• North Carolina county boards of elections conduct sample hand-to-eye counts of ballots to ensure that they were tabulated correctly
  • Though intended to be random, the way the sample is conducted makes it more likely that some types of ballots are selected for auditing than others, especially in larger counties
  • There are ways the General Assembly and the State Board of Elections can alter the hand-to-eye recount sampling method to make the sample more representative

Someone from New Hanover County wrote the John Locke Foundation expressing concern that some kinds of ballots are less likely to be included in post-election audits than others. That concern is valid and should be addressed.

State Law Requires a “Sample Hand-to-Eye Count” of Ballots

How will North Carolinians know that official vote totals reported by the State Board of Elections (SBE) are accurate?

One way is to audit the results.

As noted in the John Locke Foundations report “What Happened in 2020,” election officials perform several post-election audits. The focus here is the required sample hand‑to‑eye count.

North Carolina law (GS § 163-182.1(b)(1)) requires the SBE to “provide for a sample hand‑to‑eye count of the paper ballots of a statewide ballot item in every county.”

The partial hand-to-eye recount is designed to answer only one question: Were the ballots in election officials’ possession counted accurately?

The same state law that provides for a partial hand count of ballots also, however, unwittingly weakens how sample precincts are selected (GS § 163-182.1(b)(1)):

The sample chosen by the State Board shall be of one or more full precincts, full counts of mailed absentee ballots, full counts of one or more early voting sites, or a combination. The size of the sample of each category shall be chosen to produce a statistically significant result and shall be chosen after consultation with a statistician. The actual units shall be chosen at random.

The way the SBE interprets that requirement means that some ballots are much more likely to be selected than others, so the selection of ballots to be audited is not random (even if the units of ballots chosen are). That flaw lessens the validity of the hand-to-eye count audit, making it a less effective tool for detecting problems with the count.

The “Random” Selection Process Is Less Random than It Should Be

The selection process for which units of ballots are in the audit seems random at first glance:

The day after each election, the State Board of Elections randomly selects two sample groups of ballots for each county to count by hand. These samples can be all ballots cast at selected Election Day precincts or early voting sites, or all the absentee ballots cast in a county. The State Board then informs the counties of their sample groups of ballots and which contest on the ballots to count.

The SBE’s “sample audit input” file for the March 5 primary shows 2,616 precincts, 338 early voting sites, and 88 sets of countywide absentee ballots. All of these constitute units for the purposes of state law. According to an SBE report, 169 election-day precincts, 22 early voting sites, and nine countywide sets of absentee ballots were selected to be audited, with each county auditing two units.

In addition, counties have different numbers of units. For example, the “sample audit input file” for the March 5 primary shows that New Hanover County had 46 units (41 precincts, four early voting sites, and all absentee ballots) while Pamlico County had eight (6 precincts, one early voting site, and all absentee ballots). Larger counties have more precincts and early voting sites, but every county has only one set of absentee ballots. So, the smaller counties are more likely to have absentee ballots in their two-unit samples. From the example above, New Hanover has a 4.4 percent chance of having their absentee ballots included, while Pamlico has a 26.8 percent chance. It is no accident that most counties that included absentee ballots in their samples (Anson, Clay, Davidson, Gates, Moore, Pasquotank, Pender, Lincoln, and Martin; see the “sample audit output” file) are smaller counties.

That disparity is only somewhat mitigated by absentee and early voting units being larger than precincts.

How to Make the Audited Ballot Samples More Representative

The hand-to-eye audits must be conducted by units because ballots are tabulated by units (with, for example, all the ballots at an early voting site counted and reported to the county board of elections together).

There are ways of equalizing the chance of units, particularly absentee ballots in larger counties, being included in the audit.

Apportionment of Units by County Size

One way to address the problem partially would be to have the SBE more fairly apportion the number of units in the sample hand-to-eye count according to the number of units in each county (with every county having at least one per law). In other words, more populous counties would audit more units.

Under this system, for example, Pamlico County would need to sample only one unit, whereas New Hanover would need to sample four. The chance each county would have absentee ballots in its audit sample would be 12.5 percent for Pamlico and 9.0 percent for New Hanover. While not equal, those odds are much more balanced than they are under the current system.

Risk-Limiting Audits

A long-term solution would be to move to risk-limiting audits (RLAs). An RLA is a “hand-to-eye recount of a randomly selected sample of ballots in a contest, which provides strong statistical evidence that the machine-counted results are correct.” Since RLAs use a random sample of all ballots, every type of ballot is audited in every county after every election.

Two things are required to transition to RLAs:

  1. A change in the law to allow a move away from “unit-based” audits
  2. A transition from ballot marking devices (BMDs) to hand-marked paper ballots in all counties

The latter is important because auditors can better determine voter intent with hand-marked paper ballots:

Risk-limiting audits of a trustworthy paper trail can check whether errors in tabulating the votes as recorded altered election outcomes, but there is no way to check whether errors in how BMDs record expressed votes altered election outcomes.

That is why the ideal ballot sources for RLAs are “paper ballots preferably marked by hand, and supplemented with a ballot marking device for those who need to use one.”

Although North Carolina is moving toward hand-marked paper ballots, thirteen counties still use BMDs for all early voting and seven for all voters on election day.

There are better ways to conduct the post-election sample audit. The State Board of Elections and General Assembly should implement them.