Retrospective wisdom on the campaign

Time for another dispatch on the bizarre epistemology of campaign journalism. As I noted yesterday, the conventions of campaign journalism require journalists to obsess over the twists and turns of the presidential horse race and to ignore the fact that its outcome is highly predictable. But everything changes after Election Day. For instance, CNN.com's retrospective blithely asserts that McCain's campaign was "always a long shot":

What would have been his greatest political comeback -- to seize the White House -- proved to be too difficult. A Republican win in what is being seen as a "Democratic year" was always a long shot.

Whoever was the GOP nominee was going to have to fight against the legacy of the previous eight years of a Republican president who became highly unpopular because of the Iraq war, administration gaffes such as the handling of Hurricane Katrina, and what turned from a credit crunch into a global economic crisis.

The Arizona senator even managed to make the race appear competitive, soaring in the polls on the back of a polished convention and popular VP pick, Gov. Sarah Palin. But he made mistakes, too. Combined with the electorate's disenchantment with his party, it ensured defeat.

Did CNN let its viewers know that McCain's campaign was "always a long shot"? Or did it relentlessly hype the horse race? I don't watch a lot of cable news, but I think we all know the answer.

The irony, of course, is that CNN would have constructed a similar narrative concluding that a McCain victory was inevitable had he won. See the anecdotes about Newsweek's post-election packages from Larry Bartels and Matthew Yglesias that I cited last month. Reporters have an amazing ability to predict the past.

PS Memo to CNN: Sarah Palin was not "popular" by the end of the campaign.

Comments

[Jinchi]

journalists to obsess over the twists and turns of the presidential horse race and to ignore the fact that its outcome is highly predictable.

I agree that journalists obsess too much over the minutiae of the campaign, but as a scientist who uses models to evaluate data I think you're far too confident in the performance of your election models and too dismissive of the importance of true campaign coverage.

The models you reference are extremely underdetermined due to lack of data in a system that undoubtedly has more than 2 variables. This may not matter in most elections, but that doesn't mean that elections are highly predictable. The Bread and Peace model you cite was significantly in error in 4 of the 14 elections it covered.

I could forecast the weather in my town next week simply by looking out the window today - I'd be right more often than not, but the differences matter.

[David]

I agree with Jinchi about the models not being that terribly certain. They're based on a small number of data points. And, chances are the relationship between "fundamentals" and election results changes over time.

[Sean-B]

I agree with the others. It seems inconclusive. Agree with Jinchi about Brendan dismissing campaign coverage. He disparaged 538 in another post, but their coverage was some of the best of the campaign: visiting campaign offices across the nation and reporting on levels of enthusiasm among volunteers. And their coverage was surprisingly even handed.

Interesting analysis of the ground game to election results here.