On the first day of 2021, Colorado started requiring employers to put expected salary information into their job postings. Before the law, about 35 percent of postings in the state carried a number. Within weeks it was around half, and over the following months it reached nearly 70 percent.1

It never got close to everyone, and the way it fell short is worth a moment. Employers did not sort themselves into rule-followers and refusers. The same firm would put a salary on one posting and leave it off the next, picking where to disclose.1 Whatever else that is, it is evidence that people understood exactly what the number was doing.

The raise reached people the law was not written for

David Arnold, Simon Quach and Bledi Taska studied what happened, using job postings collected by Lightcast, a commercial labor-market data firm, from more than 45,000 websites, then checking the result against self-reported earnings from Glassdoor and against administrative earnings records in the Quarterly Census of Employment and Wages.1

Start with the employers who were already publishing salaries before Colorado made anyone do it. The law asked nothing of them. Their posted salaries went up by about 3.6 percent anyway.1

Then the earnings people actually took home. Glassdoor showed a rise of roughly 1.3 percent and the administrative records about 1.4 percent across Colorado.1 Splitting that by whether someone was newly hired or had been there all along, the authors found both groups gained about the same.1 Which means a substantial share of the money went to people who had not applied for anything, had not interviewed anywhere, and had not threatened to leave. They were paid more because the going rate had become something they could see.

Nothing about the work changed and nothing about the market changed. Visibility changed, and the price moved to follow it.

The authors put a number on the relationship. For every ten percentage point rise in how transparent an occupation became, posted salaries in it rose about 1.35 percent.1 The effect scales with how much daylight there is. That is not a compliance story, because the firms already complying moved too. It is a competition story, and competition does not check who a rule was addressed to.

A disclosure cannot be aimed

Every senior leader has approved a piece of information going out with a specific audience in mind. A band published for one job family. A promotion criterion written down for one department. A budget line shared with one team because they asked. The intended reader is easy to picture, and the mental model is that the information lands on them and stops.

Colorado is a large, clean demonstration that it does not stop. The number was published for candidates, and it re-priced employees. It was mandated for firms that had been hiding their ranges, and it moved firms that never hid anything. The people who benefited most were, in a sense, bystanders.

I have sat through enough compensation rounds to know the sentence that usually ends them, which is some version of "that is what the market pays." The Colorado result puts an uncomfortable question underneath it. If publishing the range raises what people earn without the work changing, then the number before publication was not simply the market rate. It was the market rate as understood by someone who could not check.

Not every document does this

Worth being careful here, because a written change does not always produce a real one. I wrote a couple of weeks ago about employers removing degree requirements from job postings without changing who actually got hired. Same artifact, opposite result.

The difference is what the reader can do with what they learn. Deleting a requirement tells a candidate that a door might be open, and then a hiring manager decides in private whether it is. Publishing a salary hands someone a figure they can hold against their own, and act on, without needing anyone's permission. One changes the description of a decision. The other changes the information held by the person on the other side of it.

So the question to ask about any disclosure you are weighing is not whether it is accurate or whether the intended audience can handle it. It is what a person could do differently on Monday having read it, and which people, beyond the ones you had in mind, will be reading over their shoulder.

Most organizations have at least one number that would change behavior the moment it became visible, and a settled institutional preference for not finding out which one it is. What is yours?