Field experiment finds noncompete clauses cut pay and mobility without guarding secrets

One of the largest randomized field experiments ever conducted on restrictive employment contracts has delivered a result that is likely to reverberate through boardrooms, courtrooms and state legislatures: noncompete agreements sharply reduce worker mobility and earnings while providing no measurable protection for the trade secrets they are supposed to safeguard. The study, published in The Quarterly Journal of Economics by Evan Starr, Professor of Management and Organization at the University of Maryland’s Robert H. Smith School of Business, together with Bo Cowgill of the University of Toronto and Brandon Freiberg of INSEAD, evaluated more than 14,000 job offers across two finance firms. Its authors describe the work as the clearest cause-and-effect evidence to date on a contract clause that governs the careers of tens of millions of workers, and its findings arrive at a moment when regulators on both the state and federal level are actively reconsidering how much power employers should have over their employees’ future job options.

The research design is what sets the study apart from decades of observational work. For years, economists have documented correlations suggesting that workers bound by noncompetes change jobs less often and earn less over time, but such patterns are vulnerable to confounding: perhaps firms that use noncompetes are simply different in ways that also affect their workers’ trajectories. To break that impasse, the researchers partnered with two firms that hire thousands of freelance recruiters on short-term contracts and randomized the contract terms themselves. Workers were assigned to one of three conditions: a contract with no noncompete at all, a contract with a highly salient noncompete placed prominently on the first page, or a contract with a noncompete buried mid-document where a careless reader could easily miss it. Crucially, every contract in every condition included a standard nondisclosure agreement, which allowed the team to isolate precisely what the noncompete clause added on top of ordinary confidentiality protections.

The headline result concerns mobility and pay. Removing the noncompete increased movement between competing employers by 36 to 52 percent and raised workers’ total earnings by 12 to 17 percent. Those are large effects for a single clause of contract language, and they quantify, with experimental precision, the drag that post-employment restrictions impose on the labor market. Just as striking is what the experiment did not find. The researchers detected no evidence that noncompetes reduced the leakage of sensitive business information beyond what the accompanying nondisclosure agreement already achieved. In other words, the confidentiality tool that employers already use for free appeared to do the protective work, while the noncompete added only costs, depressing both the workers’ ability to move and their ability to command higher wages.

That cost extended to the hiring process itself. Workers who received contracts containing noncompetes did not negotiate better terms, and they were not more likely to accept an offer when randomly offered higher wages alongside the restriction. The long-standing employer argument that workers must be compensated for surrendering future job options found no support in the data. Instead, the study uncovered a behavioral pattern the authors characterize as inattention: many workers skimmed past the noncompete clause in seconds, and a substantial share remained unaware of the restriction until firms contacted them after employment had already begun. The contrast between the salient and buried versions of the clause allowed the team to measure how presentation shapes behavior, and the results are consistent with a model in which workers systematically fail to attend to fine-print provisions that will later constrain their most consequential career decisions.

The experiment also surfaced a phenomenon the authors describe as adverse selection. When firms were upfront about the noncompete, placing it prominently where applicants could not miss it, recruiting became measurably harder. But the workers who did accept such contracts were more likely to go on and violate the restriction anyway. This inversion of the intended screening logic suggests that prominent noncompetes do not filter out would-be rule-breakers; if anything, they may select for a pool of candidates who are less inclined to honor contractual constraints. Buried clauses, meanwhile, avoided the recruiting penalty only by exploiting inattention, meaning that the clause’s apparent effectiveness rested on workers not knowing what they had agreed to. As the authors summarize the core tension, at least in their setting, noncompetes reduced mobility and earnings without reducing the spread of firm secrets.

One of the most policy-relevant findings concerns enforceability. Even noncompetes that were legally unenforceable, such as those used in states like California where the clauses are void as a matter of law, reduced mobility just as much as enforceable ones. Starr’s explanation is blunt: the law doesn’t matter, because the presence of the clause alone changes behavior. Workers who believe they are restricted act as if they are, whether or not a court would uphold the restriction. This chilling effect means that statutory bans, by themselves, may not fully protect workers unless the clauses disappear from contracts altogether, and it helps explain why mobility patterns differ so sharply across jurisdictions even where formal legal remedies are rarely pursued.

The study’s treatment of trade secrets deserves particular attention, because measuring information flows is notoriously difficult. As Starr notes, secrets are generally secret, and therefore hard to observe, which is why nearly all prior evidence on this question has been indirect. The experimental setting solved the problem by design: the researchers could watch firsthand what information workers shared between competing employers and compare that behavior across contract conditions. Because every contract contained an NDA, any difference in secret sharing could be attributed to the noncompete alone, and the answer was that the noncompete added nothing. The nondisclosure agreement, a far narrower and less contested instrument, provided comparable protection of sensitive information without the collateral damage to worker mobility and earnings.

These findings land in the middle of an active national debate. The Federal Trade Commission’s proposed nationwide ban on noncompetes was recently halted in court, but states continue to advance their own restrictions, and federal agencies have signaled increased scrutiny of labor-market contracting practices more broadly. The study also speaks to specific policy design questions. Early-notice rules, which require employers to disclose noncompetes before an offer is accepted, did not raise wages in the experiment; workers exposed to both prominent and buried clauses still earned less than those with no clause at all. Transparency, it turns out, reduces the inattention advantage of buried clauses but does not eliminate the economic harm. For policymakers, the authors suggest, the calculus is straightforward: any evaluation of noncompetes should weigh their limited, in this setting undetectable, benefits against their clear costs to worker mobility and wage growth.

The implications extend beyond the clause itself. Starr argues that the methodology, occupying the role of the employer and randomizing contract conditions, offers a blueprint for evaluating a wide range of restrictive terms that have so far resisted causal study, including nondisclosure agreements, training repayment agreement provisions, non-solicitation clauses and no-poach agreements. By observing how workers respond to these provisions in real time, researchers can measure the causal effects of employment restrictions and other managerial practices with far greater precision than observational data allow. For employers, the practical message is that removing noncompetes substantially increases worker mobility without raising the risk of trade-secret leakage, and that firms should weigh the perceived benefits of the clause against its real costs: candidate dropoff and the possibility that workers willing to accept restrictive terms are also more likely to violate them.

For workers, the study carries a warning about the fine print. Job seekers frequently overlook noncompetes embedded in routine onboarding documents, receive no wage premium for accepting them, and may only discover the restriction when a future employer is scared off or a former firm intervenes. Starr advises workers to recognize that agreeing to post-employment restrictions may reduce their ability to seek raises or pursue better opportunities later on, and that even seemingly ordinary paperwork can depress long-term earnings. The gap between what the law says and how workers behave, revealed so clearly by the unenforceable-clause results, underscores that the mere presence of intimidating contract language is enough to alter careers. In an economy increasingly built on human capital and the mobility of skilled professionals, the evidence suggests that the noncompete is a tool whose costs are borne by workers and the labor market at large, while its promised benefit to firms appears, at least under experimental scrutiny, to be already delivered by simpler and less damaging means.

Subject of Research: Causal effects of noncompete agreements on worker mobility, earnings and trade-secret protection

Article Title: Noncompetes suppress worker mobility and earnings without protecting trade secrets

Article References: Noncompetes suppress worker mobility and earnings without protecting trade secrets. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: noncompete agreements, labor mobility, worker earnings, trade secrets, nondisclosure agreements, field experiment, behavioral economics, contract design, FTC regulation, Quarterly Journal of Economics, adverse selection, employment law