Reply #60: The votes don't get reallocated at random: there's a pattern

[AirAmFan]

that is knowable as soon as you know the entire configuration of ballot orders at a location.

Your hypothetical corresponds to five actual locations in Cuyahoga. One of them is listed at the bottom of page 4 in http://boe.cuyahogacounty.us/boe/PDF/votinglocations.pdf . ST._SAVA_CHURCH_HALL co-located precincts with IDs 0931, 0932, 0933, and 0941 . Look up the precinct results in the big "Canvass Report" at http://boe.cuyahogacounty.us/boe/results :

0931 BROADVIEW_HTS_3-A B 1 283 251 0 1
0932 BROADVIEW_HTS_3-B W 0 456 238 0 0
0933 BROADVIEW_HTS_3-C K 0 238 179 0 0
0941 BROADVIEW_HTS_4-A D 1 263 298 0 1

(The Canvass Report has everything but the initial of the first-line Presidential candidate ("W" for Bush)).

The tiny votes for minor candidates (Badnarik and Peroutka) indicate that "caterpillar crawl" could not have been very important at this site. Each received a total of 2.

But let's plunge ahead anyway. When four precincts share one location, there are 12 ways voters could use the wrong machine: With this configuration, a voter registered in 3A could use a machine set up for 3B, 3C, or 4A, and so on. If all these ways to go wrong are equally likely, then every 12 wrong votes intended for Kerry or for Bush are split equally among the other four candidates (not for all congugurations, but definitely for this one: Do the math). An estimate of the maximum total number of votes lost to caterpillar crawl by Kerry or by Bush at this location thus is 8.