The craziest request I ever received as a researcher was to do follow-up interviews with people who had called a Viagra hotline. This was many, many moons ago, when telephone interviewing was still a thing. The client only had a phone number of the person that called. No name. No reason why they called. They imagined that we would call that number back, ask for the person who had called the Viagra hotline, and then ask them to do a quick survey. Unsurprisingly, this project never went ahead. 

Asking about sensitive topics is a common requirement in surveys. But how do you do this when we would never admit to any undesirable behaviour or bad habits to anyone, not even ourselves? 

The reason is social desirability bias, and it is a tough nut to crack. We use it not just to lie to others, but to lie to ourselves. The pressure to perform for others becomes a mental Instagram filter when it comes to ourselves. So when we are asked in a survey how much time we spend on our phones, how much alcohol we drink, or how often we really floss, we face a double dose of deception. We genuinely believe that we scroll less, drink less and floss more than we do. 

The bog-standard fix in research is promising anonymity. And while that promise may be true to the researcher, for the respondent that is largely a myth. Surveys are not scientific instruments but trust games between the researcher and the respondent. Just because you were told that all survey responses are anonymous doesn’t mean you want to be honest about how many glasses of wine you had last week or how often you really went to the gym – and risk being judged for it. In Bruce M. King’s 2022 paper, The Influence of Social Desirability on Sexual Behavior Surveys: A Review,  he shows that even when a survey is anonymous and self-administered, neither good nor bad habits are truthfully reported. 

The next best fix advocated in most research training is indirect questioning. And logically this makes sense. Instead of asking a respondent how many units of alcohol they drink per week, the researcher asks how many units of alcohol the average person drinks. The respondent projects their bad habits onto a hypothetical stranger. But this technique is not a one-size-fits-all. 

In individualistic, low-context cultures like Australia or the UK, communication is direct and explicit. People say what they mean. Projecting a bad habit onto someone else works reasonably well here. But when you try this in low-context cultures, like Japan, this can feel evasive, confusing or even passive-aggressive, as communication relies heavily on implicit cues, social harmony, and reading the room. Ingmar Ehler, Felix Wolter, and Justus Junkermann show these geographic differences in their 2021 meta-analysis, Sensitive Questions in Surveys: A Comprehensive Meta-Analysis of Experimental Survey Studies on the Performance of the Item Count Technique. So we cannot expect a Western psychological parlour trick to translate universally. 

When indirect questioning is not enough, methodologists add statistical noise. One of the methods to do this is the Item Count Technique. It’s done by asking one group how many items they have done from a list of mundane things, while the other group is asked about the same mundane list, plus one sensitive item. So respondents are only asked for the total number and never need to reveal which specific items they have actually done. The statistical logic here is to isolate the bad habit without exposing anyone’s specific answer. It sounds brilliant when you read it on paper. In practice, it is incredibly messy. Ehler and his colleagues found that the technique requires massive sample sizes because the statistical noise creates huge standard errors. It also falls apart if any of the mundane items are accidentally sensitive, or if the respondent says just yes, or just no, to every single item. 

The second major method is the Randomised Response Technique, often called the coin toss method. The respondent flips a coin, but the researcher cannot see how it lands. If it lands heads, the respondent is supposed to answer truthfully. If it lands tails, they have to say yes regardless of the truth. The researcher never knows which rule the respondent follows. Again, the statistics are elegant, but human behaviour is less predictable. Ivar Krumpal and Thomas Voss show in their 2020 paper Sensitive Questions and Trust: Explaining Respondents’ Behavior in Randomized Response Surveys that this method suffers from a fatal game theory flaw. If a respondent does not trust the researcher, both the guilty and the innocent will refuse to admit to a bad habit, just to avoid looking suspicious. The statistical noise does not overcome the social desirability bias. Both the item count and coin toss techniques fail to account for the fact that messy human beings do not behave like mathematical equations. 

This brings us to the virtue trap. Reading through all the papers to find out how we can ask people about their bad habits and get truthful answers, I noticed a blind spot. All these clever techniques are designed to catch out respondents under-reporting their bad habits, but they actually perform much better at catching out respondents over-reporting good habits.  

If we want to understand actual human behaviour, we need to stop treating bad habits as failures and start accepting that social desirability bias is not a flaw to be fixed. Human beings are complex and messy, and the pressure to perform is baked into how we talk about ourselves. There may not be a single technique that allows you to ask people why they called a Viagra hotline, and maybe that is the point. 

Ramona Daniel is a London-based independent researcher and strategist, and founder of cultural insights and strategy consultancy Twenty3, who brings a cultural lens to quantitative research. She helps organisations understand what people do and how culture shapes their choices. Before going independent, she spent more than two decades working at Ipsos, Kantar and across WPP. LinkedIn > TWENTY3 Website >

Square Holes is a cultural insight studio.

We design mixed method explorations of people and culture beyond the category,  uncovering the patterns, tensions and shifts shaping behaviour to inform strategy, inspire innovation and enable confident decisions. Our studio model brings together the right mix of thinkers, researchers and specialists for each exploration. If you’re navigating change, entering a new market, or seeking deeper understanding of people and culture, let’s start a conversation >