Gender-sensitive research and inclusion: data disaggregation, sampling, the instrument and accessibility

1. The problem: averages that hide differences

An entrepreneurship support programme reports that 70% of participants are satisfied. A digital public service shows a high take-up rate. A survey on access to healthcare shows an acceptable average picture. Each of these figures can be true and, at the same time, hide the fact that rural women use the service three times less often, that people with disabilities never reach it, or that young men only see a doctor in an emergency. Gender-sensitive research that is also inclusive is built so that these differences can be seen, measured and explained.

For donor-funded programmes, for public policy and increasingly for companies, this approach is an explicit requirement. But it does not come down to adding a "women/men" column to the tables. It changes the sample, the person who answers, the wording of questions, the organisation of fieldwork and the way results are reported. This article explains each of these changes and what to ask the agency for.

2. What gender-sensitive research is, and what inclusion adds

Two notions are often confused. Sex is a recorded characteristic of the person; gender refers to the roles, expectations and power relations that society associates with being a woman or a man. Data disaggregation by sex shows that a difference exists; gender analysis tries to explain why it exists, through roles in the household, access to resources, social norms, available time and safety.

Gender-sensitive research applies this perspective at every stage of a study: the research questions, the sample, the instrument, the field team, the analysis and the report. Inclusive research extends the same logic to other characteristics that produce unequal access: age, disability, language, place of residence, ethnicity, economic status. In practice the two are designed together, because the largest differences often appear where they intersect: older women in villages, young people with disabilities, linguistic minorities in rural areas.

3. When it is needed, and how far to go

A minimal gender perspective is useful in almost any study involving people: recording the respondent's sex and age and checking the main differences. A full gender-sensitive design is needed when the decision depends on differences between groups:

• evaluating social and development programmes, where the funder requires disaggregated indicators, usually from the baseline and endline studies onwards;
• public policies and services that must reach the whole population, not just its average;
• studies of knowledge, attitudes and practices, where behaviour differs sharply by role, as in the KAP study;
• market studies for products used differently by women and men, or chosen by one person and used by another.

The level of detail is set in the brief: which subgroups must be compared, with what precision, and which decision depends on the comparison. The rest of the design follows from that list.

4. The sample: data disaggregation needs adequate bases

The most common technical problem is simple: data disaggregation is requested in the report but was never planned in the sample. In a national sample of 1,000 people, a subgroup such as rural women over 60 may have, for example, around 120 respondents. The margin of error for a 50% estimate is then roughly ±9 percentage points, even before the design effect. With two groups of 500 respondents each, a difference between women and men must exceed about 6 points to be significant in a simple random sample, and more in a clustered one.

The solution lies in the design. Subgroups that must be compared become planned domains, and small ones receive a higher sampling fraction, with weights that restore the correct overall structure, following the procedure in the article on weighting survey data. Very rare or hard-to-reach subgroups, such as people with severe disabilities or survivors of violence, need the special designs described in the guide on hard-to-reach populations.

5. Who answers: the person, not the household

Many surveys interview "an adult member of the household", usually whoever opens the door or the head of household, and ask them for information about the whole household. For gender-sensitive research this practice produces systematic errors. A person answering for others reports their own activities better than those of others, and unpaid work, small incomes, shared decisions and ownership of assets are the most affected.

There are three practical rules. The respondent is selected at random from the eligible adults in the household, not chosen by availability, and weights correct for the unequal selection probabilities. Questions about each person's own income, time, decisions and assets are put to that person; answers given by someone else are flagged as such. And where the topic requires it, for example household decision-making, both partners are interviewed separately with identical questions, so that their perspectives can be compared.

6. The instrument and the field team

A gender-sensitive questionnaire follows a few principles. Questions about the "head of household" are replaced by questions about who decides, who owns and who does what, for each domain. Unpaid work and caring for others are measured explicitly rather than assumed. The language avoids stereotypes and wording that presumes roles: "who looks after the children in the household?" rather than "how much time do you spend with the children, as their mother?". Answer categories allow for real situations, including shared responsibilities.

For sensitive topics such as violence, reproductive health or family conflict, further rules apply. The interview takes place in private, with no other family members present. The interviewer is usually the same sex as the respondent, and training covers how to respond to a disclosure and information on the support services available. If privacy cannot be ensured, the sensitive questions are dropped, because the respondent's safety matters more than the data. The general principles are set out in the guide on research ethics and data protection.

In qualitative research, focus groups are usually held separately for women and men when the topic touches on roles or power, because mixed discussions tend to be dominated by one group. The moderator is the same sex as the participants when the subject calls for it.

7. Inclusion and accessibility in practice

An inclusive study removes the barriers that prevent some people from taking part, even when they have been selected.

• Disability. People with disabilities are identified through questions about functional difficulties (seeing, hearing, walking, remembering, self-care, communicating), such as the short set developed by the Washington Group on Disability Statistics, rather than the direct question "do you have a disability?", to which many answer no. The interview is adapted: large-print cards, extra time, a sign language interpreter, physical access to the interview venue.
• Language. The questionnaire exists in the languages of the population studied, and interviewers speak them fluently. A multilingual team prevents the quiet exclusion of linguistic minorities.
• Time and logistics. People caring for children or older relatives take part only if the interview or group is held at a suitable time, close to home, possibly with care support during the discussion.
• Interview mode. An online-only survey excludes a large share of older people and of rural residents with limited internet access. For social surveys with inclusion goals, face-to-face interviewing often remains necessary.

8. Analysis and reporting

The analysis of gender-sensitive research works on three levels. The first is data disaggregation: every key indicator is shown by sex and by the other planned variables, with unweighted bases and with statistically significant differences marked. The second is intersectional analysis: differences are examined across combinations of variables, for example rural women and men by age group, within the limits of the bases available. The third is explanation: qualitative data and questions about roles, time and resources show why the differences exist, not merely that they do.

Reporting avoids two opposite errors. The first is treating "women" or "people with disabilities" as uniform groups, when the differences within them are often larger than those between groups. The second is overloading the report with non-significant differences that look interesting and are not real. Every gender finding ends with its implication for the programme or product.

In repeated studies, such as programme evaluations, gender indicators are defined before the first wave and kept identical: the same wording, the same population, the same disaggregation. A good indicator measures not only how many women took part but what changed for them: access to a service, control over an income, available time, participation in decisions. In gender-sensitive research, choosing the indicators is already an analytical decision, because it determines which differences can be observed at the end and which will stay invisible.

9. What to ask the agency

• The list of subgroups to compare and the precision required, translated into a sample size for each subgroup.
• The method for selecting the respondent within the household and the list of questions asked individually.
• The instrument reviewed for language, stereotypes and answer categories, with questions on functional difficulties.
• The protocol for sensitive topics: privacy, interviewer matching, response to disclosures, support services.
• Accessibility measures: languages, formats, venues, hours, care support.
• The qualitative component, with separate groups where the topic requires it.
• The analysis plan: disaggregated indicators, significance tests, the combinations analysed and the rule for small cells.
• Personal data protection in line with the GDPR and Moldova's Law no. 133/2011, especially for health and disability data.

Gender-sensitive and inclusive research is not a box-ticking requirement but the condition for basing decisions on everyone they affect, not just on their average. It costs more only when bolted on at the end; planned from the start, it mainly changes how the study is thought through. If you are preparing a programme evaluation or a public policy study, our qualitative research and social surveys teams can help you build the right design.

See also: Social and political surveys Focus groups and IDI Multilingual team The KAP study Baseline, endline and impact evaluation Hard-to-reach populations

The Perspicuus research team · 19 May 2025