10 January 2008

The polls were wrong?

While voters in Newfoundland and Labrador are used to seeing utterly useless political polls reported by local news media as if they were correct, the same in not the case in the United States.

Hillary Clinton's three percentage point margin of victory in the New Hampshire primary is the talk of the pundits, since pre-vote polling showed Barack Obama decisively ahead.  Political polls are not known for being out of whack from the vote by as much as 19 percentage points, or even seven, as was the case with recent polling in this province.

salon.com writer Joe Conason blames the "mainstream" media's anti-Clinton bias for what he contends was a backlash by voters in favour of Clinton.

A more plausible explanation focuses on the polling methodology.

Start with this summary of the results from the Washington Post:

Most polls accurately reflected the large bloc of likely Democratic voters yet to make up their minds, or said they were open to switching their support in the closing days. On the network exit poll, nearly 4 in 10 said they made their final decision within the last three days; 17 percent said they decided how to vote Tuesday. Among those making up their minds on Election Day, 40 percent supported Clinton, 37 percent Obama. Clinton did even better among the half of the electorate who settled on their top choice a month or more ago.

But the late polls missed on how votes divided by gender. Pre-election polls from CNN-WMUR-University of New Hampshire and USA Today-Gallup showed Obama and Clinton about evenly splitting female voters and Obama winning men by a margin of 2 to 1. But Clinton won among women by 13 percentage points, exit polls showed, and she lost among men more narrowly than suggested, drawing 30 percent to Obama's 42 percent.

What the polls may have missed, as the Post notes, is what the undecideds finally decided to do. That's a plausible explanation but one that awaits further detailed analysis by someone with access to the data. No matter what, it's much more plausible than the so-called "Bradley effect' in which white voters are presumed to hide their real voting intentions when an African American candidate is on the ballot.

The New York Times suggests that the impact of having two historic candidates - one black, the other a woman - on the Democratic ticket, coupled with the relatively short time gap between the votes in Iowa and New Hampshire may have had an impact on poll results.  NYT also suggests that methodology - in this case the use of tracking polls and their high margin of error - had an impact if not on the result than on the reporting of the polls.

And that's where another well-known phenomenon emerged, according to NYT:

Still, reporters who cover a particular campaign face a special challenge that was documented as far back as Timothy Crouse’s chronicle of the 1972 presidential race, The Boys on the Bus: Their ability to take in all that is happening may be limited.

At 8 a.m. Wednesday, Joel Achenbach, a reporter for The Washington Post who had been covering Mr. Obama in New Hampshire, posted a mea culpa on the newspaper’s Web site.

“Count me among those who thought Obama was a runaway train, that he’d blow Clinton out of the water,” Mr. Achenbach wrote. “You had to see the crowds! Feel the energy!”

Inaccurate polls and media commentary influenced by time spent on an airplane with the candidate?

What a novel idea.


Additional links:

USA Today:  "Pollsters struggle to explain Clinton win"

Telegraph (UK): "Inquiry after pollsters miss Hillary Clinton win"

CBC Newfoundland and Labrador campaign blog 2007:  "Embedded on the big blue bus"