A Study of 14,000 Job Offers Found Noncompetes Cut Pay and Mobility Without Protecting Trade Secrets

Most noncompetes are issued on faith. A template goes into the offer packet, the legal team assumes it protects client relationships and proprietary know-how, and HR assumes it slows attrition. Almost nobody tests either assumption, because the clause is everywhere and nobody can see the counterfactual. A new study did something unusual: it randomly took the clause away and watched what happened to pay, job changes, and leaked secrets.

What the Experiment Actually Tested

The paper, "Clause and Effect," by Evan Starr of the University of Maryland, Bo Cowgill of the University of Toronto, and Brandon Freiberg of INSEAD, is published in the Quarterly Journal of Economics. According to the University of Maryland's summary, the researchers worked with two finance firms that hired freelance recruiters on short-term contracts. Each worker was randomly assigned one of three contracts: no noncompete, a prominent noncompete on the first page, or a noncompete buried mid-document. Every version included a nondisclosure agreement. The team then tracked how workers reviewed the contracts, whether they accepted, and how they responded to outside opportunities that could breach either the noncompete or the NDA.

The results run against nearly every argument employers make for the clause. Removing the noncompete increased mobility between competing employers by 36% to 52% and raised total earnings by 12% to 17%. It did not reduce the spread of confidential information beyond what the NDA already achieved. In the authors' summary, noncompetes in their setting reduced mobility and earnings without reducing the spread of firm secrets. Workers also did not receive higher pay in exchange for accepting a noncompete, which undercuts the idea that the restriction is a bargained trade.

The Clause Works Even When Nobody Reads It

The most uncomfortable finding for HR is about awareness. HR Dive's coverage of the study notes that many employees did not read their noncompetes, and a substantial share did not learn of the restriction until the firm contacted them after they had left. When the researchers compared California, where noncompetes have long been void, worker mobility looked the same as in states where they were enforceable. "The presence of the clause alone changes behavior," Starr said. In other words, a noncompete can chill job changes whether or not it would survive in court.

Transparency did not rescue the clause either. When firms made the noncompete prominent, recruiting became harder, and the workers who accepted it were more likely to violate it. That is a direct cost to talent acquisition: a restriction designed to protect the business makes it harder to fill the role in the first place, then fails to hold the people it was meant to hold.

Regulators Are Reading the Dockets

The federal rulemaking route is closed, but scrutiny is not. Foley & Lardner's 2026 overview recounts that a Texas court blocked the FTC's nationwide ban in August 2024 and the agency withdrew its appeal in September 2025. The states kept moving. Foley lists California, Minnesota, North Dakota, Oklahoma, and Wyoming as complete-ban states, and Washington's HB 1155 will make noncompetes unenforceable on or after June 30, 2027, applying retroactively. Virginia's ban already reaches every employee entitled to overtime under the FLSA, regardless of income, while Florida went the other way with its CHOICE Act for high earners.

Meanwhile, the FTC has shifted to individual cases. On May 8, 2026, Chairman Andrew Ferguson sent a public warning letter to Mortgage Connect L.P. after the company sued a former employee and her new employer in Pennsylvania state court. The public docket showed the company required all of its employees to sign noncompetes, that the employee had received no specialized training, and that the firm already protected its information through non-solicitation and non-disclosure agreements. The letter urged a comprehensive review of its employment contracts. The lesson for HR is blunt: the act of enforcing a blanket noncompete can be the thing that puts it in front of a regulator.

A Playbook for Restrictive Covenants

• Inventory every covenant you have issued. Pull the signed agreements by role, state, and pay level, because the study shows many employees do not know what they signed, and HR often does not either without a records search.
• Default to the NDA. The experiment found the noncompete added no protection for confidential information beyond the NDA, so reserve noncompetes for the narrow roles where counsel can name a specific interest an NDA and non-solicit cannot cover.
• Strip clauses from overtime-eligible and junior roles. Virginia already bans them for FLSA non-exempt staff, and the FTC letter singled out an employee with no specialized training, so these are the hardest clauses to defend.
• Map your Washington and multi-state exposure. Washington's ban applies retroactively from June 30, 2027, so identify current Washington employees and contractors now rather than discovering the gap at an exit interview.
• Check with counsel before you sue. Treat every enforcement filing as a public document that a regulator may read, and confirm the clause is tailored before relying on it in court.
• Retain people with reasons to stay. If the clause suppresses earnings rather than earning loyalty, measure regretted turnover and invest in pay, growth, and manager quality, which are the levers that actually keep strong performers.

Noncompetes survived for decades because they felt cheap and safe. This study puts a price on them: lower pay, harder recruiting, and no added protection for the secrets they are supposed to guard. For HR, the question is no longer whether a noncompete is legal in a given state. It is whether the clause is doing any work at all, or simply sitting in the file until the day it becomes evidence.