Short of spying on people in motel rooms or honeymoon suites, which academic institutions frown upon, the only way interested parties can find out about people’s sexual behaviour is self-report surveys. There’s only one problem: people lie like germ-ridden bearskin rugs. Trying to figure out how to get them to lie less is one of the uphill battles of sex research methodology.
Most of the lying on sex surveys is a function of what researchers call “social desirability bias.” In some of the first attempts at gathering sexual data from a population, by pioneering American scientist Alfred Kinsey in the 1930s, over 17 percent of white college men reported that their erect penises were more than 7 inches long. (The average is 5.1–5.9 inches.) When it comes to what you do with a 7-inch penis, the responses tended to reflect what men thought their sex lives should be like rather than what they actually consisted of.
Recently, AIDS researchers attempting to gather accurate statistics to help prevent the spread of HIV in the developing world have been comparing different ways of asking people about sex. In a recent study in the journal AIDS and Behavior, a research team from the U.S. and Botswana, headed by Philip Anglewicz of the Tulane University School of Public Health and Tropical Medicine in New Orleans, assessed how three different ways of asking questions elicited different responses from a sample of men and women in Botswana. They compared three types of information-gathering: face-to-face interviews, ballot boxes, and randomized response technique.
Face-to-face is the old-school standard, in which an interviewer meets up with an interview subject and asks questions, to which the interviewee responds out loud. The advantage here is that if your subject is sitting right in front of you when you ask the questions, they’re likely to respond to all of them. However, these surveys don’t allow for any kind of anonymity, and most people are pretty squeamish about answering questions like, “Have you ever had sexual relations with an animal?” Ballot box interviews also involve an interviewer and an interviewee in a room, but in this case, the interviewer asks the question aloud and the interviewee writes down their response on an unsigned piece of paper and puts it into a sealed container.
Randomized response technique, or RRT, is ingenious. Where face-to-face and ballot box methods are straightforward takes on question-and-answer, RRT is more like the riddle in Labyrinth with the two-headed knights, one who always tells the truth, and one who always lies. There are a few different versions of RRT, but in the Botswana study, they used this method:
Respondents were given two dice, a box into which to throw the dice (so the interviewer could not see), and a card that instructed them how to respond to questions based on the numbers thrown. If the total of the two dice rolled was 2, 3 or 4, the respondent was instructed to always respond ‘‘yes’’—even if the truthful answer would have been ‘‘no.’” Conversely, if the total on the two dice was 11 or 12, then the respondent was to always respond ‘‘no’’ regardless of the truthful answer. If the total was 5, 6, 7, 8, 9 or 10, the respondent was instructed to answer truthfully.
In each case, the interviewer doesn’t know whether a subject’s response is truthful or not. But later, back in the office, the researchers can use probability theory to determine what percentage of the answers are forced—the likelihood of a toss of two dice resulting in 2, 3, 4, 11, or 12, are finite and can be calculated. (It’s easier if you picture a coin toss. Because we know the probability of a coin toss coming up heads versus tails, once the researcher has gathered enough responses, the aggregate can tell the researchers what the breakdown really is.)
The Botswana study is the first to assess how effective RRT is for gathering self-reported data about HIV/AIDS in sub-Saharan Africa. They gathered 1,417 subjects from different geographical areas, and did face-to-face interviews with 518; ballot box surveys with 500; and randomized-response surveys with 399. Respondents rated their agreement or disagreement with statements like, “Condoms are for promiscuous people,” “Condoms actually carry the HIV virus,” “Sometimes have been too drunk to remember partner,” and, “Sex becomes boring with same person over time.”
The researchers found that people who answered questions about their sex lives through RRT gave very different responses from those participating in face-to-face interviews or secret ballots. They write: “Both men and women report less condom use in the [RRT] method compared with [ballot box] and [face-to-face]. Significantly larger percentages of respondents in the randomized response sample report still being currently involved with their past sexual partners: more men and women report still having sex with their second and third most recent sexual partners.” Essentially, when you ask the questions in this way, people report having more, and less safe, sex.
I love how this method solves seemingly intractable touchy-feely human problems—people are shy, and people are liars—with math. It lets researchers find out important information to use for medical advances, but it also releases the people answering questions from their own self-consciousness about their sex lives. Seems like there’s something freeing about rolling the dice and letting fate help you answer the questions.