Another Look at Pennsylvania

Three weeks ago, I wrote a long post about how registration data in Pennsylvania could be used to predict the outcome of the election there. I found that, even after making some generous assumptions for Senator McCain, he would still lose because there are simply way more Democrats than Republicans this year in Pennsylvania.

Well, three weeks have gone by and Pennsylvania isn’t looking quite as close as it did back in the middle of September (even my parents have mostly stopped worrying). Pollster.com now has Senator Obama ahead of Senator McCain by more than 12 points in PA. RealClearPolitics has the deficit at more than 13 points, and FiveThirtyEight.com has Obama ahead by 11. All three sites now list Pennsylvania as “Safe,” “Solid,” or “Strong,” for Obama. Quite a dramatic change from three weeks ago. But it might even be worse for McCain than that.

In the time since I wrote that initial post, registration has closed in Pennsylvania and so we have some “final” numbers to look at. If you’ll recall, my methodology is to use the relationship between county registration numbers and vote tallies from 2004 to predict the 2008 tallies. Three weeks ago, using this method, I came up with the following unadjusted prediction:

Obama – 54.6%
McCain – 45.4%

Now, between then and now, Democrats have added another 117,000 registrants while the Republicans have added only 41,000 new registrants. That’s right. In just a few weeks, the Democrats added to their registration advantage by more than half of the total margin by which John Kerry defeated George Bush. What do all these new registrations do to my prediction? Have a look:

Obama – 59.7%
McCain – 40.3%

A twenty point blow-out! Is that possible? Yes. Is it likely? Probably not, but it certainly suggests those polls showing Obama up by 15 are not outliers.

I also want to take this opportunity to address a question that readers have asked about the methodology. The questions is, “What about the independents?” As you probably noticed, I talked a lot about Democratic and Republic registration numbers, and how these numbers can be used to predict the outcome. Where are the independents? Well, they are actually already implicitly included in the model. The idea here is that the number of registered Democrats in a given county is an excellent predictor of the total number of votes that the Democratic candidate will receive in that county. Those votes can come from Democrats, Republicans and Independents, it doesn’t matter. The point is that the registration numbers are, indeed, great predictors of vote totals (at least in Pennsylvania). If there are many more registered Democrats than Republicans in a given county, it doesn’t matter how many registered Independents there are, the Democratic candidate will win that county. Why is this the case? Well, its probably the case because Independents aren’t usually all that independent. In fact, it turns out that most Independents have a fairly consisted lean toward one party or another. Moreover, and I’m speculating a bit here, it seems likely that in a county that is dominated by, say, Republicans, that the Independents are likely to be Republican leaners. Similarly, in a more closely divided county, it seems likely that the Independents are likely to more evenly divided. This may account for the fact that we can predict real vote totals based on party registrations alone.

The other question I’ve gotten is, “What about race?” Some have asked if race won’t make some of these Democrats act more like Republicans. That’s a fair question and even though I think that the events of the past few weeks will mitigate some of this behavior, I also think that there will be some otherwise Democratic-leaning voters who pull the lever for McCain because of race. But it would take a huge number of these folks to change the outcome. Indeed, for the regression model to predict a McCain victory, about 800,000 registered Democrats need to switch over and be treated like registered Republicans. For reference, 800,000 is more than 60% of the total number of voters who cast their ballot for Hillary Clinton in the primary. If McCain’s gonna pick off any Democrats its going to be among former Hillary voters. Does anyone believe that he can win more than 10% or 15% or even 20% of them?

The bottom line today is the same as it was three weeks ago. McCain will lose Pennsylvania because there are simply many more Democrats than Republicans in the state.