According to CNN, both Mitt Romney and Rick Santorum are set to win 6 delegates thanks to their close 1-2 finish in the Iowa caucuses. Ron Paul fell to third place. He didn’t win, and he fell further and further behind as the votes were counted last night.
Not one poll projected Ron Paul to drop to third, and PPP stuck to Paul being in first. I said those polls weren’t predictive. They weren’t. I was right.
I went out on the limb. It would have been easy to hedge when every poll out there disagreed with me, but I didn’t. Even when I got a wave of critical commenters via Drudge Report and the Atlantic, most telling me I was completely wrong, and that there in fact would be a surge of voters, I stuck to my guns. I denied there would be a big surge from the record 2008 turnout, and in fact turnout was at about 120,000, roughly the same as the 119,118 of last time. I kept saying it even on caucus day.
The results? PPP projected Paul 20, Romney 19, Santorum 18. Opposite order. Wrong. I said I “debunked” PPP, and the polls were indeed bunk.
InsiderAdvantage. Des Moines Register. Rasmussen Reports. NBS News/Marist. CNN/Time. Every one of them had Ron Paul in second place, down by only 1, 2, 1, 2, and 3 respectively. This was consensus in the turnout models, which I rejected to the end.
The final result was Mitt Romney 25%, Rick Santorum 25%, and Ron Paul 21%. Not one pollster had Paul in third, four off the lead. They had him +1, -1, -1, -2, -2, and -3, in first or second place. This was a total poll failure, and I took a lot of heat for saying they were all wrong.
But I called it. Was I sweating every minute Tuesday evening when the early counts had Paul ahead? Of course I was. But, in the end, there was no massive, unprecedented influx of independents and Democrats backing Ron Paul. There was no turnout doubling 2008’s record turnout. Ron Paul lost, and he finished in third.
If this happened by random chance, one would have thought at least one pollster would have had Santorum in second. At least one of the errors should have had Ron Paul with an even bigger deficit than four. No, if they were random errors, they had a 50/50 chance of being larger or smaller than the actual deficit. For six polls to all go the same way had only a 1.5% chance of occurring. Even worse when you consider IA and PPP had Paul tied or in the lead in previous polls as well. Eight polls all with a smaller deficit only happens by chance 0.4% of the time.
This wasn’t random. The polling of Iowa was systemically wrong. Polling is a useful science, but caveat (e)lector. Do sanity checks. Then you won’t be part of the herd running off the cliff, as happened with Iowa polling this year.
House of Representatives Swingometer
Electoral College Swingometer
The polls were not wrong. “+1, -1, -1, -2, -2, and -3:” these are all within what a reasonable margin of error would be for a survey of those sorts. And the difference between Romney’s 1st place and Paul’s 3rd was just over 3%.
“PPP projected Paul 20, Romney 19, Santorum 18. Opposite order. Wrong. ” Again, if you read the report it has a margin of error of 2.7%.
The Des Moines Register poll made clear that the pattern of voter support changed across the polling period. The accompanying analysis emphasised the fact that if you looked at the second half of the period, Santorum was in second and Paul in third.
That doesn’t suggest that the polling was “systemically wrong”; it suggests that support patterns changed quickly in the final days… and DMR picked that up.