Publishing behavior changed clearly and quickly: ranges are now normal in job postings well beyond the states that require them, and salary history questions have largely disappeared from application forms. The effect on pay levels themselves is far less settled, because most of the confident figures circulating come from vendor surveys of their own customers rather than from published research. Where sources genuinely disagree, that disagreement is the honest finding rather than something to average away.
Two questions that get answered as if they were one
Ask whether pay transparency laws worked and you will get an answer about pay. Ask what actually changed and the strongest evidence is about publishing. Those are different questions with different amounts of support behind them, and conflating the two is how confident claims about compression, convergence and raises get repeated without anybody checking where they came from.
This matters more than an academic distinction, because you are likely to be making a decision on the strength of one of these claims. If ranges being visible is what changed, that helps you find information. If pay levels moved, that changes what you should expect to be offered. The first is well supported. The second is not, at least not yet.
What is clearly true
Ranges appear in job adverts now in places that never legislated for them. The mechanism behind that is not mysterious: employers hiring across several states apply the strictest rule uniformly, because running one compliant posting is cheaper than running twelve and tracking which applies where. Remote roles accelerate it further, since a national advert can touch any covered state.
The second clear change is the disappearance of the salary history question. Bans on asking spread across a large number of states and cities, and application forms were rebuilt to remove the field, including by employers with no obligation to remove it. That one is easy to verify by applying for anything: the question that used to appear on nearly every form now appears on very few.
Both of these are observable without a study, which is exactly what makes them the solid part of the picture. You can open twenty postings in a state with no requirement and count how many carry a range. The answer today is very different from the answer five years ago, and you do not have to trust anybody’s methodology to see it. That is a rarity in this subject and worth leaning on.
What is claimed and less well supported
The claims needing more care are the ones about pay itself. That transparency narrows gaps within organizations, that it compresses the distribution, that it lowers average offers, that it raises them. All four are asserted regularly, sometimes in the same week and occasionally by the same outlet, and several of them cannot be true together. When a field produces confident claims pointing in opposite directions, the useful response is to look at what each one measured rather than to pick the one you prefer.
The evidence problem is worth understanding rather than just noting. A great many confident figures come from compensation vendors surveying their own customers, which is a sample of organizations that already run structured pay programs and already care enough to buy the software. That population is not the labor market, and a finding about it does not transfer.
Academic work exists and is considerably more careful. It also tends to be narrower — a single state, a single sector, a defined window of time — and that narrowness is simultaneously what makes it credible and what limits how far you can generalize from it. A well-designed study of one state’s technology sector is exactly that. The honest overall position is that publishing behavior changed fast and measurably, while pay effects are still being argued about by people with access to far better data than any article can offer.
The second-order effects that are visible
Some changes are easy to see without needing a study at all, and they are arguably more consequential than the direct effects everybody argues about. They are also the ones least likely to appear in a headline figure, because they happen inside organizations rather than in published data. Two of them are worth naming.
Internal pressure is the clearest. Publishing a range for a role people already do exposes where existing staff sit inside it, and an employee who discovers they are below the range advertised for their own job has a very specific conversation available. Employers anticipating that have had to review internal pay before advertising, which is real work that happened because of these laws and is invisible in any measure of posted pay.
The second is that ranges have become a competitive signal rather than merely a disclosure. Once most postings carry one, a posting without a range stands out, and a range that is obviously unserious stands out more. That reputational dynamic operates regardless of enforcement, which is a large part of why compliance improved faster than penalties alone would explain.
One number worth reporting honestly
The clearest illustration of the evidence problem is the count of transparency states itself, which is covered properly elsewhere in this section. Reputable sources put it anywhere between fourteen and eighteen. Every one of those numbers is defensible, because they are counting different things — posting requirements, on-request rights, at-offer disclosure, and in some compilations local ordinances as well. Nobody is wrong; they are answering different questions and not saying so.
If a question that simple produces a four-way disagreement among people with no reason to be careless, it is worth adjusting your expectations for the harder questions. What transparency did to pay involves measuring a moving target across different sectors and jurisdictions using data nobody has complete access to. A single confident percentage answering that is describing a specific study rather than a settled fact, and the honest version names the study.
What it means for you
Treat visibility as the thing you actually gained. You can now find comparable ranges for most roles in most markets, you can benchmark before an interview rather than after an offer, and you can do it without asking anybody for a favor. That is a genuine change in your position and it does not depend on any contested claim about pay levels.
Do not assume the laws have moved the number you will be offered. Nothing in the evidence supports walking into a negotiation expecting a published range to do the work on your behalf. The more useful posture is that transparency has given you better information for a conversation you still have to have yourself, which is a real gain and a smaller one than the headlines imply.
And be skeptical of any confident percentage about what transparency did to pay, particularly one without a named source and a stated population. Ask who was surveyed and how many. The answer is frequently a vendor’s own customer base, and knowing that does not make the finding worthless — it makes it a finding about a specific group of employers rather than about the market.
Why the disagreement is worth publishing
It would be easy to pick whichever figure supported a tidy conclusion and print that instead. The reason not to is straightforward: somebody who acts on a false certainty is worse off than somebody who knows the question is open. And the open question here has a practical shape rather than an abstract one — use the visibility, do not bank on the pay effect. That is an instruction you can actually follow.
That also gives you a test worth applying well beyond this subject. When two credible sources disagree about a number, the disagreement usually reflects different populations or different measures rather than one of them being careless. Working out which is a much faster route to understanding than deciding who to believe. It also tends to leave you knowing something about the underlying thing rather than something about the sources.
This is general information about the state of the evidence rather than legal advice or a forecast. The research is still arriving, and the sensible expectation is that the picture in three years will be clearer than it is today.
Common questions
What did pay transparency laws clearly change?
Publishing behavior. Ranges now appear in postings well beyond the states requiring them, and the salary history question has largely disappeared from application forms.
Why did ranges spread beyond covered states?
Employers hiring across several states apply the strictest rule uniformly, because one compliant posting is cheaper than twelve. Remote roles accelerate it, since a national advert can touch any covered state.
Did the laws change pay levels?
Much less clear. Claims about compression, gap narrowing, lower offers and higher offers all circulate, and some contradict each other.
Why are the pay figures unreliable?
Many come from compensation vendors surveying their own customers — organizations that already run structured pay programs. That population is not the labor market, so the finding does not transfer.
What about academic research?
More careful and usually narrower — one state, one sector, a defined window. That is what makes it credible and what limits how far you can generalize.
What are the second-order effects?
Internal pressure, mainly. Publishing a range exposes where existing staff sit inside it, so employers have had to review internal pay before advertising — real work that shows up in no measure of posted pay.
What should I actually do differently?
Use the visibility. Benchmark before an interview rather than after an offer. Do not walk into a negotiation expecting the range to do the work for you.
How should I judge a statistic about this?
Ask who was surveyed and how many. A vendor's customer base is a finding about a specific group of employers rather than about the market.