Everything you wanted to know about sex and gender, but were afraid to ask

If you have ever pulled together a twelve-country study and found the gender variable behaving differently in every market, you already understand why Esomar's demographics project exists. Different wordings, different response lists, different assumptions about what is being measured, lead to confusing differences in the data. Was that difference a real difference, or was it because of how the questionnaire asked the question?

The demographics project sits with Esomar's Professional Standards Committee and is delivered by a dedicated working group, led by Jon Puleston, drawn from the panel companies and the large agencies, validated by the committee and by a client sounding board, and endorsed by Esomar's Council. It works through the demographics one at a time and publishes an evidence-based recommendation for each. Age, gender, and working status are already out, and gender is the one I would encourage you to read first, because it is the most culturally sensitive of the set and the most likely to change again.

Built for work that crosses borders

The purpose of these recommendations is comparability. The aim is to describe demographic groups in as globally consistent a way as possible, so that a result from Brazil can sit sensibly alongside a result from Japan, and so that a fifteen-market tracker produces one picture rather than fifteen local ones.

For work inside a single country, local knowledge will usually serve you better, and the recommendation says so. Where your national association has published guidance on asking about gender, consult it. Your census authority and your local fieldwork partner will know the conventions, the legal position and the language that participants in that market expect. Treat the Esomar recommendation as the sensible default for the situation where no single national convention can apply everywhere.

The evidence behind it is substantial. The primary source is Cartwright and Nancarrow's 2022 IJMR paper on gender classification in international research, and the working group consulted individuals and groups in 84 countries across Europe, North America, MENA, Asia Pacific, Central and South America, and Africa.

Sex and gender are related, and they are different

The recommendation spends time on this because the terms are sometimes used interchangeably, sometimes even in legal documents. Sex is assigned at birth on the basis of primary sex characteristics, and the categories are typically binary, occasionally including intersex. Gender is largely culturally determined, expressed in terms of masculinity and femininity, and assumed at birth to follow from sex. For most of the population, the two align, and the distinction feels academic. For anyone whose gender sits differently from the sex they were registered as, it matters a great deal.

A third term gets pulled in and conflated with both. Sexual orientation is interpersonal, about who someone is attracted to, while gender is personal, about how someone sees themselves. For most studies, asking about gender alone will do the job and is less intrusive.

Mostly an online panel context

The recommendation is written with online data collection front of mind, which shapes it in useful ways. It assumes your sample provider may already hold the answer and it worries about translation and about single coding. Most of the advice carries over to face-to-face and telephone work, and the document flags where those modes need something extra, including the obvious point that an interviewer should ask rather than infer gender from appearance or from a voice on the phone.

The key tension: inclusive and representative

The document draws a distinction that is worth pinning up above your desk. To be inclusive is to design so that anyone who wishes to take part can do so, without unnecessary barriers. To be representative is to design a sample that accurately reflects the target population you are measuring. Those two goals sit comfortably together for most demographics. For gender, they pull apart, because the population data you would weight towards largely does not exist yet.

When the recommendation was published, census gender data going beyond male and female was available in Canada and in England and Wales, with plans in Australia and New Zealand, and the picture elsewhere was patchy and inconsistent. Several of those plans have since moved forward, and it is worth checking the current position for the markets you work in. So you have a growing group of participants, concentrated in younger age bands, whom you can invite and include but cannot currently weight in any accurate way.

There is a second pull alongside it. A single question asked identically everywhere gives you clean, stackable data. But in some countries, it asks participants to disclose something that is socially risky, legally awkward, or genuinely unsafe. The recommendation resolves that with a core question plus a permitted local variation, which keeps the adaptation a documented decision rather than a judgement made in a hurry on the Friday afternoon before fieldwork.

Five points to take away

1. There is one core question, with a defined fallback. Where it is safe for participants to answer: "Are you: Female, Male, Another gender, Prefer not to answer". Where a non-binary option would be unacceptable or unsafe: "Are you: Female, Male, Prefer not to answer". Markets with well-established terminology can add "non-binary" alongside "another gender". For panel registration, where a single universal question is often needed, the binary version with the opt-out is the recommended base, adjusted by market wherever you can. Studies that need more detail, for example medical research or work among LGBTQ+ populations, sit outside the scope of this recommendation, and they need care, because gender data can become special category data under privacy law once it discloses health information.

2. The wording choices are deliberate, and translation drove several of them. In English, use male and female rather than man and woman, which keeps you consistent with census terminology and with the current evidence base, and avoids words that carry an age loading. In other languages, follow census terminology where gender is asked. "Another gender" travels better than "non-binary". Avoid asking how someone "describes" themselves, since a person's gender is what they are, rather than how they portray themselves. And list the options alphabetically or randomise them, which retires the old convention of male first.

3. Ask once, then keep it fresh. Where the sample provider already holds gender and can append it, use that and leave the question out. Repeatedly asking people to restate it at the start of every survey is a reliable source of participant friction. Because gender is fluid for some people, panel records should be refreshed at least annually, with panellists able to update their own information.

4. Route and weight inclusively. Keep non-binary participants in the sample and in the routing, and avoid screening out on gender. In the absence of a robust country data source, the recommendation is to give individuals in this group a weight of 1. Think about what counts as a usable base before you report on this group.

5. Explain why you are asking. Offer a "why are we asking this?" button online, with a spoken, visual, or written equivalent offline, and consider a second explanation for why "another gender" is offered. In some countries this is a requirement. Where you ask about both sex and gender, ask them as two clearly labelled questions and explain the difference between the terms.

How to get a copy

The gender recommendation is on the Esomar site, under Codes and Guidelines, in the section on best practice recommendations for measuring international demographics:

https://esomar.org/publications/demographic-best-practice-recommendation-gender

Look at the companion documents while you are there, and keep an eye on the page, since the working group has committed to updating this guidance as often as the subject requires. If you run international studies, adopting the series as your house standard removes a whole category of argument from your next global brief.