Take the pivotal Minnesota race. Over the weekend, the Minneapolis Star Tribune poll had Walter Mondale holding a 46 percent to 41 percent over Norm Coleman, with a margin of error of +/- 3%. Meanwhile the Pioneer Press/Minnesota Public Radio poll had Coleman with 47 percent to 41 percent for Mondale, with a margin of error of 4%. Yet neither poll can be considered terribly accurate, given the small sample of the survey — neither poll had the generally accepted minimum of 1,000 respondents to bring the poll within the acceptable 3 point margin of error, although the Star Tribune came close with about 960 respondents.
“You can’t toss a poll out there with sample pool of 400 and call it accurate, or even close to accurate,” says a Republican pollster. “You have the media touting numbers that just aren’t realistic. Forget the soft money ban. Ban lousy polling.”p>