Allan Lichtman posts on spoiled ballots in Florida 2000. (Hat tip: Ralph Luker). He rightfully refers to Lott's claims about ballot spoiling as "bizarre". Lott claims:
African-American Republicans who voted were 54 to 66 times more likely than the average African American to cast a non-voted ballot (either by not marking that race or voting for too many candidates). To put it another way: For every two additional black Republicans in the average precinct, there was one additional non-voted ballot. By comparison, it took an additional 125 African Americans (of any party affiliation) in the average precinct to produce the same result.
So 50% of black Republican's ballots were rejected? How is that even possible? Can anyone even think of a mechanism? This is an obviously spurious correlation. Any normal person would decide that this meant that there was something wrong with their statisitical model, bu not Lott---he goes ahead and publishes. His nonsense was actually published in The Journal of Legal Studies, indicating some serious deficiencies in the reviewing process at that journal.
Lott's numbers don't even add up. He states that 5% of blacks are Republicans. If 50% of their votes were rejected, that means that 50% of 5% or 2.5% of black votes were rejected even if not one black Democrat ballot was rejected. But Lott claims that 1 out of 125 (less than 1%) of black votes were rejected. I guess the rejection rate for black Democrat votes must have been negative.