I rather assume the idea is to get a picture of what proportion of Americans now say they voted for George Buish, and compare it with the actual figures from the election.

This kind of exercise in other cases - for example Margaret Thatcher - has shown that once a politician gets unpopular, very large numbers of previous supporters prefer to say they never voted for them, and very probably in many cases actually believe it. Memory can play strange stricks.

I think it is pretty safe to predict that this would be the case with George Bush.