Paper Trails, in Theory and Practice
This week's electionline.org weekly features a rundown on congressional paper trail legislation. There are now several bills pending in Congress pertaining to voting technology. Among them are proposals by Rep. Rush Holt and Sen. Hilary Clinton to require electronic voting machines to have a contemporaneous paper record ("CPR," more commonly referred to as the "voter verified paper audit trail").

This legislation is still being pushed, despite the fact that even some of the strongest and most thoughtful critics of electronic voting have raised serious questions about the workability and efficacy of this method of verification, as I noted here. Moreover, as I noted in the same post, the enactment of a CPR requirement has been causing serious headaches for the State of Ohio -- and may end up causing the state to miss the 2006 deadline for replacing the punch card voting equipment used by over 70% of voters in the state.

Advocates of a CPR/VVPAT requirement have cited Nevada's experiment with a Sequoia voting machine with an attached printer. Such experimentation ought to be encouraged, but what we know about Nevada doesn't do much to support the case to mandate a CPR nationwide. The Los Angeles County Registrar-Recorder has produced this video demonstrating some of the problems that became evident during the implementation of this system in Clark County, Nevada. The video definitely has a discernable anti-CPR point of view, but raises some serious question, including 1) whether voters will actually check the paper record, 2) paper jams, 3) possible invasions of voter privacy, 4) the high cost of equipment and storage, 5) the enormous person-hours required to recount the paper records, and 6) whether or not the CPR will actually provide an effective check on fraud and error, unless all (or at least some substantial sample) of the paper records are actually recounted. The Holt and Clinton bills mandate a 2% manual recount, but this analysis by Andrew Neff suggests this won't provide much confidence in the results -- and I've yet to see any I've yet to see contrary analysis by those supporting the 2% solution. Isn't this something we ought to figure out before we legislate?

The most carefully considered version of the bills in Congress appears to be the ones proposed by Sen. Chris Dodd (S. 17) and John Conyers (H.R. 533). These would require verifiable audit records, without requiring that those records be paper. This would avoid the problem of locking in a particular technological fix -- and one that's likely to prove very expensive -- without knowing for sure whether it will actually work.

Update: Prof. David Dill of Stanford University, writes in response to the above post:

While, of course, printer breakdowns are always possible, so are failures with other kinds of equipment, including DREs (and we see them -- frequently). The critical question with contemporaneous paper records is how often problems would occur. This depends on machine design, procedures, and poll worker training. I believe it can be done well, but I don't know how current products have performed. I know there were some problems in Nevada this year, but there don't seem to have been as many as in some of the places that used DREs (based on the EIRS reports we collected).

As I have said, DREs with voter verifiable printers are an expensive and relatively risky way to have a paper trail, and precinct-count optical scan is to be much preferred. Hence, I applaud Secretary Blackwell's decision to require the use of optical scan systems (although I don't understand why he limited the vendors so severely, since almost everyone makes these machines).

My thanks to Prof. Dill for taking the time to express his views. He was indeed who I had in mind above, in referring to thoughtful critics of electronic voting, and I hope that I've not left a misimpression as to his opinions.

As for the substance of his views, I share his concern about how well the present generation of DREs with a contemporaneous paper record function. Where we may differ is in our optimism on how well such a system can be implemented -- both in terms of its efficacy in promoting integrity and its workability in real-world election environments. At least until we figure this out, it's a bad idea to enact legislation, such as that which Sen. Clinton and Rep. Holt have proposed, that would require all DREs to have this unproven device.

Prof. Dill's response is that counties can move to precinct-count optical scan. Although that may work well for some jurisdictions, particularly those without substantial non-English speaking populations, I'm not sure that it's the best answer for larger, urban counties . . . but more on that later. - posted by Dan Tokaji @ 6:28 PM