Social Science Statistics Blog

I was with you until you got to the actual curve-fit. The graph says it all: When you look at Kerry's share = 0.4, for example, there is a wide spread of actuals. r-squared notwithstanding, it looks as though the actuals at Kerry = 0.4 vary from the predicted by at least +/- 15%! I have a hard time accepting that as "an okay fit".

The other thing is obviously population. Counties don't vote in American elections, voters and delegates do. The one county you've predicted as Obama's is Philadelphia, center of the largest concentration of people in the state. After Pittsburgh and Philadelphia, the remaining counties (which fill up the majority of your colorful map) represent a comparatively small fraction of the Pennsylvania population.

Still, I do not mean to discourage. This is an interesting analysis, and I hope this criticism is taken as constructive.

The general vs. primary thing is also an issue. Centre county (that big, Woody-Woodpecker-looking thing right in the middle of PA) consists of State College and rural PA, split about 50-50 population-wise. The latter is entirely GOP territory, the former entirely Democratic -- hence the 50-50 split for Kerry in the 2004 general. The idea that that will translate into trouble for Obama *in the primary* when nearly all Democratic voters in the county are relatively well-off, highly-educated, PSU-affiliated State Collegians (a place where he drew 20K to a rally last weekend) seems like a stretch.

The larger point, of course, is that generals aren't primaries, something I wish the Evan Bayh's of the world would get through their skulls every time I hear them say how important it is that Clinton has won the "big states"...

Great post. I agree that the graph does say it all. It speaks volumes with the numbers in respect of the points made.

Hey guys! Thanks for all the comments. I guess I got a little too excited about a rather weak relationship. In reality, of course, there are other, better correlates with the Obama-Hillary vote than the Kerry-Bush 2004 vote. Perhaps rather than predicting Pennsylvania, I would've done better with a more general post about 2004 results vs. 2008 results. One amusing fact is that the only county won by Ron Paul in Iowa happened to be one of three counties won by Howard Dean in 2004. Not surprising!

Ha ha, Thats right Kevin. However it was a good post, and interesting to read.I have just visited your website politistats, which also showed some examples that are interesting to me too. Thank you.

Hi :)

Very interest post...

Regards,

This is sad for the Obamaites if the race where to be decided in this state.

Ray S.

Great thinking. It is always useful to have a sound theory when using statistic to predict the future. And I think you have some good arguments eventhough they are difficult to understand.
Thank you for some interesting and challeging thoughts!

The problem with the forecast is that the Ohio democratic primary was an open primary: républicans and independants were allowed to vote.

In Pennsylvania, the démocratic primary will be a closed one. Only democratic registered voters will be allow to go to the polls.

So, I guess that the outcome will be far better for Clinton in Pennsylvania than in Ohio.