A recent Los Angeles Times poll showing President Bush trailing Democrat John Kerry by 7 percentage points used a sample skewed toward Democrats.
According to the Capitol Hill newspaper Roll Call, the sample surveyed was made up of 38 percent Democrats and 25 percent Republicans – a difference not representative of the actual electorate.
In the past three presidential elections, the margin for Democrats in party identification ranged from 3 percent to 4 percent – well below the Times’ 13 percent.
The core problem with the survey, Roll Call said, is the Times’ Republican sample of 25 percent – a full 10 percentage points lower than it should be.
The Times defended its poll, which was released last week: “Party ID is a moving variable that changes from one election to another, and weighting by party registration makes no sense nationally because many states don’t have their voters’ register by party.”
Wrote David Winston in Roll Call: “The number that really jumped out to most analysts was the poll’s finding that the Democrats currently maintain a generic-ballot lead of 19 points. I’ve been doing congressional polling for nearly 25 years, and I’ve never seen a generic lead anywhere close to this.
“Yet despite these staggering figures, there was no story in the Times saying the Democrats were clearly going to win the House or take back the Senate. If their generic were true, that’s the kind of sea change we would see and that’s new. But no other polling operation even comes close to that number.”
Winston is president of The Winston Group, a Republican polling firm. He credited the Times with releasing the sampling data so analysts could “understand the survey’s bias.”