Over at his blog DonkeyRising, Ruy Teixeira has been running an on-going critical commentary of the Gallup poll.
Some of his criticisms I have found stronger than others. I think, for instance, that he's definitely on to something with the sharp Republican skew in the party-identification of the Gallup polls. But I wonder whether this problem skews the horse race numbers as much as Ruy says.
It's not that I disagree with his reasoning. And Ruy knows much more about public opinion research than I do. I think it's just an instinctive skepticism I have about finding arguments for disregarding polls that don't say what you want them to. Put more simply, I try to be on guard against spinning myself.
On Tuesday though Ruy came back with a further analysis of the Gallup poll which seemed to make an indisputable case that the Gallup likely voter screen clearly underrepresents minority and young voters.
One might say that minority or young voters vote less consistently than affluent whites. But Ruy shows pretty clearly that Gallup's numbers presume rates of participation that defy history and common sense.
For instance, minority representation among voters in 1996 was 17% and in 2000 it was 19.4%. Yet Gallup says it'll be 14.5% this year. That's hard to figure since, as Ruy notes, minorities are growing as a percentage of the population.
With blacks, it was 10.1% in 1996 and 9.7% in 2000. But Gallup says that it'll fall this year to 7.5%.
On young voters (18-29 year olds), it's a similar story. Young voters made up 17% of the electorate in 1996 and 2000. This year, says Gallup, they'll account for only 11%.
Finally, I would be remiss if I did not say that the demographic breakdown numbers Ruy was going on here came from Steve Soto, who has a further discussion of these demographic problems in the Gallup numbers on his website.