Audit the Vote gave us its canvassing data to check the results. It was riddled with errors

Entry type: Single project

Country/area: United States

Publishing organisation: LNP | LancasterOnline

Organisation size: Big

Publication date: 2022-05-08

Language: English

Authors: Carter Walker, reporter
Russ Walker, editor (not related)
Stephanie Zeigler, managing editor
Tom Murse, executive editor


Carter Walker is an investigative journalist in Pennsylvania whose work focuses on government and politics. He currently covers voting and elections for Votebeat, in partnership with SpotlightPA. Prior to that he worked as an investigative reporter focused on extremism, misinformation and other threats to democracy at LNP | LancasterOnline.

Project description:

This story obtained and examined data from an election denial group which was holding up the data as proof of fraud in the 2020 presidential election. Audit the Vote PA had conducted a door-to-door voter canvassing effort in search of fraud and had claimed to have discovered “anomalies” in the election. The group, however, did not release its underlying data. We convinced the group to share the data with us. Our examination of the data and investigation of the canvassing effort revealed severe flaws in the effort. Our story also remains the only public release of this data.

Impact reached:

This reporting was cited by The New York Times, Washington Post, Bloomberg, Politico, NPR, Slate, Huffington Post, Vice, The Philadelphia Inquirer, and in a timeline by the Pennsylvania House Democrats. The story became particularly important once it became widely speculated that Audit the Vote’s leader, Toni Shuppe, might have been Doug Mastriano’s pick for secretary of state. It also served to inform Pennsylvania state officials and county leaders as they considered the veracity of these and other claims made by Audit the Vote PA. The report was also cited in Maryland (The Baltimore Sun) as another group of election skeptics proposed a similar effort.

While it is impossible to prove a negative, we also suspect the story contributed to the prevention of these claims gaining traction locally. Similar claims were made in other counties, most notably Luzerne. In that county there was not the same level of media scrutiny of these claims, and the county eventually agreed to a hand recount of the 2020 election, costing thousands of dollars and several weeks worth of man hours for very little change in outcome. Carter heard personally from local GOP party members in Lancaster who were pleased the group’s claims were being thoroughly fact checked, and some local politicians became more willing to call out the group’s misinformation following this report. The type of traction Audit the Vote gained in Luzerne was not seen in Lancaster.

Audit the Votes claims represent a serious accusation against our elections system, yet they have never released their underlying data. Our story remains the only release of this data to date.

Techniques/technologies used:

Using a copy of Pennsylvania’s voter roll, Audit the Vote PA conducted a door-to-door canvassing effort asking voters a series of questions about the 2020 election. The dataset provided to LNP | LancasterOnline by Audit the Vote PA did not contain the names or addresses of the voters they surveyed. These columns had been removed before it was provided. But, information contained in a “notes” column contained enough personal information that, when combined with a column referencing a voter’s “municipality,” could be used to identify several voters by cross referencing this information in the state’s voter roll. The Audit the Vote PA dataset did not contain unique identifiers, so an SQL join could not be done on the dataset. Carter had to match the voters from one dataset to another manually. This sometimes resulted in multiple matches on the same voter. To find the correct voter he had to visit multiple homes in some cases. Once Carter had found as many voters as possible from the Audit the Vote PA data, he then reinterviewed those voters and showed the significant errors volunteers made when questioning voters and recording their responses. This involved multiple days of driving around the county to visit homes spread out over a large area.

The dataset also provided unclear information about whether the voters surveyed were registered in 2020 at the homes they were living in when Audit the Vote conducted the survey in 2021. Using OSINT investigative techniques, Walker found a photo of one of the original surveys on social media and used it to determine that the question asked was if the person surveyed was registered to vote in 2020, but not if they were registered to vote at this address. This represented a significant misstep in the survey process.

Context about the project:

Following the 2020 election, grass roots efforts to prove fraud popped up in several states. The chief group pursuing fraud in Pennsylvania was Audit the Vote PA, run by Toni Shuppe. After an unsuccessful petition campaign to try to force the state to perform a forensic audit, in the fall of 2021 the group turned its effort toward a door-to-door canvassing effort in the hopes this would prove the fraud they suspected. Audit the Vote was very actively lobbying state and county governments for election concessions, and had the ear of some elected officials, most notably state senator and GOP gubernatorial candidate Doug Mastriano. Once they had completed their canvassing effort, they began publicizing the results to elected officials and voters all around Pennsylvania claiming they had found an enormous amount of “anomalies” in the election and sought to use the results to leverage election administration concessions. Some outlets, including LNP | LancasterOnline, reported that these “anomaly” claims were being made, but until our report taking a deep look at the canvassing data and process, no outlet had critically examined these claims. Getting the data was not easy. Though Shuppe had directly challenged LNP | LancasterOnline to look at it for ourselves, it still took several weeks for us to convince her to actually provide it. Shuppe was also rumored to be Mastriano’s pick for Secretary of State, which would have given her oversight on elections across the state. This story served as an important fact check on her knowledge and qualifications for that position.

What can other journalists learn from this project?

I think this story should serve as a reminder to journalists to not be intimidated by hostile sources and to remember that our job is not simply to blindly present both sides of an argument, but rather to critically examine claims and test their validity. Prior to this story I had written about this group’s claims and tried to make contact with the group’s leader so that she could validate what they were saying. She did not respond, but posted on social media later challenging me to examine her data. Even though I knew this source was likely to be hostile to me and possibly deride me publically regardless of what my story found, I decided to take her up on her challenge. I interviewed her and examined her data in good faith, yet found many errors which I was obligated to inform the public of. I was correct in my suspicions, and she has disparaged me to her audience ever since. I have even been doxxed on social media by her followers. However, our job is not to quote one person who says it is raining and one who says it is not. Our job is to look out the window and find the truth. I am proud of myself and my editors for doing this story, especially because I did not see critical examination of their claims being done elsewhere.

Project links: