly if there are alternative value-increasing strategies. JEL numbers : G30, G34, J33, D8. Keywords : agency theory, executive compensation, strategic change. “We have created a cult of leadership that far exceeds anything that existed decades ago... What we are getting now, very danger- ously, is what I call a dramatic style of managing; the great merger, the great downsizing, the massive brilliant new strategy... So we get all these massive mergers, re, brimstone and drama, because you can’t say to the stock analysts, ‘we’re getting our logistics all straightened out, we’re going to be much more ecient at through- put to the customer.’ They start to yawn.” (Mintzberg (2000)) Supported by the EU-TMR Research Network Contract FMRX-CT960054. We thank the editor and the referee for very helpful comments. We also thank Aya Chacar, Jan Mahrt- Smith, Jukian Franks, Denis Gromb and Michael Raith for discussions and written comments on an earlier draft. We are grateful for comments from seminar audiences at Oxford, Toulouse, Amsterdam, Pennsylvaia, Chicago GSB, Northwestern (Kellog School), University College London, Pompeu Fabra, Yale SOM, the EFA 2001, AFA and Econometric Society Winter Meetings 2002. 1

Transcript of CEO Compensation, Change, and Corporate...

CEO Compensation, Change, and CorporateStrategy∗

James DowLondon Business School

Clara C RaposoISCTE

27 January 2003

AbstractWe study how CEO compensation can influence the kinds of strate-

gies that firms adopt. With performance-related compensation, CEO’smight have a tendency to look for overly ambitious, hard to implement,strategies. They do so in order to boost their own compensation, and notnecessarily shareholder value.

At a cost, shareholders can curb this tendency for excessively dramaticstrategy choice, by pre-commiting to a regime of CEO overcompensationin highly changeable environments.

Another way to avoid overambition would be to adopt a regime of com-mitment to low pay. This however might be infeasible with renegotiation,and highly costly if there are alternative value-increasing strategies.

JEL numbers : G30, G34, J33, D8.Keywords : agency theory, executive compensation, strategic change.

“We have created a cult of leadership that far exceeds anythingthat existed decades ago... What we are getting now, very danger-ously, is what I call a dramatic style of managing; the great merger,the great downsizing, the massive brilliant new strategy... So weget all these massive mergers, fire, brimstone and drama, becauseyou can’t say to the stock analysts, ‘we’re getting our logistics allstraightened out, we’re going to be much more efficient at through-put to the customer.’ They start to yawn.” (Mintzberg (2000))

∗Supported by the EU-TMR Research Network Contract FMRX-CT960054. We thank theeditor and the referee for very helpful comments. We also thank Aya Chacar, Jan Mahrt-Smith, Jukian Franks, Denis Gromb and Michael Raith for discussions and written commentson an earlier draft. We are grateful for comments from seminar audiences at Oxford, Toulouse,Amsterdam, Pennsylvaia, Chicago GSB, Northwestern (Kellog School), University CollegeLondon, Pompeu Fabra, Yale SOM, the EFA 2001, AFA and Econometric Society WinterMeetings 2002.

1

1 IntroductionThe problems resulting from separation of ownership and control have long beenrecognised in the corporate governance and corporate finance literature (Berleand Means (1932), Jensen (1986), Hart (1995), Shleifer and Vishny (1997)).Mintzberg’s analysis of the cult of leadership has three steps: (i) companiesneed to change and adapt; (ii) the CEO, rather than the shareholders, haspower to decide on the direction of change (the strategy) and (iii) the CEOmay not select the optimal kind of change (for shareholder value): it may beoverly dramatic. In this paper, we study this idea from the perspective of thecorporate governance literature, i.e. using agency theory. Note that much of thisliterature has used an incomplete contracting approach (Hart (1995)), and eachof the three steps in Mintzberg’s argument suggests contractual incompleteness.This is the approach we take here.Standard agency theory takes as given an incentive problem, the principal

and agent then agree on a (constrained) optimal contract, and then the agentgoes to work on the problem. One feature of CEO compensation that is clearlydifferent to this standard agency model is that the contract is adjusted overtime to reflect the evolution of the firm’s performance and its strategic direction.There is an annual pay setting round at which options and incentive plans arerenegotiated. Since he or she can influence the firm’s strategy, the CEO may beable to influence the compensation contract. A dramatic merger, or restructuringof the whole business, can lead to larger options grants.There are many examples in recent corporate history where radical corpo-

rate change went hand in hand with high executive compensation and optionsgrants: Coca-Cola under Roberto Goizueta, the Daimler-Chrysler merger, GEunder Jack Welch, Chris Gent and the Vodafone-Mannesmann takeover, Enron,the Glaxo Wellcome — SmithKline Beecham merger. Some of these dramaticchanges were successful, others were failures.1 Our paper is not about disas-trous dramatic change. It is about change that is expected positive-NPV at theoutset, given the available information, that may end up being either success-ful or unsuccessful depending on the outcome, but that is overly dramatic inthe sense that less radical change would have been higher NPV. We argue thatthe way executive compensation responds to changes in strategy can lead tomanagement choosing change that is more radical than the shareholders wouldoptimally prefer: the CEO of a regional electricity utility may find it personallymore profitable to create a global web-based energy market-maker. We thenlook at ways the incentive contract can be modified at the outset to anticipatethis problem.This analysis is consistent with (but not identical to) two themes that run

through much corporate finance research: free-cash-flow theory, and non-value-creating mergers. First, there is free-cash-flow theory. Jensen (1986, 2000),building on earlier analyses of managerial empire-building (Baumol (1959), Mar-ris (1967), and Williamson (1964)), has argued that managers of public corpo-

1Among the failures, some can only be criticised with the benefit of hindsight, while othersseem to have been doomed from the outset.

2

rations have a systematic tendency to overinvest.2 Overinvestment of free-cash-flow can be viewed as similar to overly dramatic change. For example, cash-generating low-growth businesses may tempt their managers to seek growththrough excessive diversification. Lang and Litzenberger (1989), Lang, Stulzand Walking (1991), Mann and Sicherman (1991), Blanchard, Lopez-de-Silanesand Shleifer (1994), Kaplan and Zingales (1997), and Lamont (1997) provideempirical evidence supporting free-cash-flow theory.Second, there is the evidence that many mergers add little or no value to

the acquirer (Asquith, Bruner and Mullins (1983), Jarrell, Brickley and Netter(1988), Bradley, Desai, Kim (1988), Jarrell and Poulsen (1989), Bruner (2002))3.Mergers are a clear example of dramatic change and hence, this evidence canhelp explain why managers may undertake dramatic acquisitions even when thisis not shareholder-value maximising.Of course, change is valuable and necessary: we do not argue that change

is bad. In this paper, we argue that firms that do change may sometimeschange in the wrong way. This may be particularly relevant because the recentyears have been a period with a high rate of corporate change. For example,the 1990’s was a decade of mega-mergers. US M&A activity over 1993-1999amounted to an annual average of 8.4% of GDP, compared with less than 4%in the 1980’s and less than 2% in the 1970’s (Weston, Siu and Johnson (2001),table 7.4). While in previous decades merger targets were typically about 10%of the size of the acquirers, in the 1990’s it became common for companies toacquire targets almost as large as, or sometimes even larger than, themselves(AOL-Time Warner and Vodafone-Mannesmann are examples).The basic outline of our model is as follows. We study a situation where

shareholders are able to set compensation optimally given incentive compatibil-ity constraints, but where top management has the advantage of formulatingstrategy. In this agency problem, after the CEO has chosen a strategy, incentivesare set or adjusted before the CEO proceeds to implementation of the chosenstrategy. The CEO’s ability to formulate strategy is part of his or her job, sothis is a natural assumption. We use a setting with incomplete contractibility4

and limited liability5 for the manager, leading to an option-like contract (re-

2He has even argued that this is so costly that the public corporation is not an efficentinstitutional vehicle for business ownership, and should be replaced by other institutions(Jensen, 1989). However, Jensen does not share the perspective offered in this paper thatstrong incentives may actually exacerbate managerial conflicts of interest.

3There is some overlap between FCF theory and research on non-value-creating mergers,since free cash flow may be spent on acquisitions. However, free cash flow may also be spenton other projects, and many acquisitions are paid for with external finance (new debt or newstock).Note also that our model can explain mergers that add little value to the acquirer, but

cannot explain mergers that are value destroying .4A detailed motivation of the form of contractual incompleteness, and discussion of this

hypothesis, is given in the appendix.5This feature of CEO compensation is sometimes criticized, but nevertheless it seems to

be almost universal (unlike small private businesses, where the entrepreneur often pledgescollateral to a bank or VC). In Tirole’s (2001) recent graduate textbook on corporate finance,all of the models make the same assumption as we do. Studying the economic rationale for

3

ward for success, no penalty for failure) conditioned on firm value. We assumethat change will require substantial effort from the CEO at the implementationstage, while business as usual requires much less. Hence the reward for successmust not only induce the CEO to put in the required effort to implement changecompared to the alternative of no effort, but also compared to the much easiertask of maintaining the status quo. Since implementation of a given strategy isnon-contractible, we show that the latter becomes the binding constraint, andthe more personally demanding are the alternatives he finds to the status quo,the higher his surplus6. By choosing a task whose success is highly dependenton his own performance, the CEO is able to extract higher surplus from hisshareholders. We conclude that high-powered incentives can encourage overlydramatic strategies.7

Anticipating this divergence, what could shareholders do? If they are con-cerned that the CEO’s incentives are not well aligned with their own, they couldsimply give him enough options or equity at the outset and the conflicts woulddisappear. In other words, if they are worried that corporate strategy may bedistorted by the CEO so as to extract larger incentive packages of equity or op-tions, they could simply hand over the large package at the outset anyway. Thisscenario, which we call “ex-ante contracting” works but is expensive: there is atrade-off to be made. In very unstable environments, in which it is likely thatchange (and particularly, dramatic change) is possible, setting such an ex-antecontract might be optimal from the shareholders’ perspective. Thus, a com-mitment to high compensation can reflect strategic discretion, without beingdirectly related to the CEO’s effort cost of implementation.The drawback of this contracting policy is that the shareholders commit in

advance to high pay (as a deterrent to overly dramatic change) even though, ex-post, sometimes dramatic change will turn out not to be an option anyway. Weassume that the firm’s strategic environment is not completely predictable inadvance, and that the CEO has an advantage in being better informed about itthan the shareholders. Hence there is a range of possible strategic choices, notall of which are always available. In some states of the world, dramatic changeis the only (positive NPV) alternative to the status quo, in other states moremoderate change is also available (and is higher NPV than dramatic change),while in other states dramatic change is not an option. In the last case, ex-ante contracting means that the CEO ends up being unnecessarily highly paidto implement a simpler strategy. Therefore in less changeable environments,shareholders may prefer a wait-and-see approach to contracting, setting the

this feature, if there is one, would be interesting but is not the subject of this paper.6Core and Guay (2002) present empirical evidence that more unstable environments tend

to be positively related to CEO pay-performance sensitivity.7Note that this is different from two other problems that may occur with high-powered

compensation: (i) straight cheating by managers; (ii) managerial rent-seeking through controlof the contracting process. Both of these problems are important and relevant: CEO’s mayuse false accounting to overstate profits and inflate the stock price, and instead of representingshareholders’ interests, boards may cooperate with management in agreeing excessive com-pensation. (The recent confrontation between board and shareholders of Glaxo SmithKlineis an illustration). However, those are not the problems we study here.

4

compensation package after the strategy is chosen while accepting that thismight cause distortion in strategy.Another way for shareholders to discourage excessively dramatic strategies

would be for them to refuse to go along with the whole notion of dramaticchange and the associated high compensation, by committing in advance not topay high salaries. It is not clear that shareholders can credibly commit, sinceex-post they may want to renegotiate, but perhaps social norms can provide amechanism for limiting CEO pay. We therefore extend the analysis to investigatethis case also. This will work in dissuading the CEO from unnecessary dramaticchange, but again there is a trade-off: it is equally effective in dissuading theCEO from dramatic change in states when it is desirable.How can the conclusions of our model be used to throw light on the evolution

of executive pay in the past 10-15 years? We believe that during this period,shareholders have become much more aware of the need for large public firmsto change. Partly this is due to the revolutionary technical changes that havetransformed the economic landscape. It may also be that shareholders simplybecame more aware of this general issue, following a period in the late 1980’swhen many firms that had become insufficiently focused on value creation wererestructured by external means (LBO’s and other hostile takeovers). In terms ofour model, we could hypothesise that before this period, there were informal po-litical and social conventions that effectively limited CEO pay (pre-commitmentto a limit on renegotiation), and this limitation either became suboptimal be-cause of changes in the parameters of the model, or it became unsustainableand broke down as the pressure for change suggested higher and higher CEOpay (ex-ante contracting).Our analysis can be used to interpret the available empirical evidence on

top management pay in the US. Hall and Liebman (1998) document that onaverage a 1% increase in stock price led to an increase of $124,000 in CEOwealth in 1994, while in 1998 this value exceeded $500,000. Murphy (1999)shows a sharp increase in pay-performance sensitivity over the early 1990’s,reaching a level almost double the 0.325% reported in the seminal Jensen andMurphy (1990) article. Conyon and Murphy (2000) report similar findings forthe UK, showing that both the level and the pay-performance sensitivity of CEOcompensation increased sharply over the 1990’s. One can interpret these studieson the 1990’s as evidence of a period of high changeability and a commitmentto high compensation (”ex ante contracting” in terms of our model).Core and Guay (2001b) describe the cross-sectional variation of executive

compensation, finding that the median large firm has options outstanding thatamount to 5.5% of common stock, rising to 10—14% for growth industries (com-puter, software, pharmaceutical) but lower at only 2-3% for low growth in-dustries such as utilities and petroleum firms. Likewise Murphy (1999) findspervasive use of stock options in most industry groupings, but not in utilities.One can interpret this, again, as evidence of higher strategic discretion of CEO’sand more changeable environments in growth industries. An alternative expla-nation, in line with standard agency theory, is that higher compensation in highgrowth industries simply reflects higher agency costs of effort (or private benefits

5

of control). Plausibly however, while managerial jobs in growth industries maybe more challenging and more onerous, this difference in effort seems unlikelyto be large enough to explain very large divergences in compensation packages.Along the same lines as the findings of the above studies, Demsetz and Lehn(1985) and Core and Guay (2003) find a strong positive relationship betweenfirm risk and CEO pay-performance sensitivity8.Core and Larcker (2003) examine firms with ”target ownership plans” spec-

ifying the minimum amount of stock that must be held by executives and showthat non-CEO executives typically hold much less equity (relative to base salary)than the CEO. We can interpret this as evidence that given the special role ofthe CEO, his or her incentives should be stronger to give good incentives instrategy formulation.We now comment on the relationship of this paper to some of the existing

agency literature. First, there is the multi-tasking version of the principal agentproblem in which the agent has two tasks to which he can devote effort, butthe output of only one of those tasks is measurable (Holmstrom and Milgrom,1991). Making incentives more high-powered relative to that output measurecan be counterproductive, reducing the agent’s effort on the other task. Thereis a similarity with our model in that incentives for one variable (effort, in ourmodel) can distort another variable (strategy choice). However, in our modelboth variables contribute to the same measured output (firm value). In ourmodel, the analysis is driven by the incomplete nature of the contracting, thesequencing of the agents’ choices, and the opportunities for renegotiation of thecontract, which are absent from the multi-tasking model.Free cash flow theory (Jensen (1986)) suggests that challenging strategies

can be beneficial for CEOs in terms of private benefits. The private benefits arenot explicitly modelled: they are exogenous. Rather than direct private benefits,we consider indirect benefits arising from the effect of managerial activity oncompensation. Hence, while free cash flow theory is rather broader than ouranalysis, our paper could be viewed as compatible with free cash flow theoryand as offering a rationalization of how private benefits can arise.Shleifer and Vishny (1989) argue that managers have an incentive to entrench

themselves by making investments that create specific human capital. They willfavour projects that they alone can operate and cannot easily be transferred toanother manager, much like a computer programmer who writes a deliberatelycryptic code that makes him or her indispensable and able to extract rents. Inour paper, there is no specific human capital. Shleifer and Vishny’s effects relyon the manager manipulating his position in the labour market, while the effectsin our paper rely on manipulating the agency problem.Prendergast (2002) considers a setting where the agent has discretion in

deciding how to solve the problem. He starts by noting that the majority ofempirical studies find a positive relationship between the risk of an agencyproblem and the strength of the agent’s incentives, contrary to the predictions

8Core and Guay (2003) contains a detailed discussion of empirical evidence on this issueand its possible interpretations.

6

of the standard model. This empirical evidence is compatible with our analysis.He then points out that in reality agents can decide how to go about solvinga problem, which is similar to our point that CEO’s decide strategy as well asimplementing it. He goes on to address different issues to the ones we addresshere, concluding that if a problem is uncertain, the principal may not know howto solve it, so he will let the agent decide that and motivate him with strongincentive pay, while if a problem is predictable the principal will know how tosolve it and will directly monitor the agent’s actions instead.The paper is structured as follows. Section 2 presents the model and the

main results of the paper. In section 3 we consider an extension in terms ofthe contracting mechanism, allowing for the firm to pre-commit to a ceiling incompensation. Section 4 presents brief concluding remarks.

2 Modelling strategy formulation and implemen-tation

In this section we start by presenting the main ingredients and structure of themodel. We assume the firm is run by a risk neutral manager with no personalfinancial resources. There are three main steps for him to take: he formulatesa strategy, takes it to the shareholders for approval, and then implements it.The outcome of the strategy formulation process - i.e., the strategy that isproposed by the manager - is influenced both by the manager’s choice and byrandom factors9. There are two strategies for change that may, or may not,be feasible. The manager learns whether change is feasible, and if so, whichkind of change, and then chooses to formulate one of them. The shareholdersdo not observe which of these two might be feasible. They are denoted Mand D, corresponding to notions of “Moderate change” and “Dramatic change”respectively. In addition the status quo option B “Business-as-usual”) remainsavailable.We assume that there is a probability p (known by all) that change will be

feasible. With probability 1− p only strategy B will be available. If change ofsome nature is feasible, the manager will learn whether it is moderate change(strategy M) that is possible (with probability qM ), or dramatic change (strat-egy D) (probability qD) or both (probability 1 − qM − qD). In this last casethe CEO will have to choose whether to investigate strategy M or strategy D.Whereas the manager observes these realizations (i.e., whether change is feasi-ble, and if so, whether M , D, or both), the shareholders will only learn whatthe CEO communicates. We assume that the CEO cannot credibly invent feasi-ble strategies when he communicates with the shareholders, but he can concealthem if he wants to.

9 If it were influenced only by the manager’s choice and there were no uncertainty, theshareholders would know as much about the firm’s strategic environment as the manager. Inthis case, it is unrealistic to suppose the choice of strategy would be delegated to the manager.Even if it were, the trade-offs betwen ex-post and ex-ante contracting, and contracting withpre-committment would become rather trivial.

7

So, at the time the CEO decides which strategy to formulate, there are fourpossible combinations of alternatives:

No alternative to B, with probability 1− p

M , with probability pqMD, with probability pqDM and D, with probability p (1− qM − qD)

In the fourth case, there is potential for the manager to make a choice ofwhich strategy to formulate and this may or may not be in the shareholders’interests. Distortion of the strategy choice means that the manager would chooseD while, in first-best, the shareholders would prefer M , or vice versa. In thefirst, second and third cases, the manager has no discretion.The next stage is for the CEO to present his strategy to the shareholders

for their agreement, and then proceed to implement the chosen strategy. Thedetails of the contracting process between shareholders and CEO are describedin the next subsection.We need to complete the description of the strategies by characterizing them

in terms of parameters, and imposing some conditions on these parameters.Strategy i (for i = B, M,or D) is characterized by effort from the CEO costinghim ei and probability πi of success, in which case the firm is worth V . Otherwise(no effort, or sufficient effort but bad luck) the firm is worth nothing.We assume that strategy M (“moderate change”) involves substantially

more effort than carrying out business as usual (eM > eB) and also a higherrate of success (πM > πB). We make a stronger assumption, that eM

eB> πM

πB.

This means that change requires higher effort than the status quo and has ahigher chance of success, but proportionately the personal effort increases bymore than the chance of success of the firm.Turning to strategy D - “dramatic change” - this represents a larger and

more radical restructuring than M and will require higher managerial effort, soeD > eM . We also assume that πD > πM . However, we suppose that thereare diminishing returns to change, so that the relationship between chance ofsuccess and effort has a concavity property:

πB + [πD − πBeD − eB

](eM − eB) < πM (1)

The assumption of diminishing returns is key to our analysis: as we shall see,the idea that really dramatic change is very costly to implement is importantin deriving our results.We impose a few more assumptions on the parameters, whose purpose will

be made clear in the following sub-sections. We assume that V ≥ eBπB. This

means that shareholders are willing to compensate the CEO for the effort inimplementing strategy B, rather than just letting the firm collapse. Similarly

8

for M and D we impose:

(V − eM − eBπM − πB

)πM − (V − eBπB)πB ≥ 0 (2)

(V − eD − eBπD − πB

)πD − (V − eBπB)πB ≥ 0 (3)

As will become clear below, condition (2) implies that the shareholders will bewilling to pay for the CEO to carry out strategy M if it is presented as analternative to the status quo B. Condition (3) performs the same role for D.We impose a condition to ensure there is a conflict of interest between the

CEO and the shareholders regarding strategy formation.

V (πM − πD)− (eM − eD) + πB

µeD − eBπD−πB

− eM − eBπM−πB

¶> 0 (4)

As will be shown, (4) means that shareholders prefer strategy M over strat-egy D, whereas (1) implies the opposite for the CEO.

2.1 Contracting between the shareholders and the man-ager

There are three main elements to contracting in our model: the manager’s lim-ited liability, the timing, and the assumptions on contractibility. Since themanager has no financial resources, the contract he agrees with the sharehold-ers can stipulate non-negative payments only. Potentially there are two timeswhen contractual arrangements could be made or renegotiated: (1) at the initialstage, when the manager is appointed and before he has had time to formulatea strategy and (2) after he has formulated a strategy but before he has imple-mented it.We assume that strategies are observable, but not verifiable. Hence, a con-

tract can stipulate a payment that is conditional on firm value, but not on thestrategy. We consider contracts that take the form of a positive payment in casethe firm is worth V , and no payment if the firm is worth nothing.10 Two keyimplications of the contractibility assumption are: (i) it is not possible to makepayments conditional on the strategic plans he proposes to the shareholders; (ii)the CEO can agree to implement one strategy, and later decide he prefers toimplement another.We will consider three types of contract, and study conditions for each to

be optimal. The first type is what we call “ex-post contracting.” In this case,the contract is set at stage (2). This allows the contract to be set in responseto the strategy that the manager has formulated. The second type is what wecall “ex-ante contracting”. The contract is set at stage (1) and can then berenegotiated at stage (2) by mutual agreement. The main part of our analysis

10Note that menus of contracts are ruled out. We consider that menus of contracts wouldbe implausible for our application to CEO compensation.

9

considers the trade-off between ex-ante and ex-post contracting, but we thenextend the analysis to consider a third type of contracting in which, as part ofan ex ante contract, shareholders commit to rule out certain types of stage (2)renegotiation. Both in the initial negotiation and in the event of renegotiation,we maximize the principal’s payoff subject to meeting the agent’s reservationutility level11 .

2.2 Setting CEO compensation once strategy is decided:ex-post contracting

To start with suppose that CEO compensation is not fixed at the time he isappointed. Instead, the owners wait until the CEO has presented his strategicplans, then they agree on the choice of strategy and set the compensation pack-age (i.e. ex-post contracting). In the simplest case (occurring with probability1− p), there is no option for change and the manager’s task is simply to imple-ment strategy B. Let mB be the contractual payment in the event the firm isworth V . Incentive compatibility requires:

mBπB − eB ≥ 0 (5)

and the solution is12:mB =

eBπB

(6)

The CEO obtains his reservation utility level (zero) in this case.The second case is when the manager reports that there is an opportunity

for moderate change M . Denote the resulting payment mM (in the event thefirm is worth V ). If the shareholders agree to implement M , then there are twoincentive compatibility conditions. The first requires expected compensation tooutweigh the effort cost:

mMπM − eM ≥ 0 (7)

and the second requires that the manager really has an incentive to choose Mover B. Since the choice of strategy is observable but not contractible, if thepayment is inadequate, he may pretend to implement M , but actually stick toB. The condition is:

mMπM − eM ≥ mMπB − eB (8)

It follows immediately from our assumption that eMeB

> πMπB

that (8) is thebinding constraint, thus the optimal contract “overcompensates” the managerrelative to his reservation wage (0). The solution is13 :

mM =eM − eBπM − πB

(9)

11This is a standard assumption.12 So long as V ≥ mB , as we have assumed. In other words B is a positive NPV project.13 So long as the shareholders are willing to pay that much, i.e. (V −mM )πM ≥ (V −mB)πB ,

as we have assumed (condition (1)). In other words M has a higher (positive) NPV than B,even allowing for the agency costs of implementing it as an alternative to B.

10

We denote his expected payoff from strategy M as UM :

UM = mMπM − eM (10)

= πB

µeM − eBπM − πB

¶− eB (11)

where the second equation follows by setting equality in (8) and substituting(9). We can use this derivation for strategy M to see the characteristics a CEOlikes in a strategy. Perversely the manager prefers strategies with a lower chanceof success and a higher effort level:

Proposition 1 So long as the parameters of the model satisfy condition (2),the CEO’s preferences for strategies are increasing in effort eM and decreasingin the chance of success πM .

Proof. Immediate by inspection of (11).The next case is when the manager reports that there is an opportunity

for dramatic change D as the alternative to B. D has higher effort and higherchance of success than M , but a less favourable effort/success ratio. Theincentive compatibility constraints are similar, the solution14 for the paymentis mD =

eD−eBπD−πB and the CEO’s expected payoff is:

UD = πB

µeD − eBπD − πB

¶− eB (12)

Again the CEO derives positive surplus (UD > 0). Furthermore, if the CEOhas a choice, he prefers to formulate strategy D over M : UD > UM :, sincediminishing returns (assumption (1) above) implies eD−eB

πD−πB > eM−eBπM−πB .

Turning to the shareholders, they prefer both M and D to B, even takinginto account the agency costs of implementing them. Both types of change havehigher NPV’s than B (which is itself positive NPV), by assumptions (2) and(3). Given a straight choice between M and D, they would prefer M so long as:

(V −mM )πM > (V −mD)πD

i.e.,V (πD − πM ) < (eD − eM ) + (UD − UM ) (13)

which follows immediately from assumption (4). This condition has a naturalinterpretation: the left hand side is the extra firm value created by strategy D;the first term on the right is the extra managerial effort cost of implementingD, and the second term represents the additional rents earned by the managerfrom ex-post contracting to implement D.For future reference, their expected payoff (averaging over the different pos-

sible strategic choices facing the manager) when the CEO chooses M is:

(1− p) (V −mB)πB + pqM (V −mM )πM + p (1− qM ) (V −mD)πD (14)

14Again, so long as the shareholders are willling to pay it, as guaranteed by the assumptionof condition (3).

11

On the other hand if the manager were to formulate D instead, their expectedpayoff would be:

(1− p) (V −mB)πB + pqD (V −mD)πD + p (1− qD) (V −mM )πM (15)

Proposition 2 So long as the parameters of the model satisfy conditions (1),(2), (3) and (4), with ex-post contracting the CEO prefers to formulate thedramatic change strategy D, whereas the shareholders would prefer him to in-vestigate the moderate change strategy M .

To summarize the analysis so far: both types of change (M and D) are pos-itive NPV and preferred by the shareholders to the status quo B (even allowingfor the agency cost of implementing them). Of the two, the shareholders pre-fer moderate change, M . Hence ex-post contracting creates a further agencyconflict: the CEO’s strategy choice is distorted because he has an incentive toseek hard strategies with an unfavorable effort/success ratio, which drive up therequired compensation. He will prefer dramatic change D.

2.3 Setting compensation to influence strategy formula-tion: ex-ante contracting

Anticipating this problem with ex-post contracting, shareholders may seek toderive a better alignment between their interests and the CEO’s by designing abetter compensation package at an earlier stage. Clearly, one way of ensuringbetter alignment of interests is just to give the CEO a large enough equity shareat the outset - if he gets 99% of the equity, this difficulty will probably beeliminated. But giving the CEO a large amount of equity, share options, or anequivalently generous bonus is very expensive. There is a trade-off. In thissection, we explore how shareholders can use this kind of device to correct theCEO’s incentives. We assume they are able to set the compensation contractex-ante, before the CEO has started to formulate strategy. Thus the contractconsists of a specified payment in the event the firm is worth V , and it can berenegotiated later when when strategy is agreed, if mutually agreeable.15

Suppose that the initial contract is set at m∗ (when the firm is worth V )and that renegotiation results in final compensation levels m∗B, m

∗M , and m∗D

(in the event the firm is worth V ) in each of the three possible combinations offeasible strategies that the manager may present to the shareholders followingthe strategy formulation stage.16 Clearly m∗B, m

∗M , and m

∗D cannot be less than

m∗; otherwise the manager would not agree to the renegotiation.15We assume here that agents cannot pre-commit not to renegotiate, an assumption that is

discussed and relaxed in section 3 below.16Although the set of possible histories is richer, it is clear that the contract could not

distinguish between events such as: (i) the manager initially had a choice between M and D,but chose to investigate D,versus (ii) the manager initially had no choice, and D was the onlyoption for change. Thus, the most that one could hope for is three final different paymentsm∗B , m

∗M , and m∗D . Even then, we shall see that not all combinations of three positive real

numbers are achievable.

12

Note that ex-post contracting can be viewed as a special case of ex-antecontracting. If a low payment is set initially (at most m∗ = mB) it will simplybe renegotiated upwards to

m∗B = mB

m∗M = mM

m∗D = mD

where mB , mM , and mD are the ex-post contracts as seen in the previoussubsection. Hence, we refer to this case as ex-post contracting, while the term“ex-ante contracting” will be reserved for the case where m∗ > mB.

Lemma 3 Ex-ante contracting results in a floor m∗ such that

m∗B = max{m∗,mB}m∗M = max{m∗,mM}m∗D = max{m∗,mD}

Proof. in appendix.The standard procedure for solving an agency problem requires us to list all

the possible actions of the agent, for each action to compute the cheapest wayfor the principal to induce the action, then to compute the principal’s payoffthat results. Finally the principal must compare his payoff across actions andpick the optimal one. For this model the action space of the agent is complexbecause the CEO needs to decide whether to chooseM or D, if there is a choice,at the strategy formulation stage; whether to choose B or the alternative, ifthere is one, at the strategy choice stage; and whether to put in the requiredeffort for the chosen strategy at the implementation stage. However, we cansimplify matters by eliminating actions where there is ever a possibility thatthe CEO does not put in sufficient effort, as proved in the appendix. Giventhis we can specify the relevant actions by describing the choices of strategyat the formulation stage and at the implementation stage (i.e., ignoring allactions without sufficient effort). The list and full description of the eight suchactions is given in the appendix, where we also show that six of these can beeliminated from consideration here17 because they involve ex-post inefficientstrategy choices. The two relevant actions are what we call:

• Moderate Change Favoured : the manager selects M over D, if there isa choice, at the strategy formulation stage, and this is then chosen forimplementation (over B). If M is the only alternative at the strategyformulation stage, it is chosen for implementation over B, likewise for D.

• Dramatic Change Favoured : the manager selects D over M , if there isa choice, at the strategy formulation stage, and this is then chosen forimplementation (over B). If M is the only alternative at the strategyformulation stage, it is chosen for implementation over B, likewise for D.

17 i.e, with renegotiation.

13

Proposition 4 The cheapest way to implement Dramatic Change Favoured isex-post contracting.

Proof. Ex-post contracting gives this action in return for contractual payments(in the event of firm value V ) m∗B = mB , m∗M = mM , m∗D = mD. If any ofthese payments were reduced, the manager would not be willing to work toimplement the specified strategy B, M , or D, therefore this is the cheapest wayto induce the action.

Proposition 5 The cheapest way to induce Moderate Change Favoured is ex-ante contracting with m∗ ∈ (mM ,mD) given by

m∗ =πD

³eD−eBπD−πB

´− (eD − eM )

πM(16)

Proof. To induce this action, we obviously cannot drop m∗B below mB , m∗Mbelow mM , or m∗D below mD, otherwise the CEO will not be willing to workto implement the chosen strategy. We have to increase one or more of thecontractual payments above the ex-post level by setting m∗ > mB, to see ifthis will induce the agent to pick M over D and if so, find the smallest m∗

that does this. The question is whether it is optimal to have m∗ ∈ (mB ,mM ],m∗ ∈ (mM ,mD], or m∗ > mD.In the event the CEO has a choice between M and D at the strategy for-

mulation stage, his expected payoff if he picks M is (πMm∗M − eM ) and if hepicks D it is (πDm∗D − eD). Since ex-post contracting induces choice of D, itfollows that m∗ ∈ (mB,mM ] will not induce choice of M because increasing m∗

from mB part-way towards mM does not affect either of these payoffs. Nextconsider m∗ > mM . If m∗ ∈ (mM ,mD], the CEO will just be willing to pickM over D if

(πMm∗M − eM ) = (πDm∗D − eD)

By hypothesis m∗M = m∗ and m∗D = mD, so substituting for mD =eD−eBπD−πB

we have

m∗ =πD

³eD−eBπD−πB

´− (eD − eM )

πMOne can verify that m∗ defined by this formula does not exceed mD, as aconsequence of our assumption that eD

eM> πD

πM(or equivalently mM < mD).

We can now complete the solution of the problem by examining the share-holders’ preferences. The shareholders will prefer to induce choice of ModerateChange Favoured using ex-ante contracting with m∗ as just derived, insteadof accepting using ex-post contracting to induce choice of Dramatic ChangeFavoured, if:

(1− p) (V −m∗)πB + p (1− qD) (V −m∗)πM +

+pqD (V −mD)πD ≥ (1− p) (V −mB)πB + pqM (V −mM )πM +

+p (1− qM ) (V −mD)πD (17)

or, equivalently:

14

Proposition 6 The shareholders will use ex-ante contracting with a guaranteedminimumm∗ (paid in the event of success) to induce the CEO to prefer ModerateChange Favoured if

p (1− qM − qD) (V −m∗)πM ≥ p (1− qM − qD) (V −mD)πD +

(1− p) (m∗ −mB)πB +

pqM (m∗ −mM )πM (18)

This ex-ante contracted payment m∗ will be renegotiated upwards to mD incase the CEO has no choice over available strategies and presents D to theshareholders.If the inequality is reversed, they will prefer the ex-post contract implementing

Dramatic Change Favoured.

This expression has a simple intuitive explanation. The term on the left-handside represents the shareholders’ payoff when the CEO chooses to formulate Mover D, which is the benefit of a regime of ex-ante contracting. For this to bedesirable it must be greater than the payoff when the CEO makes the oppositechoice (the first term on the right hand side). But it must also compensatethe shareholders for two drawbacks associated with ex-ante contracting: (i) theshareholders will have to overpay the CEO (relative to the ex-post levels) forimplementing strategy B when there are no alternatives to it (the second termon the RHS); (ii) and they will also overpay the CEO (relative to the ex-postlevels) for implementing strategyM , even when dramatic change D is not a realthreat (third term on the RHS).So, when is ex-ante contracting desirable? Clearly, if change is likely (p is

big) then ex-ante contracting is helpful to the shareholders, because one of itsdrawbacks, the overpayment for strategy B, is less relevant. Also, if moderatechange, M , is not particularly likely (low qM ) ex-ante contracting is also morefavourable because the overpayment for strategyM (the third term on the RHS)is less important. One can interpret both of these cases (bigger p and low qM )as corresponding to a more unstable, changeable business environment.The conclusion is that in a highly changeable environment, shareholders

might wish to pre-commit to give the CEO an apparently overgenerous packageof incentives at the outset. Although sometimes overgenerous relative to theimmediate task at hand, it would improve his incentives for strategic decisionmaking.

3 Extension: Setting a ceiling on compensationIn this section we consider the case where the shareholders may be able to makepre-commitments about the ex-post renegotiation of the CEO’s compensationcontract, specifically when they set an upper limit to compensation. The benefitof such a compensation scheme is the possibility of preventing the CEO fromchoosing to formulate his preferred strategy D over the shareholders’ preferred

15

strategy M . If he knows he will never be paid enough to implement an over-ambitious strategy, the CEO will lose interest in such a plan. Establishing exante a credible ceiling on compensation will also have a drawback: even whendramatic change is the only possible alternative to the status quo it will bedisregarded.The issue of pre-commitment not to renegotiate is open to debate. Clearly,

companies and CEO’s often do renegotiate compensation. The typical arrange-ment consists of an annual salary combined with a stock option grant, so eachyear the salary is renegotiated and the CEO also adds more options to an exist-ing portfolio. Hence, pre-commitment may seem unrealistic. From a theoreticalpoint of view also, it is hard to see how commitment could be enforced - per-haps by a third party, but then, the design of the contract with this third partywould have to be complex and might violate the spirit of our prior assumptionson contractibility between CEO and shareholders, and it would certainly beunrealistic.However, perhaps social norms or individual ethics could be used as a pre-

commitment device. For example, in Sweden it would probably be possible fora company to pre-commit never to pay the CEO over US $1bn (as was paid toRoberto Goizueta of Coca-Cola over a ten-year period). Even within the US, 25years ago such a payment might have been impossible. Thus pre-commitmentmay be possible at the social level. At the level of the individual company,it may be possible to pre-commit by appointing individuals who are known tobe strongly opposed to high compensation to the board or to the remunerationcommittee. Having large block shareholders with this view might also havethe same effect. For that reason, we believe it is worthwhile to explore the casewhere pre-commitment is possible. We therefore add the following assumptions:We consider the possibility of pre-committing not to engage in certain types

of renegotiation. The only pre-commitment of that type we consider reasonableis setting an upper bound on the level of compensation.18

Clearly this will not help implementation of the two actions previouslyanalysed ( Moderate Change Favoured and Dramatic Change Favoured) be-cause they require a high payment for strategy D. The interesting case nowis the possibility to implement another action which we call Moderate ChangeOnly19. This action is characterized by choosingM whenever it is available andignoring D even if it is the only alternative to B. The shareholders can correctthe manager’s preference for D by committing themselves never to pay enoughcompensation to make the CEO willing to implement that strategy. It thereforerequires m∗B ≥ mB, and m∗M ≥ mM but not m∗D ≥ mD. So the cheapest wayto implement it is to (credibly) pre-commit never to pay more than mM . The

18The last assumption rules out pre-commitment of the sort where the board commit neverto offer a payment in a set A, and the set A can have an arbitrary shape such as a subset ofthe real line that is not connected, the set of rational numbers, etc... Recall that strategiesare not contractible, hence the contract cannot be made contingent on the CEO’s announcedstrategy.19Denoted by action A7 in the appendix. In the appendix we also show that the remaining

five actions, A2, A3, A4, A6 and A8 are not optimal.

16

expected payoff for the shareholders is then:

(1− p)πB (V −mB) + pqDπB (V −mB) + p (1− qD)πM (V −mM ) (19)

We can compare this expected payoff to what shareholders would obtain incase the contracting process involved ex-ante contracting without pre-commitment,and also in case we had an ex-post contracting policy. From these comparisonswe reach the following results.

Proposition 7 The shareholders will prefer ex-ante contracting with a ceilingon compensation (inducing the CEO to choose Moderate Change Only) to ex-ante contracting with unrestricted renegotiation if:

πB (1− p) (m∗ −mB) + p (1− qD)πM (m∗ −mM ) ≥ pqD[πD (V −mD)−

−πB (V −mB)] (20)

with m∗ as defined in proposition 5.

Proof. Direct from comparison of the shareholders’ expected payoff with ex-ante contracting with a ceiling (equation (19)) and without an upper bound(LHS of inequality (17)).

Proposition 8 The shareholders would prefer the ex-ante contract with a ceil-ing (inducing the CEO to choose Moderate Change Only) to ex-post contractingif:

pqD [(V −mB)πB − (V −mD)πD] ≥ p (1− qM − qD) [(V −mM )πM −− (V −mD)πD] (21)

Proof. Direct from comparison of the shareholders’ expected payoff with ex-ante contracting with a ceiling (equation (19)) to their expected payoff withex-post contracting (equation (14)).If both conditions (20) and (21) are satisfied, the shareholders would prefer

to completely rule out the possibility of the manager formulating strategy D.We can see that setting an upper bound would be optimal if it is unlikely thatthe strategy for dramatic change, D, is the only one available (low probabilityqD) and if the probability of change (p) is also low. This is the scenario in whichthe drawback of setting an upper bound on compensation - the CEO will notdevelop strategy D (which is, by assumption, superior to the status quo) whenthis is the only option available to him - is less relevant.We can interpret this case (low qD and small p) as a very stable environment:

a low chance of any change being required, especially dramatic change. Inthis case the ceiling on compensation is a cheap way to correct the manager’spreference for dramatic change. If this is not the case, then it will be preferableto set either an ex-ante contract without a ceiling, or an ex-post contract.

17

4 ConclusionOur results are simple to summarize. In this paper we analyzed how managerialrent-seeking may distort strategy choice in favour of overambitious change. Weshowed that, in our model:

• In a highly changeable environment (change is more likely to be required)the shareholders should commit to a policy of high pay, offering the CEOa high-powered incentive package at the outset in order to improve hisincentives for strategy making, even if sometimes this results in excessive(with hindsight) payment at the implementation stage. This will correcthis incentives away from excessively drastic restructuring and towards thekind of change that is preferred by shareholders.

• Another way of curbing CEO’s tendency towards dramatic change is topre-commit never to pay the very high compensation package required toimplement dramatic change. It may be difficult to make this precommit-ment, but if it is possible, then it will be preferred if dramatic change isnot likely relative to moderate change, or if change is unlikely overall.

• Finally, in some situations it may be optimal for the shareholders sim-ply to accept some strategic distortion and to negotiate compensation tothe required level after the strategy has been selected. This will occurwhen change is not very likely or when shareholders do not have strongpreferences among the possible change options.

• Our analysis can help understand why CEO’s typically receive much higherpay and stronger incentives than other senior managers. Their pay reflectsstrategic discretion more than direct compensation for effort. Our resultsare also consistent with the empirical evidence of higher incentives in morechangeable environments.

• One can also hypothesise that the economic pressures in recent years forfirms to transform and reinvent themselves have led to “ex-ante contract-ing”, i.e., a precommitment to high pay and incentives. It is an openquestion whether mechanisms (perhaps at the social level) to precommitto upper bounds on pay are either possible or optimal.

An alternative perspective on CEO pay is that high pay is not optimal atall for shareholders and simply results from rent extraction in collusion with theboard. While we do not deny this possibility, we consider it is worthwhile ex-ploring how high pay can arise in a model with only relatively minor departuresfrom standard agency theory.

References[1] Asquith, P., R. Bruner and D. W. Mullins, 1983, “The Gains to Bidding

Firms from Merger,” Journal of Financial Economics, 11, 121-139

18

[2] Aghion, P. and P. Bolton, 1991, An incomplete contracts approach to fi-nancial contracting, Review of Economic Studies 59, 473-94

[3] Baumol, W. J., 1959, Business Behavior, Value and Growth, New York:Macmillan

[4] Berle, A. and G. Means, 1932, The modern corporation and private prop-erty, New York: Macmillan

[5] Blanchard, O. J., F. Lopez-de-Silanes and A. Shleifer, 1994, “What do firmsdo with cash windfalls?,” Journal of Financial Economics, 36, 337-360.

[6] Bradley, M., Desai, Kim, 1988, “Synergistic gains from corporate acquisi-tions and their division between the stockholders of target and acquiringfirms,” Journal of Financial Economics 21, 3-40

[7] Bruner, R., 2002, “Does M&A Pay? A Review of the Evidence for theDecision-Maker,” Journal of Applied Finance (Spring/Summer 2002) 12,48-68

[8] Conyon, M. and K. Murphy, 2000, “The Prince and the Pauper? CEO Payin the US and UK,” Economic Journal, 110, 640-71.

[9] Core, J. and W. Guay, 2001a, “Executive Equity Compensation and Incen-tives: A Survey”, working paper, Wharton School

[10] Core, J. and W. Guay, 2001b, “Stock option plans for non-executive em-ployees”, Journal of Financial Economics, 61, 253-287

[11] Core, J and W. Guay, 2003, “The other side of the trade-off: the impactof risk on executive compensation: a comment,” forthcoming, Journal ofPolitical Economy

[12] Core, J. and D. Larcker, 2003, “Performance consequences of mandatoryincreases in CEO stock ownership”, forthcoming, Journal of Financial Eco-nomics

[13] Demsetz, H. and K. Lehn, 1985, “The structure of corporate ownership:causes and consequences”, Journal of Political Economy, 93, 1155-1177

[14] Hart, O. S. D., 1995, Firms, contracts and financial structure, Oxford

[15] Hart, O. S. D. and J. H. Moore, 1990, “Property rights and the nature ofthe firm,” Journal of Political Economy, 98, 1119-1158

[16] Hall and Liebman, 1998, “Are CEOs really paid like bureaucrats?”, Quar-terly Journal of Economics, 103, 653-691

[17] Holmstrom, B. and P. Milgrom, 1991, “Multi-task principal agent analy-ses: incentive contracts, asset ownership and job design,” Journal of Law,Economics and Organization, 7, 24-51

19

[18] Jarrell, G., J. Brickley and J. Netter, 1988, “The market for corporate con-trol: the empirical evidence since 1980,” Journal of Economic Perspectives2, 1, 49-68

[19] Jarrell, G. and A. Poulsen (1989) “Returns to acquiring firms in tenderoffers: evidence from three decades,” Financial Management 18 (1989) 12-19

[20] Jensen, M. C., 1986, “Agency costs of free cash flow, corporate finance, andtakeovers,” American Economic Review, 76, 323-329

[21] Jensen, M. C. and K. J. Murphy, 1990, “Performance pay and top-management incentives,” Journal of Political Economy, 98, 225-264

[22] Jensen, M. C. and R. Ruback, 1983, “The market for corporate control:the scientific evidence,” Journal of Financial Economics, 11, 5-50

[23] Jensen, M. C., 2000, Theory of the Firm: Governance, Residual Claims,and Organizational Forms, (Harvard University Press, December 2000)

[24] Jensen, M. C. (1989) ”Eclipse of the Public Corporation,” Harvard BusinessReview (September-October), updated version, 1997, available on SSRN

[25] Kaplan, S. and L. Zingales, 1997, “Do Financing Constraints Explain WhyInvestment is Correlated with Cash Flow?” Quarterly Journal of Eco-nomics, 112, 169-215

[26] Lamont, O., 1997, “Cash flow and investment: evidence from internal cap-ital markets,” Journal of Finance, 51, 1, 83-109

[27] Lang, L., and R. Litzenberger, 1989, “Dividend Announcements: CashFlow Signalling vs. Free Cash Flow Hypothesis?,” Journal of FinancialEconomics 24, 181-192

[28] Lang, L., R. Stulz and R. Walking (1991), “A Test of the Free Cash FlowHypothesis,” Journal of Financial Economics 29, 315-335

[29] Mann, S., and N. Sicherman, 1991, “The Agency Costs of Free Cash Flow:Acquisition Activity, and Equity Issues,” Journal of Business 64, 213-227.

[30] Marris, R.,1967, The economic theory of managerial capitalism, Macmillan

[31] Mintzberg, H., 2000, “View from the top: Henry Mintzberg on strategyand management,” The Academy of Management Executive, 14, 31-39

[32] Murphy, K. J., 1999, “Executive compensation,” in O. A. Ashenfelter andD. Card (eds.), Handbook of Labor Economics, Vol. 3, North Holland

[33] Prendergast, C., 2002, “The tenuous trade-off between risk and incentives,”Journal of Political Economy, 110, 5, 1071-1102

20

[34] Salanié, B., 1997, The economics of contracts, MIT Press

[35] Shleifer, A., and R. W. Vishny, 1989, “Management entrenchment: the caseof manager-specific investments,” Journal of Financial Economics 25, 123-140

[36] Shleifer, A., and R. W. Vishny, 1997, “A survey of corporate governance,”Journal of Finance, 52, 2, 737-783

[37] Tirole, J., 2001, Lecture notes on corporate finance, unpublished manu-script, University of Toulouse

[38] Weston, J. F., J. A. Siu and B. A. Johnson, 2001,Takeovers, restructuringand corporate governance, 3rd. edition, Prentice-Hall

[39] Williamson (1964), The economics of discretionary behavior: managerialobjectives in a theory of the firm, Prentice-Hall

5 AppendixRemarks on incomplete contractibility The approach we take here ap-plies what is now a standard paradigm, following many papers in the past decadethat have studied financial contracting where some variables are observable byboth parties, but not contractible. Aghion and Bolton (1991), Hart (1995), andHart and Moore (1998) are leading examples. In general, non-contractibilitycan arise because it would be too expensive or too complex to make contractsfully conditional, or because it would be difficult for a third party to verify fulfil-ment of the conditional clauses in the contract (since contracts depend on thirdparties such as courts or private arbitrators for enforcement). For example, Hart(1995) and Hart and Moore (1998) assume that investment is not contractible.They argue that even if third parties were able to verify monetary expenditureson investment, they would be unable to tell if the funds were applied properlyto the right kinds of projects. This is similar to our assumption that strategyis non-contractible. Only “hard” variables are contractible upon.To motivate our analysis, consider the following example. Suppose that a

professor is given responsibility for a degree programme. He or she can chooseto run the programme on a business-as-usual basis, i.e. attending the usualcommittee meetings, checking the lecturing performance of his or her colleagues,monitoring the performance of the programme office in regard to admissionsand records, and taking an appropriate interest in student welfare. This isa moderately time consuming, but manageable task and the professor wouldbe entitled to expect some measure of compensation in the form of a reducedteaching load or moderate salary supplement.Now suppose the professor is tasked with restructuring the course, with a

freshly-thought out structure and syllabus that reflects current student demandand is fully up to date. The first observation we make is that engineering sucha change is extremely demanding in terms of time and energy. Many proposed

21

changes, however trivial, will meet with resistance from entrenched lecturers.Even colleagues who are enthusiastic about change will respond by making awide variety of counter-proposals that makes coordination on an improved out-come difficult. So change from the status quo can be disproportionately costlyand the professor, unless he obtains intrinsic satisfaction from the task, will notbe keen to take on this job without substantial additional compensation (gen-erally in universities, such extra compensation is unlikely!). Of course, firmsmay not be quite as resistant to change as some universities but we argue thatthe same broad feature applies.The second observation we make is that it may be easy to spot whether the

changes made by a colleague are profound or superficial, but it may be hard toprove that assessment in a way that is credible to an outsider. A professor mayfeel that the course director’s “relaunched” programme is just a minor variantof the old one, but, if he were challenged on this view, it might be hard forhim to convince a colleague in another department, a student, or the coursedirector’s lawyer. So it would be impossible to condition compensation on thesuccessful implementation of a truly improved degree programme. We arguethat in businesses generally, similar reasons are likely to make it impossible tocondition pay directly on strategy. However in businesses, unlike universities,pay can be conditioned on share price.To summarise the two key points of the example: first, change can require

disproportionately high effort to implement, relative to the status quo. Second,writing a contract conditional on “satisfactory change” is likely to be impossible.In our model we have in mind a variety of corporate strategies, of which

a possible simple example would be the introduction of a comprehensive cost-cutting programme. At first sight, it might seem that costs can be verified easilyfrom the company accounts. However, this would be a crude way of monitoringthe successful implementation of a genuinely shareholder-value increasing ratio-nalisation plan. It could likely be quite easy for the manager to slash costswith a poorly executed programme of cutbacks that would actually damageshareholder value in the longer term (through employee morale, quality controlproblems, disruption of supplier relationships, etc). Just because a strategy hasimplications for an easily quantifiable variable, it does not mean that controllingthat variable will actually verify implementation of the strategy.Holmström and Milgrom (1991) give a good illustration of this kind of prob-

lem. A teacher in South Carolina was found to have boosted her class examperformance, and hence her own performance rating, by passing the childrenanswers to the statewide geography test. Another egregious example is givenin the Financial Times of 4 February 1994: the recently appointed chairman ofAudi discovered that the previous year’s sales figures had been “bolstered by anold trick. Audi France officials confirmed yesterday that ‘several tens of thou-sands’ of cars had been parked with French distributors [and hence recordedas sales], only to be shipped back to Germany last year. Since many of themlacked airbags and ABS braking systems - regarded as essentials in the Ger-many quality car market - selling them was no easy task.” This shows whyeven apparently “hard” data such as company accounts may not be of much

22

help in writing conditional contracts.

Lemmas and proofs Proof of Lemma 3. We show thatm∗B = max{m∗,mB};a similar argument holds for the other two cases BM and BD. Let the ex-antecontract specify a payment m∗ in the event the firm succeeds and is worth V .First note that, if renegotiation occurs, it will occur upwards only, otherwisethe CEO will reject the proposal. Hence m∗B ≥ m∗. Next, note that therenegotiated payment cannot exceed the ex-post contracting level, otherwisethe shareholders could offer less and get the same behaviour from the CEO.Suppose first that m∗ ≥ mB, then it is clear we cannot have renegotiation

and m∗B = m∗.Next suppose that m∗ < mB, then it is clear (from the derivation of mB)

that renegotiation is mutually beneficial and will lead to a contractual paymentm∗B = mB.

Lemma 9 Under an optimal contract, the CEO always puts in the requiredeffort for the chosen strategy.

Proof. Consider first the case of ex-post contracting and the possibility that incase B is the only available option, the CEO does not put in the required efforteB (implying m∗B < mB). By our assumption that V > mB , this is suboptimalbecause the shareholders would prefer to pay mB and induce effort. The casesBM and BD are similar given our assumptions that the parameters satisfyconditions (2) and (3). In the case of ex-ante contracting the same reasoningapplies.In the case of a ceiling on compensation we know that the ceiling would

never be below mB, by our assumption that V > mB. Therefore if the managermoves away from implementing B with effort, it must be in order to implementanother strategy with effort.

List of Actions: We can list the actions as follows:A1: (D,MBM ,DBD) Dramatic Change FavoredA2: (D,BBM ,DBD)A3: (D,MBM , BBD)A4: (D,BBM , BBD)A5: (M,MBM ,DBD) Moderate Change FavoredA6: (M,BBM ,DBD)A7: (M,MBM , BBD) Moderate Change OnlyA8: (M,BBM , BBD).To explain the notation, take the first action on the list, (D,MBM ,DBD),

which represents the outcome of ex-post contracting. The first symbol, D,means that when the manager has a choice between M and D at the strategyformulation stage (with probability p (1− qM − qD)), he picks D. The secondsymbol,MBM , means that when the CEO presents strategyM to the sharehold-ers as an alternative to B, they pickM (to avoid lengthy circumlocutions, we do

23

not distinguish between the shareholders picking the strategy at the implemen-tation stage and the CEO picking the strategy. Obviously, since the managercan always cheat if wants to, the shareholders knows which strategy he is goingto pick so we may as well assume they both pick the same one.) Note that evenif he picks D in preference to M at the strategy formulation stage when he hasthe choice, there will sometimes (probability qM ) be occasions when M is theonly option. The third symbol, DBD, means that when the manager presentsoption D to the shareholders as an alternative to B, they pick D.The three actions that are referred to in the text are A1 (Dramatic Change

Favored), A5 (Moderate Change Favored), and A7 (Moderate Change Only).For the purposes of the analysis in Section 2 (i.e. the main analysis on

ex-post contracting and ex-ante contracting) we can rule out actions A2 to A4and A6 to A8 as suboptimal, because renegotiation would always take placeto implement the change strategy (M or D) rather than the status quo B,by our assumptions on the exogenous parameters satisfying conditions (2) and(3). The argument is similar to the previous proposition. That leaves actionA5 (Moderate Change Favored, which corresponds to the manager choosing toinvestigate M rather than the more dramatic D) in addition to A1 (DramaticChange Favored).For the extension to the analysis of pre-commitment in section 3, we cannot

exclude actions simply because they are ex-post suboptimal - after all the pur-pose of precommitment is precisely to allow ex-post inefficient actions in someoutcomes, so as to improve ex-ante incentives. However, one can readily seethat A2 and A6 are inferior to A1, A3 is inferior to A7, while A4 and A8 areinferior to all three actions A1, A5 and A7. That leaves A1, A5, and A7 asanalyzed in the text.

24