Still...did the wrong guy win? Harper's revisits the question.
The National Election Data Archive seems to think it was likely:
The discrepancies between election results and exit poll results in the 2004 presidential race have not been explained and are consistent with significant vote count errors. Examination of limited election data shows irregular patterns of vote counts in Florida, Ohio, Washington, New Mexico, and other states....Thanks to Thomas Leavitt at Seeing the Forest for the first link.
On May 15th, our National Election Data Archive Project released a scientific paper, updated September 8th, that disproves the theory that the 2004 presidential exit poll discrepancies were caused by exit poll response rates that varied by party. This discredited hypothesis was proposed by the exit pollsters and used to dismiss the exit polls. This paper follows our earlier scientific paper released on March 31st, and summary on the 2004 election exit poll discrepancies.
It has been officially confirmed (by the exit pollsters themselves) that on election night the final set of exit polls showed John Kerry defeating George Bush by 3% of the popular vote and a clear majority of 316 electoral votes. Our statisticians analyzed Edison/Mitofsky's own explanation of their exit poll discrepancies, and found serious flaws in their argument. Exit polls have been used for years to detect corruption of official vote tallies - most recently in Ukraine.