Andrew Appel on New Hampshire’s Election Audit

Really interesting two part analysis of the audit conducted after the 2020 election in Windham, New Hampshire.

Based on preliminary reports published by the team of experts that New Hampshire engaged to examine an election discrepancy, it appears that a buildup of dust in the read heads of optical-scan voting machines (possibly over several years of use) can cause paper-fold lines in absentee ballots to be interpreted as votes… New Hampshire (and other states) may need to maintain the accuracy of their optical-scan voting machines by paying attention to three issues:

Routine risk-limiting audits to detect inaccuracies if/when they occur.
Clean the dust out of optical-scan read heads regularly; pay attention to the calibration of the optical-scan machines.
Make sure that the machines that automatically fold absentee ballots (before mailing them to voters) don’t put the fold lines over vote-target ovals. (Same for election workers who fold ballots by hand.)

Tags: auditing, scanners, voting

Posted on June 15, 2021 at 10:45 AM • 23 Comments

Comments

Tatütata • June 15, 2021 11:00 AM

Make sure that the machines that automatically fold absentee ballots (before mailing them to voters) don’t put the fold lines over vote-target ovals.

“Do not fold, spindle or mutilate” in the 21st century…

tim • June 15, 2021 11:38 AM

“Do not fold, spindle or mutilate” in the 21st century…

The article says nothing of the sort.

David Rudling • June 15, 2021 11:58 AM

@tim

“Do not fold, spindle or mutilate”

You were too young to work with computer punch cards in the mid 20th century then. I seem to recall they had this stamped on them.

JonKnowsNothing • June 15, 2021 12:41 PM

@All

Not only do you have to be old enough to remember

“Do not fold, spindle or mutilate”

You also have to have been around long enough to know what a spindle is…

Tatütata • June 15, 2021 12:43 PM

Next time I will exude gratuitous silliness I shall endeavor to include a tutorial.

Clive Robinson • June 15, 2021 1:04 PM

@ Tatütata,

Next time I will exude gratuitous silliness I shall endeavor to include a tutorial.

I am reminded of the apocryphal story about mailing old style 8×6 and similar photos.

You had special envelopes where one side was stiff card board, on the brownbpaper cover side printed boldly in RED capitals maybe a third of an inch high,

PHOTOGRAPS DO NOT BEND

Someone got an envelope that had been brutaly folded and scribled unde the warning,

“Oh yes they do”.

If true or not does not realy matter it makes a statment about basic human nature when dealing with “Official Tasks” effectively “anonymously”… A lesson every one should be wise to.

Oh and if you want “sillyness” the Devils DP Dictionary has an acronym for a tecunology it calls,

Square Holes In Tape

Having worked with and actually edited programs and secure encrypted data transmissions on punch paper tape up untill the end of the 1980’s the acronym for the technology was not lost on me.

MarkH • June 15, 2021 2:04 PM

@Bruce,

Thanks for bringing this to our attention!

I suggest an omission, in the recommendations for avoiding such problems in the future: ballots can be designed with a blank region (perhaps 5 or 6 mm in extent) to either side of each “score line” intended to guide folding.

Many kinds of voting problems have been caused by poor ballot design; widespread adherence to good design practices could be as valuable as any vote-counting safeguards.

My hot takes:

• The most important recommendation is routine application of risk-limiting audits. RLAs are designed to be economical, and will with high probability detect systematic problems (whether accidental or malevolent) large enough to affect an election outcome. There’s no good reason not to do them.

• New Hampshire went through an exemplary process to investigate the discrepancies, with appropriate expertise and no political shenanigans.

• The article comments are extremely knowledgeable and informative, apparently written by people involved in the election process (with the exception, so far, of one conspiracy wingnut).

• From one of the commenters, in New Hampshire one political party (guess which) has consistently blocked legislation to bring state election practices into line with federal law. Had federal standards been followed, most of the miscounted ballots would probably have been returned by the scanners for correction.

Weather • June 15, 2021 5:04 PM

@bruce
No keep it anonymous, my brain can’t deal with it, but supposed you could do it ,I can explain what you should know. But cert etc should be able to place it.

echo • June 15, 2021 6:38 PM

It would be interesting to read a list of what election issues there are from neuro-psychology to marketing to media regulation to law to voting methods to economic models to carefully disguised use of public funds as bribes to gerrymandering. Each level (many of which interact with each other) is an opportunity to effect the result before the voter votes.

The American obsession with technology and crunching numbers has its value but there is a downside too. Has anyone yet determined whether it was communist dust or capitalist dust?

Sorry but I don’t trust these things.

benjamin shropshire • June 15, 2021 7:01 PM

That first recommendation remains me of a comment I heard pointing out that one of that major problems with the voting system is that the only people with standing to question anything are the same people who have a political ax to grind over it.

The suggestion there, if I remember correctly, was things along the lines of pull precincts out out of a hat every election and recount them, even if nobody thinks something is amiss.

Larry Sanderson • June 15, 2021 7:25 PM

It’s important to do random statistically significant hand audits of any optical machine processing for discrepancies, which can be done after the election and before it’s certified.
Most/many/some election offices only use high-speed central tabulators to count mailed-in ballots. For example, all of the Minneapolis mailed-in and overseas ballots are counted by Hennepin county’s tabulator.
All of the tabulators, including the high-speed ones, go through testing before elections, and none of the polling place tabulators should be counting folded ballots.

Jon • June 15, 2021 7:36 PM

Christ on a cracker. Democracy hangs on reader heads full of cruft, and the way a page is folded?

The times we live in…

As for the folding, it is a matter of graphic design, to anticipate how a page may be folded, and to move essential data fields away from that fold.

You tri-fold a standard letter sheet to fit in a standard letter. There will never be an opportunity to recalibrate machinery to fold pages on any other metric. Election committees may command something like a million folding machines, and I doubt that many could be recalibrated to the task of missing particular text lines, which will vary by the election. It simply can’t happen, and if it does, random constituencies will have meltdowns.

lurker • June 16, 2021 2:10 AM

At least a fold over a vote-target oval is deterministic and can be reverse engineered and corrected. I’ll take that any day against a random hanging chad…

0bserver • June 16, 2021 2:48 AM

You could just prohibit electronic voting machines and tabulation like sane countries do.

Chelloveck • June 16, 2021 9:57 AM

@lurker – It’s not really deterministic. Basically, it’s a boundary condition which causes a false sensor reading. It’s dependent on how the paper puckers at the fold and exactly where the fold crosses the oval, which may vary a little from ballot to ballot. It’s dependent on whether or not the dust mote can move around. It’s dependent on ambient light and shadow.

Even if it was deterministic the value can’t easily be backed out of the total count. Is it stuck on, always reading as if it was filled in? You can’t just disregard it, because presumably people are actually making that choice. Is it stuck off, always reading as if it was blank? Ditto. If it’s always inverting the sense of the mark you could algorithmically correct it, but that’s unlikely to be a real-world failure mode.

If the sensor picks out the fold as a false reading it’s likely to also pick up not-quite-complete erasures. Pretty much everything tabulated by this machine needs to be re-run on a different machine or counted by hand.

MarkH • June 16, 2021 11:06 AM

@shropshire, Sanderson:

Risk-limiting audits are economical. Most of the time, the sample of ballots needing to be checked is surprisingly small.

This economy makes it feasible to apply them across the board, in time to correct problems such as occurred in New Hampshire prior to certification.

I have thought of only one reason to resist such universal auditing: if you intend to steal future elections based on claims that election processes can’t be trusted, then it’s in your interest to suppress evidence that they’re usually very accurate.

Joe Fromm • June 16, 2021 11:35 AM

Optical Mark Reading is very reliable, but only under fairly controlled conditions. I worked for many years for the companies that make the majority of the test scoring machines used in the US (NCS, Pearson and Scantron). We didn’t do ballot scanning, but the principles are the same.

This page goes into the some of the rules for properly designing an OMR form: https://www.scantron.com/ScanToolsPlus/Help/v8/ScanToolsPlus/content/apicontent/scanflex/designing_forms_for_scanflex.htm

Note the reference to “fold marks”. OMR scanners historically sense the location of bubbles/targets on the Y axis (direction of travel through the scanner) by means of timing marks – black marks along the edge of the sheet. See a timing mark, shine light at the paper and read the reflectance in the mark positions. Since folds can read as marks, the rule is to make extra long timing marks at the fold locations, and to not use those marks for bubble reads. That way the fold doesn’t show up as an extra timing mark, and the fold doesn’t interfere with reading.

And dust buildup in the read head was definitely a thing, at least with high speed scanners. The big outfits had air guns at each scanner to blow the heads out every few thousand pages. The problem is worse on high speed scanners, since accelerating the paper and picking it off a stack tends to dislodge fibers. Slower scanners that are individually fed would have much less of a problem.

intind44 • June 16, 2021 2:42 PM

What do the national voting-machine standards say about how voting machines should distinguish fold-lines from intentional vote-marks?
Could a build-up of dust make the voting machine more likely to misinterpret folds as marks?
What do election administrators in other states do to avoid this problem?
Should New Hampshire throw away its voting machines and buy new ones, or throw away its voting machines and count votes by hand? Or are there measures they could take to use these same machines in a trustworthy way?
What series of circumstances led to this problem in Windham 2020, and how could those corrective measures prevent anything like this from happening again?
Could improved technology in optical-scan voting machines be less susceptible to this problem?

Okay, so if im understanding correctly we are trying to figure out how to keep the elctronic voting machines clean and working correctly?

I am very confident I remember having several large scale debates with other security engineers, and in every single debate we always came to the same conclusion that we need to get rid of electronic voting machines all together.

Need I remind everyone about the disaster of an election we just had? 6 months later they are still being re-counted, half the country doesnt believe in the results, We had an insurrection in part due to this. No one will ever trust a computers results and with good reason. Who keeps pushing this terrible idea? No one is asking for electronic voting machines, everyone agrees they are a bad choice including some of the very poeple that own the companies that made them. Yet, here we are talking about how to build them slightly better for next time.

Apparently we are going to have to get some senators involved or something. Or draft policy to ban computer voting machines. This is a threat to our democracy and im not exaggerating in that respect.

Who are these people thinking they know better than computer security and seem hell bent on “computerizing” everything, including the military is beyond me.

Who is making these decisions? Is it because there is a new administration and they didnt get the message or what?

These people dont understand the critical inherent flaws of networked computer systems. Stop purchasing them and throw them out! I believe Chris Krebs even said to get rid of them. Every election will be the same disaster over and over.

They are an “attempted” solution to a problem that never existed. In the quest for efficient and cheap election, we ended up ended up with the exact opposite. And lost confidence in the election process by many Americans.

We have the money to cover the costs of a paper election. Stop Buying Computer Voting Machines. I mean why isnt NSA and CIA stepping in here. This is a serious issue. Who is paying off who? Im going to have to write a letter to my senators. Rediculous.

Anonymous • June 16, 2021 3:25 PM

@Jon

My assumption would be the ballots were folded in ways not anticipated by the ballot designers. Humans will fold one of two ways: long axis or short axis single or double folded. Accidental mechanical folding can happen at any place on the page, and that can’t be anticipated. Either way the fold will create additional dust that will accumulate on the parts inside the reader and create different headaches for the software architects.

Unless you’ve written software for optical readers and/or designed the ballots for them, you’re not qualified to opine on what’s easy to fix in such a system. The average citizen is no different than a politician in saying “we ought to do this” when it comes to matters of policy: chances are neither are qualified.

RealFakeNews • June 16, 2021 11:43 PM

Just count them by hand, and have cameras in every center to double-check the counting process.

Why everyone is treating the issue of vote fraud as “nothing to see here” is bewildering.

Even if there was no fraud this time, it’s wide-open to it.

I’m still waiting for an explanation to all the shenanigans of votes showing up by the truckload at 3 AM.

echo • June 17, 2021 8:45 AM

@RealFakeNews

Just count them by hand, and have cameras in every center to double-check the counting process.

Why everyone is treating the issue of vote fraud as “nothing to see here” is bewildering.

Even if there was no fraud this time, it’s wide-open to it.

I’m still waiting for an explanation to all the shenanigans of votes showing up by the truckload at 3 AM.

See my comment on not trusting these things. I personally feel the low tech way is better. Pencil and paper, observers from all the vested interests, recounts, and maintaining “line of sight” throughout the whole process.

I understand complicated voting systems may require complicated counting machines. The fact that the US has a system where they are voting for everyone down to the local dog catcher tends to create a psychological and physical cost and a culture of thinking technology can solve every problem all rolls together. It’s kind of interesting that too much democracy can perhaps lead to no democracy. That should keep the number crunchers busy pondering this one.

By comparison the Russian system is so democratic you can vote as often as you like and the men in black leather jackets will help you too!

TRX • June 20, 2021 2:50 PM

My county’s old system was transparent: you got cardboard ballots from one box and a laundry marker from another. You went behind a curtain and marked them, then dropped them into a slot in a box. Every half hour or so the chief election official unlocked the box and passed the ballots to a table full of counters, who counted them right out in the open and tallied them on a piece of paper. You could look over their shoulders and make your own count if you wanted. Every hour the total was phoned in to either the capitol or the county seat, depending on what kind of election it was. After the last count, everything for loaded up for storage at the depot, and all of the hand-counted ballot information was collated and announced within the hour.

Funny, the antique paper system could do that, but systems modern electronic voting machines sometimes take a week before results are available…

Pete • June 21, 2021 3:13 AM

How about : Paper ballots, counted by HUMAN “optical scanners”?
It works perfectly fine in Denmark and many other countries.
Nobody ever cries “election fraud” around here, but then, we also have “racist Voter-ID”..So, dead people don’t vote here.

Schneier on Security

Andrew Appel on New Hampshire’s Election Audit

Comments

Leave a comment Cancel reply