Why Polls Matter
by Steve
Yesterday, I ran a post noting my doubts about a recent batch of Gallup polls showing trouble for Democrats heading into the 2010 midterms. I was questioning how plausible it was for Gallup to use a weighted registered voter sample in which 44.5% of the respondents weren’t employed, and in which Gallup assumed an equal 47% number of GOP/GOP-leaning and Dem/Dem-leaning respondents.
My concerns stemmed from my experience in the 2004 campaign when I and others noted that Gallup’s likely voter model assumed a 40%-33% Republican edge over Democrats, yielding poll results that made Gallup’s reporting of a large Bush lead an outlier amongst other pollsters.
When I circulated these most recent concerns about Gallup around to other center-left bloggers last night, they went nowhere and were shot down by the only two who responded. One well-known left-coast economist said that a 44% nonworking number amongst registered voters meant that there were 56% employed registered voters, which tracks (somehow) with the 58% of the overall working-age population that is employed. Actually, 58.6% of the working-age population is employed as of March, leaving 41.4% of the same population not working. But I’m willing to defer to more professionally trained people on the relevance of a difference between Gallup’s assumption that there are 44% nonworking registered voters, and the reality that 41.4% of the working-age population is not working.
Another blogger (who's been typically pessimistic about Obama and 2010) told me that Gallup’s methods are consistent with those of other pollsters, and that Democrats should take seriously how bad things will be for them this November. This blogger apparently disregarded what Gallup did in 2004, and isn’t aware that Gallup’s assumption of an equal number of GOP/GOP-leaning and Dem/Dem-leaning respondents amongst registered voters doesn’t track at all with the Washington Post poll of just last week, which reflects a ten-point advantage for the Democrats (51%-41%) rather than a 47%-47% split with leaners included. Even though these are different polls, such a discrepancy is significant and could lead to Gallup overstating the depth of eventual Democratic opposition at this time, or as we move towards November.
The issue I’ve had with Gallup is their ability as the country’s preeminent political pollster and media brand to set a narrative in place with a lazy media that makes a GOP rebound the conventional wisdom.