http://www.pbs.org/newshour/extra/features/elections/electionnight.html
Problems of Early Calls
This leads to the second, more serious problem afflicting the polls: the early call of Florida for Gore made between 7:49 and 8:10 P.M.EST on election night by all major networks (led by NBC) and the Associated Press based on the VNS exit poll.At the time, with other major battleground states such as NewYork,Michigan,and Illinois falling into the Democratic camp, it appeared to commentators
and observers that Gore was heading for a comfortable victory. There were a number of potential problems about this particular call. First, the result in
Florida was announced on the networks before polling stations had actually closed in all counties in the state,namely, the panhandle northwest, in the central
time zone, that tended to favor Bush. It is possible that this announcement could have discouraged some last-minute voters from casting their ballots, if they
believed that the outcomewas already cut and dried. This probably affected only a few Floridians, although, given the closeness of the eventual outcome, every vote counted. The second issue—calling the results from states on the East Coast before the polls have closed in theWest—is nothing new to the presidential election. The most famous example is President Jimmy Carter’s concession to Ronald Reagan in 1980 prior to the close of polls in California, thereby, it is argued, depressing electoral support for some Democratic House members in that state. But Republicans believe that the early call in Florida discouraged Californian Republicans from voting, for example, on their way home from work. Yet, studies remain divided about the effects of early network calls.On one hand,some argue that there is a modest “bandwagon” effect, with undecided voters
moving toward the candidate ahead in the polls.On the other hand,some suggest an “underdog” effect, with polls slightly strengthening the motivation and turnout
of those supporting the losing candidate. Given the lack of convincing evidence either way, a case can be made that both possible effects may cancel each other out.
(Taken from http://www.ksg.harvard.edu/presspol/news_events/Too%20Close%20to%20Call.pdf)