Flying Pink Unicorns

Predicting the popular vote, Take III

Here are the cliff notes:

My Final popular vote prediction:

Obama 53.8%
McCain 45.2%
Other 1%

That's Obama by 8.6%.

I'm going to take a different approach in trying to figure out the percentage margin by which Obama will beat McCain (and yes, Obama's winning the popular vote is a done deal barring (a) a truly massive racism effect, as yet completely unseen; or (b) a gigantic game changing event.

I am, for simplicity, ignoring third party candidates. Since they will in any event receive around 1% of the total vote, I'm not making any great mistake by ignoring them.

Let's start with 2004 numbers. According to CNN, in 2004, democrats and republicans were both 37% of the vote, with independents making up the remaining 26%.

http://www.cnn.com/ELECTION/2004/pages/results/president/

Kerry received 89% of the votes cast by registered democrats, Bush received 93% of the registered republican vote, and independents were split almost evenly, Kerry leading 49-48. These numbers don't quite add up to the Bush 3-point popular vote win, because they are polls rather than hard numbers, and error creeps in, but they point the way. Independents were close, but Bush got more of the registered GOP vote than Kerry did the registered democratic vote.

The key fact for purposes of predicting the 2008 election is that people registered as democrats and republicans turned out in almost exactly equal numbers. Put another way, the electorate was made up of essentially identical numbers of democrats and republicans. And, as you can see by the above numbers, the partisan makeup of the electorate is very important. Trying to guess the percentage at which these newly registered voters will vote for the candidate of their party registration is a fools errand, but its a safe guess that a very high percentage of the newly registered democrats will be voting for Obama, while a fairly high-very high percentage of the newly registered Republicans will be voting for McCain. The problem for McCain is that there were a lot more democratic registrations this year than Republican.

One of the two $64,000 questions in predicting the popular vote in the 2008 presidential election (along with racism) is what the partisan breakdown will be.

Here is one article on this issue.

http://online.wsj.com/article/SB122420151553142939.html

Each pollster essentially guesses what the party breakdown will be when all the votes are cast. Rasmussen apparently assumes 39.3% democrats, and 33% Republicans. In contrast, Zogby has it at 38% democrats and 36% Republicans. As a random example, the recent Washington Post poll which showed a 53-44 Obama lead, showed that 38% of the people polled thought of themselves as democrat, 29% Republican and 29% independent. Of the remaining people, the lean towards democrat vs. Republican was 50-50. Now how you think of yourself isn't the same as what party you're registered as, but its a decent proxy. A recent CBS national poll, showing a monster 13 point Obama lead, weighted their poll so as to capture a predicted electorate that would be 40.8% democratic, 27.6% Republican, and 31.6% independent, if I am reading their poll correctly.

http://www.cbsnews.com/htdocs/pdf/Oct08g-Saturday.pdf

Scroll all the way to the bottom.

Now if I'm reading this right, and I really think I am, the CBS poll is just horse crap. Most people are predicting an electorate that is 4-8 points more democrat than Republican. 8 seems like a HUGE lot to me. The weighted portion of the CBS poll is predicated on 13-point partisan lead of D over R, which is out and out insane.


Now perhaps you can see why the polls showing Obama up nearly double digits aren't any good. If the registered party breakdown of actual voters ends up favoring the democrats by 10 points, I'll eat my hat. It will be closer, possibly much closer.

http://www.washingtonpost.com/wp-srv/politics/documents/postpoll_110108.html

To show how significant a difference the party breakdown makes, if the dem, GOP and indy vote percentages stay the same as in 2004, but 39.3% of the electorate is democrat, 33% Republican and 27.7% are independents, as Rasmussen assumes, Obama wins the popular vote with 50.53 % of the vote. If you assume a 10-point democratic party lead, say 42-32, with 26% indys, then Obama receives 52% of the popular vote. If you assume that Obama gets 2 % more democrats than McCain gets Republicans (not insane) and that Obama wins independents by 8 (quite likely imo), then Obama gets 55.7% of the popular vote, and you have a landslide.

As for independent voters, because unlikely Kerry who tied with Independent voters, Obama has consistently polled ahead of McCain among independents. In the most recent polls, Gallup has Obama up by 5 among indys, Zogby (in a tracking poll quite favorable to McCain) has Obama up by 6 points among independents, the recent IBD poll has Obama up by 8 points among independents, a Fox News poll, otherwise favorable to McCain (Obama was up only 3 points) had independents favoring Obama by 5 points.

Most of these polls had Obama getting around 89% of democrats, and McCain getting around 87-88% of republicans (it is close).

So here's a single attempt at a prediction rather than an example. Using the above-mentioned Rasmussen estimates of the 2008 electorate (39.3% of the electorate is democrat, 33% Republican and 27.7% are independents, which sounds plausible), and assuming Obama gets 89% of democrats, McCain gets 87% of Republicans, and Obama gets 54% of independents, Obama would receive 54.19% of the vote. Taking out some votes for the 3rd party candidate, I am now predicting Obama receives 53.8% of the popular vote, McCain 45.2%, for an 8.6% win. My predictions cannot possibly be as precise as I am presenting them, I am merely "showing the work."

In reality, a 53.8-45.2% win for Obama is (a) not that unlikely, though it beats all of the reasonable polls; (b) allows for the greater enthusiasm for Obama to show through, both in the significantly greater number of democrats than republicans voting, and a slightly higher percentage of democrats voting for obama than Republicans for McCain (this is dicey given the large number of democrats in the south that will NOT be voting Obama). Just FYI, if you give McCain 89% of the GOP vote instead of 87, so that dems vote for Obama in exactly the proportion that Repubs vote for McCain, you only subtract .6% off of Obama's vote (and add it to McCain) resulting in a 7.4% win for Obama.

Final popular vote prediction:

Obama 53.8%
McCain 45.2%
Other 1%

I could be wrong by an embarrassingly large amount here, as could the pollsters. The states seem to be pretty easy to predict, the popular vote not so much.