It's Not the Weighting In Polls, It's the Loose Voter Screens

I've been talking about loose voter screens for a while.

This guy, Bob Krumm, explains it well.

He has an actual tangible conclusion: For every ten extra percentage points of registered voters a poll deems "likely," Romney's share of the vote drops 1.7%, and Obama's increases 1.7%.

Thus, a poll with a tight voter screen, like Gallup's, has Romney out front with a strong +5 point lead, but a poll like IDB/TIPP, which says that 95% or so of all registered voters are "likely voters," has Romney down 2.

The plot makes for a pretty darned straight line (with a couple of polls away from the trendline).

Jay Cost was discussing this on Twitter last night, and I assume will be posting on the subject, because he seemed pretty hot and bothered about it.

Strong likely voter screens cost money, because pollsters wind up making more calls to get their quota.

But a weak likely voter screen -- like the simple question "How likely are you to vote?" -- is cheap. Because it lets in most people (most people say "very likely," even though that's not true).

And most polls actually use that weak sort of screen.

posted by Ace at 10:04 AM

| Access Comments