Case 16 — Monmouth County, 2022

16.1 — Introduction

On November 8, 2022, voters in Monmouth County, New Jersey, went to the polls in the general election. Over 900 flash drives carried results from voting machines to the county’s election management system. On election night, county personnel inadvertently loaded six of those flash drives twice. Every vote on those six drives was counted twice.

The software was supposed to prevent this. A patch called the ECO — Engineering Change Order — was designed to detect duplicate flash drive loads and block them. But four months earlier, when an ES&S employee had uninstalled and reinstalled the election software to troubleshoot an unrelated network problem, the ECO was not reinstalled. The person who performed the work was not a technician. He had never installed software at ES&S and had never heard of the ECO. His instructions came over a mobile phone from a colleague in the sales department. No one at ES&S verified the work afterward.

The result: in four municipalities, votes were double-counted. In one — Ocean Township — the wrong candidate was declared the winner of a school board race and took office. The official post-election audit reported “tabulation accuracy was 100%.” A citizen who noticed 311 phantom votes in his town’s mayoral race was told he was comparing “apples to oranges.” It took over two months, a records request, an outside investigation by a former state attorney general, and a court-ordered hand recount before the error was fully uncovered and the correct winner was seated.

This is the only case study in this collection where a reporting error definitively changed the outcome of an election.

16.2 — Background: How Monmouth County Reports Votes

Monmouth County installed a new ES&S voting system in October 2021. The hardware includes ExpressVote XL machines — hybrid ballot marking device/tabulators that collect ballots both in paper form and on secured flash drives — along with paper-ballot tabulators for vote-by-mail and provisional ballots. The software is Electionware, an election management system that runs on a PostgreSQL database and sits on a dedicated file server not connected to the internet.

Results reach Electionware in two phases. First, on election night, ballot results are transmitted from regional sites through a process called “Regional Results.” This loads the vote totals. Second, starting election night and continuing the next day, county personnel insert the flash drives into election workstations to load ballot images — the scanned pictures of each ballot — so they can manually tabulate write-in votes. This second phase, the “Ballot Image” phase, is not designed to alter vote totals. It only adds images.

The critical safeguard is a software component called the ECO — Engineering Change Order. The ECO “optimizes” the PostgreSQL database and prevents ballots from being counted more than once. When a flash drive that has already been loaded is inserted again, the ECO detects the duplicate and displays an error message: “These Results were previously loaded. No files have changed.” The software then refuses to count those ballots a second time.

ES&S developed the ECO in mid-2020 in response to the known possibility that flash drives could be loaded twice. In the software version Monmouth County was using in November 2022, the ECO had to be installed separately from the main election software. A newer version of Electionware, which the county had considered upgrading to but decided to wait on, would have included the ECO automatically.

16.3 — The Disabled Safeguard

In late June and early July 2022, Monmouth County personnel experienced slowness and error messages on their election workstations. A July 26 special election was approaching. The county contacted ES&S for help.

ES&S sent several employees to troubleshoot. They could not initially identify the problem. Eventually, the company decided to send someone to uninstall and reinstall the election software on the county’s file server. The New Jersey-based technician who would ordinarily have performed this work was out of the office. ES&S instead sent a Customer Relations Manager — a person responsible for managing ES&S’s New Jersey accounts. He had decades of experience in account management for election technology companies. But he was not a technician. He had never uninstalled or reinstalled software at ES&S. He had never heard of the ECO.

ES&S also arranged for a Principal Technical and Account Support Engineer to provide instructions to the Customer Relations Manager over a mobile phone while he was on-site. This person had some experience with software reinstallation. But he was not a member of ES&S’s Technical Services Department either — he was part of the sales group.

On July 14, the Customer Relations Manager met with a programmer from the County Clerk’s Office, who controlled access to the election servers. Over a mobile phone, the engineer provided instructions for the uninstallation and reinstallation. No written instructions were given. The programmer from the Clerk’s Office was in the same room but could not hear the other end of the phone call because the Customer Relations Manager was using earbuds.

The Customer Relations Manager uninstalled Electionware and PostgreSQL on the primary file server. When these are uninstalled, the ECO is automatically removed as well. He then reinstalled Electionware and PostgreSQL. But the engineer on the phone did not instruct him to reinstall the ECO separately — which was required in the software version Monmouth was using. The Customer Relations Manager did not install the ECO. He did not know it existed.

The double-counting safeguard was now disabled.

The reinstallation did not fix the slowness problem, which turned out to be caused by a network IP address conflict — an issue unrelated to the election software. ES&S sent additional technicians to Monmouth in late July and August to continue troubleshooting. None were instructed to confirm that the July reinstallation had been done properly. ES&S did not verify that the ECO was in place.

16.4 — Election Night

On November 8, 2022, Monmouth County administered its general election. Starting election night and continuing the next day, county personnel inserted flash drives into election workstations to load ballot images into Electionware.

The AG investigation found that personnel followed a process of inserting flash drives one at a time to avoid duplicates. Nevertheless, some drives were inadvertently loaded more than once. In total, 19 flash drives were inserted more than once.

For 13 of the 19, the system worked as intended. The error message appeared: “These Results were previously loaded. No files have changed.” These 13 were loaded during the Ballot Image phase, where Electionware’s built-in duplicate detection still functioned.

For the remaining six, the error message did not appear. These six were loaded during the results-loading phase — the phase where the missing ECO would have caught the duplicate. Without the ECO, Electionware accepted the flash drives a second time and counted every ballot on them twice. In each case, the second load occurred less than one minute after the first.

The six flash drives were from voting districts in four municipalities: Belmar, Tinton Falls, Fair Haven (two drives), and Ocean Township (two drives).

At the end of the day on November 9, someone using a county Electionware account ran a Media Status Report — a report that tracks the load status of all flash drives, including whether any were loaded more than once. The report showed the double insertions. No one noticed.

16.5 — The Failed Checks

Every official safeguard that could have caught this error failed to do so.

Pre-election Logic and Accuracy testing — required by New Jersey law — did not include testing for what happens when flash drives are loaded multiple times. The testing verified that the machines counted votes correctly, which they did. It did not verify that the system could prevent votes from being counted twice.

On November 21, the Monmouth County Board of Elections and the County Clerk met to certify the November 2022 election results. They certified them.

The post-election audit — conducted by the Board of Elections and submitted to the Secretary of State — reported that “the tabulation accuracy was 100%”. The audit did not detect the double-counting.

The Electionware software itself contained the evidence. The Media Status Report, run on November 9, showed the double insertions. An Election Audit Events Report — which tracks all activity in Electionware — also recorded them. But county personnel stated that they were never instructed by ES&S to run these reports after elections. The County Clerk stated that she did not even know those reports existed in November 2022. The AG investigation noted that the reports are “quite lengthy and difficult to read” — the Election Audit Events Report from election day and the following day was nearly 1,000 pages, and the Media Status Report was nearly 300 pages.

The evidence was there. Nobody looked at it.

16.6 — Discovery

The same afternoon that the election was certified — November 21, 2022 — James Bean, a former city councilman in Belmar, filed a request under New Jersey’s Open Public Records Act. He had noticed that the Belmar mayoral race showed 2,823 votes counted for the two candidates, but only 2,512 people had actually voted. That was 311 phantom votes.

The Superintendent of Elections’ office was on vacation. Bean’s request was received when the Secretary returned on November 28. Over the following weeks, Bean emailed repeatedly, asking how more votes could have been counted than were cast. On December 14, an employee in the Superintendent’s office responded: “It’s like comparing apples to oranges if you will. Numbers will not exactly match.” (NJ Globe, Jan. 18)

That answer was wrong. When more votes are counted than people who voted, it is not an apples-to-oranges comparison. It is evidence of a problem.

The pattern of dismissing citizen concerns has appeared before. In Baldwin County, Alabama in 2002 (see Case 5 — Baldwin County, Alabama, 2002), the Democratic county chair was told the courthouse was closing and sent home before officials reopened the building and continued counting — after which approximately 6,300 votes vanished from one candidate’s column. In both cases, people who asked reasonable questions about the numbers were brushed aside before higher authorities took the matter seriously.

A few days later, Bean notified the County Clerk of the discrepancy. The Clerk responded that she was working on the issue with the Superintendent of Elections. But the investigation did not move quickly.

On January 11, 2023, the County Clerk contacted ES&S’s Vice President of Account Management and requested that the company escalate its investigation. At a company meeting in Omaha, a Software Support Analyst ran an Election Audit Events Report — the same type of report that could have been run in November. It confirmed that ballot results on six flash drives had been loaded twice into Electionware. The analyst also ran a Media Status Report, which confirmed the same finding.

ES&S then determined that the double insertions had affected the outcome of the Ocean Township Board of Education race. The wrong candidate had been declared the winner.

On January 17, the New Jersey Globe broke the story.

16.7 — The Race That Changed

The Ocean Township Board of Education election was non-partisan, with the top three finishers out of six candidates winning seats. In the initial certified results, Steve Clayton finished third with 3,523 votes, defeating incumbent Jeffrey Weinstein’s fourth-place total of 3,503 by just 20 votes. Clayton took office in January 2023.

On January 20, the Board of Elections and the Superintendent, represented by the Attorney General’s office, filed a petition in New Jersey Superior Court seeking a hand-to-eye recount and voting-machine recheck of the November 2022 election in the four affected municipalities. On February 2, Judge David Bauman ordered the recount to begin.

The recount showed that the double loads accounted for nearly the entire discrepancy. The initial results had shown 20,721 total votes in the Ocean Township school board race. A Results Report run without the double loads showed 20,109 — 612 fewer votes. The hand recount produced 20,104 total votes, just five fewer than the corrected Results Report, which is within normal recount variation.

In the school board race, the recount showed Weinstein with 3,408 votes and Clayton with 3,404 — Weinstein ahead by four. Clayton had lost 119 votes from his initial total; Weinstein had lost 95. The four-vote margin replaced the original 20-vote margin in the other direction.

On February 22, Judge Bauman declared Weinstein the winner, revoked the Certificate of Election awarded to Clayton, and directed a new certificate for Weinstein. Clayton had served on the school board for approximately two months, attending five meetings and casting votes — including at least one instance where his vote was in the deciding majority on a 5-4 issue.

Weinstein was sworn in at the school board’s March 7 meeting.

In the other three municipalities — Belmar, Fair Haven, and Tinton Falls — the double-counting did not change the outcome of any race. In Belmar, Bean’s 311 phantom votes in the mayoral race inflated both candidates’ totals. Democrat Gerald Buccafusco’s initial margin of 111 votes over Republican Mark Walsifer survived the correction because the double-count affected both candidates roughly proportionally.

16.8 — The Investigation

On January 24, 2023, Attorney General Matt Platkin announced a formal investigation. Because the AG’s office serves as legal counsel to the Monmouth County Board of Elections — creating a conflict — Platkin retained former New Jersey Attorney General Peter C. Harvey and the law firm Patterson Belknap Webb & Tyler LLP to lead an independent probe. Harvey, a former federal prosecutor, and his team agreed to serve pro bono.

The investigation interviewed 23 witnesses in Trenton, Freehold, Camden, Nashville, and Omaha. All relevant parties — ES&S, the Monmouth County election offices, and other witnesses — cooperated.

Harvey’s report, released September 6, 2023, provided an extraordinarily detailed forensic account of the error chain, from the October 2021 installation through the January 2023 discovery. Its conclusions were unambiguous: “We discovered no evidence suggesting that the miscount was the result of any fraudulent or willfully wrongful conduct by any Monmouth County election official or personnel, ES&S employee, or any other person. Nor did we discover evidence of double counting of ballots on ES&S’s elections system before or after the November 2022 Election in Monmouth County, or elsewhere.”

The report also contained policy recommendations for future elections, including that counties running ES&S software should generate and review audit reports after every election, that vendors should ensure those reports are easy to read, and that L&A testing should include scenarios where flash drives are loaded multiple times.

16.9 — The Vulnerability Beyond Monmouth

The Monmouth County error was caused by a specific chain of failures: the wrong person was sent to perform a software reinstallation, was given inadequate instructions, and the result was not verified. ES&S checked its help desk logs for every jurisdiction using the software version that required a separate ECO installation and found no other instances where the software had been uninstalled and reinstalled. The AG investigation accepted this finding.

But the underlying vulnerability was not specific to Monmouth County. It was architectural.

In the software version Monmouth was using, double-counting protection was a bolt-on patch — the ECO — installed separately from the core election software. Remove the core software, and the patch disappears with it. Reinstall the core software without remembering to reinstall the patch, and the safeguard is gone. The fact that it could be accidentally disabled by a single person following phone instructions from a colleague in the wrong department illustrates how fragile the protection was.

ES&S has since addressed this in its newer software version, where the ECO is built into the core product rather than applied as a separate patch. Monmouth County upgraded to this version in March 2023. But the older version was in use during the November 2022 election, and the question of how many other jurisdictions were running it — and whether any of them had experienced their own unverified reinstallations for other reasons — was answered only by ES&S’s internal review of its own help desk logs.

The ExpressVote XL machines used in Monmouth County are deployed in a handful of jurisdictions on the East Coast. But the standard ES&S ExpressVote — which works similarly and runs on the same Electionware platform — is used across the country. The underlying vulnerability was in the software, not the hardware. Any jurisdiction running the older Electionware version without the ECO would have been susceptible to the same failure mode: flash drives loaded twice, no error message, votes counted twice.

ES&S is the nation’s largest voting machine manufacturer. Its equipment is used in approximately half of U.S. counties. Seven New Jersey counties used ES&S machines in 2022. After the Monmouth story broke, five other New Jersey counties using ES&S equipment checked their own results by comparing turnout numbers to vote totals. All reported no discrepancies. But this was a voluntary, ad hoc response — not a systematic audit.

16.10 — What Actual Vote Would Have Contributed

The Monmouth County error is a reporting-layer failure. The voting machines counted correctly. The flash drives contained correct vote totals. The error was introduced when those totals were loaded into the election management system a second time, inflating the official results above the actual vote counts.

An Actual Vote comparison of poll tape totals against officially reported results would have flagged this immediately. In every affected municipality, the official results were higher than the actual vote totals printed on the tapes. In Belmar, where Bean found 311 more votes counted than people who voted, the discrepancy between tape totals and official results would have been unmistakable.

New Jersey does not require poll tapes to be posted publicly on election night, which means the standard AV workflow — a volunteer photographs the tape at the polling place — is not available in the same way it is in states like Georgia or Florida. But AV users in New Jersey can make an appointment with the Superintendent of Elections to record tapes, or submit an OPRA request to obtain them. This means AV could have caught the Monmouth error — not on election night, but within the days or weeks it takes to fulfill such a request. That would still have been far faster than the actual timeline, in which the error was not fully identified until January 2023, more than two months after the election.

The comparison between these access models is itself instructive. In states where tapes are posted on election night, AV can detect reporting-layer errors within days. In states where access requires an appointment or records request, AV can still detect them, but with more friction and delay. In states with no meaningful tape access, AV cannot operate at all. The Monmouth case illustrates what happens at each level of accessibility: Bean essentially performed a version of the AV comparison — checking reported vote totals against the number of people who actually voted — through an OPRA request. He found the error. He was dismissed. It took two more months for anyone to act on what he had found.

The broader point is that election transparency and accuracy benefit from easy access to poll tapes. Requiring tapes to be posted publicly on election night — as some states do — creates the conditions for immediate independent verification. Making access dependent on appointments or records requests preserves the possibility of verification but introduces delays and friction that allow errors to persist. The difference between these models is not abstract. In Monmouth County, it was the difference between catching a flipped election result in days and catching it in months.

What AV Would Not Have Done

AV would not have identified the root cause of the discrepancy — the missing ECO, the July reinstallation, the chain of vendor failures. AV detects discrepancies; it does not diagnose their origins. But by flagging the discrepancy immediately, AV would have triggered the investigation that ultimately identified the root cause — and it would have triggered it months earlier, before the wrong candidate had taken office and cast votes.

16.11 — What We Know and What We Don’t

What we know:

Six flash drives were loaded twice into Monmouth County’s election management system on election night and the following day. The software patch designed to prevent this had been accidentally removed during an unrelated repair four months earlier. The double-counting inflated vote totals in four municipalities and changed the outcome of one race — the Ocean Township Board of Education election, where the wrong candidate was certified as the winner and served for approximately two months before a court-ordered recount corrected the result.

The official post-election audit did not detect the error. Internal software reports showed the double loads on November 9, but no one reviewed them. A citizen who identified phantom votes through a records request was dismissed. The error was finally identified when the County Clerk escalated to ES&S in January 2023. A seven-month investigation by a former state attorney general found no evidence of fraud or intentional misconduct by anyone.

What we don’t know:

Whether the Logic and Accuracy testing gap — the failure to test for duplicate flash drive loading — has been addressed in other jurisdictions using ES&S equipment nationwide. Whether other counties running the older Electionware version ever experienced unreported software reinstallations that could have disabled their ECO protections. And whether the internal audit reports that ES&S’s software generates — reports that contained the evidence of this error — are routinely reviewed by election officials in the thousands of jurisdictions that use ES&S equipment across the country.

16.12 — Why It Matters

This case changed an election outcome. That alone sets it apart from most documented reporting-layer errors, where the discrepancies are real but the margins are large enough that the wrong candidate is not seated. In Ocean Township, the wrong candidate took office, attended meetings, and cast deciding votes on public business. The democratic process produced one result; the reporting system produced a different one; and for two months, the reporting system’s result governed.

Every official safeguard failed. The pre-election testing did not cover the failure mode. The post-election audit reported 100% accuracy. The software’s own internal reports contained the evidence and no one was told to look at them. A citizen who noticed the problem was brushed off for weeks. The error was caught not through any systematic check but through the persistence of one former councilman, the eventual attentiveness of the County Clerk, and the forensic capabilities of the vendor that had caused the problem in the first place.

The Harvey investigation was thorough, transparent, and produced clear recommendations. It is a model of how to respond to a reporting-layer failure after the fact. But it would not have been necessary if an independent check had existed at the front end — a routine comparison of poll tape totals against officially reported results, performed shortly after the election, by people outside the official chain of custody.

That is what Actual Vote provides. And the Monmouth case illustrates why the ease of access to poll tapes matters. In states where tapes are posted publicly on election night, the comparison can happen within days. In New Jersey, where access requires an appointment or a records request, the same comparison is possible but slower. The slower it is, the longer errors persist — and in this case, the longer the wrong person held public office. The strongest version of election transparency is the one where the primary evidence — the poll tapes — is available to everyone, immediately, without requiring anyone’s permission.

16.13 — Further Reading