Why Polls Can be Off
Survey USA has a new poll out showing Obama ahead of McCain in the crucial state of Ohio by 48%-39%.
This is, to say the least, an odd result, since the same group's recent poll showed Obama beating McCain by only 8% in California. When you look at the cross tabs for this survey, you discover that 52%of those surveyed were Democrats (Obama wins this group by 59%), 28% Republican (McCain wins this group by 69%), and 18% independent (McCain wins this group by 10%).
Does this distribution among the Parties sound right for Ohio? If the number of Democrats were 7% fewer, and the number of Republicans 7% higher (a 45-35 split) you get a dead heat.
Two other recent polls in Ohio show McCain up by 1% and 4% (Rasmussen and Quinnipiac). When you get a poll result that does not look right, there usually is a reason. And the reason is more often than not, a poor distribution between the number of members from each of the two parties who were surveyed.
Richard Baehr is American Thinker's Chief Political Correspondent