The post below points out that the new Ipsos-AP poll has Bush’s approval rating at just 45 percent, in rather stark contrast to Gallup’s 57 percent rating, which was accompanied by a 9 point Republican party ID advantage.
But it’s interesting to note that the Ipsos poll also has a lop-sided party ID advantage–but this time for the Democrats (+12). The fact that the Ipsos party ID figure is for RVs and the Gallup figure for all adults hardly seems adequate to account for this vast difference.
In truth, neither figure seems terribly credible and, therefore, both approval figures are probably outliers driven by the party ID composition of their samples. Certainly neither figure should be taken particularly seriously on its own, though you could average the two if you wish. In that case, you get a 51 percent approval rating for Bush, pretty much in line with other figures from recent polls.
What is to be done about these wacky partisan samples, which give such misleading pictures of current politics? Perhaps it’s time to revive “dynamic party ID weighting“, an idea whose time may finally have come. Aruguably, this is the time to pursue such an innovation, away from all the passions induced by a political campaign. And, if pollsters did so, I think it would help smooth out poll results and avoid the fake surges this way and that that are starting to erode faith in the veracity of polling.
Of course, down at Gallup and many other polling headquarters as well, the view is probably that all is fine. I can assure them that all is not fine and it is time to trade in their stone-walling for a bit of listening and openness to change.