For an even more detailed look at what goes on leading up to the caucus, Kossack desmoinesdem has provided us a very detailed description of what goes on beforehand, how the different campaigns win over supporters, and the intricacies of caucus math, in a series called How the Iowa caucuses work.
Part 1 (basic elements of the caucus system)
Part 2 (who is over- and under-represented when delegates are counted)
Part 3 (why it's hard to turn out caucus-goers)
Part 4 (more on why caucus turnout is low)
Part 5 (on second choices and caucus math)
Part 6 (on the importance of precinct captains)
FWIW, desmoinesdem herself is a precinct captain for John Edwards. What I'd like to highlight to show how bizarre the caucus system can get is from Part 5. This is the money quote.
I have told this story before, but I will tell it again because it permanently altered my thinking about the caucus system. The first time I was able to vote was in the 1988 Iowa caucuses. My brother and I flew home from the east coast (I was a college freshman, he was in grad school) to caucus for Paul Simon.
When voters divided into preference groups, Simon had a plurality in my precinct. Michael Dukakis was second, and Bruce Babbitt was third. No other candidate was viable. My precinct assigns 6 Polk County delegates. It looked as though Simon would end up with 3 delegates, Dukakis 2, and Babbitt 1.
Then the Dukakis and Babbitt people got together and realized that if enough Dukakis supporters switched to Babbitt, it would affect the caucus math. (The Dukakis supporters knew that Babbitt was not a threat to finish ahead of their candidate in Iowa, but Simon was.) When people were given the chance to make their second choice, Babbitt gained enough supporters to get a second delegate from my precinct.
Since caucus math is a zero-sum game, Simon, Dukakis and Babbitt all ended up with 2 delegates from my precinct, even though no one defected from our Simon group.
How mad was this idealistic 18-year-old, who had been excited to participate in the caucuses for the first time? Screwy caucus math erased the numerical advantage that Simon had in my precinct.
To see for yourself how a scenario like this can develop, I encourage you to read this post by Drew Miller. Drew was one of the founders of the Iowa progressive community blog Bleeding Heartland. He now works for the Iowa Democratic Party, so has had to stop blogging, unfortunately. But in his post on caucus math from earlier this year, he included a link to a "caucus calculator" he created in Excel format.
To use the calculator, you enter the number of county delegates awarded by your precinct. For most precincts, this number is between 4 and 9.
You then enter hypothetical raw vote percentages for up to six candidates. The calculator tells you how many delegates each candidate would get from the precinct under that scenario.
Prepare to waste a lot of time if you are enough of a political junkie to download this calculator. I entered 6 delegates, which is the number awarded by my precinct, and spent too much time playing around.
With three viable candidates, if candidate 1 has 42 percent of the voters, candidate 2 has 35 percent, and candidate 3 has 23 percent, then 3 delegates would be assigned to the top candidate, 2 to the second-place candidate, and 1 to the third-place candidate.
But look! If the supporters of the second and third-placed candidates pull a stunt like what happened in 1988 in my precinct, we might end up with the raw votes looking like this: 42 percent for candidate 1, 30 percent for candidate 2, and 28 percent for candidate 3. Drew Miller's calculator now tells me that each of the three viable candidates will get 2 delegates from the precinct.
Because caucus math is a zero-sum game, the gain for candidate 3 comes out of candidate 1's hide, even if no one leaves candidate 1's group of supporters.
Now, let's talk about second choices. Most voters will never need to express a second choice, because their first choice will be viable in their precinct. So it's a little misleading to look at statewide poll results and say, "Candidate X leads among second choices." I don't really care who the second choices of Edwards supporters in my precinct are, because I know Edwards will be viable. I am more interested in the second choices of the Richardson and Biden supporters in my neighborhood.
Similarly, I don't care about the second choices of college students who support Obama, because I assume that he will be viable in every precinct on or near a college campus. I am very interested, however, in knowing the second choices of the Obama supporters who live in precincts dominated by voters over 50.
Have I convinced you yet that neither you nor I nor any pollster will be able to calculate in advance who will benefit most from second choices? If not, read on.
You might assume that when it goes to second choices, caucus-goers can only choose among the candidates who were viable on the first division into preference groups. However, that is not the case. If one candidate is just a bit short of viability in the first count, it may be possible to bring over enough people to reach the threshold at the second count. That happened in my precinct in 2004. Dean was just one or two people short of the 15 percent threshold at the first count, but he did end up with a delegate in the end.
Let's say that Edwards, Clinton, Obama and Richardson are the only viable candidates in my precinct at the first count. Maybe the delegates would be split 2 for Edwards, 2 for Clinton, 1 each for Obama and Richardson. But what if the Obama and Richardson supporters have people to spare? They might get together and realize that if they transfer some of their supporters to Biden, they can bring him up to the 15 percent threshold in my precinct without falling below 15 percent themselves.
If they did that, Biden would automatically get a delegate, and Obama and Richardson would still get one each. But now either Edwards or Clinton would lose a delegate to keep the total number of delegates awarded by my precinct at 6.
If you think this kind of mischief won't happen on caucus night, think again. The best-organized campaigns (Edwards, Clinton, Obama) are going to make sure their precinct captains understand caucus math. They will have training sessions and conference calls shortly before the caucus.
They will give their precinct captains cards showing exactly how many individual supporters they need to get 1, 2, 3 or more delegates from their precinct. Once the precinct chair announces how many people have signed in, I will be able to check my card to see how many Edwards supporters I need to get to 15 percent, and how many additional people I need to be guaranteed of each additional delegate for Edwards.
What do the Clinton and Obama campaigns in Iowa have in common? They are more worried about beating each other than beating Edwards. They have decided (wrongly, in my view) that Edwards is not a threat to win the nomination, even if he wins Iowa. They would like to win Iowa outright, and they would also like their main rival to come in third or worse.
Now, no campaign is likely to admit to this publicly. But I would not be surprised if Clinton and Obama precinct captains are told privately that it's ok to mess around with the caucus math to deprive the main rival of a delegate, even if that means Edwards or Richardson getting an extra delegate.
By the same token, Obama precinct captains may be told that if Obama is not viable and can't persuade enough people to get to the 15 percent threshold on second preferences, it's better to send supporters to Edwards than to Clinton.
These tricks won't affect the delegate counts in every precinct. And those who try them had better be very sure that they know what they are doing. In the comments below the post by Drew Miller I linked to above, Bleeding Heartland user corncam wrote, "At my 2004 caucus, the Kerry people shifted some votes to Edwards, thinking that it would cost Dean a delegate, but they miscounted, and the new Edwards delegate came from Kerry, not Dean."
I'll (hopefully) clarify a bit about that weird caucus math. Because there are a fixed number of delegates per precinct, they need to be divided up so that they match up the percentage of support without overestimating their support as best as possible. In her 42%-35%-23% scenario, the best way to divide up 6 delegates is 3-2-1. The 3rd place candidate has 23%, which is greater than the 16% (1 out of 6) threshold needed to fully earn a delegate, and the 2nd place candidate has more than the 33% (2 out of 6) needed to earn two delegates. But the 1st place finisher hasn't actually earned three delegates, which would imply that person has 50% support. So there's an 8% bloc of support that the winner hasn't really earned. But if you went 2-2-2 in that scenario, then you're saying the 3rd place finisher has 33% support, which is a 10% bloc of support that hasn't been earned. Hence, 2-2-2 is a worse allocation than 3-2-1, because the 3-2-1 allocation screws over the least percentage of voters.
BUT... if caucus math mischief causes people to switch around so that it's now 42%-30%-28%, then the fairest division of the 6 delegates would be 2-2-2. The 2nd and 3rd place candidates haven't really "earned" that second delegate. So, there's a 3% overestimation of support for the 2nd place finisher and a 5% overestimation for the 3rd place finisher, for a total of 8% getting screwed. But it's "fairer" than a 3-2-1 allocation, which would overestimate the 1st place finisher by 8% AS WELL AS the 2nd place finisher by 3%, for a total of 11% of voters getting screwed. (The 3rd place guy isn't getting screwed under this rule, as he hasn't earned a second delegate because he's didn't reach the 33% threshold.)
Confused yet? :-) Yeah, that's how the caucuses will work come January 3rd. Get ready for a wild ride.