Thursday, September 16, 2004
"Both polls could be right." ... The differences are within the margins of error of the two polls. He also noted a slight difference in the times the polls were conducted; that the two polls have different ways of identifying likely voters; and that undecided voters tend to change their minds frequently.
"It's a tight race, and it can go back and forth almost every day," Frank said.
Steve's got loads of experience, so I'm willing to believe he's right about this, but it's jarring to think that you could change a poll by 11% by changing how you filter your calls down to likely voters. (If you're reading Steve, you have the floor. I'd like to learn about this.) Also, the MOE of the PiPress poll is 4% and that of the StarTrib is 3%; the difference being 11%, I am not sure Steve's right about these two polls not being statistically significantly different.
To its credit, the STrib has a description of its polling method along with the caveat that they had a higher number of people willing to respond to their callers than usual for Minnesota. The description of the PiPress/MPR poll is here. That page included the result that 23% of respondents "had concerns about what Kerry did" in Vietnam, while 14% had concerns about Bush's service.
Meanwhile, Mitch notes today that the national polls show higher negatives for Kerry than Bush; the StarTribune today states that half of its poll respondents believe Kerry flip-flops.