The conventional wisdom is that Ohio is the best reference point for understanding Pennsylvania. In this diary I use this assumption to analyze the Pennsylvania primary in terms of what Ohio told us about voting tendencies. I go on to project the race on the basis of these voting tendencies. This is not a prediction, since alot may change in the campaign in five weeks - there is much that Obama probably wishes he could change about how the campaign took place in Ohio (e.g. the dominance of the Canada story). But it is still worthwhile projecting Pennsylvania on the basis of Ohio, as the results shed some light on what the battle ahead will look like.
What we see in Pennsylvania is that the demographics and the delegate math actually favor Clinton more strongly than they did in Ohio.
Resulting Pennsylvania projection
Clinton: 60.3% popular vote, 91 delegates
Obama: 39.7% popular vote, 67 delegates
keeping reading for more details...
I figured I could sum up the results of Ohio with a few simple observations that we have all come to know, and that can be backed up by looking at any exit polls:
- Obama did much better among younger voters than older voters
- Obama did best among black voters
- Obama did better in urban areas than rural areas
- Obama did better with more educated voters
- Obama did better with more affluent voters
Given this, I figured I could come up with a useful metric for predicting his support based on these factors. All data was attained from the 2000 census (which is a little dated, but looking at more recent estimates I don't think it would make a large difference). I decided to measure these five variables with the following statistics.
- Median age of the population
- Black population (%)
- Rental housing units (%)
- Bachelor degree holders (%)
- Median household income
There is no clear distinction between urban/suburban/rural in the census, but urban areas tend to have a higher percentage of housing units rented, so I decided to use this as a rough measure.
I analyzed this data by congressional district (CD) in Ohio and I came up with a regression formula for predicting the Ohio results.
As can be seen, the model is reasonably accurate considering its simplicity. Perhaps I could introduce some other variables, but in the end there will always be effects that are impossible to model demographically (e.g. local issues, GOTV, local media, campaign visits).
Now the fun part is applying the model to Pennsylvania and looking at the projected delegate totals.
Note: There are 55 statewide delegates that are included in the total, 33-22 for Clinton. I've also added a map to put this into perspective:
Again, this is not a prediction of what is going to happen. It is merely a look at Pennsylvania on the basis of Ohio. If anything, think of it as a guide. It suggests some interesting results:
- Pennsylvania is even more representive of Clinton's coalition than Ohio
- There will be key battles in the areas around Philly to see who gets the 4-3 delegate split
- While Obama may be unlikely to win some of the rural CDs, he needs to focus there to avoid 4-1 losses (CDs 3, 4, 11, 12)
- While Obama won most of the big cities in Ohio and Texas, Pittsburgh's CD 14 is not a given
And perhaps most importantly:
- Obama has alot of work to do if he wants to win Pennsylvania
Comments and suggestions appreciated (especially from those who know the state better than I do, which is not very much).
Updates:
The Ohio demographic model predicts Maryland as a 63-37 win for Obama. The actual result was 62-38 (after normalizing it to total 100 by not to including other candidates). Unfortunately, I could not find results for Maryland by congressional district. I have county-by-county results, and the fact is that the model is all over the place (alternately overestimating and underestimating in double digits). That's partially because counties are alot smaller and there should be more variance, but no doubt the model also needs some tweaking. I'll try a few things and keep this updated.
Also, there have been a couple comments about open vs. closed primaries. This is a good point, and all else being equal a couple percentage points should probably be added to Clinton across the board in Pennsylvania.