Steuben is apparently one of the few counties in the 29th using the new technology. According to the Leader, Chemung County used lever machines. When I voted yesterday in Monroe County, I used the same lever machine technology that's been in use for 50 years.

The Leader reports on the usual screw-ups that accompany any roll-out of new technology. Tallies on the Steuben County website are wrong. A breakdown at one polling station required the use of a plan B that involves paper ballots.

Since turnout on Tuesday was a fraction of what it will be a year from now, I hope Chemung and Monroe aren't going to use that election as the first test of their new machines.

Some highlights: Five of 85 precincts reported incorrectly. One set of poll workers went home without reporting results. The "fact" that one of the candidates received zero votes in his home town was reported without question by poll workers.

New York has been using the same voting technology for generations, and poll workers still screw up. Just imagine the mess if we were trying to implement electronic voting this year.

By 2010, four years before its $65 million touch-screen machines will be paid off, Maryland expects to be back on the paper trail, following states such as Florida and California, which have also decided that all-electronic systems make it too easy to compromise elections.

This is a classic case of throwing good money after bad: once an expensive system is purchased, the pressure to retain and enhance -- rather than replace -- that system is overwhelming. It is critical that New York chooses the right voting technology the first time, because second chances are almost non-existent in government technology procurement.

Part of the the non-profit, non-partisan Election Sciences Institute analysis [pdf] of Cuyahoga's electronic voting machines was a recount of the May, 2006 primary. That recount compared the paper tape generated by the electronic voting machines with the memory cartridges that store vote counts. It found a lot of discrepancies:

• 17% of the tapes showed a discrepancy between 1-5 votes from the electronic tally. 2% of the tapes showed a difference of more than 25 votes.
• 1.4% of the tapes had missing ballots. 40 tapes (10%), were somehow compromised (taped together, illegible, blank, destroyed, etc)
• The chain of custody of the paper rolls was poorly documented: 76% of the paper rolls had incomplete labeling, and 46% of them were blank.
• When a paper tape contained results from multiple precincts, the information on the tape did not clearly identify the precinct from which the vote originated.

A lot of the errors in chain of custody, as well as with the illegible and blank tapes, could be chalked up to poor training of polling place workers. The study recommended substantial changes in the training of those workers. Even so, the study pointed out that the paper tape mechanism is a source of great risk to the validity of the whole voting process. When a tape jams, runs out of ink, or otherwise stops working, the physical record of the vote is lost.

In New York State, lever voting machines sometimes malfunction, and over the years we've accepted the risk of occasional glitch as acceptable given the other benefits of lever machines. But the promise of electronic voting with paper record is the best of both worlds: the convenience of a lever machine with a paper audit trail. The Cuyahoga experience shows that the state-of-the-art audit trail is suspect at best, and requires a level of training and diligence from poll workers well above current expectations.

This last point is critical, because it's easy to mandate more training for poll workers as a cure-all for problems with the voting system. But turning them into election machine repairpeople is just not practical. Poll workers are people with other jobs and interests who work elections a few times per year. They are generally older and less comfortable with technology. Expecting them to set up and verify the function of voting machines, and to repair them when they break, is putting a major new set of tasks on their plate.

In Cuyahoga, the study found that 1/3 of the election workers had trouble setting up the machines, 45% had problems closing out the machines at the end of voting, and 38% had problems repairing the paper tapes. Over half said their training was inadequate, and almost 60% said they needed more hands-on training.

In a scanned ballot system, the role of most poll workers is to make sure that voters get ballots and know how to mark them. The main technology involved is a sharpie. Only a few poll workers per polling place need training in the mechanics of the scanning system, and if the scanner breaks, the integrity of the paper ballots is not compromised. Cuyahoga has shown that electronic voting machines with paper tape are not only inaccurate, they're also a heavy burden for poll workers.

New York is in an enviable position when it comes to procuring voting technology. A number of other states have gone first, and we can learn from their mistakes. Cuyahoga County, Ohio, which includes Cleveland, has been on the bleeding edge, and their experience has made them want to ditch electronic voting machines and turn to scanned ballots. This post will look at one aspect of the Cuyahoga experience: the economics and scalability of machines versus ballots. The next post will take a closer look at other election issues raised by a 234-page study [pdf] of Cuyahoga's 2006 primary. That report, which was issued by an independent, non-partisan testing body, raises serious doubts about electronic voting machines, including those with a paper trail.

The first simple fact about voting machines is that they are expensive. New Yorkers for Verified Voting (NYVV), a non-partisan, non-profit organization, has conducted an in-depth, county-by-county comparison [pdf] of the cost of voting machines versus scanned paper ballots. Overall, electronic machines would cost $116 million more than a scan system. In Monroe County alone, the difference is over $6 million.

One might think that voting machine costs are front-loaded, since the scan system requires the printing of paper ballots for every election. The NYVV report shows that this probably isn't the case. Touchscreen voting machines, which take a beating from voters, aren't warranted for more than five years, whereas scanners last much longer (14 years in some cases). This faster replacement cycle might make up for the higher recurring cost of paper ballots. In addition, the NYVV estimates do not include the cost of paper and toner need for the paper trail produced by electronic machines, which adds a significant recurring cost to the electronic option.

Since electronic machines cost $8,000 each, there won't be a lot of extras hanging around. Yet, in the Cuyahoga study, almost 4% of the voting machines suffered a failure and had to be shut down on election day. A failed machine might mean lines in the polling place. The study also found another potential source of lines: ballot length. Because this length varies by jurisdiction, the study recommended that Cuyahoga buy more machines and create a deployment strategy to position machines to better handle voter demand.

In a paper-ballot system, voting booths cost roughly $250. It's much cheaper to have a few extra hanging around to handle higher demand, and voting booths won't have a 4% failure rate. No fancy studies are required to allocate resources when the resource is cheap and reliable.

In the paper system, each polling place will usually have one scanner. If that scanner fails, the polling place can still open (or stay open) because ballots can be voted without being immediately scanned. Unlike electronic voting machines, the easiest-to-break part of the scanned ballot system is not critical to the operation of the polls.

Cuyahoga might still keep their electronic voting system. After all, it's hard to throw away a $17 million system and start over from scratch. New York is lucky that we can learn from their mistake and choose scanned paper ballots.