Salary History Bans Worked. That Tells You How Pay Really Gets Set.

When governments stopped employers from asking what you used to earn, pay went up. Job-changers in ban jurisdictions earned 3.9% more than comparable workers elsewhere, with women up 6.2% and non-white workers up 5.8%. That single question was setting your price. Remove it, and the price moved.

Here's the part nobody tells you. The number in your head that you call "my salary" is not a measure of your worth. It's a measure of what one past employer decided to pay you, recycled forward into every offer since. The salary history question was the mechanism that kept recycling it. The ban experiments are the cleanest proof we have of how pay really gets set: downstream of history, not downstream of value.

What actually happened when 22 states banned the question?

Massachusetts passed the first statewide salary history ban in August 2016. As of 2026, 22 US states and 24 local jurisdictions had enacted some form of it. Different states, different years, different industries covered.

That staggered rollout handed economists something they almost never get: a natural experiment. Ban the question in some places, leave it in others, then compare the workers. No survey opinions, no theory. Just wages, before and after, on both sides of a state line.

Multiple independent research teams ran the comparison with different methods. Boston University's team used Current Population Survey and Burning Glass data across 200 million job postings. An NBER team used synthetic controls. A separate study tracked recession-scarred cohorts. The directional finding held across all of them for private-sector job-changers: pay rose, and it rose most for the people the old system underpaid.

How much did pay actually go up, and for whom?

The headline numbers are specific, and they cluster around the same story.

Finding	Effect	Source
All job-changers in ban areas	+3.9%	Bessen, Denk & Meng
Women job-changers	+6.2%	Bessen, Denk & Meng
Non-white job-changers	+5.8%	Bessen, Denk & Meng
Gender earnings ratio (ban states)	+1%	Hansen & McNichols
Recession-scarred workers, weekly earnings	+5%	Mask

The number that should stop you is this one: those gains closed 40 to 67% of the residual wage gap for disadvantaged groups. Residual means the part of the gap that survives after you control for occupation, education, experience, and industry. The part that was supposed to be a mystery. Almost half to two-thirds of it vanished by deleting one question from the interview. That chunk had nothing to do with anyone's productivity. It was pure anchor drag.

For context on the size of the problem: in 2024, women earned 85 cents for every dollar men earned overall. The earlier working-paper estimate found roughly half the residual gender gap disappeared for job-seeking women under a ban.

Why did deleting one question move the market?

Anchoring. The first number on the table colonizes the final number. In a classic negotiation study, the first offer was a strong predictor of the final settlement price. Whatever lands first does most of the work, which is exactly why the old advice to never say a number first misreads the game.

Your salary history was that first number. When an employer learns you made $72,000, every figure they consider after that orbits $72,000. A "generous" offer is $78,000. Nobody in the room is asking what the role is worth. They're asking how little they can raise your old number and still get a yes.

Here's the cleaner version of the mechanism, from a two-sided audit study. Researchers sent 2,048 fictional software engineering applications to 256 recruiters. Identical resumes. Same school, same four years of experience. The only difference: some disclosed a salary, some didn't.

Disclosers: mean offer of $103,993. Non-disclosers: mean offer of $96,521.

A $7,472 gap, built from one number. Recruiters assumed the silent candidate had 9% weaker competing offers. The silence itself became a new anchor.

And the discounting is selective. An extra dollar of someone's salary that came from the gender wage gap raised recruiter valuation by only $0.42, versus $0.64 to $0.70 when the dollar came from working at a higher-paying firm. Employers half-know your history is contaminated. They use it anyway, because it's the anchor on the table.

Weak anchor vs strong anchor: what this looks like for one person

Take a woman who spent three years underpaid on a male-dominated team, sitting at $72,000.

Weak anchor. She interviews in a non-ban state. They ask what she makes. She says $72K. They offer $78K, an 8% bump that feels like a win. Her starting point was her own underpayment, so her ceiling is her underpayment plus a tip.

Strong anchor. She interviews in a ban state. They can't ask, so they post a range: $90K to $110K. She negotiates to $95K. Same skills, same week, same person. The $23K difference isn't her getting better at her job between Tuesday and Thursday. It's the anchor being reset from her history to the market.

Now the recession version. Imagine a worker who graduated into a downturn and started at $38K instead of the roughly $52K the same skills would have commanded in a normal year. Every later offer anchored to $38K. Years on, still trailing peers who graduated two years later with identical credentials. In a ban state, the research finds that move delivered a 5% weekly earnings boost above baseline, partially healing the scar. The scar was never about skill. It was about a number that wouldn't let go.

So is the ban a clean win? No. Here's the cost.

This is where honest reporting earns its keep, because the evidence isn't unanimous.

The public sector ran backwards. In the public sector the evidence runs the other way: at least one study found new-hire pay declined under bans, with no gender-gap improvement. The likely reason: stripped of history, employers facing more uncertainty hedge low. The private-versus-public split is real, and you should know which evidence applies to you.

People disclose anyway, and it isn't random. Even under a ban, 25% of workers always disclose their salary unprompted, only 58% comply, and the always-disclosers skew male. If high earners keep volunteering numbers while underpaid workers stay quiet, employers infer that silence means a low number. The ban's protection erodes for exactly the group it was built to help. The fix is posted ranges, so that silence carries no penalty.

The ban only fires at the moment of hire. If you're already in a role, you're still anchored to your history. Incumbents see little direct benefit. The question gets disrupted when you change jobs, not while you sit still.

Some of the gain may be transparency, not the ban itself. Salary-posting ads jumped 20 percentage points in a single year after bans passed, though the transparency laws themselves cut both ways. Posted ranges and the deleted question are tangled together in most studies. The trade-off named plainly: a ban with no posted ranges and no enforcement is a partial fix that can quietly backfire. The full effect shows up when the question is gone and a real range is on the table.

What this means for you right now

Your salary history is a fact about your last employer's willingness to pay. It is not a fact about your value today. The ban experiments prove the gap between those two things is large enough to move entire wage distributions.

You don't get to wait for a ban. So borrow the mechanism the bans use: walk in with a market number as your anchor, not your history.

• Get a market range before any conversation. What the role pays now, for your level, in your city. That range, not a single number, is your anchor. Your old salary is not.
• If they ask what you make, answer with the market. "I'm targeting roles in the X to Y range, based on what this role pays now." You answered. You just refused to hand over the old anchor.
• Treat a posted range as the floor of the negotiation, not the offer. The disclosers in that audit study cleared the posted midpoint. Silence cost the others 9% in assumed bargaining power.
• If you took a recession or first-job discount, name it and reset it. That scar follows you only as long as you keep quoting the scarred number.

The consistent move beats the heroic one here. You don't need to win a single dramatic negotiation. You need to stop carrying a stale number into every room, year after year, letting it compound against you. Reset the anchor once and every future offer builds from the market instead of from your past.

Want to know your real market number before your next conversation, not your salary history? Tell Praxy your role and city on WhatsApp, and we'll pull the range that should anchor the deal.