The vast majority of dead people voting come from reports of states comparing voting records to death records. In one example South Carolina reported that 953 people voted after being deceased between 2005 and 2012.
However when they investigated 207 cases from the 2010 election they found that '106 cases were the result of clerical errors by poll managers; 56 cases were the result of bad data matching, meaning that the person in question was not actually dead; 32 cases were “voter participation errors,” including stray marks on lists erroneously indicting they had voted; three cases were absentee ballots issued to registered voters who cast ballots and later died before Election Day; and 10 cases contained “insufficient information in the record to make a determination.”'
Or they don’t believe your estimates of low levels of voter fraud.