I've been scratching my head, wondering what on earth was going on with Gallup's presidential tracking poll. Except for the post-convention bounce (since faded, by their lights) Obama's lead has been consistent but tiny. Gallup's a poll of registered voters, rather than likely voters. Given Nate Silver's research that RVs tilt democratic by ~2% compared with LVs the Gallup numbers have always made me a wee bit nervous.
I haven't done a detailed analysis, but comparing Gallup's sample to the US Census data on RV's at the time of November 2010 helps clear things up: Gallup has way too many old folks in their sample.
Age Gallup % US Census %
18-29 9.1 16.2
30-49 22.5 34.2
50-65 32.3 29.0
65+ 35.0 20.6
(NB: Gallups percentages do not add to 100. I can't tell you what's going on that their N only divides to 99 percent, but those are their numbers.... weird).
Given that Gallup has all voters younger than the 65+ age group favoring Obama I think we can square the results with other polling that we're seeing. I'll do that tomorrow.