MOVEON MOVED VOTERS. Joel Middleton and Donald Green from Yale have just come out with a paper (PDF) looking at the effect of MoveOn.org's GOTV canvassing operation in 2004. MoveOn organized its canvassing by precinct, so the paper compares turnout between voters living on one side of a street in a precinct that was canvassed and those living on the opposite side of the street in a precinct that was not canvassed.
Overall, canvassing increased turnout by about 7 percent. The effect was constant across states, highly statistically significant, and, because of the study's design, hard to dispute. Apparently, previous literature tended to assume that GOTV was less effective in what they call "high salience" elections, but the 2004 election was about as high salience as it's possible for an election to be.
As effective as MoveOn apparently was, it seems clear that Republican efforts were as successful if not more. That, along with the importance of the election, was why overall turnout increased from 107.4 million votes in 2000 to 123.5 million in 2004. This study gives democratic donors all the evidence they need that making turnout efforts even more extensive in 2008 is both necessary and worthwhile.