January 20, 2004

Yet More Iowa Analysis

After sleeping on it, I believe Iowa came down to two main ingredients.

First was the wildcard of Dean's supporters. No one knew how well the polls represented them. Were they a small number of very activist supporters - narrow, but deep? Were they the only visible parts of a vast army? No one knew the answers to this, so the press and pundits felt an excuse to dismiss many of the polls. This is why Dean's finish was considered such a disappointment even when he wasn't in the lead in any poll beforehand. It turned out that Dean was relatively popular, but the supporters were inexperienced, so were unable to effectively horse-trade at the caucuses.

Second was the nature of caucuses. I saw a study that looked at a hypothetical caucus. Say you have four candidates, each with equal support. You need 15% to be viable, but every precinct has only 59 delegates, which mean only the top three for that precinct become viable - essentially random which one that would be.

It comes down to second choices. Two of the candidates have supporters that hate each other and never choose each other as second choice. The other two candidates have second choices evenly distributed.

The end result is that the two candidates that hate each other would end up with 17% apiece, while the other two would end up with 33% apiece - even though they came in with equal support.

A version of this is essentially what happened. Dean and Gephardt went negative on each other. Kerry and Edwards either stayed positive, or in Kerry's case, acted like they were positive.

Combine this with the deal that Edwards and Kucinich had - where if they weren't viable they'd go to the other guy - and you've basically got the caucus. The nature of the caucus basically multiplied the effect of all the negative coverage that happened to Dean. I don't believe that Dean's statewide support is only 18% though.

Posted by Curt at January 20, 2004 05:08 PM