Post-Election Procedures and Audits
Results on election night are unofficial. Canvass is the official process of determining the votes have been counted and tabulated correctly, resulting in the authentication of the official election results.
For close elections, the canvass period is especially important. During this time, elections officials count absentee ballots that came in before the deadline and research provisional ballots to determine whether they should be counted.
In every county, the canvass meeting when the results are certified is 10 days after Election Day. Because elections thrive on transparency, the canvass meeting is open to the public. Unless the county board unanimously votes for another site, the meeting will be held at the county board of elections office, per N.C.G.S. § 163-182.5.
Before the canvass meeting, the board must ensure that the sample hand-to-eye audit count and any discretionary or mandatory recounts have been completed.
Additionally, county boards confirm all eligible ballots have been counted, including:
- Ballots that were unable to be read in the precinct (ballot jams, torn ballots, etc.)
- Absentee ballots
- Provisional ballots
At the meeting, the board will prepare and sign all relevant, permanent public documents. The county board will retain one copy; a copy will be filed with either the clerk of Superior Court or, in the case of municipal elections, with the municipal clerk; and a copy will be provided to the State Board of Elections.
The state and county boards of elections conduct audits after every election. Audits can detect problems such as equipment tampering, ballot stuffing and voting machine or counting errors.
The State Board of Elections constantly strives to improve its post-election audits using best practices among elections experts and experiences of other states. After each election, elections officials conduct the following audits.
Types of Post-Election Audits
Six types of audits are:
- Voter History Audit
- Manual Entry Audit
- Provisional Audit
- Sample Audit
- Close Contest Audit
- Risk-Limiting Audit (Coming Soon)
When voters check in at a polling place or early voting site, they fill out an authorization to vote (ATV) form or one-stop application during early voting. This results in a voter history record for each individual. When ballots are inserted into the tabulator, the tabulator displays a ballot count for the number of ballots cast on the tabulator.
The voter history audit compares the number of ATV forms and one-stop applications with the number of physical ballots cast. These two numbers should generally match, but may be slightly off for valid reasons, such as if a voter checked in and then decided not to vote.
This audit is designed to identify certain problems or fraud, such as ballot stuffing, fraudulent manual entries, tampering with media cards, or certain ballot coding issues.
County election officials occasionally must enter election results by hand, rather than from results media, directly into tabulation software. The manual entry audit can catch inadvertent mistakes in transcribing numbers from tally sheets and election result tapes, as well as intentional manipulation of data.
North Carolina has updated its election reporting system to include an automated process that detects almost all transcription errors in real time as results are entered by hand.
Several data validation rules are run against manual entries, effectively eliminating transcription errors, within an error rate of a fraction of one percent.
Voters cast provisional ballots when there are questions about their qualifications or eligibility to vote in certain contests. Those ballots are held aside pending research by county boards of elections as to whether they should be counted.
This audit of provisional voter data analyzes data from several sources, including the Division of Motor Vehicles’ database, an incomplete queue that catalogs registration attempts that were deemed incomplete, and the current registration database as of Election Day.
Data analysts work to determine whether provisional voters were eligible to vote in the counties where they presented to vote. Additionally, geocoding services are used to geocode the voters’ addresses to confirm that they resided in the county at the time they voted. Audit results are sent to county boards of elections, which analyze them and, where appropriate, amend their decisions to reflect any changes. This audit often results in additional ballots being counted as required by law and aims to ensure that counties across the state treat provisional ballots uniformly.
The sample audit count is a test to ensure voting equipment read the voter’s choices accurately. It compares the machine counts with hand-to-eye counts conducted by elections officials in randomly selected voting sites. The sample audit count is open to the public and is completed before canvass.
The hand-to-eye counts required for this process are not recounts, although they are similar processes.
The day after the election, the State Board of Elections informs each county of their assigned contest and the two randomly selected samples (Election Day precinct, one-stop site, or absentee by mail ballots) to audit.
For a presidential election, the contest audited is always the presidential contest.
Selected ballots are hand-counted by a bipartisan team of trained volunteers. The hand-counted results are compared to the tabulated results and any variances are noted. Permitted variances include the following situations: (1) The write-in oval was not filled in, but a candidate’s name was written in, or (2) the machine did not count a choice that was represented by check marks or Xs or that was poorly shaded.
The county sends the machine counts and hand counts to the state along with an explanation of any discrepancies.
This audit compares the margin of victory in any contest to the sum of the aforementioned variances (based on the other audits) to ensure that election outcomes are justified.
The State Board soon plans to pilot the state’s first Risk-Limiting Audit. A Risk-Limiting Audit, or RLA, requires election officials to sample batches of ballots, count those ballots by hand and compare the results to the machine count. This is done until an acceptable confidence interval is reached, such as 95%, that an election outcome is reliable.