William Saletan and Matt Schiller, writing in Slate (Jan. 16, 2004):

More than 100,000 Democrats will go to precinct caucuses to select a nominee for president. Which candidate will get the most votes that night? If the race remains close, you'll never know....

Every viable group gets at least one delegate. The bigger your group, the more delegates you can earn. But there are two catches. First, the number of delegates to be distributed in the room depends on how many Democrats voted in your precinct in the most recent gubernatorial and presidential elections. If you're new in town, and the turnout in your precinct was lousy four years ago, your vote effectively counts less than it would have if you'd moved to a high-turnout precinct. Second, if your group is bigger than another group in the room, that doesn't guarantee you'll get more delegates. Let's say the chair has six delegates to distribute, and there are four viable groups. That leaves two extra delegates, which will probably go to the two biggest groups. If you're in the third-biggest group, and you've got more people than the fourth group does, tough luck. You each get a delegate, and that's that...

On caucus night, the Iowa Democratic Party will release the delegate count. Here's when the party will release the raw vote count and the realigned vote count: Never. The party won't compile or even record them, except as a temporary step in most precincts so that the caucus chair can determine how many delegates each candidate gets. The party doesn't want raw votes compiled and released, because it wants the caucuses to be a collaborative activity, not a tally of individual preferences. That's all well and good, if you like the party's communitarian version of democracy. But if you want to know how many voters stood up for John Edwards, you're out of luck.

This wasn't a problem four years ago, because Al Gore thumped Bill Bradley in the Iowa delegate count by a 2-1 margin. It was a two-man race, and Gore had clearly won. In 1996, Bill Clinton faced no real opposition. In 1992, Sen. Tom Harkin, D-Iowa, was such a prohibitive favorite that other Democrats skipped the caucuses. To create a dangerously high risk that the winner of the delegate count isn't the winner of the raw vote, you need two things: a big field, so that there will be plenty of nonviable groups to redistribute at the precinct caucuses, and a close race.

You have to go all the way back to 1988 to find an Iowa Democratic presidential contest in which both of those factors applied. What happened that year? Gephardt beat Sen. Paul Simon, D-Ill., by four percentage points in the state delegate count. Did Gephardt win the raw vote? A media consortium called the News Election Service said he did, by three percentage points. But four months later, an article revealed that the NES had reported vote counts from only 70 percent of the caucuses and had botched so many of those that its numbers couldn't be trusted. A separate caucus-night projection by NBC News, aborted and never disclosed on air, had Gephardt leading by only half a percentage point. Given the closeness of the race, there was no way to know whether, when Iowans stood up to be counted, Gephardt got the most votes.