He’s Ahead! He’s Behind! Which Polls Can You Trust?
Midterm elections are coming, and each day brings another poll or forecast. How believable are they?
1. Mason-Dixon works with newspapers in states across the country for its research. This firm is methodologically sound and shoots straight.
2. Selzer and Co., which polls independently in Iowa, Indiana, and Michigan on behalf of media clients, has its states’ demographic quirks down pat.
3. The Wall Street Journal’s print reporters are aghast at how media outlets and campaigns conflate the WSJ Online Zogby Poll with the real Journal poll, which is conducted by a bipartisan team of pollsters in conjunction with NBC News. For the online version, John Zogby’s firm culls its data from a geographically and demographically weighted self-selected pool of Internet respondents. There is no hard evidence that the method is valid enough even to be interesting.
4. Thumbs-down to the autodialers. These are the robo pollsters that use the phone book or another database to find respondents who answer questions by punching in digits on a telephone. First the bad: Scott Rasmussen’s Rasmussen Reports. His demographic weighting procedure is curious, and we’re still not sure how he prevents the young, the confused, or the elderly from taking a survey randomly designated for someone else. Most distressing to virtually every honest person in politics: His polls are covered by the media and touted by campaigns that know better.
On the plus side, while skepticism is warranted, SurveyUSA’s poll seems to be on the leading edge of autodial innovation. Its numbers generally comport with other surveys and, most important, with actual votes.
5. Quinnipiac College is an A-list, established, academic polling outfit, but the farther away from Connecticut, the worse its polls become. We don’t know why or how Quinnipiac acquired the experience to poll Florida, a complicated, diverse state that vexes even the most experienced pollsters. Like the autodialers’, Quinnipiac’s results can fluctuate wildly even when nothing indicates the race should be in flux.