4/17/12

Omens, Augurs, and Polling

Let's talk about superstition. If you wanted an unconventional -- but still accurate -- definition of superstition, you could say that it's the belief that unscientific methods can yield scientific results. Walking under a ladder predicts misfortune, while a rabbit's foot can skew results in your favor. Given this definition, a very short post at Buzzfeed may qualify as superstitious.

"Modern presidents who got re-elected were all leading in the polls at this point in their presidencies — as were some who lost anyway," wrote Zeke Miller yesterday. "Obama is in a statistical tie with Romney in the first Gallup daily tracking poll of the general election, and that might not be enough." The post goes on to spell out Obama's obvious electoral doom with a graphic.

But Nate Silver has a must-read post up about reading polling that pretty much blows this thinking out of the water. And when I say must-read, I mean go read it. In fact, bookmark it and check it later, when some partisan starts waving around some favorable finding or the media starts running wild with their close race narrative.

Silver's post consists of twelve rules to apply when reading about polling and it's the final rule that knocks out Buzzfeed's pseudo-scientific reading of history and the Gallup poll.


Don’t over-learn the lessons of history. A final and more general point is that there have been only 16 presidential elections since World War II. That simply isn’t a lot of data, and overly specific conclusions from them, like “no recent president has been re-elected with an unemployment rate over 8.0 percent” or “no recent incumbent has lost when he did not face a primary challenge,” are often not very meaningful in practice and will generally not carry much predictive weight.


I don't bring this up to criticize Miller or Buzzfeed specifically, but to point out that we see way too much of this crap in the media. If we take history as predictive, then we have to assume that Barack Obama could not possibly have won either the nomination or the presidency, because history shows that an African-American had never won either.

This is what the media does. Sometimes because of carelessness and other times intentionally. They pretend that all information is useful, even when it's not, and that every crystal ball works. And if it promotes the tight race narrative, even better. In this case, Obama's ahead in most polling but -- oops! Lookie here; history offers another way of looking at them. In this case, a way that allows them to discount the results of most polling -- e.i., a way of looking at them that's not scientific at all.

"This race is way tighter than the polling would suggest," a talking head might say. "Make sure to check in later -- things can change in a heartbeat!"

And that's what all this misreading of polling and the search for signs and portents is all about; eyes on the page, eyes on the screen. Often times, the media is biased. But it's not a left/right bias. It's a bias in favor of ratings or hits or circulation. A tight race is an exciting race and an exciting race draws eyes. So -- no matter what the reality of the contest might be -- a tight and exciting race it's going to be for as long as they can keep that story going. If that means disemboweling birds to read their entrails, then bring out the knife.

Maybe the best way of looking at this is by using an old truism -- there's a first time for everything. Basically, this means that historical trends are always true right up to the point that they're not anymore. An African-American never won the Democratic nomination, until one did. An Irish Catholic never became president, until one did. This is all as silly as telling the Wright brothers that since humans had never flown in heavier than air flyers before, it was impossible that they ever would.

Polls are factual data (most of the time). It's those who interpret polling you should look out for. When it comes to history and the desire for ratings, they become a very superstitious lot.

-Wisco

No comments:

Post a Comment