Print Page | Contact Us | Sign In | Become a Member
Quarterly: Winter 2019 - Megan Ko and Brooke Leary

Auditing a vote-by-mail System: reducing risk, removing barriers

By Megan Ko and Brooke Leary


Every vote counts, but does every vote get counted? To make sure that democracy functions fairly, local elections offices work to make elections accessible for all eligible voters. This article discusses the methodology and results of a 2017 audit we did of elections in King County, Washington — which are 100 percent vote by mail. The article is organized based on our two audit objectives focused on risk and accessibility.


King County is the largest jurisdiction in the United States to conduct all of its elections by mail. As of 2019, there are three states — Washington, Oregon, and Colorado — that vote exclusively by mail, and 28 more states with vote-by-mail options. Even if your state or county does not vote by mail, our audit can provide useful lessons for identifying control deficiencies and removing barriers to voting.

King County voters receive their ballots in the mail at least two weeks before Election Day. Voters can cast their ballots as soon as they arrive. Election Day is the deadline for posting a ballot (no stamp needed) or returning it to one of more than 60 drop boxes located throughout the county. By allowing people more time and places to vote, vote-by-mail systems can remove barriers to voting such as transportation to a polling place and lost wages associated with standing in line.


Ahead of the 2017 general election, we set out to assess whether ballot processing by King County Elections had sufficient internal controls in place. Our first audit objective asked: To what extent does King County Elections have processes in place to make sure that elections are accurate, fair, and efficient? In other words, what was the elections department doing to ensure it counts votes as cast, and what gaps in control remained?

We relied on direct observation as a methodology. Our team of three auditors spent many hours watching ballot processing staff do their work — sometimes as early as 6 a.m. to see the first batch of ballots arrive at the processing center. To document any potential control gaps or weaknesses, we used a checklist we developed based on a review of the Green Book and took lots of photos.


The following pictures and text are three examples of control gaps we found, all relating to staff being alone with ballots.

Sole custody of uncounted ballots

While employees are directed to work in pairs on the ballot processing floor, we observed instances of employees working alone with envelopes that contained uncounted, freshly delivered ballots. The length of time an employee was alone tended to be brief and offer little opportunity to destroy a ballot. The mailroom, however, was a different story. We observed that a single employee would often be sent to get the mail, which included ballots that did not make it into the day’s bulk delivery. The room was visually separated and had a secure shred bin, which is designed to allow staff to drop paper in but not to look in or take things back out.

An employee works alone in the mailroom near uncounted ballots and a secure shred bin
Source: King County Auditor’s Office photograph of King County Elections’ mail room scanner

Opportunities to alter paper ballots

Ballot opening employees had pens since part of their work involves taking some notes. The elections department mitigates the risk that an employee would use their pen to change a vote by requiring staff work in pairs. However, we observed some staff alone at their work stations when their partner was on break.

Employees alone at workstations with open ballots and pens

Source: King County Auditor’s Office photograph of King County Elections’ ballot processing floor

Opportunities to alter votes

Once opened, paper ballots are fed into a scanner and uploaded into a program that translates the marks on the scanned ballots into votes. To make sure that votes are counted correctly, staff review some scanned ballots, such as when two selections are made in a race where only one selection is valid. For example, if a voter accidently picks two options in one race, they may cross one out to show which is their true vote. However, the scanner is not as capable as the human eye in differentiating between a cross out and an actual vote. Therefore, staff can overwrite how the computer logs a vote. The elections department mitigates the risk of improper vote alteration by having staff work at computers in pairs, spot-checking ballot adjustments, and having a supervisor on site to answer questions. We observed instances, however, where staff were on computers alone, increasing the risk that ballots could be modified without detection. Moreover, elections staff did not have individual computers logins and there was no recurring review of software logs for suspicious patterns of alteration (e.g., all votes changed the same way in a single race).

Employees alone at computers that can alter votes
Source: King County Auditor’s Office photograph of King County Elections’ ballot processing floor

Even with good staff, controls are necessary. Every elections staff member we interviewed was committed to ensuring that ballots were processed efficiently and accurately, and management could not recall instances of staff intentionally mishandling ballots. However, even within organizations that have a strong culture of accountability, people are people. We make mistakes and have our own internal drives that require organizations to have their own controls to ensure that ballots are counted as cast. We made recommendations to ensure ballots are not altered or left in the custody of a single person and to train staff on the criminal penalties for ballot mishandling.


Our second audit objective asked: To what extent has King County Elections identified and minimized barriers to voting among eligible voters? In other words, we wanted to know what barriers voters faced and what the elections department was doing about them.

Not sure where the barriers are in your community? By looking at U.S. Census data in conjunction with ballot return information, we were able to get a sense of which demographic characteristics corresponded to higher or lower voter engagement at the local level. (Less data-intensive approaches to finding barriers include interviewing stakeholders and reviewing federal voter protection laws enforced by the Department of Justice.)

We used data from the American Community Survey (ACS) and Public Use Microdata Sample (PUMS) to get demographic information on age, race, English language ability, disability, income, education, and veteran status. We then requested that our elections agency provide us with deidentified data disaggregated by census tract. Using elections data, we calculated for each census tract:

  1. Registration rate (the number of registered voters divided by the number of U.S citizens over age 18);
  2. Voter turnout rate (ballots returned divided by voters registered);
  3. Challenge rate (voters informed by KCE of an issue with their ballot divided by ballots returned); and
  4. Cure rates (challenged ballots accepted divided by total ballots challenged).

We created scatter plots to look for correlations between the demographic makeup of census tracts and their voter engagement rates. Each dot on a scatter plot was for a census tract. The y-axis showed one of the four rates listed above, while the x-axis showed a demographic characteristic. For example, for age we used the percentage of the population over age 65. For race we used percentage of the census tract who identified as people of color, and for income we used median income.

Our scatter plots showed that higher income was correlated with higher registration rates but not with signature challenge rates.
Source: King County Auditor’s Office analysis of King County Elections and U.S. Census data

Plots that looked like a mess of scattered dots indicated no strong correlations, while those that looked like upward or downward sloping lines indicated that some populations may face greater barriers to voting. Finally, we conducted linear regressions to determine whether the correlations were statistically significant.


Knowing what groups of people are underrepresented at the polls can help decision makers craft programs that reduce barriers to voting. One of King County Election’s priorities is to identify barriers to voting and to work actively to remove them at the individual and community levels. The elections department has a Disability Advisory Committee to help advise on access to voting for people with disabilities. During our audit we interviewed members of the committee and toured accessible voting centers, which offer audio and large print ballots. We also reviewed guidelines for making polling places accessible set forth by the Americans with Disabilities Act checklists.

Language access is also important. Section 203 of the Voting Rights Act requires certain jurisdictions to offer language assistance to people with limited English proficiency. Auditors can review requirements under this law and ensure that their jurisdiction is meeting the mark and whether there are compelling reasons to go beyond the federal minimums. In King County, Section 203 requires elections to provide materials in Chinese and Vietnamese in addition to English. The County recently decided to roll out Spanish and Korean materials as well. To understand whether there are areas for improvement, auditors can look at how these materials are distributed and disaggregate voter turnout by ballot language. For example, we found that 24 percent of voters who opted to receive a Spanish-language ballot voted in the 2017 general election, compared with 44 percent of voters who asked for a Korean-language ballot.

Uniformed and overseas voters are also a protected group who faced barriers to voting by virtue of being outside of the country. We found that in the 2017 general election, King County Elections received about 1,500 electronic ballots, with 71 percent of these ballots from military and overseas voters. Elections department staff accessed these ballots using a standard version of Microsoft Outlook. The email program did not record how many ballots it received or have technical controls in place to prevent the ballots from being deleted, increasing risk that uniformed and overseas voters will not see their vote counted. We recommended that elections work to ensure that emailed ballots are not deleted.


King County Elections has a number of strong controls in place to ensure that every vote gets counted as cast. However, even strong elections stewards can benefit from an outside look to help them spot what they might be missing and get more voters involved in the democratic process.


Megan Ko is a Senior Management Auditor with the King County Auditor’s Office (Seattle, WA) and a member of ALGA’s Diversity, Equity, and Inclusion (DEI) Committee. Megan grew up in Iowa and Texas and worked in the private and nonprofit sectors in mainland China and Hong Kong before earning a Master of Public Administration from the University of Washington. She is a proponent of equitable, effective, and transparent governance.

Brooke Leary is a supervisor with the King County Auditor’s Office in Seattle, Washington. In addition to her audit work on elections, she has audited topics including law enforcement oversight, sheriff’s office overtime, and employee domestic violence policy. She has worked on three Knighton Award-winning audits. Brooke began her auditing career with the U.S. Government Accountability Office in its Seattle field office and made the move to local government in 2013.