The Million Dollar Question

Over at Redstate, Adam C raises what is probably the most important question of the election season for pollsters:

Let's take a look at PA polling and party ID. Today, Franklin and Marshall released a new poll showing Obama +5. This is in line with other recent polls showing a comfortable Obama cushion (RCP Ave.: Obama +6.8%) that is similar to his national lead. But the F&M poll is a sample of 50D/38R or D+12. I thought that looked a bit more Democratic than I would expect so I looked into.

In 2004, PA exit polls showed the electorate was D+2 (41D/39R). Surely it is more Democratic now for the well-known reasons: R registration drop nationally, D registration increase nationally, close D primary led to more registration/participation, and national party ID numbers have gone from D+4 to D+10 regardless of actual registration. But a shift of 10 points in four years would be quite devastating for the GOP in PA.

The last F&M poll showed a D+14 partisan makeup and showed Obama winning by 6. In both F&M polls Obama only wins 68% of Democrats but he leads the race because the polls show an overwhelming partisan advantage for the Ds.

So do other polls in PA show a similar margin? Well, I cannot find party ID breakdowns for the Rasmussen, Quinnipiac, and Strategic Vision (R) polls. So it's hard to tell. I find it unlikely that Ds have gone from 41% of the electorate to 50%. But if that is true, the GOP is going to face a steep hill in winning back Congressional seats in addition to the Senate and Governor races in coming years. And a 50% Democratic electorate may be insurmountable for the McCain campaign.

As I point out in the comments, the official Pennsylvania registration numbers are 51% D, 38% R.  In 2006, the turnout numbers were 43% D, 38% R, 19% I.

The big question then is what this change represents.  You will notice that R numbers are more-or-less consistent from 2004-2008.  What has happened is that Independents have swung D, and did so in large numbers from 2006-2008.

Are the Independents who swung to D from 2006-2008 -- who make up about 15% of PA Demcorats today -- true Democrats who are creating a party registration edge that is almost insurmountable for the GOP to overcome?  Or are they merely independents who wanted to play a role in the Obama/Hillary fight?  And of that latter group, how many are independents who became registered Democrats to stop Obama?

The answer to that question is important.  If the latter is the case, most of the polling we are seeing out of PA may be badly skewed in the Dems' favor when it is weighted by the polling companies.  This may be true in other states as well.

UPDATE:  I really should have thought of this myself.  In the comments over at Redstate, someone dug up the historic registration numbers in PA.  In 2004 it was 47-40 D/R.  In 2006 it was 48/40 D/R.  Now it is 51/38 D/R.

In other words, Dems in PA typically run 5-6 points behind their party registration.  F&M shouldn't be normalizing to 50% Dem makeup; they should be normalizing to ~45% Dem makeup of the electorate.

Now this year may be different.  Dems are probably fired up, especially AA Dems in Philly.  But I'm not certain that Rs won't be similarly fired up by election day.  I'm also much less than certain that all those D's are REALLY D's; a substantial number may be "4/22/08 only" D's who haven't bothered to re-register as I's (or R's).

Incidentally, if you take the F/M numbers and normalize to a 45/38 turnout split, you get a narrower 48.5 %- 45% Obama lead.