Sunday, February 24, 2008

Questionable Journalism

The Associated Press has a policy of not citing automated polls. It's a silly, backwards rule with companies that use IVR showing more and more by the year that they're just as or more accurate than traditional live interviewers. But that's their right and it's not anything that the main political writer for the AP in North Carolina, Gary Robertson, is responsible for.

Robertson is responsible, though, for misleading stories like this.

The Elon Poll said that a third of North Carolina voters are undecided in the Governor's race so Robertson wrote a lengthy story about it, basically citing it as fact without any evidence to the contrary.

PPP's most recent poll showed only 20% of likely voters undecided in the Democratic primary. Survey USA, which has had the most accurate numbers of any company in the country for the Presidential primaries, showed 22% undecided.

The Elon poll is great for gauging where public opinion in North Carolina is about various political issues, but because of the way it screens for likely voters (it doesn't), it's not a reliable barometer for election results. And I don't think they would claim that as their role- that's what companies like PPP and SUSA specialize in both in North Carolina and at the national level.

Yet Robertson wrote an entire story about the Governor's race relying on one poll that probably is not a great indicator for what's going to happen in May. He is on our media distribution list, so it's not like he doesn't know what the other polls are saying.

It's not Robertson's fault that he can't cover the polls of most companies regularly talking to North Carolina voters. But it is his fault if he writes lengthy feature stories trumpeting one poll that is showing significantly different results from companies that specialize in election polling without granting that the individual poll may be an outlier.

Robertson didn't even mention in the story that the margins of error on the polls for the Governor's race were abnormally high for a publicly released poll.

I don't think writers should extrapolate entire stories about political races based on any individual poll, including ours. But they certainly shouldn't do it for polls that don't attempt to get a sample representative of what the primary electorate will be and let respondents skip back and forth between parties within the poll either.

It results in a misleading story, and that's a disservice to the small papers across the state that rely on the AP for their state political coverage.

1 comment:

Anonymous said...

Yep, pretty much, Tom. Chris Hayes made a similar point over at the Civitas blog.

Web Statistics