So we had two major polls going into the Colorado primaries: SurveyUSA and Public Policy Polling were on the line.
How’d they do against the actual results of Ken Buck over Jane Norton 52-48, Dan Maes over Scott McInnis 51-49, and Michael Bennet over Andrew Romanoff 54-46?
If I had a large sample size to use, I’d evaluate a pollster by taking the root mean square difference of all his results, but I don’t, so all I can do is take the results I have, nod gravely, and make a few comments. Let’s go.
Senate: For the Democrats, PPP had a two candidate split of 53-47 in favor of Bennet, while SUSA went 48-52 in favor of Romanoff. The actual was 54-46 Bennet, so PPP got this one rather close. Point.
For the Republicans, PPP said 51-49 Norton, and SUSA said 45-55 Buck. The actual was 48-52 buck, so SUSA was in the neighborhood, and PPP wasn’t close this time. Call it a tie so far.
And the tiebreak of course is the GOP Primary for Governor. This had to be a tough race to poll, too, with major, late-breaking news. PPP put McInnis up 51-49, and SUSA put Maes up 52-48. SUSA gets the right winner, and is ever so slightly closer to the actual totals, so SUSA wins Colorado 2-1.
House of Representatives Swingometer
Electoral College Swingometer
Actually, for GOP Senate, PPP and SUSA were off by the exact same amount: each poll was off by 3 points for each candidate. The only difference was that PPP was off by 3 points in the direction of the loser (which changed the result), while SUSA was off by 3 points in the direction of the winner.
Um, are you using the raw totals or are you factoring out the undecideds?
If you’re leaving the undecideds in that’s pretty shaky, since there’s no ‘undecided’ total in the final vote.
I was using the numbers given in your post above.
Buck- Actual:52%, PPP:49% (-3%), SUSA:55% (+3%)
Norton- Actual:48%, PPP:51% (+3%), SUSA:45% (-3%)
All those numbers come from your post. Am I misinterpretting what you wrote?
Oh I see.
Well I think getting the winner right counts for something, though in a pure mathematical analysis it may not.