Monday, 29 December 2003

Dean's lead

While Stephen Green points out that the latest Zogby numbers show Howard Dean in statistical dead heats in both Iowa and South Carolina (but with a commanding lead in New Hampshire), James Joyner retorts that Dean’s lead is more durable than the numbers indicate:

...this is now Dean’s race to lose. While he’s not running away with anything, he’s got a huge lead in New Hampshire and a small one everywhere else. Meanwhile, there is no consistent number two. More importantly, he’s absolutely dominating the money primary.

More to the point, the rules are such that you can effectively discount anyone not named Al Sharpton* unless they get double-digits, because the Democrats’ delegate allocation system works at the congressional district level—and, as I keep pointing out, you have to get 15% in a congressional district to win delegates from it. This effect will massively inflate Dean’s standing at the convention.

Like James, I’m becoming more convinced than ever that South Carolina will be pivotal. And, barring a seismic shift after Iowa, I can’t see any S.C. scenario that Dean can’t spin as a win—realistically, he needs to be blown out by 10% or more by a credible candidate (at this point, either Clark or Gephardt), which just ain’t happening with Edwards still in the race and a lot of Republicans coming out to vote for Sharpton. A narrow ABD victory gives Dean the line that he “polled well in the South,” even if he only gets a third of the primary vote.

* Sharpton will likely pick up a decent share of delegates in the so called “majority-minority” congressional districts due to racial gerrymandering, even though his statewide numbers in most states are anemic and he hasn’t received many endorsements from establishment Democrats. This will be further enhanced because congressional districts that more heavily favored Gore in 2000—many of which were majority-minority—receive extra delegates.

Also, one minor demerit for Zogby: since the poll numbers are measured with error, it’d be nice to see the 95% confidence region on the bars so the reader can decide whether or not the differences are meaningful.