In Cambridge, we use an elaborate electoral system of proportional representation for electing the city council and school committee. Voters get to specify their preference for each race (first choice, second choice, etc) for 9 at-large city councillors and 6 school committee members. The counting system then processes the ballots as follows. To be elected, a candidate needs to get 1 more than 1/10 or 1/7 of the ballots. As a result some candidates get more than are needed and every n'th ballot is transferred to the 2nd place candidate. Then they start from the bottom distributing the ballots to the next position until all 9 or 6 are elected. This system is pretty good - you don't need to worry about wasting your vote for an unpopular candidate (such as voting for Nader instead of Nader 1), Gore 2) etc.). The ballots used to be counted manually and it took a week for the ~20,000 ballots. I was on an Elections advisory committee (still am) that helped computerize this a few years ago. The computerized system involves ballots on which voters fill in ovals to indicate choices, scanning each ballot at the polling place and then processing the results from the scanning at a later time in a central place.
In the 2001 election the school committee election was very close (in the final round the number 5 person had 2220 ballots, the number 6 had 2219, and the number 7 had 2213 and therefore lost.) A hand recount was requested as a result. The problem is that this distribution of the ballots is order dependent so the paper ballots had to be put in the same order as they had been processed. When they drop into the scanner bin they aren't necessarily in the same order as they had been scanned and recorded. What they did then was to create facimiles of the ballots from the computer record which were ordered correctly and then match them up with the original ballots so that the recount could proceed. Matching was done in the public with candidate's representatives challenging some of the ballots. Sloppily marked ballots don't always get scanned correctly. This process took 7 days. The recount took 4 days. (The end result elected the same people with slightly different numbers.)
Yesterday's meeting was for a new procedure which simplified the ordering by scanning the original ballots a second time but preserving the ballot order the second time and then doing a computer match of the two scanned results. This leaves a few unmatched ballots that have to be done manually (In a test of 919 ballots there were 35 that had to be matched manually.) This meeting approved this new process.
The state election law doesn't allow the obvious fixes such as the scanner printing a sequence number while scanning or having prenumbered ballots because of its privacy provisions. It also doesn't allow the scanner to reject overvotes or undervotes and allow the voter to fix his or her ballot. Undervotes are often caused by sloppy ballot marking such as drawing a circle around the oval instead of filling it in. If a ballot has two number 1s they cancel out each other and the number 2 is treated as the number 1. These overvotes are sometimes caused by a slip of the pen or by someone x'ing out an error. Interestingly enough, the new federal law will require rejecting overvotes and undervotes to allow the voter to fix the ballot.
The second meeting was the Mid Cambridge Neighborhood Association annual meeting. Unfortunately extremely bad presentation and organization caused the meeting to take 1 1/2 hours to discuss some by-laws changes. It is amazing how people can worry about the consequences of what seemed to be simple changes. I'm sorry I bothered to go since the topic of substance which was about a neighborhood community afterhours school that was closing hadn't come up by the time I got too bored to stay any longer.
Tonight's meeting is about the public library where we (I am a member of the Design Advisory Commitee) see the revised plan.