Greg Palast has an article at TomPaine.com arguing that Kerry really did win the election. According to him:
Although the exit polls show that most voters in Ohio punched cards for Kerry-Edwards, thousands of these votes were simply not recorded. This was predictable and it was predicted.
Once again, at the heart of the Ohio uncounted vote game are, I'm sorry to report, hanging chads and pregnant chads, plus some other ballot tricks old and new.
The election in Ohio was not decided by the voters but by something called "spoilage." Typically in the United States, about 3 percent of the vote is voided, just thrown away, not recorded. When the bobble-head boobs on the tube tell you Ohio or any state was won by 51 percent to 49 percent, don't you believe it ... it has never happened in the United States, because the total never reaches a neat 100 percent. The television totals simply subtract out the spoiled vote.
And not all vote spoil equally. Most of those votes, say every official report, come from African American and minority precincts.
Palast has done some good work on spoiled ballots in the 2000 election (based in part on my analysis of the issue), but I don't think the facts support him here. First, there are just over 100,000 residual or spoiled ballots (residual votes=total number of voters who went to the polls minus the total number of votes cast for president). Bush won Ohio by nearly 140,000 votes. Even if every residual ballot had gone to Kerry, a highly doubtful possibility, it still wouldn't have eliminated Bush's margin. Furthermore, there's no statistically significant relationship between the spoilage rate and the race of a county. Finally, the spoilage rate actually went down the better Kerry did in a county. Like it or not, Bush won Ohio.