Many industries – especially research heavy fields like high tech and biotech – are riven with “non-compete agreements”, where you sign a contract when you’re hired banning work for a competitor firm in a similar area for some amount of time after you quit. These are controversial. Indeed, California comes close to banning them altogether (more on this in a future post). This seems like a great deal for the employer. If your employee develops any industry-specific human capital, you ensure that they won’t use the knowledge against you by working for a competitor. This immediately raises another question, then: why doesn’t every employer use NCAs?
Theory comes to the rescue, in the form of an extension of Holmstrom’s (may he win his deserved Nobel!) career concerns. Think of your income as having three components: wages, bonuses and “implicit payments”. Wages are set salaries. Bonuses are payments conditional on some verifiable goal. Implicit payments are increases in your total expected lifetime wages and bonuses as a result of some action. For instance, a firm pays a young trainee very little, less than her outside option, but promises that the deal is worth it because the trainee position will develop human capital in such a way that future job offers will be at a high wage.
Kräkel and Sliwka show how this can lead firms to avoid NCAs. Consider a model where you work with a firm on an invention. With probability p, you invent it. With probability q, if you don’t invent, the firm invents anyway. Exactly who came up with invention is not contractible (a common problem with team effort!). The agent’s effort is costly. After the invention is made, if the agent stays with the firm, a big surplus results. Alternatively, if the agent was responsible for the invention, she may receive an outside offer. The initial contract is a triple: wage, bonus conditional on the invention being made, and potentially a non-compete clause which precludes the agent from taking any future outside offer. Total surplus is assumed to be highest when the firm and the agent stay together.
Intuitively, the agent knows if they work hard and are responsible for the invention, and there is an NCA in place, then after the invention is made, the firm is going to claim that it was not the agent’s doing. Hence the incentives for individual effort are fairly low-power. If, however, there is no non-compete clause, the agent gets an outside offer if she is, in fact, the inventor. For the firm to keep the agent, then, it must give her an extra bonus. This outside offer implicitly incentivizes the agent to work harder than she would if only the bonus and wage were available. Further, it is less susceptible to free-riding on the agent’s part, since the outside option only comes about if it was the agent, and not the firm, who made the invention: the bonus when an NCA is in effect has such ability to distinguish, hence the incentive to free-ride is stronger. The model shows this intuition is correct. A non-compete clause will only be imposed when there is a very-high probability that the agent can get an outside offer, and when the relative value to the firm of keeping the agent after an invention is small. Indeed, there is a large range of parameters where I don’t pay any bonuses and I don’t provide an non-compete. The “implicit bonus” that I will have to match the agent’s outside option is enough to encourage effort. A short extension shows that if I can use clawbacks or pay workers set amounts not to compete with me, I prefer to do that always over using an NCA since the incentives can be tuned even finer.
This isn’t to say that noncompete agreements aren’t worrying from a social policy perspective; there are other reasons we should be concerned about them, as I’ll discuss sometime soon. But this result shows again the value of thinking through problems theoretically. In general, the answer to “why doesn’t X screw over Y?” turns out to be “because in equilibrium, it is not in X’s interest to do so”!