Are 2008 polls off because white folks lie?
The folks over at Politico have an interesting piece this morning reminding us of California political history as they try to figure out why political polling in this cycle has been so off. One answer is something the pros refer to it as the "Bradley effect," the phenomenon of white poll respondents telling pollsters they'll vote for a black candidate -- but not doing so once they get into the privacy of the polling booth. Some think that was behind the gap between polls and results in former Los Angeles Mayor Tom Bradley's loss in the 1982 governor's race (and thus the name of the phenomenon).
Pollsters are trying to sort some of this out during their annual convention in New Orleans this week -- three out of five said they wouldn't get an answer (sorry, couldn't resist). And a study is underway to try to figure out what's gone wrong, looking at such arcana as whether pollsters' samples are not reflective.
Our guess? People have been genuinely undecided, and small events pushed them one way or the other. New Hampshire was the first notable poll flop, and a close look at the numbers could hold a key to the explanation.
All the polls showed Barack Obama with a healthy margin, but Hillary Clinton won by a 2.6% margin. But dig into the polls. The Suffolk/WGBH poll, for instance, found Obama leading but with 8% undecided; 6% of those who had decided were "very likely" to change their minds, and another 18% said they might change their minds. With that much volatility, it's hard to measure the impact of even something so small as a tear.
By the time the Indiana primary rolled around -- polls there gave Clinton an aggregate 5-point lead; she won with a 1.4% margin -- the pollsters had largely stopped asking about the depth of voter commitment. Maybe they ought to add that question back in.
But then, 40% of you don't care (you have to love a poll about polls).
-- Scott Martelle