There's been a lot of poll analysis in the last couple of weeks. Today, Pollster.com posted a discussion of where their method of averaging available polls failed. The 29th was one of those races. No particular conclusions were drawn about the 29th, but it is clear that this district was very lightly polled. When there's hardly any data to start with, any average is going to be more suspect.
Turning to the actual polls, Rochesterturning has been posting graphical interpretations of election results. Today's post shows that a significantly smaller percentage of voters voted for Kuhl in 2006 than in 2004 in almost every county. This is especially interesting because the Conservative party fielded a candidate in 2004, as did the Independence party. Both of those candidates should have taken some votes from Kuhl in '04 that he regained in '06. In this analysis, and the other county-by-county post, keep in mind that the graphs are unweighted. So, for example, when you see that Kuhl lost roughly 15% of his vote in Yates county, remember that Yates had just under 7,000 votes cast in 2006. The 29th had about 200,000 voters in '06.
Rochesterturning also looked at correlations between census data and voting patterns. They found a positive relationship between income, education and a tendency to vote for Eric Massa. There are two ways to interpret this data. A partisan spin for Democrats is that smart people voted for Massa. Republicans could argue that the Democrats are turning into the party of the elites, rather than the party of the "average working man". I don't buy that. I agree with RT that those numbers show that more educated people are more likely to inform themselves about the candidates in an election, and therefore are more likely to vote for a less-well-known challenger like Massa. Also, the income split mirrors the North/South split in the 29th, with Monroe being the most wealthy county. By registration, that county is also the most Democratic of the bunch.