In August I discussed the possibility that people were trying to use a bit of money to fiddle with the poll of polls. My hypothesis was that now that averages of polls are getting more attention than the well known polls taken alone (a reasonable change in emphasis) it is possible to make the average poll for a candidate look better by setting up a polling organisation with no track record and credibility and issuing fiddled polls (so that it will never have any credibility). Since political operatives all agree that a poll which would be good news if value is good for a candidate, the bogus polls would be designed to give the favored candidate high mis-measured support.
Since this requires no lack of money and a lack of scruples, I would guess that the new pollsters who only hope to get tossed into an average favor Republicans.
My August list started with Foster McCollum White Baydoun which for which Nate Silver (of fivethirtyeight) estimated an 11% pro Republican house effect. It also included “We Ask America” and “Purples Strategies.” The CEO of Purple Strategies is Alex Castellanos who is not only a political operative but also notorious for the most recent (now blessedly decadesold) ad which directly appealed to white anger about race (the white hands ad). Now he and a Dino who handled pr for BP are giving fair and balanced polling (fair and balanced being used in the Fox News sense).
Rasmussen is different. I think their success in pulling perceptions towards Republicans in spite of their demonstrated statistical bias (3.8% in 2010) is the cause of the new strategy.
After that too long introduction, I have an addition to the list Gravis Marketing. People much better informed than I about polls and pollsters haven’t heard of them. They have a Republican house effect. according to Nate Silver Importantly, Silver only partially removes house effects from his averages, and other aggregators don’t remove them at all. Back to the first link to Silver discussing Foster McCollum White Baydoun
We do not subtract out the entire 11-point house effect from the polling firm’s results — the model allows polling firms to retain some of their house effect — but the model does adjust the poll substantially, treating it as about a 7-point lead for Mr. Romney rather than a 15-point one
Aggregators TPM and Real Clear Politics who (very very defensibly) use simpler easier to understand approaches have Ohio shifting from a small Obama lead to a tiny Obama lead. Two of the 6 polls in the RCP average are from members of my list of suspect pollsters,, there is also a purple strategies poll) and one is from Rasmussen. So I suspect that the average is half bogus. IIRC the Gravis poll caused enough of a shift to change the TPM rating from leans Obama to Toss up with dramatic effects on their automatic race summary (no link as it updates continuously).
This, I think, is the whole purpose of the brand new Gravis marketing political polling operation.
Update: But I don’t know this. I don’t have enough data from Gravis Marketing to make any firm inference. I can’t tell much about any single pollster. My view is that there is a brand new pollster group effect which is significantly Republican.
I can’t tell anything with confidence about any single brand new pollster (except that Foster McCollum White Baydoun sure doesn’t seem to be doing a very good job).