Demographic survey questions: inclusive examples, coding and privacy

Demographic survey questions describe characteristics such as age, location, education, income, disability, race or ethnicity, sex, gender, language, household composition or work role. Ask only variables tied to a defined analytical or operational decision, use standards appropriate to the population and jurisdiction, let respondents self-identify, include a genuine nonresponse route where appropriate, and test whether combinations of answers could reveal a person.

What are demographic survey questions?

Demographic survey questions collect respondent or household characteristics used to describe a sample, evaluate coverage, compare subgroups, set quotas, weight results or route services. A variable belongs in the questionnaire only when the research plan names its purpose, required precision and reporting treatment.

The Office for National Statistics treats ethnicity, religion and national identity as self-identification measures and advises that adults answer directly. U.S. federal race and ethnicity standards changed in 2024, including a combined question, multiple category selection and a new Middle Eastern or North African minimum category. The Census Bureau plans to introduce the updated approach in the 2027 American Community Survey and 2030 Census, which demonstrates why teams must check both the current standard and the program's implementation status instead of copying a list.

Purpose-before-question matrix

Question-design rules

1. Name the analysis or decision that needs the variable.
2. Use the respondent's own identity rather than interviewer inference.
3. Separate analytically different concepts, such as sex and gender, when both are genuinely required.
4. Use categories from the correct current standard when comparability is required.
5. Allow multiple selections where identities are not mutually exclusive.
6. Add a write-in only when the team has a coding, privacy and reporting plan for it.
7. Use “prefer not to answer” as a valid response, not an error.
8. Distinguish refusal, don't know, not applicable and routed-not-shown in the data.
9. Explain why a sensitive question is asked when context would help respondents.
10. Pretest labels, translations, order and mobile presentation with the target population.

Reusable question patterns

These are structural patterns, not universal category lists.

Age

For an exact-age need, ask “What is your age in completed years?” Use numeric validation, a documented minimum and maximum, and “Prefer not to answer.” Derive reporting bands from the retained exact value only when collection is necessary and policy permits; otherwise ask the analysis bands directly.

Location

Ask for the least precise geography that supports the analysis: country, region or the first part of a postal code. Do not collect a full home address for regional reporting.

Education or income

Anchor categories to the relevant local education system, currency, period and unit. “Household income before tax in the past 12 months” is a different measure from personal monthly take-home pay.

Race or ethnicity

Use the current jurisdictional standard required by the study and allow self-identification. In the United States, check the 2024 SPD 15 implementation status for the program. In the UK, use the appropriate country-specific harmonised question rather than assuming one list fits England, Wales, Scotland and Northern Ireland.

Sex and gender

Ask only the concept the analysis requires. Explain the reference when needed, allow self-description under the applicable standard, and do not use one variable as an undocumented proxy for another.

Disability

Measure the relevant activity limitation, access need or official concept rather than asking for a diagnosis list merely to appear comprehensive. Accessibility accommodations should not depend on disclosure of unnecessary medical detail.

Work role

Define whether the study needs occupation, job level, department, employment status, industry or purchasing responsibility. “What is your role?” without a frame mixes unrelated concepts.

Worked coding example

A study needs broad age bands for reporting and must retain refusal separately. The displayed choices are 18–24, 25–34, 35–44, 45–54, 55–64, 65 or older, and Prefer not to answer. Store stable codes 1 through 6, code refusal as 98, and reserve blank for routed-not-shown. Do not code refusal as zero or include it in an average.

If the reporting plan later merges 18–24 and 25–34, store the derived band in a new field with its transformation version. Do not overwrite the collected response. This preserves the original instrument and makes every published grouping reproducible.

Disclosure-risk review

Each variable may look harmless alone while combinations identify people. A rare job title, small location, narrow age band and language can point to one employee. Before release, create the intended tables, count each cell, test differencing across related tables and apply the organization's suppression or disclosure-control rules. Do not promise anonymity when the design retains linkable identifiers.

Eighteen-field variable specification

Store variable_id, concept, purpose, population, jurisdiction, standard_name, standard_version, question_text, response_mode, displayed_options, stored_codes, multiple_select, write_in_rule, missing_codes, routing_rule, retention_period, reporting_rule and disclosure_control. Link translations to the same variable and version without assuming literal wording is equivalent.

Twelve release checks

1. Every demographic variable has a named analytical or operational purpose.
2. The current standard and jurisdiction are recorded.
3. Respondents self-identify; no characteristic is inferred from appearance or name.
4. Concepts such as sex, gender, nationality and ethnicity are not conflated.
5. Multiple selection works where the specification allows it.
6. Write-ins have a coding and privacy plan.
7. Refusal, don't know, not applicable and routed-not-shown have distinct codes.
8. Labels and codes match across questionnaire, export and reporting.
9. Translations were reviewed with target-language respondents.
10. Mobile and assistive-technology presentation exposes every option.
11. Planned subgroup tables pass disclosure-risk review.
12. Derived groupings preserve the original response and transformation version.

Transparent method

We rechecked the sources on 21 October 2026. The inclusion and reporting principles use current UK Government Social Research guidance. The self-identification and country-specific category boundary use current ONS harmonisation guidance. The U.S. example uses the Census Bureau's current description and implementation status for the 2024 SPD 15 revisions. The purpose matrix, question patterns, eighteen-field specification and release checks are editorial synthesis.

Limitations

Inclusive wording does not guarantee representative sampling or adequate subgroup precision. Official standards can lag community language, yet custom categories can break comparability. Some programs are legally required to collect or report specific variables; others should not. Sensitive characteristics, small groups, minors, health research and employment settings can require additional legal, ethical and community review. This guide does not supply a universal list or legal determination.