IN THE MATTER OF THE PETITION OF CELESTE FERNANDEZ, ETC. (L-3848-20, ATLANTIC COUNTY AND STATEWIDE)

NOT FOR PUBLICATION WITHOUT THE APPROVAL OF THE APPELLATE DIVISION

SUPERIOR COURT OF NEW JERSEY APPELLATE DIVISION DOCKET NO. A-1139-20

IN THE MATTER OF THE PETITION OF CELESTE FERNANDEZ FOR A RECOUNT OF THE VOTES CAST AT THE ELECTION HAVING APPROVED FOR PUBLICATION OCCURRED ON NOVEMBER 3, June 29, 2021 2020 FOR THE 2020 ATLANTIC APPELLATE DIVISION COUNTY COMMISSIONER ELECTION, ATLANTIC COUNTY, NEW JERSEY, and CELESTE FERNANDEZ,

Petitioner-Appellant. ______________________________

Submitted June 8, 2021 – Decided June 29, 2021

Before Judges Yannotti, Haas, and Mawla.

On appeal from the Superior Court of New Jersey, Law Division, Atlantic County, Docket No. L-3848-20.

Herman Law Offices, LLC, attorneys for appellant Celeste Fernandez (Robert D. Herman, on the briefs).

Cooper Levenson, PA, attorneys for respondent John W. Risley, Jr. (Randolph C. Lafferty and Rebecca C. Lafferty, on the brief).

Gurbir S. Grewal, Attorney General, attorney for amicus curiae Attorney General of New Jersey (Melissa H. Raksa, Assistant Attorney General, of counsel;

Nicole E. Adams, Deputy Attorney General, on the brief).

The opinion of the court was delivered by

YANNOTTI, P.J.A.D.

Celeste Fernandez appeals from an order entered by the Law Division on December 22, 2020, which accepted the results of the limited recount and recheck of the election results for one of two positions of Atlantic County Commissioner at Large (CAL) and denied her request for a more expansive recount.1 We reverse.

I.

We briefly summarize the pertinent facts and procedural history. On November 3, 2020, there were four candidates in the race for Atlantic County CAL: two Democrats and two Republicans. Each voter was permitted to vote for up to two candidates, and the two candidates who received the most votes would be declared winners.

The Clerk released the following unofficial election results for the CAL race, as canvassed by the Atlantic County Board of Elections (Board): Caren

1 In the November 2020 election, the position was listed on the ballot as "Freeholder." However, after the election the office was renamed "County Commissioner." See L. 2020, c. 67 (eff. Jan. 1, 2021) (amending N.J.S.A. 1:1- 2, N.J.S.A. 40:20-1). A-1139-20

Fitzpatrick and John W. Risley, Jr. placed first and second, respectively, with Fitzpatrick receiving 67,600 votes, and Risley 66,427. Fernandez placed third with 66,046 votes, and James Toto came in fourth with 64,586 votes. The difference between Risley and Fernandez was 381 votes out of a total of 132,473 votes cast for either candidate, or a margin of .3%.

Although Atlantic County voters cast just over 143,000 ballots in the November 2020 election, not all ballots included votes for the CAL race, and some ballots contained only one vote for that position. A ballot with one or no votes was labeled an "undervote."

A ballot in which only one CAL candidate was selected counts as one undervote, whereas a ballot in which no candidate for that office was selected counts as two undervotes. The Board's initial canvass recorded 21,263 undervotes for the CAL race.

Unlike an undervote, an "overvote" is a ballot in which the voter selects more than the permitted number of candidates. This results in a rejected vote, recording zero votes for all candidates. The Board's initial canvassing recorded 740 overvotes for the CAL race.

On November 20, 2020, Fernandez filed a petition in the Law Division seeking a recount. Thereafter, at the direction of the New Jersey Secretary of

A-1139-20

State, the Board conducted a hand-to-eye audit of voter-verifiable paper records for several races, including the CAL race. The Board conducted the audit in accordance with procedures established by the State's Division of Elections (DOE).

On December 1, 2020, Fernandez filed an amended petition in the trial court, and the following day, the Law Division judge conducted an evidentiary hearing on the petition. The judge heard testimony from Evelyn Caterson, the Board's Chairperson, and from Susan Sandman and William Sacchinelli, the Board's Republican and Democratic clerks, respectively.

Sandman testified that the State-mandated audit process consisted of a hand-to-eye recount of 767 randomly selected audit units, or "batches" of 200 ballots. Sacchinelli testified that the Board was still reviewing the form of the audit report for accuracy and would release the results within a couple of days.

Caterson testified that the audit, which was conducted by twelve workers, took about five-and-a-half hours, not including the subsequent Board review. She was unable to say how long it would take for a hand recount of all 143,000 ballots cast in the election.

Caterson stated, however, that such a recount would require the Board to hire forty staff members and its supervisors would have to perform additional

A-1139-20

tasks. She estimated that the cost of hiring additional staff for a two-week period would be "close to $90,000."

Caterson added that the hand recount of 143,000 ballots could take "several weeks" without additional staff, and that a hand recount of fewer ballots would take less time. She did not offer any testimony regarding how long it would take, how expensive it would be, or how many workers would be required for a limited hand recount of only the ballots registering overvotes and undervotes after using the machine to identify those ballots.

The judge granted the petition for a recount. The court scheduled the matter for another hearing on the size and scope of the recount.

On December 4, 2020, the Board released the results of the State-mandated audit. Risley's margin of victory over Fernandez was reduced by one vote to 380.

The court conducted another hearing on December 7, 2020. Sandman testified that the machine scanner the Board used was capable of "identify[ing] specific ballots which contain overvotes . . . [and] undervotes." She stated that because of the decrease in overvotes and undervotes tallied in the audit, the total votes cast for the CAL position in the batches of ballots recounted increased

A-1139-20

from 5,158 to 5,184, which, according to Sandman's calculation, resulted in an "error rate" of about .5%.

Sacchinelli testified that the entire audit process, including the review by Board staff, had taken between seven and eight hours. He said that, in general, the machine scanner could read approximately 3,000 ballots per hour at maximum capacity. He explained that after the scanner identified ballots as overvotes or undervotes, the staff could then separate those ballots from the other ballots in the batch for a hand recount.

Sacchinelli further explained that although a staff member would have to separate the overvote and undervote ballots physically from the rest of the batch, "the scanner would pick up [the ballots that included] undervotes and overvotes" to facilitate separation of the ballots. He added that with the vendor's assistance, the machine could segregate the ballots that contained undervotes or overvotes from the rest of the batch.

At the December 7 hearing, the judge asked Fernandez's attorney whether Fernandez was seeking a hand recount of all 143,000 ballots cast in the CAL election. Counsel replied:

No, Judge. Given the nature and circumstance, I mean during a normal election year where we utilize the machine to a greater level, then the answer is yes. But this year . . . we’re looking at error rates. In particular,

A-1139-20

we're looking at the overvotes and the undervotes because that seems to be where you have a [fifty percent] error rate and then the other one somewhere between, you know, [an error rate of] .75 to 1.5 or maybe higher, we're not even sure . . . . It seems that the better [approach is to have] a hand count for the undervotes and the overvotes and a machine total for the remainder, Judge.

Counsel added that Fernandez was requesting that the Board run all the ballots through the vote-tallying machines, have the machine separate the overvotes and undervotes from the rest, and have a hand recount of the overvotes and undervotes.

On December 11, 2020, the judge ordered the Board to conduct a further audit of two additional two percent batches of ballots for the CAL race, using the same procedures that the Board used in the State-mandated audit. On December 15, 2020, Fernandez filed a motion seeking reconsideration of the December 11, 2020 order.

Fernandez asked the court "to expand the recount and recheck to include the machine rescanning of 143,000 ballots cast and the hand counting of all ballots containing undervotes and overvotes." On December 16, 2020, the court denied Fernandez's motion for reconsideration "at this time."

On December 18, 2020, the Board released the results of the recount, which included 5,855 randomly selected ballots in two separate two percent

A-1139-20

batches. Before the recount, these ballots had contained a total of 10,825 recorded votes, 38 overvotes, and 839 undervotes. After the recount, the total amount of overvotes had decreased by eighteen and the total number of undervotes had decreased by forty-four, with most of these newly recorded candidate votes going to the two candidates who were not parties to the lawsuit.

In the first batch of hand-counted ballots in the recount, both Fernandez's and Risley's vote tallies increased by five, the number of recorded overvotes decreased by five, from twelve to seven, and the number of undervotes decreased by twenty-three, from 446 to 423. In the second batch of hand-counted ballots, both Fernandez's and Risley's vote tallies increased by eight, the number of overvotes was cut in half, from twenty-six to thirteen, and the number of undervotes decreased by twenty-one, from 393 to 372.

In all, combining the 8,646 ballots and 15,983 votes hand-counted in the three batches included in the audit and the recount, the total number of overvotes had decreased by 24, from 50 to 26, and the number of undervotes had decreased by 50, from 1,247 to 1,197. Combined, 1,297 votes originally tallied as either overvotes or undervotes had been recounted, and 74 had been changed from undervotes or overvotes to recorded votes for candidates.

A-1139-20

On December 22, 2020, the trial court entered the order, which denied Fernandez's motion for reconsideration and accepted "as sufficient the results of the [r]ecount and [r]echeck completed." The court ordered the Clerk to "adjust and recertify the final results of the 2020 Atlantic County [CAL] Election in accordance with the results of the audit and additional recount." This appeal followed.

On February 18, 2021, the Clerk, the Board, and the County Superintendent of Elections advised this court they would not be participating in the appeal. Thereafter, we granted the Attorney General's motion to appear as amicus curiae.

II.

On appeal, Fernandez argues that under N.J.S.A. 19:28-1, an unsuccessful candidate is entitled to a full or partial recount of the votes cast in an election, and the trial court erred by interpreting the statute to give the court discretion in determining whether to order a recount. Risley contends the statute does not mandate a recount on request. The Attorney General agrees and argues that N.J.S.A. 19:28-1 provides the court with discretion to determine if a recount is required.

A-1139-20

Issues raised regarding the interpretation of statutes are questions of law which we review de novo. Kocanowski v. Twp. of Bridgewater, 237 N.J. 3, 9 (2019) (citing State v. Fuqua, 234 N.J. 583, 591 (2018)). Therefore, the trial court's "interpretation of the law and the legal consequences that flow from established facts are not entitled to any special deference." Manalapan Realty, L.P. v. Twp. Comm. of Twp. of Manalapan, 140 N.J. 366, 378 (1995).

"The overriding goal" of statutory interpretation "is to determine . . . the intent of the Legislature, and to give effect to that intent." State v. Hudson, 209 N.J. 513, 529 (2012). "The inquiry thus begins with the language of the statute, and the words chosen by the Legislature should be accorded their ordinary and accustomed meaning." Ibid. Courts should "apply to the statutory terms the generally accepted meaning of the words used by the Legislature," Patel v. N.J. Motor Vehicle Comm'n, 200 N.J. 413, 418 (2009), "read . . . in context with related provisions so as to give sense to the legislation as a whole." DiProspero v. Penn, 183 N.J. 477, 492 (2005).

Section one of the recount statute provides in part that:

When any candidate at any election shall have reason to believe that an error has been made in counting the votes of that election, the candidate may, within a period of [seventeen] days following such election, apply to a judge of the Superior Court assigned to the county wherein such district or districts are located, for

A-1139-20

a recount of the votes cast at the election in any district or districts.

[N.J.S.A. 19:28-1.]

Section two of the recount statute, governing costs, requires, in pertinent part, that the recount applicant must place a sum "not exceeding twenty-five dollars" per district for which a recount is sought with the court, "as security for the payment of the costs and expenses of the recount in case the original count be confirmed." N.J.S.A. 19:28-2. Section three of the recount statute states that a "judge shall be authorized to order upon such terms as he deems proper a recount of the votes as he may determine, to be publicly made under his direction by the county board." N.J.S.A 19:28-3.

A separate statute, which was enacted in 2005, creates a vote-auditing process, pursuant to which the Attorney General is required to appoint a "professional audit team" to "oversee, in each county, random hand-to-eye counts of the voter-verifiable paper records that are to be conducted by appropriate county election officials . . . for each election held for . . . county and municipal offices selected by the Attorney General." N.J.S.A. 19:61-9(a). The statute requires that, "the audit shall be conducted in at least two percent of the election districts in which each audited election appears on the ballot." Ibid. The statute further provides:

A-1139-20

Nothing in this section shall be construed to prevent a candidate or other applicant from requesting a recount pursuant to R.S.19:28-1 et seq. or any other law. In the event that such a recount is held in any election district that has been audited pursuant to this section, the official result from such election district shall be applied to the recount in lieu of conducting a subsequent hand count of the audited election district unless a court, at the request of a candidate or other applicant who requested the recount, so orders.

[N.J.S.A. 19:61-9 (d).]

Furthermore, on August 14, 2020, Governor Murphy issued Executive Order 177, titled "An Order to Protect Public Health By Mailing Every Active Registered Voter a VBM [Vote-By-Mail] Ballot Ahead of the General Election." Exec. Order. No. 177 (Aug. 14, 2020), 52 N.J.R. 1701(b). Among other things, the order directed that, "[t]he November General Election shall be conducted primarily via vote-by-mail ballots, which will be sent to all 'Active' registered voters without the need for an application to receive a vote-by-mail ballot." Ibid.

Thereafter, on August 28, 2020, the Legislature enacted three separate election-related laws: a law designed "to modify and establish various voting procedures," for the 2020 election and beyond, including provisions relating to the "curing" of mail-in ballots (The Ballot Cure Act), L. 2020, c. 70; and two laws that codified several of the vote-by-mail procedures for the 2020 election directed by Executive Order No. 177, and amended various other statutory

A-1139-20

provisions, L. 2020, c. 71; L. 2020, c. 72 (collectively, the 2020 election statutes).

The 2020 election statutes provided that the Legislature did not intend to disturb the existing scheme for election laws in New Jersey, except as otherwise provided, stating that "[t]he November 2020 General Election shall be conducted in accordance with Title 19 except as set forth below." N.J.S.A. 19:63-31(a). Among the new election procedures implemented were a codification of the directive in Executive Order No. 177 that the election would be "conducted primarily via vote-by-mail ballots," N.J.S.A. 19:63-31(a), as well as the following:

[T]o account for the increase in vote-by-mail ballots and to ensure that registered voters' efforts to vote are not impacted by delays in the postal service, every vote- by-mail ballot that is postmarked on or before November 3, 2020, and that is received by November 10, 2020, at 8:00 p.m. shall be considered valid and shall be canvassed, assuming the ballot meets all other statutory requirements. . . .

[N.J.S.A. 19:63-31(m).]

Prior to the election, the Secretary of State determined that the Commissioner elections in Atlantic County would be among those subject to a mandatory hand-to-eye audit pursuant to N.J.S.A. 19:61-9. The State DOE issued procedures for the audit, which provided in part that each county board

A-1139-20

of elections, shall "conduct a hand-to eye manual count" in "at least [two per cent] of the election districts in which each audited election contest appears on the ballot . . . of the voter-verifiable paper records and compare those results to the results produced by the voting machine."

Here, the trial court found that there is no automatic right to a recount under N.J.S.A. 19:28-1. The court noted that under New Jersey law, there is no objective measure that would trigger a recount. 2 The court therefore determined that to secure a recount, the candidate must establish that the board of elections erred in its tabulation of the votes, or that the results of the election were sufficiently close that denial of a request for a recount would undermine the integrity of the electoral process.

The trial court noted that the Board had initially reported a 381-vote difference between Risley and Fernandez. The court observed that this difference was "extremely close," equaling a twenty-nine-vote difference for

2 The Attorney General notes that other states require recounts of ballots cast if the returns show a margin of victory that is less that a specified number or percentage. See Del. Code. Title 15, § 5702(e) (requiring recount if number of votes separating candidates for certain offices is less than 1,000 votes or one- half of one percent of all votes for those candidates); 25 Pa. Stat. § 3154(g)(1)(i) (requiring a recount if candidate for public office was defeated by one-half of one percent or less of the votes cast for the office). A-1139-20

every 10,000 votes cast. The court found that based on the closeness of the results, plaintiff was entitled to a recount.

We are convinced that the trial court correctly found that N.J.S.A. 19:28-1 does not provide a candidate with an automatic right to a recount. Moreover, the statute does not expressly authorize a court to order a recount due to the closeness of the reported results. As noted, the statute provides that a candidate may apply for a recount if he or she "shall have reason to believe that an error has been made in counting the votes of that election . . . ." Ibid.

Therefore, to secure a recount pursuant to N.J.S.A. 19:28-1, a candidate must present sufficient competent, credible evidence showing there is reason to believe an error was made in the tabulation of the votes. The court then must determine whether the claimed error warrants a recount. In making that decision, the court must consider whether the claimed error has the capacity to affect the outcome of the election.

In her initial petition, Fernandez asserted that, upon information and belief, there were reasons to believe an error was made in counting the votes in the CAL election. She asserted that certain machine, vote-by-mail ballots, provisional ballots, and emergency ballots had not been counted properly. She claimed that if the ballots had been properly counted, she would attain the votes

A-1139-20

needed for her election to the CAL position. Fernandez provided no certifications or affidavits to support these claims.

Fernandez repeated these same allegations in her amended petition. She further alleged that, upon information and belief, errors had been revealed in the State-mandated audit. She contended there was an "unusual" number of overvotes and undervotes in the election. She did not submit any affidavits or certifications to substantiate her claims.

However, the results of the State-mandated audit provided sufficient support for Fernandez's request for a recount. The audit encompassed a hand count of 2,791 randomly selected ballots, and only changed the difference between Fernandez and Risley by one vote, with an overall 99.45% accuracy rate. Even so, the audit reduced the number of overvotes in the batch audited from twelve to six, and the number of overvotes from 408 to 402.

Thus, the audit showed an error rate of 50% for overvotes and about 1.5% for undervotes in the batch. It is reasonable to assume, for purposes of determining the scope of the recount, that the same error rates would apply to all the initially reported overvotes and undervotes. Based on this analysis, the number of erroneously recorded overvotes and undervotes could be sufficient to affect the outcome of the CAL race between Risley and Fernandez.

A-1139-20

We therefore conclude that under N.J.S.A. 19:28-1, there is no automatic right to a recount and a recount is not required merely because the reported results of the election are close. The candidate seeking a recount under N.J.S.A. 19:28-1 must present the court with sufficient competent, credible evidence showing there is reason to believe there was an error in the count. If the claimed error could alter the results of the election, the court should order a recount.

Here, Fernandez presented the trial court with sufficient evidence to show there is reason to believe there was an error in the count of overvotes and undervotes. Moreover, the record shows that the claimed error could affect the outcome of the CAL race between Risley and Fernandez. Therefore, the court did not err by ordering a recount.

III.

Fernandez argues that that the trial court mistakenly exercised its discretion by denying her application for a hand recount of the overvotes and undervotes in the CAL election. In response, Risley contends that trial court properly exercised its discretion to order a limited audit of two additional two percent batches of votes. The Attorney General takes no position on this issue.

"A trial court's exercise of discretion is 'entitled to respectful review under an abuse of discretion standard[.]'" Thabo v. Z Transp., 452 N.J. Super. 359,

A-1139-20

368 (App. Div. 2017) (quoting Serenity Contracting v. Fort Lee, 306 N.J. Super. 151, 157 (App. Div. 1997)). Typically, an abuse of discretion will only "arise[ ] when a decision is 'made without a rational explanation, inexplicably depart[s] from established policies, or rest[s] on an impermissible basis.'" Flagg v. Essex Cnty. Prosecutor, 171 N.J. 561, 571 (2002) (quoting Achacoso-Sanchez v. Immigr. and Naturalization Serv., 779 F.2d 1260, 1265 (7th Cir.1985)).

Although this standard of review "defies precise definition," the "functional approach . . . examines whether there are good reasons for an appellate court to defer to the particular decision." Ibid. Reviewing courts should "accord substantial deference to the trial court's findings of fact provided that they are 'supported by adequate, substantial and credible evidence[,]' and also give deference to the trial court's conclusions and 'discretionary determinations that flow from them.'" Pitney Bowes Bank, Inc. v. ABC Caging Fulfillment, 440 N.J. Super. 378, 382-83 (App. Div. 2015) (alteration in original) (quoting Cosme v. Borough of E. Newark Twp. Comm., 304 N.J. Super. 191, 202 (App.Div.1997)).

Here, the trial court found that "there [wa]s no justification to order a full recount of all ballots," due to "the significant burden" that a hand recount of all 143,000 ballots received would "impose[ ] on the Board." The court reasoned

A-1139-20

that because the State-mandated audit had resulted in "an error rate of only about a half of a percentage point," a full hand recount was not warranted.

As stated previously, the court instead ordered a randomly selected hand recount of an additional four percent of the ballots so that the combination of the State-mandated audit and the court-ordered recheck would result in a hand recount of six percent or "about [8,100] ballots." The court stated that this would "provide even more confidence in the accuracy of the results consistent with legislative intent."

When it ordered the recount, the court did not address Fernandez's specific request to have a machine scan and identify the ballots marked as overvotes or undervotes and then conduct a hand recount of those ballots only. The court found that it would be an "immense task" for "the Board to rescan all 143,000 ballots and then identify and separate the ballots with undervotes and overvotes for a hand recount of the ballots containing undervotes and overvotes." The court noted that the State-mandated audit had revealed only "minor inaccuracies," and had only reduced Risley's margin of victory by one vote.

Following the recount of the additional four percent of the ballots, the court issued another order denying Fernandez's request for an expanded hand recount. In the written decision accompanying the order, the court again stated

A-1139-20

that the audit had "provide[d] a reliable method for recounting the ballots," and had resulted in "a reliability rate of over 99.498% accuracy based on the testimony presented."

The court found that because the additional recount "did not change the vote difference between [Risley] and [Fernandez] . . . the audit and the additional recount provided enhanced reliability and confidence in the election results in accordance with the statutory scheme." Due to the "greater confidence in the election results" that resulted from the court-ordered recount, the court found no justification for "ordering a further recount."

On appeal, Fernandez contends that the trial court erred by declining to order the hand recount she had requested because, following the audit and the limited recount, the margin of victory had narrowed between Fernandez and Risley,3 and there was an error rate of forty-eight percent on the overvotes already recounted, and four percent on the undervotes. She contends that applying those error rates to the unrecounted ballots would result in 331

3 As discussed, the court found the margin between Risley and Fernandez, after the court-ordered recount, remained at 380 votes, the same margin that separated the candidates after the initial State-mandated audit. Fernandez claims, without explanation, that the gap is now 377 votes. However, the margin of 380 is consistent with the results of the audit and partial recount. A-1139-20

additional countable votes which had been mischaracterized as overvotes and 800 votes that were mischaracterized as undervotes.

Fernandez asserts that the trial court erred by focusing on whether a review of the overvotes and undervotes would weigh overwhelmingly in her favor and change the result. She maintains the court should instead have only considered whether the uncounted overvotes and undervotes were "sufficient in number to effect an outcome" or "raise concerns as to the certainty of the vote total."

As noted, the court's decision to limit the recount to a second audit was based in part on its finding that a hand recount of the overvotes and undervotes would pose an "immense" or "significant burden" on the Board. It appears that in making that finding, the court relied on Caterson's testimony that a full hand recount of all 143,000 ballots could take several weeks and could require the Board to devote extensive resources to the effort, at considerable expense.

However, the record shows that Fernandez was not seeking a full hand recount of 143,000 ballots. As stated previously, Fernandez's attorney told the court that Fernandez was asking to have all ballots run through the machine scanner again so that the overvotes and undervotes could be identified and

A-1139-20

separated. The Board's staff then could conduct a hand recount of the overvotes and undervotes only.

When the court determined the scope of the recount, there remained 20,855 unrecounted undervotes and 728 unrecounted overvotes, or 21,583 combined. In requesting the ballots containing those rejected votes be recounted, counsel referenced Sacchinelli's testimony regarding the Board's ability to bring in the machine vendor to better facilitate the separating of overvotes and undervotes from the remainder of the ballots. There was no testimony from Caterson, Sandman, or Sacchinelli indicating that such a process would present the Board with a logistical or financial burden.

The hand recount of 5,585 ballots, and 11,702 votes, as ordered by the court, included only 877 votes that Fernandez had asked to be recounted by hand (the overvotes and undervotes in the two batches recounted) and 10,825 votes that Fernandez had not asked to be recounted by hand (the remainder). The hand recount of more than 11,000 votes had been completed within one week of the court's order. Accordingly, the court's finding that a hand recount of 21,583 votes would pose a significant and "immense" burden on the Board was not supported by sufficient credible evidence in the record.

A-1139-20

Moreover, the trial court's apparent misapprehension that Fernandez was seeking a full hand recount seems to have led the court to conclude the "error rate" in the machine's tabulation of overvotes and undervotes was .5%. The court based this on Sandman's testimony that the "error rate" of the aud it was .5%, and the results 99.498% accurate.

The court relied upon the .5% error rate for its conclusion that Fernandez's request for a machine recount of every vote was unnecessary. But that error rate was irrelevant to Fernandez's application which was for machine scanning of all votes cast in the CAL election and a hand recount of the ballots identified as having overvotes and undervotes. Fernandez's application was premised on the claim that the machine scanner had erroneously discarded valid votes, which had been mistakenly recorded as overvotes and undervotes.

In determining the "error rate," or amount of false positive overvotes in the CAL election, the numerator should be the number of votes amended from uncounted overvotes to recorded candidate votes either in the audit (6), the recount (18), or both combined (24). Similarly, for the error rate of the undervotes in the election, the numerator should be the number of votes amended from uncounted undervotes to recorded candidate votes following t he audit (6), the recount (44), or both combined (50).

A-1139-20

Furthermore, the denominator for determining the error rate of overvotes should be the number of votes originally recorded as overvotes in the sample audited (12), recounted (38), or both combined (50). Similarly, the correct denominator for the error rate of undervotes was the number of votes originally recorded as undervotes in the sample audited (408), recounted (839), or both combined (1,247).

Considered together, as overall false positives, the numerator for overvotes and undervotes from the audit and recount combined would be 74 and the denominator 1,297. This equation yields an overvote false positive rate of 50% in the audit, 47.37% in the recount, or 48% overall; an undervote false positive rate of 1.47% in the audit, 5.24% in the recount, or 4.01% overall; and a combined false positive rate of 5.71% overall.

Applying the appropriate cumulative error rates to the 690 unrecounted overvotes and the 20,016 unrecounted undervotes, a hand recount of those ballots would be expected to result in approximately 331 additional overvotes, 1,143 undervotes, and 1,474 combined additional votes.

We note that there can be no certainty as to how these votes would likely be distributed among the four candidates for the two CAL positions, and even less certainty as to the likelihood that counting these votes would diminish

A-1139-20

Risley's margin of victory over Fernandez. However, no testimony or other evidence was presented to the trial court as to whether the 380-vote gap between Risley and Fernandez could be overcome if an additional 1,474 votes were considered.

We conclude that the trial court erred by finding that the results of the State-mandated audit and the court's initial limited recount showed the election results for the CAL positions were sufficiently reliable and obviated the need for a hand count of the overvotes and undervotes. Rather, the results of the audit and initial recount established there was reason to believe an error had been made in the tabulation of the overvotes and undervotes, and the error could affect the outcome of the contest between Risley and Fernandez for the CAL position.

We therefore conclude the court erred by denying Fernandez's application for a machine recount to identify and segregate the overvotes and undervotes in the CAL election, and a hand recount of the overvotes and undervotes in the election. We remand the matter to the trial court for the issuance of an order requiring such a recount.

Reversed and remanded to the trial court for further proceedings in accordance with this opinion. We do not retain jurisdiction.

A-1139-20