I was thinking about how many people (especially on the right) are saying that the polls are skewed by unrealistic party ID weighting, i.e. if the sample contains 10% more Democrats than Republicans, of course it'll show Obama ahead of Romney.
I did a little research which confirmed my suspicion that self reported party identification is the most reliable predictor of voting behavior. It's better than race, age, income, anything. So it occurred to me to compare past party ID to election results and extrapolate 2012 results from current (mid-Sept. '12) party ID data. I'm just cutting and pasting from Excel so it may be ugly, but here's what I found:
- In 2004, Republicans outperformed their party ID by 4.7%
- In 2008, Republicans outperformed their party ID by 0.4%
- Today, Republicans have a 4.3% advantage in party ID
- If this party ID vs. vote pattern holds, there will be a Republican landslide in 2012 of between R+4.7% and R+9.0%