nh primary: Old Machines & No Audits

So Much Better Than Iowa

It's such a relief to have results after an election that we almost forgot to take a look at the New Hampshire primary and say "What could be better, here?" 

First, let's take note of the positive. Many of the towns in New Hampshire count their ballots publicly, in the precinct, by hand, using rigorous counting techniques that have been employed for decades. Usually 4 people look at each ballot. If you'd like to see an example of this, I included footage of it in the first documentary that I made. 

Secondly, even towns that are using machines, are using the type of system that security experts recommend: hand-marked paper ballots, counted by scanners, with a ballot-marking device available for voters with disabilities. 

Internationally renowned hacker Harri Hursti was in New Hampshire for the primary, and liked what he saw, so that is reassuring. 

But NH is not a "dream" scenario for security and accuracy for two reasons.

One is that their technology is very old. Many of the towns are voting on optical scanners that are about 25 years old. Warren Stewart from Verified Voting was kind enough to email me and confirm, "They have been using the AccuVote ES2000 scanners ... since the mid 90s."

The other point of concern is that they are not performing audits. It's great to have paper ballots, but it doesn't mean much if you don't audit the results to confirm that the machines counted the paper ballots correctly. Scanners are computers. They can be hacked, as I demonstrated in a DEF-CON investigation, where hackers took control of a scanner used widely across the country in a matter of minutes. And they can make mistakes. In 2016, I reported on scanners in Wisconsin with a 2.5% error rate. Far greater than the 1% margin of victory that Trump won by there. 


Some optical scanners in WI had a 2.5% error rate in 2016

This data is from the hand-count audit conducted by Wisconsin Election Integrity

the need for audits


NH Secretary of State calls audits "unnecessary expense"

David Brooks at the Concord Monitor reports that the New Hampshire Secretary of State is opposed to audits "as an unnecessary expense." The article says however that a bill pending in the New Hampshire House "would allow voters in any AccuVote community to request a post-election audit, a request currently limited to when results are very close or somehow contested." Security experts support robust post-election audits as the only way to verify whether computerized voting machines are accurately counting results.

Audits are "justifiable and a recommended procedure"

I spoke with Dr. David Bader, of the New Jersey Institute of Technology about potential problems with optical scanners and he said that audits are, "justifiable and a recommended procedure for any technology, especially vulnerable and old technologies," saying that "end of life failures or external attacks – could hurt the integrity of the system." 

He described optical scan machines as being like any other type of mechanical device. "It will have motors that are driving the ballot through. It will be using light and lasers to optically read the ballots ... Maybe there's graceful failure. Maybe there's harder failures, but the only way to know is to test."      

New Technology is not Always Better Technology

As we saw in the Iowa Caucuses, new untested technology can bring risks. Many of the new voting systems on the market are bringing additional risks by using touchscreen machines, counting votes with barcodes or QR codes, and running paper ballots back under the printer after the voter casts the ballot. Security experts warn about the serious risks posed by these elements, and recommend that as many voters as possible use hand-marked paper ballots. They recommend voters with disabilities use secure ballot-marking devices that offer them privacy and independence but do not tabulate votes. 

When New Hampshire towns replace their current voting systems, it will be up to NH voters to make sure they choose a system that meets those standards. 


2008 nh recount

In 2008 New Hampshire Recounted Some Precincts by Hand

Because New Hampshire has both hand-count and machine-count precincts, it's possible we could learn about whether those methods of counting votes are contributing in any way to the outcome of the election. Dr. Bader expressed interest in looking at that type of analysis saying, "It would be interesting to look at the differences between the machine-count vs. the hand count in the precincts that were counted - to see if there's any systematic differences or biases between those two." Our #CountTheVote team hopes to work with him on that.  

A 2008 recount - results shown above - illustrated many discrepancies between machine-count precincts and a public, hand recount. In the majority of precincts there were differences between the machine count and the hand count. Some of the larger discrepancies were explained as "human error" but other discrepancies of as many as 16 votes per precinct would still be significant in a close contest - as the current Democratic primary is turning out to be.   

Dr. Bader suggested that the most useful type of data would be collected over time comparing different technologies in multiple election cycles. 

Dr. Stephanie Singer, a former election official who consults with Verified Voting expressed some frustration at the lack of patience for this type of long-term study examining the accuracy of elections. "The attention window is the day after the election," she said in a phone interview, adding that there are insights to be gathered from retrospective studies of election results, but at that point, "resources are thin and attention has moved on."

Dr. Singer is trying to remedy that situation. She's received a sizable grant from the National Science Foundation and plans to provide tools to help election officials and others better assess the accuracy of election results. "The research will help transform the way data science is used to protect elections ... and can give Americans more justified confidence in election results."

She has a GitHub repository that those interested can explore. Research is in preliminary stages, not Beta she joked, but "Delta." Currently the code "can find the worst looking bar chart for each contest," if the correct data is plugged in. The code is being tested on past elections, starting with North Carolina. Those who have a burning desire to have the results of a particular election examined can request it by opening an issue on GitHub. 

She plans to be publishing some of her research soon. 


SMART Elections is collaborating with other organizations to establish a corps of trained volunteers to help ensure an accurate vote count. Members of a community can connect with local security experts and election officials to advocate for secure voting systems that use hand-marked paper ballots – for those able to mark their ballots by hand – and secure ballot-marking devices for those unable to. They can implement robust post-election audits, and engage in pro-active monitoring of election data and results. 

We encourage everyone concerned about fair elections to join us, or to find their own way to get involved in this mission-critical job. 

To join our #CountTheVote 

team, send an email to contact@SMARTelections.us or get on our mailing list by clicking the link below.

Post by Lulu Friesdat - Co-Founder of SMART Elections. 

Additional research by Suzanne O'Keeffe.