Three Counties in Pennsylvania

This report is an analysis of election results in Pennsylvania, with a focus on in three counties:

• Philadelphia County
• Allegheny County
• Erie County

These three counties were selected based on multiple factors, including but not limited to:

• Data Availability: In Pennsylvania, election results are reported at both the state and county level – but with different degrees of granularity. Pennsylvania does report precinct-level data at the state level, but this data is not separated by vote type. While some counties report precinct-level data, others do not. Philadelphia, Allegheny, and Erie Counties all report precinct-by-precinct data separated by vote type.
  
• High Population Counties: In any analysis, having more data allows for a better understanding of the topic at hand. In this instance, Philadelphia and Allegheny have over 1,700 and 1,300 precincts, respectively. This high number of data points for us to analyze improves our confidence in these results.
  
• Bellwether County: When a smaller geographic area (e.g. a county) is seen as likely to vote in alignment with a wider geographic area (e.g. a state or country), that area can be called a “bellwether” or “electoral bellwether.” There have been 25 statewide Pennsylvania elections since 2008; in 23 of 25 of those elections, Erie county has voted for the state-wide winner. (Source: Mercyhurst University.) Bellwether counties are therefore observed more closely during and after elections to assess whether it has lived up to its reputation.  

In addition to Philadelphia, Allegheny, and Erie counties, our team has undertaken preliminary analysis of many other Pennsylvania counties to investigate voting patterns. This process can be challenging, as counties report their voting data individually – meaning that the formats and data structure are not standardized. Multiple members of our team have worked to verify that the data sets used in this analysis are accurate to data sets posted by state and local governments. 

-

Pennsylvania reports voting data returns for each race broken down by the voting type. Individual counties each report precinct level breakdowns as well. Unlike Clark County, NV, which reports voting results broken down by tabulator, Pennsylvania counties only report their data by precinct. This means: 

• The counties with fewer precincts provide fewer data points than larger, more populated counties; and,
• Data sets provided by Pennsylvania are not as detailed or in-depth as tabulator-level data would be. 

For each of the three counties examined (Philadelphia, Allegheny, Erie), this analysis drills down into two of the three vote types:

• Mail-In Voting (which includes “On-Demand Mail-In Voting”); and,
• Polling Place (i.e. Election Day)

Because Provisional Ballots make up only 0.98% of ballots cast for President in Pennsylvania, they are excluded from this analysis.

Mail-In Voting

 Pennsylvania has unusual laws regarding how mail-in ballots can be cast; specifically, what counts as part of the ‘Mail-In Voting’ category is slightly different in Pennsylvania than in other states. 

Across Pennsylvania, mail-in ballots can be submitted one of two ways: 

1. By Mail: Like other states, residents can request a mail-in ballot to be mailed to them. When received, they fill out their ballot and seal it in a secrecy envelope, which is then inserted into an outer envelope that the voter then signs. Once completed, they can turn in their ballot by mail, at a drop box, or other certified location.
2. On-Demand Mail-In: Residents can also vote by mail at a county elections office. This process allows voters to request a mail ballot, complete their mail-in ballot, and turn it in at a county elections office on that same day. 

This means that a portion of ballots considered to be “mail-in ballots” in Pennsylvania were submitted using a process that is roughly equivalent to Early Voting in other states. Absentee ballots, which are similar to mail-in ballots in Pennsylvania but require the applicant to provide a reason (such as disability or illness) they cannot physically access the polls, are generally grouped together with mail-in ballots. 

By law in Pennsylvania, mail-in ballots cannot be counted at polling places but must instead be processed and counted in a centralized location. County officials are not allowed to open and begin counting mail-in ballots until Election Day. The processing and counting of mail-in ballots is conducted with bipartisan teams of canvassers present. (Source: Election Facts PA) Envelopes are opened and steps are taken to open and count ballots, removing those with abnormalities. Potential abnormalities may include stray markings on the envelope used to ensure the vote cannot be linked to the voter’s name or address (the ‘secrecy envelope’) as well as ballots that were mailed without a secrecy envelope at all. 

Per Verified Voting, paper mail-in ballots are then tabulated at a central location using a batch-fed optical scanner. 

Because of the strict procedures surrounding the counting of mail-in votes, it is a relatively secure process that would be more challenging to manipulate than electronically-tabulated vote totals. (Source: PA Government)

Issues with mail-in ballots failing to arrive on time for voters to cast them by mail and lack of clarity around required mail-in ballot paperwork resulted in some last-minute changes to voting rules in Pennsylvania, potentially resulting in an up-tick in provisional ballots.

1,928,593 mail-in ballots were cast and counted in the 2024 presidential race in Pennsylvania.

Election Day ("Polling Place")

In Pennsylvania election data, ballots that were cast in-person on Election Day are cast at “polling places” – as opposed to Mail-In Ballots, which are turned into county election offices. In official Pennsylvania election result data, Election Day votes are captured under the category “Polling Place”. 

Voters are required to vote at the specific polling place where they are registered to vote. Polling place votes can be cast by:

1. Hand-marked paper ballot, which is then tabulated electronically, such as by physically feeding the paper ballot into a scanner (tabulator); or,
2. Ballot-marking device, an electronic system (such as a touch screen) that the voter uses to select candidates, print the ballot, and then feeds the paper print-out into a scanner (tabulator). 

The exact equipment used to process votes on Election Day in Pennsylvania is listed by county on the Pennsylvania government website. Votes cast on Election Day are electronically tabulated, not counted by hand. If there are grounds for a recount, each County has its own rules surrounding how that shall be handled. In most cases, this is done by using different tabulator machines than those used on election day, and in some counties a hand recount can be done; however, it isn’t always required.

4,969,749 Polling Place (Election Day) ballots were cast and counted in the presidential race in Pennsylvania, representing the majority of votes in the state. 

Provisional Ballots

In Pennsylvania, a “provisional ballot” records a vote and allows the county board of elections to determine after the fact whether or not that vote can be counted. There are several circumstances in which a Pennsylvania voter may have the option of being issued a provisional ballot.

This includes, but is not limited to:

• Name not in the poll book or supplemental poll book.
  • For example, voter reported to the wrong precinct; or
  • Voter did not report a recent change in residence to the county election office.
• Voter could not show the required ID. 
• Voter eligibility was challenged by an election official.
• Voter was issued an absentee or mail-in ballot but believes they did not successfully vote that ballot, and the voter did not surrender that ballot and outer return envelope at the polling place to be spoiled.
• Voter returned a completed absentee or mail‐in ballot that was rejected.
• There is a special court order with respect to a voter’s registration status.
• There is a special court order related to extending the hours of voting.

Due to issues with mail-in ballot requirements and mail-in ballots failing to arrive in time, more voters than usual may have needed to cast a provisional ballot at a polling place on Election Day.

Despite this potential increase, approximately 68,008 provisional ballots for Harris and Trump were cast and deemed by county officials to have counted in Pennsylvania for the 2024 U.S. Presidential Election – representing just 0.97% of all votes cast in Pennsylvania (or 0.98% if third-party candidates are included). Given that this represents only a fraction of Pennsylvania votes, Provisional Ballots are excluded from this analysis. 

Pre-Election Lottery
In October 2024, the America Political Action Committee (America PAC) launched a tour of Pennsylvania, during which America PAC creator Elon Musk promised to give away $1 million USD daily to registered voters in swing states in exchange for their name, street address, and phone number on a petition. (Source: AP News) The language in the petition stated that its purpose was to indicate support for the right to free speech and to bear arms, the first two amendments to the U.S. Constitution. 

Musk promised to pay people $100 for signing the petition, and an additional $47 for every additional referred signatory. (Source: Forbes, Newsweek) Being registered as a voter in a swing state was a requirement for signing the petition.

Ultimately, Musk’s lawyers admitted in court that the winners of the “sweepstakes” were pre-selected. (Source: CNN) With no actual element of chance, it was therefore not a lottery. Philadelphia District Attorney Larry Krasner characterized the petition’s signatories as being “scammed for their information”. (Source: PBS)

The America PAC FAQ emphasized the importance of signatories names and addresses matching, exactly, the name and address on their voter file. Failure to provide an identical name and identical address to match the signatory’s voter file would result in the potential signatory receiving notification by email of a mismatch, and an opportunity to correct the information to ensure a perfect match.  

The underlying intent for collecting this information remains unclear. One potential use case for this data is the micro-targeting of swing state voters with advertisements and/or get-out-the-vote efforts. This would not explain the insistence from America PAC that names and addresses must exactly match the voter file, to the point where a correction would have to be submitted by the signatory in order to be eligible to receive funds. 

Another potential use case for which name/address exactly matching a voter file would be important is for the purpose of falsely inflating voter records in order to conceal or obfuscate artificially inflated vote counts. As 2020-2024 election security breaches implicated electronic pollbooks (also known as E-Poll Books), which contain voter registration information and are used by election officials during the voting process.  

Bomb Threats
Across Pennsylvania, over 30 bomb threats were sent to election offices, courthouses, and other election-related buildings, almost entirely via email accounts linked to Russian email domains. (Sources: FBI, USA Today Network) These threats, all deemed false, forced officials to move election materials and workers to other facilities. This hampered efforts to efficiently and securely conduct the election counting processes. 

These types of disruptions are a particular concern for election proceedings, as having to evacuate a polling or ballot-counting location can result in election equipment (e.g. e-Poll books, ballot-marking devices, scanners/tabulators) and/or actual ballots being left unattended for extended periods on election day.

There were only a few instances in Pennsylvania on Election Day 2024 when the bomb threats resulted in court-ordered extensions of voting hours. 

Machine Failures
In addition to disruptions caused by bomb threats, Pennsylvania counties reported a high volume of voting system malfunctions on Election Day, resulting in delays and other impacts on both casting and counting votes. (Sources: Politico, Cybernews, WTAJ)  A large portion of the error reports concerned ballot scanner/vote tabulator machines failing to scan ballots. 

In Cambria County, ballot-printing errors have since been determined to be a key factor in machine failure to scan ballots, leading to last minute reprinting of ballots. Crawford County reported at least four instances of machines failing to read memory cards. Issues with tabulating scanners occurred in multiple counties, including Delaware, Elk, Lycoming, Montgomery, Westmoreland, and York counties. (Source: All County Voting System Report, 2024) 

On Election Day in Pennsylvania, the same machines that scan the ballots are frequently used to tabulate them. (Source: Pennsylvania Voting Equipment by County) Generally, when a voter submits their completed ballot for scanning, digital images of the ballot are created and stored on memory cards that are subsequently collected from machines after the election and used to tabulate official results. The scale of disruption caused by voting equipment errors varied across the affected counties, ranging from minor delays to – in the case of Cambria County – extended voting hours and the decision to duplicate ballots rather than undertake a hand recount. (Sources: NBC Philadelphia, Altoona Mirror)

Of the counties examined in this data package, only Philadelphia reported Election Day failures of their electronic voting infrastructure. These failures appear to have had minimal reported impact, as polling locations that experienced failures had additional machines functioning properly. (Source: All County Voting System Report, General Election 2024)

Malfunctioning voting equipment does not necessarily mean that these malfunctions were nefarious or intentional. However, any kind of disruption to electronic voting equipment can provide a window of opportunity for malicious actors. 

J. Alex Halderman, professor of computer science and engineering at the University of Michigan and one of the foremost election security experts in the U.S., provided the following Congressional testimony in relation to Russian interference in the 2016 U.S. Presidential election: 

“Both optical scanners and [direct record electronic] voting machines are computers. Under the hood, they’re not so different from your laptop or smartphone, although they tend to use much older technology—sometimes decades out of date. Fundamentally, they suffer from security weaknesses similar to those of other computer devices. I know because I’ve developed ways to attack many of them myself as part of my research into election security threats. [...]

Cybersecurity experts have studied a wide range of U.S. voting machines—including both [direct record electronic voting machines] and optical scanners—and in every single case, they’ve found severe vulnerabilities that would allow attackers to sabotage machines and to alter votes.” (bold added)

Ultimately, about 30% of counties reported machine errors. As this reporting requirement was introduced in 2023, there are no past presidential election records to compare with 2024. (Source: PBS) 

A full PDF of all post-election machine error reports can be found here on the Pennsylvania government website. 

“Vote-Counting Computers”
The Election Truth Alliance did not originally plan for Pennsylvania to be the second “swing state” of focus, in part due to:

1. Data Limitations: Limitations with Pennsylvania’s reporting of election result data (described in more detail below); and,
2. Other Priority Swing States: Other swing states appeared to demonstrate more conspicuous indicators of potential vote manipulation when looking at all vote types (e.g. Mail-In, Election Day, etc)

However, comments made by then-President Elect Trump at a pre-inauguration rally in Washington, D.C. on January 19, 2024 – specifically, a reference made to Elon Musk’s familiarity with “vote-counting computers” and “winning Pennsylvania like in a landslide” – prompted a shift in our focus. (Source: C-SPAN, ETA YouTube)

In our initial analysis of Clark County, Nevada, robust and comprehensive election result reporting from the Clark County Election Department – including a “Cast Vote Record” down to the tabulator level – allowed the ETA to identify trends consistent with potential vote manipulation in data from tabulators that had processed over a certain number of ballots. As such, the remark about “vote-counting computers” in a different swing state seemed particularly noteworthy. 

These remarks prompted us to dive into Pennsylvania election result data, including data related to specific vote types.

In these three counties, Harris received 1,065,943 votes to Trump’s 496,772 votes. As these counties contain major cities (including Philadelphia and Pittsburgh), this is in line with the general expectation that Democratic candidates receive more votes in urban areas. This is backed by research from Pew Research Center.

However, the drop-off votes for Harris in these three counties account for over 80% of her total drop-off votes in the state, even though she only received about 31% of her total votes in the state from those counties.

We have separated the drop-off rates for the three counties into two charts to showcase the difference between Mail-In and Election Day votes.  

-

 Hypothesis 1: Different Voting Types = Different Voting Behavior

• In most states, mail-in voting has a degree of flexibility. In their own homes, voters may take more time to research down-ballot candidates and fill out their preferences. This could in theory lead to a higher rate of full ballot completion vs. on Election Day, when in-person lines and perceived time pressure could result in a more hastily filled-out ballot. 

• This hypothesis is complicated by Pennsylvania’s “on-demand mail-in” model. As regular mail-in and on-demand mail-in ballots are combined in Pennsylvania’s reporting, it is unknown how many mail-in ballots were cast in county election offices with a similar perception of time pressure.

 Hypothesis 2: Candidate Popularity Difference

• An enthusiasm gap or popularity gap between the relevant candidates can result in a higher or lower drop-off rate. 

• Some single-minded voters may choose to vote exclusively at the “top of the ticket” for their favored candidate, or leave certain races blank if voters determine that a given candidate is not deserving of their vote. Split-ticket voting to avoid voting for an undesirable candidate can also widen the gap between Presidential and down-ballot candidates of the same party. 

• Conventional wisdom in U.S. politics is that Republican voters (and/or Trump voters specifically) are more likely to vote in higher numbers for the top of the ticket than Democratic voters. 
  • This trend is present in Pennsylvania in 2008 and 2012 when John McCain and Mitt Romney had a higher drop-off rate than Obama 2008 or Obama 2012. 
  • However, this trend reversed in both races in which Trump ran for president, where Trump 2016 and Trump 2020 both had a lower dropoff rate than Clinton or Biden. 

 Hypothesis 3: “More Republicans Just Turned Out on Election Day”

• This hypothesis may be initially appealing in that it appears, at first, to align with the expectation that Democrats are broadly more likely to vote by mail and Republicans are more likely to vote in-person on Election Day. 

• However, drop-off is about the difference between votes received by the Presidential candidate and votes received by the next down-ballot candidate of the same party. More Republicans voting on Election Day does not necessarily equate to more voters either casting a ballot for Trump in isolation, or splitting their tickets to vote for Trump as President and Casey for Senator. 

• While Election Day voters in theory could be comprised of more zealous Trump enthusiasts, resulting in a high drop-off rate for Trump, this does not align with the trends in the last two presidential elections where Trump himself was a candidate. In 2016, he received 0.63% more votes than the down-ballot Republican. In 2020, he received 1.59% fewer votes than the next down-ballot Republican.  

Hypothesis 4: Vote Manipulation With Specific Impacts to Presidential Candidates on Election Day

• To be clear: in isolation, a difference in drop-off rates between candidates is not necessarily indicative of manipulation. 

• However, as highlighted above, it is a known pattern that can emerge “if protagonists of fraud fail to interfere into multiple races to equal extents”. (Source: Behrens 2023) 

• Unexpected differences between ‘top of the ticket’ and down-ballot races can be one of the simplest and earliest indications that election results may warrant further scrutiny. 

• Drop-off analysis is best used as one vector of analysis among many.