Post-election brain dump
I'm not going to spend a lot of time on any one topic but instead just get a bunch of stuff out that's twirling around in my head. Each of these topics is probably deserving of it's own post, but that will come later. Right now I have to get this stuff out of my head so I can move on with my life.
Outcome Prediction System
While returns were coming in I was plugging the numbers into a formula to try and determine a final vote split. The last update I made on election night predicted the final outcome as being (unofficial results in parenthesis):
Mark Dayton: 44.9 (43.6)
Tom Emmer: 41.8 (43.2)
In the end it didn't work quite as well as I had hoped, although it wasn't a complete failure. It missed Mark Dayton by a point and Tom Emmer by a point and a half, but it was the final margin that I cared about and in that regard it missed by enough for me to be unhappy with the results.
If I plug the numbers in right now it comes out a little bit closer, but the point of the system was to provide a real time election night projection of the final vote split, not to be correct after the fact. If I attempt this type of thing in the future it will have to be at a precinct level instead of a county level.
Polling and the GPI
The polling showed a small but consistent Mark Dayton lead throughout the month of October, the end result though, was basically a dead-heat, with less then half a point separating the two candidates. Let's take another look at the final GPI.
Just like in the 2002 and 2006 Governor's race the final polling average significantly underestimated the performance of the GOP candidate. This was something I was aware of, but had no way of knowing weather it was a fluke or a trend. Now that we have three races in a row that this has happened, I have to conclude that it's a trend.
So then the question becomes, why? I don't have anything more than theories at this point; it could have something to do with the nature of three-way contests or it could have something to do with the mid-term electorate in Minnesota or even the likely voter screens employed by the pollsters.
The pollster who came closest to predicting the final margin, SurveyUSA, was the same pollster who saw CD8 as close and got CD1 pretty close. It's possible that Minnesota has a shift in the electorate between Presidential and mid-term elections that is different than normal and other pollsters aren't picking up on.
This topic certainly deserves more attention, but it won't be in this post.
Turnout
Secretary of State Mark Ritchie predicted 60% turnout, which fits with historical trends. I predicted it would be a little higher 62.5%. We were both wrong. Right now turnout stands at just over 55%, which is about what it was in 1994.
The conventional wisdom is that lower turnout benefits Republicans while higher turnout benefits Democrats and this election certainly won't do anything to change that perception. Based on the results of the Minnesota Senate and House I would guess that the 5% that stayed home were Democrats or Democratic leaning voters.
MN House and Senate Predictions
Some of these races will go to a recount (In 15B, for example, King Banaian is ahead by just 10 votes) and obviously none of the results are yet finalized, so this analysis assumes that everyone who is currently ahead, stays ahead.
Now is where I get to toot my own horn a little bit, not that I'm happy about it. Every Senate seat that I labeled either Safe or Tilt stayed with that party except one. In SD51 I had Don Betzold as Tilt-DFL and he lost to Pam Wolf. Amazingly the GOP won every single Senate race I had rated as a toss-up.
On the House side there were four seats that I had labeled as Tilt DFL that the GOP won; 2B, 27A, 30B and 41A. Three of those races probably should have been labeled as Toss-up, and the scoring system I used to give me my ratings labeled them as such. But there were things about either the candidate or the district that made me move the race to the Tilt column.
In 27A Robin Brown got killed in fundraising, but that's what Robin Brown does, she gets killed in fundraising, it didn't stop her from winning before. 30B was Andy Welti's seat, which he took from an incumbent who had been regularly been getting around 70% of the vote and while Welti had always had close races, he had been gaining vote share in every election. In 41A, well, that was a completely subjective call on my part.
Even with those four mis-calls, the House ratings were still rather accurate, I got all the Safe seats correct (some barely) and all the Tilt seats except those four already mentioned. It was the Toss-ups that the GOP killed us in. I identified 32 house races as Toss-ups and the GOP won those races 24-8.
It looks like the GOP will have a Senate majority of 37-30 and a House majority of 74-60, not overwhelming, but also not razor thin.
As far as the ratings go, I didn't push as many of the "Toss-up" races into Tilt categories as maybe I should have, but that had more to do with uncertainties in the model and not wanting to be too aggressive in pushing races. My plan was to do an update before the election with the new fundraising info, but I just didn't have the time.
The other thing I want to do in the future, and what would have helped make these ratings more accurate, is to include generic ballot numbers.
Conclusion
I don't really have much in the way of wrap up. It was a bad election night all around. I did not conceive of the possibility of Mark Dayton winning the governor's race, the DFL winning all of the statewide constitutional offices and the GOP taking over both houses of the state legislature.
And now we get to go through another recount process, with the state GOP having an even greater incentive than last time to drag this out for as long as possible. Oh happy day.
No comments:
Post a Comment