What you have to keep in mind are the kind of errors which occur when you have such small groups in a poll. Occasionally people give messy answers. I know this b/c I know people who work with polls. Some institutes use double questions, and can then see that some people might first state they are males and then that they are females or something similar, and then they are leaved out. It seems unlikely that Gore voters would have gone for Bush without Nader it doesn't make much sense to me. Is there a logical explanation?
Good point, in a normal poll the margin of error is often +/- 3% or more, in part because of the small sample size. But also in part because some in the poll who say they intend to vote don't. However, in an exit poll 100% are actual voters, and MUCH larger numbers of voters are interviewed. For 2000 over 13,000 nationwide were asked who they would vote for in a 2 way race. The answer was the same as FL, Bush 49%, Gore 48%, 2% non-voters, again because 2% of Gore's voters woudl switch to Bush without Nader running!!