Not Quite a Blog

Warning: This is a very long post.

Introduction

I've spent the past week looking over the voting equipment problems captured by the Election Protection Coalition's 25 nationwide call centers into the Our Vote Live database. There were around 1900 such incidents in the database, although that number is probably closer to 1700 taking into account duplicates.

Before I launch into the analysis, a few caveats:

• This is voter-reported data, which means it can be inaccurate.
• There has been no attempt to control for multiple reports from a single precinct.
• In many cases it is hard to tell what exactly happened as the incident reports were taken by mostly non-technical legal volunteers from mostly non-technical voters and volunteers in the field.
• Given the unbelievable popularity of the 1-866-OUR-VOTE hotline, OVL was unable to capture all incidents that people wanted to report.
• There are undoubtedly incidents that were not reported to the OVL hotline for a variety of reasons.

Bottom-line: This is useful for qualitative notions of what went wrong on election day.

While I've worked to make this post accessible to an audience that may not be familiar with the vagaries of voting technology, I just don't have enough time to explain everything. In that sense, I encourage you to ask questions (joehall@berkeley.edu) and I can amend this document to clarify as needed.

The Big Picture

Somewhere close to 85-90% of all voting equipment incident reports from the OVL database are very simple and report some combination of: broken equipment, long lines, and/or emergency ballots being handed out and/or auxiliary bins on optical scan systems being used (many optical scan systems have a bin incorporated into the design of the machine where ballots can be placed in the event the system ceases to function or the power remains out for hours). If we can do anything to improve the experience of the average voter facing a machine problem, it should be reduce the amount of time they spend in line.

Another curious feature of the data is the voters' uniformly negative attitudes toward contingency or back-up plans. Whenever the primary mode of voting is affected, whatever the contingency plan to keep people voting, it appears that voters are often upset and mistrustful. They seem to think that there is a possibility that their vote will not count if cast via a contingency plan. This is unfortunate as with any critical technical system, there should be a contingency plan in place that covers what to do if that system fails and how the system may continue or recover from that failure. For example, when an optical scanner goes down, the standard contingency plan is to place ballots in an auxiliary bin or container. Voters reported concerns with this kind of accommodation and were worried that their votes might not count. In a number of cases, voters refused to place their ballots into auxiliary bins and waited for hours before the scanner could be serviced or a new replacement scanner brought in. It's clear from this that we need to do a better job of educating voters as to what to do and what to expect in case of emergencies.

Ok, enough of the big picture; now on with the details!

Read more »

[cross-posted to the OurVoteLive blog]

I'm going to try and do today what I didn't have time to do yesterday: some discussion of voting machine and voting technology incidents. Since I was based out of OVL's Simpson Thatcher call center in NYC, I'll first talk about the main states we covered: New Jersey and Michigan.

New Jersey - Early on we were getting tons of reports about broken machines and emergency ballots. In a number of cases poll workers were confused and were offering provisional ballots instead of emergency ballots. In follow-up on a few cases, this was simply poll workers issuing emergency ballots and having voters place them in provisional envelopes without asking the voters to fill out the provisional information (affidavit?). Places we followed up with further here ensured us that they were no longer using the provisional envelopes past the mid-morning. Unfortunately, some of these polling places will undoubtedly run out of emergency ballots by the end of the day. Two other consistent problems with the AVC Advantages are "premature voters" and "fleeing voters". In these cases voters either: 1) report accidentally casting their ballot -- pressing the "VOTE" button -- after making only a few selections on the ballot (casting an incomplete ballot) and 2) report either entering a booth with an uncast ballot on it or realizing later that they neglected to cast their ballot. A few voters reported what might be broken buttons or dead lights on the Advantage... they report having pushed a button next to their candidate of choice and the button not lighting up. It's hard to tell here if the button is broken -- vote not registered or cast -- or the light is broken -- vote cast but not apparent.

Michigan - The big story in Michigan was that optical scanners were breaking down en masse. I have a feeling this is related to the fact that some MI jurisdictions have not been able to perform preventive maintenance on their machines since they were purchased three years ago due to contractual warranty restrictions (see this letter to EAC from the Oakland County, MI Clerk: http://tinyurl.com/64hrmn). Most of the MI reports concerned broken, dysfunctional or nonfunctional optical scanners. Some voters were surprised that they have to cast their ballots into auxiliary bins or ballot boxes instead of the scanner; some refuse to do so and wait, causing very long lines. The other story in MI is a lack or shortage of privacy sleeves. These are essentially large manila folders and it appears that voters have to wait in line with their ballot in a privacy sleeve before casting their vote, thus the shortage given long lines.

More later...

[Cross-posted to OurVoteLive blog]

In my last voting machine incident roundup, I talked briefly about new voting machine incidents from the OurVoteLive call database. Three days later, there are about 100 or so incidents specifically tagged by call center operators as being a possible machine issue.

Here are some incidents that recently caught my eye and why (these are in order of decreasing seriousness, in my opinion):

• Disability access problems. We've seen a distressing number of incidents reported about various problems with accessible voting systems. Two incidents from earlier this month (OVL#2619 and OVL#3559) report an audio ballot that is "reading the wrong names" and that accessible voting machines in Whitefish Bay, Wisconsin are unavailable during early voting, only on election day. Two more recent incidents (OVL#18139 and OVL#19495), report a non-functioning audio ballot and what appears to be a case of "vote-flipping" -- although I don't like using that term -- in an audio ballot interface.

  I know from reports from colleagues with disabilities that disability voting support has been very inconsistent this year. In some cases, the machines aren't set up until a voter with a disability arrives, in other cases machines simply aren't working in a way that a disabled voter could vote privately and independently. In my own work, I was once testing a voting machine and able to inadvertently get the audio ballot into an infinite loop; the technician on-site had to reset the machine. It's unfortunate that accessible voting isn't more usable by people with disabilities after so much R&D and training has gone into its development and deployment.

• Bleeding pens. An incident report from Lake Forrest, Illinois (OVL#15104) reports that the pens used to mark optical scan ballots ballots are bleeding through the paper. Black ballpoint pens are a good bet for any optical scan ballot, but felt-tip or other very wet pens could very well bleed through to the other side of the ballot. If there are only contests on one side of the ballot, there's not a big problem. If there are contests on both side of the ballot, it's conceivable that the bleed-through might change how the optical scan system reads the ballot.

• Capacity issues. Among the multitude of reports of very long lines in early voting, incident OVL#15857 shows just how bad high turnout can get. The caller reports after waiting in a long early voting line, she was confronted with an optical scan machine where "ballots were crammed in and spilling out of the machine." After being told that she was "holding up the line" by a pollworker, she cast her ballot and was unsure if it would be read properly. This is unfortunate and there would come a point where, if a replacement machine hadn't arrived, pollworkers should consider either securely emptying the ballot box to make room for more ballots or temporarily accepting ballots without scanning them in an auxiliary ballot box. Yikes.

• Possible evidence of selective calibration. A caller from Manassas, Virginia reports (OVL#14318) a possible miscalibration of the machine; however, in this case the extent of the miscalibration seems to vary across the screen. In the EVEREST study of voting systems in Ohio (I was part of that team), we pointed out (see pg. 71) that selective miscalibration could be used as a way to frustrate voter intentions. However, Manassas uses the Sequoia AVC Edge, and in the EVEREST report that attack was specific to the ES&S iVotronic which has a more complicated calibration algorithm (involving 20 individual calibration targets instead of a handful). Anyway, this isn't direct evidence that this is happening or that anyone intentionally miscalibrated a machine, it's just the first evidence I've seen of a voter noting a difference in miscalibration across the screen.

• Where's my receipt? A number of incidents report voters being frustrated with not getting a receipt after voting (OVL#4998, OVL#5668, OVL#17022). One voter from Pasco County, Florida, which uses a very new piece of optical scan equipment -- the DS200 -- from ES&S, reports "Machine was designed to dispense receipts and voter saw evidence of past receipts in machine." What is happening here is that 1) the voter mistakenly thinks he should receive a receipt for his cast ballot and 2) the voter mistakenly thinks the receipt printer on the DS200 is for that purpose. Voters frequently think they should receive some sort of receipt for voting. Of course, the "I Voted" sticker is as close as most people get: allowing people to leave the polling place with an official record of how they voted could be used for vote buying or coercion. As for the receipt printer on the DS200, that's actually an event log -- something that records all errors, messages, etc. from the machine and is used to print zero tapes at the beginning of the day and totals tapes at the end of the day.

• Usability/Education issue. A voter from North Carolina (OVL#16759) mistook a message from a voting machine that was asking him to "confirm" that he intended to not make choices in a few contests for it asking him to "confirm" that he wanted to go back and fill in the rest of his ballot. This misunderstanding meant that he cast his ballot before fully filling it out. Voters need to practice working with these kinds of machines, they need to be educated about what to expect when they vote and they need to read the instructions very carefully.

• GOTV-related issue. A voter in Texas (OVL#15441) calls to report that he voted early but is still getting calls from campaigns saying he didn't vote. To support Get Out The Vote (GOTV) campaigns, many jurisdictions publish a list during early voting and on election day of registered voters who have not yet voted. Campaigns can then selectively contact those voters to persuade them to get out and vote. Here, it appears that he is listed as not yet having voted. This could mean that he signed in on the wrong line, or that a pollworker forgot to cross off his name. We can't be sure. There's little chance that his vote was not counted, however.

• Fleeing voter. Similar to the premature voter I pointed out in the last roundup (OVL#10678), we now have a report of a fleeing voter (OVL#13960). This voter from West Virginia reports that he left his machine in the "review mode", or the last page of the ballot that summarizes the voter's selections, instead of casting his ballot. Jurisdictions have different procedures for dealing with "fleeing voters" and I'm uncertain as to what they do in Wheeling West Virginia with their ES&S iVotronic machines. I would hope that the next voter in line would have alerted the pollworkers and that they would have cast the ballot.

For a bit of levity, here are a couple of ones I found a bit funny:

• Lolcats? A voter from Cataret County, North Carolina (OVL#14507) is concerned given how his county has made national news in the past with machine problems. The incident description concludes: "Voter wondering is he can request watchdogs for the polling places[?]". That phrasing is very lolcat-esque and makes me wonder if this isn't a gang of punk rock kittehs waiting to pounce on "watchdogs"... Arf!

• Simpsons? A caller from Florida (OVL#6287) seems like they've watch the leaked Simpsons clip from tomorrow night's halloween special one too many times: the voter claims that the voting machine said, "Thank you for voting for McCain" after he attempted to vote for Barack Obama and cast is ballot.

[cross-posted at the OurVoteLive blog.]

Since I'm an expert in election technology, I'm going to focus my comments here over the coming week about those kinds of issues. I'll do a periodic "Incident Roundup" where I highlight things interesting to an academic from the new incidents showing up in the OurVoteLive database.

Here are some that caught my eye this morning:

• OVL#11385: In a twist to the typical "vote-flipping" complaint, this Texan caller reports that the final screen said that her "vote may be not be valid". This voter likely voted on a Hart eSlate DRE which has an iPod-like "dial selector" and "enter" button instead of the typical touchscreen voting machine. I've never heard of an eSlate displaying a message like that. It could have been an error message from the voting system saying something about storage space, etc.
• OVL#10678: A caller from Illinois, who is also voting on an eSlate, complains that she accidentally cast a blank ballot. This is unfortunate and one of the first cases I've heard of where a voter on an eSlate accidentally cast a blank ballot. If I recall correctly, it should be pretty difficult to do this on an eSlate, with multiple warnings and clear evidence that your ballot is blank. Most of the reports we hear of a voter accidentally casting a vote early occur with older "first generation" DRE voting machines like the Danaher/Guardian ELECTronic 1242 or the Sequoia AVC Advantage.
• OVL#8984: A caller from New Orleans reports a possible case of vote-flipping. I wouldn't usually highlight such an incident as it appears pretty normal. However, the caller makes a good point: if you have a problem while you're voting and the poll workers on-hand aren't much help, what are you supposed to do? Well if you could use a phone in a polling place, you could call 1-866-OUR-VOTE. Alas, phones are not allowed in many polling places, for very good reasons. I'd recommend that the voter request to make a phone call outside of the polling place and come back later to complete voting.
• OVL#8482: A Floridian calls and reports that their optical scanner is broken at that the ballots are being placed in a "black box" and that the poll workers will feed the ballots through the machine later in the day, once it is fixed. While this might seem alarming, especially if it is your ballot at stake, this is a pretty typical back-up plan in case the optical scan system goes down or the power goes out (and the system's battery doesn't last). When the scanner breaks, there may not be much of a chance to fix it unless the problem is relatively simple and there are technicians on-hand that can do the work. Jurisdictions plan for scanner failures by either buying scanners with integrated "emergency" ballot boxes or by providing a ballot box to each precinct that can double as an "emergency" ballot box. In some jurisdictions, these emergency ballots will be cast on that scanner later in the day when it is fixed or they will take the ballots to election headquarters and scan them using a high-capacity (central) optical scanner. So, don't fret if this happens to you. If you're concerned, you can ask the poll workers questions like: When will these ballots be counted? Is there a security seal on the ballot box that would show tamper-evidence if someone were to attempt to mess with these ballots? Are two people (hopefully from different parties) with these ballots at all times until scanned and delivered to election headquarters? (And, remember to be nice to your poll workers, at all costs... they work very very very long hours and are horribly underpaid.)
• OVL#8016: A caller from Georgia reports a troubling but increasingly common practice: poll workers "helping" people vote. Of course, poll workers are there to help; it's their job to help. But selecting choices and casting ballots should never be something that a poll worker engages in unless the voter cannot possibly perform these actions without assistance. Most jurisdictions allow someone to have this level of assistance if they request it, but poll workers should not do these things unrequested. As this incident highlights, it can be very upsetting if someone casts your ballot for you without your permission.