Alternate title: "Surprise (Not): Barone Exposes How Exit Poll Samples Are Typically Biased."
Early this morning, at the Washington Examiner, Michael Barone casually put out what is apparently a well-known fact in polling circles. I'm thinking that it's not at all well-known to the general public (bold is mine):
But if you think the exit poll (from the Milwaukee Journal Sentinel, showing a dead-heat race earlier in the day -- Ed.) was 4% too Democratic—and that’s in line with exit poll discrepancies with actual vote results over the last decade, as documented by the exit poll pioneer, the late Warren Mitofsky* — that result looks more like 49%-47% Romney. Or assume the remaining Milwaukee County precincts whittle Republican Governor Scott Walker’s margin over Democratic Milwaukee Mayor Tom Barrett to 53%-47%, which looks likely, the Obama-Romney numbers would look like 48%-48%.
Really? Supposedly "random" exit polls are skewed 4% Democratic? Who knew?
The explanation of Barone's asterisk:
Mitofsky found that the biggest WPE, Within Precint Error, where exit poll results tilted most heavily toward Democrats compared ithe (sic -- it appears that it should say "with" -- Ed.) actual vote, were in precincts where the interviewers were female graduate students. Go figure.
Well, there you have it. I guess exit polls will continue to skew towards Democrats as long as female graduate students do the polling. So maybe we need a demographic breakdown of polling personnel before we can evaluate a poll's credibility.
What's interesting to me about this (maybe not to anybody else, but work with me on this) is that I definitely recall the networks telling us for many years that exit polling wasn't "random" at all, but instead was based on polling at "key precincts." When did it change to "random"?
It turns out that some people, like the person writing things up here at MSNBC's FirstRead, who specifically refers to an example involving 50 "key precincts," that it's still how things are done. Apparently not, according to this research paper, which says that the "key precincts" approach went away in 1986. But the First Read item does raise the valid point that the candidates favored by absentee and early voters my differ from those favored by Election Day voters.
Zheesh. What a mess. That would argue for not bothering with exit polling at all.
Barone makes perfect sense when he claims that Wisconsin is in play in November, no matter what the rest of the delusional press wants us to believe.
Cross-posted at BizzyBlog.com.