Accountability Watch: Who Is Predicting What

By James Fallows

A fascinating interview this afternoon ran much, much longer than expected, which is a good thing. But if I am to get back to the pace of the one-day-only, all-day Festival of Election Eve™ updates, I can't spend much time actually setting up items. So I won't explain why I think prediction is usually the least useful thing journalists can do -- even though huge amounts of pundit time and effort go to tell us what they think is going to occur. I refer you to the classic text on the subject for more. And, as I'll explain further another time, I see a difference between prediction in the tactical sense -- "I really think Hillary is going to run next time" -- and trying to assess the long-term consequences of different choices or policies. (Eg this and this from The Atlantic.)

Here's why political prediction is morally inferior to sports-line wagering or other kinds of normal betting: In pundit-world, the losers never have to pay off. You can assert with blowhard certitude that this or that candidate looks strong, this or that voting bloc is going to turn out, this or that strategy will be effective. If you're right, you play up that fact. If you're wrong, no one seems to notice or care. In Vegas, you have to pay up. In pundit land (or "we need to invade Iraq, now!" land), you just move on. That's why, to give yet another argument in shorthand, I think it's good rather than bad if people who are making a big deal of their predictive talent are willing to back their views with actual bets. I will flesh out that argument some other day.

For now, two data points:


1) Let's talk with these people in a few days. A number of people whose claim to public attention is that they know about politics are flatly predicting a blow-out win for Mitt Romney tomorrow. There's a nice (if maliciously) illustrated guide to them at The Blaze. Another roundup is here. These names and electoral-vote predictions include, starting with the four pictured above:

  •  George Will: 321-217 electoral vote landslide for Romney
  •  Michael Barone: 315-223 for Romney
  •  Dick Morris: "a landslide"
  •  Karl Rove: "at least" 279 for Romney, meaning at most 259 for Obama
  •  Joe Scarborough: close race with Romney in a better position.
  •  Charles Krauthammer: a "very close" win for Romney, which means 270 electoral votes or just above
  • Peggy Noonan, "a [Romney] win"

Remember, these people's claim to fame -- especially in the case of (in their respective primes) Barone, Morris, and Rove -- is that they know something special about politics. If they are putting their names behind these predictions, presumably they would like us to take them seriously. We'll see what happens in the next day or two: If they are right, all appropriate credit. But if they are not, this should be remembered, rather than just blown off. And similarly, if the "quants" who are unanimously predictable a sizable Obama win prove to be wrong, they should be made to explain.

(Why am I not making my own predictions? Because that is not the business I have ever been in.)

2) Avast! When I used PC's I used Avast! anti-virus software, from a company in Prague. (They now have a Mac offering, which I will check out.) They recently surveyed a very large number of users on presidential views -- and the results highlighted the ever-fascinating "preference versus expectations" divide.

The preference of Avast! users, and their own personal voting plan, was that Mitt Romney should become the next president:

Thumbnail image for map-who-will-you-vote-for.jpg

But their expectation was that Obama would win:
Thumbnail image for map-who-do-you-think-will-win.jpg

The "who we expect to win" map gives Obama some 349 electoral votes. We'll see how this matches up against the "experts."

This article available online at: