Luther Weeks | CTVotersCount.org

Report: Security Analysis of the Dominion ImageCast X

Report released this week on vulnerabilities of the Dominion ImageCast, used for the vast majority of the votes in Georgia <Report>

The report was actually submitted to a court on July 1, 2021 – the court considered the information so dangerous to elections that is has largely been suppressed until now!

However in two years, Dominion has made several fixes, yet Georgia Secretary of State Brad Raffensperger is in no hurry to update Georgia machines at least until after the 2024 election.

Note: After planning for a couple of months, I launched CTVoters Count in late 2007, Little did I know that the California Top To Bottom Review would be release at that time! Many are claiming that this report may rival the impact of that California report.

Report released this week on vulnerabilities of the Dominion ImageCast, used for the vast majority of the votes in Georgia <Report>

The report was actually submitted to a court on July 1, 2021 – the court considered the information so dangerous to elections that is has largely been suppressed until now!

However in two years, Dominion has made several fixes, yet Georgia Secretary of State Brad Raffensperger is in no hurry to update Georgia machines at least until after the 2024 election.

You do not need to read the whole report. Better to start with a summary from the author, Prof. Alex Halderman. <Summary Report>

…we discovered vulnerabilities in nearly every part of the system that is exposed to potential attackers. The most critical problem we found is an arbitrary-code-execution vulnerability that can be exploited to spread malware from a county’s central election management system (EMS) to every BMD in the jurisdiction. This makes it possible to attack the BMDs at scale, over a wide area, without needing physical access to any of them…

The report was filed under seal on July 1, 2021 and remained confidential until today, but last year the Court allowed us to share it with CISA—the arm of DHS responsible for election infrastructure—through the agency’s coordinated vulnerability disclosure (CVD) program. CISA released a security advisory in June 2022 confirming the vulnerabilities, and Dominion subsequently created updated software in response to the problems. Georgia Secretary of State Brad Raffensperger has been aware of our findings for nearly two years, but—astonishingly—he recently announced that the state will not install Dominion’s security update until after the 2024 Presidential election, giving would-be adversaries another 18 months to develop and execute attacks that exploit the known-vulnerable machines…

The right solution is Voter-Verified-Paper-Ballots and sufficient post-election audits, recounts, and sufficient ballot security. Then even with election systems subject to errors and fraud, election results can be verified and corrected.

Testimony on two small, instructive bills

Last week I submitted testimony on two bills before the GAE (General Administration and Elections Committee.) (Read my testimony here)

This is likely the last time I will testify this year. Both of the bills seem minor, yet offered and opportunity to highlight errors and inconsistencies in the law that are overlooked and not addressed.

The first about collecting envelopes from drop boxes. There is no requirement for more than one person to collect the envelopes. There is no requirement that the collection and materials be logged. Who supports that ballots and other materials should be collected and transported by only one person, at any time?

The other making minor changes to the recanvass law, including requiring a training video from the Secretary of the State. I suggested several other changes, such as notifying all candidates, sending the video link along (so that everyone involved know the rules, and that one observer should be allowed per counting team.

When submitting testimony one can specify Support, Oppose, or General Comments. When signing up to speak the choices are Support or Oppose. I often wrestle with this. I know that some look just at how many support or oppose a bill. Here there is much missing, so I choose oppose.

Last week I submitted testimony on two bills before the GAE (General Administration and Elections Committee.) (Read my testimony here)

Testimony opposed to six bills on RCV and RLAs

On Monday I testified against five bills on Ranked Choice Voting (RCV) and one on Risk Limiting Audits (RLAs),

As I said,

I am not opposed to the concepts of Risk Limiting Audits (RLAs) or Ranked Choice Voting (RCV) but I am opposed to all six of these bills as they are insufficiently detailed. They also provide no guarantees of transparency..

Both of these concepts involve detailed technical and computational issues. Neither are as simple as looking at marks on ballots and simply counting votes. Officials, candidates, the public, and the SEEC need to know exactly what is expected of officials, so they can perform as expected and such that all can determine if they are doing what is required, uniformly across the state…

On Monday I testified against five bills on Ranked Choice Voting (RCV) and one on Risk Limiting Audits (RLAs),

As I said,

Both of these concepts involve detailed technical and computational issues. Neither are as simple as looking at marks on ballots and simply counting votes. Officials, candidates, the public, and the SEEC need to know exactly what is expected of officials, so they can perform as expected and such that all can determine if they are doing what is required, uniformly across the state.

Just like we need paper ballots to avoid trusting voting machines and software, we need transparency to judge all the counting and calculating required for RCV and RLAs.

In both RCV and RLAs the counting rules details are critical. Especially, in close contests, where the result is dependent on the nuances in the rules.

Here is a link to my complete testimony, also including a link to extensive comments on the RLA bill <read>

Testimony on Early Voting and Absentee Voting Bills

Yesterday I submitted testimony on four bills before the GAE (General Administration and Elections Committee.) (Read my testimony here)

I was pleased to learn that Secretary of the State, Stephanie Thomas generally agreed with me and that she called out my testimony in hers!

It was clear before I spoke that the Committee understood my main points, so I asked them to read the testimony and spent my three minutes discussing additional thoughts:

Yesterday I submitted testimony on four bills before the GAE (General Administration and Elections Committee.) (Read my testimony here)

I was pleased to learn that Secretary of the State, Stephanie Thomas generally agreed with me and that she called out my testimony in hers!

It was clear before I spoke that the Committee understood my main points, so I asked them to read the testimony and spent my three minutes discussing additional thoughts:

That if Absentee Ballots are made FOIable (I thought they were), we should post them to the Internet to save citizens from filing FOI requests and officials from processing them.
That my surveys show that our voting machines are NOT failing at an increased rate, and that this might be the worst time in history to replace them. (Because newer VVSG 2.0 machines will be available soon, making anything we buy today obsoleted soon as well.
Emphasizing that if they General Assembly had listened to me in 2013 then we could have avoided many of the problems, some that still remain with Election Day Registration – and that many of the same issues can occur with Early Voting and Same Day Registration, if they ignore similar suggestions in today’s testimony

Nov 2022 Post-Election Audit Report

From the Press Release:

Watchdog Group: 24 Audits Since 2007 with Little Improvement

Independent Observation and Analysis of Connecticut’s Nov 2022 Post-Election Audit

HARTFORD: We conclude, based on citizen observations and analysis of official municipal post-election vote audit of the November 2022 election, that it failed to meet basic audit standards

After 16 years with disappointing, locally performed, hand-count audits, we recommend replacement of all local hand-count audits with sufficient and efficient electronically assisted manual audits utilizing the UConn Audit Station.

The non-partisan Connecticut Citizen Election Audit has provided volunteer observation and post-election audit reports since the adoption of optical scanners statewide in 2007. Without the hours and mileage incurred by these volunteers after every election nobody but a few election officials would know the actual quality of the audits performed, while officials would have less motivation toward credible audits.

From the Press Release:

Watchdog Group: 24 Audits Since 2007 with Little Improvement

Independent Observation and Analysis of Connecticut’s Nov 2022 Post-Election Audit

HARTFORD: We conclude, based on citizen observations and analysis of official municipal post-election vote audit of the November 2022 election, that it failed to meet basic audit standards

The audits were not conducted and reported as required by law. The Secretary of the State’s Office continues to fail to take responsibility for that failure by local officials.
Human error was still considered an acceptable explanation of differences between machine and manual counts. This defeats the purpose of the audits.
Weaknesses in ballot chain-of-custody and security procedures necessary for confidence that ballots were not tampered with between the election and the municipal audit counting sessions.
The short schedule for audits and dates for electronic audits not announced sufficiently in advance cause both registrars and the Citizen Audit to scramble to conduct and observe audits – they should be added to the annual election calendar months in advance.
There were at least three municipalities with new registrars, neither of which had previously performed audits. This resulted in various failures to follow procedures and in one case failure to allow transparency required by the procedures.

The public, candidates, and the Secretary of the State should expect local election officials to be able to organize audits and produce accurate, complete audit reports. The public should expect the Secretary of the State’s Office to take the lead in ensuring that the audit is scheduled in advance, complete, and publicly verifiable.

We are pleased with the following developments:

Electronic audits again included random manual verification comparing some paper ballots to Cast Vote Records produced by the audit station.
There was a significant reduction in incomplete forms.

We emphasize that this report does not question any election official’s integrity.

All reports and backup data are available online at: https://www.CTElectionAudit.org.

CT Secretary calls for 10 days of early voting starting in 2023

https://portal.ct.gov/SOTS/Press-Releases/2023-Press-Releases/Secretary-Thomas-Presents-Legislature-with-Recommendations-for-Early-Voting-Program From no ePollbooks currently approved and no early voting in place, this will be a tall order for the two elected registrars in each of our 169 towns. Likely tripling the number of pollworker days in most towns, requiring lots of recruiting and training and loads of novice pollworkers. This will also put a lot of work on the Secretary’s small staff to develop procedures, provide training, and approve all sorts of plans in a short time.

Editorial:..

Editorial:

Meanwhile in addition to likely approving new ePollbooks, the State has also recently acquired a new CVRs to be implemented. And presumably changes to our current Election Night Reporting system or replacing it with a new one in the new CVRS.

Also, quite a challenge for the Legislature to draft, finalize, and pass such a bill by March 31.

If this plan is approved, I can’t imagine anything but chaos.

Early Voting in Connecticut – Part 5 – Choices and Disappointments

This is the fifth in a series on Early Voting in Connecticut. See <Part 4 – Electronic Pollbooks>

In this post we will cover the choices for implementing Early Voting facing the General Assembly along with the disappointments associated with each choice.

Disappointments are based on the expectations outlined in our first post. See <Part 1 – Expectations>

Option 1 – Fourteen or So Long Days of Early Voting Places…
Option 2 – Four to Six Days, Six to Seven-Hour Early Voting Days…
Option 3 – In-Person Absentee Voting…

Why follow California and Colorado to massive early in-person early voting for just 5% of voters who could all easily choose to vote by mail or on Election Day? Why not benefit/save from their experience, before they do? Start slow, gain experience, add mail-in voting, and learn from our own experience…

This is the fifth in a series on Early Voting in Connecticut. See <Part 4 – Electronic Pollbooks>

In this post we will cover the choices for implementing Early Voting facing the General Assembly along with the disappointments associated with each choice.

Disappointments are based on the expectations outlined in our first post. See <Part 1 – Expectations>

Option 1 – Fourteen or So Long Days of Early Voting Places

By this we mean, in general, following the request of the ACLU and the provisions of the For the People Act. Perhaps ten days or twenty-five. Likely requiring voting on at least two weekends including the Saturday and Sunday before Election Day. Open in the mornings and evening hours, at least during the week. Uniformity across the State. An early voting place similar to a polling place, i.e. checkin, ballot clerks, machine tenders and scanners – giving the voters who make overvotes the opportunity to spoil a ballot and vote another. Similar to what we see in other states, including CA, CO, GA, etc.

Disappointments:

Registrars and other election officials. In a small town with one polling place today. It will change the number of pollworker days from perhaps 8-10 to about 100! Finding citizens willing to work that many 16 to 17 hour days will be a challenge. Staffing, presumably, with inexperienced pollworkers will add to the challenges. Further it will be more difficult to find voting locations, especially during the week, that are not already used for other things and that have sufficient parking. It will greatly increase the stress and work for registrars as they do the normal work preparing for Election Day – especially tight preparing pollbooks between that last Sunday and Election Day. We expect many will join those that resigned in 2022. As an experienced pollworker, I doubt I could serve more than one or two 16 hour days in addition to Election Day.

Not much different in mid-size towns like mine, with six polling places and central count absentee, about 70 Election Day officials, this would add perhaps another 140 early voting polling place staffing days. That is plenty of novice officials, plenty of work for registrars, and stress. Including problems finding appropriate venues. Even for large cities it will be significant, perhaps doubling staffing.

The Public. Presumably many will want to try out early voting the first day that seems convenient and the last. So, there will likely be lines as we see in every election in Georgia. The public has been promised no more lines in Connecticut – yet there are none today, except for big problems like missing pollbooks, or at Election Day Registration. In fact, every option we discuss may have that same problem. It is especially likely for any option that opens early or closes late. Especially on that last Sunday if “Souls to the Polls” materializes in Connecticut.

Perhaps we should start slowly and if early voting is popular, work up to Option 1. More on that in the next two options, and in our final comment.

Option 2 – Four to Six Days, Six to Seven-Hour Early Voting Days

This is a compromise between satisfying the public’s expectations for polling place like voting places, but with fewer and shorter days for officials.

Four days, starting 10 days before Election Day: Saturday and Sunday 9:00am to 3:00pm, Tuesday 6:00am to 1:00pm, and Wednesday 2:00pm to 8:00pm. This would provide the possibility of election officials to do multiple days – a single Moderator (or with an Assistant) might be able to cover all these days and tabulate to votes on election night. Registrars would have more time to prepare pollbooks and otherwise for Election Day. Voters would have an opportunity to vote at any convenient time, weekend, early morning, after work, and at lunch time.

Six days, starting 10 days before Election Day: Saturday and Sunday 9:00am to 3:00pm, Monday and Tuesday 6:00am to 1:00pm, Wednesday and Thursday 2:00pm to 8:00pm. A bit more work, but more opportunities for voters.

Disappointments:

The Public. Who were expecting more days and will still see lines.

Election Officials. Who may be somewhat relieved but still face quite a bit of work and other challenges.

Option 3 – In-Person Absentee Voting

This would be voting very similar to the absentee voting that occurs today when one goes into a municipal clerk’s office, makes an application, votes, and hands in their absentee ballot. Except that any registered voter could do it. Clerks would need to be open on at least one weekend and in many cases open more hours than today, perhaps hiring one or two additional staff.

It would be more and less work for registrars and pollworkers. More checking-off of pollbooks between the end of early voting and election day. Today that happens on Friday and Monday before Election Day. As long as early voting ended before Friday, the only change would be more staffing for absentee checking and counting, with a somewhat smaller polling place staff on Election Day.

Some have suggested going to the registrars’ office and voting via machine for many hours of availability, yet that is really a variation on option 1, perhaps for small towns. It would still require Democrat and Republican Registrars or Deputies be present to correct registration errors, perhaps a checker, ballot clerk and definitely a machine tender all visible to each other and voters – not an option when you expect a volume of voters.

Disappointments:

The Public. Who were expecting that polling place like voting experience, the opportunity to be protected from overvoting, and the opportunity for correcting erroneously not being on the voters registration list – presumably that would require a call to the registrars office, visiting the registrars office when it was open, or requiring the registrars office to be open all the hours of early voting.

Once again, long lines are possible, especially if early morning or evening opportunities are limited and if on that last Sunday if “Souls to the Polls” materializes in Connecticut.

Finally, A Concern – The Experiences of California and Colorado

Many say the ideal for Connecticut should be California and Colorado. I would not emulate everything they do. They have ten or more days of early voting in vote centers, people can vote where they live or where they work, they have absentee voting, in fact they now send ballots to every voter to mail-in or drop off. Pretty close to the final draft of the For the People Act. They did not arrive there overnight. It has been perhaps 20-30 years in the making in each state.

Contrary to what the ACLU and Brennan Center would have us believe, we should not be trying to emulate the likes of Georgia, Florida and many other southern states that have the “Highest Early In Person Voting Rates” (see the map on page 3 of Brennan Center report, and the exclusion of California and Colorado from much of their report.) Those states have high in-person early voting apparently because they try to suppress absentee voting and do not provide enough polling places and voting machines on election day.

On the other hand, California and Colorado both seem to have the same experience. Until recently, the voters have chosen to use 70% main-in voting, 20% Election Day voting, and just 10% early voting. The latest trends in both states are closer to 90% mail-in voting.

As we have said, early in-person voting is expensive. If we are trending our policies toward California and Colorado, especially if we pass the 2024 constitutional amendment for no-excuse absentee voting in Connecticut. Why follow California and Colorado to massive early in-person early voting for just 5% of voters who could all easily choose to vote by mail or on Election Day? Why not benefit/save from their experience, before they do? Start slow, gain experience, add mail-in voting, and learn from our own experience.

Early Voting in Connecticut – Part 4 – Electronic Pollbooks

This is the fourth in a series on Early Voting in Connecticut. See <Part 3 – New Voting Machines>

In this post we will cover Electronic Poolbooks – Why, How, and When we should add electronic pollbooks. Next time we will cover the alternatives for early voting in 2024. Hint: they all have advantages and disadvantages.

Our understanding is that UConn, under the direction of the Secretary of the State’s (SOTS) Office is already evaluating electronic pollbooks. Presumably they could be selected by the SOTS sometime in 2023.

Why Electronic Pollbooks (ePollbooks)

The answer here is not as simple and clear as many would suggest. There are two advantages often touted for ePollobooks which are not actually true:

This is the fourth in a series on Early Voting in Connecticut. See <Part 3 – New Voting Machines>

Why Electronic Pollbooks (ePollbooks)

The answer here is not as simple and clear as many would suggest. There are two advantages often touted for ePollobooks which are not actually true:

First, they are often touted as speeding up checkin such that fewer pollworkers are required. Actually, they don’t significantly speed up checkin. Sometimes they speed it up a little, sometimes they slow it down a bit. There are several variables here: The model of electronic pollbooks employed, what the State checkin requirements are, checkers capabilities, checker training, and how the pollbooks are implemented (including are they connected to a central database which may slow down the process). Suffice to say that no matter if 1, 2, 3, 4, or more checkers and lines are required today, likely that number will still suffice and be required.

Most states require voters to sign a pollbook or sign electronically with an electronic pollbook, Connecticut does not. Some print a pass for the correct district for each voter – now Connecticut may need to do that to tell ballot clerks what ballot to give to a voter – especially for early voting. All these things take extra time on the part of checkers – or maybe save time in other states over what they were doing without ePollbooks.

Second, they save paper because paper pollbooks do not have to be printed. In fact, they still need to be printed as a backup for when electronic pollbooks fail. They fail for many reasons: software glitches, download glitches, power failures, hardware failures etc. So not only are paper backups required, the checkers must be ready to immediately fallback on paper pollbooks.

There are advantages:

Any voter can go to any line as ePollbooks are usually connected within a polling place and coordinate a master checklist between them. That can speed the process a bit and also allow for easily reducing the number of checkers in slow periods and similarly facilitate adding checkin lines.

A more comprehensive voter search may be available. Easily finding and restoring voters moved off the rolls for not voting recently, finding voters added supplementally, and finding which polling place to which a voters is assigned.

All checkers can handlevoters in multiple districts in a polling place (especially early voting places).

Especially for early voting places, if connected to the Central Voter Registration System (CVRS), then more than one polling place can handle voters in a municipality. Also, even when not connected a single polling place, they speed the setup for Election Day voter lists. They can simply be uploaded to the CVRS to then update pollbooks for Election Day.

Finally, they can save registrars a lot of time updating the CVRS with who voted after the election. Today it is a time-consuming manual process, to data enter who voted from the paper checklists.

How to Add Electronic Pollbooks

Most important in evaluating them is testing them with actual typical pollworkers and people acting like a range of average voters. How fast can the average pollworker type in the addresses and select voters? How fast can they do searches when the voter is not initially found? What does it take to setup the systems in a typical polling place? What additional time does it take to provide voters with receipts for which district they are in?

ePollbooks should be tested for compatibility, ease of use, and performance in downloads, uploads, and connectivity to the CVRS. And what happens during Election Day or early voting when connection is lost? Our assumption though is that especially on Election Day many, if not most, polling places will not be connected centrally, since many polling places do not have internet and may also be in poor cell communications territory.

And like new scanners, ePollbooks should be purchased and maintained by the State. Each polling place should have one station (likely a laptop) for the maximum number of checkers in that polling place, plus one or two extras. That is perhaps $500-1000 per station, plus annual maintenance.

Like implementing new machines and early voting it takes procedure development, planning, training of registrars and pollworkers.

But unlike some other changes ePollbooks can be implemented in stages: Start with a few polling places using them in parallel with paper pollbooks or just using them officially, verifying their effectiveness and improving procedures before a full rollout. Like other changes, they can best be implemented in lower volume elections or primaries.

When to Add Electronic Pollbooks

As we discussed in Part 2, change should be limited to one big thing at a time, if possible, done in part first, and avoiding a big change in even year elections.

There is good news for electronic pollbooks here. It would be possible to do a small test either in the September 2023 Municipal Primaries, the November 2023 Municipal Elections, or both. Then expended and made universal across the State in the 2024 Presidential Primaries or the August Primaries, so every municipality and checker could be experienced in time for the November 2024 election with some of their benefits available for early voting.

Early Voting in Connecticut – Part 3 – New Voting Machines

This is the third in a series on Early Voting in Connecticut. See <Part 2 – Implementing Change> See <Part 4 – Electronic Pollbooks>

In this post we will cover New Voting Machines – Why, How, and When we should implement new voting systems.

Why New Voting Machines

The simple answer is for two reasons..

Stay tuned, we plan at least one more post before we get to the choices for implementing in-person Early Voting.

This is the third in a series on Early Voting in Connecticut. See <Part 2 – Implementing Change> See <Part 4 – Electronic Pollbooks>

In this post we will cover New Voting Machines – Why, How, and When we should implement new voting systems.

Why New Voting Machines

The simple answer is for two reasons:

First, Connecticut’s voting machines are ageing, aging both in technology and physically. For the most part they were acquired in 2007, two machines for each polling location and for central count absentee locations. A couple of years ago SOTS Denise Merrill acquired a stash of extras to have on hand to replace any that were beyond repair – they are also available for a song, used. A recent article articulates the view of many registrars, mostly through the eyes of a novice registrar. Her views contrast in several ways to ours: CT’s voting machines are ‘past their useful life’ and in need of replacing <read>. Let us annotate some of the statements in that article.

On Election Day this year, there were reports from several towns of malfunctioning tabulators.

That has been true for every election and primary since 2007. As far as we know, there are no official or unofficial statistics for the rate of failure and the causes, except those of CTElectionAudit.org which has random sample reports going back to 2007. (I am also Executive Director of the Citizen Audit.)

Here is a brief history of AccuVoteOS scanner problems in Connecticut:

Shortly after the scanners were deployed there were extensive problems with the memory cards used on each machine to hold the program and totals for each ballot for each machine. The state wisely had ordered two machines for each polling place, along with four memory cards. Most errors were found as each machine and all four memory cards were tested before the election, others immediately on election day, quickly replaced by the backup scanner or one of the extra cards. Several years later the State purchased a compatible more modern memory card. Since that purchase there have been very very few hardware problems with memory cards – there have been problems with incorrect programming or incompatible ballot printing, once again, usually discovered in pre-election testing.

According to the CTElectionAudit reports about 4-5 years ago there were growing problems with scanners failing, mostly due to wearing out of rollers which grab and move ballots through the scanners. The root cause was poor routine maintenance by the distributor, LHS Associates. LHS Associates is contracted by the State to program the memory cards and to perform scanner maintenance. Soon the problem went away and the CTElectionAudit statistics for roller and scanner problems went back to normal.

Preliminary CTElectionAudit results from the November 2022 election show that scanner roller problems remain at the normal level, with no significant additional scanner problems reported.

In Norwalk, for example, the sole tabulator at Brookside Elementary School was broken for about a half an hour, Democratic Registrar of Voters Stuart Wells said that day.

That is interesting, experienced registrar Stuart Wells and his staff were able to fix the scanner in half an hour. That is about what it takes to fire up a backup scanner or replace a memory card with a backup. A more interesting question is why that polling location did not have a backup scanner at the ready, since the State purchased more than enough to provide two to each polling place? And the number of polling places keeps shrinking with each redistricting. (I once worked for the Dean of Connecticut Registrars, the late Judy Boudreau. She made us fire up two scanners before the polls opened, such that a failing scanner could be replaced by the ready and waiting backup. Sounds like a good plan to me if registrars are concerned – the cost is an extra ballot box to hold the waiting backup scanner.)

Cara Gately, the Republican registrar in Darien, said “every town had some impact.”

During the primary in August, a rubber roller that pulls the paper ballot into the tabulator started to melt in the summer heat, Gately said, a problem she said people have called “melting tabulators.”

Those things are rubber and so in the heat in August they started to get gummy,” she said. “The texture changed, and there was residue on the ballot, which then gums things up and then they just stopped working.”

Every town is a large exaggeration! That was August. Many towns had problems like that. Many did not. The cause was simple, heat above the stated temperature maximum for operation of the scanners. Apparently, the failing machines were well maintained in time for November. I expect they cooked many pollworkers as well. Polling places should be air-conditioned and heated, machine specifications should be followed!

In order to fix broken tabulators, parts must be cannibalized from old ones from nearby states that have already replaced theirs.

We are not sure exactly which states those are. According to the Verified Voting Verifier they are still in use in about six states including New Hampshire and Massachusetts.

This is Gately’s first year as a registrar of voters, but she was told that “the current tabulators that the state has was the year that model got decommissioned.”

Not sure what that means or who was the authority that told Gately that. The AccuVoteOS is no longer manufactured due to no demand, however, it still meets the 2002 Voluntary Voting Systems Guidelines (VVSG). Those standards are still in effect for purchasing new voting machines today. However, they will be superseded by a recent major upgrade effective November 2023 to VVSG 2.0 standards. More on that later.

[SOTS Chief of Staff] Rosenberg said the process has begun to replace them, but that “it has to be done in a transparent way.”

“We’ll take requests for proposals and take a lot of public input,” he said.

We are hopeful that it will be transparent with a lot of public input!

[Gately] said she hopes that the tabulators will be replaced before the 2024 presidential election, or shortly thereafter. One year, she said, is doable, but five years “would make me nervous.

We will address timing later in this post.

Secondly, the second half of 2023 will be an ideal time to begin evaluating new scanners. That is because in 2020 the Election Assistance Commission approved VVSG 2.0 the first update to the 2002 VVSG standards which are in effect today. However, those will be effective in November 2023. That means that after that no one can purchase new jurisdiction wide machines that do not meet the new standards (Old machines may be purchased to add to existing machines in use.) So, during the second half of 2023 the three vendors will likely all be submitting machines to testing and then offering machines that meet the 2.0 standards.

We have been arguing for years that newer and better machines will become available and that Connecticut should, if possible, wait for then to purchase products that will be new rather than old out of the box. Now it is clearer when that will be possible.

How to Update Voting Machines

As we said in Part 1: In 2005 the SOTS Office initiated evaluation of voting systems for Connecticut with UConn testing, followed by public demonstrations of machines in four locations around the State, also with focus groups of registrars, those with disabilities, and technologists providing feedback on the machines. In late 2005 machines were selected. Then in November 2006 those machines were used in 25 municipalities in the even-year State election. Procedures were developed in 2006, refined in 2007 followed by registrar, pollworker, and public education, then implemented statewide in the September 2007 municipal primary. Still various problems, concerns, and complaints were found in the November 2007 elections.

That is the kind of process we favor. Note machines were evaluated starting as early as 2004, with the evaluation taking most of 2005. Then a few were used in November 2006 after some planning. Even more planning, procedure creation, training, and voter education occurred in 2007. Pretty much three full years.

There is probably less need for voter education this time as we are already use paper ballots. Perhaps a little less planning and official training, yet from the standpoint of officials the machines will be different, plans and training must cover new and changed features. On the other hand, there may be more need for thorough evaluation, testing, and research, since last time Connecticut was a little late to the game and many other states had already evaluated, purchased, and deployed systems meeting the 2002 standard. We might be one of a few acquiring them out of the gate this time.

We do have some inside information from other states regarding the current systems available from the three vendors, presumably the 2023 machines will have some of the same advantages and concerns. The machines are significantly slower, dependent on the length of ballots, ballot sides to be voted on, and on the number of bubbles on each side. That is partially because, unlike our current scanners, they make an image of each ballot side and then interpret that image, creating a Cast Vote Record (CVR) of the votes on each ballot. Otherwise, that is a benefit facilitating certain kinds of audits and recounts, along with making Ranked Choice Voting feasible. Yet, for those benefits we will need an extra one or two scanners in many polling places. For instance, in my town, Glastonbury, we have six polling places where in a tight presidential election we can expect 3,500 to 4,000 voters. It’s doubtful one scanner can handle that. However, perhaps firing up and using two, and having a backup delivered if one of them fails would be sufficient. Some towns, like Greenwich have huge ballots for municipal Representative Town Meeting elections, may need more machines to handle those elections.

Beyond that some central count absentee ballot locations may benefit from high-speed scanners. They are expensive and may not be necessary if enough regular speed scanners are purchased.

Finally, we are aware of one brand where the rejection of over voted ballots may not work as our current scanners work – in a much less acceptable way. That needs to be evaluated and perhaps a fix negotiated with the vendor(s). Maybe there will be other issues uncovered. The sooner they are uncovered the better. We do not want to find them in an election after a huge long-term purchase!

Perhaps obviously, the State needs to pay for the acquisition, maintenance, procedure development, and training for the new machines as they did last time.

A couple of years ago, SOTS Denise Merrill estimated $20 Million for new machines. I estimated $12 Million. We could both be right, just estimating different things such as including long term maintenance. On the other hand, I had available information that exposed the actual selling price of the equipment available then, much lower than the list prices.

When to Implement New Machines

As we discussed in Part 2, change should be limited to one big thing at a time, if possible, done in part first, and avoiding a big change in even year elections.

Last time it took about three years from the beginning of the evaluation to the final implementation. Maybe two years or so will be enough this time. Since machines to test may not be available until mid-2023 it seems the natural time to implement new machines would be the 2025 municipal elections – starting with the September 2025 municipal primaries, giving a short time to tweak things for a full roll-out in November 2025.

Changing a machine will also likely require changes in how the machines are programmed. Right now, our Election Night Reporting System is completely separate from the voting machines. If that stays the same, then there will be less change at once!

What about 2024 or 2023? Both are too soon for a thoughtful evaluation, procedure development, and training. In 2023 VVSG 2.0 machines may not even be available to evaluate until after the election. Also 2024 is the worst possible time since it is a presidential election, and also likely to have the first running of in-person Early Voting.

Stay tuned, we plan at least one more post before we get to the choices for implementing in-person Early Voting.

Early Voting in Connecticut – Part 2 – Implementing Change

This is the second in a series on Early Voting in Connecticut. See <Part 3 – New Voting Systems> or <Part 1 – Expectations>

In this post we will cover Implementing Change – how election changes have been implemented in Connecticut and the risks of doing too much too fast and at the most challenging times. In the future we will address more specific issues associated with some of the changes coming. Then get to the tradeoffs in implementing in-person early voting.

Implementing Change

As we discussed last time, one of the big changes coming is in-person early voting.

In addition to that:

In the last few months, the Secretary of the State (SOTS) and his office have selected and are presumably beginning implementing a replacement for the Central Voters Registration System (CVRS)…

Do one change at a time, test as much as possible, then test the change on a small scale, and implement it system wide at the least disruptive time...

This is the second in a series on Early Voting in Connecticut. See <Part 3 – New Voting Systems> or <Part 1 – Expectations>

Implementing Change

As we discussed last time, one of the big changes coming is in-person early voting.

In addition to that:

In the last few months, the Secretary of the State (SOTS) and his office have selected and are presumably beginning implementing a replacement for the Central Voters Registration System (CVRS). That same system may replace the current Election Night Reporting System (ENR)
UConn under the direction of the SOTS has begun evaluating various electronic pollbooks (ePollbooks).
There are calls for replacing our aging scanners with newer models. Actually, the second half of 2023 would be an ideal time for UConn to begin a technical evaluation.

Not just for the State, but for any institution there are good methods for implementing change: Do one change at a time, test as much as possible, then test the change on a small scale, and implement it system wide at the least disruptive time.

Yet, when it comes to election systems, some changes can’t be done piecemeal: Early voting must be available to all voters at the same time or it would be a civil rights/equality issue. The CVRS must be implemented statewide at the same time (it can be implemented between elections, but some of its functions must work on election day and the days before – functions that are not generally done at other times of the year, under circumstances of demand for creating pollbooks right before the election and registration checking on election day.

Using new voting machines or new ePollbooks can be done in a few municipalities first.

Connecticut has a good recent record of planning such changes, but not a great record in the actual implementation.

In 2005 the SOTS Office initiated evaluation of voting systems for Connecticut with UConn testing, followed by public demonstrations of machines in four locations around the State, also with focus groups of registrars, those with disabilities, and technologists providing feedback on the machines. In late 2005 machines were selected. Then in November 2006 those machines were used in 25 municipalities in the even year State election. Procedures were developed in 2006, refined in 2007 followed by registrar, pollworker, and public education, then implemented statewide in the September 2007 municipal primary. Still various problems, concerns, and complaints were found in the November 2007 elections. There will few problems. Many of those complaints and concerns were normal for the transition from lever machines to scanners – a couple of years later New York went through a similar process with a different brand of scanner with the same problems, concerns, and complaints. This was a positive example of how extensive evaluation, planning, and training can have a great result.

In 2013 the state implemented Election Day Registration, with several problems which we had predicted – mainly that by law being in line by 8:00 was insufficient to be allowed the opportunity to register and vote. But those problems were not addressed. In 2014 those problems resulted in long lines at EDR locations with many voters turned away, including those still in line at 8:00pm. Despite news stories across the State those problems were not addressed in law for several years. The long lines somewhat reduced still remain, however, now anyone in line by 8:00 can have the opportunity to register and vote. This is an example of inadequate planning and a deaf ear by the SOTS, the General Assembly, and election officials, many of whom really did not have sympathy for voters who failed to register by the regular deadlines.

In 2012 the SOTS and registrars implemented the new Election Night Reporting system in parallel with the regular system. It was dead on arrival as polling place moderators refused or were incapable of inputing reams of data after a long day via their smart phones (many did not even have or understand smart phones at the time.) Once again this was all predicted. The system was designed completely in isolation from the real world. There was no feedback from actual officials, no testing with the actual polling place officials. Worse the SOTS Office attributed the failure to the officials (Even today I would be challenged to enter that much data from my cell phone, after a long day. Others are in polling places with poor cell coverage and no internet.) Finally, years later the system was redesigned and after a couple of elections was working well with officials in town halls putting in the data from desktops and laptops. Once again, a predictable failure abetted by tone deaf State officials.

Lessons We Hopefully Will Follow

Do one change at a time, test as much as possible, then test the change on a small scale, and implement it system wide at the least disruptive time.

There are a lot of calls for multiple changes in a very short period, from the public, from advocates, and from election officials. In subsequent posts we will address some of those changes in the light of these pressures and the lessons we hopefully will follow. Then get to the tradeoffs in implementing in-person early voting.