Jen Rubin wonders about the sampling of the Los Angeles Times/Bloomberg poll.
I won’t quibble much with their results within the parties (Hillary and Giuliani both up big), although when your Democratic sample is 469 respondents and your Republican sample is 364 respondents, you’re going to have a larger margin of error in the results among Republicans.
#more#
But in the head-to-head matchups, in which the sample is a random half of all registered voters polled, I’m detecting an odor:
The Los Angeles Times / Bloomberg Poll contacted 1,209 adults nationwide by telephone October 19 through 22, 2007. Included are 1,039 registered voters, 469 Democratic Primary voters and 364 Republican Primary voters. “Primary Voters” include both registered party members and eligible others who plan to vote in the party primaries. The 2008 presidential election match-up questions were split among two random sub-groups of 512 and 522 registered voters respectively.
I’m a little wary of their head to head numbers, because in the pool of registered voters, 35 percent of the sample identified as Republicans and 45.1 percent of the sample identified as Democrats.
In 2004, I remember a Los Angeles Times poll that had Kerry up big, but had a ludicrously skewed sample. (Take a trip down memory lane into the Kerry Spot archives. Good times, man, good times. Anyway, that sample was 38 percent Democrats, 25 percent Republicans.)
Having said that, the Republican “brand” was in much better shape in early 2004 than it is today. So I’m not going to gripe about a poll sample that has more Democrats than Republicans; the question is, how many more Democrats can you include in your sample until it no longer resembles who is going to show up on Election Day? When you have ten percent more Democrats, I get a little skeptical.