Results from Prognostications #6 and #7: The non-binding caucuses in Colorado and Minnesota

There was only one company who tried to poll in Colorado and Minnesota prior to yesterday's caucuses, and given that lack of data Nate Silver declined to make a prediction.  While I consider myself a better mathematician than Silver is, I am not so foolish as to think there is nothing I can learn from him.  My prediction accuracy was stunningly bad, worse by far than anything I had done before.  There are several factors at play, but having so little data is certainly a major factor.  Without more fresh polling data, I will not make a prediction in future contests.

Let us first see the gory results. My updated predictions are better than my earlier stuff with less data, but none of them are even close to good. The actual numbers are first, my guess is second, the difference between them is in parentheses.

Colorado caucuses

Santorum 40.2% 28% (off by 12.2%)

Romney 34.9% 36% (off by 1.1%)

Gingrich 12.8% 22.5% (off by 9.7%)

Paul 11.7% 13.5% (off 1.8%)
75.2% accurate, a C on the standard grading scale

===========

Minnesota caucuses

Santorum 45.0% 34% (off by 11%)

Romney 16.9% 23.5% (off by 6.6%)

Gingrich 10.8% 22%(off by 11.2%)

Paul 27.1% 20.5%(off by 6.6%)
64.6% accurate, a D on the standard grading scale

As Charles Barkley might say, I was turrrribull last night.  Let me list some of the reasons I sucked so bad.

1.  Not enough data.  I brought this up already.
2. Caucuses! We hates 'em!  There are so few voters at caucuses, they are hard to find by the methods pollsters use and all other things being equal, polls do a better job of estimating the votes in a primary, where the percentage of people casting votes is dramatically greater.
3. The elusive Santorum caucus voter.  In the caucuses in Iowa, where there was a ree-donkulous amount of fresh data, no poll told us Santorum was a front runner. Even the polls that put him in second put him a distant second. When all the votes were counted, he beat Romney in a photo finish. In two caucuses last night, again Santorum over-achieved spectacularly over the polls' guesses.  Maybe people who vote for Santorum don't answer phones.  They must be at prayer meetings or something.

And so we have a caucus in Maine that ends this weekend, but so far nobody has any poll data. I won't make a prediction without three different companies giving their best guesses by this Friday.  if that doesn't come to pass, the next contests that count are February 28 in Arizona and Michigan.  I'm certain there will be boatloads of polls in those two states, so my next prognostication against Nate Silver will probably have to wait until then. I can boldly predict that I have no clue right now what will happen then.  Former British P.M. Harold Wilson said a week is a long time in politics, and he was not subjected to the 24 hour cable news cycle.  Three weeks might as well be an eternity now.