Academic tenure, posttenure effort, and contractual damages.

Dnes, Antony ; Garoupa, Nuno

I.INTRODUCTION

In this article we explore the incentive properties of academic tenure relative to alternatives that might be substitutes for tenure. We do not present a comprehensive theory of academic tenure, but rather emphasize the role of tenure in providing an incentive for incumbent faculty to reveal the characteristics of recruits and to maintain their own posttenure performance. Academic tenure often implies that its holder cannot be removed from post except for good cause, usually based on gross moral turpitude or gross incompetence (McPherson and Schapiro 1999). Such removal is historically characterized by a costly procedure governed by organizational or state statutes (Tepker 1993). Our results suggest that tenure is just one of many mechanisms able to achieve honest revelation and that it has important interactions with other mechanisms.

There is no question that tenure is a constraint to administrators and limits the discretion of university managers. The objective of our article is not to provide a new explanation for academic tenure but to show how tenure could in principle be replaced by contractual damages and how it more generally interacts with damages.

We believe it is worth theoretically exploring alternatives to tenure because universities appear to ignore tenure statutes on occasions. In particular, it is now common for schools to ignore orders to reinstate dismissed, apparently tenured faculty and to choose to pay damages instead (Tepker 1993). Carmichael's (1988) model suggests that damages can achieve efficient information revelation where the value of academics' next best occupations are known. A damages payment equal to the difference between current expected earnings and next-best earnings, that is, restoring the victim's expectations (ignoring nonpecuniary benefits for simplicity) will then make an incumbent indifferent between remaining and being sacked and happy to reveal the skills of entrants.

Carmichael rules out the payment of "expectation" damages and concentrates on the use of tenure, which creates a requirement for specific performance by the university. We contend that the frequent award of expectation damages by courts faced with cases of refusal to reinstate suggests that more attention should be given to the damages option, (1) and indeed to possible interactions between tenure and its alternatives. Our contention is reinforced by the observation that many universities are currently softening tenure by introducing devices like five yearly reviews for tenured faculty. Note also that the role of tenure looms large once it has been awarded, directing our attention toward a model of posttenure performance, particularly if it could be lost possibly subject to the payment of contract damages.

Our model explores the role of contractual damages in replacing academic tenure (i.e., reinstatement) as well as addressing the problem of revelation. We depart from previous literature in considering that it is now common practice for university departments to use rankings and publication records to evaluate faculty, thus reducing the asymmetry of information (with respect to research and productivity) detected by Carmichael (1988, p. 460). We focus on posttenure performance in the shadow of the threat of dismissal followed by payment of contract damages.

Dnes and Seaton (1998, 2001) examine the reform of academic tenure in the United Kingdom by the 1988 Education Reform Act, which allowed softer tenure to be chosen by universities. (2) The authors tested the hypothesis that softening tenure encourages incumbent academics to consolidate their hold on academic life. It seems that after 1988 the distinction between hard-tenure and soft-tenure universities persisted, that is, many British universities chose to continue operating a hard-tenure system. Allowing softer tenure has not hindered the improvement of academic performance in the context of increased pressures from a research quality exercise that is periodically carried out by the Higher Education Funding Council. Li and Ou-Yang (2003) also confirm that academic tenure does not seem to have a substantial effect on research productivity (measured by publications) and impact (measured by citations) when comparing performance before and after tenure promotion.

Our model also allows risk-neutral academics to be fully compensated for risk bearing by a salary and expectations-based contractual damages for dismissal. In such a world the academic is indifferent between holding a job and being dismissed, and tenure is important if salaries cannot compensate for risks and contractual damages are too low. This aspect of our work draws attention to the interaction between tenure, compensation, and salaries. It is often asked why we do not simply pay academics more to bear the risk of dismissal if their effectiveness falls off later. Our model sheds some light on the role of salaries vis-a-vis tenure and related matters.

II. A MODEL OF ACADEMIC TENURE

The department has I incumbent professors. Each one of them is ranked according to a research and teaching productivity index: a [member of] [??].

The index is common knowledge and can be regarded as defined by a tenure committee of senior faculty within the department. The department has to decide on giving tenure to a candidate with ability a. The entrant's ability is judged by the tenure committee and becomes common knowledge. Under the assumption of risk neutrality, the model would still follow through if a were an expected value of a distribution.

The construction of an index of research and teaching solves some of the problems discussed by McPherson and Winston (1983). In particular, they argue that an academic job has a high monitoring cost and a high degree of specificity. By considering an index, we address these observations in the sense that it becomes quite easy for a department to assess individuals. It is clear, however, that it excludes variables that are not strictly research or teaching. (3)

Each tenured faculty member j = 1 ...., I has a research and teaching productivity index [a.sub.1] when the tenure committee decides whether to grant tenure. This initial index is exogenous in our model and should have been set by the department to guarantee efficient pretenure effort and productivity. (4) After tenure is approved, each faculty member exerts effort e to provide further research output. The index [a.sub.2] is given by [a.sub.1] + e - [epsilon], where [epsilon] is a random variable that measures a mismatch between the professor's effort and research output after tenure distributed according to g([epsilon]) with support in [??], and zero expected value. In our model each faculty member j = 1, ..., I is characterized by two values for a: one (exogenous) before the tenure review ([a.sub.1]), and another one (endogenous) after the tenure review ([a.sub.2]). Ours is a therefore model of posttenure and not pretenure effort and productivity. (5) Whereas the initial index is used for the tenure decision and ultimately guarantees efficient pretenure effort, the second might be used in a posttenured review (in soft-tenure universities), or in reviewing salaries and promotion, aiming at efficient posttenure effort.

Our model accepts both explicit and implicit incentives for exerting research effort after tenure is granted. Explicit incentives include an increase in salary w. Following a first-order approach to optimal contracting, we opt for the use of the most common solution to participation contracts with w = [alpha] + [beta][a.sub.2], where [alpha] [member of] [??] and [beta] [member of] [0, 1], both determined by the department as employer.

The implicit incentive is based on an external constraint: The monetary value of academic reputation (thus consultancy revenues or research grants) is increasing on [a.sub.2] such that an academic has an extra incentive to research and teach with quality even after a tenure contract has been signed. (6) This possibility is introduced in our model by assuming that each faculty member has an extra benefit given by R([a.sub.2]), with R' > 0 and R" < 0. (7)

There are three stages in our model:

Stage 1: The tenure committee decides whether to give tenure to a risk-neutral untenured professor j, based on the observable index [a.sub.1]. A salary w is also negotiated.

Stage 2: The new tenured risk-neutral professor j decides his or her research effort e, not observable by the department.

Stage 3: The department decides whether to dismiss the tenured risk-neutral professor j, based on the observable index [a.sub.2] = [a.sub.1] + e - [epsilon]. If not dismissed, a salary w negotiated in stage 1 is paid.

We must solve the game backward for sub-game perfection. Let us start with the last stage. The payoff for the tenured professor, if not dismissed, is the salary w plus the extra benefit R([a.sub.2]). Regarding the payoff of the department, much discussion in the literature has focused on the kind of objective function a university department has or should have (see McPherson and Winston 1983; McCormick and Meiners 1988; McKenzie 1996; Brown 1997). We define the department's payoff as [a.sub.2] - w, where research and teaching output is mapped into gains for the department.

Now consider the possibility of dismissing tenured faculty. There will be litigation between the two parties. (8) Let us define the payoffs of a tenured professor and of the department if the case goes to court as [S.sub.j] and [S.sub.d] respectively:

[S.sub.j] = qw + (1 - q)(v + d) + R([a.sub.2])

[S.sub.d] = q([a.sub.2] - w) - (1 - q)d,

where d stands for damages awarded by the court, v is the income from working elsewhere, and q measures the probability of a tenured professor being reinstated by the court (specific performance). A model that takes into account litigation costs is presented in the Appendix.

The probability q measures the importance of tenure (in terms of reinstatement). In hard-tenure universities, we expect q to be high and close to 1 (specific performance dominates contractual damages). (9) In soft-tenure universities, we expect q to be low and close to 0 (contractual damages dominate specific performance). Thus, increasing q can be interpreted as hardening tenure.

A tenured professor is indifferent between hard tenure and contractual damages if d = w - v. This is the particular case of expectation damages (that is, damages are awarded so that the professor will be in the same position as if the tenure contract had been honored). When d < w - v, the professor is undercompensated (and is punished if dismissed) and would prefer the tenure contract to be honored (specific performance). If d > w - v, the professor is overcompensated and prefers contractual damages to specific performance (the department is punished by dismissal).

Let us define p = d + v - w, where p [member of] [??]. The parameter P measures how much actual damages deviate from expectation damages. When [rho] = 0, we have expectation damages. However, if [rho] < 0, damages undercompensate the dismissed professor. On the other hand, if [rho] > 0, damages overcompensate the academic (we could say that the department has to pay punitive damages for dismissal). Thus, we can rewrite:

[S.sub.j] = w + R([a.sub.2]) + (1 - q)[rho]

[S.sub.d] = q[a.sub.2] - (1 - q)([rho] - v) - w.

The department decides to dismiss the professor if and only if [S.sub.d] [greater than or equal to] [a.sub.2] - w. Re-arranging we can write:

[epsilon] [greater than or equal to] [a.sub.1] + e + [rho] - v = [[epsilon].sup.*].

Thus, if the random variable is higher than [[epsilon].sup.*], a dismissal will follow. However, if [epsilon] < [[epsilon].sup.*], the department will decide not to dismiss a tenured professor. A dismissal is more likely if contractual damages are low.

Given what happens in the last stage of the game, we should solve the second stage, that is, incentive compatibility constraint of this exercise. There is a probability G([[epsilon].sup.*]) that no dismissal will take place and a probability 1 - G([[epsilon].sup.*]) that a dismissal will take place, where G(.) is the cumulative distribution.

A tenured professor exerts effort so as to maximize his expected payoff:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

where C(e) is the effort cost, with C' > 0 and C" > 0, and EIR([a.sub.2])] is the expected value of the extra benefit from consultancy and research grants (given that at this stage [a.sub.2] is a random variable because effort is decided before the random variable [epsilon] being revealed).

The first-order condition is given by:

[differential][U.sub.j]/[differential]e = [beta] + [differential]E [R([a.sub.2])]/[differential]e - g([[epsilon].sup.*])(1 - q)[rho] - C' = 0

and the second-order condition is supposed to be satisfied under the usual concavity assumptions. Consequently the optimal effort is given by [[epsilon].sup.*](q, [rho], v, [a.sub.1], [beta]). It balances the marginal benefit from exerting more research effort (marginal increase in the salary, marginal expected revenues from consultancy and research grants, marginal reduction in the expected loss from being dismissed) against the marginal disutility of effort. We should expect tenured faculty members to exert significant research and teaching effort as long as (a) salary varies substantially with research output (when [beta] is near 1), (b) marginal expected gains from consultancy revenues and research grants are important, or (c) tenure is soft (q is low) and damages undercompensate the dismissed faculty member ([rho] is significantly negative). The optimal research and teaching effort after tenure is likely to be 0 if (a) salary does not vary too much with research output (if [beta] is negligible), (b) marginal expected gains from consultancy revenues and research grants are unimportant, and (c) tenure is hard (q approaches 1) or damages do not undercompensate the faculty member (p is not too negative). (10) Notice that in the particular case of expectation damages, the optimal effort is independent of the nature (soft or hard) of tenure. In other words, under a rule of expectation damages, tenure is irrelevant to determine research and teaching effort.

Having solved the second stage, we should look at the first stage. First, let us solve the participation constraint. A professor who does not accept the tenure offer receives y, which is the utility given by the best alternative at that moment (likely to be higher than the best alternative if dismissed afterward, v). Thus, the reservation wage for a professor is given by:

[alpha] = y - (1 - G[[[epsilon].sup.*]])(1 - q)p + C([[epsilon].sup.*]) - E[R([a.sub.2])] - [beta]([a.sub.1] + [[epsilon].sup.*]).

The department's expected payoff is given by:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

The department decides to give tenure as long as [U.sub.d] is greater than the opportunity cost, which is normalized to 0 after the appropriate choice of salary. Let us first notice that the optimal research effort for the department is

[differential][U.sub.d]/[differential]e = 1 - (1 - q)(1 - G[[[epsilon].sup.*]]) + [differential]E[R([a.sub.2])]/[differential]e - g([[epsilon].sup.*])(1 - q)[rho] - C' = O,

where the second-order condition is assumed to be satisfied under the usual concavity assumptions.

The problem for the department is, of course, that the tenured professor ignores the gain for the department from researching. Therefore, the department should set [beta] in such a way that the choice of research and teaching effort will be optimal. Looking at both first-order conditions we can easily see that

[beta] = 1 - (1 - q)(1 - G[[[epsilon].sup.*]]).

On the left-hand side of the expression we have the residual rights of the academic over his or her research and teaching gains. On the right-hand side, we have the marginal benefit to the department from his or her research and teaching effort. We can see immediately that under a regime of hard tenure (q is high), the tenured professor should have more residual rights over the gains produced by research and teaching. However if we have a regime of soft tenure (q is low), the tenured professor should have fewer residual rights.

Given that faculty member and department are both risk neutral, we have the common solution of high-powered incentive contracts. As a consequence, the fixed component of the salary [alpha] is negative according to the participation constraint. There might be several restrictions to the implementation of high-powered incentive contracts in academia even under the assumption of risk neutrality, including statutory constraints. (11) Therefore, in the remainder of the article, we allow for the possibility that a suboptimal contract is implemented. In this suboptimal contract, the residual rights of the academic (our [beta]) are less than the marginal gain from his or her research and teaching effort (our 1 - (1 - q)(1 - G[[[epsilon].sup.*]])), As a consequence, the choice of effort by the tenured faculty member will be less than efficient.

After setting the salary (eventually the one that aligns the interests of the tenured professor with those of the department), we see how the tenure regime (q) and damages (p) increase the likelihood of granting tenure to a candidate with an initial index given by [a.sub.1].

Let us start by considering the decision to harden tenure:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

The first term is negative (unless [rho] is substantially negative, which would mean that the dismissed professor would have to pay damages to the department, which seems highly unrealistic). The rationale is that as tenure become harder, it will be more difficult for the department to dismiss the tenured professor if there is a mismatch in effort and research output.

As for the second term, it is 0 if the optimal high-powered incentive contract can be implemented. Otherwise, it depends crucially on contractual damages. If damages are less than expectation damages (9 < 0), it is negative because the effort exerted by the tenured professor is less than efficient and decreasing as tenure becomes harder (due to the fact that there is a lower probability of being under-compensated and so the marginal loss from being dismissed is reduced). If damages overcompensate a tenured professor ([rho] > 0), the second term is positive because the effort exerted by the tenured professor is less than efficient but increasing as tenure becomes harder (because there is a lower probability of being overcompensated and so the marginal loss from not being dismissed is reduced). In the particular case of expectation damages, the second term is 0 because the choice of effort is not affected by tenure.

In general, hardening tenure makes the department worse off because of diminished incentives for research and high-quality teaching (not a problem if the conflict over optimal effort has been solved by a high-powered incentive labor contract or if expectation damages are implemented) and increased departmental costs in cases of a mismatch between research effort and research output. However, there is a small caveat (probably not empirically very relevant) to this result: Hard tenure could increase research and teaching effort if a large overcompensation for dismissal is in place (with soft tenure, tenured professors would be looking for the opportunity to be dismissed and claim damages, thus they would exert less effort to increase their chances of being overcompensated in the future).

When we have expectation damages (p = 0), a professor is indifferent between being given tenure or being sacked with compensation. However, we have just shown that the department prefers dismissal with compensation over tenure (because the department is worse off as tenure is hardened unless, of course, dismissed professors are overcompensated). The department is better off by paying expectation damages for dismissing faculty than bearing the cost of effort mismatching.

An immediate consequence of our results is that under a regime of hard tenure (specific performance), we should expect the department to impose a higher initial index [a.sub.1] to grant tenure to overcome the expected costs of a mismatch between research effort and research output.

The relationship between the payoff of the department and damages is given by

[differential][U.sub.d]/[differential][rho] = -g([[epsilon].sup.*])(1 - q)[rho] + [differential][U.sub.d]/[differential]e x [differential][e.sup.*]/[differential][rho].

The sign of the first term depends on p. If positive (the department must overcompensate the dismissed faculty member), the department is worse off; if negative (the tenured faculty member is undercompensated), the department is better off. The sign of the second term is 0 when a high-powered incentive contract is implemented, and it will be negative otherwise (because the effort exerted by the tenured professor is less than efficient and decreasing on compensation for dismissal).

In general, higher contractual damages make the department worse off, except if the dismissed professor is substantially undercompensated. In the particular case of expectation damages and a high-powered incentive contract, the sign of the relationship is 0. Thus, a marginal increase of damages when a rule of expectation damages applies does not affect the payoff of the department, but a professor will prefer being sacked with compensation than tenure. In summary, soft tenure with a rule of expectation damages (even damages with marginal undercompensation) comes out as more appropriate than hard tenure.

III. SIGNALING AND REVELATION

As pointed out before, the most compelling argument for tenure is that those who are responsible for recruitment or tenure decisions should have an incentive to reveal information about candidates or give tenure to the best candidate. Consider the situation where the research and teaching index is only observed by the members of the tenure committee (for example, the tenure committee can construct an index of research and teaching productivity as they wish and will only choose the appropriate index if they have an incentive to do so).

Suppose that if tenure is given to the best candidate K, the probability of dismissal is 1 - [G.sub.K](.), and if given to the worst candidate W, the probability of dismissal is 1 - [G.sub.W](.), such that [G.sub.W](.) > [G.sub.K](.).

Giving tenure to candidate W can also be interpreted as rejecting K when the decision concerning tenure promotion is taken on an individual basis. The expected payoffs are, respectively:

[U.sub.jW] = w + [E.sub.W][R([a.sub.2])] + (1 - [G.sub.W][.])(1 - q)[rho] - C(e)

[U.sub.jK] = w + [E.sub.K][R([a.sub.2])] + (1 - [G.sub.K][.])(1 - q)[rho] - C(e),

where we allow for the possibility of expected gains from consultancy revenues and research grants being affected by the tenure decision, positively (a better department fosters extra gains for the faculty members) or negatively (better tenured professors make the market for extra gains more competitive, thus reducing the expected payoff for the incumbents).

As long as the department implements a policy such that [U.sub.jK] [greater than or equal to] [U.sub.jW] , the decision maker has no incentive to hide information or not hire the best candidate. Given the same choice of research and teaching effort, we can see that such policy must satisfy

([G.sub.W][.] - [G.sub.K][.])(1 - q)[rho] + ([E.sub.K][R([a.sub.2])] - ([E.sub.W][R([a.sub.2])] [greater than or equal to] 0.

For the moment, disregard the extra gains from consultancy revenues and research grants. The sign of the left-hand side of the expression will depend crucially on [rho]: It is positive if [rho] > 0, negative if [rho] < 0, and 0 if [rho] = 0. In the particular case of expectation damages, an incumbent is indifferent between dismissal and staying at the department, thus there is no need for tenure. However, if an incumbent is undercompensated if dismissed, then the department should implement hard tenure for the committee members.

This result must be readjusted once we include the extra gains from consultancy revenues and research grants. If giving tenure to a better candidate reduces expected extra gains for incumbents, then even expectation damages or hard tenure would not solve the revelation problem, damages overcompensating tenured incumbents would be necessary. However, if giving tenure to a better candidate improves the standing of incumbents in the market for consultancy and research grants, soft tenure with undercompensating contractual damages could solve the problem.

A different kind of problem also addressed by this set-up is when tenure decisions are subject to voting by all tenured professors. It is not a question of information revelation (the quality of the candidate is common knowledge) but due to the fact that a tenure promotion is decided by secret ballot, a majority of the incumbents might vote against tenure if they lose by granting the deserved promotion. Using previous notation, the department might avoid this problem by simply compensating incumbents in at least [U.sub.jW] - [U.sub.jK]. Once more specific performance and appropriate contractual damages are perfect substitutes in solving this problem. One consequence of the model is that by only allowing the best tenured incumbents (those less affected by a new tenure) to vote for a tenure promotion, the department would save on compensation costs.

IV. DISCUSSION

Our model of faculty hiring focuses on decision-making incumbents, contract damages, academic wages, and tenure. There is, in principle, a trade-off between damages and tenure, the wage being endogenous in our model. In addition, we find that it is important to focus on the faculty personnel committees that actually make hiring recommendations. This last point corresponds well with academics' perceptions of departmental life.

One implication of the model is that universities could focus to their advantage on the hiring decision makers in a department. Their wages could be raised to a level that compensates them for the risks of honestly revealing the skills of entrants. Equally, it would be possible to create tenure just for this group. In practice, universities use uniform tenure policies that do not narrowly target decision makers. They may use higher wages for decision makers, although this may be a corollary of using the recognizably more able to judge the attributes of entrants.

Using only the most able departmental members to judge entrant's skills may be the lowest-cost approach that a university could take to encouraging honest revelation. In terms of our model, if the group taking decisions could always be reshaped to keep it a long way from the hiring margin ([G.sub.W] = [G.sub.K] in the model), there need be no protection in terms of tenure, higher wages, or contract damages. The process could be managed by judgment to avoid the problems flowing from fear of displacement. Also, the information flow might well be highest using such committees. This process in action is the impression gained from observing a typical U.S. hiring committee over time. Being on the committee is a signal that faculty are highly regarded in the rank order. Our model shows why this might be important.

Tenure may very well be a failsafe for academics. It cannot be known in advance which individuals will drop in and out of departmental hiring committees over time. Therefore, tenure may be the way in which an incentive for revelation is maintained over the long run. When an incumbent is in favor, he or she need not worry about vulnerability over time (when dropping out of favor). This observation based on creating long-run incentives may explain the uniformity of tenure, when, apparently, short-run considerations would not necessarily indicate a need for it. Also, it gives an explanation, related to the revelation problem, for why academics would want tenure. (12)

However, the unqualified supply of tenure might be a problem for universities. Tenure creates property rights in jobs, and this could give rise to hold-up possibilities. Whenever a university wished to restructure, it might find itself paying academics almost all of the resulting cost saving, rather than just their losses. (13) Bargaining costs in the face of hold-up would probably deter useful restructuring. The university would require some safeguard of its other interests, that is, would wish to avoid supporting information revelation at the cost of creating huge restructuring costs.

The safeguard appears to be present in the United States. As Tepker (1993) observes, it is relatively common for courts to require the reinstatement of dismissed tenured faculty but for schools to refuse. The courts then award expectation damages, the standard award for breach of contract that aims to put the victim in the position he or she would have been in had the contract been honored. (14) This broadly requires payment of the difference between academic earnings and those in the next best occupation. Presumably, redundant academics who try to hold out for more than their expectation induce universities to dismiss them, knowing that courts impose expectation damages for breach. Courts may also be useful devices for assessing expectation damages and its components, like the value of alternative employment.

Our model shows considerable possible interaction between wages, tenure, and damages. In relation to the discussion, there is always an upper limit on damages that protects universities from hold up. However, damages may be the true force behind tenure, which defines property rights so that expectation damages can be awarded in the event of dismissal. That protection will encourage participation in hiring committees and honest revelation of entrants' skills. We are broadly in agreement with Carmichael's (1988) approach but do not think it is correct to separate tenure from contractual damages. Tenure reform may not be the panacea it is often thought to be and, according to our discussion, may remove an important basis for defining damages for breach of contract that in turn supports information revelation. Indeed, in the case of the United Kingdom, it is becoming clear that statutory reform of tenure yielded little change because the universities have chosen not to change their existing practices, presumptively because these were efficient.

Finally in this section, we speculate about the failure to pursue the logic of performance wages, contractual damages, and tenure standards differentially across the academy. We do not see vast differences in salaries and other differentiable rewards, comparing departments or levels of efficiency attainable for a department. There is some separation of the kind consistent with the analysis of this article, for example, rewards in humanities tend to be significantly less than in business, medicine, or law schools. The impression is that the differentiation does not go far, and it is unusual for a university administration to offer efficiency bonuses to departments. Our view on this is that it is costly to differentiate performance and may invite costly rivalries within the organization, particularly if errors were made in allocating bonuses.

V. CONCLUSION

Our results suggest that tenure is just one of many mechanisms able to achieve honest revelation in university hiring procedures. Interactions between alternative mechanisms based on wages, damages, tenure, or restricted hiring committees are of significance. It appears possible to link damages to tenure in a manner that supports revelation without creating incentives for hold up. We point to a delicate balance within the academic labor contract.

APPENDIX: MODEL WITH LITIGATION COSTS

Let [L.sub.j] be litigation costs borne by the academic if dismissed, and let [L.sub.d] be litigation costs borne by the department if dismissal takes place. The payoffs should be rearranged:

[S.sub.j] = w + (1 - q)[rho] - [L.sub.j]

[S.sub.d] = q[a.sub.2] + (1 - q)(v - [rho]) - [L.sub.d] - w.

We should notice that [L.sub.j] plays an important role in determining the credibility of the dismissed professor's threat to sue. If [L.sub.j] is reasonably low, the lawsuit by the dismissed professor is credible, and the qualitative results are not very different from the main text. The department decides to dismiss the professor if

[epsilon] [greater than or equal to] [a.sub.l] + e + [rho] - v + [L.sub.d]/(1 - q) = [[epsilon].sup.*],

where higher litigation costs for the department reduce the likelihood of dismissal.

After the appropriate choice of a labor contract, the expected payoff of the department is

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

The first-order conditions with respect to q and [rho] are (assuming a high-powered incentive labor contract is implemented)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

Litigation costs are a loss for both parties. As a result, avoiding litigation becomes more important. Thus, specific performance (hard tenure) becomes more attractive because it saves on litigation costs. The importance of this effect is overestimated in the model due to the fact that settlements are not allowed.

Consider now the situation where [L.sub.j] is large enough to deter a lawsuit filed by the dismissed professor. The department decides to dismiss the professor if [a.sub.2] < w or

[epsilon] [greater than or equal to] [a.sub.1] + e - [alpha]/(1 - [beta] = [[epsilon].sup.*].

After the appropriate choice of a labor contract, the expected payoff of the department is

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

The parameters q and [rho] are irrelevant in this case because there will be no litigation.

REFERENCES

Brown, W. O. "University Governance and Academic Tenure: A Property Rights Explanation." Journal of Institutional and Theoretical Economics, 153(3), 1997, 441-61.

Carmichael, H. L. "Incentives in Academics: Why Is There Tenure?" Journal of Political Economy, 96(2), 1988, 453-72.

Chatterjee, J., and R. C. Marshall. "A Model of Academic Tenure." Working Paper. Pennsylvania State University, 2001.

Dnes, A. W., and J. S. Seaton. "The Reform of Academic Tenure in the United Kingdom." International Review of Law and Eeonomics, 18(4), 1998, 491-510.

"The Research Assessment Exercise and the Reform of Academic Tenure in the United Kingdom." Contemporary Economic Policy, 19(1), 2001, 39-48.

Li, S., and H. Ou-Yang. "Incentives, Performance, and Academic Tenure." Working Paper. Duke University, 2003.

McCormick, R., and R. Meiners. "University Governance: A Property Rights Approach." Journal of Law and Economics, 31(2), 1988, 423-42.

McKenzie, R. B. "In Defence of Academic Tenure." Journal of Institutional and Theoretical Economics, 152(2), 1996, 325-41.

McPherson, M., and M. O. Schapiro. "Tenure Issues in Higher Education." Journal of Economic Perspectives, 13(1), 1999, 85-98.

McPherson, M., and G. Winston. "The Economics of Academic Tenure: A Relational Perspective." Journal of Economic Behavior and Organization, 4(1), 1983, 163-84.

Siow, A. "Tenure and Other Unusual Personnel Practices in Academia." Journal of Law, Economics and Organization, 14(1), 1998, 152-73.

Tepker, H. F. "Good Cause and Just Expectations: Academic Tenure in Oklahoma's Public Colleges and Universities." Oklahoma Law Review, 46, 1993, 205-17.

(1.) See, for example, Bruno v. Detroit Institute of Technology, 215 N.W. 2d. 745 (Mich. Ct. App. 1974) (expectation damages should be awarded for loss of tenure, if reinstatement is resisted by the college). More recent cases buttressing the argument in the text include Gray v. Mundelein College (Loyola University), 695 N.E.2d. 1379 (Ill. App. 1 Dist., 1998) (confirming the right to bring an action for breach of contract after a merger removed tenure); and Johnson v. Colorado State Board of Agrieuhure, 15 P. 3d. 309 (Colo. App., 2000) (concerning the retroactive introduction of posttenure review). In Mundelein, the court was very clear that custom and practice did not undermine the contract, and that expectation damages equal lost net earnings absent the breach. The duty to mitigate losses, for example, by taking an available job, was also confirmed.

(2.) Tenure in Britain was easier to receive and harder to lose compared with the United States prior to 1988. The act allowed dismissal for financial reasons, such as closing a department, subject to protections on academic freedom. British cases also show a history of paying damages in lieu of reinstatement, as per Bruno in the United States.

(3.) For example, Brown (1997) addresses tenure from a property rights perspective where the ownership of ideas and the freedom of speech are the driving forces. Both variables are usually excluded from an index of research and teaching.

(4.) For a model of pretenure effort, see Chatterjee and Marshall (2001).

(5.) See Li and Ou-Yang (2003) for an investigation of 249 economists before and after their tenure promotion.

(6.) Note that Slow (1998) conversely argues that tenure is used to induce older academics, whose research productivity has fallen, to do less research. A tenure contract alleviates the misallocation of effort.

(7.) An empirical validation of this setup can be found in Li and Ou-Yang (2003).

(8.) We ignore the possibility of settlements and golden parachutes to avoid litigation. The qualitative results we present are not affected by this assumption.

(9.) There is of course always the possibility that the whole department is closed, therefore even with hard tenure q is strictly less than 1.

(10.) These results are consistent with the empirical analysis provided by Li and Ou-Yang (2003).

(11.) For example, in most continental European countries, academics are essentially public servants, salaries are fixed by the government or the legislator, and high-powered incentive labor contracts are strictly forbidden.

(12.) Our explanation is a demand-side one. A complementary supply-side explanation is that uniform tenure is the result of competition between universities to attract top scholars. Uniform tenure is the equilibrium of such game because there is a cost to be borne by those defecting from the equilibrium (a reputation effect).

(13.) That is, compensation would exceed expectation damages, which are explained shortly.

(14.) The compensating variation for a price fall.

ANTONY DNES and NUNO GAROUPA *

* We are grateful to one anonymous referee, the editor, and Benito Arrunada, Antonio Cabrales, Andreu Mas-Colell, Luis Corchon, Eduardo Rodes, Wolfgang Weigel, seminar participants at the Central European University (where Dnes is a regular visitor) and participants at the annual conference of the European Association of Law and Economics, Vienna, in September 2001 for helpful suggestions.

Dnes: Professor, Business School, University of Hull, Kingston-upon-Hull, HU6 7RX, England. Phone 44-1482-465875, Fax 44-1482-466216, E-mail a.dnes@ hull.ac.uk

Garoupa: Research Affiliate, CEPR, London; and Professor, Faculdade de Economia, Universidade Nova de Lisboa, Campus de Campolide, P-1099-032 Lisboa, Portugal. Phone 351-21-3801600, Fax 351-21-387033, E-mail [email protected]