Problems in estimating turnout
If you compare the figures shown in the Census Bureau spreadsheets in the post below and the figures from the Alabama Secretary of State, you will notice that the Census has 9 to 17% more people who say they voted than did vote for president in Alabama. I think it is highly unlikely that many people went to the polls to vote in 2000 and 2004 and found the race for president so uninteresting that they did not bother to vote, but did vote on a constitutional amendment, a county commission, a judge, or a member of the Public Service Commission. Don't you?
The problem is caused by the methodology of the Census. It asks people if they voted; no double check of actual election results is done. Social scientists know that many people misrepresent whether they voted when asked in surveys. They give the answer expected.