Grover's Take on Things

Anyone remember Nate Silver?

Of course you do — or at least you should. He is the 36-year-old statistician who did what no other pollster or analyst seemed able to do: he accurately predicted the 2008 Presidential voting results in 49 of 50 states (Indiana tripped him up by 1 percentage point); and in 2012, when all other pollsters missed (Republican Party polling disastrously so), Silver correctly predicted the results in all 50 states.

So, with exactly 25 days left until election day, what is the stats wunderkind saying about the 2014 mid-terms?

Silver is not a pollster, nor does he commission polls from others. He obtains the results from existing professional, journalistic, and university polling organizations, he combines and analyzes their results, and he uses those results to make his own probability-based predictions. His genius is in (1) the relative weightings he gives each source, and (2) the (closely guarded) secret of how he has determined those weights and how they can translate into highly calibrated probabilities.

Silver’s latest graph (from last night; shown below) plots the evolution of his predictions of the likelihood of control of the Senate (Dem blue vs. Repub red), beginning back on August 10. As of right now, Silver’s projection is that there’s a 58% likelihood that Republicans will win control of the Senate, with only a 42% chance that the Democrats will keep it.

 
In the North Carolina race, he is currently predicting a 48-45 percentage point Kay Hagan (D) victory over Thom Tillis (R). He calculates a 79% likelihood of her winning, and a 21% chance of a Tillis win. (Please note that when I use the shorthand “Silver predicts victory…,” I am saying that in that race, his percentage probability of victory for that candidate currently sits at greater than 50%.)

Of the 36 Senate races being decided in 25 days (including 3 special elections in Hawaii, Oklahoma, and South Carolina), Silver is currently predicting Republican wins in 22 states and Democratic wins in 14. These include 8 Democrat-to-Republican pick-ups in Alaska, Arkansas, Colorado, Iowa, Louisiana, Montana, South Dakota, and West Virginia. The only Republican seat he is predicting they will lose is in Kansas.

The Maroon 6: Silver has identified six states that he calls the “Maroon 6,” with Senate races to watch: Alaska, Arkansas, Georgia, Iowa, Kentucky, and Louisiana. His calculations tell him that if Republicans win all 6, they have a 99% chance to take the Senate. If they win 5, their chance slips slightly to 84%. With anything less than 5 wins, however, Republicans’ chances to control the Senate drop precipitously, to levels that strongly predict Democrats’ keeping the Senate. Right now, on his state-by-state predictions, Silver is predicting Republican victories in all 6 of the Maroon states:

Alaska	76% chance of R win	predicting win by 3 pct. points
Arkansas	74% chance of R win	predicting win by 3 pct. points
Georgia	72% chance of R win	predicting win by 2 pct. points
Iowa	65% chance of R win	predicting win by 2 pct. points
Kentucky	75% chance of R win	predicting win by 3 pct. points
Louisiana	75% chance of R win	predicting win by 5 pct. points

 
Because he has a history of being uncannily correct, and because he was in the right place at the right time to call major Democratic Presidential wins, he has emerged as that party’s darling — almost a talisman or harbinger of Blue victories. But I understand he is taking flak this year from some Left/Liberals for his predictions about the Senate races. Some attacks have been so shrill it sounds as if they’re saying, “How dare you predict bad things for us?!”

As we all know, there is always the possibility of one or more major October Surprises, and polling data fluctuate all the way up until the day of the selection. So on the next 3 Fridays before the election, I will give an update here on the evolution of seer Nate Silver’s projections, and then I will do a “how did he do?” review after the election.