Ah, I remember it well…It was 1990 and I was in my second year of graduate school at Duke. I was working with a couple of people at UNC, and Tony Oberschall’s running seminar on social movements was in full stride. We were giddy with excitement. Friends working over at the Carolina Pop Center and the research center there were certain, absolutely certain, that Harvey Gantt (an African American businessman and former Mayor of Charlotte) was going to win the Senate race against Jesse Helms in a landslide. He was up on Jesse Helms by 8 points with only a few weeks to go. Helms dismissed the polls saying “The only people who are in favor of my opponant are a bunch of liberal, communist, homosexuals from Chapel Hill.” We had t-shirts made up, proclaiming “Another Liberal, Communist, Homosexual from Chapel Hill for Gantt.” Of course, Helms wound up pasting Gantt. To this day, everyone misreads why. The popular explanation by political pundits and people in the polling industry is that people misstate their preferences. Some have taken to calling it the “Bradley effect” to note only one of MANY instances where prospective polls have shown a clear lead for an African American candidate, while the reverse won out in the election.
Pew followed up on this with some interesting research on “reluctant respondents”, only the Pew guys (I’ve never met a Pew woman, have you?) are too interested in whoring out science for media attention and contracts to correctly interpret the importance of this research. What they found is that non-respondents are more conservative and more racist than people who answered the questions voluntarily (they bribed the reluctants to get them to come into the sample). Well, that was more than a decade ago, and almost two decades since Helms-Gantt. Since then, response rates from political and commercial polls have gone down the toilet. People even try to pass this crap off as science–with Pew, Gallup, Baylor, and others having press conferences to report their astounding new findings…based on a survey with a response rate under 25% (and a very generous method of calculation to get to even that pathetic total). It’s even more hilarious when they then start making a fuss about “margins of error” in these hack polls. The old Carolina Poll that had Gantt solidly in the lead (nearly double the margin of error) was a pretty good effort, with a 55% response rate. And it got it WRONG. Way Wrong. These schlocky Gallup/Pew/Baylor efforts aren’t qualifiable as science, but I’d bet they have response rates at least 5 times as great as those of the standard Rasmussen\Zogby\insert your beltway bandit firm here.
Everything we know about survey non-response bias (which isn’t much) tells us that the people who aren’t responding are more conservative. Conservatives are assholes. They hate the media. They hate liberals, and they most of all hate liberal women. When some liberal communist homosexual media type (as they think) calls them on the phone during their white bread and fried chicken dinner, they hang up. Liberals feel sorry for the poor sap on the other end of the line, and may even think that public opinion is something of a civic duty.
Fortunately, the response rates for these crap polls are so low that we can have some hope that the normal biases won’t hold. That means that the polls are completely and utterly worthless. But, if they have any inkling of value, then Obama is going to lose big.