The Effects of Noncompetes on Workers and Employers

We propose to work with a firm to hire professional HR workers, who will be tasked with reviewing and screening resumes. For HR workers who respond to our job ad, our experimental design involves three dimensions of randomization. In the job offer we will randomly assign the use of a noncompete, the salience of the noncompete, and wages. We describe these more in the hidden fields in order to keep the interventions private from potential subjects.

2023-05-23

2023-11-01

The outcomes we aim to measure from the first randomization are below:
--Job Acceptance and Wages: We then measure who accepts the job offer. The randomization permits us to infer the effects of noncompete agreements on the composition of hired workers, the job acceptance rate, compensating differentials for noncompetes, and the length of employer search.
-- Performance. For workers who accept the job, we will send recruiting work and instructions to review sixteen resumes or job applications. We will measure who complies with these instructions for all workers as a measure of their quality. Exploiting the salience intervention allows us to address selection into second-stage performance.

Other outcomes we aim to measure are below:
-Other Employment Experiences and Earnings: During the prohibited noncompete period (6 months), we will also gather information from the platform on other work that the worker has completed (i.e., number of other employers) and their cumulative earnings. We will supplement this information with any new information on their LinkedIn profile related to employment.
-Testing Willingness to Violate the Noncompete: During the prohibited noncompete period (6 months), we will assume the role of a poaching firm and reach out to the workers we hired previously. We will make them a job offer to measure if (a) those bound by noncompetes are less willing to accept our offer, and (b) whether they require higher wages to violate the provision.

For HR workers who respond to the job ad, our experimental design involves three dimensions of randomization related to the job offer. We describe these more in the hidden fields in order to keep the interventions private from potential subjects.

We will randomize using a computer script and a random number generator.

The randomization will occur at the individual level.

No

We expect to hire approximately 2080 workers. Anticipating a 20-25% acceptance rate, this is approximately 8-10k workers in the initial outreach.

It depends on the stage of the survey. At the outset, we hope to hire 2080 individuals, out of approximately 10k job offers.

In the second stage, we expect to have:
240 individuals in each of the 2 control conditions without noncompetes.
400 individuals in each of the 4 treatment conditions with noncompetes.

The power calculations are complicated by the fact that the first stage outcomes and second stage outcomes are linked. In the first stage we examine who accepts the job offer, and in the second stage we examine the effect of noncompetes among those who have accepted the job offer. If noncompetes make it exceedingly unlikely to accept a job offer, our treatment alone will affect our power in the second stage. Accordingly, our power calculations were done via backward induction from the second stage to the first stage. We do not have precise information on several key variables to perform a power analysis for all of them. But based on prior experience and research we expect a 25% job acceptance rate in the first stage in the control group and a 15% job acceptance rate in the treatment groups. For the second stage, we expect a 70% job acceptance rate for the control groups and a 35% acceptance rate for the treatment groups. The upper envelope of our power calculations imply that we need to hire approximately 400 individuals in each of our four treatment cells, and 240 individuals in the control cells to have power of at least 80% in both stages.