What people have signed, and what it did to them
The OECD ran surveys of firms and employees in fifteen countries - Belgium, Canada, France, Germany, Italy, Japan, Korea, Mexico, New Zealand, Poland, Portugal, Spain, Sweden, Switzerland and the United Kingdom - and asked, in effect, what is written into people's contracts about the job they take next.
The most common answer is a non-disclosure agreement, covering around half of private-sector employees. Then the non-compete clause, binding between a fifth and a third of the workforce depending on the country. Then, at a tenth to a fifth, clauses that forbid approaching former clients and colleagues, or that require repayment of a bonus or of training costs on departure.
Two findings about non-competes are worth separating from the rest, because they are what makes the number surprising rather than merely large. The first: the clauses have spread well beyond knowledge-intensive occupations, covering many low-skilled employees who have no access to confidential information - and the report finds little variation in their use by a firm's training, growth or innovation activity. They are, in its words, applied indiscriminately and rarely a bargained outcome. The second: their duration and scope are often broader than the national regulatory framework would support, partly because firms are unaware of what the law requires.
| Measure | Share of employees | What it restricts, or what it prevented |
|---|---|---|
| Non-disclosure agreement | ~50% | What you may say. The most common restraint, and the one least linked to lost productivity |
| Non-compete clause | 20-33% | Where you may work. A range across the fifteen countries surveyed, not a single figure |
| Non-solicitation or repayment clause | 10-20% | Approaching former clients and colleagues; or repaying a bonus or training costs on departure |
| Stopped from changing jobs | 5% | Reported by employees as having actually happened, not as a term they signed |
| Stopped from starting a business | 3% | The same question asked of founding a firm rather than joining one |
| Blocked by a deal between employers | ~17% | A no-poaching agreement between firms - not in the worker's contract, and usually illegal |
The measured five per cent
Most of the debate about non-competes is about what they might do. The survey asked what they did. Five per cent of all private-sector employees report having been stopped from moving jobs because of a non-compete clause. Three per cent report having been stopped from starting a business.
Those are small percentages of a very large denominator, and they are the floor rather than the ceiling of the effect: they count only the people who tried and were blocked. The report's term for what happens to everyone else is a chilling effect - and it names the mechanism precisely. Because a worker cannot easily tell whether a clause is legitimate or enforceable, the clause works whether or not it would survive a court. In countries where non-competes are more tightly regulated the economic bite is more modest, but still negative, which is what a chilling effect looks like in data.
And there is a second restraint that never appears in anyone's contract. Almost half of the firms surveyed report being aware of no-poaching or wage-fixing practices within their industry - agreements between employers not to hire each other's staff, which antitrust authorities generally treat as illegal. One-sixth of employees report having been prevented from joining another firm because of one, and a quarter report having heard of the practice. A worker blocked this way has signed nothing at all.
What it costs, and what the authors decline to claim
The chapter's economic estimate is the part that most needs its hedges carried with it. A 10-percentage-point increase in how common non-compete clauses are within an industry is associated with a 1.9% lower level of aggregate business-sector labour productivity.
The decomposition is more interesting than the total. Only 0.3 points of it comes from the channel most people would name first - weaker labour reallocation, the extent to which workers flow towards more productive firms, which the report estimates could decline by one-sixth. The other 1.6 points comes from slower knowledge diffusion: the rate at which lagging firms catch up to the global frontier falls by an estimated 5%, worth 1.8% of productivity growth for the median firm. The damage is less about the individual who cannot move than about the ideas that stop moving with them.
The authors are careful in a way that summaries of them often are not. They write that this analysis is "not identifying a purely causal relationship", and the chapter describes its own evidence as "preliminary". Both hedges are theirs and both belong in any honest report of the number. What the chapter does show, consistently, is a direction: where these clauses are more common, reallocation and convergence are weaker.
One contrast within the same data supports the reading. Non-disclosure agreements - the most widespread restraint of the three - have a much weaker link to productivity than non-competes. That is what you would expect if the mechanism is about mobility rather than secrecy, and it suggests, as the report puts it, that there may be less distortive alternatives for protecting a firm's legitimate interests.
| Channel | Associated change in aggregate productivity |
|---|---|
| Total | -1.9% |
| Slower knowledge diffusion | -1.6% |
| Weaker labour reallocation | -0.3% |
The constraint a matching engine cannot see
We score jobs against people on skills, experience, location, salary, contract type and work style. A non-compete clause is invisible to every one of those dimensions. It lives in a document we never see, it binds a worker to a restriction that may or may not be enforceable, and it can veto a match that is otherwise perfect on all six.
This is not a flaw we can engineer away, and pretending otherwise would be the dishonest move. What a matching system can do is stop treating the labour market as though the only obstacles are the ones it measures. If between a fifth and a third of the workforce carries a clause about where they may work next, then a match score is a statement about fit, and never a statement about availability.
It is also a reason to be careful with a metric this desk has criticised before: time-to-hire, application-to-offer ratios, and every funnel number that treats a candidate who does not proceed as a candidate who was not interested. Some of them were not free to proceed. That is a different failure, and it is invisible in the same data that reports them as a drop-off.
We publish how our own matching is tested, including where it is weakest, at our bias audit. If a number here is wrong, that page is where the correction should start.