Friday, May 4, 2012

Swift Boating? Probably a Dud in 2004

"Swift Boating" seems to be in the news this week, so here's my quick contribution:

Swift Boating  -- and the George W. Bush campaign in 2004 in general -- was probably a dud. The most important thing to know about the 2004 election, when it comes to electioneering effects, is that Bush underperformed the "fundamentals" models.

Nate Silver has conveniently gathered some of the numbers on this. Of the 15 prediction models he collected, 14 picked Bush as the winner, and of those 13 predicted a larger margin of victory than the 2.4% that Bush won by. The three predictor systems (Hibbs, Abramowitz, Wlekien and Erikson) that have performed best over the years all picked Bush, and by anywhere from 3.4% to 7.5%. I don't really put a whole lot of weight on the other systems, but for whatever it's worth they generally erred even more on the side of a big Bush win.

It's far short of proof, and of course it's certainly possible that one part of the campaign was a success but that another part was an even larger failure. But overall it suggests to me that anyone who believes that the Bush campaign did a better job than the Kerry campaign in general, or that the Swift Boat smear in particular helped Bush in any significant way, must meet an even higher burden of proof than if we didn't have this additional evidence. In particular, it just won't it to say "Bush won" as if it's an argument clincher. What really appears to need explanation is why Bush did worse than he should have, not why he won.

I'm going to get into this in more detail later, but that's the bottom line: there's no good reason to believe that the George W. Bush campaign in 2004 was responsible for his victory.


  1. I've looked over the ANES data from 04 and my guess would be that Bush underperformed because of foreign policy. I don't know the numbers off the top of my head. But I don't recall him getting good marks regarding the two wars. I'm sure the models account for this in some manner. But perhaps people were weighing their interest in these issues a bit more than normal and that caused the underperformance.

  2. I know it would be silly to say the campaign caused Bush's victory, but wouldn't it be fair to say it at least served some role as a necessary condition for his victory?

    To put it another way, the models seem to build in an assumption of a basically competent campaign, or at least campaigns that aren't extreme outliers. Because of sample size issues and also the fact that political campaigning is probably a reasonably efficient system, it's unlikely we'd see a campaign on the extreme low end of things. I'm not talking about not doing enough polls or not spending enough on TV in Ohio. I mean huge mistakes like if the campaign arranged for Bush to speak at a KKK rally or Karl Rove told Tom Brokaw to 'eat a dick' on live TV. And on and on and on.

    So the models are useful in an 'all else being equal' sense because for the most part all else is equal. The candidates have professional campaign staff who do all the things that modern political campaigns entail. And while campaign effects are rarely going to give a candidate much of an advantage, at least in general presidential elections, we can at least thought experiment our way to a scenario where a campaign is bad enough to tank an otherwise electable candidate. So it seems fair to say campaigns do have an effect, just not a particularly interesting one.

  3. I have always thought that Kerry ran a good campaign after Labor Day, at which point he called the Iraq war a terrible mistake, and that Dems turned on him cruelly in the bitterness of their disappointment. Nice to see some verification.

  4. Two armies meet. The pundits run statistical models and all predict team red will win. Team red deploys mustard gas. Despite this, team blue nearly wins. But, according to the logic of this article, the mustard gas was irrelevant, because team red was expected to win anyway. No. If team blue came close despite the mustard gas, it would have probably won without it, and the base issue is that on this one the pundit's models were wrong.

    1. Anon,

      What evidence do you have that it was "mustard gas"? What evidence do you have that the Swift Boat campaign was effective?

    2. No statistical evidence of either, although as an interested observer my money is that the SB campaign, and lots of equally ugly stuff did matter. My point, less polemically is just that I think Nate, and to a lesser extant you, are updating a long way on the basis of pretty noisy models (the models in my profession are pretty noisy too, this is not intended as a slight), and that an equally plausible story is that on this particular election, the models over-predicted how well GWB would have done absent the dirt, and that the dirt really mattered.

  5. Mr. Bernstein...the fact that you still call the Swifties telling of the truth a "smear" 8 years later proves that their campaign worked.


Note: Only a member of this blog may post a comment.

Who links to my website?