Real Clear Bias
Some have accused Real Clear Politics, an avowedly right-leaning site, of favoring polls that are good for McCain. They have no stated rules for which polls go into their average, especially on the state level, so it seems likely that their “gut feeling” approach to which polls to include would lead to such results. But on some things they do have clear rules — for instance, it’s consistently been the case that a state where a candidate is leading by 9.0 or more is “solid” for that candidate.
Imagine my surprise, then, when I looked at the map this morning and saw Pennsylvania listed as “leaning.” Have the polls fallen sharply? Turns out the answer is no: in fact, Obama is still up by 9.5 in the averages. More disturbingly, that average itself seems to be a lowball. Looking at the Pennsylvania page, one can see that there are several polls that fall close to the range of the average (10/23-10/26) but are excluded — for instance, the Quinnipiac (10/22-10/26) and AP/GfK (10/22-10/26). A one-off poll for 10/26 is included, but these two — both of which have bigger sample sizes, substantially bigger in the case of Quinnipiac — do not make the cut for some mysterious reason. Coincidentally, the excluded posts would bring up Obama’s average, while the included one brings it down slightly.
The questionable choice of individual polls is perhaps irrelevant, however, in the face of their apparent shift in policy on which ones are “leaning” or “solid.” It was easy to pull it off in this case because Pennsylvania is currently the only state within the 9.0-9.9 range, so less obsessive readers likely wouldn’t notice it and would assume that 10 has been the cutoff all along — and perhaps the exclusion of the two polls listed above ensures the sub-10 average as well. Overall, it’s hard not to ponder the possibility that they’re manipulating their results to lend some credence to the idea that McCain is making serious inroads in Pennsylvania.
Sorry, the comment form is closed at this time.