How Do You Collect LGBTQ+ Self-ID Data at Work? A 2026 Employer Guide

Self-ID is the voluntary process by which employees tell their employer whether they are LGBTQ+, so the organisation can measure representation, pay, progression and attrition by group. In the UK and EU, sexual orientation is special category data under Article 9 of the GDPR — and the lawful route most employers rely on is not consent, but the equality-of-opportunity condition at Schedule 1, Part 2, paragraph 8 of the Data Protection Act 2018, which permits aggregate monitoring and expressly forbids using the data in decisions about individuals.

That single distinction settles most of the design questions below. If the data can never touch an individual decision, it has to be firewalled from pay, promotion and performance by construction rather than by good intentions.

Why bother? Because HR is the last to know

The Human Rights Campaign Foundation's Equality Rising report, fielded in 2023 across 3,044 US workers, found that 84% of LGBTQ+ workers are out to at least one person at work and 69% are out to their direct supervisor — but only 48% are out to their HR department. People are out to the colleagues who affect their day and closed to the function that holds the records.

That gap is the entire argument for self-ID. Without it, your headcount reporting reflects who felt safe telling HR, which is roughly half of the people it is meant to describe, and any inclusion metric built on it understates the population by a factor nobody can quantify.

It is worth being precise about what different studies measure, because the numbers are not interchangeable. Deloitte's LGBT+ Inclusion @ Work, fielded in early 2023 across 5,474 LGBTQ+ employees in 13 countries, found 43% globally were comfortable being out at work — a different construct from HRC's "out to at least one person", and not comparable to it. Stonewall's UK Work Report found 18% of lesbian, gay and bi people were not open with anyone at work, but that fieldwork dates from 2017 and Stonewall has published no newer UK-wide workplace survey.

What is the correct question wording?

Do not write your own. Harmonised wording exists so that your results can be compared against national data, and the UK standard warns that even small alterations may substantially affect comparability.

Sexual orientation, UK. The Government Analysis Function harmonised standard asks: "Which of the following options best describes how you think of yourself?" with the options Heterosexual or Straight, Gay or Lesbian, Bisexual, Other, and Prefer not to say. The "prefer not to say" option belongs in self-completion surveys, which is what an employee survey is.

Sexual orientation, US. The federal recommendations issued by the Office of the Chief Statistician in January 2023 ask "Which of the following best represents how you think of yourself?" offering Gay or lesbian; Straight, that is not gay or lesbian; Bisexual; I use a different term; and I don't know. The guidance is explicit that "don't know" respondents are qualitatively different from those choosing "something else" and should be reported separately.

Gender identity. Treat this as unsettled and say so internally. The UK's harmonised standard was archived in December 2024 and replaced with interim guidance that states plainly there is no agreed best practice yet, with finalised standards expected in autumn 2026. In the meantime it offers two validated forms: the England and Wales Census 2021 question, "Is the gender you identify with the same as your sex registered at birth?", and the Scotland Census 2022 question, "Do you consider yourself to be trans, or have a trans history?" The US federal approach is a two-step: sex assigned at birth, then current description, with a confirmation prompt.

Why "prefer not to say" is the most important option on the form

Without an explicit opt-out, a blank is ambiguous: it could be refusal, an oversight, or none of the categories fitting. With one, refusal becomes a measurable signal you can track over time — and a rising "prefer not to say" rate is one of the few early indicators that trust is moving in the wrong direction.

For a sense of scale, the Behavioural Insights Team ran a randomised trial across 24,581 first-time job applicants in 2021 and found "prefer not to say" running steadily at 7.1% to 7.7% across every condition. The more useful finding for anyone planning a communications push: none of the three message treatments the trial tested — addressing mistrust, reassuring applicants it would not harm their application, or explaining the value of disclosing — produced a significant effect. Copy was not the lever. System design and trust were.

What response rate should we expect?

There is no published benchmark for corporate self-ID response rates, and any figure presented as an industry average is almost certainly extrapolated. What does exist is a long real-world series: the UK Civil Service's declaration rate for sexual orientation rose from 38.1% in 2015, the first year it was collected, to 74.4% by March 2026. Of those with a known orientation, 7.4% identify as lesbian, gay, bisexual or other.

The shape of that curve is the lesson. Eleven years of sustained, repeated asking roughly doubled the declaration rate. A first-year result in the thirties is not a failed programme, and a campaign that expects to arrive at eighty per cent in one cycle is going to be judged a failure for the wrong reason. The Government of Canada's guidance, for what it is worth, sets an operational aim of at least 80% while stressing that participation must be genuinely voluntary and free of pressure.

How small is too small to report?

This is where invented numbers circulate most freely, so here is what is actually documented. There is no LGBTQ+-specific published threshold.

• The ICO sets no numeric threshold at all, holding that the risk of singling someone out depends on context. It does cite the NHS standard for strong anonymisation, where k is set to five.
• UK Government Statistical Service disclosure guidance suppresses or combines cells with one or two contributing units, and treats cells of size one and two as inherently risky.
• The most common HR platform default — Microsoft Viva Glint — requires five respondents to show a score and ten to show open-ended comments, and applies cascading suppression so that a group above threshold is still hidden if publishing it would let someone back-calculate a group below it.

Cascading suppression is the part most in-house dashboards miss. Publishing four of five regions and withholding the fifth tells anyone with a calculator exactly what the fifth contains.

The legal basis, and the three things it obliges you to do

Article 9(1) of the UK and EU GDPR names sexual orientation explicitly as special category data. Gender identity is not separately listed, and is commonly treated as special category by inference rather than by the text — a distinction worth getting right in your own documentation.

Consent is a poor basis in an employment context, because the power imbalance makes freely-given consent hard to establish. That is precisely why paragraph 8 of Schedule 1, Part 2 of the Data Protection Act 2018 exists, permitting processing necessary for identifying or reviewing equality of opportunity between groups, with sexual orientation named among the specified categories. Three conditions come attached, and each translates into something concrete:

• The processing must not be carried out for measures or decisions affecting particular individuals — so self-ID data must be architecturally separated from pay, promotion and performance systems.
• An appropriate policy document must be in place at the time of processing. This is a legal precondition, not documentation you write up afterwards.
• An employee may prohibit the processing by written notice, so you need a route for that which does not require a conversation with their manager.

A data protection impact assessment is the expected accompaniment for special category processing at scale. In the United States there is no federal equivalent, but California is a significant exception: under the CCPA as amended by the CPRA, sexual orientation is sensitive personal information, and the employment-related exemption expired at the end of 2022, bringing HR data into scope from January 2023. Other state privacy laws vary in whether they exempt HR data — verify state by state rather than generalising.

The question multinationals have to answer first

IBM's programme, documented by the UMass Amherst Center for Employment Equity, runs self-ID in 40 countries covering 87% of its global workforce — and deliberately excludes the 27 countries where being LGBT+ is criminalised. Only four people in the company can ever see self-ID data linked to employee names, all under non-disclosure agreement. Spain was removed from the programme following a change in data privacy law.

Whatever you think of the specific line IBM drew, the design principle is the transferable part: decide the exclusion list before launch, on published legal criteria, and put a very small number of named people between the raw data and everyone else. Our guide to the countries that criminalise being LGBTQ+ covers the current list and how quickly it moves.

Frequently asked questions

Is sexual orientation special category data under GDPR?

Yes. Article 9(1) of the UK and EU GDPR names "data concerning a natural person's sex life or sexual orientation" explicitly. Gender identity is not separately named and is generally treated as special category by inference.

Can we use consent as our lawful basis for self-ID?

You can, but most UK employers do not, because the employment power imbalance makes freely-given consent difficult to establish. The equality-of-opportunity condition in Schedule 1, Part 2, paragraph 8 of the Data Protection Act 2018 is the route designed for this purpose. Take your own legal advice; this article is not legal advice.

What is a good self-ID response rate?

No verified corporate benchmark exists. The UK Civil Service moved from 38.1% in 2015 to 74.4% in 2026, which suggests judging year one against year three rather than against a target.

What is the minimum group size for reporting LGBTQ+ data?

There is no LGBTQ+-specific standard. UK statistical guidance suppresses cells of one or two, the NHS anonymisation benchmark sets k at five, and the most common HR platform default is five for scores and ten for free-text comments.

Before you launch

Self-ID measures how many LGBTQ+ people you employ. It does not measure whether your organisation is built to support them, which is the question the data is usually being gathered to answer. Our Active Allyship Index scores that in about twelve minutes across four dimensions, and it is a useful thing to run before a self-ID launch rather than after — if your risk-and-response score is low, your declaration rate will tell you so anyway, twelve months later.

myGwork is the global business community for LGBTQ+ professionals, graduates, inclusive employers and anyone who believes in workplace equality. Figures in this article are cited to the HRC Foundation, Deloitte, Stonewall, the UK Government Analysis Function, the ICO, the Behavioural Insights Team and UK Civil Service Statistics, with the date of each source given in the text. It is general information, not legal advice.