Debunking PPP in Iowa

I’ve been telling people PPP’s polling Iowa was making nonsense predictions. The pollster has doubled down.

I’m going way out on a limb here, and if the actual results refudiate what I’m saying then I’m going to have to take some taunting, but I just don’t see how this poll remotely reflects reality, and I’m flatly saying it’s not predictive of the caucus results.

The facts: 597 LVs, no mention of mobile handling. MoE 4.

The results of this poll are going to get huge play. While they predict less for Mitt Romney than the apparent Rasmussen outlier for Romney, PPP seems to show an apparent rush of support from Newt Gingrich to, well, every other candidate. Six candidates are in double figures, three at precisely ten: Rick Perry, Michele Bachmann, Rick Santorum tie for fourth. Gingrich comes in third at 14, his lowest Iowa poll result in six weeks. Mitt Romney checks in at 20, and Ron Paul finishes first at 23. Between the Paul lead and the Gingrich crash, this poll will generate lots of web hits for PPP.

The problem with the poll is that it’s just not likely to be true, though. We have a benchmark for evaluating this poll: 2008 Iowa caucus entrance polls. The partisan alignment is all wrong: In 2008 the caucuses, being closed of course, included 86% self-identified Republicans, 13% self-identified Independents who presumably registered Republican to caucus, 1% Democrats, 1% “Other.” PPP’s poll drops the Republican proportion to 75%, raises Independents to 19%, and raises Democrats to 5%. Guess who’s helped by both of those shifts, which are far outside the Margin of Error and so predict genuine, large shifts in the partisan makeup of the closed Iowa caucuses. That’s right: Ron Paul, who wins 40% of Democrats, 34% of Independents, but only 19% of Republicans according to the poll.

There are three broad possibilities: The 2008 entrance polls are wrong. The 2012 Republican caucuses will find huge new turnout from independent voters showing up and registering Republican. The PPP poll has systemic issues and is not meaningful.

Other suspicious bits: Do we really believe that 36% of “Very liberal” Iowa caucusers went for Mike Huckabee and not Rudy Giuliani or John McCain? Do we believe the Republican Party’s makeup has shifted so that John McCain would have tied for second in Iowa in 2008? That’s what PPP says: Huckabee 26%, Romney and McCain 19%. Remember that the actual result was Huckabee 34, Romney 25, Fred Thompson 13, McCain 13.

Polls with solid fundamentals can be tweaked with post weighting to correct the results. However the rule of Garbage In, Garbage Out applies. The PPP results are disconnected with any sort of reality we can compare them with. Reports of a Ron Paul takeover in Iowa are nonsense until proven otherwise. Because fundamentally: Do we really think a horde of Independents who hate the Republican Party will show up at the last minute, register Republican, and caucus for hours with Republicans? I don’t buy it, and neither should Iowa Republicans.