In early May I wrote The Silence of the Phones observing the lack of polling calls after cutting off the land line. I wondered if polling was skewed by not including the increasing population of cell phone only Americans. Since that time I’ve received repeated assurances that pollsters make sure their polls are correct by matching their respondent demographics to the target poll group.
In three major election surveys comparing results with and without cell phone users in polls the Pew Research Center has observed “a consistent difference of two-to-three points in the margin.” The article states
My question now: when do “small, and not statistically significant, differences” of two to three percentage points become recognized as real and significant? If you get the same two point spread over and over does it not illustrate a two point bias in your method?
The Pew study strongly suggests that Obama has been doing two to three points stronger than our pollsters report. The professionals need to revisit their weighting techniques.
People write …