Large UK employers have published gender pay gap data every year since 2017. In March 2026, the government confirmed it will extend the same basic requirement to ethnicity and disability, for any employer with 250 or more staff. The plan borrows heavily from what already exists: the same six metrics, the same snapshot date, the same online reporting service. On paper, this looks like an expansion of a familiar exercise into two new categories.
It is not. The gender regime is built around a two-group comparison. Ethnicity isn't. Pay for different ethnic groups does not move together at all. Disability changes the picture even more sharply. Most of the story it needs to tell happens to people who are not in paid work in the first place. A pay gap measured only among the employed cannot see them at all. Both characteristics share a further problem gender reporting never had. Employers do not reliably hold the data. Disclosing ethnicity or disability at work is voluntary. Take-up has historically been patchy. Employers who treat this as “gender pay gap reporting, with two more columns” are underestimating the exercise on every count that matters.
What the New Regime Requires
The government's March 2026 response confirmed the shape of the new regime, though not yet its start date. Employers with 250 or more staff will report the same six measures already used for gender: the mean and median gap in hourly pay, the mean and median gap in bonus pay, the proportion of employees receiving a bonus and the split across pay quartiles. Three things go beyond that. Employers will have to publish a breakdown of their whole workforce by ethnicity and disability, not just the pay gap itself. They will have to publish declaration rates: the proportion of staff who chose not to say. And they will have to publish an action plan setting out what they are doing to close the gaps they report, brought in alongside similar action plans already becoming mandatory for gender from 2027. The requirement is no longer just to produce a number. It is to produce a number, understand what is driving it and show what is being done about it.
These additional requirements exist because the underlying data does not work like gender data. Sex is recorded as standard employment information. Ethnicity and disability are not. Employees will be asked to self-declare, with a “prefer not to say” option built in. Nobody yet knows how many will answer.
The legislation itself, the Equality (Race and Disability) Bill, had not been formally introduced to Parliament as of mid-2026. The government has been clear that mandatory reporting is coming. It has been far less clear about when the first reports will actually be due.
One Ethnicity Pay Gap Does Not Exist
Start with what “the ethnicity pay gap” is actually made of. The government's own impact assessment for the new reporting regime uses 2022 earnings data broken down by specific ethnic group. The picture is not one gap. It is several, moving in different directions.
Black African, Caribbean and Black British employees earn a median of £13.53 an hour, against £14.35 for White employees, a gap that has held steady since 2012. Pakistani and Bangladeshi employees earn less again. Chinese and Indian employees earn more than the White British median, comfortably so, running in the opposite direction from what most people picture when they hear “ethnicity pay gap.”
Aggregated into a single figure, these stories cancel each other out into something misleadingly reassuring. The government's own minimum reporting standard for the new regime is a binary comparison: the White group, including White Other, against everyone else combined. That is precisely the aggregation that flattens three genuinely different stories into a single, far less informative number: a persistent gap for Black employees, a severe gap for two specific South Asian groups and a pay advantage for two other Asian groups. Employers are free to break the comparison down further. The default they have been given will not do it for them.
There is a second layer of complexity. When the ONS adjusts these raw gaps for other characteristics that affect pay, such as occupation, qualifications, geography and age, most of them narrow. Some reverse entirely. A raw gap and an adjusted gap are answering different questions. One says who earns what today. The other asks how much of that gap survives once you account for where people work and what job they do. Reporting the first number without the second invites exactly the kind of oversimplified reading that got the aggregate “ethnic minority” figure into trouble in the first place.
The Disability Pay Gap Is the Smaller Story
The headline disability pay gap looks, at first glance, like a more contained problem than the ethnicity picture. It stood at 12.7% in 2023, roughly stable since 2014. Disabled employees earned a median of £13.69 an hour against £15.69 for non-disabled employees.
That number only describes people who are in paid work. It says nothing about everyone who is not. That is where the real story sits. The government's own data describes the long-term improvement in the disability employment gap as having stalled. In the latest quarter, 52.8% of working-age disabled people were in employment, against 82.5% of non-disabled people, a gap of 29.7 percentage points. A pay gap measured only among the employed captures an increasingly small and unrepresentative slice of the disabled working-age population. The people furthest from being paid at all do not appear in it.
The pay gap that does get measured is also far from uniform. It varies substantially by type of health condition. Researchers point to specific structural barriers behind these differences: limited access to flexible and part-time roles, inadequate workplace adjustments and patchy tailored employment support. Fewer than one in ten jobs advertised above £35,000 a year in the UK offer part-time working, at a time when flexible hours are, for many disabled people, the difference between being able to work at all and not.
Put the two facts together. The pay gap headline is real but comparatively modest. The employment gap behind it is close to 30 percentage points. It has stalled rather than closed. Any employer reading only the pay gap figure is reading the smaller and more comfortable half of the story.
The Data Most Employers Don't Actually Have
Gender pay gap reporting works, mechanically, because employers already hold the underlying data. Sex is recorded for payroll, pensions and a dozen other administrative reasons, whether an organisation ever intended to report a pay gap or not. Ethnicity and disability data does not arrive that way. Employees have to be asked. Under the proposed regime, they are entitled to decline.
This matters more than it sounds like it should. Legal commentators advising employers on voluntary ethnicity reporting describe low disclosure as the single biggest obstacle to producing a meaningful figure at all. The government has built a specific metric to capture this: the declaration rate, the proportion of staff who did not say. One law firm's assessment put it plainly. Non-disclosure rates may end up more revealing than the pay gap figures themselves, since a high rate of “prefer not to say” answers is itself information about how safe employees feel being honest with their employer.
None of this is a reason to expect leniency if the numbers come out badly. UK equal pay law already treats statistical evidence of pay disparity between comparable roles as legally significant. The government is also proposing to extend statutory equal pay rights to ethnicity and disability for the first time. A published gap, once it exists, is a standing invitation to ask why. An employer with clean data and a defensible answer is in a completely different position from one that can only offer the number itself.
What Being Ready Actually Means
Three different problems point to the same conclusion. Ethnicity data hides group-level variation behind one aggregate figure. Disability data hides most of the real story outside the workforce altogether. Both hide behind a disclosure rate nobody can yet predict.
Publishing a number is the easy part of this regime. It is not even the part that matters most. The new regime does not stop at a number. It requires an action plan: a statement of what is driving the gap and what the organisation is doing about it. Almost any organisation with 250 or more staff can run the calculation once the data exists. Writing an honest, specific action plan is a different task entirely. It requires an employer to understand what is actually driving the gap before they can say anything credible about fixing it. A gap can exist for entirely defensible reasons: a different age or tenure profile in the workforce, a different mix of roles, genuinely different qualification levels. It can also exist because pay decisions have quietly tracked assumptions nobody ever examined. From the outside, the two look identical. Only one of them survives scrutiny.
Telling the two apart requires the same thing equal pay law has always required for gender: a documented, consistent basis for what any given role is paid, applied the same way regardless of who holds it. Multiply that requirement across numerous ethnicity groups moving in different directions at once and a wide range of disability types, with much of the underlying data collected voluntarily and with real gaps in coverage. The administrative bar rises sharply compared with a single binary gender comparison. Employers who already have a structured, evidenced way of showing what a role is worth will be able to answer the question the moment a number is published. Employers who do not will be publishing a figure they cannot yet explain, at the exact moment the law is being extended to let more people ask them to.
Conclusion
The reporting mechanics for ethnicity and disability pay gaps have been copied, almost line for line, from a framework built for a much simpler comparison. Gender pay gap reporting works reasonably well because it compares two large, roughly balanced groups using data every employer already holds. None of that is true here. Ethnicity encompasses multiple groups moving in different directions at once. Disability's biggest story happens to people outside the workforce altogether. Both rely on data that has to be asked for. A meaningful share of employees may choose not to give it.
None of that makes the exercise pointless. It makes the exercise considerably harder than the version employers have been living with since 2017. Done properly, it is also considerably more revealing than a form filled in for compliance's sake. The organisations that handle this well will not be the ones with the tidiest spreadsheet. They will be the ones that already know, in detail, why every role is paid what it is.
