Thursday, October 11, 2007

You drink how much?

So, how many cigarettes did you smoke last week? How many times did you "engage in sexual intercourse"? What do you earn?

Social scientists have a problem on their hands: People don't answer questions on sensitive subjects accurately. One survey of women wound up underreporting the number of abortions occurring nationally by about half, for example. (The answers could be cross-checked against hospital data.) And some researchers have found that people are even less likely to speak honestly about income than sex. Often, they just won't answer. (A curiosity about those sex questions: Asking them how often they "have sex" leads to better answers than references to "sexual intercourse.")

According to Roger Tourangeau and Ting Yan, survey specialists at the University of Maryland and Michigan respectively, the problem is getting worse, because of homo sapiens' growing annoyance over surveys, period. (Think: call screening.) "The need for methods of data collection that elicit accurate information is more urgent than ever," they write in the latest issue of Psychological Bulletin. [Subscribers only.]

The key to accurate data collection, they conclude based on a review of the literature, is to depersonalize the process as much as possible. The goal is to decrease the shame factor. Machines help -- but not phones. Respondents are as unlikely to tell the truth to a fellow human during a phone call as they are to a human sitting on their couch.

When an interview takes place in person, asking the respondent to write down the answer and place it into an immediately sealed envelope helps. Having them type the answer into a computer is even better, some research suggests.

Yet Tourangeau and Yan add that as our relationship to technology changes, survey-takers will have to stay one step ahead of the game. Evidence suggests that survey-takers do not trust networked computers, for example. And computerized voices -- anything that makes the computer seem "human" -- elicit some of the same defensive mechanisms as do queries from live researchers.

There have always been some oddball solutions to these survey challenges: Asking a teenager to blow into a tube and informing him that his answers about drug use over the past month will be verified against the contents, to take one example, tends to elicit truthful responses. But that's a lie, so the technique faces ethical hurdles. Same with the old "I'm carrying a lie detector in my pocket" gambit -- which, surprisingly, works, too. The quest for honest answers about embarrassing matters continues ...

(How much do you make again?)

Posted by Christopher Shea at 02:13 PM

Print | E-mail to a friend | Permalink | Subscribe via rss