Do Noncompete Agreements Help or Hurt Workers’ Wage Growth and Job Stability?

Potter, Kurmann, and Hobijn (2026) examine how noncompete agreements affect workers’ wages and job tenure and whether these effects differ by education. They use National Longitudinal Survey of Youth 1997 data, comparing workers who change jobs and noncompete status with similar job changers whose status remains unchanged. They find sharply different wage effects. Among workers without a four-year college degree, signing a noncompete produces significantly slower wage growth over the following four years. College-educated workers instead experience significantly higher wage growth. Noncompete jobs initially pay about 15% more for less-educated workers and 22% more for college-educated workers, although these differences largely reflect worker sorting.

Why This Article Was Selected for The Policy Scientist

Noncompete agreements raise broad questions about how employment contracts shape wages, worker mobility, bargaining power, and firms’ incentives to invest in employees. The topic is especially timely given continuing policy attention to restrictions on worker mobility. Potter, Kurmann, and Hobijn make an important contribution by examining actual noncompete incidence rather than relying solely on state enforceability laws, building on a substantial literature to which researchers in this area, particularly Starr, have contributed extensively. The NLSY97 is a high-quality longitudinal dataset that permits workers to be followed across jobs. Their local-projections difference-in-differences design is a strong causal-inference approach and substantially more persuasive than conventional multivariate regression. Generalizability beyond the United States remains uncertain because employment law and labor-market institutions differ across jurisdictions. Publication in AEA Papers and Proceedings, a respected American Economic Association publication, further increases the study’s visibility within economics.

Full Citation and Link to Article

Potter, T., Kurmann, A., & Hobijn, B. (2026). The impact of noncompetes on wages and job tenure: New evidence from NLSY data. AEA Papers and Proceedings, 116, 273–277. https://www.aeaweb.org/articles?id=10.1257%2Fpandp.20261078

Central Research Question

Noncompete agreements restrict employees from taking certain jobs with competing employers after leaving their current positions. Potter, Kurmann, and Hobijn examine how these agreements affect two central labor-market outcomes: wage growth and job tenure. Their analysis focuses on whether workers experience different wage trajectories and employment duration after moving into jobs covered by noncompete agreements. A particularly important component of the research question is whether these effects differ according to workers’ educational attainment.

The study addresses two competing explanations for noncompete agreements. Under an investment explanation, firms may use noncompetes to protect investments in workers, including firm-specific knowledge and training. If firms can prevent competitors from capturing the benefits of these investments, they may have stronger incentives to invest in employees. Workers could consequently experience greater productivity and wage growth. Alternatively, noncompetes may restrict workers’ outside employment opportunities. Reduced mobility could weaken workers’ bargaining positions and suppress wage growth. The authors investigate which explanation is more consistent with observed outcomes and whether the answer differs between workers with and without four-year college degrees.

Previous Literature

Previous research has established that noncompete agreements are common in the American labor market, including among workers who are not highly paid. Starr, Prescott, and Bishara (2021), for example, document the prevalence of noncompete agreements throughout the U.S. labor force. Their research is particularly relevant because it demonstrates that noncompetes remain common even in states where their legal enforceability is weak. Consequently, studying state laws alone may not capture the full effects of these agreements.

Other research helps explain why formally weak enforcement may nevertheless affect workers. Marx (2011) examines noncompete agreements and the mobility of technical professionals, while Prescott, Bishara, and Starr (2016) study workers’ understanding of these contracts. This literature suggests that employees may comply with noncompetes even when their enforceability is questionable because workers may be uncertain about their legal rights or concerned about potential litigation.

Several studies investigate the relationship between enforceability and labor-market outcomes. Starr (2019) examines training, wages, and enforceability, while Lipsitz and Starr (2022) specifically study low-wage workers. Both provide evidence that stronger enforceability can negatively affect wages among workers with less education or lower earnings. Johnson, Lavetti, and Lipsitz (2025), however, find that negative wage effects associated with restrictions on worker mobility are concentrated among college-educated workers. These differing results suggest that the consequences of noncompetes may vary across labor markets and worker characteristics.

Rothstein and Starr (2022) are especially important to the present analysis because they previously used the National Longitudinal Survey of Youth 1997 to examine noncompetes, bargaining, and wages. They found positive associations between noncompetes and wages, but those relationships were highly sensitive to statistical controls, indicating substantial worker selection into noncompete jobs. The present study builds on that work by using workers’ longitudinal employment histories rather than relying primarily on cross-sectional comparisons.

Data

The analysis uses the National Longitudinal Survey of Youth 1997, or NLSY97, a longitudinal survey of 8,984 American men and women born between 1980 and 1984. Respondents were initially interviewed annually from 1997 through 2011 and subsequently every two years. The longitudinal structure allows researchers to connect jobs across survey waves and reconstruct individual employment histories.

Beginning in 2017, respondents were asked whether their jobs were subject to noncompete agreements. The authors combine this information with detailed measures of wages, job tenure, educational attainment, and other employment characteristics. They divide workers into two groups: low-education workers, defined as those without a four-year college degree, and high-education workers, defined as those possessing at least a four-year degree.

The analysis concentrates on job-to-job transitions, defined as situations in which one job ends and another begins within three months. The sample excludes self-employment, nontraditional employment, military jobs, government employment, and unpaid work. Workers must generally work at least 30 hours per week, and hourly wages must fall between $2 and $250. These restrictions create a sample designed to permit meaningful comparisons of wage trajectories surrounding job changes.

Noncompete coverage is substantial in both educational groups. Approximately 12–13 percent of jobs held by workers without four-year degrees involve noncompetes, compared with approximately 17 percent among college-educated workers.

Methods

The primary methodological challenge is selection. Workers who sign noncompetes may differ systematically from workers who do not. Similarly, jobs requiring noncompetes may offer different compensation, training, occupations, or career opportunities. A simple comparison of workers with and without noncompetes therefore cannot establish the causal effect of these agreements.

The authors address this problem using workers’ employment histories and a “clean controls” local-projections difference-in-differences methodology developed by Dube, Girardi, Jordà, and Taylor (2025). The approach is designed for circumstances in which treatment occurs at different times and treatment effects may differ across individuals and time horizons. It also avoids problems that can arise from conventional two-way fixed-effects difference-in-differences models under staggered treatment.

Identification comes from comparing workers who begin with the same noncompete status but experience different changes when switching jobs. For example, a worker moving from a job without a noncompete into one with a noncompete is compared with workers who move between two jobs without noncompetes. Similarly, workers leaving noncompete jobs are compared with workers who begin with noncompetes and remain covered after changing jobs.

The analysis estimates wage changes at the job transition and two and four years afterward. The authors also examine the two years preceding the transition as a placebo test for differential preexisting wage trends. They conduct robustness analyses that distinguish entry into and exit from noncompetes and account for changes in occupation and industry.

Findings/Size Effects

Simple descriptive comparisons initially suggest that noncompete jobs pay considerably more. Among workers without four-year degrees, average log wages are 2.99 in noncompete jobs compared with 2.84 in jobs without noncompetes, corresponding to an approximately 15 percent difference in the reported descriptive comparison. Among college-educated workers, the corresponding values are 3.60 and 3.38, representing an approximately 22 percent difference. These gaps, however, should not be interpreted as causal effects. Once worker characteristics and job sorting are considered, much of the apparent wage advantage disappears.

The causal analysis produces a strikingly different pattern across educational groups. For workers without four-year college degrees, there is no statistically significant compensating wage increase when they enter jobs with noncompetes. Over the subsequent four years, however, workers signing noncompetes experience significantly slower cumulative wage growth than comparable workers entering unrestricted jobs. The graphical estimates indicate a four-year cumulative log-wage effect of roughly −0.2, although the article does not report a precise numerical coefficient in the text. The result is consistent with noncompetes reducing outside employment opportunities and weakening workers’ subsequent earnings growth.

College-educated workers experience the opposite pattern. They also receive no clearly significant immediate wage premium when signing noncompetes, but their subsequent wage trajectory becomes more favorable. By four years after the transition, the estimated cumulative wage effect is positive and statistically significant, with the figure suggesting a magnitude of roughly 0.2 log points. The evidence is therefore more consistent with noncompetes accompanying productive firm-worker relationships or human-capital investments for highly educated workers.

The authors find considerably weaker evidence regarding job tenure. Noncompetes do not produce statistically conclusive changes in the probability that workers subsequently leave their jobs for either educational group. The estimates are imprecise, however, meaning that economically meaningful effects cannot be ruled out.

Robustness analyses using average hourly compensation, including signing bonuses and other nonbase earnings, generate similar patterns. Controlling for occupation and industry changes also leaves the central results largely intact, although statistical uncertainty increases in some specifications.

Conclusion

The study demonstrates that the labor-market consequences of noncompete agreements cannot be adequately characterized by a single average effect. Educational attainment substantially changes the relationship between noncompetes and workers’ subsequent wage trajectories. Workers without four-year degrees experience slower wage growth after entering noncompete jobs, while college-educated workers experience faster wage growth. Neither group receives a clearly significant immediate wage premium for accepting the restriction.

These findings help distinguish between competing explanations for noncompete agreements. For less-educated workers, the evidence provides little support for the proposition that mobility restrictions generate wage benefits through greater firm investment. Instead, the wage trajectory is consistent with diminished outside opportunities and bargaining power. Among college-educated workers, the positive wage trajectory is more consistent with noncompetes operating alongside firm-specific investment, although the empirical results do not directly establish that mechanism.

The analysis also demonstrates why noncompete incidence and legal enforceability should be treated as related but distinct subjects. Workers may respond to contractual restrictions regardless of whether courts would ultimately enforce them. By following individual workers across job transitions and applying a modern difference-in-differences estimator, the study provides evidence about what happens when workers actually move into or out of jobs governed by noncompete agreements. The resulting educational differences suggest that the consequences of these contracts depend substantially on workers’ positions within the labor market rather than operating uniformly across the workforce.