VOTE COUNTING METHODS: Three main vote counting methods are used by the state of New Hampshire as reported on the Secretary of State's website as of January, 2004: ACCUVOTE, OPTECH, and "PBHC" - Paper Ballots, Hand Counted.
PBHC are used in approximately 50% (149) of all precincts, but the smaller size of the precincts means this method is responsible for only 21% (142,234) of the overall votes counted. PBHC account for 27% (19) of the "out of trend" precincts, but comprise only 8% (20,192) of potentially affected votes.
OPTECH is used in approximately 8% (25) of all precincts, and accounts for 12% (82,094) of all votes counted. The 8 "out of trend" precincts are 11% of the total affected precincts, and only 14% (35,839) of the potentially affected votes. There should be some concern for the equipment in Category E, which has a 42% "out of trend" precinct rate. The 50% ranking for both Categories F and G may simply be a statistical anomaly due to the small sample size.
ACCUVOTE is currently used only 42% of the time (in 126 precincts), but accounts for 67% or 453,235 of all votes counted. This method has the highest rate of "out of trend" precincts: 44 precincts, or 62% of the total "out of trend" precincts use this method of vote counting; of more concern is the fact that, due to the larger size of the precincts involved, this represents 78% of all votes in affected "out of trend" precincts, for a whopping 200,269 potentially miscalculated votes -- nearly 30% of all votes cast in the 2004 presidential election in the state of New Hampshire.
http://www.invisibleida.com/New_Hampshire.htm All are therefore paper ballots.
She concluded that the biggest anomalies were in large precincts using Accuvote. It was the precincts with the largest anomalies that were recounted.
One thing worth noting - the largest precincts in NH are HUGE - way bigger than precincts elsewhere. We know that the exit poll redshift tended to be larger where the interviewing interval was high (e.g. every 10th voter), and interviewing intervals were high in large precincts. So it is at least possible that the high WPE in NH was due to departures from random sampling in large precincts.
Just sayin'