Reply #124: This is certainly a difficult problem because there are no overt clues in

[L. Coyote]

the vote record, no non-votes, no 3rd party votes. Some deductions might be possible from vote results, given adjacent precincts tend to vote the same. But this will not be valid with small sample sizes. An overarching trend would be convincing of something being wrong.

Even with the assumption that crosss-voting one-way can be used to quantify cross-voting other ways, you still need to use a probability index to determine the true multiplier. If in 2 ballot/2 precinct locations the number of crawl 1 and crawl 4 ballot collocations is greater that the 2 and 3s, then the probabilities of major candidate switches increases, and the evidentiary trail of non- and 3rd-party-votes is proportionally diminished. This is the case in Cuyahoga.

Another factor is the actual demographic inequality of the collocated precincts. The number of potential cross-voters in each precinct is a factor in equations assuming equal cross-voting. How many are in precincts with a 1.0 probability of a Kerry to Bush switch, how many are in the precincts with a Kerry to disqualified switch probability. Only after this is known can an appropriate multiplier be arrived at for the location (or subset) in question.

In locations with 2 ballot orders and 3 precincts (2\3 in notation form), the probability matrix is more complex and a different formula prevails. At some of these locations, from one precinct you have 1.0 probability of a Kerry to Bush switch, at others you have a 0.50 probability because of KKB collocation instead of KBB. With KKB Kerry voters can cross-vote for their own candidate, with KBB they cannot.

And it just gets more complicated with 4\4, 2\4, 3\4, etc.