Don't Trust California's Exit Polls [CORRECTED]
CORRECTION Feb. 6, 1:05 p.m.: Well, I was wrong. Contrary to my original post, California's exit poll data did include a sample of early voters polled via the phone. From Associated Press' account of the methodology:
There
were 17,454 interviews of Democratic primary voters, and 11,205 GOP voters.
Results included 1,005 telephone interviews of Democratic absentee voters and
813 GOP absentee voters in Arizona, California and Tennessee.
Overall sampling error was plus or minus 1 percentage points for both parties.
We gleaned our information from CNN, which has a misleading exit poll description of its own. What CNN says about their exit polls:
Exit polls are a survey of selected voters taken soon after they leave
their voting place. Pollsters use this sample information, collected
from a small percentage of voters, to track and project how all voters
or a specific segments of the voters sided on a particular race or
ballot measure.
No mention of an early voter phone survey. The entire country uses the same exit poll data, which means CNN uses the same data the AP uses, both from Edison Media Research. I spoke to someone in their office who confirmed the AP's account that a poll was conducted by phone with early voters. We regret the error. The original piece is below.
This item was cross-posted at Slate's Election Scorecard.
Poll junkies beware: California exit polls are not to be trusted.
California
has issued 5.5 million absentee ballots for today's primary, reaching
more than one-third of the 15.7 million total voters registered in the
state. As of yesterday, 3 million ballots had already been returned,
and state officials expect about 75 percent of the ballots to be
returned by the close of polls—that's 4.125 million people who voted
without pulling a lever. (These numbers include both Democratic and
Republican ballots.) The remaining ballots are expected to be turned in
at polling stations today, just like you drop off a movie rental.
For
our purposes, it's the 3 million ballots that have already been sent
back that may play havoc with expectations tonight. Exit polls, as
their name implies, measure only the opinions of residents who go to
the polls and submit a ballot. If you don't show up to the voting
booth, you're not going to be part of an exit poll.
Conventional wisdom suggests early voters chose Hillary Clinton over Barack Obama, since Clinton led in the polls until recently. In the most recent SurveyUSA poll,
34 percent of respondents told the pollster they had already voted,
which echoes the predicted vote-by-mail participation rate. Of those
voters, 54 percent favored Clinton and 37 percent favored Obama. We
should note that SurveyUSA shows a stronger Clinton lead than other
polls (Zogby, Rasmussen) released in the last 48 hours.Given
all this, exit polls that show a slight lead for Obama actually may be
bad news for Barack. If SurveyUSA's poll is correct—no sure thing—then
Obama will need a strong lead (around 10 percent) among today's voters
to win California.