Over at Hub Blog, Jay Fitzgerald has a good post linking to two commentaries by pollsters. In the New York Times, Andrew Kohut argues that a subtle version of the Bradley effect may have made Barack Obama’s support in New Hampshire look stronger than it was — lower-income white voters who might tend to reject a black candidate are also less likely to answer polling questions in the first place.
But John Zogby, in the Huffington Post, says (at least I think this is what he’s saying) that the pollsters didn’t get it wrong so much as they ran out of time. Uh, OK.
Discover more from Media Nation
Subscribe to get the latest posts sent to your email.
Maybe I missed reading this in the coverage– haven’t been reading much beyond Slate– but if 8,000 votes swung the other way, there’d be much reckoning in Camp Hillary, yet the number of delegates awarded wouldn’t have changed — 8 to 8.Sure, there’s the psychological push of a win, but that’s manufactured by the media, of course.
Kohut writes, “Poorer, less well-educated white people refuse surveys more often than affluent, better-educated whites.”Wait a second. If said people refused to participate in the survey, how did the surveyors make conclusions on their net wealth and level of education?
How can my focus group tell how poor and racist you are if you won’t answer the phone? Plz advise.
Lissa: Kohut explained it quite well, which is why I recommended his column.
No need to get testy, Dan–that was, er, tongue in cheek. Kohut makes a good point. But although he puts it delicately, he’s making a pretty big claim–that poor white people are a lot more racist than rich ones. Is that true? Where’s it coming from? And how does he know, if they won’t talk to his pollsters?Once you start going down the survey-error rabbit hole, all kinds of questions crop up. Don’t young people disproportionately not respond to surveys, too? What about poor black people? Is it a race effect, or an education effect, or an income effect? Are we better at estimating support for black candidates today because people are less racist, or because pollsters have gotten better? There’s a lot of guesswork here.
Lissa: Kohut didn’t say none of them will talk to pollsters. He said fewer of them will talk to pollsters. So you survey the ones who will talk to you and then try to extrapolate and adjust.Personally, I suspect that Clinton never actually lost her lead in New Hampshire, and that we’ll never know why the polls were so wrong.
I get that. Feh. I should have said “if they _disproportionately_ won’t talk to his pollsters.” Kohut’s argument smacks of “I believe this, therefore I massaged my data into supporting it.” There are a truckload of other plausible explanations for why polls screw up, none of them particularly empirically verifiable.