A Simple, Fundamentals-Based Prediction of the Presidential Election

I was at a wedding this past weekend, and the question kept coming up; who do you think is going to win? My answer, “I don’t know.” 

Predicting politics is a hazardous activity, and people are often wrong for some of the right reasons and right for the wrong reasons. But I think it’s important to make predictions and then conduct a postmortem, win or lose. We can learn a lot by proposing and testing hypotheses.

I am deeply uncertain about this election because there are many conflicting pieces of evidence and a huge number of important unknowns (both known and unknown). 

So let’s strip all of this down to the basics; 7-point partisan identification now, the shift in the partisan breakdown from all registered voters to ACTUAL validated voters in 2008, and the Obama/McCain vote share of each partisan category in 2008. 

I’m using the last NBC/WSJ poll from October for the 7-point Party ID because they report the breakdown then for all registered voters, not their stab at “likely voters.” The reported “likely voter” numbers are guesswork, not sampling science, but I think the polls of all registered voters as a starting point are pretty accurate.

I then get my prediction of the actual electorate this year by looking at the shift in partisan makeup in 2008 from all registered voters to all actual, validated voters using data from the CCES project. In other words, what was the actual drop-off in the percentage of pure independents in 2008? How much did the Strong Republican share increase? 

Here’s my prediction of what the national electorate will look like this year:

Party ID is hugely predictive of the vote. So I looked at the two-party vote for each of the 7-point party ID categories:


Finally, I assigned the same vote share to the predicted 2012 electorate. Obama wins, by a hair, if turnout is the same as 2008, if every category of partisan votes in the same proportions as 2008, if all that has changed is the mix of partisan identification.

But I don’t think those are very reasonable assumptions. So, what happens if turnout is down for the Dems and up for the Republicans, or if vote preferences have shifted on top of Party ID?

The table below lays out two, I think fairly conservative, scenarios where Romney wins, just barely. 


The basic facts make this election a tossup, and I think the mood of the country (not to mention the evidence from 2010) make a 3-point swing in partisan turnout AND a 3-point swing in the pure independent vote more likely than not.

My prediction: Romney wins, narrowly.

My caveats:

  1. Although I think it is unlikely, Romney could lose the election by losing in Ohio due to Obama’s dominance in the ad wars (both in terms of number of ads aired and their effectiveness in moving the vote), as well as the far more sophisticated and effective GOTV efforts on the Left.

  2. Although I think it is unlikely, Romney could win while losing Ohio for some of the same reasons; the Left’s ads and GOTV game pushes them over the finish line in Ohio, but in one or more “safe” Blue states where the Left has been mostly absent, the natural, underlying shift in turnout and preferences is able to deliver a winning Republican surge. In other words, Romney could lose Ohio but win with Pennsylvania or Wisconsin because the Left worked it’s magic in the former and not in the latter.
  3. Finally, again unlikely, Romney could fall just short in Ohio and the closing “Blue” states, losing the election but winning the popular vote.