Stanford CIS

How I spent my Election Day

By Stanford Center for Internet and Society on

I spent fifteen hours on Election Day in a conference room in a Manhattan law firm, surrounded by dozens of attorneys, fielding questions from what felt like all of them. It wasn't the callback interview from hell, though: I'd volunteered to help work the Election Protection Coalition's 866-OUR-VOTE callin hotline.

The hotline has existed for some years, but after 2000, it became much more well-publicized and well-funded. Its goal is to allow poll watchers, voters, and election workers to call in with questions, comments, and especially their observations of anything that looks like irregularities in the electoral process. The line connects callers to banks of volunteer attorneys and law students at call centers all over the country. Even though there was a call center in San Francisco, I'd been tapped to fill a last-minute hole in a phone bank responsible for handling calls from Pennsylvania.

(much more below the fold...)I started with a warmup day on Monday. The calls came in slowly, mostly from people who wanted information -  whether they could vote, where they should go, what sorts of documentation they'd need to bring to the polls. For the most part, they were easy questions with straightforward answers.

The hard part started right away in the morning on Tuesday, right after the Pennsylvania polls opened at 7 AM. We started getting rapid-fire calls from voters in Philadelphia complaining that their poll sites were closed or backed up with long lines because of malfunctioning voting machines. People were leaving without voting. My first preconception about electronic voting was smashed in the first hour of voting on Tuesday. I'd thought that since Philly's machines were an older model of direct-recording electronic machine, built on dedicated hardware and already used for some years, the reliability problems would have had time to shake out. Older also means more prone to failure, though, especially after having been kept in cold storage for months between elections.

The complaints slowed later in the morning, indicating that replacement machines were being deployed - but hours of precious voting time had already been lost. The polling places didn't have an adequate backup plan for failing over to paper ballots if the machines failed. Most had only a few dozen "provisional" ballots on hand - intended for use by voters whose registration status was in question - and polling places were inconsistent as to whether they permitted regular voters to vote provisionally while the machines were down. A fully registered voter casting a provisional ballot - which has to go through an extra stage of review before the vote counts - is a bad idea, since it leaves open the possibility that the vote won't count at all. Those polling places that did permit provisional voting as a backup scheme quickly ran out of paper ballots and had to halt voting entirely until their replacement machines arrived.

The lesson from the morning of November 2nd in Philadelphia: you've got to have a reasonable backup plan, and that plan has to be consistent across every polling place. Equipment failure is inevitable and unpredictable, and hours-long lines while the machines are being replaced are unacceptable. How hard would it be, really, to keep a stack of paper ballots handy? Counting a few ballots by hand is not nearly as bad as turning away voters.

The really big equipment failures came in Mercer County, a rural area in the far western part of the state. Several - perhaps most or even all, it's not clear - of their precincts were voting on paper following the apparent failure of their Unilect Patriot touchscreen DREs. The actual cause of the problem isn't well-understood, but it may have been widespread voter error. The Patriot system's interface is set up so that, to change a vote, you first touch the presently selected choice to deselect it, then choose again. The problem comes when you first choose a straight-party ballot on the first screen, causing every candidate of your chosen party to be selected on subsequent screens, then go through and manually touch every candidate. The result is that you've manually voided every vote you initially cast by selecting a straight-party ballot, which means that the machine will record no votes at all. That's consistent with the reports we were getting of vote summary screens indicating no votes at all. By the end of the day, the Mercer County folks had figured this out and were cautioning voters not to accidentally clear out their choices after voting straight-party.

My experience Tuesday changed my beliefs about the best way to vote. I used to think that an electronic voting machine with a paper trail was the right answer - the best of both worlds in terms of accessibility and verifiability. The big fear about paperless e-voting is that the system will fail in a way that causes votes to be subtly miscounted. That remains a serious concern. But I didn't consider that complicated computer systems, whether auditable or not, are much more likely to fail in a way that prevents votes from being cast at all. A computer-based paper trail system certainly meets the requirement of producing a voter-verified paper trail - but so does a simple optical-scan paper ballot filled out by hand. Op-scan ballots don't crash, aren't susceptible to power failures, and they're largely idiot-proof on both the pollworker and voter side. For disabled voters who need computer-mediated systems in order to vote unassisted and in private, the right answer is a ballot marking system. These devices have touchscreens, audio interfaces for the blind, sip-puff tubes for those without full use of their hands, and various other features making them usable by a wide range of disabled voters - but instead of recording votes to memory, they print out a machine-readable optical-scan ballot, which can be read right along ones marked by hand with a pen. Every pollsite needs only one of these; nearly every voter won't need it. If there's enough voting booths, then there need not be long lines. If there's a computer failure, it will not close down the entire polling place.

The Election Protection Coalition has its entire database of election issues available online, searchable and sorted by state, county, and type of problem. You can see what calls we got about your state. (For the record, Pennsylvania had the most incidents reported of any state at 2200. When the folks in charge announced that, just after close of polls, a ragged, ironic cheer went up from the hotline volunteers.)

If you've come this far, and you're still awake, you might also want to check out the E-Voting Experts blog, run by a group of computer security experts interested in electronic voting, including my undergrad advisor Dan Wallach.

Published in: Blog