According to one estimate, 18 percent of U.S. workers currently work under a non-compete agreement and 37 percent have been subject to such an agreement at some point during their career.

Job-hopping is, unfortunately, on the decline in the United States. While we don’t fully understand the reasons for the decline yet, we should start thinking about economic and policy changes that may help workers switch more readily between jobs. Which leads us to non-compete agreements.

Non-compete agreements are contracts between employers and employees that determine how long a worker has to wait after leaving a firm before he or she can go work for a competitor. Some of the logic behind non-competes makes sense, as employers might want to protect trade secrets or the agreements may give employers an incentive to invest more in their workers. But there’s mounting evidence that non-competes have expanded too far and pose a problem for workers and the U.S. economy.

A new report from the U.S. Department of the Treasury looks at the extent and impact of non-compete agreements in the labor market. For a labor market institution just now gaining significant attention from researchers and policymakers, non-competes are fairly common. According to one estimate, 18 percent of U.S. workers currently work under a non-compete and 37 percent have been subject to such an agreement at some point during their career. You might think that these agreements are mostly for highly educated or high-income workers, but 15 percent of workers without college degrees and 14 percent of workers making less than $40,000 a year are working under non-competes. As the infamous example of Jimmy John’s shows, it’s unlikely that these workers have trade secrets they’ll spill to their new employers.

So why are employers are using these kinds of agreements? One possibility is that employers want to hold on to the workers they invest in, so non-competes give employers the security to invest in the human capital of their workers. As the Treasury report notes, there is some evidence for this effect as the probability of firm-sponsored training increases in states where non-competes are enforced more, but only by 2.4 percent for high litigation occupations relative to low ones. Another possible reason is that workers aren’t aware of these agreements when they take a job and then the agreement is used as a means to suppress workers’ bargaining power and therefore their wages.

One way to sort out this question is to look at how wage growth across states that enforce non-compete agreements differs from other states. First, states that have stronger enforcement of non-competes have lower worker job mobility. Furthermore, stricter enforcement is associated with lower initial wages and lower wage growth over the source of one’s career. This second result is particularly troubling for the job-training story, because we’d expect wage growth over a career to be higher if non-compete enforcement increased training relative to areas where the agreements aren’t strictly enforced. Instead, the opposite happens, strengthening the argument that the agreements are about shifting bargaining power to employers.

But given this information, how should states reform these agreements moving forward? The Treasury report suggests reforms such as states specifying the exact extent to which the agreements can be enforced or making firms give “consideration” to workers in the form of payout. Or perhaps the agreements should be banned outright, as some commentators such as Jordan Weismann of Slate have argued. Regardless, whether we scale them back or end them outright, it’s clear the right direction is backward when it comes to non-competes.