Over at the Atlantic, James Warren brings us shocking news from a recent meeting of the Midwest Political Science Association: sometimes people cheat on internet surveys. When they correctly say that John Roberts is Chief Justice of the Supreme Court, it might only be because they googled it or asked their daughter-in-law:
[Brad Gomez of Florida State University] noted the tendency for a relatively small but apparently rising number of survey respondents to cheat on Internet and mail surveys. When it comes to the Internet, "It's pretty clear people are cheating," especially when they at first don't know an answer. Robert Luskin of the University of Texas referred to "cheating on steroids" when it comes to both Internet and mail surveys, with respondents perhaps googling a response or asking a nearby family member for help. "You're not necessarily getting the respondent who was randomly selected," he said.
Via Jonathan Bernstein, Lynn Vavreck calls a foul. She did a study last year that recruited subjects at the CBS Research Facility in the Las Vegas MGM Grand Hotel. Half were interviewed face-to-face, and half answered questions on a computer connected to the internet:
Of the 505 people who completed the survey on a computer, only 2 people cheated by looking the answers up on-line. That’s less than one-half of one percent of the respondents....Plenty of people had a hard time answering our fact-based questions, and they knew they were on the Internet, yet very few of them took the time to look up the answers — in fact, almost none of them.
....In the spirit of the popular television show Myth Busters, consider this myth busted!
Hmmm. What do you think of this? Vavreck is right that if you want to claim there's cheating going on, you need to produce some evidence. At the same time, I don't find her study very convincing. There's a big difference between a formal setting like hers and the more relaxed confines of your own house and your own computer. I'd expect a whole lot less cheating in her study than in real life. Hell, I'd probably cheat more at home.
Still, I'm curious what evidence Gomez and Luskin have that cheating is on the rise. As near as I can tell, Americans have an abysmal knowledge of just about everything, and it seems to be abysmal on nearly every survey ever done. What's more, most of us don't seem to give a damn enough about this stuff to even bother cheating. Still, I think we could find out for sure. Vavreck could do a study where half the respondents were interviewed in person (or over the phone) and the other half got questionnaires via mail or the internet. As long as all of them were recruited over the internet, and chosen randomly, better scores from the internet half would suggest cheating, wouldn't it?