Tag: statistical analysis of voting

Bonus Ballots for Biden: A Case Study of Election Lawfare Excess

Aug 27th, 2024
As the 2024 Presidential campaigns ramp up, election lawfare is underway as well, with the political parties seeking to exploit ambiguities in voting rules or bend them in their favor, and we can only hope that its role will remain limited and not affect trust in the election. This article highlights the problematic nature of election lawfare, focusing on a controversial case from the 2020 Presidential election in Pennsylvania.

In 2020, in states with newly expanded mail-in voting processes where the parties battled in court over what would be permissible under the new rules and what would not be, with Pennsylvania a particularly active lawfare battleground. The case discussed and analyzed here resulted in extension of Pennsylvania’s originally mandated deadline for receipt of mail-in ballots, and likely had a material effect on the final vote tally.

The analysis, which is in part data-driven, calls into question the primary, stated rationale for extending the deadline, which was to ensure sufficient time for return of mail-in ballots for those who had been relatively late in requesting a mail-in ballot. It also offers a rough quantification of the net vote gain for Joe Biden arising from the extension.

While this number may have been material–possibly as many as 9,000 votes, it was far from enough to swing the election. Nonetheless, in combination with other perceived shortcomings of the process such as lax security around ballot drop boxes, the case contributed to a broader erosion of trust in the election outcome.

The Democrats’ legal challenge to the ballot return deadline

No-excuse mail-in voting was introduced into Pennsylvania by PA Act 77, which was legislated in 2019 with bipartisan support in the state legislature and the blessing of the Democrat governor (Tom Wolf). Although the Democrats in the state’s General Assembly in 2019 mostly opposed Act 77, the party had a change-of-heart by the summer of 2020. Conversely, at the national level during 2020, President Trump and other Republican leaders increasingly were expressing fears that no-excuse mail-in voting might facilitate election fraud.

Observations on voter turnout during the 2020 Primary in Pennsylvania might have helped motivate the Democrats’ embrace of mail-in voting. No-excuse mail-in ballots accounted for almost 40 percent of total votes in the 2020 Primary Election, underscoring the preference of many voters for the mail-in option. Although data on mail-in ballot shares by party are not publicly available, it is a reasonable presumption that Democrats in the Primary were more likely to vote by mail than Republicans, as turned out to be the case in the general election, and that this fact was known to strategists in each party. ¹

Moreover, it seems that, because of the perception that liberalized vote-by-mail favored their party, Democrats were not content with a literal or narrow reading of Act 77. They adopted a strategy of lawfare to try and stretch the limits of mail-in voting as far as possible in order to maximize their perceived vote potential.

On July 10, 2020, the state Democratic Party, joined by some individual Democratic officials and candidates, filed a petition for review in the Commonwealth Court against Secretary of State Kathy Boockvar and all 67 county boards of elections. The petition requested declaratory and injunctive relief “to protect the franchise of absentee and mail-in voters.”² Specifically, the plaintiffs asked the Court to rule in favor of a number of requested flexibilities, among which was to grant an exception to the deadline in Act 77 for receipt of mail-in ballots, such that any ballot postmarked by 8:00 p.m. on election day (November 3) and received by 5:00 p.m. on Tuesday, November 10 can be counted.

Technically, Boockvar was an opponent of the petitioners in the case. However, she herself was a Democrat, and on August 13, she filed a response that objected only to the requested seven-day extension of the received-by deadline, primarily on logistical grounds, proposing instead a three-day extension. Her response strongly concurred with the petitioners’ other requests.

On August 16, Boockvar filed an application requesting that the state’s Supreme Court exercise extraordinary jurisdiction over the petition for review; that is, take over for the intermediate appellate Commonwealth Court, in order that the resolution of these matters be expedited to preclude any disruption of the 2020 General Election. On August 19, the petitioners filed an answer to the Secretary’s application, indicating no objection to the Supreme Court exercising extraordinary jurisdiction. The Pennsylvania Supreme Court, where 5 of the 7 justices are Democrats, then took over the case.

The state Supreme Court ultimately agreed to an extension of the ballot receipt deadline by three days so that any ballot cast by 8 p.m. on Election Day, (Nov. 3, 2020) and received by 5 p.m. on Nov. 6 would be counted. The Court’s decision, which was based on majority opinion, was announced on September 17.

Review and critique of the petitioners’ arguments

The petitioners had argued that special circumstances arising from the Covid-19 pandemic justified a seven-day extension to the received-by deadline for mail-in ballots. Essentially, they contended that, due to the preference of many voters to stay away from polling places on account of the ongoing pandemic, the volume of mail-in ballot applications would likely overwhelm county resources, such that bottlenecks and delays in the processing of ballot applications and distribution of ballots would likely arise. Secondarily, they noted the potential for delays in the United States Postal Service (USPS) delivery system as an aggravating factor.

They argued that under the state constitution’s Free and Equal Elections Clause, which prohibits interference in or prevention of “the free exercise of the right to suffrage”, the state was obligated to take action to remedy the anticipated disruption to mail-in voting. As described in the Court’s written opinion:

Petitioner interprets this provision as forbidding the Boards from interfering with the right to vote by failing to act in a timely manner so as to allow electors to participate in the election through mail-in voting.

A key component of the petitioners’ argument was a claim that the kinds of obstacles to the timely processing of ballot applications that were encountered prior to the June 2^nd Primary Election would recur and be even more severe. As summarized in the Court’s written opinion:

Petitioner avers that the difficulties encountered by Boards processing the ballot applications prior to the June Primary will only be exacerbated in the November General Election. It emphasizes the continued grip of the pandemic and a potential second wave of infections…. It highlights that the Secretary estimates that 3 million electors will seek mail-in or absentee ballots for the General Election in contrast to the 1.5 million votes cast by mail at the Primary.

However, the petitioners offered no substantive analysis to support this claim—only gripes about difficulties during the Primary and a poorly supported, inflated estimate of increased processing burden for the General Election.

During the Primary, many counties had indeed been inadequately prepared for the unexpectedly large volume of ballot applications they received, as this was the state’s first encounter with no-excuse mail-in voting. Thus, in some counties, most notably in Bucks and Delaware Counties, a significant number of ballots were mailed out too late for voters to return them by the deadline.

But clearly, county boards would have learned from their experiences in the Primary and taken action to shore up their capabilities in anticipation of needs for the General Election. The Court did not adequately consider what remedies might already have been implemented or what preparations were currently underway.

At least anecdotally, such ongoing efforts were substantive. A press report in late August contained the following assessment:³

Democratic leaders also contend they’ve remedied the issues that plagued them in the June primary when election officials were overwhelmed by a deluge of mail-in ballot requests during the coronavirus pandemic. It was the first time the state had conducted a mail-in election since a new law authorized the procedure last year.

For example, according to this report, Montgomery County had “used federal funds from the Cares Act to add eight staffers and more technology to help process the vote faster” since the Primary.

Another press report in early September included the following snippet:⁴

We were learning on the fly from February to June,’ says Jeff Greenburg, the director of elections in Mercer County until August, when he stepped down to work for The National Vote at Home Institute. ‘I really think Pennsylvania is in a better position now. … To me, we have a much better chance of succeeding because we now know what to do.’

Not only did the petitioners fail to back their claim that county boards remained inadequately prepared, but they overstated the anticipated increase in application processing volume. As evidence of increased burden, the petitioners had alluded to the gap between 1.5 million “votes cast by mail” in the Primary and 3 million anticipated applications for mail-in or absentee ballots for the General Election. This comparison, referenced in the Court’s written opinion, is misleading, as it is not apples-to-apples. The 3 million applications anticipated for the General Election is more appropriately measured against the total number of applications for mail-in or absentee ballots received for the Primary, which was 2 million.

Even that corrected numerical comparison, however, materially overstated the potential increase in processing burden, because much pre-processing of applications had already been completed. In fact, about 1.1 million individuals who had requested and been approved for a mail-in ballot in the Primary had signed up to be automatically sent a mail-in ballot for the upcoming general election, a highly pertinent fact missing from the discussion.

No further processing was required for the latter requests aside from the eventual automated printing and mailing of the ballots. Netting these out from the projected 3 million implies 1.9 million new applications for the General Election, not surpassing the 2 million processed for the Primary.

Review and critique of Boockvar’s arguments

Secretary Boockvar did not concur with the petitioners’ claim regarding the lack of preparedness of county boards, nor their request for a one-week extension. Instead, she advocated for a three-day extension based on the contention that those applying for mail-in ballots on or close to the statutory deadline for doing so, October 27, would face a high likelihood of missing the receive-by deadline for returned ballots if they chose to return them by mail.

The USPS delivery standards in effect at the time specified 2-5 day delivery for domestic First-Class Mail and 3-10 day delivery for domestic Marketing Mail. As described in the Court’s written opinion, “the USPS had advised that voters should mail their ballots no later than October 27 in order to meet the received-by deadline”. Boockvar argued that under these standards, a voter applying for a mail-in ballot on Oct. 27 would face a substantial risk of not meeting the November 3^rd deadline:

…even if a county board were to process and mail a ballot the next day by First Class Mail on Wednesday, October 28th, according to the delivery standards of the USPS, the voter might not receive the ballot until five days later on Monday, November 2nd, resulting in the impossibility of returning the ballot by mail before Election Day, Tuesday, November 3rd…This mismatch [between the USPS’s delivery standards and the Election Code deadlines] creates a risk that ballots requested near the deadline under state law will not be returned by mail in time to be counted…

Further, Boockvar argued that mail delays associated with the pandemic would make the problem worse, selectively disadvantaging those who are applying for mail-in ballots shortly before the October 27 deadline and therefore necessitating a one-time extension of the received-by deadline:

The Secretary emphasizes that the remedy sought here is not the invalidation of the Election Code’s received-by deadline, but rather the grant of equitable relief to extend temporarily the deadline to address ‘mail-delivery delays during an on-going public health disaster’.

This argument, however, is specious. To begin with, it rests on the illusory notion that the application and received-by deadlines in Act 77 were originally flawed because of inconsistency with standard USPS delivery times.

The legislature could have opted for an earlier application deadline, whereby voters returning their ballot by postal mail would be more assured of meeting the received-by deadline. Instead, the legislature opted to provide more flexibility on application timing with the later application deadline.

Act 77 grants voters the choice to wait up to a week before Election Day to apply for a mail-in ballot, with the implicit, common sense understanding that voters will bear an increased risk of missing the received-by deadline the longer they wait to apply. To mitigate this risk, Act 77 also provides voters the option of hand delivering their mail-in ballots, including the option of applying for and submitting ballots in-person at an election office on the same day. To claim the deadlines were flawed is disingenuous.

Moreover, Boockvar’s argument presumes that voters have little or no agency — that those who apply close to the application deadline and consequently miss the receive-by deadline do so by fate — suggesting an obligation to provide such voters with flexibility on the receipt deadline when mail delays occur. However, voters can choose when to submit a ballot application. The possibility of pandemic- related mail delays should just change their calculation, motivating them to apply sooner rather than later.

Additionally, the argument fails to consider factors that would mitigate the effect of any pandemic-related mail delays. Drop boxes are one obvious mitigant. Surely, the availability of drop boxes for returning a mail-in ballot limits the risk that late applicants will miss the receive-by deadline. Another is to inform voters about the potential for mail delays and to encourage them to apply early, which was, in fact, the sensible advice being disseminated by the USPS at the time.

The Court decides

The Court’s written opinion indicates no awareness of any of these weaknesses in the arguments presented by the petitioners and Boockvar. The judges apparently were swayed by those arguments:

In light of these unprecedented numbers and the near-certain delays that will occur in Boards processing the mail-in applications, we conclude that the timeline built into the Election Code cannot be met by the USPS’s current delivery standards…The Legislature enacted an extremely condensed timeline, providing only seven days between the last date to request a mail-in ballot and the last day to return a completed ballot. While it may be feasible under normal conditions, it will unquestionably fail under the strain of COVID-19 and the 2020 Presidential Election, resulting in the disenfranchisement of voters…this Court can and should act to extend the received-by deadline for mail-in ballots to prevent the disenfranchisement of voters.

In short, the Court undertook to alter the deadline ensconced in plain print in the legislation, potentially skewing the vote outcome in favor of one party. The decision lacked factual support, and instead relied on an inaccurate and misleading estimate of processing volume and on nonsensical griping about a “condensed timeline.”

Quantifying the effect on the vote tally

The Court’s decision to grant a three-day extension to the ballot return deadline favored Biden due to the overwhelming ratio of Democrats relative to Republicans anticipated to vote by mail. It can also be seen as implicitly disfavoring those who were planning to vote at the polls, insofar as they were not extended any comparable flexibility, such as expanded hours at polling places or more than one day of in-person voting.

How many mail-in and non-overseas civilian absentee ballots did the Democrats manage to gain during the extended timeframe, net of ballots returned by Republicans? To address this question, we turn to a dataset from the Pennsylvania Department of State which has detailed information on each approved application for an absentee or mail-in ballot for the 2020 General Election. A link to the dataset, which is housed within the Department’s online Open Data Portal, was shared with the author in response to a “Right-to-Know Law” (RTKL) request.⁵

Individual data items include the date the application was received by the county in which the voter is registered. In addition, the data distinguish between traditional absentee ballots available to voters and the newly available, no-excuse mail-in ballots (henceforth to be referred to simply as absentee and mail-in, respectively) and further differentiate among types of traditional absentee ballots, such as military, overseas civilian, regular civilian, and so on.

The database also provides the date the ballot was received back by the county election office (the ballot return date). Non-return of a mail-in ballot that was sent to a requester is indicated by missing information in the return date field: if the ballot was not recorded as returned, then no return date is present.

Chart 1 shows the counts of returned mail-in and non-overseas civilian absentee ballots received in the two weeks prior to and including November 3, the statutory deadline for returning these ballots, and the return counts for the three-day extension period granted by the Court. The counts are separated by party affiliation of the voters. During these three days, 9,432 more ballots were returned by Democrats compared to the number returned by Republicans.

While some of these ballots from Democrats might have been votes for Trump and some from Republicans might have been for Biden, there is no way to determine how many of each, and a best guess is to assume that such cross-party votes are few and mutually offsetting. Under the further assumption that Independent and other voters split equally between Trump and Biden, the net additional vote count for Biden is simply this difference between the returned ballot count of Democrat versus Republican-affiliated voters.

Does the extension seem justified in retrospect?

Chart 1 also indicates that the number of ballots returned from October 21 through November 3 swamps the number returned during the added three days. The number returned during the added days sums to 45,761, just 2.3 percent of the total number (1,987,793) returned on or after October 21.⁶ This small share seems inconsistent with the notion that pandemic-related mail delays would present a major obstacle to the timely return of mail-in ballots—the primary rationale put forth to justify the extension.

The three-day extension was motivated by a perception that those applying close to the application deadline would require this flexibility due to the potential for “systematic mail delays”. Chart 2 shows how ballots with November 4-6 return dates were distributed by application period. As seen in the chart, over half of the mail-in and non-overseas civilian absentee ballots with extension period (November 4-6) return dates were associated with applications filed prior to October 15; that is, applications filed well in advance of the deadline. Thus, most of those benefiting from extension of the deadline had not been late filers.

Another question worth asking is, how many of the ballots with application dates close to the October 27 application deadline would not have arrived in time had the extension not been granted? If there were many such ballots, then this could provide some ex-post justification for the Court’s decision.

Chart 3 shows the percent returned during the November 4-6 extension period by date of application for each application date in the late October (October 15-27) application period. Not quite 10 percent of individuals applying within three days of the deadline had return dates on November 4-6, only a slightly higher percentage compared to individuals applying four to seven days before the deadline.

These outcomes seem inconsistent with the anticipated “systematic mail delays” put forth to justify the extension, and quite plausibly can be attributed instead to procrastination on the part of some individuals or to idiosyncratic factors affecting ballot return times. In retrospect, there seems to have been little justification for extending the deadline, especially as some individuals intending to vote in person may have missed their opportunity to vote due to idiosyncratic factors, and it seems unfair to have provided extra accommodation only for the mail-in voters.

Not all ballots with recorded return dates in November 4-6 are votes that would have been lost

If the Court had left the statutory deadline unchanged, then it seems quite plausible that some of the individuals whose votes came in during the added three days would have behaved differently, returning their ballots sooner.

In other words, voters’ behavior cannot be presumed independent of the stated deadline—voters would have adapted their behavior in response to changing the deadline. It is not possible, however, to determine how many voters with return dates on November 4-6 would have met the original deadline had it not been extended.

Likewise, the actions of those whose duty it is to collect and record ballots may have adapted to the extension of the deadline. In some cases, those responsible for collection may have left some ballots uncollected as of the close of business on November 3, knowing that there would be additional time for completing the task. Even if all ballots had been collected on the 3^rd, some officials might have delayed opening and recording them until the next day, at which point they might be assigned a November 4 return date.⁷ The actual net gain to Biden due to the extension would depend on the total number of ballots incorrectly attributed to the extension period and how they divide up between Biden and Trump, which are also unknown quantities.

Enabling underperformance?

A far disproportionate share — nearly half — of the mail-in and non-overseas civilian absentee ballots with return dates on November 4-6 were from voters in just two counties, Bucks and Delaware. Chart 4 shows the distribution of ballots with November 4-6 return dates across the top 10 counties ranked by number of such ballots. Delaware County alone accounted for a third of all ballots returned during the three-day extension period. Bucks had about the same share of these ballots as much larger Philadelphia.⁸

These same counties, Bucks and Delaware, were notorious for having had especially severe delays in distributing mail-in ballots for the Primary in June. Uniquely among all Pennsylvania counties, these two had gone to their county courts to seek an extension to the ballot return deadline for the Primary. In fact, the experience of these two counties in the Primary was highlighted by proponents of the November extension as portending more systematic delays in the General Election. As described in the state Supreme Court’s written opinion regarding the November extension:⁹

…in Delaware County, thousands of ballots were ‘not mailed out until the night’ of the Primary, making timely return impossible… Bucks County apparently experienced similar delays. To remedy this situation, the Election Boards of Bucks and Delaware Counties sought relief in their county courts… the Courts of Common Pleas of Bucks and Delaware Counties extended the deadline for the return of mail-in ballots for seven days, so long as the ballot was postmarked by the date of the Primary.

Hence, one might reasonably conclude that the primary accomplishment of the Supreme Court’s granting the three-day extension was to relieve these two counties from responsibility for making needed improvements to their ballot distribution and collection processes. Surely, in retrospect, this is insufficient justification for the Court’s decision.

Concluding remarks

The troubling nature of election lawfare is exemplified by a particular case from the 2020 Presidential election in Pennsylvania in which the Democrats convinced the state’s Supreme Court to extend the deadline for receipt of mail-in ballots. In short, the Court undertook to alter the deadline ensconced in plain print in the legislation, potentially skewing the vote outcome in favor of one party.

The decision lacked factual support. Rather, the judges (the majority of whom were Democrats) were swayed by inaccurate and misleading information that was presented to them, and by a nonsensical, emotionally driven argument about a “condensed timeline” for mail-in voters arriving late to the game.

The Court’s decision was wrongheaded, because of the weak reasoning brought to bear. Overriding the explicit requirement of the legislation without adequate justification amounted to judicial overreach.

A bit of irony is evident here, considering the lawfare conducted in 2020 by the Democrats against the Green Party in Pennsylvania. A Green Party official chose to fax (as opposed to hand deliver or mail) a candidate affidavit, which was technically a violation of candidacy requirements. This minimal infraction, during the state’s Covid lockdown when office hours were limited and direct contact between persons discouraged, could have merited forbearance, but the Democrats succeeded in expelling the Green Party candidate from the ballot. Whereas the Democrats would grant no leeway to the Green Party, they self-servingly pursued a far greater degree of forbearance regarding the statutory ballot receipt deadline, based on flimsy justifications using the pandemic as a pretext.¹⁰

Analysis of mail-in voting data from this election casts further doubt on the notion that extending the deadline was necessary to ensure sufficient time for return of mail-in ballots for those who had been relatively late in requesting a mail-in ballot. In particular, the data show that most of the ballots that arrived during the three-day extension period were not from late applicants.

The extension of the ballot receipt deadline yielded a material net vote gain for Joe Biden resulting from the extension–possibly as many as 9,000 votes–and alongside the Green Party’s expulsion and other factors contributed to a broader lack of trust in the election outcome. Let us hope that in 2024, experiences such as this, where lawfare is used as a means of acquiring an unfair advantage, are not repeated.

Footnotes
1. In the 2020 Presidential Election, 57.7 percent of votes cast for Biden, but only 17.6 percent of votes cast for Trump, were by mail-in or absentee ballot, according to the state’s official election results. ↩︎
2. The case history discussion is drawn from the Opinion recorded by Justice Baer for the Pennsylvania Supreme Court’s September 17, 2020, decision on this case. ↩︎
3. Roarty, Alex, and David Catanese. August 27, 2020. “Despite uncertainty, Democrats bet big on mail voting in Pennsylvania.” McClatchyDC.com. Link. ↩︎
4. Bryant, Christa Case. September 4, 2020. “Why Pennsylvania is ground zero for mail-in voting debate.” The Christian Science Monitor. Link ↩︎
5. The link was provided in a RTKL response letter dated March 25, 2021; however, the dataset itself might already have been publicly available. The version currently accessible through this link may incorporate some updates relative to the dataset used here. ↩︎
6. Broken out by party affiliation, the shares received during November 4-6 were 1.9, 3.5, and 2.7 percent for Democrats, Independents or other, and Republicans, respectively. ↩︎
7. In fact, it is almost certain that some ballots recorded as returned on the 4^th actually were returned on the 3^rd. For five outlier counties, the data show zero or near zero ballots returned on the 3^rd but a far larger number on the 4^th, which is not credible. These five outlier counties and their ballot counts for November 3^rd compared with the 4^th are: Clinton (6 versus 169); Luzerne (5 vs. 2,076); Sullivan (1 vs. 33); Lebanon (0 vs. 729); and Mifflin (0 vs. 513). In the rest of the state, more than three times as many ballots had a return date of the 3^rd relative to the 4^th. ↩︎
8. Philadelphia had about 2⅓ times as many registered voters as Bucks County. Montgomery had less than a third the number of ballots with November 4-6 return dates relative to Bucks, despite having about 40 percent more registered voters. Also note that the shares for Luzerne and Lebanon shown in the chart are inflated for the reason described in the previous footnote – zero or near-zero ballots with November 3 reported return date. ↩︎
9. See pages 21-22 the Pennsylvania Supreme Court’s September 17, 2020, decision. ↩︎
10. For a wider ranging discussion of lawfare cases initiated by Democratic Party activists during the 2020 election around the nation, see chapter two of Fund, John, and Hans von Spakovsky. 2021. Our Broken Elections: How the Left Changed the Way You Vote. New York: Encounter Books. ↩︎
Data from the 2020 Pennsylvania Election Confirm: Elections Can Be Won or Lost on the “Supply Side” of Mail-in Voting

Aug 4th, 2024
Voting at a polling place or filling in and returning an absentee ballot can be regarded as the “demand side” of the voting process. Until 2020, elections in Pennsylvania, and in other states without a no-excuse mail-in voting option, primarily played out on the demand side.

Most voters had only one option, which was to go to the polls and pull the lever for their preferred candidate. Absentee voting was limited to active military, and to civilians specifying a pertinent reason they would be unable to go to their polling places on election day. Therefore, the process was controlled almost entirely by citizens’ decisions about whether to vote and for whom to vote.

However, the introduction of no-excuse mail-in voting in 2020 in Pennsylvania and other states greatly expanded the “supply side” of the election process. The added supply-side dimensions encompass the mechanics and procedures by which voters apply for mail-in ballots; how these applications are processed; and how the ballots are then distributed, returned, and tallied.

While in some states mail-in ballots are automatically distributed to all registered voters, in many states, including in Pennsylvania, voters must apply, either online or paper, to receive a mail-in ballot. Where this is the case, efforts to encourage and facilitate of mail-in ballot applications from voters, so-called “application harvesting”, can become a critical, “supply-side” activity affecting the outcome of the election.

This article analyzes data from the 2020 Presidential election in Pennsylvania, finding that spikes in paper applications for mail-in ballots from registered Democrats that occurred in the latter half of August into the first week of September may have produced enough votes to swing the election to Joe Biden. The analysis demonstrates the importance of the supply side to the outcome of the election, as the most plausible source of the spikes was application harvesting activity by partisan actors.

Forms of application harvesting

Nonprofit, tax-exempt, and technically non-partisan “get-out-the-vote” organizations as well as the political parties themselves historically have focused on promoting voter registration. With the expansion of mail-in voting, they have allocated increased attention and resources to application harvesting activities.

One important tactic the nonprofit organizations have employed is the mass distribution of paper applications for mail-in ballots. For example, it is well known that two nonprofit “sister” organizations, the Center for Voter Information and the Voter Participation Center, had together engaged in a massive effort to promote mail-in voting for the 2020 general election, by sending out millions of paper applications for mail-in ballots via postal mail to voters in multiple states.

But get-out-the-vote organizations are not the only possible conduit for application harvesting. Other organizations that are ostensibly nonpartisan such as employee associations, professional organizations, and the like, might distribute paper applications for mail-in ballots to their members.

While permissible in Pennsylvania and many other states, such organizations generally are required to conduct their activities in a non-partisan manner. Sixteen states currently restrict or prohibit third-party individuals or groups from distributing absentee and mail-in ballot applications, according to the National Conference of State Legislatures.

The political parties themselves may engage directly or through surrogates in application harvesting activities, including through mailbox stuffing, door-knocking campaigns, or personal, interactive outreach campaigns by party loyalists. Conceivably, personal outreach might be more effective than mass mailbox-stuffing.

For instance, party activists might be supplied with paper applications to personally distribute to family, friends, or neighbors who share their political leanings. The activists not only would seek to persuade these relatives, friends, and neighbors to apply for a mail-in ballot but could supply them with their own set of applications and encourage them to reach out to others in the same way. Such a campaign, if sufficiently scaled up, could be quite effective at generating additional votes.

Is application harvesting legit?

Application harvesting is a legitimate election-related activity provided it is conducted in accordance with established rules and norms. For instance, when engaged in by non-partisan not-for-profit organizations (where permissible by state law), it can lose legitimacy if conducted in a manner that excludes voters from a disfavored party or if coordinated with a favored party.

The legitimacy of application harvesting efforts by political parties or their surrogates could be open to question to the extent they might incorporate overly aggressive or intrusive tactics or violations of privacy. Indeed, it is conceivable that application harvesting activity in some cases might cross the line into questionable tactics.

For instance, an application might be filed by one individual on behalf of another, either directly without informing the prospective voter, or by asking him or her to sign a pre-filled form without adequate informed consent. A nursing home employee might instruct a resident to sign a ballot application that the employee had filled in, a clear violation of trust or fiduciary responsibility, for example.

A fundamental concern around mail-in voting in general is that the voter’s privacy could be compromised—persons voting by mail may be subject to intrusion into their decision-making, from the initial choice whether to vote by mail right up to the point that they seal their ballot. Such a concern encompasses application harvesting. For example, an intrusive family member might apply for a mail-in ballot on another’s behalf, a form of voter impersonation, as part of a scheme to pressure the voter.

Introducing the data

The source of the data for the analysis is a dataset from the Pennsylvania Department of State with detailed information on each approved application for an absentee or mail-in ballot for the 2020 General Election. A link to the dataset, which is housed within the Department’s online Open Data Portal, was shared with the author in response to a “Right-to-Know Law” (RTKL) request.

Individual data items include the date the application was received by the county in which the voter is registered. The data also indicate the means by which an application was submitted — either online, or as a paper application delivered by postal mail or in person at an election office.

In addition, the data distinguish between traditional absentee ballots available to voters and the newly available, no-excuse mail-in ballots (henceforth to be referred to simply as absentee and mail-in, respectively) and further differentiate among types of traditional absentee ballots, such as military, overseas civilian, regular civilian, and so on. The data also provides the voter’s date of birth and party affiliation.

Counts of applications according to the type of application filed indicate that more than 1.2 million paper and about 1.4 million online applications for the no-excuse mail-in ballots were submitted, far outnumbering requests for traditional absentee ballots. Online applications for domestic, civilian absentee ballots were the third largest category, totaling about 326,000.

Chart 1 shows the distribution of applications for mail-in ballots by type of application separately for three age ranges: under 35 (younger voters); over 70 (older voters); and 35 to 70 (mid-age voters). Younger voters favored online to paper applications by roughly a 2 to 1 margin, with the reverse being the case for older voters. Mid-age voters also favored applying online, but by a smaller, 35 percent margin.

Chart 2 shows counts of applications for no-excuse mail-in ballots by application type (online or paper) across five time periods defined as follows:
- Applications received prior to June 1 would mostly (perhaps entirely) have been from voters who had applied to vote by mail in the early June Primary and had opted (by checking a box on the Primary application) to automatically receive a mail-in ballot for the General Election. Thus, it is natural to distinguish this period, which we refer to as the Primary period.
- Ballot printing in Pennsylvania did not move forward until the state’s Supreme Court resolved a dispute over the Green Party candidates would appear on the ballot. That decision was announced on September 17, which we adopt as the end date of the second application period and shall refer to that as the “summer period”.
- The remainder of September constitutes the third period.
- The state’s deadline for applying for a mail-in ballot was October 27, motivating a division of October into two roughly equal-sized periods: the 1st through the 14th, and the remaining days up to the application deadline.
As seen in this chart, the number of online applications associated with the Primary period was about double the number of paper applications associated with that period. The number of online applications dropped off sharply between the Primary and summer periods, but there was just a small decline in the number of paper applications between these periods. Note that Primary period applicants were restricted to voters registered as Democrat or Republican, because in Pennsylvania voters must be affiliated with the party to vote in its primary, while later applicants additionally included those not affiliated with either of the two major parties.

A surprising divergence

Chart 3 disaggregates Chart 2 based on party affiliation: the number of applications by period and type are shown in Panel A for Democrats and in Panel B for Republicans.

For the Primary period, online outnumbered paper applications by more than two-to-one among Democrats, whereas applications were about evenly divided between online and paper among Republicans. These differing shares of online versus paper might reflect socioeconomic and demographic differences by party affiliation among voters who participated in the Primary.

It also is evident from Chart 3 that online applications from voters of both parties as well as paper applications from Republicans dropped sharply between the Primary and summer periods. However, paper applications from Democrats increased between the two periods.

Thus, the path of paper applications from Democrats exhibits a remarkable divergence. It is now evident that the contrasting dynamics seen in Chart 2 for online versus paper applications reflect the divergent path of paper applications from Democrats.

Specifically, paper applications from Democrats rose by nearly 20 percent, whereas those from Republicans plummeted by 70 percent and online applications from both parties fell by about 60 percent. If paper applications from Democrats instead had dropped only half as much as online applications; that is, by 30 percent, Democrats would have had about 100,000 fewer mail-in ballot applications.

That would translate into 90,000 fewer votes (assuming a 10 percent non-return rate of mail-in ballots, which is the non-return rate observed in the data for summer period, paper applications from Democrats.) That number of votes would have been more than enough to swing the state’s Presidential election. In a sense, then, the 2020 election outcome was a done deal by summer’s end, determined by the summer period’s heavy volume of paper applications for mail-in ballots from Democrat voters.

It seems a safe bet that some form of organized campaign had put paper applications in the hands of registered Democrat voters during this time and then encouraged them to fill out and submit the applications. In other words, the divergent path of paper applications likely was a supply-side phenomenon, not a spontaneous wave of voter enthusiasm for Joe Biden. Otherwise, one would expect equally strong momentum for online applications from Democrats.

Drilling down further—weekly application dynamics

The Department of State data contain sufficient detail to support a more granular examination of the timing of mail-in ballot applications, including further exploration of the summer period surge in paper applications from Democrats. The data also provide the ability to drill down to the county level, enabling an analysis of weekly application rates by county. It is advisable to disaggregate this way because a state-level analysis might mask significant differences across individual counties.

The approach we pursue here is to track weekly application rates (measured as each week’s average daily count) by application type, county, and party affiliation, from the week ending June 7 through the week ending October 27. Because overall application volume can differ greatly across counties and between Democrats and Republicans in any given county, the approach involves normalizing the weekly application rate for each party in each county relative to a benchmark week, in order to facilitate visual comparisons of the timing of activity. The week ending September 20 is selected for this normalization.

The dynamics of paper applications

It turns out that counties can be segmented into groups based on visual comparison of weekly application rates for paper applications from Democrat voters, such that counties within the same group have similar dynamics that are distinct from those in other groups. All but one of these groups exhibit comparatively rapid acceleration in applications from Democrats during mid-to-late August, with a peak application rate occurring no later than the week before Labor Day. The primary distinction among the several groups exhibiting this characteristic pattern regards the timing of the peak—the week in which the paper application rate from Democrats reached its apex.

Thus, in these counties, which comprise most of the state, paper applications from Democrats are front-loaded relative to paper applications from Republicans. The largest of these groups, consisting of 19 counties, is characterized by a spike in the former that peaks in the week ending August 30, as shown in panel A of Chart 4.

For the sake of brevity, we shall limit our discussion of application rate dynamics to the latter cohort (as broadly representative of all groups characterized by front-loading of paper applications from Democrats) and the single cohort not exhibiting this pattern; that is, no front-loading. The dynamics of paper applications for the 18 counties comprising the latter cohort are shown in Chart 4, panel B.

Map 1 distinguishes the latter cohort based on distinct, lighter green shading, and likewise differentiates the other counties by degrees of darker shading, according to the week in which the rate of paper applications from the county’s Democrat voters was at its apex. Due to data irregularities, three counties remain unclassified.

Panel A of chart 4 shows a surprisingly steep and voluminous surge in paper applications from Democrats within the last two weeks of August and into the first week of September in the group of counties represented. This compares to a more gradual and modest rise for Republicans. For voters unaffiliated with either of the two major parties (self-declared Independents and others), paper application rates were situated between those of Democrats and Republicans during the weeks of the spike. Thus, we find that the Democrats reaped a bountiful, late summer harvest of paper applications for mail-in ballots in these (and in most other) counties within this three-week period.

In stark contrast, as seen in Panel B, the no-frontloading group exhibits nearly identical dynamics between Democrats and Republicans for paper applications through the week ending September 20. In both cohorts, application rates from Republicans continued to rise after mid-September; those from Democrats did not. As a result, the paper applications from Republicans were comparatively back-loaded.

The dynamics of online applications

No such surge occurred in the case of online applications. In contrast to paper applications, the dynamics of online applications through the week ending September 20 were very similar between Democrats and Republicans, within both county groups, as seen in the two panels of Chart 5.

Therefore, whatever caused the late summer surge in paper applications from Democrats in most counties was not present in the case of online applications or in the case of the “hold-out” group represented in Panel B of chart 4.

Circling back

These weekly dynamics can be mapped back to the pattern observed earlier in the article mail-in ballot for application volumes across the five application periods. Charts 6 and 7 revisit that analysis, now differentiating between the two county groups. Panel A of each chart tracks the volume of applications from Democrat voters, separately for paper and online applications, while panel B does the same for applications from Republican voters.

As seen in the two panels of Chart 6, the county group characterized by the August 30 peak is characterized by a steep rise in the volume of paper applications from Democrats between the Primary and summer periods, and the opposite (a large decline) for Republicans. Thus, the comparatively rapid acceleration in paper applications from Democrats during mid to late August in these counties yielded a correspondingly large increase in application volume relative to the Primary.

In contrast, as seen in the two panels of Chart 7, the “no front-loading” counties where this rapid acceleration did not occur show a substantial decline in paper applications from Democrats between the Primary and summer periods, comparable to that observed for Republicans.

Thus, it is clear that the statewide aggregate jump in paper applications from Democrats between the Primary and summer periods, described earlier in the chapter, is tied to the counties that experienced a surge in these applications during mid-to-late August. In the case of counties that had no such surge, application volumes across the five periods followed similar paths (only the scaling is different) for Democrats versus Republicans, whether for paper or for online applications.

Summing up

Spikes in the rate of paper applications from Democrats during the last two weeks of August and first week of September, distinct from a more gradual ramping up of paper applications from Republicans, occurred in most Pennsylvania counties. In contrast, the dynamics of online applications were similar between Democrats and Republicans.

It seems implausible that these observed spikes in paper applications from Democrats were voter-initiated. In other words, it is difficult to imagine that among Democrat voters, enthusiasm for Biden, the impulse to get an early start, and the inclination to use a paper application simultaneously surged to generate these unique dynamics. If that were the case, why would the zeal have been confined to filing paper applications, while those preferring the online channel lagged behind? Moreover, how would one then explain the presence of a “holdout” group of 18 counties exhibiting no such wave of enthusiasm?

Therefore, we can be fairly certain that the impetus came from the supply side. In other words, some form of application harvesting in the counties exhibiting the spikes almost surely drove these dynamics. However, the particular activities responsible for the spikes, and the number of additional votes that resulted, have not been determined.

And as for the impact on the election outcome? On the one hand, undoubtedly some of the paper applications “harvested” during the summer period would have been from individuals who would have voted anyway, by applying for a mail-in ballot later on or voting in person. On the other hand, almost surely some would have been from individuals who might not have voted at all if not for the harvesting activity, implying some effect on the election outcome. Based on the data presented here, it is reasonable to conjecture that this effect was material.