The public polls over the last several days show lots of variation in today's two primary states. As Mark Blumenthal at Pollster.com points out, the North Carolina numbers find Obama leading anywhere between 4%-14%. The Indiana polls demonstrate even more variation. They find Obama leading by 2% (technically it's a statistical tie) to losing by 12%. Lots of room for scenario building around those results.
The most recent Indiana numbers reported by Zogby (a statistical tie) are the most provocative, suggesting a late surge by Obama in the Hoosier state. On one hand these results look like outliers compared to some of the other earlier polls showing a larg Clinton lead in that state. On the other, this poll is one of the latest out of the field and could be measuring a late-Obama surge. Keep in mind Zogby nailed the Pennsylvania numbers very accurately.
Late deciders are also critical in predicting the final outcome, particularly in Indiana, where the race appears to have tightened. But Hillary Clinton has done better than Obama with those deciding in the last few days in recent primaries. Blumenthal has a good piece on the subject of how pollsters deal with undecided voters here.
Bottom line: Pre-election polls include "undecided" voters--elections do not. One of the tricks of the trade in polling is how hard you "push" those who don't initially reveal a preference to pick a candidate. Survey organizations use different methods and those tactics can shape their final numbers. I'll be on the lookout for the exit polls later tonight and post something when they become available.
