Post-Election Audits

10/25/2019

Table of Contents

Contact

Introduction

A post-election audit checks that the equipment and procedures used to count votes during an election worked properly, and that the election yielded the correct outcome.

In states that conduct post-election audits (see table below for more details) it is usually a statutory requirement. Legislatures can decide whether or not to require post-election audits in their states.

While the phrase "post-election audits" can be used to mean a variety of election validation efforts, as a term of art it refers to checking paper ballots or records against the results produced by the voting system to ensure accuracy. 38 states + DC currently have a post-election audit as defined here. Paper records used in an audit may include voter-marked paper ballots, voter-verified paper audit trails produced by direct-recording electronic voting machines (DREs) or paper ballot records produced by ballot-marking devices. Typically only a sample of the paper records are examined, so in effect a post-election audit is a partial recount of results to verify that the voting system is accurately recording and counting votes.

Although post-election audits can lead to a full recount if errors are detected, they differ from a recount in that they are conducted regardless of the margins of victory. Recounts are most often triggered or requested if there is a tight margin of victory. See NCSL’s webpage on recounts for more information.

Below, this webpage provides a detailed table on states with post-election audits and the variations within them. 

A photo of a pencil resting on top of an audit checklist

Why Do a Post-Election Audit?

If an audit process is in place, it can inform election officials of any bugs or errors in the system, and can act as a deterrent against fraud. Proponents of post-election audits argue that they can also help avoid a full recount by revealing when a recount is necessary to verify the correct election outcome. And, ultimately, a robust post-election audit can increase confidence in the results of an election.

Post-election audits can be time-consuming for election administrators, and most post-election audits contain an element of hand counting paper records, which is an error-prone process. Even with good procedures, people physically looking at votes on a piece of paper are less reliable at tabulating than a machine, although humans are better at discerning voter intent. 

Issues Legislators Might Need to Address

  • Are audits required? Are they voluntary or only conducted under certain circumstances?
  • What type of equipment is used in the state? A paper trail is needed to conduct a post-election audit, and some older systems do not contain this paper trail. Most newer voting systems, however, have the capacity to undertake a post-election audit and produce the necessary paper trail.
  • How are the audited ballots, precincts, machines or districts selected?
  • How many races will the audit include?
  • Who determines what contests are selected to be audited? What are the factors in determining the selection of audited contests (for example, the closeness of the reported outcome, the geographical scope of the contests or any cause for concern regarding the accuracy of the reported outcome)? When is this decided? Is it before the election, after unofficial results are available, or after the election has been certified?
  • How robust is the audit? Does a risk-limiting audit make sense? (See below for details on risk-limiting audits)
  • Which categories of ballots are audited? Does the audit include absentee ballots, early voting ballots, ballots sent from citizens overseas, and provisional ballots, or just votes cast in person on Election Day? 
  • When is the audit conducted? Some states call for post-election audits after election results are finalized, with the intent of providing information to improve future elections. Other states conduct an audit before results are finalized, which may allow for the resolution of any discrepancies.
  • Does a requirement for post-election audits put an extra burden on election officials? Post-election audits can be labor and time-intensive. In the longer term, though, states may find cost savings if audits make expensive recounts less common. And, with time better procedures will develop.
  • Who conducts the audit? Is it done by the states, local jurisdictions, or an independent audit board? Who provides the resources and staff for conducting the audit? Who participates and who can be present when the audit is conducted?
  • Who conducts the audit? Is it done by the states, or the local jurisdictions? Who provides the resources and staff for conducting the audit?
  • What happens if the audit shows a discrepancy? When is a full recount triggered?

Types of Post-Election Audits

There are several variations of post-election audits used in states.

Traditional Post-Election Audits: Most audits look at a fixed percentage of voting districts or voting machines (see table below for details) and compare the paper record to the results produced by the voting system. Even in a landslide election, they will count the same number of ballots as they would in a nail-biter election. 

  • 34 states and DC require a traditional post-election audit: Alaska, Arkansas, Arizona, California (counties have the option of conducting a risk-limiting audit), Connecticut, Delaware, District of Columbia, Florida, Georgia, Hawaii, Illinois, Iowa, Kansas, Kentucky, Maryland, Massachusetts, Michigan, Minnesota, Missouri, Montana, New Jersey (although the state currently does not have machines that produce a paper record and therefore cannot yet meet this requirement), New Mexico, New York, North Carolina, Ohio (risk-limiting audits are recommended but not required), Oklahoma, Oregon (counties have the option of conducting a risk-limiting audit), Pennsylvania, Tennessee, Texas, Utah, Vermont, Washington (counties have the option of conducting a risk-limiting audit), West Virginia and Wisconsin.

Traditional post-election audits are usually conducted manually by hand counting a portion of the paper records and comparing them to the electronic results produced by an electronic voting machine. However, some states have a process by which some or all of the audit can be conducted electronically. This may be done with the assistance of a computer or a tabulation device other than the one that was initially used to tabulate results. And, some traditional post-election audits use a "tiered" system, which means a different number of ballots are reviewed, depending on the margin of victory.

A case study of a traditional post-election audit, from West Virginia, is provided below, as well as a tiered post-election audit, from New Mexico, and a partially electronic post-election audit in Connecticut.

Risk-Limiting Audits: In recent years, researchers have developed statistically based audit techniques that cut down on the number of ballots to be audited, while also providing statistical confidence that the election result is correct. As defined in Washington, "A "risk-limiting audit" means an audit protocol that makes use of statistical principles and methods and is designed to limit the risk of certifying an incorrect election outcome." If the margin is larger, fewer ballots need to be counted. If the race is tighter, more ballots are audited. For more details see NCSL's webpage on Risk-Limiting Audits.

  • 4 states have a statutory requirement for a risk-limiting audit: Colorado, Nevada (pilots in 2020, all counties in 2022), Rhode Island, and Virginia.
  • Georgia and Indiana have statutory requirements to conduct risk-limiting audit pilots.
  • California, Ohio, Oregon and Washington provide options for counties to run different types of audits, one of which being a risk-limiting audit.

A case study of a risk-limiting audit, from Colorado, is provided below and details on other statse can be found in the State Requirements Table.

Other Post-Election Options

Some states have other variations, which do not fit the definition used here of a post-election audit, although these processes are valuable in terms of confirming a well-run election.

Procedural Audits

States may have a process for ensuring that the correct process and procedures were followed during the course of the election. This is referred to as a “procedural audit” and may be conducted instead of or in addition to a post-election audit. Procedural audits vary in their scope and comprehensiveness, but almost always include a ballot accounting and reconciliation process. This isn’t a check that the software in the voting machine is working correctly, but rather a check on the human processes.

The first step in accounting for and reconciling ballots occurs at the polling place, when poll workers may be required to fill out a ballot accounting log indicating the number of blank ballots received at the beginning of the day, or the “zero tape” from the voting machine from the beginning of the day that shows no ballots have yet been recorded. Later, the number of ballots voted during the day or the results tape from the voting machine, and any incidences that may have occurred throughout the day, are collected. (For example, if a voter signed in on the poll book but subsequently left without casting a vote, this would be recorded as an “incident”.) Election officials can then see if there are any discrepancies in the number of voters that signed in at a polling place and the number of votes that were cast. If there is a discrepancy, election officials can investigate it. Did all of the votes get uploaded correctly? Were votes from one machine mistakenly counted more than once? Were any voted ballots mistakenly not counted?

A procedural audit also may include a reconciliation of the provisional ballots, early ballots and absentee ballots, and a review of security and chain-of-custody procedures. Chapter V in the report, Counting Votes 2012: A State by State Look at Election Preparedness, contains details on state ballot accounting and reconciliation procedures.

For an example of a detailed procedural audit, see Michigan’s Post-Election Audit Manual, based on legislation passed in 2012, to require a thorough procedural audit at both the state and county levels that covers numerous pre-election, Election Day and post-Election Day tasks. 

South Carolina runs a procedural audit to compare the tabulated results of the election with the raw data in the electronic audit files of each voting machine using a series of computer applications written to detect anomalies. See Description of Election Audits in South Carolina. 

Post-Election Audits Under Certain Circumstances

Some state laws do not require a post-election audit to be run after each and every election, but instead require them in certain circumstances. For example:

  • Idaho conducts a post-election audit only when a recount is required (Idaho Code §34-2313).
  • Indiana requires a procedural audit under some circumstances, and is piloting the use of risk-limiting audits in 2020 (Indiana Code §3-12-13). The secretary of state may require a procedural audit if there is an investigation relating to the election or a recount (§3-12-14).If the total number of votes cast (on the electronic voting system and by absentee ballot) and the total number of voters in the poll book differ by a pre-defined "audit threshold," the county must conduct an audit of that precinct (Indiana Code §3-12-3.5-8).
  • Nebraska doesn’t have a statutory requirement for post-election audits, but they may be conducted by the secretary of state’s office.

Post-Election Logic and Accuracy Testing

At least two states (North Dakota and Wyoming) conduct a repeat of the pre-election logic and accuracy test after the election to ensure that voting machines are still tabulating accurately. Before an election, election officials create a “test deck” of ballots (a stack of all ballots styles with different iterations of marked ballot selections) that are run through tabulators to ensure races are being accurately recorded and tabulated. In these states, the same test deck is run through the machines after the election, to once again test the accuracy of the machines.

Case Studies

Traditional post-election audit: West Virginia

In West Virginia the post-election audit is conducted as part of the canvass. At least 3 percent of precincts are chosen at random, and voter-verified paper ballots are counted manually to compare the manual count to the automated equipment count. If the manual count differs by more than 1 percent from the automated equipment count, or if the outcome of the election would change due to the discrepancy, then the discrepancies are immediately disclosed to the public and all voter-verified paper ballots are manually counted. Note that whenever there is a difference between the vote totals from the automated equipment and the manual count, the manual count of the voter-verified paper ballot is the vote of record (W. Va. Code, § 3-4A-28).

Risk limiting audit: Colorado

Risk-limiting audits (RLAs) were first enacted in Colorado in 2009, and implemented in November 2017. According to the statutory language, the goal of the legislature in enacting RLAs was “to ensure effective election administration and public confidence in the election process” and to “provide a more effective manner of conducting audits than traditional audit methods in that risk-limiting audit methods typically require only limited resources for election races with wide margins of victory while investing greater resources in close races” (Colo. Rev. Stat. §1-7-515).

In a nutshell, an RLA provides strong statistical evidence that the reported outcome of an election is correct. It also provides a high probability of discovering and taking appropriate action in the case of a wrong outcome. State and local election officials worked closely with academics, experts and advocates to establish a strong set of rules and procedures for the implementation of RLAs that can be a model for other states. (See Colorado Secretary of State Election Rule 25).

The first step is to establish a “risk limit,” the largest chance that an incorrect reported tabulation outcome of a contest will not be corrected by the audit. In Colorado this risk limit is set by the secretary of state’s office before the election. A risk limit of 10 percent means that there is a 90 percent chance that the audit will correct a wrong outcome. The smaller the risk limit, the greater the number of ballots to initially audit. The audits are managed centrally at the secretary of state’s office, but conducted locally by a bipartisan audit board at the county election official’s office.  

RLAs conclude in one of two ways. The audit either (a) stops when it finds strong evidence that the reported outcome is correct, or (b) fails to find strong evidence that the outcome is correct, and evolves into a full hand count of ballots.

For more details on the process works, see this blog from Colorado election officials in preparation for the first RLA in 2017.

Tiered post-election audit: New Mexico

New Mexico has a tiered traditional post-election. This means that a greater number of precincts are audited if the margin of victory is narrower. An independent auditor, hired by the secretary of state, oversees the audit and assists counties with its conduct. Statute contains a table that instructs the auditor how many precincts, out of a pool of all precincts, should be selected according to the winning margin between the top two candidates for a given race. At the low end, if the margin of victory is greater than 15 percent no precincts are tested for that office. At the high end, if the margin is 0.5 percent or less,165 precincts are tested. 

Similar to a risk-limiting audit, precincts are randomly selected “in a process that will ensure, with at least 90 percent probability for the selected offices, that faulty tabulators would be detected if they would change the outcome of the election for a selected office” (N.M. Stat. Ann. §1-14-13.2).

A hand count is conducted for all federal offices, the governor, and the statewide office unofficially won by the smallest margin in the selected precincts. If there is a high error rate between the difference of the hand counts and the original vote totals reported by the voting machines, another sample may be selected. If the error rate between the first and the second samples is more than 90 percent, a full recount is ordered.

Electronic post-election audit: Connecticut

At least 5 percent of all voting districts used in the election are randomly selected for a post-election audit in Connecticut. The races included in the audit vary by the type of election, and the audit must be completed no later than two business days before the canvass. 

The secretary of state partners with the University of Connecticut’s Center for Voting Technology Research (VoTeR Center) to conduct post-election audits. The two organizations jointly developed a semi-automated audit to augment the hand count audit. The option to conduct audits using either a manual audit or electronic equipment was enacted in 2015 legislation (SB1051).

A VoTeR Center report on Computer Assisted Post Election Audits explains that the semi-automated approach was developed because hand-counted audits are “expensive, time-consuming, labor-intensive and error-prone.” According to the report, “It takes two auditors up to five hours to examine 1,000 ballots, while it was observed that hand counted audit returns routinely show up to 2 percent error.” The report also notes that fully automated audits have associated risks and do not adequately provide confidence in election outcomes, so a combination of the two approaches was desirable.

With the help of a grant from the U.S. Elections Assistance Commission, the VoTeR Center developed a computer “audit station” that presents scanned ballot images on a projector. This allows those auditors to examine the ballot more fully than if they handled the paper ballots themselves, and also offers more transparency since the public can also see what the auditors are seeing. The system identifies whether a given ballot has questionable votes or marks, or if the voter’s intent appears to be unclear. It then allows the human auditors to make the final decision on how the marks on the ballot should be interpreted—i.e. who the voter intended to select. The audit is conducted in small batches so that auditors can easily pull a problematic ballot and examine the physical ballot more closely.

Recent Legislative Action

  • 2019
    • Delaware SB 121 established a traditional post-election audit. 
    • Arkansas SB 524 established a procedural audit and required the state board of elections to conduct pilot audits of various counties to inform a more comprehensive post-election audit plan for the state.
    • Georgia HB 316 established a traditional post-election audit to be put in place as soon as possible but no later than the November 2020 general election, and a risk-limiting audit pilot project in one or more counties by December 31, 2021.
    • Indiana SB 405 established pilots for risk-limiting audits and required procedural audits of counties in cases of an investigation of the election or a recount.
    • Maine HP 756 required the secretary of state to conduct a study of best practicies in post-election ballot audits, recounts, and the intersection of audits with recounts. 
    • Nevada SB 123 required risk-limiting audits to be phased in. A pilot program for conducting risk-limiting audits will be conducted in November 2020 and each county clerk must conduct a risk-limiting audit beginning in 2022. Nevada previously had a traditional post-election audit in place.
    • Oklahoma SB 261 established a traditional post-election audit, with methods, procedures and timing to be determined by the secretary of the state election board.
    • Oregon SB 944 permitted county clerks to choose to conduct a traditional or a risk-limiting audit.
  • 2018
    • Michigan SB 1238 made conforming changes to statute to comply with a voter-approved constitutional amendment that required and audit of election results. 
    • California AB 2125 permitted counties to conduct a risk-limiting audit, beginning at the statewide primary election in 2020, in place of the 1% manual tally. See table below for details.
    • Kansas HB 2539 established a traditional post-election audit of 1% of precincts or at least one precinct in the county.
    • Maryland HB 1248 modified its existing post-election audit process to also include an automated and manual audit conducted by the state board of elections. 
    • Indiana HB 1383 modified its procedural audit language to include an "audit threshold" for when there is a discrepancy between the total number of votes cast on an electronic voting system and the number of voters who cast ballots.
    • Washington HB2406 amended post-election audit requirements, giving county auditors the choice of conducting audits of duplicated ballots using several different methods. The bill states,"The legislature wants to maximize this locally run benefit by adding options to the auditing process for local elections administrators. Multiple jurisdictions, with multiple options for ensuring election outcomes will increase the transparency, integrity, and trust of our elections process."
    • Iowa HB 2252 amended language regarding those who may observe the audit. Previously the language read that representatives from two political parties shall be present. The modified language clarified that representatives of political parties shall be invited, and the state commissioner informed if the invited representatives do not attend. 
  • 2017
    • Rhode Island HB5704/SB413 required risk-limiting audits beginning in 2018.
    • Virginia SB1254 required "risk-limiting audits," though the law did not define the term or contain details on how they will be implemented.
    • Washington HB 1507 enhanced election reconciliation reporting requirements.
  • 2014-2016
    • Connecticut SB252 (2016) reduced the audit requirement from 10 percent to 5 percent of voting districts.
    • New York SB5908 (2015) specified that post-election audits could be performed either manually or “via the use of any automated tool authorized for such use by the state board of elections which is independent from the voting system it is being used to audit.
    • Connecticut SB1051 (2015) authorized the use of electronic equipment for the purpose of conducting audits after January 1, 2016.
    • Massachusetts HB3788 (2014) established post-election audits.
    • Virginia SB456 (2014) made the state’s post-election audit pilot program a permanent feature of the state’s election law.
    • Vermont SB86 (2014) made the state’s post-election audits mandatory, rather than at the discretion of the Secretary of State. 

State Requirements Table

State

Audit Type

Scope and Method

Who Conducts the Audit?

Timing

What Happens if a Discrepancy is Found?

Who Can Observe?

Alabama

None

N/A

N/A

N/A

N/A

N/A

Alaska

Alaska Stat. §15.15.420

§15.15.430

§15.15.440

§15.15.450

§15.10.170

Traditional

One randomly selected precinct from each house district that accounts for at least 5% of ballots cast in that district.

Local election officials and State Review Board members, with oversight from the state election director.

Before the canvass. Begins as soon as practicable, and completed no later than 16 days after the election.

If discrepancies of more than 1% are found by local officials, the state election director hand counts all the ballots from that precinct by hand. The director certifies in writing and publishes online any changes resulting from this count.

Appointed representatives from political parties are required to observe; other observers may be present as well.

Arizona

Ariz. Rev. Stat. §16-602

State of Arizona Elections Procedures Manual

Traditional

2% of precincts in county or 2 precincts (whichever is greater). For counties that use vote centers, 2% of vote centers or 2 vote centers (whichever is greater).

 

1% or 5,000 early ballots in the county (whichever is less) is also included in the audit.

County election officials and party representatives appointed by the county party chairperson, with oversight from the Secretary of State.

Before the canvass. Begins within 24 hours after the polls close, and completed before the canvass.

Before the election a committee (made up of persons with expertise in math, statistics or voting systems) establishes margins for each contest to be used during the audit to determine when the audit should be expanded.

Audits are conducted publicly, but political party observers have additional access to the process.

Arkansas

Ark. Code § 7-4-101, § 7-4-121, § 7-5-702

Procedural audit with a pilot program for an election equipment audit.

 A sufficient number of counties, polling sites, early voting locations and vote centers are selected by lot to obtain a meaningful sample. Temporary language in SB 524 (2019) would create a pilot program for auditing each election system using a ballot marking device and tabulator in various counties in the state to inform a more comprehensive plan to audit the state’s election equipment. State Board of Elections.

No less than 60 days after the general election.

A report is filed that has no legal effect on the outcome of any election subject to the audit. Counties that refuse to provide the requested documents forfeit state funded election expenses for a period of up to two years.

Not specified.

California

Cal. Elec. Code §336.5

§15360

§15365 et seq. 

(West 2015)

Traditional, with an option to conduct a risk-limiting audit instead beginning in 2020

Traditional audit: 1% of the precincts, randomly selected, and 1 precinct for each race not included in the randomly selected precincts.

Risk-limiting audit: each contest fully contained within the county’s borders, and partial risk-limiting audits for each cross-jurisdictional contest.

A ballot-level comparison audit or ballot polling audit is conducted with a 5% risk limit. The Secretary of State shall adopt regulations to implement the risk-limiting audit requirements.

Local election officials, with oversight from the Secretary of State.

During the canvass.

Traditional audit: discrepancies are required to be reported, though no additional action is mandated. The elections official conducting the risk-limiting audit shall publish a report on the results of the risk-limiting audit in the certification of the official canvass of the vote.

Audits are conducted publicly. 

Colorado

Colo. Rev. Stat. §1-7-515

Colo. Sec. of State Election Rule 25 on risk-limiting audits. Colo. Sec. of State Election Rule 8 on watchers.

Risk-limiting

The Secretary of State selects the “risk limit” that applies for a risk-limiting audit (RLA) for that election before the election. Different risk limits may be established for comparison audits and ballot polling audits.(See Colorado case study for additional information)

County election officials, with oversight from the Secretary of State. The secretary of state selects the “target contest” for the RLA, including at least one statewide contest and one other contest.

Before the canvass. Audit reports must be submitted to the Secretary of State’s Office by 5 p.m. one day before the canvass deadline.

The RLA will continue until the risk limit for the target contests is met or until a full hand count results. The audit report submitted to the Secretary of State must include any discrepancies found and the corresponding ballot images.

At least two members of the canvass board of the county observe the audit, and members of the canvass board may serve as members of the audit board. Eligible watchers may also observe the post-election audit. Members of the public have explicit access to certain parts of the process, and may observe the entire process at the discretion of state and local election officials.

 

 

 

 

 

Connecticut

Conn. Gen. Stat. §9-320f

Traditional

 

At least 10% of randomly selected voting districts. May be conducted by a hand count or electronic count.[3]

Local registrars of voters. The Secretary of State randomly selects offices to be audited.

Before the canvass. Between the fifteenth day after the election and two business days before the canvass.

A re-canvass is ordered if the difference between the manual and machine counts is more than 0.5% and cannot be resolved through other means.

Audits are conducted publicly. Audit reports are filed with the Secretary of State and subsequently analyzed by the University of Connecticut, and are also open to the public.

Delaware

Del. Code. Title 15 § 5012A

Traditional

Within 48 hours of certification of results, 1 randomly selected voting device in each county and 1 randomly selected election district in the City of Wilmington. Within 60 days of certification, 1 randomly selected statewide race in 1 randomly selected election district in each county and 1 randomly selected election district in the City of Wilmington, different from the previously selected districts.

State Department of Elections

Within 48 hours after the certification of results for the first audit, and within 60 days of the certification of results for the second audit.

A report containing the results of the audit must be published within 60 days of the audit and the Department shall develop procedures if an audit reveals a discrepancy, including the threshold that triggers specific action and what corrective actions need to be taken.

Audits are conducted publicly.

District of Columbia

D.C. Code Ann. §1-1001.09a

Traditional 

At least 5% of precincts with precinct-level vote-tabulation machines and at least 5% of the voter-verifiable records that are tabulated centrally.

At least 3 contests are examined: 1 district-wide contest and at least 2 ward-wide races.

Board of Elections.

Before the election is certified. Date is announced no later than 3 business days after tabulation has been completed, but no fewer than 24 hours in advance of the audit.

If there is a discrepancy with an error rate greater than 0.25% or 20% of the margin of victory, whichever is less, and the discrepancy is not attributed to marking errors, a 2nd count shall be conducted. If the 2nd count confirms the discrepancy, another precinct in each ward where the contest appeared and an additional 5% of centrally tabulated ballots are audited. If the discrepancy is still there, all ballots with that contest are recounted.

Audits are conducted publicly. Results of the audit are also made publicly available before the election is certified.

Florida

Fla. Stat. Ann. §101.591

Traditional

Two options: 1) manual audit of votes for one randomly selected race in 2% of precincts containing that race, or 2) public automated tally of the votes cast across every race that appears on the ballot in at least 20% of randomly-chosen precincts.

County canvassing board or local board responsible for certifying the election.

Immediately following certification of the election. Results must be made public no later than 7 days after certification and reported to Dept. of State within 15 days after certification.

The audit report to the Dept. of State must contain a description of discrepancies, the likely cause, and recommended corrective action to avoid/mitigate such circumstances in future elections.

Audits are conducted publicly and results are made available to the public.

Georgia

Ga. Code Ann. §21-2-498

Traditional, with a pilot program for risk-limiting audits.

 Manual inspection of random samples of paper ballots, including those cast in person, by absentee ballot, early voting and provisional ballots.

Local election superintendents

Before the final certification of the contest.

Not specified.

Conducted in public view and details of the audit are made public within 48 hours of completion.

Hawaii

Hawaii Rev. Stat. §16-42, Haw. Admin. Rules § 3-172-102

Traditional

10% of precincts using the electronic voting system.

 

Chief election official and bipartisan audit team.

Before the election is certified.

If discrepancies occur, the chief election official conducts an expanded audit and to the extent possible resolves misreporting problems.

Not open to the public, but “observers may request to conduct a manual audit.”

Idaho

Idaho Code §34-2313

Other

A post-election audit is only conducted when a recount is required. For a statewide or federal office or a statewide measure, two precincts in each county are audited. For other offices or measures, the greater of 100 or 5% of ballots cast are audited.

County election officials.

A recount may be requested within 20 days of the canvass. The post-election audit is the first step of the recount.

For statewide or federal office or a statewide measure, a manual count will be conducted if the difference is greater than O.25%; for other officers or measures, a manual count will be required if the difference is greater than 1% or two votes.

Recounts are conducted in plain view of the candidates or their representatives.

Illinois

Il. Rev. Stat. ch. 10 §5/24A-15,

ch. 10 §5/24C-15

Traditional

Where in-precinct counting equipment is used, 5% of precincts and 5% of voting devices used in early voting.

County election officials, with oversight from the State Board of Elections, which also conducts the random selection of precincts.

Before the Canvass.

If any error is detected, the cause shall be determined and corrected, and an errorless count shall be made prior to the official canvass. If an errorless count can’t be conducted, a written report explaining the errors is made available for public inspection.

The State Board of Elections, the State's Attorney and other appropriate law enforcement agencies, the county chairman of each established political party and qualified civic organizations may observe the audit. Audit results are made public.

Indiana

Indiana Code

§3-12-13

§3-12-14

§3-12-3.5-8

 

Other - procedural audits in certain circumstances, with a risk-limiting audit pilot in 2020

A risk-limiting pilot program has been authorized. The secretary of state may require a procedural audit of a county if there is an investigation or a recount. A procedural audit must also be conducted if

if the total number of votes cast and the total number of voters recorded in the poll book differ by the "audit threshold." 

County election officials. The secretary of state designates risk-limiting audit pilot counties, and supervises the procedural audit in cases of an investigation or recount.

For a procedural audit due to meeting the audit threshold, within 30 days after the election.

Not later than ninety (90) days after each election in which a procedural audit meeting the audit threshold was conducted, the secretary of state shall publish a report stating whether the results of each audit indicate that the discrepancy was the result of human error, intentional violations of election laws, unknown causes, or a combination of these factors.

The public is notified of the time and place of a procedural audit meeting the audit threshold at least 48 hours before it begins.

Iowa

I.C.A. § 50.51

Traditional

Number of counties and precincts to be audited are determined by the state commissioner and selected by lot.

The state commissioner of elections, with the cooperation of the county commissioners.

The audit report must be transmitted to the state commissioner of elections no later than 20 days after the election. The results of an audit do not change the results of an election.

An administrative recount may be ordered if the results of the audit require an administrative recount, for example if the commissioner suspects that voting equipment used in the election malfunctioned or that programming errors may have affected the outcome of the election.

A representative selected by each of the two political parties whose candidates received the highest number of votes statewide in the preceding general election.

Kansas

K.S.A. 25-3009

 

Note: These provisions apply to elections after Jan. 1, 2019

Traditional

1% of all precincts, randomly selected, with a minimum of one precinct within each county. The audit includes a manual review of all paper ballots and the races examined differ depending on if it is a presidential election year or not.

County election officials.

Before certification of the election.

If a discrepancy is reported between the audit and the unofficial returns and cannot be resolved, the county election officer or the secretary of state may require audits of additional precincts. Once the audit has been completed, the results of the audit shall be used by the county board of canvassers when certifying the official election results.

The audit is conducted publicly. Any candidate or entity who is authorized to appoint a poll agent may appoint a poll agent for the audit.

Kentucky

Ky. Rev. Stat. §117.383

§117.305 §117.275(9)

Traditional

Randomly selected precincts representing 3-5% of the total ballots cast in each election. Note that some counties in Kentucky use DREs with no paper trail, making a manual audit in these counties impossible.

State Board of Elections.

Part of the canvass.

If a discrepancy is discovered, a re-canvass is required and errors corrected.

Representatives of candidates, political parties, and members of the news media may be authorized to observe.

Louisiana

None

N/A

N/A

N/A

N/A

N/A

Maine

None

N/A

N/A

N/A

N/A

N/A

Maryland

Code of Md. Regs. §33.08.05.00 et seq., Md. Election Law §11-309

Traditional

A procedural audit is conducted on the local level. The greater of 3 precincts with at least 300 registered voters, or 5% of all precincts used in the election are audited by comparing voting machine results with those reported in the election management system. Following each statewide general election, the state board shall conduct an automated software audit of the electronic images of all ballots cast (required for primaries too) and a manual audit of voter-verifiable paper records (optional for primaries). The manual audit includes at least 2% of precincts statewide, including on randomly chosen precinct in each county, and at least 1% of the statewide total of early votes, absentee votes, and provisional votes in the previous comparable election.

Local election official and board of canvassers conduct the procedural audit. The state board of elections conducts the audit of ballot images and the manual audit of paper records. 

Procedural audit is conducted before local election officials certify the election. The manual audit by the state board must be completed within 120 days after the election. It does not have any effect on certified results, but shall be used to improve the voting system and process for future elections. 

For the procedural audit, if there is a discrepancy of five or more between the number of voters checked in to vote and the number of ballots cast, local officials must hand count signed voter cards and review other materials to determine the cause of the difference. Discrepancies discovered during the procedural audit must be resolved and the resolution approved by the State Administrator. If a discrepancy is discovered by the manual audit the state board may expand it, or take any other actions necessary to resolve the discrepancy. 

The manual audit conducted by the state shall allow for public observation to the extent practicable. 

Massachusetts

Mass. Gen. Law Ann. ch. 54 § 109A

 

Traditional

3% of all precincts in the commonwealth. Note that an audit is only conducted after a presidential election.

Board of registrars or election commission. The Secretary of State selects the precincts to be audited.

The audit must be completed no later than 14 days after a presidential election.

If there is a discrepancy that reasonably leads to doubt about the outcome of the election or systemic failure to accurately count ballots, the Secretary of State may order audits of additional precincts, offices or ballot questions as necessary to ensure that the outcome of the election is accurate and that the cause of the systemic failure is identified.

Audits are conducted publicly and results made public.
 

Michigan

M.C.L.A. § 168.31a

Post-Election Audit Manual (2018)

Traditional and procedural

An extensive procedural audit is conducted. It is required to include an audit of the results in at least 1 race in each precinct randomly selected for audit and at least 1 statewide race or ballot question for statewide elections. The audit reviews procedures performed before, during, and after the conduct of an election.

Counties and state-level election staff.

After the canvass. The audit does not change any certified election results. 

Discrepancies discovered as part of the procedural audit will be used to develop training for the future. The audit of voted ballots will reinforce accuracy and security of the voting system.

The audit takes place in a public location.

Minnesota

Minn. Stat. Ann. §206.89

Traditional

Depends on county size. At least two precincts for smaller counties. For larger counties, four precincts or 3% of precincts (whichever is greater).

County election officials, with oversight from the Secretary of State. The county canvassing board randomly selects precincts to be audited.

Before the canvass (no later than 2 days before the state canvassing board meets to certify the election).

If there’s a difference greater than 0.5% (or greater than 2 votes in a precinct where 400 or fewer ballots were cast) an additional review of at least 3 precincts is conducted. If there is still a discrepancy of the sort listed above, all precincts must be reviewed within that district. If the audit results show an error in 10% of the total votes cast in the election in one or more counties, a manual recount of all districts where the error occurred is ordered.

Audits are conducted publicly.

Mississippi

None

N/A

N/A

N/A

N/A

N/A

Missouri

15 Mo. Code of State Regs.

§30-10.090

§30-10.110

 

Traditional

At least 5% of precincts.

Local election officials, with oversight from the Secretary of State, who also randomly selects precincts to be audited.

Before certification of results.

If results differ by more than 0.5%, discrepancies are investigated and resolved.

Random selection of precincts is open to the public, and results of the audit are public. No other guidance for observers is provided.

Montana

Mont. Code Ann.

§13-17-501 -

§13-17-509

Traditional

At least 5% of the precincts in each county or a minimum of one precinct in each county, whichever is greater; and the audit must include an election for one federal office, one statewide office, one legislative office, and one ballot issue.

County audit committee, with oversight from the Secretary of State. The State Board of Canvassers selects the precincts to be audited.

Before the canvass.

If there is a discrepancy of more than 0.5% of total ballots cast or 5 ballots (whichever is greater) and the discrepancy is due to machine error and not administrative or user error, the machine involved in the discrepancy may not be used in another election until it has been examined and tested by a computer software expert in consultation with a voting system vendor and approved by the secretary of state. At least 3 additional precincts must be audited.

Audits are conducted publicly, and results are made public.

Nebraska

Source: Nebraska Secretary of State's Office

 

Other

A post-election audit is not required by statute, but may be conducted at the discretion of the secretary of state. A minimum of two percent of precincts are randomly selected, and the audit includes one each of federal, statewide and local races. 

Secretary of State’s Office

Not specified

 Any discrepancies are checked and noted in a report provided to the Secretary of State’s office.

Not specified.

Nevada

SB 123 (2019)

Nev. Admin. Code 293.255

Traditional, with pilots of risk-limiting audits in 2020 and all counties conducting risk-limiting audits in 2022.

Traditional audits: If county population is greater than 100,000, the audit looks at 2% of the machines. If the county population is less than 100,000, 3% of the machines are audited.

Risk-limiting: The secretary of state adopts regulations on procedures and scope.

County clerks, with oversight from the Secretary of State. The secretary of state sets regulations for post-election audits, including risk-limiting audits.

Traditional audits: Before the canvass. Results must be sent to the Secretary of State within 7 working days after the election.

Not specified.

“Any member of the public who observes the postelection certification audit shall not interfere with the conduct of the audit.”

New Hampshire

None

N/A

N/A

N/A

N/A

N/A

New Jersey

N.J. Stat. Ann. §19:61-9

Traditional

At least 2% of election districts. Note: The statutes referenced here are dependent on implementation of new voting systems that produced voter-verifiable paper records. As of May 2018 those machines have not yet been purchased. The current voting system uses paperless DRE machines that produce no auditable record, so post-election audits are currently not being conducted in New Jersey.

Local election officials, with oversight from an independent, professional audit team appointed by the Attorney General.

Before certification of election. Within a “reasonable period of time after the final vote count.”

If a discrepancy is discovered, the audit is expanded to include additional districts or audit units. Criteria to be employed to trigger an expansion of the audit are established before the election.

Audits are conducted publicly and results of the audit are published.

New Mexico

N.M. Stat. Ann. §1-14-13.2 et seq.,

N.M. Admin. Code 1.10.23

Traditional, with a tiered system based on the margin of victory

Audit of electronic voter tabulators in randomly selected precincts is conducted for all federal offices, government and statewide elective offices. The number of precincts to be tested depends on the winning margin for each office.

County clerks, with assistance and oversight from an independent auditor hired by the Secretary of State.

A random sample of precincts is selected by the independent auditor no later than 12 days after the election. The clerks shall report their results to the auditor within ten days of being notified which precincts to audit.

If there is a high error rate between the difference of the hand counts and the original vote totals reported by the voting machines, another sample may be selected. If the error rate between the first and the second samples is more than 90 percent, a full recount is ordered.

 

 

The selection of the sample to be audited is conducted publicly and audit data is reported to the public. Accredited county canvass observers may be present during the audit process.

New York

N.Y. Election Law § 9-211 (McKinney 2015)

9 N.Y. Comp. Rules & Regs. 6210.18

Traditional

Random selection of 3% of machines.

County Board of Elections, with oversight from the State Board of Elections.

Within 15 days of each general or special election and within 7 days of every primary or village election.

If there is an unresolved discrepancy of even a single vote, the manual count is conducted again. An expanded audit is required if discrepancies would alter the vote share by 0.1% or if discrepancies are found in at least 10% of the machines audited. When an expanded audit is required, an additional 5% of machines are audited, then an additional 12% if the discrepancy still exists, and if the discrepancy is still found, a full manual recount.

Each candidate, political party or independent body entitled to appoint watchers to a polling place is entitled to appoint watchers to observe the audit.

North Carolina

N.C. Gen. Stat. Ann. §163-182.1

Traditional

Manual count of the paper ballots or paper records of a statewide ballot item in every county; precincts are randomly selected. 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.

County election officials, with oversight from the State Board of Elections, which also selects the random sample.

Selection of precincts to be audited must take place at the latest 24 hours after polls close. A deadline for the audit is not specified.

In case of a discrepancy, the hand count shall control, except where paper ballots have been lost or destroyed or where there is another reasonable basis to conclude that the hand count is not the true count. If the discrepancy between the hand-to-eye count and the mechanical or electronic count is significant, a complete hand-to-eye count shall be conducted.

Not specified.

North Dakota

N.D. Cent. Code 16.1-06-15

Other

Random testing of the voting system programming for one precinct in each county in the state according to logic and accuracy testing procedures.

County election official.

Before the canvass.

Not specified.

Not specified.

Ohio

Secretary of State Directive 2017-14

Traditional, with risk-limiting audits recommended

Audit of at least 3 contests: the “top of the ticket” contest (i.e., President or Governor); at least one other statewide contest to be selected at random by the secretary of state’s office after Election Day; at least one non-statewide candidate contest to be selected by the board of elections. Board of elections determine the “units to be audited,” either by precinct, by polling location or by individual voting machine but are encouraged to audit the smallest unit available to the board. A sufficient number of units must be audited so that the number of votes cast on all selected units equals at least 5% of the total number of votes cast in the county. This includes all categories of ballots.

Local election boards

No sooner than 6 business days after the local election board certifies election results, and no later 21 days after certifying the official results of the election, unless a recount must be conducted. If a recount is conducted, the post-election audit must be completed no later than 14 days after certification of recount results.

A county is required to escalate the audit if its accuracy rate is less than 99.5% in a contest with a certified margin that is at least 1% (calculated as a percentage of ballots cast on which the contest appeared), or less than 99.8% in a contest with a certified margin that is less than 1%. Escalation entails drawing a second random sample of at least 5% of votes cast, selected from units that were not audited in the original sample, and auditing the ballots using the same procedures. If, after the second round of auditing, the accuracy rate from the two samples is below 99.5%, the county shall investigate the cause of the discrepancy and report its findings to the Secretary of State’s Office. In such cases, the Secretary of State’s Office may require a 100% hand-count.

The post-election audit must be open to the public, the media and to duly appointed observers.

Oklahoma

Okla. Stat. §26-3-130

 Traditional

Manual or electronic examination of a limited number of ballots.

County election boards, supervised by the secretary of the state election board.

Timing shall be determined by the secretary of the state election board.

Procedures shall be determined by the secretary of the state election board.

Not specified, but a report of the findings is available to the public.

Oregon

Or. Rev. Stat. §254.529

§254.535

Traditional, with a tiered system based on the margin of victory, or risk-limiting.

Counties may choose to conduct a traditional hand count poste-election audit that depends on margin of victory: if margin of victory is less than 1% of the total votes cast, 10% of all precincts are hand counted; if margin of victory is between 1% and 2%, 5% of all precincts; margin of victory 2% or greater, 3% of all precincts are hand counted. Or a risk-limiting audit.

County clerks, with oversight from the Secretary of State.

Traditional audits: begin no later than the 21st day after the election and completed by the 30th day after the election.

Risk-limiting: before an election contest is certified.

Traditional audits: If there is a discrepancy of greater than 0.5%, the sample is audited again. If the second audit show a discrepancy of 0.5%, all ballots for that system are audited.

Risk-limiting: rules and procedures to be determined by the secretary of state. 

Audits are conducted publicly and results made public.

Pennsylvania

Pa. Cons. Stat. tit. 25 §3031.17 §2650

Traditional hand count or electronic

Recount of random sample of the lesser of 2% of votes cast in each county, or 2,000 ballots.

County board of elections.

Part of the canvass.

Not specified.

Candidates and their appointed watchers and attorneys.
 

Rhode Island

§ 17-19-37.4

Risk-limiting

The state board shall determine what local, statewide and federal contests are subject to a risk-limiting audit. A risk-limiting audit shall begin with a hand tally of the votes in one or more audit units and shall continue to hand tally votes in additional audit units until there is strong statistical evidence that the electoral outcome is correct.

State board of elections in conjunction with local board of elections.

Before certification of results.  

In the event that counting additional audit units does not provide strong statistical evidence that the electoral outcome is correct, the audit shall continue until there has been a full manual tally to determine the correct electoral outcome of the audited contest.

Not specified.

South Carolina

Description of Election Audits in South Carolina

Other

The audit process compares the tabulated results of the election with the raw data collected in the electronic audit files by each voting machine on a flash card. The State Election Commission has developed a series of computer applications that compare the tabulated returns reports with the raw audit data. If the audit application detects an anomaly it lists it in one or more audit report.

County election officials or staff of the state election commission.

Before the county certifies the election, the county assures that all ballots have been counted, including accepted provisional ballots. A second audit is completed before the state election commission certifies the election.

Not specified

Not specified

South Dakota

None

N/A

N/A

N/A

N/A

N/A

Tennessee

Tenn. Code Ann. § 2-20-103

Traditional

Automatic audit of at least 1 precinct for small counties and at least 5 precincts for large counties. 

County election commissions.

Begins before 1:00 p.m. on the day after the election.

If after the automatic audit there is a variance of more than 1% between the unofficial election results of the top race and the automatic audit, a hand count of 3% of the precincts is conducted.

Audits are conducted publicly.

Texas

Tex. Election Code Ann. §127.201 (Vernon 2015)

Election Advisory No. 2012-03

Traditional

Not more than three races in 1% of precincts or three precincts, whichever is greater; does not apply to tabulation of DRE machines.

Local election officials, with oversight from the Secretary of State.

Begins within 72 hours after the polls close and completed no later than the 21st day after the election.

If there are discrepancies in the audit, the election official shall continue the audit until it determines the cause of the discrepancy.

Candidates or their appointed representatives

Utah

Election Policy Directive from the Office of the Lieutenant Governor, Utah Code Ann. §20A-3-201

Traditional

Vote-by-mail counties audit 1% or 1,000 mail ballots, whichever is less. Batches to be audited are randomly selected by the Lt. Governor’s Office (LGO). One accessible voting machine (DRE) per 100 deployed in every Utah House District, selected randomly by the LGO, are also audited.

Local election officials, with oversight from the Office of the Lieutenant Governor.

Before the canvass.

Election officials should ascertain and record the reasons for any differences.

Appointed poll watchers.

Vermont

17 Vt. Stat. Ann.

§2493

§2581 - §2588

Traditional

Determined by Secretary of State, with a mix of electronic and hand count auditing.

Secretary of State.

Within 30 days of the election.

Not specified.

The public may observe the process, and results are made public.

Virginia

Va. Code § 24.2-671.1

Risk-limiting

Post-election risk-limiting audit of ballot scanner machines. (Note that risk-limiting audit is not defined, nor is additional information on scope.)

Local board of registrars, with procedures provided by the state department of elections.

Annually, after the election has been certified and the period to initiate a recount has expired. An audit shall have no effect on the election results.

The local jurisdiction issues a report, which includes a comparison of the audited election results and the initial tally for each machine audited, and an analysis of any detected discrepancies.

Candidates and political parties may appoint observers.

Washington

Wash. Rev. Code Ann. §29A.60.185, §29A.60.170, Wash. Admin. Code 434-262-105

Traditional, with option of conducted a risk-limiting audit

County auditors choose (at a minimum) one of the following methods to audit duplicated ballots: 1) audit of DREs or ballot marking devices when at least 10 votes have been cast on all devices. Up to 4% of devices are selected by lot. 2) a random check 3) a risk-limiting audit whereby the scope and "risk limit” (the largest statistical probability that an incorrect reported tabulation outcome is not detected) are set by the secretary of state.

County auditors, with oversight from the Secretary of State.

Before the certification of the election.

For each audit method, the secretary of state must adopt procedures for expanding the audit to include additional ballots when audit results in a discrepancy, under what circumstances the discrepancy leads to an audit of additional ballots, the method to determine how many additional ballots will be selected, and how to investigate the cause of any discrepancy found during an audit.

Appointed observers.

West Virginia

W. Va. Code, §3-4A-28

Traditional

At least 3% of precincts, chosen at random.

 

Board of Canvassers

During the canvass, before certification.

If there is a discrepancy of more than 1% or that would result in a different outcome, it is immediately disclosed to the public and a full manual recount is ordered.

Audits are conducted publicly.

Wisconsin

Wis. Stat. Ann. §7.08(6)

Wisconsin Elections Commission 2018 Post-Election Voting Equipment Audit

Traditional

At least 5 percent of statewide reporting units (a minimum of 183 total reporting units). No more than two reporting units will be chosen per municipality, and municipalities selected as part of the audit will be chosen randomly. At least one audit will be conducted in each of Wisconsin’s 72 counties. Reporting units will be selected to include a sample from each piece of voting equipment that records and tabulates votes.

Local election officials, with oversight from the Wisconsin Elections Commission (WEC).

No later than 2 weeks after certification of results. 

Discrepancies are reported to the WEC. If the discrepancy cannot be reasonably explained, WEC staff will request that the voting equipment manufacturer investigate and explain the reasons for differences between the machine tally and the paper record tally. Should the vendor fail to provide a sufficient written explanation, including recommendations for preventing future occurrences, within 30 days of notification, WEC staff will suspend approval of the affected voting system in Wisconsin.

Audits are conducted publicly.

Wyoming

W.S. 22-11-104,

Wyo. Admin. Rules Secretary of State Election Procedures Chapter 25

Other

The pre-election logic and accuracy testing is repeated after the election. A random audit of ballots is conducted by processing the pre-audited group of test ballots on 5% of the automated tabulating machines for that county.

County election officials.

Within 30 days of the election.

If a voting machine fails the post election audit test, the county clerk notifies the secretary of state in writing of the audit problem.

The county clerk shall notify each political party of the time and date of the post election audit. Other witnesses may also be present.

This table is based on the U.S. Election Assistance Commission’s Statutory Overview, with additional research conducted by NCSL, and with assistance from the Verified Voting State Audit Laws Searchable Database.

Additional Resources

Contact NCSL's elections staff at 303-364-7700 for more information.