In 1982, Californians voted for their next governor. Exit polls favoured Tom Bradley, the African American Mayor of Los Angeles[i]. However,George Deukmejian(his white opponent) ultimately won the race. Experts attributed the unexpected outcome to the “Bradley effect”[ii]– the idea that some voters lied to pollsters about their intentions to avoid the appearance of racism (a “social desirability bias”[iii]). Other voters may have simply refused to answer the question out of discomfort.
Unfortunately, market research surveys with “sensitive questions” are vulnerable to the same biases! Sensitive questions can relate to racial attitudes, political affiliations, drug use, sexual behavior, personal health/hygiene, income, age, gender, marital status, and beyond. They can also vary by region and culture (e.g. Muslims may recoil from questions about alcohol consumption).
But there are tricks to mitigate such biases, or at least lessen their impact on the broader survey. When posing sensitive questions, be sure to:
- Guarantee anonymity and privacy: Respondents are “more willing to share sensitive information”[iv]when they know the survey is anonymous and private. So make this clear at the start of the survey and before any sensitive questions. You may even want to emphasize all the safeguards in place to ensure anonymity and privacy. Op4G, for example, storespersonal and survey data in physically separate servers, delivers data to clients via a ‘pseudonymous’ framework, and encrypts all data using Secure Socket Layer technology[v].
- Use online surveys: Research suggests that surveys administered via the web (notusing a human interviewer) greatly reduce sensitivity effects. Respondents feel less pressure to “project a positive image”[vi]than they would in front of an actual human being, making them more likely to reveal “private and socially undesirable information about themselves”[vii].
- Pay attention to language: With thoughtful wording, a question can make a supposedly taboo answer appear more normal or acceptable (known as “normalization of the sensitive issue”[viii]). For instance, instead of asking “Do you shower daily?”, try “Some people don’t have the time, or need, to shower daily. How often do you shower?”
- Place questions strategically: If you put a sensitive question near the startof the survey, respondents may be turned off and abandon the survey completely[ix]. Similarly, if you put a sensitive question at the end, respondents may be left with a negative or suspicious feeling, leaving them less inclined to participate in follow-up studies[x]. Hence, the sweet spot for sensitive questions is half to two-thirds of the way through. That way, the researcher can “draw the respondent into the response process and build a base level of trust before sensitive topics are raised”[xi].
- Pose indirect questions: Rather than asking respondents questions about themselves(or loved ones), try asking them questions about others, such as acquaintances[xii]. Assuming that the respondents know the acquaintances’ history/behavior, they are likely to provide more accurate answers. After all, the respondent is not vulnerable to personal judgment! Sample questions include “Does your neighbor recycle?” or “Is your friend divorced?”
- Validate answers: If you have actual data on the respondent, you may be able to confirm or disprove their given responses[xiii]. To join Op4G[xiv], for example, all panelists must answer questions on income, age, gender, marital status, political views etc. This information can serve to validate but also to filter respondents from the outset - so you never have to pose the sensitive question!
If these tips aren’t enough, here’s the next best solution: add an “escape response” like “I’d rather not say”[xv]to your options. If respondents choose this option, you will not gain information on the sensitive topic. However, you may spare respondents embarrassment and they may be more inclined to complete the rest of the survey.