Covenants not to compete, which prevent workers from quitting their employer and moving to a competing employer, have ambiguous effects on welfare, according to economic theory. Noncompetes may enable employers to invest in assets by protecting those assets from expropriation by workers or competitors, but noncompetes also interfere with labor market competition. Recent research, however, suggests that the anticompetitive effects of noncompetes are greater than any benefits, and thus that the existing common law regime for regulating noncompetes is too weak. A plausible step forward is to strengthen the antitrust law on noncompetes. A better antitrust law would shift the burden, allowing employers to use noncompetes only when they can prove that their use of noncompetes results in higher wages in the affected labor market.
Eric Posner1
New research in economics suggests that employee covenants not to compete play a significant role in the cartelization of labor markets and the suppression of wages. This research uses the variation in the strictness of the common law test for enforceability across states as a lever for determining the impact of noncompetes on wages, job mobility, and related factors. The studies show that in states where noncompetes are broadly enforceable, workers earn lower wages than do workers in states where noncompetes are subjected to greater restrictions or are prohibited. The explanation appears to be that workers subject to broader noncompetes lack a credible threat to quit if they are unhappy with their wages, and so lack bargaining power necessary to win raises from their employers. Because workers are effectively removed from the labor market, fewer employers will compete for them, and with fewer workers available, firms may be deterred from entering markets in the first place, further exacerbating the problem. This reduction in competition is, or ought to be regarded as, an antitrust violation. Yet courts have been unreceptive to antitrust challenges to noncompetes. This needs to change.
Employers have used noncompetes for centuries, and early on courts recognized that they could interfere with competition. In the common law, an employer could use a noncompete only if it could identify a “protectable interest” — usually trade secrets or customer lists — and prove that the noncompete was no broader than necessary to protect that interest. The common law test was not toothless but the sanction — merely nonenforcement — was. That meant that noncompetes were usually challenged only when they applied to relatively highly skilled employees of great enough value to a competing employer that the litigation costs of challenging the noncompete could be justified.
Until recently, it was thought that noncompetes were used mainly for such highly skilled workers. Whether or not a noncompete was overbroad in a particular case, the traditional understanding was that they were “customized” in the sense that they were negotiated individually to reflect the specific circumstances of a particular employee, as indeed the common law seemed to require. But this turns out not to be true. It appears that many people take jobs without reading sometimes lengthy employment contracts or reading but not fully understanding the terms of those contracts. In many cases, the noncompete is presented to employees after they accept employment. Reluctant to quit after having just accepted a job, uncertain about the legal effect of the noncompete, and optimistic about their future at their new employer, workers appear to be unwilling to demand additional compensation or to refuse to sign the noncompete.2 Other workers who are initially unaware of noncompetes may eventually learn about noncompetes if their employers warn them against taking jobs with competitors or hear about coworkers who are prevented from doing so, but no doubt many workers never do learn about the noncompetes to which they are subject.
A survey conducted by the economist Evan Starr and his colleagues revealed that 18.1 percent of all workers in 2014 were subject to a noncompete, and that nearly 40% of all workers had at one time been subject to a noncompete; among workers without a college education the comparable figures were 14.3 percent and 26.6 percent.3 The vast number of workers subject to noncompetes, including numerous low-skill workers, suggests that most noncompetes are standard terms in form contracts imposed by an employer on hundreds or thousands of employees. The employees, like consumers who are confronted with mass consumer contracts, can take it or leave it; they are not allowed to negotiate over boilerplate. The phenomenon is illustrated by the controversy over the Jimmy John’s sandwich chain, which imposed noncompetes on its low-skill sandwich workers as well as managers and other employees.4 These noncompetes were not negotiated, and they were likely in violation of the common law. The franchise withdrew them to settle claims brought by state attorneys general, and is now contesting a class action.
Employers benefit from noncompetes in two ways. First, a noncompete protects certain assets and investments from expropriation by workers. For example, a worker with access to trade secrets or customer lists may be lured away by a competitor who is willing to pay a high wage in return for that information. If an employer is not permitted to protect its assets with a noncompete, it may be unwilling to invest in these assets in the first place, to the detriment of the economy. The noncompete is used to prevent this type of expropriation. If this theory is correct, employers will pay workers subject to noncompetes higher wages than those who are free to leave in order to compensate the first group for lost future employment opportunities.
Second, a noncompete weakens the bargaining power of employees by preventing them from credibly threatening to quit if compensation stagnates or working conditions worsen. While in theory employees should bargain for a higher wage in return for giving up future employment opportunities, employees may be uninformed of the noncompetes in their contracts or unwilling to bargain with a new boss. And even if employees do demand wage premiums, other workers will pay the price. If noncompetes are widely used, they will prevent firms from entering a labor market because the relevant labor force is tied up. The reduction in competition among employers will hurt workers by enabling incumbent employers to charge below-market wages, leading to a reduction in employment and overall production.
From the standpoint of public policy, only the first benefit counts in favor of enforceability of noncompetes. The noncompete is a contractual device for protecting valuable assets when other means — for example, enforcement of trade secrets law — fall short. The second benefit produces a deadweight cost for society, based on the standard understanding of monopsony. The employer’s gain is less than the loss to workers and consumers because the employer must reduce production in order to suppress wages. This point is just the flipside of product-side antitrust theory: monopoly profits for the seller does not justify the deadweight cost for society generated by the contraction of production needed to keep prices high.
The social value of any particular noncompete, then, depends on the tradeoff between asset protection and reduction in competition. While one cannot in the abstract determine whether the benefit or cost is higher in any particular case, employers have strong incentives to use broad noncompetes because they obtain most of the benefit while the cost is externalized on society. The common law acknowledged this problem by allowing noncompetes only for the purpose of protecting assets, tolerating the inevitable reduction in competition only to the extent that this purpose was advanced. But courts never undertook or demanded from employers a serious empirical account of the two effects, and, as noted, imposed a weak sanction for overbroad noncompetes.
The extensive use of noncompetes, coupled with employers’ incentives to overuse them, and the weak common law sanctions, has led many commentators to suggest that noncompetes do more harm than good. Indeed, a few states have banned employee noncompetes — most significantly, California, which enjoys the most innovative economy in the world despite the theoretical prediction that noncompetes are essential to innovation.5 But until recently, there was no hard evidence.
In the last couple of years, however, the evidence has rolled in. The new studies consistently find that noncompetes lower wages. For example, Evan Starr and Michael Lipsitz found that a 2008 Oregon law that banned noncompetes for low-income workers increased average hourly wages by 2-3 percent, implying that the hourly wages of the fraction of workers subject to noncompetes increased by even more—as much as 14-21 percent.6 Lipsitz, Kurt Lavetti, and Matthew Johnson found that wages increase with the restrictiveness of noncompete laws across states and across time. An increase in their measure of restrictiveness from the 10th to the 90th percentile is correlated with a 3-4 percent decline in annual wages.7 In yet another paper, Natarajan Balasubramanian and a group of coauthors found that when Hawaii introduced a ban on noncompetes for tech workers, the workers’ wages increased 4-5 percent compared to the wages of similar workers in states with the mean level of noncompete enforceability.8 These papers come on the heels of earlier papers that found similar results, although noncompetes are associated with high wages for some groups, including physicians and CEOs.9
The evidence of wage reduction suggests that the main effect of noncompetes in American labor markets has been to reduce competition rather than spur innovation. The implications are troubling in light of another wave of new research that finds that American labor markets are far more highly concentrated than economists once believed. Researchers have calculated concentration levels for every labor market in the country and found that a large majority of them are dangerously concentrated, particularly in rural and other thinly populated areas.10 The widespread use of noncompetes may help maintain these high levels of labor market concentration — by preventing new businesses from finding enough workers to enter existing labor markets. They are also likely to be effective at suppressing wages in labor markets that are highly concentrated since workers may have trouble finding alternative employment even if they are willing to take jobs that are outside the coverage of the noncompetes that bind them. The high level of concentration, accompanied by increasing use of noncompetes, may account for some of the most visible problems in the American economy: the decline in job mobility, the stagnation of wages below the highest percentiles, rising inequality, and stagnation of economic growth.11
Noncompetes are a problem for which antitrust law would seem to be the solution. As courts have recognized, noncompetes, as contractual restraints of trade, fall comfortably within the requirements of Section 1 of the Sherman Act. The noncompete is a contract in restraint of trade, as common law courts recognized, and it resembles exclusionary contracts on the product market side that courts have frequently struck down on antitrust grounds. But a survey of the cases reveals only a handful of antitrust challenges, virtually none of them successful.12
The lack of success can be attributed to a number of factors. Courts that have evaluated noncompetes under the antitrust laws agree that the rule of reason — rather than the per se approach — applies. The rule of reason is said to apply because noncompetes are ancillary to employment contracts rather than “naked” restraints, and they are vertical relationships to boot. More important, the long common law tradition recognizes that noncompetes, if narrow, can advance legitimate economic interests. Thus, courts have rejected arguments that noncompetes are per se illegal. But once the rule of reason is triggered, the obstacles to proof in typical cases where a single employee seeks to escape a noncompete are insurmountable. It is unlikely that a single noncompete could have a marketwide impact on wages and output, as the rule of reason requires. The employer’s normal defense — that it needs the noncompete to protect its investment in assets or training — will often strike a court as just common sense.13
Employees might conceivably overcome these obstacles by bringing class actions. But here the problem seems to be that employees rarely know the terms of the employment contracts of their colleagues. If class action lawyers do not know which workers are subject to noncompetes, they cannot propose a class. This is a major distinction between antitrust litigation in labor markets and antitrust litigation in product markets, where contractual terms are either public or easily accessible through purchase of the goods and services offered by suspected monopolists. And even if the lawyers can identify the workers who are subject to the noncompetes, the different relationships of each worker with the employer — seniority, role in the organization, pay level, etc. — may defeat class certification. Because the ubiquitous role of noncompetes in employment relationships has become common knowledge only in recent years, it may never have occurred to lawyers that class-action claims against noncompetes were possible.14
In reaction to the Jimmy John’s scandal and the academic research about noncompetes, several states have passed laws that ban noncompetes for low-income workers and restrict them in other ways. Some commentators have argued that California’s ban on all employment noncompetes should be extended to the entire country. While these laws and proposals are understandable, the first approach is insufficient and the second is overbroad. Even if noncompetes on average suppress wages, tailored noncompetes in appropriate circumstances may benefit workers — and the economy at large — by enabling employers to make investments that result in higher wages as well as lower prices. Moreover, a ban on noncompetes only for low-income workers does not stop employers from imposing overbroad noncompetes on the rest of the workforce.
A better approach would be to strengthen the antitrust laws. The antitrust laws already ban exclusionary terms with anticompetitive effects, and so it does not require new legislation to apply antitrust principles to noncompetes. The way forward is clear. Courts already treat contract terms as presumptively illegal under the antitrust laws (either under the “per se” or “quick look” approach) when experience teaches that they are normally used for anticompetitive purposes. On the product market side, the most common per se violation is a price-fixing or output-fixing agreement. On the labor market side, the analogous violations are wage-fixing and no-poaching agreements. A no-poaching agreement involves two employers who agree not to hire away each other’s employees. A noncompete is very similar. Although a noncompete binds the employer and employee, rather than two employers, it has a greater anticompetitive effect than a no-poaching agreement since the noncompete prohibits the employee from working for any of the employer’s competitors while a no-poaching agreement prohibits employees from going to work with only one of the employer’s competitors. Moreover, because most employees have little bargaining power, an employer can more easily negotiate or impose favorable noncompete agreements with employees than no-poach agreements with competitors. Thus, if no-poaching agreements are presumptively illegal, noncompetes should be as well.
It might be argued in response that employees can protect themselves from the harmful effects of noncompetes by demanding higher pay or refusing to take the job, whereas they are helpless when their employer enters a secret no-poaching agreement with another firm. But there are two problems with this argument. First, as noted earlier, employees are rarely able to protect themselves through bargaining. Second, even if they are, employees have no reason to take into account the effects on third-party workers who are deprived of work because of the contraction of the labor market. These third-party effects may be as bad as, or worse than, those of no-poach agreements.
Employers should be permitted to defend themselves against antitrust challenges to their noncompetes by showing that a noncompete has resulted (or will result) in higher wages. However, courts should be skeptical of such a defense when a noncompete binds low-skill or lowwage employees, and in other cases should demand high-quality empirical proof from the employer. Legal reform that shifts the burden of proof to the employer should be sufficient. A worker who seeks to escape a noncompete should prevail unless the employer can make a positive case that the noncompete does more good than harm.
Courts should be receptive to a stronger antitrust regime for noncompetes because it flows naturally from existing antitrust principles, and is based on strong empirical research. If they are not, then legislation may be necessary.
1 Kirkland & Ellis Distinguished Service Professor, University of Chicago Law School.
2 Matt Marx, The Firm Strikes Back: Non-Compete Agreements and the Mobility of Technical Professionals, 76 Am. Soc. Rev. 695(2009).
3 Evan Starr, J.J. Prescott & Norman Bishara, Noncompetes in the U.S. Labor Force, 16 (U. Mich. L. & Econ. Research Paper No. 18-013, 2019).
4 Dave Jamieson, Jimmy John’s Makes Low-Wage Workers Sign ‘Oppressive’ Noncompete Agreements, HuffPost (Oct. 13, 2014), http://www.huffpost.com/entry/jimmy-johnsnon-compete_n_5978180.
5 Ronald J. Gilson, The Legal Infrastructure of High Technology Industrial Districts: Silicon Valley, Route 128, and Covenants Not to Compete, 74 N.Y.U. L. Rev. 575, 578 (1999).
6 Michael Lipsitz & Evan Starr, Low-Wage Workers and the Enforceability of Non-Compete Agreements (2019), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3452240.
7 Matthew S. Johnson, Kurt Lavetti & Michael Lipsitz, The Labor Market Effects of Legal Restrictions on Worker Mobility (2019), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3455381.
8 Natarajan Balasubramanian et al., Locked In? The Enforceability of Covenants Not to Compete and the Careers of High-Tech Workers (U.S. Census Bureau Center for Econ. Studies Paper No. CES-WP-17-09, 2018), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2905782.
9 For a summary of the empirical literature, see Eric A. Posner, The Antitrust Challenge to Covenants Not to Compete in Employment Contracts, Antitrust L.J. (forthcoming), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3453433.
10 See, e.g. Jose Azar, Ioana Marinescu, Marshall Steinbaum & Bledi Taska, Concentration in US Labor Markets: Evidence From Online Vacancy Data, NBER Working Paper No. 24395 (2018); Efraim Benmelech, Nittai Bergman & Hyunseob Kim, Strong Employers and Weak Employees: How Does Employer Concentration Affect Wages?, NBER Working Paper No. 24307 (2018); Brad Hershbein, Claudia Macaluso & Chen Yeh, Concentration in U.S. Local Labor Markets: Evidence from Vacancy and Employment Data (unpub., 2018); Yue Qiu & Aaron Sojourner, Labor-Market Concentration and Labor Compensation (unpub., 2019).
11 See Suresh Naidu, Eric A. Posner & Glen Weyl, Antitrust Remedies for Labor Market Power, 132 Harv. L. Rev. 536 (2018).
12 See Posner, supra.
13 For some representative cases, see United States v. Topco Assocs., 405 U.S. 596, 606 (1972); Bradford v. New York Times Co., 501 F.2d 51, 59 & n.5 (2d Cir. 1974); Newburger, Loeb & Co. v. Gross, 563 F.2d 1057 (2d Cir. 1977); Lektro-Vend Corp. v. Vendo Co., 660 F.2d 255, 269 (7th Cir. 1981); Consultants & Designers, Inc. v. Butler Service Group, Inc., 720 F.2d 1553, 1560-1561 (11th Cir.1983).
14 Arbitration clauses also pose a significant barrier to litigation, including class litigation.
