Examining the effects of a 2008 ban on noncompete agreements for low-wage workers in Oregon, a recent paper finds that the ban increased average hourly workers’ wages by up to 6 percent.
In a moment of deep political polarization and weak income growth for the lowest-earning workers, bipartisan agreement is coalescing around one way to raise wages: ban employment provisions which prohibit low-wage workers from joining or starting competing firms (often referred to as noncompete agreements). On the heels of media reports showing that sandwich makers and camp counselors earning minimum wage had signed noncompetes, US Senators Elizabeth Warren, Todd Young, Chris Murphy, Marco Rubio, and many others have proposed such bans, believing that these agreements actively and directly prevent low-wage workers from climbing the income ladder.
Simple economic theory, however, suggests that workers do not voluntarily agree to contracts that make them worse off. In fact, if noncompetes allow firms to fearlessly invest in training or client acquisition (by ensuring that workers will not leave for a competing firm), workers may actually benefit from the increased productivity that comes with these investments.
So, do noncompetes hurt or help low-wage workers? And what are the implications for economic theory and the recent bills that seek to ban low-wage noncompete agreements?
We address these questions in a recent paper by examining the effects of a 2008 ban on noncompete agreements for low-wage workers in Oregon—a bill that was recently used as a model for Massachusetts’ revised noncompete law. Prior to 2008 Oregon enforced noncompete agreements in a manner similar to most other states: if the contract was deemed reasonable in its coverage and the employer could demonstrate a protectable business interest, then the contract would be likely to be enforced. The new law, which took effect in January 2008, rendered all new noncompete agreements unenforceable for select types workers: hourly-paid workers, workers earning less than the median income for a household of four, and workers in certain occupations.
Using Oregon’s 2008 ban on low-wage noncompetes as a natural experiment, we compare how hourly-paid workers’ outcomes in Oregon change following the ban, compared to a variety of control states whose noncompete laws did not change. We find that relative to control states, the noncompete ban increased average hourly workers’ wages by 2-3 percent, rising to 6 percent five years after the ban was implemented.
Moreover, because only 14 percent of the hourly-paid population signs noncompetes—and we cannot distinguish which ones in our data—our estimates of the earnings effect is an average across those who did and did not sign noncompetes. If we scale our estimates by the proportion of workers who actually signed noncompetes, the effect on noncompete signers may be as large as 14-21 percent, though spillover effects in the labor market likely mean that the true effect was somewhat lower.
We also examine whether these positive wage effects are confined to certain sectors or groups. We find that while the positive wage effects are strongest in jobs where noncompetes are most likely to be found, there is evidence of positive wage effects across industries, occupations, age groups, and education levels. We also find that while men benefitted from the ban on low-wage non-compete agreements, women had disproportionately large wage gains. The positive wage effects are also present throughout the wage distribution, rather than being concentrated among the highest earners.
We bolster these baseline results by examining how the ban impacted job-to-job mobility. We estimate that hourly workers in Oregon experienced a 12-18 percent increase in mobility due to the ban. This finding is important in its own right, as high potential job mobility may lead to better labor market matches. It also strengthens the credibility of our main estimates by illustrating the three-way link between the noncompete ban, worker mobility, and the earnings effect.
Recognizing that firms could respond in potentially unexpected ways to the noncompete ban, we examine whether firms cut worker hours, laid workers off, or moved hourly-paid workers to salaried compensation structures (to avoid the ban entirely). We find little evidence of the first two responses. Hours worked remained unchanged after the ban, such that the increases in hourly wages translated directly into increases in actual earnings (as opposed to being offset by fewer hours worked).
Furthermore, rather than firms laying off workers, we find that workers in Oregon were marginally less likely to be unemployed following the ban (by about 1 percentage point), and there was no decline in labor force participation. This suggests that taking away employers’ ability to use noncompete agreements caused them to, if anything, hire more workers.
We do find evidence that after the ban, relative to control states, workers are 4-5 percent more likely to be paid a salary (versus an hourly wage), and this effect is especially pronounced for those earning just above the low-wage threshold determined by the law. This side-effect allowed employers to use noncompetes for workers who may have otherwise been exempt from the ban (decreasing its effectiveness), but may have improved job quality if salaried jobs are more stable, come with better benefits, or have more predictable work hours.
Additionally, since hourly workers at the top end of the distribution are more likely to become salaried after the ban, our hourly wage estimates are likely to be attenuated, meaning that the true impact on hourly wages may be greater than we report.
| “Overall, our study of the Oregon noncompete law suggests that for the 40-plus states that do not currently have these laws in place, low-wage workers will benefit from banning noncompetes.” |
One key challenge with this empirical design is that anything that changes hourly wages in Oregon (but not in control states) after the noncompete ban is going to confound the estimates. We do show our results are robust to controlling for some obvious concerns (measures of the recession and the minimum wage), but we cannot rule out every possible Oregon-specific shock. Accordingly, our approach is to construct an array of results that are theoretically consistent but would be difficult for an alternative story to fully explain. For example, we highlight that any confounding event is unlikely to explain why the wage effects are stronger in jobs most likely to use noncompetes, why job-to-job mobility also rises, and why firms begin moving workers to salaried jobs right around the low-wage threshold established by the law.
Taken together, our results address a broader question regarding the degree of competition in low-wage labor markets. Traditionally, economists have viewed low-wage labor markets as largely competitive: workers without highly specialized skills should be able to easily find new jobs if their wages are arbitrarily cut. However, recent work has shown that low-wage labor markets are plagued by employer monopsony power, contributing to inefficiencies and poor worker outcomes. In the case of noncompetes, restrictions on worker mobility grant firms monopsony power by allowing them to ignore wage offers from competitors in the future. But, if the labor market is competitive, then workers should either be unaffected if they can easily circumvent the restrictions by moving to another industry, or they should be compensated for lower future wage growth such that they are better off under a noncompete.
The fact that wages rise after noncompetes are banned suggests that, prior to the ban, workers were not being appropriately compensated for signing noncompetes, implying that some aspect of perfect labor market competition failed. We cannot identify precisely where the economic theory gets it wrong, but we can highlight some likely avenues. Low-wage workers have little bargaining power, rarely negotiate over noncompetes (4.5 percent of the time), and may be unable to afford an attorney to help them interpret a contract before signing it, or to fight a contract if it goes to court. They are also less likely to possess the highly specialized information typically used to justify the use of a noncompete (and thus the efficiency gains that go with that argument).
Taken together, this means that low-wage workers may be unable to fight noncompetes that may have been unenforceable in the first place.
Overall, our study of the Oregon noncompete law suggests that for the 40-plus states that do not currently have these laws in place, low-wage workers will benefit from banning noncompetes.
In addition to this baseline policy implication, however, two additional points bear noting. First, prior work has found that noncompetes are found even where they are unenforceable and that they can still impact workers. We cannot verify that the actual use of noncompetes went down following the Oregon noncompete ban, but we think this is likely because Oregon paired the reform with a “garden leave” provision that requires firms to pay workers during the prohibition period. This “garden leave” provision made it costly for firms to use noncompetes, and thus likely contributed to a reduction in their use in Oregon among low-wage workers after 2008.
A second important provision of the 2008 Oregon law was to require “early notification” of the noncompete (2 weeks before the commencement of employment). Such notice provisions ensure that workers have a clear idea of what they are agreeing to before they start the job, preventing firms from foisting noncompetes on workers at unexpected moments when they have no choice but to agree. Prior literature has indeed found that firms regularly delay asking workers to sign noncompetes, and that such delays are associated with worse outcomes for workers. As it relates to our estimates, if such delays were common in Oregon before 2008, then the comparison our work reflects is between noncompetes that are implemented on a delayed basis and a ban on noncompetes.
Although we cannot tease out the independent effects of each of these policies (because they were simultaneously adopted), our work does provide policymakers with an opportunity to improve the plight of the lowest-earning workers: Free them from constraints that limit their ability to take better jobs in the industry. Just as importantly, with policymakers on the right and on the left side of the political spectrum proposing low-wage noncompete bans in recent years (indeed, yesterday), as well as consideration by the Federal Trade Commission, this may be a rare opportunity for policymakers to cross the aisle to help the lowest earning workers finally get a raise.
The ProMarket blog is dedicated to discussing how competition tends to be subverted by special interests. The posts represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty. For more information, please visit ProMarket Blog Policy.
