Organisations should continue to use gender data for gender pay gap reporting

By Jeni Tennison

The main message for this post is in its title. If you are a UK organisation that is reporting gender pay gap data, you should continue to use the existing gender data that you have in your HR system, because it will:

• Make very little difference to your reported gender pay gap figures.
• Make no difference at all to what you do as an organisation to address your gender pay gap.
• Reduce your legal risks by avoiding potential discrimination and unlawful data sharing.
• Make a difference to your trans employees because you will be treating them with dignity and respect.

Background

The gender pay gap is the difference between the average pay of men and women in an organisation. Since 2017, companies with over 250 employees and public bodies have been required to report their gender pay gap by The Equality Act 2010 (Gender Pay Gap Information) Regulations 2017 and The Equality Act 2010 (Specific Duties and Public Authorities) Regulations 2017. These reports are all published through the government’s gender pay gap service.

Following the Supreme Court ruling on the interpretation of “male” and “female” in the Equality Act 2010 and the publication of a new draft of the EHRC’s Code of Practice on 21st May 2026, the government has made changes to the guidance to employers for preparing data in order to calculate the gender pay gap, specifically around how to identify someone’s gender for the purposes of gender pay gap reporting. Prior to the changes, this guidance said:

The new version of the guidance says:

(Note that “biological sex” here is undefined, but in practice it uses the same definition as that in the updated EHRC Code of Practice, namely the sex recorded on someone’s original birth certificate. For the rest of this post, I will use the term “sex recorded at birth”, which has a clearer definition than “biological sex”.)

The guidance for gender pay gap reporting also tells employers to ignore gender recognition certificates, which generally apply “for all purposes” but now no longer apply when using law based on the Equality Act 2010, such as that around gender pay gap reporting.

Following the UK government’s revised guidance would mean that many organisations would need to keep three different records for someone’s gender/sex:

1. Their self-identified gender, which as a matter of dignity and respect, should be used in most circumstances, such as when sending someone an email or letter.
2. Their legal gender, which is used in payroll reporting to HMRC (which continues to recognise gender recognition certificates).
3. Their sex recorded at birth, used only for gender pay gap reporting.

This guidance follows the letter of the law, which it should, being guidance from the government. However, as we will see, in practice organisations have to consider the wider ramifications of collecting gender pay gap data and balancing different approaches and methods with other impacts and implications.

Makes very little difference to the figures

The charity Close the Gap works with policymakers, employers and employees to influence and enable action to address the causes of women’s inequality at work. They carried out a scenario-based study reported in their submission to the Equalities, Human Rights and Civil Justice Committee on the Gender Recognition Reform (Scotland) Bill in May 2022 in which they looked at what would happen to gender pay gap figures when trans women were counted as women rather than according to their sex recorded at birth (male) in three examples:

• Example A: Large male-dominated, private sector employer with 823 employees with an existing gender pay gap of 26%, employing three trans women working in the lower, middle and highest grades of the organisation – no change to the gender pay gap, which remained at 26%
• Example B: Large public sector employer with 10,000 employees with an existing gender pay gap of 9%, employing 28 trans women employees working across all eight grades with the most senior person being a trans woman – no change to the gender pay gap, which remained at 9%
• Example C: Medium-sized, female-dominated private sector employer with 280 employees with a gender pay gap of 21%, with one senior trans woman – a 2% reduction on the employer’s gender pay gap to 19%

Close the Gap concluded:

To build on the scenarios tested by Close the Gap, I ran a set of simulations (using what’s called the Monte Carlo method) to see what difference it would make for a typical organisation in the 250-to-499 employee range. I chose this size of organisation because the plurality of companies reporting gender pay gaps (43%) are of this size. The full methodology is described at the end of this post, but you can also take a look at the spreadsheet (and copy it to run your own similar simulations if you like).

I based the simulations on the median 250 to 499 person company that had a link to more details about their analysis – ENGIE POWER LIMITED (a company specialising in sustainable energy supply and services). As far as I was able, I used the same size of workforce (374 people), average salary (£71k) and top salary (£220k). Then I simulated 200 different versions of this company, with roughly the same mix of men and women as they reported in 2025-26 and roughly the same gender pay gap (for ENGIE POWER this was 11.45%), but with different numbers and distributions of trans employees (decided randomly with a 0.63% chance of being trans – the percentage of trans people in the working age population, derived from 2021 Census data).

The average mean hourly pay gap over those 200 simulated versions of the company are shown in the following table:

This shows that the gender pay gap statistic that’s reported by a company changes very little whether you use self-identified gender or sex recorded at birth as the basis for the calculation. That’s because trans people are a very small percentage of each workforce, and because there are roughly as many trans men as trans women.

The number and distribution of trans people does make a difference. If you have no trans people in your organisation, of course it makes no difference to the figures. In the simulations, the biggest differences came when trans men or trans women were in the most highly paid positions:

• The biggest reduction in mean pay gap (from 9.52% when using self-identified gender to 7.31% when using sex recorded at birth) was in simulation no. 152 with two trans men with high salaries and one trans woman with a low salary.
• The biggest increase in mean pay gap (from 11.52% when using self-identified gender to 13.41% when using sex recorded at birth) was in simulation no. 38 with one highly paid trans woman, one trans woman in the top middle quartile, and two trans men and one trans woman in the bottom quartile.

This bears out the Close the Gap analysis: the method you use to calculate the gender pay gap has minimal effect. It has the most effect when trans people are in the highest paid positions, but even in these cases it’s only around two percentage points.

This modelling does not reflect some of the realities of employment for trans people, who are less likely to be in employment, and more likely to occupy lower paid roles. TransActual’s Trans Lives 2025 survey found:

All the evidence points to the conclusion that the method of measuring the gender pay gap will make very little difference to the figures.

Makes no difference to what you do

Data collection and analysis always has a purpose. If you don’t pay attention to that purpose, you’re unlikely to make good, proportionate decisions about what data to collect and how to use it.

In the case of gender pay gap reporting, companies are complying with the law, but there’s a reason that law is in place. The explanatory memorandum accompanying the Equality Act 2010 (Gender Pay Gap Information) Regulations 2017 makes this clear:

The exact figures for the gender pay gap for a company are not all that important. There are no thresholds you are required to reach and no penalties for exceeding a certain gap. Companies are supposed to analyse why they have a gender pay gap and change their policies to address it, but there is no expectation that it will be eliminated. The trend is important – year by year, is it going down, staying about the same, or going up? But the actual figures aren’t important, and certainly not significant enough that a change of less than one percentage point – or even 2-3 percentage points – is going to actually change what a company does.

Like many other organisations, the company on which my simulations were based, ENGIE POWER, publishes their gender pay gap report. In it, they list their achievements and the actions they’re taking around recruitment, flexible working, salary reviews and so on. None of these will be meaningfully affected by the very small changes to the gender pay gap figures brought about by making the analysis be based on sex recorded at birth rather than gender.

When there are multiple methods for defining, collecting or calculating data, and the precise value of a data point doesn’t meaningfully change how you act, or a decision you take, it’s more than reasonable to take other impacts into account. This is actually the case in many situations, especially when you’re analysing data about groups of people. Ethical statisticians, researchers and other data professionals will usually be aware of, and seek to balance, the trade-offs between different data definitions and methods of data collection and other impacts of data collection and use.

Reduces your legal risks

Reporting gender pay gap data is a legal obligation, and some of the calculations for doing so are part of the law, but organisations also have other legal obligations, including compliance with the Gender Recognition Act 2004, the Equality Act 2010 and the UK GDPR. The updated guidance tries to help organisations navigate the interactions between these obligations, albeit not explicitly. For example, the aim of the following recommendations is to help organisations to avoid complaints about discrimination based on gender reassignment (or perception of gender reassignment):

• “We recommend that you have a policy or process to collect this data. This should be the same for all employees, regardless of their sex or gender identity.”
• “You should not single out individual employees and ask them about their sex or gender identity.”
• “We recommend that you do not ask for documentation to confirm an employee’s biological sex.”

Similarly, the guidance around handling of data about who has a gender recognition certificate reflects the very real risk of highly confidential information leaking out through the gender pay gap reporting process, against both data protection and gender recognition law.

There is a balance of risks here, but it is clear that in practice, gathering data about sex recorded at birth is the more risky path. If you use self-reported or legal gender, you have the risk of reporting gender pay gap data that is fractionally different from what it would be if you reported gender pay gaps based on sex recorded at birth. There is no legal penalty for this and even if there were it would be impossible for anyone without intimate knowledge of your calculation process to detect.

If you use sex recorded at birth, on the other hand, you need to manage the risk of contravening equalities, gender recognition and/or data protection law. Any issues with your approach – such as data breaches – are likely to be visible to the people affected, and there are real penalties attached to violations of equality, gender recognition and data protection laws.

Even the official guidance doesn’t say you have to gather data about people’s sex recorded at birth. It just says “take reasonable and proportionate steps to obtain [that data]”. The reasonable and proportionate step is to use the existing data you have about people’s gender. Make that your policy.

Supports your trans employees

As TransActual’s Trans Lives 2025 survey starkly illustrates, the UK is already a hostile environment for trans people. According to an Amnesty International analysis, there has been an organised backlash against trans people’s rights in the UK starting in 2017/18, with a growing gender critical movement and the media normalising anti-trans rhetoric. From being top of the European Rainbow Map Index in 2015, the UK is now ranked 22nd, On laws relating to the recognition of trans people’s identities, the UK is now ranked 45th out of 49; the only other European countries with a legal block on effective recognition of trans people’s identities are Bulgaria, Georgia, Hungary and Russia.

In this context, as a trans person, being asked about your sex recorded at birth, and especially to ignore your legal gender in providing a response, is intrusive and threatening. Employers should instead aim to make work a safe and inclusive space for their trans employees; it is not too much to ask to respect their gender.

Implications beyond gender pay gap reporting

This post has shown how, in the vast majority of cases, when you are comparing hundreds of men and hundreds of women, whether statistics are based on self-identified gender or sex recorded at birth makes little difference to the data you get, because trans people are such a small proportion of the population.

Accuracy and precision matters for data about individuals where that data is used to support decisions about that individual. For example, whether they’re a man or woman or non-binary, you really want to have accurate data about whether someone has breasts or not to know whether to invite them for breast cancer screening. You need to know someone’s legal gender to be able to apply the correct tax and pension rules to them.

But when you design research studies or create statistics, the requirements are different. You usually have multiple potential definitions for the data you want to collect, and multiple potential methods for collecting it. So you need to weigh these up alongside the ease of collection of data, and the impacts on the individuals and groups about whom data is being collected. “Good enough” is good enough.

This is one of the places where the Sullivan Review goes wrong in its analysis. When it states that “It is important to remember that the purpose of survey data collection is to gather data about populations rather than to provide an opportunity for each individual to express the full complexity and richness of their identity” it fails to recognise that there are ethical choices involved in how data is collected. If the method you use won’t make much difference to the results – as it won’t when you have a small population – it is reasonable and proportionate to choose the data definition and method of collection that treats people with dignity and respect.

About the author

Jeni Tennison is the founder and executive director of Connected by Data which campaigns for communities to have a powerful say in decisions about data and AI.

Methodology

I ran a set of simulations to see what kind of difference it would make for a typical organisation in the 250-to-499 employee range (chosen in part because the plurality of companies reporting gender pay gaps (43%) are of this size and in part because having a relatively small number of employees would make running the simulations easier).

I downloaded the reported pay gap data for 2025-26 from the gender pay gap service and filtered to companies of that size that provided their own gender pay gap reporting link, as I figured these companies would be most transparent about other aspects of their business. I then identified the company within that set whose gender pay gap was closest to the median for those companies to use as the basis for the simulation. That company was Engie Power Ltd.

I found the latest accounts for Engie Power Ltd from Companies House, which were from 2024 – not quite the period for the gender pay gap data, but close enough. From there I identified the number of employees, the total wage bill (from which I calculated the mean hourly wage), and the highest salary for the company (they only have one director, which made this easy). This enabled the modelling of salaries within the company, based on a log normal distribution. (I didn’t model bonus pay because it would have added unnecessary complexity and wouldn’t meaningfully change the result of this analysis.)

I then simulated 200 versions of Engie Power Ltd, in each of which gender and trans statuses for each employee was assigned randomly, based on the probability of each gender in each quartile (as reported in the gender pay gap data) and the probability of a working age adult being trans. Each employee’s wage was adjusted slightly based on their gender – men up and women down – to result in roughly the same gender pay gap as observed for Engie Power Ltd (a lot of the gender pay gap is actually a result of higher waged employees being more likely to be men than women).

Trans people made up about 0.54% of the 16+ England and Wales population and 0.44% of the 16+ Scotland population, according to their 2021 censuses. For working age people, this is a little higher because of the increasing prevalence of trans people amongst more recent generations: 0.63% of 16-64 year-olds in England and Wales, and 0.54% of 16-64 year-olds in Scotland. In a 250-person organisation, you’re likely to have one or two trans staff total; in one of the around 70 organisations with over 20,000 employees, you’re likely to have roughly between 100 and 130.

The results of the trials are shown in the following table:

As you can see from this table, the gender pay gap statistic that’s reported by a company changes very little whether you use gender or sex as the basis for the calculation. That’s because trans people are a very small percentage of each workforce, and because there are roughly as many trans men as trans women.

Even these small differences are likely to be even lower in real companies. The calculations in the simulations described above were based on the percentage likelihood of self-identifying as trans as a working age person in England or Wales (0.63%). However, for the purpose of gender pay gap reporting, companies are probably going to use information already in their HR system and specifically payroll information. This will include gender as reported to HMRC as part of each month’s payroll. HMRC only respects gender recognition certificates (GRCs) as proof of changes of gender, so instead of comparing the figures based on self-reported gender with those based on sex recorded at birth, we should be using figures based on legal gender.

The proportion of people with a gender recognition certificate (GRC) is just 0.014% of people and 0.018% of people of working age. That means only about one in 20 organisations with 250 people will employ someone with a GRC. An organisation with 20,000 staff is likely to employ 3-4.

The same table using this “trans likelihood” percentage looks like:

So, using your employee’s legal gender (which you already have in your HR system because you report it in payroll every month), rather than their sex recorded at birth will give almost exactly the same figures for your gender pay gap.