The auditing processes in AZ, GA, MI, NC, PA, and WI
Hi everyone, and a happy early Thanksgiving!
While most vote counting has wrapped up across our six pilot states, they are still undergoing the canvassing process, of which the audit is a key step.
Audits are a routine post-election process to check the accuracy of election results. There are two main types: risk-limiting audits (“RLAs”) and fixed-sized post-election audits.
In fixed-sized post-election audits, election officials select a particular number of units (either a fixed number or percentage) to audit at random: precincts, voting machines, polling locations, etc. The results of that unit are then reviewed. [Verified Voting, “Coordinating Audits and Recounts to Strengthen Election Verification” at 4 & 5, 11/18/22]
A risk-limiting audit (RLA) looks at the margin of victory and total number of votes cast to determine the number of ballots to be audited, and is considered the “gold standard” of post-election audits. RLAs are particularly useful in close contests because they ensure more ballots will be reviewed in those races, increasing the chance of detecting any errors. [Brennan Center for Justice, “A Smart and Effective Way to Safeguard Elections” 11/18/22]
Here are the auditing processes in AZ, GA, MI, NC, PA, and WI.
Arizona 🌵
Arizona conducts a fixed-sized post-election hand count audit of two percent or two precincts, whichever is greater. If the margin of the initial hand count audit is below a certain statutory threshold, the scope of the hand count audit is expanded. While uncommon, this process can be repeated several times until the audit may reach the entire jurisdiction.
Audit type: Fixed-sized post-election hand count audit
Geographic scope: Performed by counties, largely statewide (in counties where the local parties have designated sufficient people to conduct the hand counts)
Occurrence: Between 24 hours after poll close (Wednesday, November 9) and before county canvassing, which starts on Monday, November 14. Typically counties conduct the hand count audit by Sunday, November 13 (the end of the first weekend following the election).
Description of audit process
Designate hand count board members
At least 14 days before an election, the Officer in Charge of Elections (OCE) estimates the number of hand count board members needed and notifies the county party chairpersons to designate hand count board members. [2019 EPM at 213]
At least seven days before an election, the county party chairpersons designate hand count board members and alternates. The hand count will be canceled if there are not enough board members. [2019 EPM at 213, 214]
Select precincts/polling places and types and quantities of ballots to hand count
The OCE must hand count regular (not provisional or conditional) ballots from at least two percent of the precincts, or two precincts, whichever is greater. [A.R.S. § 16-602(B)(1) & 2019 EPM at 215]
After all ballots voted in the precinct polling places have been delivered to the central counting facility, the county party chairpersons take turns randomly drawing precincts, vote centers, or consolidated polling places for the hand count.
Select contests to be hand counted: The county party chairpersons randomly select the races.
Procedures for conducting the hand count
The OCE ensures that each board is properly assembled and has the necessary forms and supplies, and then apportions ballots to each board.
The inspector or one of the judges fills out and signs a hand audit tally sheet for each race.
For optical/digital scan ballots, the hand count board separates ballots with write-in votes from the other ballots and hand counts each group of ballots following the "stacking method."
The hand count boards list the totals for each candidate on the appropriate hand audit tally sheet and compare the hand count results for the batch with the machine count results. The auditor initials the tally sheet to indicate that there are no discrepancies.
The OCE creates a Precinct Hand Count Margin Worksheet for each hand-counted race and calculates the hand count compared to the machine count margin in the presence of the county party chairpersons.
If the calculated margin is less than the margin designated by the Vote County Verification Committee (three votes or one percent, whichever is greater), then the hand count concludes and the results of the electronic tabulation is the official count for that race. The OCE indicates on the Hand Count Cumulative Sheet that the count is complete and two members, one from each political party present, will sign the sheet. [2019 EPM at 225-227]
If the margin is equal to or greater than the designated margin, a second precinct hand count of that race and ballots is performed. [2019 EPM at 227]
If the second precinct hand count results in a margin that is equal to or greater than the designated margin, the precinct hand count will be expanded to twice the original number of precincts/polling locations) (selected by the same method). [2019 EPM at 227, 228]
If the expanded precinct hand count results in a margin that is equal to or greater than the designated margin, the precinct hand count will be expanded to include the entire jurisdiction for that race.
The full hand count must be repeated for a particular race until the results are identical to the results of another full hand count for that race. [2019 EPM at 228]
Hand count of recounted elections
A precinct hand count audit must also be conducted following a court-ordered recount. The same procedures must be followed except that the OCE must conduct a hand count of at least five percent of precincts for the recounted race. [2019 EPM at 234]
Reporting results of the hand count
Within two business days of the completion of the precinct hand count and before the county canvass, the OCE must electronically submit a report to the Secretary of State that includes information and summary of the hand count. [2019 EPM at 234]
Sources: Arizona Elections Procedures Manual, 12/2019; Post Election Procedures, accessed 11/21/2022
Georgia 🍑
Georgia conducts an RLA of a single race selected by the Secretary of State each year. The selection is based on several criteria, as laid out in rules promulgated by the State Election Board, and can be found at Ga. Comp. R. & Regs. r. 183-1-15-.04(1)3. This year, the Secretary of State selected his own race to be audited. This audit is currently underway, with results expected this week.
Audit type: Risk-limiting audit, ballot-polling audit that has a statutorily limited risk of 10%.
Geographic scope: The RLA is performed by all 159 counties.
Occurrence: The risk-limiting audit kicked off on November 16, and counties began conducting the audit the next day. The Secretary of State’s target date for state certification of the November 8 General Election is Monday, November 21, 2022.
Guidance: The Secretary of State distributes instructions to all county election superintendents and coordinates with counties on managing and conducting the audit.
Portion of ballot audited: The Secretary of State chooses a single race to be audited; this year, the Secretary of State chose his own race.
Description of audit process
On November 16, twenty 10-sided dice were rolled to create a random seed number. The seed number is used to determine the random selection of batches to be audited.
County election officials began auditing selected batches on November 17. The date, time, and location of each county’s audit was posted to each county’s website, as required by law.
Every county audits at least one batch of ballots from Election Day or advanced voting, and another batch from absentee or provisional ballots. Each county will have audit teams that work in pairs to first determine the quantity of ballots in a batch, and the number of votes for each candidate.
Copies of the tally sheets of ballots, as well as the results of the audit, should be made available to the public this week.
The Secretary of State partnered with a company called VotingWorks to use their open-source program, called “Arlo,” to conduct the statistical calculations.
Sources: GPB, 11/17/2022; The Carter Center, 03/31/2021; Georgia Secretary of State, 11/15/2022; Bartow County, Board of Elections and Voter Registration, 11/18/2022
Michigan🧤
Michigan conducts a fixed-sized audit. Although it conducted a pilot RLA in 2020, it did not extend that pilot into 2022.
Audit type: Fixed-sized post-election audit of election procedures, materials, and results
Geographic scope: Statewide, conducted by auditors selected and reviewed by the county clerks, under the supervision of the Secretary of State.
Occurrence: Within 30 days after completion of canvass and no later than December 22 unless a recount has been ordered. Additionally, each county clerk who conducts an election audit will provide the results to the Secretary of State within 20 days after the audit.
Guidance: The Secretary of State is responsible for developing the audit program with detailed procedures and a list of which ballots and documents require review [MCLS § 168.31a]. The right to have the results of statewide elections audited is included in the State Constitution and was added in 2018 after Michigan voters approved a constitutional amendment.
Portion of ballot audited: The focus of the audit is the inspection of election procedures and materials used prior to the election and on election day. Materials included in the audit: election notices, election inspector appointments and training, e-pollbook security, test deck procedures, military and overseas voter applications, and the pollbook and ballot containers. Additionally, the results of up to three contests in a General Election and one contest in other elections in each precinct will be audited.
Description of audit process
Training of audit official
The Secretary of State and County Clerks conduct election audits. The Secretary of State trains and certifies County Clerks and their staffs for the purpose of conducting audits of precincts randomly selected by the Secretary of State. The Secretary of State supervises each County Clerk in the performance of election audits.
Precinct/contest selection
The Bureau of Elections (BOE) randomly selects precincts and contests for county and state audits the day after an election. The County Clerks then notify and coordinate the scheduling of the audit and appointment of auditors with their respective local jurisdictions.
Conducting the audit
The auditor inspects the following documents, materials and information:
Public notices (notice of registration, notice of election, public logic and accuracy test)
Weekend hours at the clerk's office prior to the election
Election Commission minutes relating to the appointment of precinct and receiving board inspectors
ePollbook's security features
Procedure manual for tabulators and voter assist terminals'
Logic and Accuracy testing "test deck"
Applications to vote for absentee, military, and overseas voters
Pollbooks
Provisional ballot forms
Ballot containers' seal and spoiled and original ballot envelopes
Conducting the voted ballot hand count audit
The final step is a review by hand tally of the voted ballots.
Two people count the number of the ballots to ensure that they match the number tabulated and/or the number of voters.
One person counts the ballots and places them in a stacks of 25 and separated into piles based on the votes cast in the contest being audited. This is done while the other person observes so that the ballots do not stick together and that the count is accurate.The audit of the results is not a recount and does not change any certified election results.
Once confirmed, the auditors complete a comprehensive worksheet where they will note any discrepancies. After the audit is concluded, the worksheet will be reviewed by the County Clerk. The worksheet data must be submitted electronically within two days of the audit.
Consequences of the audit
Discrepancies and deficiencies found as a result of the post-election audit are used by the Bureau of Elections to determine additional training needs at the municipal, county and state level.
Sources: 2020 Michigan Bureau of Elections' Post-Election Audit Manual, Michigan Election Code (MCLS § 168.31a) and Michigan Constitution (Mich. Const. 1963, art. II, §4(h)
North Carolina 🏀
Although North Carolina conducted a pilot RLA in 2021, the pilot did not continue into 2022. Its current audit process is more similar to a fixed-sized post-election audit, but it does factor in statistical significance in determining its sample size.
Audit type: Fixed-sized post-election audit. The statute does not require a specific sample size, though it must be “statistically significant.”
Geographic scope: State officials select a sample from each county that randomly chooses a group of precincts, absentee/early voting ballots, or a combination of both. For example, a group of only precincts in a county could be selected. Or a group of county precincts and absentee ballots cast in that county could be selected. [N.C. Gen. Stat. § 163-182.1]
Occurrence: After every election, before certification. The Board randomly selected the precincts on November 9, (24 hours after polls closed) or after the initial count of election returns. The Board approves in an open meeting the procedures for randomly selecting sample precincts. [N.C. Gen. Stat. § 163-182.1]
Guidance: Issued by State Board of Elections
Portion of ballot audited: The State Board of Elections selects a statewide ballot item for the audit. In presidential election years, the sampling must be the presidential ballot. If there is no statewide ballot item, the Board will determine a process for selecting a district or local ballot item to sample. [N.C. Gen. Stat. § 163-182.1]
Description of audit process:
The size of the sample is chosen in consultation with a statistician and must be statistically significant. [N.C. Gen. Stat. § 163-182.1]
State elections officials get a 10 digit “seed number” by rolling 10 different 10-sided dice. The 10 digit number is put into a software program, which randomly generates two groups of ballots for each county.
State Board gives the list of the ballot groups to be audited to the County Board of Elections. County Boards are assigned a specific contest to audit.
A bipartisan team hand-counts the selected ballot group in each county. They compare the results to the local voting machines. Any variances are noted.
The State Board must submit the rationale and findings for the audit to the Joint Legislative Oversight Committee and the Joint Legislative Oversight Committee on General Government 10 days after the audit is completed. [N.C. Gen. Stat. § 163-182.12A]
Sources: NCSBE, accessed 11/17/2022; Audit Results for July 26, 2022 Elections, accessed 11/18/2022; News Observer, 11/09/2022
Pennsylvania 🔔
Pennsylvania has two separate audit processes, a fixed-sized statistical sampling required by statute and an RLA created by executive order, first piloted in 2020. Though the 2% statistical sample and the post-election RLA are two separate processes, the board of elections may use the same group of randomly selected ballots to support both the statistical sample and the post-election audit. Both must take place before certification. Below we have outlined the RLA process, for which more robust statewide guidance is provided.
Audit type: Batch comparison RLA. In a batch comparison RLA, the auditing board manually examines every ballot in the randomly selected batches of ballots and compares the vote totals to the machine count.
Geographic scope: Statewide, conducted by each county
Occurrence: No later than November 28, 2022.
Portion of ballot audited: The Governor’s race
Description of Audit Process:
Identify auditors: The County Board of Elections must identify bipartisan teams of auditors (canvassers). These auditor teams can include members of the public, elections office staff, and precinct election officials. There must be at least two, but three members is recommended.
Create a Ballot Manifest
A Ballot Manifest is a detailed description or accounting of how the county’s ballots are stored. A county official will do this on a simple spreadsheet.
After the counties upload their manifests to the audit software, a random seed number is generated, and the audit software selects the random list of ballots to be retrieved by each county.
Create a Contest Totals by Batch file: The Contest Totals by Batch file outlines each batch of ballots and how many votes were cast for each contest (or candidate) in a particular batch. A county official will do this on a simple spreadsheet.
Ballot retrieval & tally: The County Board of Elections will retrieve the randomly selected batches and convene the meeting of the auditing board for the purpose of manually examining the ballots, tallying the votes and recording the vote totals on tally sheets.
Reporting tallies: The County Board of Elections will report the results of the post-election audit in the manner prescribed by the Secretary of the Commonwealth, usually using a secure website.The audit results must be transmitted to the Department of State concurrent with the county’s certification of official election returns. In addition to reporting the results of the audit, the county must note any anomalies or errors found during the audit process.
Sources: PA Department of State, Risk Limiting Audit Directive, 09/30/2022; Post Election Audits, accessed 11/21/2022; WITF, 11/18/2022
Wisconsin 🧀
Wisconsin audits its voting equipment after each election using a fixed-sized method. Voting machines are selected at random across the state by the Wisconsin Elections Commission (WEC), who ensures that a minimum of 10% of machines are included in the audit. This is not an audit to confirm results, but rather to ensure that all voting machines work properly.
Audit type: Fixed-sized voting machine audit
Geographic scope: Statewide, must include at least one machine in each county, conducted by the municipalities.
Occurrence: These audits begin after all county canvasses have been completed. This year, because the Secretary of State race is still within the recount margin, audit activities have been postponed until 5 pm on the third business day following the completion of the last county canvass, to be delayed further if a recount petition is filed. Therefore, the earliest deadline for audited municipalities to submit materials to the Commission has been moved to December 2, 2022 (from November 25, 2022). If a statewide recount is requested, the deadline to submit audit materials will be December 23.
Portion of ballot audited: A minimum of 4 offices are audited, including the top race on the ballot. Three federal or state contests are also randomly drawn to determine the election contest and included in the audit. This year, the Governor/Lieutenant Governor, Attorney General, Secretary of State, and Representative to the Assembly races have been selected.
Description of audit process:
Selection of audited machines: Wisconsin Elections Commission (WEC) determines the number of machines to include in the audit each year (this year 368 reporting units were selected) and then uses a randomization tool to select them, with additional voting machines selected to ensure at least 5 of each type of voting system used and that each of Wisconsin’s 72 counties is included in the audit.
Notice: The audits are open to the public and notice must be provided 48 hours in advance.
Conducting the audit: A minimum of 2 individuals conduct the audit. They each tally votes by hand and compare their totals. If the auditors’ totals agree, the totals are then compared to the results generated by the voting system. Any discrepancies are recorded and explained in the minutes of the audit.
Reports: Each municipality conducting an audit submits the designated reporting forms and supporting documents from the audit, including tally sheets, to WEC staff via email.
Review: WEC can review the audit and may conduct further investigation into the voting machines if there is an unexplained discrepancy.
Sources: Voting Equipment Audit, 2022; WEC guidance, 11/15/2022