Sam Trenholme's webpage
Support this website

Wednesday election prediction


October 17 2012

In tonight's prediction, I am using Nate Silver's numbers from yesterday (they appear to be pre-debate; Nate hasn't updated his numbers yet) which can be seen at to see how likely each candidate is to win, using two different models.

No correlation between states

Assuming there is no correlation between the states, we get the following numbers. In other words, I am assuming that Obama getting Ohio does not affect his chances of getting Florida, and vice-versa. Indeed, as recently as 2000, Bush fairly easily won Ohio but won Florida by a much narrower margin.
Romney   19.48%
Tie       1.32%
Obama    79.21%

The big three: Florida, Virginia, and Ohio

The three key swing states are Florida, Virginia, and Ohio. That in mind, I have looked at how Romney's victory or loss in these three states affects the election results. To summarize:
Obama gets OH
Romney    9.47%
Tie       1.10%
Obama    89.42%
Romney gets OH
Romney   40.56%
Tie       1.77%
Obama    57.68%

Obama gets FL
Romney    1.40%
Tie       0.22%
Obama    98.38%
Romney gets FL
Romney   28.77%
Tie       1.88%
Obama    69.35%
In the above blog, we look at the election probabilities based on various conditions. For example, if Obama wins Ohio, Romney has only a 9.47% chance of winning the election.

As per yesterday's blog, if Obama gets Florida, there are very few paths to victory for Romney. A friend of mine who lives in Florida told me that Romney supporters are very busy putting up signs down there -- and with good reason.

Despite the low chances, my simulation today found some 263 unique paths to victory for Romney if Obama wins Florida. Most of them (178) give Romney under 280 electoral votes. Here is one 270-electoral vote victory for Romney where Obama gets Florida:

Correlate the states

If one clicks and looks at the above map, it is apparent that the model assumes that there is no correlation between the states. Since Nate's model gives Romney a 9% chance of winning Pennsylvania, it is possible to have a scenario in this model where Romney has Pennsylvania while Obama wins Florida.

Realistically, Pennsylvania is more liberal-leaning than Florida, and, if Obama wins Florida, it's really unlikely he will lose Pennsylvania to Romney. The political climate of the nation as a whole affects how the candidates will do in a given state. Since Florida leans more to the right than Pennsylvania, Obama would probably win Florida after already securing enough other key battleground states to win the election.

This in mind, I ran another model against Nate's numbers: What happens if we have full correlation between the states. In other words, I used a model where Obama gets all of the states with more than a (for example) 50% chance to vote Obama on November 6 (my birthday, guys). I ran this model with all 1000 possibilities from 0% to 100%.

In this model, using Nate's numbers, Obama has a 62.7% chance of winning the election. In other words, if Obama gets all of the states that he currently has a 62.7% chance or higher of getting, he will win the election. The key state right now is, of all places, Iowa.

In order to reduce spam, comments for this entry are now closed