Inequality, Discrimination, and the Covenant Not to Compete: A Commentary on Orly Lobel, Exit, Voice & Innovation

I. Introduction

I very much appreciate being asked by the editors of the Houston Law Review to comment on Professor Orly Lobel’s Frankel Lecture entitled Exit, Voice & Innovation: How Human Capital Policy Impacts Equality (& How Inequality Hurts Growth).^[1] As anyone familiar with Professor Lobel’s other work would expect, her article ranges widely and offers insightful connections among what are often considered to be very distinct fields of inquiry. Her take-off point is Albert Hirschman’s famous essay, Exit, Voice, and Loyalty,^[2] but as her article impressively shows, the economist’s “exit” and “voice” are in fact complex ideas with a dense legal structure.

Within this complexity, I want to talk about the parts of the article that connect the law of human capital with gender inequality. Professor Lobel’s thesis in this regard is straightforward: “Restricting exit and voice harms all workers, but its harm to women and minorities is disproportionately greater. Moreover, the specific paths of exit and voice are patterned by inequality. That exit and voice restrictions can entail disproportionate harms on women should be understood as rising from several interrelated factors.”^[3] In considering the impact of these “interrelated factors,” I am going to focus especially on the most salient restriction affecting an employee’s power to exit: the covenant not to compete against a present employer for a set number of years once the present employment has ended.^[4]

II. Employee Covenants Not to Compete

I should say right off that, with regard to these noncompete clauses, Professor Lobel and I fundamentally agree as to what social policy should be. We do not like employee covenants not to compete, especially in the profligate ways they are now being used. We are both among those who recently submitted a petition to the FTC to use its regulatory authority over unfair trade practices to prohibit such clauses.^[5]

One can be against these clauses for many different reasons. One might claim they are bad economic policy because they lead to a less innovative, less creative economy. One might claim they are transactionally unfair because, in the typical case, employees get very little in exchange for what they are asked to give up. My opposition, however, is fundamentally on liberty grounds. I believe in free labor.

Put in Professor Lobel’s terms, the freedoms at stake are matters of both “exit” and “voice.” As she very nicely sets out, despite the sometimes negative connotations of “quitting,” exiting a job is often a positive act—the beginning of a new, more valuable use of an employee’s talents and labor.^[6] But as Amartya Sen, among others, has emphasized, any freedom requires more than an abstract right.^[7] The “right to quit” also requires the capability to quit and move on, which is greatly diminished when a worker is prevented from making a living doing the kind of work she normally does, which is, of course, exactly what these covenants do. Professor Lobel also presents a good case, I think, for recognizing that these sorts of limits on exit also produce limits on voice and thereby reduce what freedom there is within the modern workplace. And I would add that it seems to me that not being able to quit—not having the capability as well as the right to quit—conditions the entire emotional spirit one has in going to work each day (which in most situations means going to work and following orders). All of these points stand on top of the fundamental dubiousness of the deal many workers have to accept—that the covenants mandate a year or two of idleness, or trouble, or destitution even if the employee can be fired “at will,” indeed even if the employee has good cause to quit; that the covenants are often adopted without any additional consideration and enforced without any severance pay; and that, as the practice has mushroomed, they are extracted from workers doing the most ordinary sorts of work,^[8] such that it is a victory for freedom for a state legislature to say that they cannot be applied to those whose right to a minimum wage and overtime need to be protected by the federal Fair Labor Standards Act.^[9]

On the assumption that the usual reader of this journal is a member of the legal profession, it is perhaps worthwhile to give the following hypothetical illustration of the baneful workings of the covenant not to compete. It is a common problem, when a lawyer moves from one firm to another, to have to cordon off the lawyer from specific matters at the new firm in order to prevent leakage of secret or privileged information gathered at the old firm. Suppose legal employers, to guard against such leakage, began demanding that their legal employees sign a covenant not to compete. What would then be the position of the law school graduate taking his or her first job with such a firm? Of course, that job might work out permanently, but it is a common experience these days that it will not. The new lawyer may not fit the firm, or the firm may not fit the new lawyer, or even if they do fit each other, there may be a better fit across the way. But the covenant would likely say something like: the lawyer who leaves for whatever reason cannot work for a competing firm in the area covered by the old firm’s business for one, or perhaps two, years. Unless the old and new firms are both boutiques dealing in different fashions, they are likely competitors, and they likely see themselves as potentially active at least in their entire state. Does our young lawyer have to find his or her next job out of state (and move and take a new bar exam)? Does our young lawyer have to find a new employer willing to pay him or her to sit out unproductively for a year or two? Does our young lawyer have to find a new employer dealing in a specialty not covered by the former employer and lose what expertise he or she has? Fortunately, we will never find out which of these highly restrictive alternatives will come to pass. The rules of professional conduct prohibit contracts that, in the words of the applicable ABA Model Rule, “restrict[] the right of a lawyer to practice after termination of the relationship.”^[10]

III. Inequality or Discrimination?

But the subject of this Commentary is not what I think, or what Professor Lobel advocates as a matter of policy, but rather her claim, to repeat it, that “the specific paths of exit and voice are patterned by inequality.”^[11] Professor Lobel gives us a lot of data and reports a lot of studies in her article to the effect that one or another legal or contractual restriction on employees has a greater, more negative impact on female employees than on male ones. Whether these studies are sound—whether the data add up to prove the asserted gendered impact—is addressed in another commentary to Professor Lobel’s lecture by Amy C. Madl and Lisa Larrimore Ouellette.^[12] What I want to focus on is the question: Assuming the evidence shows what the article claims, what does it add up to?

Is Professor Lobel’s argument in her article a general welfare argument? Does she want to argue that we would be better off in economic terms if covenants not to compete disappear because there will be better, more economically rational utilization of the female workforce? She has made that argument in other articles,^[13] and she suggests it here, for example, in her claim that the effect of noncompetes is to reduce innovation.^[14] But I do not take that to be the gravamen of her article. I take her instead to be claiming that the present situation is not just wasteful but wrong.

Wrong in what sense? There are two broad possibilities: giving legal credence to these contractual terms is wrong because they contribute to social inequality; or it is wrong because going beyond inequality per se, their use partakes of discrimination. Although we often carelessly conflate these two positions, they are not the same thing. Take, for example, the claim that covenants not to compete are bad because in our society families are more likely to move for a new job for the husband than for a new job for the wife, with the consequence that a restriction on the career possibilities for a woman within her current geographical area—the way noncompetition covenants have traditionally been framed—is more likely to harm her employment prospects than to harm a man in comparable circumstances.^[15] That is certainly a claim that the covenants work out unequally, and unequally in a patterned fashion, but is it a claim of discrimination?^[16] Professor Lobel does not spend time on this question. She more commonly uses the language of inequality, and only occasionally speaks of discrimination, and it is certainly possible to be opposed to inequality on its own terms. But, as I will try to show, I do think it is important to mark the difference with some clarity and to consider how close the evidence she produces comes to demonstrating the stronger claim of discrimination. To be clear, I am not asking this as a matter of positive law—of what would be needed to bring any particular set of facts within any particular statute or rule—but as a matter of what the broad conceptual framework ought to be.

The starting point for such an analysis would seem to be the cases from the mid-1970s that set forth the constitutional structure for analyzing this question in the context of race: Washington v. Davis^[17] and Arlington Heights v. Metropolitan Housing Development Corp.^[18] In these cases, the Supreme Court distinguished between official action as to which all that is shown is “racially disproportionate impact” and official action as to which a “racially discriminatory purpose[]” can be shown to have been an actual “motivating factor in the decision.”^[19] The Court’s principal argument for requiring proof of “discriminatory purpose” was that “disproportionate impact” created too broad a category to be considered discriminatory; it could be true of all sorts of official action taken for what we would consider to be innocent reasons.^[20] Presumably the Court was also responding, in its focus on actual intent, to a sense that “discrimination” as legally relevant includes not just differentiation, but an element of wrongfulness: subordination, or denigration, or stigmatization.

If we take our conceptual universe to include only these two choices—either we can prove discrimination as intent, or we have only an insufficient claim of disparate impact—it seems clear that the gendered results of noncompetition agreements do not make a case for a finding of discrimination. Showing mere disproportionate impact is not enough, and the facts that these covenants long predate the substantial influx of women into the modern workforce and seem now to be extracted from employees without regard to their gender prevent any finding of an actual intent to discriminate against women. Of course, if in a particular case an employer enforced the clauses only against its female employees, that might count as discrimination, but that is hardly the model case.^[21]

IV. An Intermediate Possibility: Looking at “Intent”

If, however, we imagine “disparate impact by gender” and “subjective intention to subordinate women” not as the only possibilities, but rather as the extreme opposite ends of a spectrum, there is a lot of room between them, and there have been many suggestions that at least some of this middle ground should also be thought to constitute discrimination. One group of suggestions moves away from the “subjective intention” end of the scale by rejecting the claim that the law should require proof (including of course reasonable inference) of what the relevant actors consciously thought. For example, there is the claim that action that looks discriminatory to an outside observer—action that “objectively” carries a discriminatory meaning—ought to sometimes meet the test in place of subjective intent. This claim may be right—at least I made it myself in 1994.^[22] Professor Robin Lenhardt’s thesis, that what matters is not official intent but rather meaningful stigmatization,^[23] is not far away from the same proposition, and it is similar to the suggestion of Professor Charles Lawrence that sometimes proof of disparate impact itself, without more, ought to be enough to justify an inference of discriminatory intent because bias can be unconsciously held even by the actors themselves.^[24]

Whichever, if any, of these proposals is convincing, none of them seems to fit the range of facts presently at issue. Because of the history mentioned above, there seems to be nothing about a covenant not to compete that carries to the reasonable person an odor of discriminatory intent, or suggests stigmatization of a group, or invites an inference of unconscious discrimination against women. We need to move closer to the end of the spectrum that starts by looking at the impact of the action.

V. Another Intermediate Possibility: Looking at Justification

This quest invites consideration of the Supreme Court’s handling of several antidiscrimination statutes, where the Court has held that such statutes cover action that has a disparate impact on specified categories of people without proof of discriminatory intent, but only when no substantial nondiscriminatory justification for the action can be given.^[25] That the Supreme Court wants in these cases to establish an intermediate stance seems clear: clear in the Court’s construction of the statutory language to reach cases of differential impact even where discriminatory intent cannot be shown, and clear in the Court’s addition of various rules, some substantive and some procedural, to make sure that impact analysis only goes so far. But as to what is the theory underlying this stance, it is hard to say.

Justice Kennedy’s opinion of the Court in the relatively recent case in this line, Texas Department of Housing & Community Affairs v. Inclusive Communities Project, Inc.,^[26] provides a good example of both the stance and of its opacity. There the Court held that there was disparate-impact liability under the Fair Housing Act but that “housing authorities and private developers” had to “be allowed to maintain a policy if they can prove it is necessary to achieve a valid interest.”^[27] This, said the Court, was analogous to the earlier holding under Title VII that “an employer may maintain a workplace requirement that causes a disparate impact if that requirement is a ‘reasonable measure[ment] of job performance.’”^[28]

Putting to one side how it dealt with questions of legislative history and textual interpretation, the Court offered many different justifications for its two-part holding. One possibility is that it was treating the Fair Housing Act as directly concerned with inequality per se. On this view, the Court’s emphasis on allowing a defendant an opportunity to justify the disparate impact is not part of showing that there is or is not discrimination, but is rather a precaution taken to make sure that the quest for equality is not socially or economically destructive: “An important and appropriate means of ensuring that disparate-impact liability is properly limited is to give housing authorities and private developers leeway to state and explain the valid interest served by their policies.”^[29] Here, the phrase “properly limited” suggests that the quest for equality is a legitimate part of the statute’s goals but is not the be-all of public policy. But if this interpretation is adopted, it seems to follow that once a disparate impact is shown, the entire balance of social advantages and disadvantages is a matter for judicial weighing. As a matter of the allocation of institutional competency and authority, that seems to create a doubtful doctrine. It seems better to view the combination—proof of disparate impact plus a chance to justify—as searching for at least some element of discrimination as well as of inequality.

That element might be seen in two different ways, also suggested by Justice Kennedy’s opinion. One strand is evidentiary. Disparate impact itself, he says, “plays a role in uncovering discriminatory intent: It permits plaintiffs to counteract unconscious prejudices and disguised animus that escape easy classification.”^[30] On this view, the additional search for valid reasons for what the defendants did could be seen as part of the same inquiry, with a valid reason serving to negate an inference of discriminatory intent. This, of course, would still be a judgmental standard, but not one directed simply to overall social welfare.

Returning to the question of covenants not to compete, it seems to me that the usual justifications would be adequate to meet this test. There are two primary reasons that employers (or their attorneys) give for using covenants not to compete: first, that their trade secrets really are at risk when employees go to work for competitors, but the common law protection for trade secrets will not by itself catch up with the potential harm; and second, that employers cannot adequately invest in training employees without some assurance that the training they pay for will not immediately be employed on behalf of their competitors. These are sensible propositions. To my mind, the claims of employees to their freedom, already stated, are more important. But if the question to be posed is, could employers enforce covenants not to compete for what they, and we, reasonably understand to be nondiscriminatory business reasons, it seems to me the answer is yes.

There is another possible reason for looking at the proffered justifications, also suggested, although perhaps not highlighted, by the Inclusive Communities opinion: that actions which are not otherwise discriminatory, but only unequal, become discriminatory because of the discriminatory quality of the social reality in which they come to happen. Before addressing the Fair Housing Act’s terms, Justice Kennedy spends considerable time on its historical background, including “both open and covert racial discrimination” that had “prevented black families from obtaining better housing and moving to integrated communities.”^[31] Later, in describing the Act’s “central purpose,” it is these “discriminatory practices” that he seizes upon.^[32] “These unlawful practices include zoning laws and other housing restrictions that function unfairly to exclude minorities from certain neighborhoods without any sufficient justification. Suits targeting such practices reside at the heartland of disparate-impact liability.”^[33] On this view, the question posed to the excuse being proffered to justify disparate impact is whether it is part of a general system of discrimination or not.

The connection between finding present discrimination and preexisting social practice is made even more strongly in the first disparate-impact-plus-chance-to-justify case, Griggs v. Duke Power Co., decided under Title VII of the Civil Rights Act of 1964.^[34] There, the employer made applicants for employment meet educational and standardized-test requirements for jobs as to which the requirements had no shown job relationship; these requirements had a disparate impact to the disadvantage of black applicants, but, although the employer had in the past been overtly discriminatory by race, there was no showing of a discriminatory purpose in the adoption of these new policies. The Court held that these not job-related requirements could not be used. “Under the Act, practices, procedures, or tests neutral on their face, and even neutral in terms of intent, cannot be maintained if they operate to ‘freeze’ the status quo of prior discriminatory employment practices.”^[35] While this language might be thought to be conditioned by the fact that this very employer also continued to promote those it had already hired, in its overtly discriminatory past, to positions for which the educational attainment for new hires had since been increased, the opinion of the Court clearly went further. For the Court proceeded to treat as relevant the impact, on preparedness for the tests, of the fact that black applicants (in North Carolina as of the 1960s) had gone to school in inferior, segregated school systems, ultimately concluding that “[t]he Act proscribes not only overt discrimination but also practices that are fair in form, but discriminatory in operation.”^[36]

Considered in terms of carrying out a statutory mandate, cases like Griggs and Inclusive Communities have much to recommend them. It seems natural to assume that Congress passes legislation in light of existing circumstances in order to cure an existing evil. When it selects a particular kind of existing inequality to outlaw, and connects it to particular groups that it thinks needs that protection, it is reasonable to assume that it is making just the kind of connection between a pattern of past subordination and present practices that we have been discussing. But, absent that reliance on Congress as a sociological fact-finder, relevant to the construction of a particular statutory text, how would we justify making the connection?

VI. Context as a General Approach

The carryover from social context to a determination of discrimination is easy to see when an otherwise neutral action is superimposed on a reality that was created by the discriminatory acts of the very same actor—when, for example, an otherwise unexceptional seniority system is superimposed on a set of employees for whom opportunities to gain that very seniority in desirable positions were previously assigned by gender or race. Can this argument legitimately be extended to more general, although discriminatory, prior social conditions, such as the inferior schooling relied upon in Griggs? A very good recent piece by Deborah Hellman of the University of Virginia Law School, “Indirect Discrimination and the Duty to Avoid Compounding Injustice,” says yes.^[37] Hellman argues that there is a general duty to avoid compounding the injustice that someone else has already suffered.^[38] Hellman gives as an example of such compounding, an insurance company that decides to raise a woman’s life insurance rates because she has been a victim of domestic abuse. Assuming that the company would ordinarily be allowed to set its own rates, and assuming that data show that battered women are likely to die at a younger age, Hellman argues that it is still wrong for the company to do this.

[T]he insurer bears some responsibility for the effects of the original injustice because the insurer takes the injustice or its effects as a reason to charge battered women higher rates. In addition, the insurer amplifies the original injustice because the harms caused by the battery now include not only the physical and mental suffering caused by domestic violence but also the economic loss of high priced insurance.^[39]

More generally, to constitute such a compound injustice, she says, “an actor must amplify the harm of the prior injustice and must take the fact of victimi[z]ation or its effects as her reason for acting.”^[40] When the prior injustice was itself discriminatory, she goes on to say, then the compounded injustice may also be considered a form of discrimination.^[41]

An example from another forthcoming article of Professor Lobel that would exemplify this claim is the practice of asking prospective employees about their prior pay rates.^[42] If we assume that the prior pay of women was lower than it should have been because of a general social practice to subordinate women by paying them less than men, then even if the particular employer who asks the question did not participate in that practice, and even if it uses the information about prior rates in a completely equal way, say by offering each lateral hire 10% more than he or she was making previously, the net result will be to carry forward in the now-offered salaries the past discrimination. This practice could, therefore, itself be considered discriminatory.^[43]

Can we make a similar argument against covenants not to compete in particular, or more generally, about the range of practices Professor Lobel discusses in her article? In my view, we are not there yet. A claim like Professor Lobel’s, that covenants not to compete differentially reduce the prospects of women because women are less inclined to be risk-takers in developing new employment opportunities,^[44] does not seem to me to be sufficient. Without a showing that this attitude is itself the consequence of prior discrimination, the point comes too close to the “merely differential impact” end of the spectrum. But if, for example, in scientific and technical job categories, the data that Professor Lobel presents about women in those jobs not getting lab space, not getting published, and not getting publication credit,^[45] can be shown to make a covenant not to compete more of a hindrance for women in these fields in finding a new employer, the case might move closer to showing discrimination.

In short, I think Professor Lobel presents a convincing case that various legal mechanisms that structure employee “exit” and “voice” have an impact that is unequal by gender. Whether they are “discriminatory” seems to me not yet shown.

VII. Conclusion: Why It Matters

Does it make a difference whether we understand the differential outcomes created by various social or legal mechanisms to be discriminatory or merely unequal? If by unequal, we simply mean the differences between one person’s circumstances and another’s, taken as individuals, that stands very far away from the type of discrimination we are addressing: subordination or denigration based on social categories. But once inequality is shown to be patterned, to differentiate between one group of people and another—once it shows up as a statistical regularity—the apparent difference between inequality and discrimination begins to shrink. Still, I think it is important to draw a line between the two categories, even if, as I have suggested, the line is more complicated than the distinction between “bad intent” and “only disparate impact” would have it. It is important for at least the following three reasons.

First, the concepts of “nondiscrimination” and “noninequality” portray very different ideal societies. A society in which no group is discriminated against—that is, no group is subordinated—is an attractive ideal that leaves a great deal of room for individuality. A society in which there is no inequality, even just no group inequality, is not an attractive ideal because of the great social constraints that are needed to reach it.^[46] It is also unattractive because, even as seen by the least advantaged members of society, pure equality might well be counterproductive.^[47]

Second, the two concepts have very different cultural valences in our society. Americans are notoriously accepting of social inequality, even patterned social inequality, while yet strongly opposed to invidious discrimination. Social theorists—and, for that matter, practical politicians—can argue that this cultural separation is too strong, that inequality is more often the result of discrimination than we usually credit. But in the popular imagination—and in the workaday decisions of the law^[48]—being down and out is a very different matter from being subordinated and denigrated by others.

Third, as a practical matter, a determination that some practice or rule is discriminatory calls for a different remedy from a determination that it unfairly contributes to inequality. If it is discriminatory, we ought to extirpate it; its very existence is an offense against what we want our society to be. If it contributes to inequality but is not discriminatory, our remedial options are far greater. We can get rid of it, or we can let it be and make up for the inequality elsewhere in our legal or economic system.^[49] A combination of many actions might well create serviceable, long-term equality or near equality as a unit; by contrast, offsetting one discriminatory action against another represents, at best, an extraordinary intervention.

This question of remedy has significance for those like Professor Lobel, who want to argue that there are inequalities that need correction. Merely demonstrating that a specific legal practice contributes to a specific type of inequality—in this case, to a gendered inequality—is not enough to demonstrate what should be done. But although Professor Lobel’s article might be said to veer close to making that mistake, surely her work as a whole, which discusses a panoply of considerations applicable to refashioning our human capital law, does not. Accordingly, we should take this fine article as a warning rather than a program—a warning about the intricate ways in which legal practices can affect the social relationships of our society if we do not pay enough attention to what we are doing.