2012 election predictions are coming

Were you reading this blog during the 2008 election season? If so, you may remember my election prediction posts that took you over to Hominid Views. This year, I’ll post the election predictions here.

I’ve spend the past week collecting polls, updating the software, creating a clickable cartogram for the 2012 electoral college, updating the FAQ, and figuring out how to make it all work on Horsesass. The first analysis for the presidential election will be posted later today.

At this point, I am only doing analyses of an Obama versus Romney general election. As much as I would like to see one of the weaker candidates take the G.O.P. nomination, I’m pretty certain Republicans will, as they did in 2008, act rationally, and chose the candidate that performs best against Obama in head-to-head polling. That is currently Mitt Romney. As the Republican primary circus continues, I’ll reassess. If, say, Santorum trickles on up to the front (eww!) or there is a crazy surge for Ron Paul, or the Mittster takes a tumble after unintentionally tweeting a photo of his underwear, or Rick Perry challenges the rest of ’em to a duel (and wins), I’ll switch do doing analyses for the new front-runner(s).

Later in the election season, I add senatorial and gubernatorial analyses as well.

When I post these analyses, there are occasionally naysayers. They complain that polls are meaningless, the analysis is flawed, or the results are not predictive, or “can’t we just wait for the ‘real poll'”, blah, blah, blah. I’ll repeat my counterargument.

It works the same way as the score at a sporting event. The first quarter score in, say, a basketball game doesn’t typically allow you to determine the eventual winner. The score, the spread, the amount of playing time remaining, and the recent changes in scoring momentum gives a good feel for how the game has progressed, who might win if things continue in the same vein, and what each team needs to do to attain victory. Somehow I think fans would not appreciate basketball scores being hidden until the the game has concluded.

Same with the election analyses. They aren’t predictions of the outcome on election day. Instead, they show the score so far. And the currency is a probability of winning, if the “game” ended now.

Another point from the naysayers in 2008 is that Nate Silver, now at the NY Times, does similar analyses, and everyone knows he is the best. Mr. Silver is an entertaining writer, and does a very nice job with graphics and site design. And since he does this stuff full time, he is quite prolific. Here is the problem I had with his 2008 analyses. He used a complicated (nearly proprietary) analysis that involved using information beyond simple polling data. This ancillary information was included as a pseudo-poll in his analyses. My preference is for a straightforward, data driven analysis that makes the fewest assumptions necessary.

But the proof of the pudding is in the eating…so how did we each do in 2008? Here is his last pre-election post and here is mine. We both missed a single state—Indiana. The late polling in Indiana gave McCain a sliver of a lead, and the “big poll” came down in favor of Obama by a 1% margin.

For the electoral college, Mr. Silver projected a 349 to 189 victory for Obama and I projected a 364 to 174 victory for Obama. The actual result was 365 to 173. I was off by one vote.

This one vote discrepancy, in fact, reflected a weakness of my analysis. I ignored the possibility that either Nebraska or Maine might split their electoral votes. Nebraska’s 1st district did split in 2008, giving one of the state’s electors to Obama. If I had included this little detail, my projection would have been spot on. For the 2012 elections, I have already added separate analyses of Nebraska districts, and will do so for Maine when some district-level polling data becomes available.

For more information on the methods used, please visit the draft of the new simulation FAQ. Also if you have recommended changes or have additional questions for the FAQ, please mention them in the comment thread.

Update: The Obama-Romney analyses can be found here. There is now a side-bar blurb, too.