A moderately liberal Democraticly-themed blog
I don’t know how many people remember from 2008, but I decided to do an election prediction algorithm that I called EV Status back then. I even had my own domain for it, though I can’t use it now because someone else grabbed it up. But you can still see an archived version of my final predictions on archive.org.
In any case, I was thinking of reviving it for 2012, though it’s a little early to get started, and I haven’t decided if I’m going to do it here, do it as it’s own wordpress.com blog, or get a domain for it again. There are advantages as doing it as it’s own domain such as having all the PHP that calculates everything online and being able to format things the way I want. Of course, the downside is that I have to pay for it.
In any case, my project in 2008 was a success. It predicted every state right except for two: Indiana and Missouri, which were the two states with the closest margins in my projections (0.6% and 0.2% respectively) and, well, a lot of people guessed wrong on Indiana and Missouri. The only thing I didn’t get was the Nebraska electoral vote, mainly because I wasn’t trying to track it.
One thing I don’t think I have done yet is actually go through my final projections and compare them to the actual final margins in each state. In fact, I still haven’t, so as you see them on here, they’ll be as new to you as they are to me…
So jump below the fold to see the results
On my EV Status site, I had a “weighted average” and a “weekly average.” As I recall, I reduced the weight I gave the poll based on how old it was, so as a poll got older, it got weighted less. Then I did a week average where I think I just averaged the polls released in the last week. The weekly average actually only missed Indiana, but it uses a smaller sample, and not even all states had one, but I’ll compare both to see if it made a difference when it came to predicting the margin.
I’m getting the official margin from The Green Papers, rounded to the nearest 1/10th of a percent. If you don’t want to read the table, just jump down to the analysis
| State | Official Margin | Weighted Average | Difference | Weekly Average | Difference |
|---|---|---|---|---|---|
| Alabama | McCain – 21.6% | McCain – 23.8% | McCain +2.2% | McCain – 25% | McCain +3.4% |
| Alaska | McCain – 21.5% | McCain – 12.9% | Obama +8.6% | McCain – 14.2% | Obama +7.3% |
| Arizona | McCain – 8.5% | McCain – 4.6% | Obama +3.9% | McCain – 3.6% | Obama +4.9% |
| Arkansas | McCain – 19.7% | McCain – 10.1% | Obama +8.6% | McCain – 8.5% | Obama +12.2% |
| California | Obama – 24.1% | Obama – 22.3% | McCain +1.8% | Obama – 23% | McCain +1.1% |
| Colorado | Obama – 9% | Obama – 6.3% | McCain +2.7% | Obama – 6.8% | McCain +2.2% |
| Connecticut | Obama – 22.4% | Obama – 20.5% | McCain +1.9% | N/A | N/A |
| Delaware | Obama – 25% | Obama – 24.5% | McCain +0.5% | Obama – 30% | Obama +5% |
| D.C. | Obama – 85.6% | Obama – 69% | McCain +16.6% | N/A | N/A |
| Florida | Obama – 2.8% | Obama – 2.1% | McCain +0.7% | Obama – 2.2% | McCain +0.6% |
| Georgia | McCain – 5.2% | McCain – 4.2% | Obama +1% | McCain – 3.6% | Obama +1.6% |
| Hawaii | Obama – 45% | Obama – 37.8% | McCain +7.2% | N/A | N/A |
| Idaho | McCain – 25.4% | McCain – 24.8% | Obama +0.6% | McCain – 23% | Obama +2.4% |
| Illinois | Obama – 25.1% | Obama – 22.9% | McCain +2.2% | Obama – 22% | McCain +3.1% |
| Indiana | Obama – 1% | McCain – 0.6% | McCain +1.6% | McCain – 1% | McCain +2% |
| Iowa | Obama – 9.5% | Obama – 13% | Obama +3.5% | Obama – 15.4% | Obama +5.9% |
| Kansas | McCain – 15% | McCain – 16.5% | McCain +1.5% | McCain – 21% | McCain +6% |
| Kentucky | McCain – 16.2% | McCain – 13.4% | Obama +2.8% | McCain – 13.5% | Obama +2.7% |
| Louisiana | McCain – 18.6% | McCain – 10.4% | Obama +8.2% | McCain – 8% | Obama +10.6% |
| Maine | Obama – 17.3% | Obama – 15.6% | McCain +1.7% | Obama – 16.5% | McCain +0.8% |
| Maryland | Obama – 25.5% | Obama – 18% | McCain +7.5% | N/A | N/A |
| Massachusetts | Obama – 25.6% | Obama – 19.6% | McCain +6% | Obama – 17% | McCain +8.6% |
| Michigan | Obama – 16.5% | Obama – 12.6% | McCain +3.9% | Obama – 13% | McCain +3.5% |
| Minnesota | Obama – 10.2% | Obama – 11.2% | Obama +1% | Obama – 12% | Obama +1.8% |
| Mississippi | McCain – 13.1% | McCain – 11.1% | Obama +2% | McCain – 10.5% | Obama +2.6% |
| Missouri | McCain – 0.1% | Obama – 0.2% | Obama +0.3% | McCain – 0.6% | McCain +0.5% |
| Montana | McCain – 2.3% | McCain – 2.7% | McCain +0.4% | McCain – 2% | Obama +0.3% |
| Nebraska | McCain – 14.6% | McCain – 20.5% | McCain +5.9% | N/A | N/A |
| Nevada | Obama – 12.5% | Obama – 5.3% | McCain +7.2% | Obama – 6.8% | McCain +5.7% |
| New Hampshire | Obama – 9.6% | Obama – 10.8% | Obama +1.2% | Obama – 11.4% | Obama +1.8% |
| New Jersey | Obama – 15.6% | Obama – 15.5% | McCain +0.1% | Obama – 15.8% | Obama +0.2% |
| New Mexico | Obama – 15.1% | Obama – 9.8% | McCain +5.3% | Obama – 10.5% | McCain +4.6% |
| New York | Obama – 26.6% | Obama – 28.7% | Obama +2.1% | Obama – 29% | Obama +2.4% |
| North Carolina | Obama – 0.3% | Obama – 1.2% | Obama +0.9% | Obama – 1.4% | Obama +1.1% |
| North Dakota | McCain – 8.6% | McCain – 0.7% | Obama +7.9% | McCain – 1% | Obama +7.6% |
| Ohio | Obama – 4.5% | Obama – 4.2% | McCain +0.3% | Obama – 4.8% | Obama +0.3% |
| Oklahoma | McCain – 31.3% | McCain – 28.7% | Obama +2.6% | McCain – 29% | Obama +2.3% |
| Oregon | Obama – 16.4% | Obama – 15.6% | McCain +0.8% | Obama – 15.6% | McCain +0.8% |
| Pennsylvania | Obama – 10.3% | Obama – 9.1% | McCain +1.2% | Obama – 9.8% | McCain +0.5% |
| Rhode Island | Obama – 27.8% | Obama – 18.9% | McCain +8.9% | N/A | N/A |
| South Carolina | McCain – 9% | McCain – 11.9% | McCain +2.9% | McCain – 9.5% | McCain +0.5% |
| South Dakota | McCain – 8.4% | McCain – 9% | McCain +0.6% | McCain – 9% | McCain +0.6% |
| Tennessee | McCain – 15.1% | McCain – 15% | McCain +0.1% | N/A | N/A |
| Texas | McCain – 11.8% | McCain – 11.8% | Exact | McCain – 11% | Obama +0.8% |
| Utah | McCain – 28.2% | McCain – 24.9% | Obama +3.3% | McCain – 24% | Obama +4.2% |
| Vermont | Obama – 37% | Obama – 21.2% | McCain +15.8% | Obama – 21% | McCain +16% |
| Virginia | Obama – 6.3% | Obama – 5.6% | McCain +0.7% | Obama – 5.3% | McCain +1% |
| Washington | Obama – 17.2% | Obama – 15.2% | Obama +2% | Obama – 15.5% | Obama +1.7% |
| West Virginia | McCain – 13.1% | McCain – 8.3% | Obama +4.8% | McCain – 12% | Obama +1.1% |
| Wisconsin | Obama – 13.9% | Obama – 11.6% | McCain +2.3% | Obama – 13.5% | McCain +0.4% |
| Wyoming | McCain – 32.1% | McCain – 24% | Obama +8.1% | McCain – 25% | Obama +7.1% |
As you can see, the numbers are all over the place, but here is the quick roundup, ignoring which way I was off by:
The average miss for the weighted average was 3.61% while the median miss was 2.2%. The average miss for the weekly average was 3.4% while the median miss was 2.25%. Of course, not all states had a weekly average, so the average and median for the weighted averages only in states that also had a weekly average were 3.09% and 2.05% respectively.
So based on the average, the Weekly Average was better – unless you only looked at states that had both averages, in which case the Weighted Average was better. However the median for the Weighted average was better in both cases, but got better when only comparing states that had both.
However, this includes all states, including states that weren’t polled often, but happened to have a poll in the last week. What if we narrow it down to, say, the closest 20 states (or in this case, closest 21 states since there was a tie at 20th). The Weekly Average does better here, with an average miss of 2.07% and a median miss of 1.1%, where the Weighted Average had an average miss of 2.13% and a median miss of 1.2%.
But what if we get down to the nitty gritty – the 10 closest states, which in this case are states which had a final margin better than 9%. This is once again mixed. The Weighted Average had a better average miss of 1.74% vs. the Weekly Average, which had an average miss of 1.94%. However, the Weekly Average had a better median miss of 0.45% than the Weighted Average, which had a median miss of 0.5%.
In any case, at the end both averages performed pretty similarly and, I would say, reliably. The worst miss by either average in the any of the 8 closest finishes, was a miss of 2% in Indiana’s Weekly Average. The Weighted Average projected the final margin within 1% in 7 of the 8 closest states, while the Weekly Average projected the final margin within 1% in 6 of the 8 closest states. And the 10th closest state – North Dakota – should probably have an asterisk next to it since almost all of the polls done there were from Research 2000, who may very well had been making shit up at the time.
