Many polls have a huge Republican bias in there polling data. However, a closer inspection shows something quite different. I just finished reading about the Sestak poll in Pennsylvania today. It showed a huge lead for Tommey. However, there appeared to be a serious flaw in the sample group.
The poll indicates that this sample group favored McCain by three points in 2008. They currenty favor Toomey by 8. That would indicate a five point shift towards the Republican since 2008. However, in 2008 Pennsylvaniz voted for Obama by seven points. In other words had this polling sample been used in 2008 the results would have been 10 points in error. Why should we believe this same group is not as skewed towards the Republicans this year.
Another way to look at this is to suggest this group moved five points more Republican this year. Okay now assume all of Pennsylvania moves five points more Republican that still leaves a two point Democratic edge. When you apply any time of norming to Pennsylvania's actual demographics the Toomey lead disintegrates.
When I looked into a PPP poll on the same race I found a similiar situation. The polls showed a four point Republican shift from last year. When normed for Pennsylvania's actual Demographics Sestak appears stronger.
This all indicates that the polling is double counting the Republican shift this year. If that is the case in Pennsylvania and in other states as well it might indicate that Democrats are doing better than the media expects