This morning Nate Silver posted an analysis, based on data from the American Action Forum, suggesting that generic ballot polls may be underestimating Democrats’ strength. AAF asked people in 31 closely contested House districts if they would vote for the Republican or the Democrat, then asked the same respondents about the actual candidates they’d encounter on the ballot. In these races, named Democrats are outperforming the “generic Democrat” by an average of 4 points. Silver’s hypothesis is that Democrats’ attempts to localize close races (since the national environment is stacked against them) might be succeeding in these districts.
Henry Olsen of AEI combs through the same data and counters that much of the discrepancy can be accounted for by Republican challengers’ lower name identification. His hypothesis is that as the election approaches and the name-ID gaps close, the discrepancy Silver highlights is likely to fade.
In general, I’m skeptical of generic ballot polls, and when I make election night predictions I always focus on polls that name candidates. But given that, as Olsen notes (citing Silver!), the generic poll has generally overstated Democratic strength, Democrats are probably fooling themselves if they take much comfort in the data Silver cites.