Wednesday, October 1, 2014

SanityIID #Senate #model for October 1, 2014

Welcome to October. Just a little over a month until the U.S. elections.

The SanityIID Senate model run for this morning shows the Republicans solidifying their lead in the short term and making the long term predictions more stable.

As you can see, the most likely probability is that Democrats will win 49 out of 100 U.S. Senate seats, giving control of the Senate to the Republicans. But remember, this model assumes that Orman will caucus with the Democrats. If he decides to caucus with the Republicans due to their taking control of the Senate, then the most probable outcome will be for the Democrats to win 48 out of 100 seats.

Since my last writing, I have added a few more predictions to the SanityIID model. These predictions use the same polling data, but they force the outcome into a win-lose-tie decision. While not as helpful overall as a probabilistic outcome like the above graph, it does help to see where the polling is going over time.

I have broken the data down into long term, medium term, and short term deterministic predictions. Here are the deterministic predictions from the SanityIID model:

Short Term

Democrat Republican Tie
Assumed Wins 44 45
Model Predicts 4 7 1
Total Seats Per Party 48 52

Med. Term

44 45

4 6 1

48 51

Long Term

44 45

4 5 2

48 50

As you can see, using the long term data, the Republicans win 50 seats, the Democrats 48 seats, and two seats are considered a tie. Medium term data gives one of the tie seats to the Republicans for a 51-48 split with one tie remaining. And finally, with the short term data, the final tie is given to the Republicans showing them at 52-48.

While these aren't necessarily the way I expect the elections to come out (I much prefer the probabilistic model as shown by the graph), it does show that the more recent the polling, the more likely they show the Republicans in the lead. Is this a trend? If so, will the trend continue? We only have a little over a month to wait to find out.

No comments:

Post a Comment