2024 U.S. ELECTIONS RAPID RESEARCH BLOG
There are often good explanations for anomalies and outliers in a dataset.
By Ashlyn B. Aske and Kate Starbird
Center for an Informed Public
University of Washington
This is part of an ongoing series of rapid research blog posts and rapid research analysis about the 2024 U.S. elections from the University of Washington’s Center for an Informed Public.
On the evening of March 31, an X user [1] posted a series of more than 10 tweets[2] containing data they attributed to the Washington Voter Registration Database (VRDB). The user identified anomalous voter registrations within the database, implying that these anomalies are suspicious and reflect untrustworthy election processes. The first tweet in the poster’s thread made a striking claim that over 200 people are registered to vote at an address that appears to be a small residential home. That tweet received significant engagement, garnering 121,300 views and 2,000 retweets by the following afternoon (April 1). Notably, several of the tweets in the thread alleging suspicious addresses were removed for violating X Rules on April 2, but the user maneuvered around those removals by posting new tweets that excluded the addresses in the tweet text but linked to Truth Social posts with the original content.
The implication that the alleged registration anomalies uncovered by the poster reflect election insecurity and/or “bad data” is false. The truth is that these anomalies where large numbers of people are registered to the same address almost always have simple — and in many cases mundane — explanations. According to Washington state law, people who are eligible to vote but do not have a residential mailing address — for example, people experiencing homelessness or people who live on boats moored in a marina — can register to vote at various locations where they can receive mail. These locations are typically businesses or organizations that provide mail services, resulting in a large number of people being registered to vote at a centralized address.
One of the primary ways that false election rumors take shape is through misinterpretations and/or mischaracterizations of “real” evidence. The tweet thread about anomalies in Washington state voter registration demonstrates two dimensions of those misinterpretations. The first is a data analysis error — i.e., people misinterpret and then mischaracterize data, including data that may be partial or imperfect. The second involves a lack of knowledge or confusion about election law and procedures. People misinterpret how the data relate to election processes, for example, assuming that many people registered at the same address indicates an error or fraud. These misinterpretations can later be compounded by selective and confusing presentation of the evidence used to support them.
In this blog post, we first provide some context on the history of claims misrepresenting voter registration data, discuss the intricacies of the claims in this particular tweet, and then briefly show how the rest of the poster’s thread follows the same pattern of misrepresentation.
Context: Rumors about voter registration data
Misinterpretations of voter registration data are common in rumors about election integrity. Narratives suggesting that voter registration data anomalies indicate untrustworthy elections and/or fraud take various forms.
Sometimes, voter registrations appear suspicious due to errors or flaws with registration data management systems. For example, in 2020, The New York Times reported on a viral social media post that claimed people in Michigan were voting under the names of dead people. The post showed data from the Michigan Voter Information System, which seemed to suggest that people over 100 years old were active voters. As it turned out, some birthdays were listed incorrectly in the database due to a computational error. Though the post did not uncover a fraudulent scheme to impersonate deceased voters, it did highlight real imperfections with the Michigan Voter Information System.
In other instances, seemingly suspicious irregularities may arise from an individual’s data analysis errors. In September 2022, KING5-TV in Seattle reported that earlier that summer, a group of individuals submitted a report of 239 alleged voter anomalies to Mason County, Washington officials, drawing from voter registration data analyses and canvassing efforts. Research errors explained some of the anomalies — e.g., the report claimed a family was registered to vote at an address for a vacant lot, but the writers had misread the address and made a wrong turn while canvassing. Here, the confusion does not arise from an error in the database itself but from an error made in the analysis.
Registrations may also appear strange to analysts unfamiliar with relevant federal, state, and local election laws. In the Mason County report, other noted anomalies were explained by misinterpretations of election law. For example, the report claimed an individual was registered to vote at an address where he no longer lived after joining the Coast Guard, implying a violation of Washington state law. However, the record was in compliance with the Uniformed and Overseas Citizens Absentee Voting Act, a federal law. The report’s writers, seemingly unaware of the relevant federal law impacting this person’s record, did not consider all possible legal explanations for an anomaly.
Election processes can be complex, and the data and procedures around them are resultantly nuanced. As demonstrated in these three examples, these complexities and nuances create space for rumoring, misinterpretations, and mischaracterizations of election-related data and procedures.
Case study: A misleading X thread about numerous registrations at single addresses in Washington state
The first tweet in the thread reported observing a high concentration of registered voters at a single address while querying the Washington State Voter Registration Database. The author stated that there are 208 active voter registrations at “77 ST S, Seattle, WA.” The tweet includes a photo of a relatively small residential home, implied to be located at this address. After several commenters mentioned that “77th ST S” is not a valid address, the poster corrected the address to “77th Ave S.”
Neither “77 ST S” nor “77 Ave S” are proper addresses — while the numeric street name makes it less obvious at first glance, both are missing the house number. “77 Ave S” refers to an entire street, and the house pictured is the Google Maps image result associated with searching for this street. Yet, the creator’s posts frame this as a single residential address.
A handful of commenters, including Garrett Archer, a journalist who frequently fact-checks election rumors, suggested that the original address may be a typo for a third address.[3] That third address is the location of an organization in the Pioneer Square neighborhood of Seattle that provides services for people experiencing homelessness. Our analysis of the poster’s dataset[4] supports this conclusion.
Figure 1: Garrett Archer, a journalist who frequently fact-checks election rumors, suggested that the original address may be a typo for a third address, an organization that provides services to people experiencing homelessness.
Under Washington state law, people who do not have a traditional residential address, such as those who are homeless, may register to vote at an address where they receive mail. Thus, it makes sense that a large number of people would be registered to vote at an address associated with an organization providing services to people experiencing homelessness; services likely to include a place to receive mail — homeless people can legally register to vote and need a mailing address to do so.
The poster makes several other similar claims about a large number of people registered to vote at single addresses, which demonstrate a similar misunderstanding of the law. They find it suspicious that over 160 people are registered to vote at each of two different locations that provide daily temporary housing and mail services for homeless people — but this is clearly permitted under the law.
Another tweet makes a similar claim that there are 144 registered voters at 7001 Seaview Ave. NW #160 in Seattle, which they described as “a mailbox store.” This address houses Dockside Mail, a business at the Shilshole Bay Marina where people living on boats moored in the marina can rent a post office box. Under the same law, residents living on boats can also register to vote where they receive mail.
Analyzing two different layers of misinterpretations
This thread illustrates two different layers of misinterpretations: First, the typo in the first tweet’s address and associated image wrongly suggests that a large number of people are registered to vote from a single residential home. Second, the implication that a large number of people registered to vote at any one address is suspicious is also incorrect, as it is legal under Washington law. These misinterpretations fit a larger pattern of erroneous voter registration database analyses, common in election integrity rumoring.
As we explain the rumors surrounding voter registration data in this blog post, it’s important to acknowledge that mistakes can happen. This applies to both election officials and analysts who study registration data. Errors in the data, misunderstandings about what the data mean, and confusion about election processes can all contribute to the spread of false rumors. Voter registration data may reflect the nuances of differing federal, state, and local laws. Exceptions for particular circumstances, like the nontraditional address provision, may not be well understood. For independent analysts searching for evidence to support an existing expectation of election insecurity and/or voter fraud, those expectations may contribute to misinterpretations — a dynamic our team explored in previous work on the influence of frames.
On social media, conditions that contribute to rumoring are exacerbated. Discussions cross jurisdictional lines, where laws and procedures differ, so claims about election processes that are true in one context may be false in another. Posts tend to be short, making it harder to convey the complex contexts voting data reflect. Content creators are motivated to produce and interpret “evidence” in ways that support existing political narratives, such as a “rigged election.” As such, when viewing claims about voter registration data on social media, users should keep in mind that there are often many legitimate explanations for why patterns and outliers occur in a dataset.
Notes
[1] We anonymize account names to protect the identity of accounts that may have a reasonable expectation of privacy. However, we do not anonymize public figures, professional journalists, verified accounts, and accounts that have massive online followings (>250k followers).
[2] A Note on Terminology: Following a common practice in academic scholarship, we refer to the different elements of X (tweets, retweets, likes, follows) by the terms used in the community itself. When referring to the platform itself or associated policies, we use the term “X,” the legal name of the platform.
[3] In line with media best practices and to protect privacy, we are not publishing the actual address or the name of the organization.
[4] The poster shared a link to the dataset they used to run their analyses in another post. We used the dataset they provided to run our own queries seeking to replicate their results. We found that the majority of the numbers they provided were roughly supported by the dataset they shared, but the dataset returned no results for querying either the “77 ST S” or “77 Ave S” addresses. Querying for the address shared by Garrett Archer did return slightly over 200 active registration records, as claimed by the poster. As we were not able to verify the source of the poster’s dataset and did not have access to official Washington voter registration data, we cannot make any representations about whether these statistics truly reflect Washington voter registration records.
- Ashlyn B. Aske is a CIP graduate research assistant and a Master of Jurisprudence student at the UW School of Law.
- Kate Starbird is an associate professor in the UW Department of Human Centered Design & Engineering and a Center for an Informed Public co-founder.
- Photo at top: A person places a ballot into an official ballot dropbox in King County, Washington. (via King County Elections / Facebook)
