Saturday, May 14, 2005

How to Fix the Wrong Problem 

Via Zemblan patriot J.M.: At last, an explanation for the mysterious epidemic of exit polls that showed John Kerry winning the popular vote and carrying key swing states, but later turned out to be wrong, all wrong, outside the statistical margin of error and always to the President's advantage:
Better training of interviewers to get a proper sample of voters after they cast ballots will be key to improving the performance of exit polls, one pollster who handled the 2004 election surveys said Saturday.

Exit polls on Election Day 2004 overstated support for Democrat John Kerry overall and in many key states, which led to widespread confusion that day about the election eventually won by President Bush.

The exit polls contacted more supporters of Kerry than of Bush because of "the failure of interviewers to follow the selection rate," said Warren Mitofsky, who conducted the exit polls along with Joe Lenski of Edison Media Research.

In exit polling, properly trained interviewers are supposed to follow a carefully designed strategy of contacting voters, such as every fourth or fifth voter, to get a random sample.
The problem being that, if you interview every third voter, or every sixth voter, or two in a row from a group of eight, you'll invariably wind up with a massive oversampling of Democrats, while Republicans will be hugely underrepresented.

Always. In every last case. As we saw in 2004.

Any decent statistician will tell you that.

Got it?

| | Technorati Links | to Del.icio.us