On January 9, Ryan Nunn, The Hamilton Project's Policy Director, spoke at the Federal Trade Commission about his research concerning non-compete contracts. Watch the full testimony here.
Thanks very much to the FTC for having me here today to talk about these fascinating and important economic policy issues. My name is Ryan Nunn—I am the policy director at the Hamilton Project and a fellow in Economic Studies at the Brookings Institution. I was previously an economist in the Office of Economic Policy at the U.S. Treasury Department, where I first began working on non-competes policy. What I am about to say will draw on that work as well as work that the Hamilton Project, Alan Krueger, Eric Posner, Matt Marx, and I have done in recent years.
What I would like to do now is give a quick description of what we know about non-competes and particularly what we know about the comprehensive evidence on non-compete contracts. What we now know is that non-competes are surprisingly common throughout the labor market. Previous work had focused on particular occupations and industries, for instance on CEOs or electrical engineers, rather than on the workforce as a whole (e.g., Marx 2011; Schwab and Thomas 2006; Garmaise 2011). Beginning with a survey by Evan Starr and his coauthors—who you will hear from later—followed up by some more recent surveys, we learned from workers themselves that nearly one fifth of workers have a non-compete on their current job, with substantially higher fractions reporting having had a non-compete in certain specific occupations.
So this finding itself has jumpstarted a lot of interest among economists and policymakers as to what accounts for the widespread use of non-competes and what might those labor market effects be.
The larger economic context for all of this is a growing understanding that labor markets do not often look like the classic competitive models that we once applied. They are actually characterized by a considerable amount of market power. Where non-competes would seem unimportant or perhaps irrelevant in a mostly classical labor market, we now can see them as instruments of (or at least symptoms of) employer market power.
And then we also have evidence that there is labor market concentration in many markets and that this concentration has effects on labor market outcomes (Qiu and Sojourner 2019; Rinz 2018; Azar et al. 2019; Hershbein et al. 2019). Moreover, we know that wages for median workers have grown extremely slowly over the last 40 or 50 years—much slower than productivity—and this is leading many to reappraise the labor market institutions that affect how wages are bargained for and that affect the relative position of workers and firms.
Empirical work demonstrating that employers are not price-takers in the labor market is also very useful here. This figure draws on the work of Webber (2015) and shows that, across sectors, the labor supply elasticity that firms face is often quite low; in other words, an employer could lower the wages it pays without losing all of its employees to other employers. This indicates that firms have a substantial degree of power in the labor market. Moreover, this labor market power is associated with lower wages.
It is also important to briefly describe the policy context in which we are learning about non-competes. The first thing to note, in my view, is the dramatic decline in private-sector union density. Unions bargained on behalf of many workers and helped set standards for the rest of the labor market. Over the last 50 years we have seen union membership decline from about 24 to just over 6 percent of the private sector workers (Nunn, O’Donnell, and Shambaugh 2019). This decline may have opened the door to contracts that unions would never have agreed to, or perhaps would have agreed to only under more favorable circumstances.
Another type of arrangement that’s often compared to non-competes are so-called no-poach agreements—between franchisees and franchisors—which restrict the ability of franchises to hire away workers from each other. Those agreements have also been shown to be quite common and are now under a great deal of legal scrutiny (Ashenfelter and Krueger 2018; Krueger and Posner 2018). Finally, other types of restrictive covenants, like non-solicitation agreements and IP assignment, are often used in conjunction with non-competes (unpublished work by Nunn and Starr).
Now I want to talk more specifically about non-competes and organize my discussion of the evidence on non-competes by talking about what we think they do in the labor market. Again, in the modern understanding of the labor market there is considerable scope for employers to exploit and even extend their market power. By “extend” I have in mind Evan Starr’s (slightly modified) phrase “the intertemporal conduit of market power,” whereby an employer exploits a moment (the beginning of an employment spell) during which it has market power to cement that power in later periods. On their face, non-compete agreements appear to be a way for employers to do so.
But non-competes could serve other purposes with more social value. I think we need to apply both economic theory and evidence to sort through these different accounts of non-competes. The potential explanations that emphasize social benefits are (most notably) about trade secrets and training. Then there are explanations that emphasize the employer benefits and the lack of social benefit. The first is the intertemporal conduit of market power that I mentioned. The second is about workers’ lack of understanding, either about non-competes themselves or the enforcement regimes that govern them. There is also a potential explanation in terms of screening for low-mobility employees, but very little evidence on this (except evidence from Starr, Prescott, and Bishara’s 2014 Non-compete Survey Project that workers with non-compete agreements (NCAs) don’t believe they are more likely to job hop than other workers). I should also say that “explanation” and “justification” are somewhat different; in what follows I will focus on the latter concept because I want to assess whether, to what extent, and how the social planner would want to implement NCAs and their enforcement regimes.
The trade secrets justification starts with the premise that trade secrets litigation is protracted, costly, and difficult for employers to win. Non-competes may represent a more effective (or at least lower-cost) way to prevent the loss of trade secrets than would a more narrowly targeted law that simply sanctions the exposure of trade secrets. The underlying idea is that it is necessary to prevent secrets from being divulged in order to induce an employer to share that information in the first place. The idea is that employers share the trade secrets with the worker, facilitating their joint production and contributing to social welfare, because they know that the information will not be divulged outside the firm.
A couple caveats to this justification: first, it is limited to those workers who plausibly have meaningful trade secrets and second, it depends on whether the employer has much of a choice about sharing trade secrets. For some types of production, there may be no other good option, in which case employers will not adjust their behavior very much whether or not they have assurances about the safety of their information. An important note that is not often emphasized here: client lists are not equivalent to trade secrets for this purpose. While trade secrets can have a positive-sum aspect, client lists often do not. Arguably there is less or no social interest in facilitating employer efforts to develop client lists.
But what do we actually know about non-competes and trade secrets? Workers who report having trade secrets are roughly 25 percentage points more likely to have a non-compete. But most workers with non-competes report not possessing trade secrets, so this isn’t the whole story (Starr, Prescott, and Bishara 2019). And several studies including that one have shown non-competes to be common among workers with low pay and/or educational attainment, for whom trade secrets are less likely to be relevant.
The training justification is a little more complicated. It starts with the widely held premise that worker training is undersupplied. On the worker side, liquidity constraints or lack of information about the quality of training will limit their willingness or ability to finance training by accepting initially lower wages. On the firm side, the expectation that workers will at some point leave, or bargain for higher wages after training, limits the employer’s willingness to pay for training. What a non-compete can do is give the employer more assurance that this won’t occur after their training investments. There is some evidence to support this: firm-sponsored training is more common in states with more-stringent NCA enforcement (Starr 2019; Jeffers 2019). For example, in states that allow courts to modify and enforce over-broad contracts, researchers see more employer-sponsored training. But there are two caveats to the training justification: first, many policies that reduce worker bargaining power should have this effect and are not socially beneficial on account of this effect (indeed, quite the opposite). Second, there are other contracts that can be used to protect training investments. For example, a worker could agree to a contract that requires repayment of some fraction of training costs in the event of an “early” departure from the firm. This could protect the employer investment without unnecessarily restricting workers.
Now I turn to the explanations that emphasize private benefits for employers—that is, how NCAs benefit firms at the expense of workers or the overall economy. First, non-competes can in principle cement an employer-friendly bargaining situation at the beginning of an employment spell. A worker who has just accepted an offer (possibly turning down others) or has not yet accepted, but has gone through time-consuming interviews, etc., is in a tough spot. She may just have gone through an unemployment spell, and we know that labor markets are not kind to those with longer unemployment durations (Kroft et al 2013) and job search is costly and uncertain. And more generally, workers just before and after job acceptance often have little leverage. NCAs can be imposed in a moment of worker weakness and used to maintain employer advantage.
We do not know as much as would be helpful here. We do know that non-competes are often presented to workers after the job offer was accepted or even on/after the first day of work (Marx 2011; Marx and Fleming 2012). The figure below, showing results from Marx (2011), demonstrates this for a sample of electrical engineers. Nearly half of workers report having signed their non-competes on or after the first day of work. But more evidence and theory in this area would be very useful.
Moreover, workers are not likely to be compensated for something they don’t understand is bad for them (or that they don’t know they signed). There is much worker confusion over whether and how NCAs are enforced, and workers in states that enforce less stringently—or not at all—are generally unaware that this is the case (Starr, Prescott, and Bishara 2019 and unpublished work by Prescott and Starr). Indeed, roughly as many workers report having NCAs in states that do not enforce them—states like California—as in states that do enforce them. Consequently, it is unsurprising that few workers report bargaining over their NCAs (Starr, Bishara, and Prescott 2019).
State enforcement regimes do differ widely. The figure below shows two important dimensions of state enforceability: whether or not non-competes can be enforced at all, and for those states that do enforce them, whether or not they allow for judicial modification of overbroad (or otherwise legally noncompliant) non-compete contracts.
The lack of NCA (or NCA enforcement) salience does not mean that NCAs are invisible to workers at economically important moments. A NCA can be non-salient from a worker’s perspective until an employer brings it to a worker’s attention, as they might do after the worker receives a competing offer, for example. At such a moment, the NCA (and any threats of employer enforcement) can have a chilling effect on the worker’s behavior. Note that this does not require that actual litigation ever ensue, but merely that workers adjust their behavior in response to the NCA.
Bringing all of this evidence together, what should we expect to see if NCAs tend to be mutually beneficial for workers and firms? When NCAs more common or more enforceable, we should expect a combination of more employer-sponsored worker training, more business investment, and higher wages.
What do we actually observe? There is limited evidence on all of these points, but what research we do have suggests that stringent non-compete enforcement is associated with somewhat more worker training (Jeffers 2019; Starr 2019). It may also be associated with more investment at incumbent firms (Jeffers 2019), but this is offset by diminished firm entry and startup performance (Ewens and Marx 2017; Jeffers 2019; Samila and Sorenson 2011).
On wages, some suggestive evidence is found in the fact that states with more stringent NCA enforcement on average have lower age-wage profiles (Treasury 2016). More rigorous evidence was generated by Oregon’s recent ban of NCAs for hourly workers. Lipsitz and Starr (2019) found that wages were higher after the ban, and other research has shown that less stringent enforcement is also associated with higher wages (Johnson, Lavetti, and Lipsitz 2019).
But assessing the social welfare consequences of NCAs and NCA enforcement is not just about the employer-employee relationship. It is also necessary to consider spillovers to the rest of the labor market as well as broader consequences for entrepreneurship, innovation, and labor market dynamism. A comprehensive summary of evidence on these outcomes is not possible here, but NCAs and/or relatively stringent NCA enforcement appear to have negative spillovers for entrepreneurship (Ewens and Marx 2017; Starr, Balasubramanian, and Sakakibara 2017), innovation (Belenzon and Schankerman 2013), and the mobility of workers who do not have NCAs (Starr, Frake, and Agarwal 2019). These implications should be assessed in the broader context of declining labor market and business dynamism, which has put downward pressure on wage and productivity growth (Shambaugh, Nunn, and Liu 2018).
If one believes that NCAs and their enforcement are on the margin producing net harms, what might be done to change policy? The following are some of the key options that are currently being contemplated by state and federal policymakers, in roughly descending order of ambition. (See also Starr 2019 for an overlapping set of policy options.)
- Ban NCAs altogether and/or render them unenforceable
- Make enforcement less employer-friendly by
- Removing the possibility of ex post judicial modification of contracts (Marx 2018)
- Tightening the scope and shortening the duration of NCAs (White House 2016)
- Requiring that workers receive legal consideration (beyond continued employment) for signing NCAs (White House 2016)
- Requiring so-called garden leave during a period of NCA enforcement (Treasury 2016)
- Enhanced transparency and notification (Marx 2018; Treasury 2016)
More evidence will help to inform policymakers’ choice of a subset of these options. But we already have ample justification for moving the needle substantially in the direction of a more worker-friendly set of non-compete policies. For low-wage workers in particular, the case for allowing non-competes is weak, and efforts to eliminate those NCAs are supported by the best research currently available.