During the last (pre-hiatus) discussion group meeting, Paul [Gurian] brought up the idea that in typical (Super Tuesday era) nomination campaigns, caucus winners usually gain support as the number of steps in the process increase. There are a few issues at work here:
1) Some (or all) of the candidates, who won support initially, have dropped out of the race. By the time the second step rolls around, those free agents move over to the leader (or presumptive nominee).
2) If the nomination is still in doubt as subsequent caucus steps begin, the first step winner's supporters use that original plurality/majority to tweak the numbers for the next step even further in their direction. In a competitive environment then, this majority builds a modest level of momentum throughout the process, giving the original winner a delegate total greater than the projection following the first step.
As best Paul and I can ascertain, there really isn't any literature addressing this question directly, only anecdotal evidence from campaigns past. In the Super Tuesday era though, this hasn't been an issue because the field had been winnowed significantly ahead of any second step caucus meetings. This anecdotal evidence then would come from the elections after reform but prior to 1984 or 1988. In other words, the Carter vs. the field in 1976, Carter/Kennedy and Mondale/Hart match ups. Carter was a steady force throughout the 1976 nomination phase. After doing well early in 1980, Carter's support faded down the stretch as Kennedy made gains. Mondale, after having lost in early caucus rounds to Hart in 1984, gained delegates in subsequent rounds to take a significant pledged delegate lead into that year's convention.
There are two hypotheses that emerges from this:
1) The winner of the first step gains delegates in subsequent steps.
2) The frontrunner (but not necessarily the winner) gains delegates as to process progresses.
Both require controlling for either the level of competition or the amount of candidate winnowing that has taken place. And in the frontloaded era, that winnowing has been rapid enough that the second steps fell after the point at which a presumptive nominee had been determined.
[The question then becomes one of data collection and this is the tricky part, simply because the transparency of the caucus process is less than that of a primary election. The reporting just isn't the same.]
The 2008 campaign though, fits the pre-Super Tuesday era model in that the competition has extended beyond the massive, early clustering of state contests. Given Obama's success in the caucuses then, it stands to reason that he would gain even more ground in the delegate count over Clinton as the next steps are held. However, there may actually be some evidence to the contrary: that Clinton has made some gains in the lead up to the second step caucuses. The Monkey Cage (via Enik Rising) has shown that in Colorado, the post-caucus numbers have fluctuated some in the time after the precinct caucuses were held on February 5. This whole thing is speculative, but it is a means through which the Clinton campaign could make strides in the delegate disparity between the two Democratic candidates. In one state, flipping a delegate or two won't make that much of a difference in the grand scheme of things, but if this is happening across all the caucus states, then those changes could become significant.
How then does this fit in with the research question posed above? Well, it adds another layer to consider. Things get more complicated as a factor like specific candidate strategy to protect or steal delegates as the process progresses. Much of this would depend again on the competitiveness of the race and how each candidate is positioned in relation to the other. That we have witnessed a virtual tie in the 2008 Democratic race is something of an anomaly compared to races past (even competitive nomination races). And that is where the extra layer--the possibility of delegate shifting--originates. If it is perceived that the options are still open, then delegates are as a result more likely to entertain the idea of shifting. Whereas, if the race was signaling the emergence of one candidate over the other (no matter how small that lead), delegates would be less likely to move.
This is an interesting question that even just a case study of how the 2008 race (in caucuses) would provide some enlightening answers.