Crunching the Nyah Nyah Numbers
Back in mid-September, I began the LibertyPell Nyah Nyah Senate Majority Index. It was so named because of a belief, both then and now, that the outcome of the midterm election would make no difference to our daily lives. It would only determine whether the playground taunts were to be delivered by the Democrats or the Republicans.
Each week I looked at nine pundits and averaged the results. There was no great magic to the sources as most would have selected The Cook Political Report, the Rothenberg Political Report, Larry Sabato’s Crystal Ball, Nate Silver 538, the Monkey Cage (The Washington Post Election Lab), Upshot (The New York Times), Princeton’s Sam Wang, Real Clear Politics and an aggregator of British bookmaking websites. Part way through the process I began following the predictions of The Huffington Post, the Daily Kos and Politico.
I chose eight states — Alaska, Arkansas, Colorado, Georgia, Iowa, Kansas, New Hampshire and North Carolina — then began to follow Louisiana and Kentucky when they got interesting. Seven of the first eight went for the Republicans, as has one of the last two. These were the margin of the Republican win.
It was astonishing how accurate were the sources of information and how consistent they were with each other. I spoke with an exit pollster on Tuesday morning after watching him get rebuffed by most of those who are leaving the polls and hurrying to work. We discussed my standard practice of hanging up on pollsters when they call me at home and agreed that it was extremely difficult to gather information.
Then there is the math, which I still do not understand except to respect its complexity.
Though most people wonder how the pollsters get some of them wrong, I am amazed at how many they get right beyond the 50% you could expect from flipping coins.
With that as backdrop, there is no reason to go through which sources correctly called which states. Most did most of the time.
Out of 100 final picks, 77 were correct. The consensus approach cancelled out all of the errors except North Carolina, where everyone was wrong and Kansas, where only the Monkey Cage was correct. Here were a few mistakes.
Across the board, everyone was wrong in predicting the Democrats to win North Carolina. Sam Wang and Rothenberg get points from calling it a tie at the end.
Assuming the Alaska result holds with the Republican leading by about 4% (8000 votes), Wang loses a point for calling that one a tie.
Arkansas must’ve been too easy because everyone called it right all seven weeks. Same with New Hampshire.
Rothenberg loses the point he gained in North Carolina by saying Colorado was too tough to call.
Rothenberg and Wang also lose points in Georgia for calling a January runoff and a tie, respectively.
Another Rothenberg loss in Iowa for describing the race as a tie for six weeks then too tough to call at the end.
In Kansas, only Monkey Cage called it correctly. Rothenberg and Sabato called it a tie, which is presumably better than the others who got it wrong.
Of the eight states, LibertyPell got six right and two — Kansas and North Carolina –wrong. We added the correct call on Kentucky and maybe half a one for calling the Louisiana December runoff.
The bookies never wavered in their firm belief that the Republicans would take the Senate, but they had money on it.