01 November 2008

That Zogby Result

By now many of you have heard the news that John Zogby's tracking showed John McCain up by a point in last night's interviews --- which narrowed the gap to about five points in the three day roll. As willing as I am to grasp at straws at this point, and as hopeful as I know this news makes many of you feel (except you, J.D.), I need to throw a bit of cold water on the story.

There is one glaringly obvious fact that was missing from Zogby's press release: Last night was a Friday. And Halloween.

What difference does this make? All the difference.

As many of you know, in my day job I am a political scientist and public opinion researcher (sad to say, but farming does not pay the family's bills). Before going into business for myself, I spent many years working for a Republican political polling firm. In the last three or four weeks leading up to every election day, we would be in the field nearly every night conducting tracking interviews.

Every night, that is, except Friday and Saturday. Well, occasionally Saturday. But we almost never interviewed on Friday. And during the rest of the year, when a regular poll was in the field, we almost never interviewed on a Friday night. There is a simple reason for this: the kind of person who is home, and willing to sit around answering a series of questions about politics, on a Friday night, is...well...not all that representative of the electorate as a whole.

And then there's Halloween. No matter what day of the week Halloween fell on, we tended to regard that night's interviews with skepticism. If we didn't have to be in the field that night, we didn't interview on Halloween. Simply put: many young and single people are out at parties that night. Go to any big or not-so-big city on Halloween night, and you'll find lots of twenty and thirty-somethings out and about, dressed in bizarre costumes. You'll also find, pretty much anywhere in America, single parents out taking their kids door-to-door. Who's home to answer the phone when the pollster calls? Disproportionately (and that's the key word), married parents of young children (home to answer the door for trick-or-treaters while the other parent takes the kids door to door), or single people with no social life, or empty nesters whose children have grown.

A pollster can weight his sample all he wants to make it representative in terms of age, gender, race, and partisanship. But when the available sample on a certain night is fundamentally skewed on certain other dimensions, that's something more difficult to weight into proportion --- assuming the pollster is even able to identify those other biases in his sample that night.

I'm not saying John Zogby shouldn't have interviewed last night. Heck, we interviewed on Halloween (at least when it wasn't a Friday) at my firm. Rather, what I'm suggesting is that there's a strong possibility last night's sample is fundamentally unrepresentative of the electorate as a whole. I'd love to be proven wrong. But until Zogby reports a similar result in Saturday's interviews, I'm going to remain highly skeptical.

No comments: