Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
The States Ordered Left to Right: How to Predict the Election Early Tuesday Night
Certain “red” states, few will dispute, almost surely will vote for Romney Tuesday night. Here are 18 such states, ordered according to the two-party vote share they gave to McCain in 2008: Wyoming (66.6), Oklahoma (65.6), Utah (64.5), Idaho (63.0), Alaska (61.1), Alabama (60.9), Arkansas (60.2), Louisiana (59.5), Kentucky (58.2), Tennessee (57.6), Kansas (57.6), Nebraska (57.6), West Virginia (56.7), Mississippi (56.6), Texas (55.9), South Carolina (54.5), North Dakota (54.4), and South Dakota (54.3).
Similarly, here are 13 “blue” states—again, ordered by their 2008 two-party McCain vote share—that will almost surely vote for Obama: New Jersey (42.1), Maine (41.2), Washington (41.2), Connecticut (38.7), California (37.7), Delaware (37.4), Illinois (37.3), Maryland (37.1), Massachusetts (36.8), New York (36.4), Rhode Island (35.8), Vermont (31.1), and Hawaii (27.0).
Finally, here are 19 “purple” states for which there is at least some uncertainty about how each will vote. Next to each state, I list the probability—which I have gathered from a betting market—that it will vote for Romney. (The data come from Intrade.com. Essentially, I use the average of the bid and ask prices for the markets for each state.)
State Pr{State Elect. Cum. Elect.
Votes Votes Votes
Romney}
Indiana .970 11 151
Georgia .968 16 167
Montana .950 3 170
Missouri .923 10 180
Arizona .925 11 191
N. Carolina .765 15 206
Florida .700 29 235
Colorado .463 9 244
Virginia .436 13 257
New Hamp. .363 4 261
Ohio .324 18 279
Iowa .309 6 285
Wisconsin .269 10 295
Nevada .182 6 301
Penn. .180 20 321
Mich. .175 16 337
Minn. .152 10 347
Oregon .129 7 354
New Mex. .077 5 359
In the above table I also list each state’s electoral votes as well as its cumulative electoral votes. To compute the latter, I assume that Romney will win all the electoral votes of the 18 red states I list above (140), plus the particular state in the list, plus all the states that appear above it in the list.
My experience is that the betting markets are incredibly accurate at predicting the ordering of states in terms of how they vote. That is, I believe the above ranking will be very close to the actual ranking that occurs election night. Although we may see a couple states switch positions with a neighboring state on the list, we will not see large deviations. Thus, for instance, although New Hampshire might switch with Virginia or Ohio, it almost surely will not switch with much more conservative states such as North Carolina, Missouri, or Arizona; nor will it switch with much more liberal states such as Nevada, Pennsylvania, or Minnesota.
Assuming that the above ordering will be similar to the actual ordering of states on election night, we can make an important prediction: Ohio, as many are predicting, will indeed be the key swing state. That is, Romney (or Obama) will win the election if and only if he wins Ohio. To see this, note that Romney needs to win 269 electoral votes to win the Electoral College. (Technically, he needs to win 270 to get a majority. But I’m assuming that if he and Obama both win 269 votes, thus tying, then Romney will win the tie-breaker in the House of Representatives.) And to win 269 electoral votes, Romney needs to win Ohio and all the states that come above it in the list. Further, note that—because Ohio has so many electoral votes—this will be true even if Ohio switches places with New Hampshire or Iowa in the list.
I have transformed the above probabilities, using some standard statistical techniques, to predict Romney’s (two-party) vote share in each of the above states. Here are my predictions:
State Romney Vote Share
Indiana 56.8
Georgia 56.7
Montana 55.9
Missouri 55.3
Arizona 55.2
N. Carolina 52.6
Florida 51.9
Colorado 49.7
Virginia 49.4
New Hamp. 48.7
Ohio 48.4
Iowa 48.2
Wisconsin 47.8
Nevada 46.7
Penn. 46.7
Mich. 46.6
Minn. 46.3
Oregon 45.9
New Mex. 44.9
As a final exercise, I perform the following thought experiment: “Suppose Ohio votes exactly 50-50. If so, what will be the vote shares of the other states?”
To perform this, note that my prediction for Romney’s vote share in Ohio is 48.4%, which differs from 50% by 1.6%. To perform my thought experiment, I simply add 1.6% to all of the above vote shares. I have done a similar exercise for the red and blue states that I mentioned at the beginning of this post, except instead of using probabilities from betting markets, I use vote shares from the states in the 2008 election. Given those calculations, here are my predictions for Romney’s vote share for all of the states in the Eastern Time Zone—conditional upon the event that Ohio votes exactly 50-50:
State Romney Vote Share if Ohio Votes 50-50
Connecticut 41.7
Delaware 39.4
Florida 53.5
Georgia 58.3
Indiana 58.4
Maine 44.2
Maryland 40.1
Massachusetts 39.9
Michigan 48.3
New Hampshire 50.4
New Jersey 45.2
New York 39.5
North Carolina 54.3
Ohio 50.0
Pennsylvania 48.3
Rhode Island 38.9
South Carolina 57.6
Vermont 34.2
Virginia 51.1
West Virginia 59.7
Accordingly, if, on election night, Romney is outperforming most of the above predictions, then this is a sign that Romney’s vote share in Ohio will be higher than 50%. If so, I’ll go to bed early, confident that he’ll indeed win the election.
[I have corrected an earlier version of this post. In the third table, in the column “Romney Vote Share if Ohio Votes 50-50,” I had wrongly recorded 56.7 and 56.8 as the numbers for Georgia and Indiana. I had forgotten to add 1.6% to the earlier numbers I computed for those states. The new numbers, 58.3 and 58.4, are the correct numbers.]
Published in General
Looks like the mathematician/political scientist’s version of a crossword puzzle. –Joy
My head hurts!
I can’t look at numbers anymore! (I hope I’m not tutoring math today.)
Romney, 341…..or more.
I see Ohio going Romney 52/49. Based upon the potential for monkey business in Cuyahoga County being drastically diminshed by the voter roll clean-up (300k+ names of ineligible and/or dead voters removed).
Without the Cleveland ghost votes, there are not enough genuine Democrats to drag this miserable failure over the finish line.
Did you account for systemic fraud? It can’t be close in key states. If recounts are ordered, Obama wins – his legal stormtroopers will “find” enough votes to put him ahead.
Your methodology using the ranking, if what you say about the historical record of this data, is probably more reliable, but I couldn’t help computing a simple expectation from your numbers, as this appears to be the most obvious statistical prediction. We give 140 electoral votes to Romney from the Red States and add the expectation for each state, and arrive at 150.02 electoral votes — advantage Obama; not quite a landslide, but a sound victory.
This is assuming that the betting market produces realiable probabilities, and that those probabilities are statistically independent. Big assumptions, so I’m not putting any stock in this prediction. Still, I think one must take it as the null hypothesis (for anyone starting with this data).
Looking at the Senate numbers at Rassmussen, it appears that it is a wash. There are 10 Republican seats up, and the Republicans stand to win 10, not gaining the 3 they need to control that house. In the Congress there will be a minor deterioration in Republican control. If Romney loses the situation will be worse than before the election in terms of distribution of power in Washington. The Tea Party is needed more than ever.
For good measure I did a similar calculation using the Rassmussen “share of vote” numbers for each state. This is methodologically even worse than my expectation calculation above, but has the advantage of working from more verifiably reliable data. I give Romney all the votes leaning his way and weight the toss-up states’ share for Romney according to the percentage of their votes in his camp multiplied by the EC votes up for grabs. I arrive at 253.04. Again, not a Romney win, but slightly more optimistic than the one above.
I note that Gallup froze their reporting earlier than in most election years, happening to hit a date on which Obama’s approval number had spiked to 51%. I wonder what it has done in the last week? Let’s hope Romney’s election day ground game is as good as people are saying it is. Obama has a pretty large standing army.
Your headline sort of rhymes.
Awesome analysis. You initially threw me off with your use of Intrade values, but then I saw that it was to rank the states, which effectively cancels out the loopiness of the values.
It seems that Intrade is largely using the forecasting skills of MSNBC’s core audience, which is why they’re predicting a blowout Obama victory. I was unlucky enough to buy my Romney shares at $4.07… at least I’m going to get $5.93 profit per share. But I could have bought them at $3 and change! :-)