A Critical Evaluation of The FTC’s Empirical Evidence That Prohibiting Non-Compete Clauses Will Increase Earnings
Introduction
On January 19 the Federal Trade Commission (FTC) proposed the Non-Compete Clause Rule (the Rule) which would prohibit businesses and workers from using a non-compete clause (NCC) as part of their employment relationship. An NCC is a voluntary agreement between an employer and employee that prevents the employee from working at competing businesses in a certain geographic area and time period after leaving the employer. The Rule claims that recent research has shown that “the use of non-compete clauses by employers has negatively affected competition in labor markets, resulting in reduced wages for workers across the labor force – including workers not bound by non-compete clauses.”[1]
In a recent New York Times opinion essay, Lina M. Khan, chair of the FTC, wrote “a body of empirical research shows” that NCCs “inflict major harm across the economy.” She claims a “staggering finding” of a recent study is that NCCs “systematically drive down wages, even for workers who aren’t bound by one.” In contrast, former FTC Commissioner Christine S. Wilson, in her Dissenting Statement concerning the rule states that the academic studies about the economic effects of NCCs are “scant, contain mixed results, and provide insufficient support for the scope of the proposed rule.” She also states that an ”objective review” of whether NCCs negatively affect competitive conditions reveals “a mixed bag,” and an interpretation of this evidence is that “the scientific evidence is still muddled as to who is helped and who is harmed” by NCCs.
This article critically evaluates the empirical evidence in the studies cited in the Rule that led the FTC to conclude that NCCs reduce wages across the labor force. Commissioner Wilson is correct; the evidence of a causal relationship between NCCs and lower earnings in these studies is, at best, inconclusive. Predictions of the impact of a nationwide NCC ban based on these studies would be inappropriate because they either examine the effect of small differences in NCC enforceability between states or over time, or they study two statewide NCC bans that are limited to certain jobs. The cited studies cannot be extrapolated to reliably predict the effects of the Rule on earnings. In addition, the empirical finding of a correlation between lower NCC enforceability and slightly higher earnings is also consistent with reduced incentives to hire workers who require training and may not be the result of a decline in monopsony power.[2] Because NCCs increase incentives for employers to provide training, the Rule could have adverse consequences for workers who lack the skills that employers demand. Finally, few of the studies cited in the Rule, that purport to show the impact of NCCs on earnings, include information about which employees in their data work under an NCC.
Economic Consequences of Non-Compete Clauses
The Rule argues that NCCs restrict worker mobility and limit the competition for employees who work under an NCC. The Rule also claims there is empirical evidence that all employees in a state with strong NCC enforceability earn less, on average, even if they do not work under an NCC.[3] The research upon which this claim is based is inconclusive because it may conflate differences in NCC enforceability and other policy differences between states. For example, there are many reasons why the pay distribution differs between California and Florida other than the fact that Florida has the strongest enforceability of NCCs and California does not enforce NCCs.
The Rule, however, does not focus on the fact that skilled workers value mobility after leaving their current employer so that firms that require NCCs must pay more than firms that do not, all else equal, i.e. a “compensating differential.” Businesses that benefit the most from NCCs, for example, to protect trade secrets and intellectual property, pay enough additional compensation so that their employees agree to an NCC and limit their future options. Most of the studies cited in the Rule do not compare the pay of employees working under an NCC to employees with similar skills who do not. One study cited in the Rule finds that CEOs are paid 18.4% more in total compensation, all else equal, if their contract includes an NCC and that the pay differential is larger in states where NCC enforceability is stronger.[4]
Competition among businesses for skilled workers is how the free market limits the use of NCCs that are not mutually beneficial to employers and employees. An employer that requires an NCC but does not pay a compensating differential will find it more difficult to hire skilled workers. In the current economic environment, where it is challenging to attract qualified workers to fill job vacancies,[5] requiring an NCC without additional compensation would make it even more difficult to recruit qualified workers.
The situation is different for workers who lack the skills that employers demand. Employers are reluctant to invest in on-the-job training for workers who are free to leave to a competing business.[6] NCCs can include terms that reduce worker mobility for a period of time, thereby increasing incentives for firms to invest in training because firms are more likely to retain trained workers and receive a return on their investment.[7] Workers can share in training costs through lower initial pay, and part of their return for agreeing to an NCC is employer-provided training. A study cited in the Rule finds that greater NCC enforceability encourages training; “the incidence of training is 14% higher in an average enforceability state relative to a non-enforcing state. The positive relationship between enforceability and training is strongest when the training content is meant to upgrade skills and when it is firm-sponsored.”[8] NCC enforceability increases incentives for businesses to hire less experienced workers and provide them with firm-sponsored training.
Economic theory predicts an ambiguous relationship between NCCs and earnings. Workers are expected to bargain for higher compensation to agree to an NCC, but after the NCC is in place the worker has less bargaining power. A limitation of the empirical studies that attempt to quantify the impact of NCCs on earnings is that NCCs also change the incentives to hire less experienced workers. As a result, the earnings differences in these studies may reflect differences in workers’ relevant experience and skills rather than differences in bargaining power.
For example, a study cited in the Rule examines Hawaii’s ban on NCCs in the tech sector in 2015.[9] The study finds that the average earnings of new hires in the tech sector increased by about 4% relative to other benchmarks (tech jobs in other states and other jobs in Hawaii) after the NCC ban. The Rule appears to interpret this as evidence that a newly hired worker with the same skills and experience was paid 4% more after the NCC ban than they would have received prior to the ban. However, the study is based on the average salary for new hires and does not allow a comparison of the relevant experience, education, or qualifications of newly hired employees in the tech sector before and after the NCC ban. Some job applicants may lack experience or a degree in software engineering, but can be trained by the employer and grow into the job. If NCCs are banned, fewer tech firms will be willing to hire less experienced workers and invest in their training, because they are less likely to receive a return on their investment. If worker skills and relevant experience are not accounted for in an empirical model, it can appear that an NCC ban (or lower NCC enforceability) increases bargaining power when the real effect may instead be to reduce incentives for firms to hire and train less experienced, lower-wage workers.
Non-Compete Clauses Can Benefit Lower-Wage Workers Despite the FTC’s Claims
The Rule suggests that NCCs for lower-wage workers may be more “exploitative and coercive” than NCCs for higher-wage workers with no empirical support for this claim.[10] Despite this suggestion, NCCs may benefit lower-wage workers by increasing incentives for employer-provided training. However, whether NCCs benefit or “exploit” lower-wage workers cannot be determined by the data because, other than for CEOs, the studies that purport to show the impact of NCCs on earnings cannot identify which employees work under an NCC.[11] Most of the employees, other than CEOs, in the studies cited in the Rule, do not work under an NCC because the survey with “the broadest and likely the most representative coverage of the U.S. labor force” finds that only about 18% of employees in the U.S. work under an NCC.[12]
Surveys also show substantial differences between employees who are more likely to work under an NCC and those who are not. More educated and higher-paid employees are more likely to work under an NCC. One study cited in the Rule finds that workers in a “high NCC use” occupation group are more educated and earn 60.8% more, on average, than workers in a “low NCC use” occupation group.[13] NCCs for low-wage occupations are relatively less common and are for different purposes than NCCs for executives and tech workers.
Brief Review of the Empirical Evidence on the Enforceability of Non-Compete Clauses and Earnings
The Rule cites six studies that purport to examine the direct effects of differences in the enforceability of NCCs on earnings.[14] Because NCC enforceability is not a simple yes/no outcome in most states, these studies use enforceability indices for each state and year that have been constructed by researchers and are based on both legislative changes and court decisions regarding key aspects of NCCs.
Limitations of these six studies include:
- The only study relying on changes in enforceability across all states and workers examines the impact of very small changes in NCC enforceability.
- The only nationwide study examining the impact of differences in NCC enforceability across states only measures the change in the earnings differential between “high NCC use” and “low NCC use” occupation groups.
- Two studies examine NCC bans by state, but the tech workers in Hawaii and hourly workers in Oregon that are affected by these bans are not representative of the employers and workers affected by the Rule.
- Two studies examine the impact of differences in NCC enforceability on CEO compensation, but CEOs are not representative of workers affected by the Rule.
- Interpreting the correlations between earnings and NCC enforceability in these studies is speculative because the statistical models in these studies explain very little pay variation so that earnings differences may be due to skill differences rather than differences in bargaining power.
The cited study that the FTC says has the “broadest coverage” is an unpublished (not peer-reviewed) paper that is co-authored by an FTC employee.[15] It uses annual national data from 1991 to 2014 to study the impact of “within-state” changes in NCC enforceability. The study finds a one standard deviation decrease in enforceability is associated with a pay increase of three-tenths of one percent. A hypothetical change from the fifth strictest to the fifth most lax state for NCC enforceability is equivalent to a change of more than 11 standard deviations, and a nationwide NCC ban would mandate even larger changes in many states. Both hypothetical changes are far outside what is observed in the study. Predicting the Rule’s expected impact on earnings from this study would be inappropriate because it requires unreasonable extrapolations. For comparison, if a pharmaceutical study found a small benefit from a small dose of a proposed drug, it would be inappropriate to conjecture that taking 11 times the dose will provide 11 times the benefit.
Even if this study could be used to extrapolate an overall average effect of NCC enforceability on earnings, the study itself finds that NCC enforceability “has little to no effect on earnings for non-college-educated workers.” An implication of the study is that a reduction in the enforceability of NCCs would increase (slightly) the earnings differential between college-educated and non-college-educated workers. Despite this empirical finding, Commissioner Khan states in her New York Times essay that NCCs are “routinely invoked” against “fast-food workers” and “manual laborers,” implicitly suggesting that NCCs, rather than a lack of marketable skills, are the reason for low wages in these occupations.
The other cited study that uses national data for all occupations examines the period from 1996 to 2008 and relies on between-state differences in NCC enforceability because “few states changed their policies between 1991 and 2009.” [16] The study examines the impact of NCC enforceability on the pay differential between workers in high-NCC-use and low-NCC-use occupation groups. The study finds that a one standard deviation decline in NCC enforceability across states leads to a one percentage point increase in the pay differential between high NCC use and low NCC use groups. This change in the pay differential is relatively small given the large pay difference between occupation groups and suggests that implementation of the Rule might increase earnings inequality.[17]
The Empirical Studies Cited in the Rule Cannot Identify Workers with Similar Skills and Qualifications
In addition to lacking information about whether a worker is employed under an NCC, the data for both the “broadest” national study and the Oregon NCC ban include no information about workers’ company tenure, length of time in current position, or relevant prior work experience. Because the data contain so little information about workers’ qualifications and skills, the pay models only explain about 36% of pay variation in the national study and about 22% of pay variation in the Oregon study (using a similar government survey). The study of between state differences in NCC enforceability and earnings uses survey data with information about company tenure, but its model explains only about 39% of pay variation. Because of the low explanatory power of the pay models in these studies, workers in the same occupation and industry, with the same educational attainment and age, are characterized by large pay variation. The pay variation is likely due to substantial differences in productivity, skills, and qualifications among workers that the pay model considers observationally equivalent. Consequently, higher earnings for a worker of the same age and educational attainment, within the same industry and occupation as another worker, may simply be because one worker is relatively more qualified, based on relevant prior experience and other factors, than the other worker.
The inability to compare workers with similar skills and qualifications across jurisdictions (or time periods) with differing NCC enforceability causes the interpretation of the correlation between NCC enforceability and earnings in these studies to be speculative. For example, the study of the Oregon NCC ban for hourly workers in 2008 finds 2.3% higher wages for hourly workers in Oregon after the ban relative to the pay of hourly workers in other states. But in this study, 78% of the pay variation is unexplained by the model. Earnings may be slightly higher after the NCC ban in part because businesses hire slightly more experienced and qualified workers after it becomes more difficult to obtain a return on investments in employee training.[18]
Are The Cited Studies Valid Natural Experiments?
In her Dissenting Statement, former Commissioner Wilson poses a question for possible commenters:
The NPRM describes papers that exploit natural experiments to estimate the effects of enforcing non-compete clauses. While this approach ensures that the estimates are internally valid, it reflects the causal effects of non-compete agreements only in the contexts within which they are estimated. What should the Commission consider to understand whether and when these estimates are externally valid?
The Rule refers to studies that examine the impact of changes in state laws as “natural experiments” and asserts that this type of study “allows for the inference of causal effects since the likelihood that other variables are driving the outcomes is minimal.” This is an oversimplification. While changes in state laws can “allow for” causal inference in certain situations, the estimated impact of the change in state law on earnings may be the result of changes in economic conditions and policies other than changes to NCC enforceability. There is more than a “minimal” likelihood that a correlation between a change in NCC enforceability and earnings is due to factors not accounted for in the model making the results of the study invalid even internally. That is why the authors of the cited academic studies conduct sensitivity analyses to attempt to convince readers that their results are not due to factors other than a change in NCC enforceability.
The “natural experiments” cited by the Rule are unlikely to be externally valid for the impact of a nationwide NCC ban. The contexts for the cited natural experiments are much different than the mandated change in the Rule. The nationwide and broadest natural experiment only examines the impact of very small changes in NCC enforceability and is therefore unreliable for the enormous changes mandated by a nationwide NCC ban. Two studies of state-specific NCC bans examine new employment agreements for tech workers in Hawaii and hourly workers in Oregon at the onset of the Great Recession. These groups are not representative of the employers and workers affected by the Rule. This is relevant because many studies find a differential impact of NCC enforceability across education and occupation groups and it would be inappropriate to extrapolate from these narrow bans on NCCs to the sweeping change in the Rule. The other “natural experiment” cited by the Rule examines changes in NCC enforceability and CEO compensation and is certainly not representative of workers affected by the Rule.
Conclusion
The empirical studies cited in the Rule are not useful for predicting the impact on earnings of a nationwide NCC ban. The study that receives the most attention from the FTC estimates the impact of small changes in NCC enforceability within the same state over time, which is much different than a nationwide NCC ban. Other studies cited by the FTC include examinations of statewide NCC bans that apply to selected occupations in Hawaii and Oregon. Even when these studies find a statistically significant correlation between NCC enforceability and earnings, it may have little to do with differences in bargaining power because a reduction in NCC enforceability discourages investment in employer-provided training. Because the pay models in the cited studies explain little variation in pay, the studies inevitably compare workers with different skills and qualifications before and after a change in NCC enforceability. The risk of a nationwide NCC ban is that it may harm workers who lack the skills and qualifications to obtain high-wage jobs by reducing firms’ incentives to hire and train inexperienced workers.
[1] P.3482 of the Non-Compete Clause Rule.
[2] Monopsony power is the market power enjoyed by an employer that does not compete for the services of a worker in a free market.
[3] This article focuses on the direct effects of NCC enforceability on earnings.
[4] Kini, O., Williams, R. and Yin, S. “CEO noncompete agreements, job risk, and compensation.” The Review of Financial Studies, 34(10), 4701-4744. The authors reviewed individual CEO contracts to determine who was employed under an NCC. Unfortunately, CEOs are not representative of the workers impacted by the Rule.
[5] Over the past two years, there has been an average of 9.4 million private sector job openings, about 2.3 times as many job openings, on average, as during the previous two decades.
[6] Many of the costs of training are not measured or recorded, because they involve the use of supervisor and co-worker time and effort to train new employees and are opportunity costs.
[7] In general, labor economists expect the costs and returns of training to be shared by employer and employee. Employers pay some of the training costs by paying more to workers than they generate in revenue while being trained.
[8] Starr, E. (2019). Consider this: Training, wages, and the enforceability of covenants not to compete. ILR Review, 72(4), 783-1044. In addition, the study: Johnson, M.S., & Lipsitz, M. (2022), “Why are low-wage workers signing noncompete agreements?” The Journal of Human Resources, 689-700, finds that hair salons that use NCCs invest more in training, on average.
[9] Balasubramanian, B., et. al., (2022), “Locked In? The Enforceability of Covenants Not to Compete and the Careers of High-Tech Workers.” The Journal of Human Resources, S349-S396.
[10] PP. 3512-3513 of the Rule.
[11] One exception is the CEO study referenced in footnote 3.
[12] P. 3485 of the Rule.
[13] Starr, E. (2019), referenced in footnote 7.
[14] The Rule puts more weight on studies of NCC enforceability and earnings and “minimal” weight on studies that correlate earnings with NCC use because the NCC use is a choice and does not provide evidence of a causal relationship.
[15] Johnson, M.S., Lavetti, K., & Lipsitz, M. (2020). “The labor market effects of legal restrictions on worker mobility.”
[16] This statement seems to undermine the validity of the study referenced in footnote 11, cited by the Rule, that relies completely on small within-state changes in NCC enforceability.
[17] For example, while workers in high NCC-use occupations earn 60.8% more than workers in low NCC-use occupations, in a state with NCC enforceability one standard deviation below the national average, the pay differential is 61.8%.
[18] As noted earlier, the study of the NCC ban for tech workers in Hawaii did not control for age, educational attainment, or any individual worker characteristics.
Experts
Partner