Reply #78: Actually, this has been done and it is detectable. [View All]

At any given location with several precincts, the voting pattern can be analyzed. Adjacent precincts have similar voting patterns, unlike the urban/suburban schism. You simply need to value each precinct in a subset with probabilities for how cross-voting results, then sort and analyze.

The discussionin in the the 2004 Ohio Election - Presidential Ballot Orders and Cross-Voting page discusses this. Scroll down to the table with the heading: "2\3 Locations with Crawl 1 and Crawl 4 Ballots." Here the ballots have been separated into the two groups within the subset, precincts with 1.0 probability of Kerry cross-votes switching to Bush, and those with 0.0 probability at the same locations. The difference in non-vote percentage is 2.2% higher in the P = 0.0 group, the Kerry vote is 3% lower in the P = 1.0 group.

There is further discussion above and below the table, including how the vote-switching probability was skewed by non-random, higher numbers of precincts with collocation of Kerry and Bush.