# Sensitive Survey Questions: Clear Wording, Privacy, and Choice

A practical checklist for sensitive survey questions: explain purpose and data handling, ask neutrally, minimize detail, allow a skip, and test the wording.

- Canonical URL: https://www.harmate.com/en/blog/sensitive-questions-how-to-get-honest-answers-without-triggering-defense
- Author: Harmate Team
- Published: 2025-12-14
- Updated: 2026-10-19T22:58:34.042840+00:00
- Language: en

## Content

# Sensitive Survey Questions: Clear Wording, Privacy, and Choice

A well-worded question cannot guarantee an honest answer. People may have different reasons to answer, skip, or interpret it differently. The task is to ask only what is needed, explain why it matters, describe how responses will actually be handled, and leave participants a meaningful choice.

The American Association for Public Opinion Research recommends that surveys explain sensitive questions, allow respondents to skip or decline them, and avoid making answers required when that can push people to exit ([AAPOR Best Practices](https://aapor.org/standards-and-ethics/best-practices/)). Wording and setting also matter in ways that are not always intuitive: a U.S. Bureau of Labor Statistics experiment using fictional vignettes found that “forgiving” phrasing was sometimes rated as more sensitive than direct phrasing, especially when judged from an interviewer’s perspective. That study tested perceived sensitivity in specific vignette conditions; it does not show that direct wording is always preferable ([BLS Office of Survey Methods Research, 2015](https://www.bls.gov/osmr/research-papers/2015/st150170.htm)).

## A five-part review before you field the question

**1. State the decision and purpose.** What will the answer help someone decide? If you cannot name a use, remove the item or make the need explicit before asking.

**2. Reduce the request to what is necessary.** Ask one thing at a time. Set a specific period or event when recall matters. Avoid names, exact dates, or details that the analysis does not need.

**3. Explain the real handling.** Say who will see raw answers, how they will be stored or reported, and whether comments may be quoted. Do not call a survey anonymous if responses can be linked to a person. “Confidential” is also a promise about actual access and handling, not a calming adjective.

**4. Preserve choice.** Let people skip, decline, or choose “not applicable” where appropriate. If the question is required for a necessary process, explain that requirement and consider whether there is a safe alternative route. Do not imply that an optional answer is consequence-free unless that is true.

**5. Test interpretation and context.** Ask a small number of people similar to the intended respondents what they think the question means, what they would need to know before answering, and whether any wording is unclear. A pretest can reveal problems; it cannot certify that future answers are unbiased.

### Fictional example: raising a workload concern

A fictional team wants to learn whether its escalation process needs changing. Instead of asking, “Is your manager the reason you are overloaded?”, it explains that the survey concerns workload escalation and who will review results, then asks: “In the past month, what made it easier or harder to raise a workload concern? Please do not include names. You may skip this question.” The team keeps both positive and negative accounts and does not treat the answers as a measure of any person’s motives.

The rewrite narrows the request to a process and a time period. It does not make disclosure risk disappear or establish how common the concern is. If the survey cannot protect raw comments as described, change the data handling or the wording before collecting answers.

## Keep interpretation modest

Use open-ended answers to understand the range of experiences and the language people use. Preserve excerpts, differences, and unanswered questions. Do not present a small set of comments as representative of everyone or claim that a neutral phrase removed bias. If you need a prevalence estimate for a sensitive behavior, a formal indirect survey method may be relevant, but it requires a separate statistical design; see [uses and limits of indirect open-ended questions](/en/blog/measuring-sensitive-topics-without-naming-them-the-art-of-indirect-open-ended-question-design).

For an additional check, compare your draft with the guidance on [social-desirability bias](/en/blog/social-desirability-bias-why-your-surveys-look-great-and-fail-anyway). A careful question makes its purpose, limits, and choices clear; it cannot promise the “truth.”