Sunday's national polls once again are all over the lot. Nine polls are reported on Real Clear politics. They run the gamut from a 2% lead by Obama to one with a double digit lead for Obama. There are 9 polls with 7 different results. Only in two cases do polls agree. Again, with two days left until the election, one would think that the polls would cluster around the likely result. Instead, they continue to get further apart. The only realistic conclusion is that there is something more at work here than statistical variation. Indeed, the likelihood that these polls are varying as they do just due to statistical variation is infinitesimal.
It is true that some of the polls use a different bias in their samples. Those which have drunk the Obama cool ade and decided that hordes of young or new voters will appear at the polls have weighted their results to reflect those groups. Others have use a more conventional analysis. Still, this is not enough to explain the major variations in the results. We do not see two groups of clustered results, rather, the results do not cluster.
I cannot wait to see what actually happens on Tuesday and then watch as the pollsters explain how they were right all along.
No comments:
Post a Comment